Advances in Immunology Volume 51

ADVANCES IN Immunology VOLUME 51 This Page Intentionally Left Blank ADVANCES IN Immunology EDITED BY FRANK J. D...

Author: Frank J. Dixon

14 downloads 800 Views 26MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

ADVANCES IN

Immunology

VOLUME 51

This Page Intentionally Left Blank

ADVANCES IN

Immunology EDITED BY

FRANK J. D I X O N Scripps Clinic and Research Foundation La Jolla, California

ASSOCIATE EDITORS

K. FRANK AUSTEN LEROYE. HOOD JONATHAN W. UHR TADAMITSU KISHIMOTO FRITZMELCHERS

VOLUME 51

A C A D E M I C PRESS, INC. Publishers

Harcourt Brace Jovanovich,

Son Diego N e w York Boston London Sydney Tokyo Toronto

This book is printed on acid-free paper. @ Copyright 0 1992 By ACADEMIC PRESS, INC. All Rights Reserved. No part of this publication may be reproduced or transmitted in any form ox by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher.

Academic Press, Inc. 1250 Sixth Avenue, San Diego, California 92101-4311 United Kingdom Edition published by ACADEMIC PRESS LIMITED 2428 Oval Road, London NWl 7DX Library of Congress Catalog Card Number:

61-17057

International Standard Book Number: 0-12-022451-8 PRINTED IN THE UNITED STATES OF AMERICA 9 2 9 9 9 4 9 5 9 6 9 7

BB

9 8 7 6 5 4 3 2 1

CONTENTS

Human Antibody Effector Function

DENNIS R. BURTON A N D J E N N M. Y WOOF

I. Introduction 11. Antibody Structure: Considerations for Effector Function

Ill. Complement Activation by Antibodies IV. Human Leukocyte Fc Receptors V. Catabolism of Antibodies VI. Bacterial Fc Receptors VII. Conclusions References

1 3

7 29 54 56 62 64

The Development of Functionally Responsive T Cells

ELLENV. ROTHENBEKG I. Properties of Mature T Cells: The Destination of the Process 11. The Thymus and Its Seeding 111. Intrathymic Transitions in Outline

IV. Functional Maturation of Thymocytes V. Selection: Questions of Mechanism VI. Summary References

85 110 117 139 164 189 194

Role of Perforin in Lymphocyte-Mediated Cytolysis

HIDEOYACIIA,Moroui NAKATA,AKEMIKAWASAKI, YOICHIS H I N K A I , KO OKCMURA

AND

215 216 219 224 232 233

1. Introduction 11. Structure of Perforin

111. Expression of Perforin IV. Role of Perforin in Cytolysis V. Perspectives References

V

vi

CONTENTS

The Central Role of Follicular Dendritic Cells in Lymphoid Tissues

FOLKESCHRIEVER A N D LEEMARSHALL NADLER I. Introduction 11. Distinction of Follicular Dendritic Cells from Other Dendritic Cells 111. Solution of Technical Difficulties IV. Antigen Expression of Follicular Dendritic Cells V. Cellular Origin of Follicular Dendritic Cells VI. Regulation of the Normal Germinal Center by Follicular Dendritic Cells VII. Differential Expression of Follicular Dendritic Cells in Malignant Lymphomas VIII. Follicular Dendritic Cells as Targets of the Human Immunodeficiency Virus IX. Concluding Remarks References

243 244 246 257 261 263 274 275 278 281

The Murine Autoimmune Diabetes Model: NOD and Related Strains

HITOSHI KIKUTANI AND SUSUMU MAKINO I. Introduction 11. Development of the NOD Strain 111. Clinical and Histopathological Characteristics IV. Genetics V. Autoimmunity VI. Intervention of Autoimmune Diabetes VII. Concluding Remarks References

285 287 289 290 294 306 310 312

The Pathobiology of Bronchial Asthma JONATHAN

P. ARMA N D TAKH. LEE

I. Introduction 11. Pathology of the Asthmatic Mucosa

111. Eicosanoids and the Pathophysiology of Asthma IV. Aspirin-Sensitive Asthma V. Corticosteroid-Resistant Bronchial Asthma VI. Summary References

323 324 340 353 358 361 363

INDEX

383

CONTENTS OF RECENT VOLUMES

393

ADVANCES IN IMXIlJNOLO(;Y. VOL. 51

Human Antibody Effector Function DENNIS R. BURTON't AND JENNY M. WOOFt * Departments of Immunology and Moleculor Biology, The Scripps Research Institute, La Jo11a, Colifornio 92037

t Krebs Institute, Department of Molecular Biology and Biatechnology, The University, Sheffield, 510 2TN, England

I. Introduction

A molecular explanation of antibody effector function requires the description of multiple antibody molecules cross-linking an array of antigen molecules to multiple effector molecules. The antigen molecules are likely to be on a cell surface and the effector molecules are either large, as for complement, or also on a cell surface, as for Fc receptors. Therefore this is a complex problem. As regard antibodies, we have crystal structures for Fab fragments, for Fall fragments coniplexed with a number of antigens (reviewed in Davies et al., 1990), and for Fc from IgG (Deisenhofer, 1981; Sutton and Phillips, 1983).We also have the low-resolution structures of two mutant whole IgG molecules (Silverton et al., 1977; Sarnia and Laudin, 1982; Rajan et ul., 1983). These mutant molecules lack the hinge region and generally show poor effector activity (Burton, 1985). They crystallize and show a complete diffraction pattern probably because the loss of the hinge has reduced their flexibility. Native IgG molecules are flexible and do not give diffraction from the Fc part of the molecule (Huber et al., 1976; Ely et al.,1978). Therefore, our best picture of the whole IgG molecule is probably built up by combining the crystal data on the fragments with techniques such as electron microscopy (EM) and solution studies giving information on whole antibody conformation and flexibility. For the other antibody classes we have less data and generally have to make some fairly broad extrapolations from IgG in order to have a working model. A small number of studies provide us with some ideas as regards the arrangement of antibodies in arrays such as might be found triggering effector systems. NO effector molecule has been crystallized. However, the gross shape of the complement C l q as a bunch of tulips is well known, and the dimensions of some of the other complement components interacting with antibody are emerging. Many of the Fc receptors have now been cloned and sequenced and shown to belong to the immunoglobulin supergene family. Coupled with recent data on the sites on anti1

Copyright 0 1YY2 by Academic Press, Inc. All rights 01 reproduction 111 a n y kiriii rrserved.

2

DENNIS R. BURTON AND JENNY M. WOOF

FIG.1. The structure of Fc. The structure was solved for the Fc fragment from pooled approximatecenters of carbohuman IgG (Deisenhofer, 1981);0,a-carbon positions; 0, hydrate hexose units.

body molecules interacting with effector molecules, one can therefore place some useful constraints on how the three molecules, i.e., antigen, antibody, and effector, could be arranged in space relative to one another. Such arrangements then need to be placed into the context of arrays, and here we are largely at the stage of suggesting models for experimental investigation. In this review we shall consider effector functions in turn, concentrating first on IgG, for which molecular information is most detailed,

FIG. 2. The structure of IgG. This picture has been generated by taking the known structures of a human F(ab’)2fragment and a human Fc (Marquart et al., 1980; Deisenhofer, 1981) and constructing the hinge of human IgG,. The heavy chains are shown in dark blue and yellow, the light chains in green, and the cH2 carbohydrate chains in light blue. The proposed Clq-binding site discussed in the text (Gln 318, Lys 320, and Lys 322) is shown in red on one of the CH2 domains. The proposed Fc receptor (Fc,RI) binding region (Leu 234-Gly 237) is shown in white on both heavy chains. We thank Drs. Peter Artymiuk and Geoff Ford for permission to use this picture genented in collaborative work.

FIG. 3. The structure of IgE. A stPreodmwing of the a-carbon tlace of the revised model of Fc, from Helm et al. (1991) based on sequence homology with Fc of IgG (Deisenhofer, 1981) and taking into account the parallel nature of the inter-€ chain disulfide bonds. The C,2 domains (red) are at the top, the C,3 (green) are in the middle, and the C,4 (magenta) are at the bottom. Carbohydrate side chains (yellow) as found in IgG, (Deiseuhofer, 1981) are drawn attached to Asn 394 between the C,3 domains. The intmdomain and interchain cysteines (Cys 241 and Cys 328) are indicated by large circles (yellow).The white segment on the left €-chain highlights the location of an Fc,RI-binding peptide (residues 301-367).

HUMAN ANTIBODY EFFECTOR FUNCTION

3

and then on the other antibody classes where relevant. We shall attempt to build from unit interactions, e.g., one Clq head binding to one Fc site, to the more physiological situation, e.g., complement activation at an array of antigen-bound antibody molecules. There will be many instances wherein it will not be possible to be so tidy. Because there is a growing interest in the ability to engineer or design antibodies for specific effector function, we shall seek to highlight to what extent this is currently possible and describe some of the potential problems.

II. Antibody Structure: Considerations for Effector Function

A. STRUCTURE OF Fc

The crystal structures of human (Deisenhofer, 1981) and rabbit Fc (Sutton and Phillips, 1983)from IgG have been determined to intermediate resolution and analyzed in detail in terms of potential interacting sites elsewhere (Burton, 1985). Nevertheless, the Fc is so central to effector function that it is worth summarizing a few central points. The structure of human Fc is shown as an a-C trace in Fig. 1 and space-filled in Fig. 2 (compare also the Fc from IgE in Fig. 3). The two C,3 domains form a classical immunoglobulin domain pairing. There are extensive lateral van der Waals contacts between the domains as well as several hydrogen bonds between polar side chains and a pair of salt bridges. This close interaction results in approximately 1000 A2 of surface from each domain being removed from solvent contact. Each Cy3domain is linked by a loosely folded segment (Ser 337 to Gln 442) to the Cy2 domain. The C,2 domains are not paired in the usual fashion and indeed the polypeptide chains have no contact with one another until the hinge region. The hydrophobic face of the domain normally involved in pairing is partially covered by a branched N-linked carbohydrate moiety attached to Asn 297, which helps to stabilize the domain. Longitudinal interactions between residues in Cy2 (residues 247-253 and 310-314) and in C,3 (376-379 and 428-433) further serve to stabilize both domains. About 500 and 750 A' of surface area, respectively, are removed from solvent in these interactions. The carbohydrate chains of the C,2 domain are not a single oligosaccharide moiety but consist of a set of about 20 structures based on a mannosyl chitobiose core, which can be represented as shown in Fig. 4. The possible role of carbohydrate in ef'fector function is often investigated using aglycosylated IgG prepared by growing hybridomas in

4

DENNIS R. BURTON AND JENNY M. WOOF +_ +- (6‘) (5’1 (4‘1 Siaz2 --t %alp I -.4GlcNAcfl I -B 2Man a1

I I

f‘”czi

(2) 6 ( 1 ) 6 (3) G l c N A c f l I - r 4 M a n ~ I ~ 4 G l c N A c /-4GlcNAc,,, ~I 3

f

Siar2

-

f (6)

(5)

(4)

&alp I -,4GlcNAcfl I -. 2 M a n ~I

FIG. 4. The CH2 domain N-linked carbohydrate of IgC. As shown, four types of niannosyl chitobiose cores are found ( ? bisecting N-acetylglucosamine/+ fucose) and other chain variants include the presence or absence of galactose and sialic acid.

tunicamycin or by engineering Asn 297, the carbohydrate attachment residue, to another amino acid. The question then arises as to what effect loss of carbohydrate has on Fc structure. Proton nuclear magnetic resonance (NMR) (Matsuda et al., 1990) comparing native and recombinant Fc indicates that the overall structures are quite similar; there is no evidence for example of inward “collapse” of the two Cy2 domains. However, some differences are monltored, most notably by a reporter group (His 268) on a loop in spatial proximity to Asn 297. Small differences are also sensed by His 433 and His 435 (Cy3domain) at the junction of the Cy2 and C,3 domains and remote from the glycosylation site. One caveat here is that the NMR studies refer to Fc fragments and the presence of Fab arms could affect an aglycosylated Fc. Finally, the protein in Fig. 1 begins at residue 238 whereas Fc is generated by papain cleavage N-terminal to Cys 226. The intervening residues are not defined in electron density maps of human Fc presumably because of flexibility. This region is probably better considered as part of the hinge (the “lower hinge”; see below) even though it is coded for in the C,2 exon. B. CONFORMATION AND FLEXIBILITY OF IgG: THEHINGE Both immunoglobulin flexibility and the conformation of antibodies have been recently reviewed (Nezlin, 1990; Burton, 1990a,b; Schumaker et al., 1991).Again only major points will be summarized. Flexibility of antibodies is classically equated with Fab arm flexibility (Y to T shape changes), which should allow “variable reach” for the antibody and therefore bivalent recognition of differently spaced antigens. However, the IgG molecule has available a number of modes of flexibility as illustrated in Fig. 5. It would seem very likely, although it is not formally demonstrated at this stage, that this flexibility would be


5

FIG. 5. Flexibility in the IgC molecule. The modes of flexibility have been most convincingly demonstrated by electron microscopy, as reviewed by Burton (19901~).

used in the allotted task ofantibody to cross-link antigen and effector in a variety of situations. The potentially most interesting flexibility is that associated with the hinge. Indeed the hinge is the region of greatest difference between isotypes of IgG, and because the isotypes also show great differences in effector function activity, there has developed a notion that activity and hinge flexibility are intimately linked. In simplified terms, one tends to associate good effector activity with flexible isotypes and poor activity with less flexible isotypes. We shall try to evaluate this association during the course of this review, The hinge region can be conveniently divided into three parts (Fig. 6). The middle or core hinge contains a variable number of cysteines forming interheavy chain disulfide bonds and is connected to the folded Fc by the lower hinge and to the Fab arms by the upper hinge. NMR studies of human IgGl and isolated hinge peptides (Endo and Arata, 1985; Ito and Arata, 1985) indicate that the core hinge (Cys 226 to Cys 229) has a conformation little affected by the presence or absence of Fab and Fc and quite similar to that found in crystals of human IgG, Kol protein. The lower hinge (at least the part comprising Pro 230 to Leu 234) is suggested to be flexible and to have an extended conformation independent of the presence of the Fab arms but critically dependent on the intact core. The upper hinge (Lys 222 to Thr 225) is suggested to be the most flexible part, adopting the helical structure found in Kol crystals in the presence of the Fab arms. Nanosecond fluorescent depolarization experiments have been used to show a cor-

6

DENNIS R. BURTON A N D JENNY M. WOOF

Hinge

Antibody

I Upper

Lower

Middle

I

Mouse IgG3 Rat IgG2C Human IgG3 Mouse IgG2b G.pig IgG2 Human IgGl Rat IgG2b Mouse IgG2a Rabbit IgG Rat IgGl Human IgG4 Rat IgGaa Mouse IgGl G.pig IgGl Human IgG2

I

EPRIPKPSTPPGSS EPRRPKPRPPTDI ELKTPLGDTTHT EPSGPISTINP EPIRTPZBPBP EPKSCDKTHT ERRNGGIGHK EPRGPTIKP APSTCSKPM VPRNCGGD ESKYGPP VPRECNP VPRDCG QSWGHT ERK

I

csc CPRCP(r.u.)3 CPPCKECHKC CTCPKC CPPC CPTCPTCHKC CCPPKC C CKPCIC CPPC CGC CKPCIC CPPCIPC CCVECPPC

I

I IgG1 (1.2 X lo4) > IgGz (0.6 X lo4) > IgG4 (0.4 x lo4)was found. There was considerable error in the last value. The current consensus implies that there is no appreciable affinity of IgG, in associated form for C l q (Garred et al., 1989; Horgan et al., 1990; Tan et al., 1990). In contrast to IgG, 1gM occurs in its native form in an associated state, primarily a pentamer (Fig, 8). Uncomplexed, the pentamer expresses a single Clq-binding site with an affinity estimated as 5 x lo5 M-' (Feinstein et al., 1983) or as lo4 M-' (Poon et al., 1985). Complexed with antigen the affinity of the Clq-IgM interaction increases to of the order of 5 x lo7 M-'. This increase arises in antibody excess (IgM binding muItivalently to the same molecule expressing multiple epitopes) or in antigen excess (IgM cross-linking different antigen molecules). However, only the former case leads to C 1 activation (Feinstein et al., 1983). Therefore, it is suggested that the functionally important increase in affinity arises from the exposure of new C l q binding sites on a single IgM molecule on complexation rather than the spatial association of monovalent pentamers. The ability of single, complexed IgM molecules to activate C1 supports this view (Borsos and Rapp, 1965a,b; Ishizaka et al., 1968; Feinstein et al., 1983). Similarly, on a cell surface it is found that complement activation only occurs when more than one antigen-binding site in the IgM molecule is occupied (Borsos et al., 1981).


11

Interestingly Fc5, isoIated by proteolytic removal of the F(ab‘)z units, shows the same binding affinity for C l q as native uncomplexed IgM. Hence the creation of new C l q sites cannot simply be the result of antigen binding somehow moving F(ab‘)z units to “unblock” sites on an unaltered Fc5. The implication is that some sort of conformational change in Fc5 must accompany functional IgM binding to antigen (see below).

3. C l q-IgG lnteraction at the Molecular Level A number of studies have indicated the importance of charged groups in this interaction (Hughes-Jones and Gardner, 1978; Burton, 1985).The most detailed (Burton et al., 1980; Emanuel et al., 1982) indicated that about 12 ions are released into solution on the binding of one molecule of C l q to an IgG immune aggregate. Because the structural information on Fc is relatively so good, many studies have approached the molecular details of the interaction by attempting to localize the site on Fc binding to C l q . A definitive delineation of the site would likely require cocrystallization of an F c - C l q head complex but protein engineering of mouse IgGzb (Duncan and Winter, 1988) gives indication that three charged residues, Glu 318, Lys 320, and Lys 322, constitute the essential binding motif. This motif is part of a site earlier proposed (Burton et al., 1980). Figure 2 shows the localization of the motif and Fig. 9 shows the mutations that were made in the protein engineering approach. For the purposes of design one would like to know whether the motif is the entire binding region, and, if so, any importance of the context of the motif. The ability of a peptide mimic of the motif to inhibit the activation of complement with an inhibition constant close to that observed for IgG (Duncan and Winter, 1988) would suggest that the motif is sufficient. Certainly many mutations were carried out in the proximity of the motif (Fig. 9) with no effect on C l q affinity. The involvement of three charged groups if interacting with three similar groups on C l q would imply the release of six ions per C l q head bound. The observation of 12 ions released would therefore be consistent with two heads bound per C l q molecule to an immune aggregate, which is in agreement with the thermodynamics. Other mutations in the motif implied that a positive charge is required at positions 322 (Arg can substitute for Lys) and probably a hydrogen bond at position 318 (Thr can substitute for Glu). Position 320 will accept either Arg or Gln with retention of C l q binding, but the latter mutation abrogates complement activation. This underscores the existence of requirements additional to C l q binding for effective complement activation. It should

12


A

340

B 29


13

also be noted here that experiments with IgG molecules with heterologous heavy chains have been used to suggest that only one heavy chain is essential for Clq binding (Clark et al., 1989a). Other studies are suggestive of some importance for the context of the motif. A substitution of Gln by Glu at position 324 in mouse IgGz, adjacent to the motif abrogates C l q binding and complement activation (Nose et al., 1988; Nose and Leanderson, 1989). The behavior of different IgG isotypes also suggests the importance of context. The human IgG subclasses all possess the motif and yet IgG, does not significantly bind C l q . The favored interpretation of this has been that the Fab arms of IgG4 obstruct the Clq-binding site because of the “restricted” hinge of IgG, (Burton, 1985). The key experiments here are the description of Fc4 (Fc fragment of IgG,) binding C1 with an affinity comparable to that for F c l and, in aggregated form, activating complement (Isenman et al., 1975). The inability of hinge-deleted human IgGl mutants to bind C1 or activate complement seem to further provide examples of just such Fab arm obstruction (Klein et al., 1981).New protein engineering studies (Tan et d.,1990; Norderhaug et al., 1991)provide a challenge to this interpretation of IgG, function. In particular, an IgG, molecule wherein the hinge exon of IgG, is replaced with those found in IgG3 fails to bind C l q or activate complement. An IgG3 molecule with the genetic hinge of IgG, still binds C l q effectively (Tan et al., 1990).IgGS molecules with an IgG4 hinge or an IgG, hinge and C H 1 domain activate complement more effectively than wild type IgG3 (Norderhaug et al., 1991). Therefore, it would appear that there is a structural lesion in Fc4 that leads to its inability to bind C l q . The data of Isenman et al. would require that this lesion is sensitive to the presence of the Fab arms. The

FIG.9. Mutations made in a protein engineering strategy to localize the Clq-binding site on IgG (after Duncan and Winter, 1988). The diagram shows an a-carbon trace ofthe CH2 domain from human Fc (Deisenhofer, 1981).(A) Groups of residues that had been proposed as Clq-binding sites. A, Residues 274-281 (Boackle et QZ., 1979); B, residues 282-292 (Lukas et al., 1981); C, residues 290-295 (Brunhouse and Cebra, 1979); D, residue 318,320,322,331,333,335, and 337 (Burton et al., 1980). (B) Residues in which exposed side chains were altered by site-directed mutagenesis in the homologous mouse IgG,,, antibody. Residues altered to Ala were S239, K248,1253, D265, S267, D270,Q274, E283, H285, Q290, E294, N297, K317, E318, K320, K322, K326, E333, T335, S337, and K340. P331 was mutated to Gly and E235 (not shown) was mutated to Leu. Open circles show mutants that were still lytic; closed circles show mutants that were nonlytic and are suggested to comprise the Clq-binding site, i.e., E318, K320, and K322. The carbohydrate attachment residue, N297, is shown shaded.

14

DENNIS R . BURTON AND JENNY M. WOOF

positions in the CH2 domain distinguishing IgG4 from the other subclasses have been discussed previously (Burton, 1985; Jefferis, 1986). The sequence Ser 330-Ser 331 (Ala-Pro in the other subclasses, Fig. 1) in close proximity to the motif stands out but there are other differences, such as Phe 234 (Leu) and Gln 268 (His). The ability of an IgG4/IgG, switch mutant (CH1 to residue 291 of CH2 from IgG4; residue 292 to end of CH2 and CH3 domain from IgG1) to activate complement implies that residues 292-340 in the CH2 domain contain the amino acids responsible for the inability of IgG4 to activate (Tao et

al., 1991). The case of human IgG4 may have implications for the C lq-binding and complement-activating patterns of all the isotypes. Basically there have been two schools of thought. Both saw the Clq-binding site as being present on most isotypes but modulated by the Fab arms. The hinge, and in particular, the upper hinge, was seen as crucial. The first school tended to emphasise the necessity for flexibility in this upper hinge in binding C l q and pointed to a correlation between C l q binding and segmental flexibility (Oi et al., 1984; Dangl et at., 1988).The second school tended to associate “restricted” hinges, which could arise from shorter upper hinges, with structural accommodations placing Fab arms closer to Fc (Burton, 1985).It pointed to solution data indicating more compact conformations for isotypes such as human IgG2 and IgG4 with their inferior Clq-binding ability compared to the more extended IgGl and IgG3 (Gregory et al., 1987).The data of Tan et al. show that upper hinge flexibility per se (at least in the nanosecond time range) is not necessary for C l q binding. Thus, for example, two “rigid” mutant IgG3 molecules are able to bind C l q as effectively as the “flexible” wild-type IgG3 molecule. They also show that close approach of the Fab arms is probably not the reason for the failure of IgG4 to bind C l q unless CH2 folded domain sequences were somehow controlling hinge conformation and therefore Fab arm disposition. What is the origin of the poor C l q binding of isotypes such as mouse IgGl and guinea pig IgG1 and the loss of C l q binding associated with hinge deletion (Klein et al., 1981; Michaelsen et al., 1990; Tan et al., 1990)? Fab arm obstruction is an appealing explanation brought into question by the human IgG4 results. Other unidentified structural lesions are a possibility. In particular, hinge deletion may subtly perturb conformation. Interestingly, for mouse IgGI, replacement of the C,2 domain with that of mouse IgGzb generates C l q binding and complement lysis at a level comparable to that for IgGzb (Clackson and


15

Winter, 1989).This again militates against the importance of the upper hinge (hinge exon) but leaves open the role of the lower hinge (Cy2 domain exon). Mouse IgGl has a restricted hinge in terms not only of a short upper hinge but also of a short lower hinge where it lacks the characteristic Gly-Gly sequence. It should be noted, however, that introduction of the mouse IgGZb lower hinge into IgGl does not generate a lytic antibody (T. Clackson, personal communication). The exceptionally long hinge of IgG3 is something of an enigma. Two groups have now reported that most of the hinge can be deleted without any detrimental effects on C l q binding or complement activation (Sandlie et al., 1989; Michaelsen et al., 1990; Tan et al., 1990). In fact, an IgG3 with a single hinge exon is more effective at C l q binding and a molecule with an extra exon is less effective. Aglycosylation is a context to which C l q binding is sensitive to varying degrees. Aglycosylation of mouse IgGz, or mouse IgGZb produces only a small (roughly threefold) reduction in C l q binding affinity (Leatherbarrow et al., 1985; Duncan and Winter, 1988). Again, a small reduction (twofold) has been reported for human Fcl (Matsuda et al., 1990).In contrast, complete abolition of C l q binding has been reported for aglycosylated human IgGl and a dramatic decrease for IgG3 (Morrison et al., 1989).Even for those cases where C l q binding is minimally affected, whole complement activation, where looked at, is abolished. Recombinant hybrid molecules in which Fc is linked to other proteins afford an opportunity to look at the effect of context on effector function. Two groups have reported on recombinant CD4immunoglobulin molecules (Capon et al., 1989; Traunecker et al., 1989; Byrn et al., 1990). Replacing the VH domain of human y l chain by the first two or all four of the CD4 cytoplasmic domains gives molecules expressed as dimers in a eukaryotic cell line in the absence of light chains. Both molecules bind gp120 and FcRI, but neither binds C l q . However, a molecule in which both V H and C,1 domains of mouse IgG,, are replaced by the first two domains of CD4 does bind C l q . It would seem that the presence of the C y l domain, perhaps unable to pair with the CL domain, is detrimental to C l q binding. In this vein, attempts to graft Clq-binding sites into different molecules will undoubtedly b e revealing as regards to site requirements. Considering the interaction site on C l q , there are charged regions on each of three chains in the heads, e.g., the sequence Glu 198-XAsp 200-Lys 202 on the A chain, which might interact with the charged motif on IgG (Reid et al., 1982).

16

DENNIS R. BURTON AND JENNY M . WOOF

4 . C1 q-ZgM Znteraction at the Molecular Level As for IgG, it appears that charged groups are important in the interaction of IgM and C l q (Hughes-Jones and Gardner, 1978; Poon et al., 1985).It is estimated that 12 ions are released into solution on the binding of C l q to uncomplexed IgM in solution (V. N. Schumaker, personal communication) and 8-9 ions are released on the binding to IgM interacting with a cell surface. Given the differences in binding affinities described earlier, the Clq-binding site on IgM is not expected to be identical to that of IgG. Indeed, the IgG motif is not found on IgM. The Clq-binding site(s)are located on the Fc pentamer (Fig. 8) but the precise domain (C,3 or C,4) has been debated. An interesting mutant IgM molecule with a single amino acid change (Pro + Ser 436) in the C,3 domain has been isolated and shown to have decreased affinity for C l q (Wright et al., 1988).This has been taken as evidence to support the C,3 domain as binding Clq. It is noteworthy that the residue analogous to Pro 436 in IgG is Pro 331 in the C,2 domain, which is on the edge of the proposed IgG-binding site. Further, a Pro- Ser mutation at this position is prominent in the non-Clqbinding IgG4 isotype. The effect of the mutation in IgM is complex in that it renders half of the mutant molecules incapable of binding C l q and the other half capable of binding but with lowered affinity. A model based on an equilibrium between different IgM conformations has been proposed to explain these observations (Wright et al., 1988). Another mutant (Asn + Gln 402, comparable position Asn 297 in IgG, the C,2 glycosylation site), which does not glycosylate, shows decreased complement-activating ability compared to wild-type IgM (Muroaka and Shulman, 1989). The binding of IgM to antigen has been studied by electron microscopy and a model has been suggested for how extra Clq-binding sites would become available on antigen complexation (Beale and Feinstein, 1976; Feinstein and Richardson, 1981; Feinstein et al., 1983, 1986). Briefly, pentameric uncomplexed IgM generally appears crudely as a star-shaped molecule with F(ab’)2 unites emerging in various orientations from a planar Fc5 unit. On the binding of specific IgM to Salmonella paratyphi flagella, new staple-like IgM molecules are observed in which the F(ab’)z units are all dislocated out of the plane of the Fc5 disk and are bound to the flagella. Similarly, in an antidextran IgM/dextran system, the ability to activate C1 correlates with the appearance of IgM bound to single molecules ofdextran in the staple form. Therefore, it is suggested that some distortion introduced


17

into the IgM molecule in the star-to-staple transition reveals extra C l q sites and triggers complement activation. However, as commented above, removal of the F(ab')z units does not reveal extra sites, implying the star-to-staple transition generates some change in Fc5 conformation. Again, the problems of transmitting changes through the domain structure of immunoglobulins have been used to argue against an allostei-ic mechanism. Instead it has been suggested (Feinstein et al., 1986) that pivoting about the inter-C,3 bridges and readjustment of the spatial relationship of neighboring C,4 dimers might reveal extra C l q sites. Certainly protein engineering experiments indicate the importance of the inter-C,3 bridges involving Cys 414, because the pentameric IgM Ser 414 mutant is unable to activate complement. Perhaps one of the most intriguing developments in this area recently has been the demonstration in a number of laboratories of the propensity of IgM to form hexamers, particularly in the absence of J chains (Cattaneo and Neuberger, 1987; Davis et ul., 1988; Randall et al., 1990). The hexamer is found to activate whole complement 10- to 20-fold more efficiently than the pentamer (Davis and Shulman, 1989; Randall et al., 1990). Because J chains are not necessary for either assembly or secretion of polymeric IgM from B cells, it has even been suggested that their function may be to regulate the lytic efficiency of IgM by controlling the pentamer : hexamer ratio (Randall et al., 1990). The existence of both C l q and IgM as hexaniers is striking and at least suggestive of some involvement of symmetry in the triggering process.

5 . Clq-Associated 1gG at the Moleculur Level The most detailed description available of associated IgG comes from two-dimensional crystallization studies of a monolayer of a dinitrophenol (dnp)-lipid binding a mouse monoclonal anti-dnp IgG, antibody (Reidler et al., 1986). The antibodies form hexagonal arrays in which the Fcs are dislocated out of the plane of the Fabs to generate an angle of about 80". The monomers interact with one another via both Fc and Fab regions. We have constructed graphics representations based on this idea (Burton et al., 1989; Burton, 1990a) when association of Fc regions was made most readily though interaction of the large hydrophobic patch at the CH2/CH3 domain interface. In fact, although the precise amino acids in this region are not conserved between IgGs of different species and subclasses, there is a strong conservation of character as an extensive exposed hydrophobic patch. The patch is the recognition site for staphylococcal protein A, and the common reactiv-

18


ity of protein A with IgG is comprehensible in terms of its conservation (Burton, 1985). Two major features of the two-dimensional crystal studies are that the IgG molecule is viewed as dislocated and in Fc-Fc interaction. Independently, from studies on the interaction of IgG and Fc receptor, we suggested that IgG might be dislocated and in Fc-Fc interaction on an antigenic surface (Burton, 1986). We highlighted a possible link between IgG and IgM in their mode of binding Clq and complement activation (Fig. 10). Thus, IgM is normally in an associated state but must dislocate to bind Clq effectively. IgG is normally monomeric but on binding antigen it would form a defined polymer in a dislocated conformation, which would trigger complement. The complementactivating molecular species in the two cases is very similar according to this model. It is made more plausible by the description of hexavalent IgM, which, in a dislocated conformation, would be expected to closely resemble the two-dimensional crystal view of hexameric IgG. Interestingly, hexagonal symmetry was also described some time ago from E M studies of Fc crystals (Pinteric et al., 1971). Experimental evidence on associated IgG conformation in less arti-

FIG. 10. Schematic representation of C l q binding to an array of IgG molecules. The IgG molecules are arranged in a hexameric array as suggested by two-dimensional crystallization studies (Reidler et al., 1986). The Fcs are dislocated (Burton, 1986) and roughly at right angles to the plane of the corresponding Fabs. For clarity, the IgG molecules are shown schematically in a ribbon format: the two-dimensional crystallization studies indicate lateral Fc-Fc and V domain interactions. The C l q molecule, roughly to scale, is shown interacting with two IgG molecules. Distortions of C l q would be required to recognize adjacent IgG molecules.


19

ficial situations than an E M grid is sparse. There is evidence of Fc-Fc interactions in the formation of immune precipitates (Moller, 1979; Rodwell et al., 1980; Easterbrook-Smith et al., 1988). Fc-Fc interactions, reflected in a cooperative binding, have also been observed for the binding of mouse IgG3 to a surface antigen (Greenspan et al., 1987). However, we have observed no cooperative binding of human anti-rhesus D IgGl or IgG3 to red cells or of human anti-NIP (3-iodo-4hydroxy-5-nitrophenylacetate) IgCl or IgG3 to NIP-coated lymphoid cells (Woof et al., 1992). Howard and Hughes-Jones (1988) have focused on synergistic lysis of red cells to propose a model for the interaction of C l q and IgG on an antigenic surface. Briefly, as first reported by Elliot et al. (1978), antibodies bound together to a single antigen molecule on a cell surface have unusually strong lytic activity. For example, single monoclonal IgGs are often poor agents of complement lysis, but the addition of a second monoclonal IgG to a different epitope on the same antigen can cause a dramatic increase in lysis to a level much greater than the sum of that seen with the individual antibodies (Howard et al., 1979; Bindon et al., 1987). Howard and Hughes-Jones reported that in a synergistic situation of two rat monoclonal IgGZb antibodies binding to MHC class I antigen on a red cell surface, the stoichiometry of C l q binding is two C l q molecules bound per monoclonal antibody pair. The “autonomous” model of Fig. 11, in which two C l q molecules are bound bivalently to opposite faces of the same pair of IgG molecules, themselves bound to the same antigen molecule, was proposed to explain this data. Additional support for the model was derived from the ability of pairs of functionally monovalent antibody (one Fab arm inactivated b y association with a nonfunctional light chain) to show a full synergistic lysis effect. This excludes cross-linking of different antigen molecules as an important factor in triggering red cell lysis in this case. The autonomous model is clearly of some considerable interest and one would like data from other systems to examine its validity. A study (Bindon et ul., 1987) of C l q binding to synergistic rat IgGZb monoclonal antibodies binding to human leukocyte common antigen is consistent with an antibody:Clq stoichiometry of about 3: 1, but the experimental errors here could be quite large (G. Hale, private communication). An earlier study (Hughes-Jones et a.?.,1983) of synergistic rat IgG binding to MHC class I antigen on red cells found that the total number of Clq-binding sites was approximately the same whether the synergistic pair was composedpftwo IgGzb molecules or an IgGzb with an IgGz, molecule (which by itself did not significantly bind Clq). The

20


FIG.11. The autonomous model of‘ complement activation (Howard and HughesJones, 1988).The scale model shows the binding of two Clq molecules to two adjacent Fc pieces. The two IgC molecules are bound to two different epitopes on the same antigen molecule.

binding constant was decreased in the latter case. Furthermore, an F(ab’)z fragment of the IgGz, could not establish a full synergistic effect. These data imply some role for the Fc part of the IgGz, molecule in C l q binding and are perhaps indicative of stabilization of an antibody array rather than autonomous binding. The human IgG subclass pattern seen in the binding of C l q to associated IgG roughly mirrors that seen for monomeric IgG. More C l q becomes bound to IgG3 associated on hapten-coated red cells or on immobilized hapten-bovine serum albumin than to similarly associated IgGl (Bruggemann et al., 1987; Bindon et al., 1988a; Garred et al., 1989). However, the extent of IgG3 superiority depends on epitope density and complement concentration. The difference appears to reside in the number of C l q sites made available in the two cases, the binding constants being very similar. More sites are made available by the IgG3,(g) than the IgG3m(b)allotype. There appear to be some very subtle antibody con formational or geometric requirements for the generation of a site capable of binding C l q multivalently. Another study of two IgGl antibodies against different epitopes on the same synthetic branched polypeptide showed that, with equal amounts of antibody


21

bound to the polymer, one antibody bound severalfold more Clq than the other. No difference in sequence in the hinge or C,2 domains between the two antibodies was found (Horgan et al., 1990).This study again implies preferred antibody conformations or arrangements for productive C l q binding. Relatively significant C l q binding to associated IgGz is observed at higher epitope densities and complement concentrations whereas at lower ones such binding virtually disappears (Garred et al., 1989). These observations may account for earlier apparent inconsistencies in the literature. Associated IgG, is never observed to bind C l q significantly. Finally, there are now two reports to show that associated rat IgA is capable of binding C l q and of activating C1, but this does not lead to C4 deposition or cell lysis (Bindon et al., 1990; Hiemstra et al., 1990). There are conflicting reports about the ability of human IgA to take part in classical pathway activation (Iida et al., 1976; Burritt et al., 1977; Romer et al., 1980; Jarvis and McLeod Griffiss, 1989). B. C1 ACTIVATION Attempts to understand how antibody activates C l are facilitated by increased information on the structure of C 1 (reviewed in Cooper, 1985; Weiss et al., 1986; Arlaud et al., 1987; Schumaker et al., 1987). Briefly, there is ample evidence to indicate that the Clr&ls, tetramer is very extended as an isolated molecule but that it compacts considerably when complexed to C l q in C1. Symmetrical models have been proposed in which the Cls-Clr-Clr-Cls tetramer is wound within the stalks of C l q in an S or figure 8 shape (Fig. 12). Such models have the advantage that they bring the catalytic domains of C l r and C l s into contact with one another, thereby making it conceptually easier to understand how activated C l r can activate C l s . An asymmetric model (Fig. 13) that also achieves this feature has the tetramer folded at its midpoint and wrapped around the outside cone of C l q (Cooper, 1985). A filrther set of asymmetric models has been proposed (Perkins, 1985). The interesting question for this review is how C l q binding to associated IgG (or antigen-complexed IgM) leads to C 1 activation. The evidence seems to favor a distortive model in which binding to an array of Fcs distorts the cone formed by the spreading C l q arms. This leads to autoactivation ofClr, which in turn activates C l s . The best evidence for this viewpoint comes from the description of a mouse monoclonal IgGl antibody [or F(ab’)z fragment], which is against an epitope on the collagenous arms of C l q and which is able to activate C 1 (Hoekzema et al., 1988).

22


FIG. 12. Model of Complement C1 (adapted from Arlaud et al., 1987). The upper diagram shows a model of the extended conformations of Clr2Cls2 and how a "figure 8-shaped" conformation could be acquired on complexation with Clq. Ir, Is, Interaction dcmains of C l r and Cls; Cr, Cs, catalytic domains of C l r and Cls. The lower diagram models the Cls-Clr-CIr-Cls tetramer interacting with Clq.

This study shows that (1)bivalency of the antibody is a requirement for C1 activation but not for binding to C l q ; (2)increasing the segmental flexibility of the antibody by reduction and alkylation of hinge disulfides destroys the ability to activate C1; ( 3 )an antibody against the C l q heads inhibits C1 activation by associated antibody but not b y the anti-Clq arm antibody; (4)isolated C l q stalks ( C l q with the heads digested away) are still activated by the monoclonal antibody, indicating the heads are not the origin of the activating signal; and (5) C 1 activation is optimal at a monoclonal antibody : C l q ratio of 3:l. The data can be readily interpreted in terms of a symmetrical model wherein the monoclonal antibody distorts a pair of C l q arms to bring C l r and C l s catalytic subunits together in space. In this context it is known that dimers of IgG will activate complement (Wright et d., 1980) and therefore by implication the binding of two heads on the same C l q molecule is sufficient for C l activation.

23


Cir,Clr,

FIG.13. Alternative model of C1 (adapted from Cooper, 1985). The ClrzClsz tetramer wraps around the arms of Clq rather than being intertwined.

A second suggestion has been that associated antibody may activate C1 by release from the action of C1 inhibitor. According to this theory C1 inhibitor, normally regarded as functioning by actively disassembling activated C 1 to give a ClInh-Clr-Cls-ClInh complex, binds to unactivated C1 and prevents activation. Antibody displaces C l inhibitor and therefore activation proceeds. However, the ability of C 1 inhibitor to bind to unactivated C l has been questioned and earlier observations have been interpreted in terms of C1 inhibitor binding solely to activated C1 (Bianchino et al., 1988). Generally, it seems that C1 can autoactivate by both inter- and intramolecular catalysis, but there has been controversy over the years about how much importance to attach to the mechanisms. A recent study (Hosoi et al., 1987) suggests that intramolecular spontaneous activation is very slow but intermolecular activation can be rapid. In this second case, typically, contaminating proteases convert a little C1 to activated C1, which then cleaves further C1 molecules. C1 inhibitor acts solely on the latter process. Tight C l q binding is not a guarantee of C1 activation. For example, in the case of erythrocytes sensitized with IgG or IgM, both bind C1 equally well but the rate of activation of C1 is far greater in the latter case (Colten et al., 1969). Glutaraldehyde-cross-linked IgG binds C l q as effectively as immune complexes but fails to active C 1 (Folkerd et al., 1980). One of the most detailed investigations of the relationship between C1q binding and C 1 bindinglactivation is that of Bindon et al. (1988b). These authors compared the ability of rat IgG isotypes binding to the human lymphocyte antigens CAMPATH-1, MHC class I, and LCA (leukocyte common antigen) to bind C l q , bind and activate

24


C1, and mediate cell lysis. These antigens have comparable surface densities but show marked differences in lytic ability (CAMPATH-1, MHC class I >> LCA). It was found that C l q binding was roughly proportional to antibody binding and dependent on antibody isotype. However, the lytic antibodies were able to bind and activate more C1 than were the poorly lytic ones. C3 activation and whole lysis patterns then propagated these C1 activation differences. The authors suggested two factors that might play a role in this “antigen effect,” i.e., preferential C 1 binding associated with antigens promoting lysis. One was the possibility of antibody-Clr&lsZ interactions (see below). The other was the mode of presentation of antigen-complexed antibody to C1. The authors noted the lower arm flexibility o f C l relative to C l q (Schumaker et al., 1987), which might place more rigorous binding requirements on the former. Therefore, C l q binding and C1 activation need not correlate when comparing different antigens. For a given antigen, from the limited studies to date, they generally do correlate, e.g., human IgG isotypes binding to hapten-coated red cells (Bindon et al., 1988a), rat IgG isotypes binding to LCA (Bindon et al., 1987), and rat IgG isotypes binding to hapten-coated red cells (Bindon et al., 1990). In particular, human IgG3 binds C l q and activates C1 more effectively than does kG1.

As above, a feature of C1 activation receiving some interest is the possibility of an interaction between antibody and the ClrzClsz tetramer. A review of the literature in 1985 (Burton, 1985)showed no direct evidence for such an interaction involving IgG, and this is still so. Similar conclusions have been reached for IgM (Poon and Schumaker, 1991). However, a number of interesting observations have been made. First, as described initially by Reid et al. (1977), the binding affinity of C l q for Clr&lsZ is increased by about an order of magnitude on binding to immune compIexes (Cooper, 1985). Second, the dissociation rate of C l q in the activated C 1 complex from sensitized red cells is about 10-fold slower than that for C l q alone. This applies to a high antibody density, the difference narrowing at lower densities (Okada et al., 1985).Third, the rate of C1 activation on a red cell surface is dependent on antibody density and is independent of antigen or C1 density (Hughes-Jones et al., 1985). Fourth, the lowered dissociation rate for C l q in activated C 1 is not found when rabbit Facb (lacking the C,3 domains) is used instead of IgG. The dissociation rate of C l q alone is the same for IgG and Facb (Okada et al., 1985). Fifth, the Facbbound activated C 1 complex is more susceptible to C1 inhibitor inactivation than is the IgG-bound complex. The results have been inter-


25

preted to indicate a direct, albeit weak, interaction between ClrzCls2 and the C,3 domains of IgG, with these domains protecting the activated tetramer from C l inhibitor (Okada and Utsumi, 1989). An alternative is that in a C1-activating situation the conformations of both antibody (IgG) and C l q (in C1) become constrained in the mutual interaction. The antibody becomes arranged in a defined array in which Fc-Fc interactions play a part so that Facb is less effective. Array formation is facilitated at higher antibody densities. The C l q undergoes a distortion that increases its affinity for ClrzClsz. The distortion is different depending upon the presence or absence of ClrzClsz bound to C l q , leading to a difference in dissociation rates as described above. These two alternatives are not mutually exclusive, i.e., there could be tetramer binding to antibody and array formation. Finally, recent sedimentation studies show that activated C1 binds much more tightly to IgM than does C l q alone (Poon and Schumaker, 1991). It is argued that the binding of activated Clsz to C l q , either alone or together with activated Clrz, induces a conformational change in C l q that results in additional C l q heads binding to complementary sites on IgM. Cryptic sites on IgM, transitorily exposed by random thermal motion, might be “captured” by activated C1, forming a new complex that could mimic activated C1 attached to cell-bound IgM. C. C4 ACTIVATION, C 3 ACTIVATION, AND CELLLYSIS After C 1 activation, the next step in the classical complement pathway is the activation of C4 through proteolytic cleavage by activated C l s (schematic structures for these molecules are shown in Fig. 14). The major fragment, C4b, can attach covalently to a suitable surface via an activated acyl group. The Fab of IgG (Campbell et al., 1980;Alcolea et al., 1987) and the antigen (Garred et nl., 1990) have been implicated as the surface for C4b deposition in immune complexes. Studies on antibody-coated red cells have found C4 deposition primarily at the cell membrane (Circolo and Borsos, 1982; Bindon et al., 1988a). A number of studies have investigated isotype patterns of C4 activation. Bindon et al., (1988a) looked at the amount of C4b deposited on NIPcoated red cells in complement activation by chimeric human anti-NIP antibodies. They found that IgGl deposited far more C4b than did IgG, under conditions in which IgG3 was more efficient at C l q binding and C 1 activation, as described earlier. The authors showed that the poor C4b deposition was due to poor C4 activation rather than inability to reach the cell surface, i.e., C l s activated by IgG3 appeared inefficient at C4 activation. Possible explanations suggested included easier access of C1 inhibitor in the IgG3 case, favored association of C4

26

D E N N I S R. BURTON A N D JENNY M. WOOF

c1n

m1

c4. c3

FIG.14. Schematic representation of the molecules involved in the classical pathway activation of complement. The molecules are drawn roughly to scale. Dimensions and shapes are taken from Reid and Porter (1981), Perkins (1985), Perkins et al. (1990a,b), and Odermatt et al. (1981).

in the IgGl case, and restricted access ofC4 to C l s in the IgG3 case. No C4b binding was detected with either IgG2 or IgG,. Garred et al. (1989)looked at C4b deposition on NIP-BSA immune complexes formed with the chimeric human anti-NIP antibodies. In particular, they studied the IgG subclass patterns as a function of epitope density (N1P:BSA) and the concentrations of NIP-BSA, antibody, and complement. They found that IgGl and IgG3 were comparable in C4b deposition at higher epitope density and BSA-NIP concentration but that IgGl was far less effective at lower values of these parameters. IgGz produced significant C4b deposition at higher epitope density and BSA-NIP concentration but IgG4 was ineffective under any conditions. In this study, C4b deposition patterns followed those of C l q binding. It is unclear why the two studies produce a different rank order for IgGl and IgGS. It could be that IgGl is generally more effective for cell surface activation and IgG3 for immune complex activation. This interpretation is rendered unlikely by the studies of Michaelsen et al. (1991) on isotype patterns in cell lysis (see below). A more probable explanation lies, at least in part, in the manner of presentation of the hapten, and we shall return to this point.


27

Bindon et al. (1990) looked at C4b deposition on NIP-coated red cells using different isotypes of rat anti-NIP antibody. They found that IgM and IgGa, were both very efficient, reflecting their efficacy in C l q binding and C1 activation. However, of IgG,, IgGZ,, IgG,,, and IgA, which bound C l q and activated C 1 roughly equivalently, only IgG2, was effective at C4 activation under the conditions employed. Fixed C4b interacts with C2 to form the classical pathway C 3 convertase, which in turn cleaves C3. The C3b product can bind covalently to an appropriate surface in a manner analogous to C4b. Again this surface can be antibody (Gaddand Reid, 198lb; Brown et d . ,1983; Takata et al., 1984), the antigen part of an immune complex (Garred et al., 1990), or a cell membrane (Circolo and Borsos, 1984; Bindon et ul., 1988a). The available studies generally indicate a good agreement between efficiency of C4b deposition and efficiency of C 3 deposition and cell lysis or terminal complement complex formation (Bindon et al., 1988a, 1990; Garred et al., 1989). However, there is a striking exception. Clark et al. (1989a) found that a rat monoclonal anti-CD3 IgG21, antibody with one nonfunctional light chain (and therefore monovalent) generated similar levels of cell-bound C3 but gave more lysis than did the parent divalent antibody. It would seem, therefore, that the triggering antibody molecule can influence the complement cascade at a stage after C 3 deposition. Considering the interesting question of the relative abilities of IgGl and IgG3 to mediate cell lysis, the following observations have been made. IgGl is considerably more effective than IgG3 in mediating lysis of hapten (NIP)-coated red cells (Bruggeniann et at., 1987; Bindon et al., 1988a). IgG, is more effective at mediating lysis of dansyl-coated red cells (Dangl et al., 1988). An IgG, directed against a surface antigen is more effective at lysis of lymphocytes than is IgG, (Riechmann et al., 1988). An IgG, against a tumor cell line mediates lysis (Liu et nl., 1987), whereas an IgG, against a different cell line does not (Shaw et al., 1987). The picture apparently emerging is therefore of IgGl being generally superior for lysis. However, the studies of Michaelsen et ul. (1991) on red cells labeled with hapten (NIP)-anti-red blood cell (RBC) F(ab') suggest that the conditions of lysis, i.e., antigen density and antibody and complement concentration, are an important consideration here. Thus, they found anti-NIP IgGl to be superior to IgG3 at high antigen concentration, but this order was reversed at lower antigen concentration. The amount of IgG3 bound was less than that of the other subclasses so its relative potency is even greater. IgGz was only effective at the highest antigen densities. This latter observation is interesting in view of the preponderance of IgGz antibodies in anticarbohydrate responses and the

28


tendency of carbohydrate antigens to be presented in high concentration on the surface of microorganisms. These studies did not determine where in the C l q binding, C 1 activation, and C3 and C4 activation steps the relative efficacy of the subclasses is established. The previous studies on immune complexes (Garred et al., 1989) make the Clq-binding step a likely candidate. Can the various studies on the relative efficacies of IgG3 and IgGl be reconciled simply on the basis of different experimental conditions? We believe probably not and the differences probably arise from differences in presentation of the antigen to the complement system. In particular, Bruggemann et aE. (1987) and Bindon et al. (1988a) used NIP attached to kephalin and therefore, presumably, the hapten was close to the red cell surface, whereas Michaelsen et al. (1991) attached NIP to cell surface proteins directly or via Fab' at greater distances in a different local environment. Therefore, in conclusion, it does not yet seem possible to assert that IgGl or IgG3 is the most effective isotype for complement lysis. This may depend on the antigen involved and the precise experimental conditions employed. Finally, it should be noted that in the case of lysis mediated by a pair of rat IgGs against the human leukocyte common antigen, efficient lysis required the alternative as well as the classical pathway (Bindon et al., 1987).

D. ANTIBODIES AS ACTIVATORS OF THE ALTERNATIVE PATHWAY Alternative pathway activation, in its simplest form, involves the generation of the alternative pathway C3 convertase (C3bBb) from C 3 and factor B in the presence of factor D. The activation, once initiated, has the potential for positive feedback and amplification. Factors I and H in the fluid phase act to regulate C3 convertase and prevent amplification. Properdin acts as a positive regulator. In the presence of a suitable activating surface, stabilization of the C3 convertase tips the scales in favor of amplification. A typical activating surface is that of a microorganism, although it is suggested that associated antibody may also function in this way (reviewed in Ratnoff et al., 1983). In particular, it is often assumed that IgA can activate the alternative pathway, although this has been controversial (Kilian et al., 1988). Recent literature maintains the controversy. Hiemstra et al. (1987) reported that red cells coated with chemically aggregated human IgA were lysed by the alternative pathway. Rits et al. (1988)found that both soluble and insoluble rat IgA immune complexes activated the alternative pathway of homologous rat complement. Hiemstra et al. (1988) reported that human IgA1, IgAs, secretory IgA, and the F(ab')s frag-


29

ment of IgAl coated onto microwells were able to activate the alternative pathway. Fab and Fc fragments were not. Schneiderman et al. (1990) reported that a series of chimeric rabbit mouse IgA antibodies bound to antigen activated the alternative, but not the classical pathway. Valim and Lachmann (1991) found effective alternative pathway activation by immune complexes formed between BSA-NIP and chimeric human IgAz anti-NIP antibodies. The same complexes did not trigger the classical pathway. IgC, was also found to activate the alternative pathway less effectively at high epitope density and equivalence or antibody excess. On the other hand, Imai et al. (1988) found that neither human IgA immune complexes nor covalently cross-linked IgA activated the alternative pathway. Russell and Mama (1989) reported that although human IgA coated onto plastic surfaces activated the alternative pathway in a dose-dependent manner, IgA bound to antigen did not. In contrast, IgG antibodies, either bound to antigen or coated directly onto plastic, activated complement, mainly by the classical pathway. The authors concluded that the complexation of IgA with antigen is insufficient to trigger the alternative pathway and argued rather that denaturation plays a key role in IgA activation. Jarvis and McLeod Griffiss (1989) found that human IgAl was unable to mediate alternative pathways lysis of Neisseria meningitidis but did mediate classical pathway lysis. The other antibody often associated with alternative pathway activation is rabbit IgG when C3b has been found bound to Fab and the C,3 domain of IgG (Gadd and Reid, 1981a; Anton et al., 1989).

IV. Human leukocyte Fc Receptors

Receptors specific for the Fc region immunoglobulins are found on the surface of a variety of human leukocytes. The presence of F c receptors confers on these immune cells the ability to mediate a number of effector mechanisms important in the humoral response. Recent cloning and sequencing studies have revealed that the large majority of mammalian Fc receptors have evolved as part of the immunoglobulin gene superfamily. Here those Fc receptors specific for IgG (Fc, receptors) will be discussed first, followed by those specific for other classes of immunoglobulin.

A. Fc, RECEPTORS Three classes of human Fc, receptor have thus far been described: Fc,RI, Fc,RII, and Fc,RIII (Unkeless et al., 1988). All three ap-

30


pear capable of mediating a number of protective functions against antibody-coated infectious agents (Pound and Walker, 1990; Van de Winkel and Anderson, 1991). Studies using hybridomas expressing surface anti-Fc,R monoclonal antibodies as targets have demonstrated that Fc,RI, Fc,RII, and the macrophage/NK cell form of Fc,RIII can mediate antibody-dependent cell-mediated cytotoxicity (ADCC) (Fanger et al., 1989). These same receptors can also mediate phagocytosis (Huizinga et al., 198%; Anderson et al., 1990a), whereas Fc,RI, Fc,RII, and the neutrophil form of Fc,RIII have been shown to trigger an oxidative burst (Anderson et al., 1986; Crockett-Torabi and Fantone, 1990; B. A. M. Walker et al., 1991). 1. Fc,RI Human Fc,RI (CD64) is a 72-kDa glycoprotein expressed constitutively on monocytes and macrophages. It may be induced on neutrophils in vitro by treatment with interferon-y (IFN-y) (Perussia et al., 1983).IFN-y treatment also up-regulates Fc,RI expression on mononuclear phagocytes on the Fc,RI-bearing monocytic cell lines, U937 and HL60 (Guyre et al., 1983). Three cDNA clones for human Fc,RI have been isolated using a ligand-affinity cloning technique (Allen and Seed, 1988, 1989). Two clones represent polymorphisms whereas the third has a shorter predicted intracytoplasmic domain (Fig. 15).In each case, the deduced amino acid sequence indicates an integral membrane protein with a single hydrophobic membrane-spanning region. The extracellular portion is composed of three immunoglobulin-like domains, the first two of which exhibit homology to the two extracellular domains of human Fc,RII and Fc,RIII (see later). The third domain, nearest to the membrane, is less closely related.

2 . Fc,RI-IgG Interaction at the Molecular Level Fc,RI is sometimes referred to as the high-affinity Fc, receptor, because of the three human receptors it is the only one displaying appreciable affinity for monomeric IgG. It binds monomeric human IgGl and IgG3 with a K , of -5 X lo8 M - l (Fries et al., 1982; Kurlander and Batker, 1982). The affinity for human IgG4 is approximately 10-fold lower and human IgGz does not bind (Woof et al., 1986). Human Fc,RI appears to bind aggregated IgG, at least in heat or chemically crosslinked forms, with similar affinity to monomeric IgG (Cosio et al., 1981; Carter et al., 1982; Kurlander and Batker, 1982; Woof et al., 1986).This finding, together with the fact that the high serum concentration of monomeric IgG presumably results in constant saturation of

31


FcERI

Fc R I I

II

Fc R l l l

r

FIG.15. Schematic representation of the structures of human Fc, receptors. T h e extracellular immunoglobulin-like domains, shown as oval shapes, each have an internal disulfide bond. No structural information is yet available for cytoplasmic domains, but their relative lengths are indicated. PIG, Phosphatidylinositol-glycan.

FcyRI, raises the question of how FcyRI is able to distinguish antibodycoated infectious agents. One possibility is that FcyRI may play an important role at tissue sites where monomeric IgG is limiting. Alternatively, up-regulation of FcyRI expression by IFN-.), at inflammatory sites may influence the function of the receptor. The high affinity of FcyRI for monomeric IgG has facilitated attempts to localize the FcyRI interaction site on IgG. Earlier studies had suggested that the site lay in the Cy3domain of IgG (Okafor et al., 1974; Ciccimarra et al., 1975). However, the use of highly immunoaffinity-purified IgG fragments and domain-deleted IgG paraproteins indicated that this was not the case (Woof et al., 1984). Further, loss of the N-linked carbohydrate moieties from the Cy2 domain of IgG was shown to result in a marked reduction (>50-fold) in affinity for FcyRI. This suggested an important role for the C y 2 domain in the interaction, because aglycosylation was deemed more likely to perturb Cy2 structure than that of Cy3 (Leatherbarrow et al., 1985; Walker et al., 1989a). As mentioned earlier, recent NMR studies have verified this assumption (Matsuda et aZ., 1990). Direct evidence for the involvement of the Cy2 domain of IgG was provided by experiments in which a panel of antihuman IgG monoclonal antibodies were assessed for ability to inhibit the IgG-FcyRI inter-

32


action and to bind to receptor-bound ligand (Partridge et al., 1988). Only those monoclonal antibodies recognizing epitopes at the Nterminal end of the Cy2domain both blocked IgG binding to FcyRI and failed to bind receptor-bound IgG. Recent “domain swap” experiments in which the Cy2 and Cy3domains of human IgG1, either singly or together, substitute for one or both of the homologous C,3 and C,4 domains of mouse IgE have generated chimeric IgG/IgE molecules (Shopes e t al., 1990). In the series of mutants produced, all those lacking Cy2 domains did not inhibit the FcyRI-IgG interaction, whereas three out of four IgG/IgE constructs possessing Cy2 domains did inhibit. These results confirm the importance ofthe Cy2domain for interaction with FcyRI. Futher, calculations of the relative energetic contributions of each domain revealed that the Cy2domain contributes about 73% of the overall drop in free energy seen on binding. In contrast, the Cy3 domain was shown to contribute maximally about 25% of the free energy drop upon binding. This is consistent with a requirement for the presence of the Cy3 domain of IgG for FcyRI binding, earlier suggested by the lack of reactivity of a C4-deleted human IgGl paraprotein with the receptor (Woof et al., 1984). The C$3 domains would appear to serve to stabilize the overall structure of the Fc, necessary for optimal FcyRI binding. Another approach assessed the ability of various IgG molecules, from different species and of different subclasses, to inhibit the interaction of radiolabeled human IgG with FcyRI. Amino acid sequence comparison of the Cy2 domains of these IgGs then outlined potential FcyRI-binding sites that fulfilled the requirements of solvent accessibility and conservation in only the tight binding IgGs. This procedure highlighted a region at the N-terminal end of Cy2comprising residues Leu 234-Ser 239 which appears to be critical for interaction with FcyRI (Woof et al., 1986). This region, although encoded by the Cy2 exon, structurally forms part of the lower portion of the hinge, lying mostly beyond the extent of the solved crystal structure of Fc, which stops at residue 238 (see earlier). The proposed site, Leu 234-Leu 235-Gly 236-Gly 237-Pro 238Ser 239, is present in all IgG isotypes with high affinity for FcyRI, namely, human IgG1 and IgG3, mouse IgGz,, rat IgGzb, and rabbit IgG. In the weaker binding human IgG4, residue 234 becomes Phe and in mouse IgGzb residue 235 becomes Glu. This latter IgG displays little or no binding to human FcyRI. However, a mutant mouse IgGzb molecule, generated b y site-directed mutagenesis in which Glu 235 is converted to Leu, displayed a greater than 100-fold increase in affinity for the receptor. Indeed, Scatchard analysis of direct binding measure-


33

ments showed it to have an affinity comparable to that of human IgGI (Duncan et al., 1988). More recent niutagenesis experiments have concentrated on the high-affinity human IgG3 subclass. Substitutions at residues 234-237 inclusive resulted in reductions in ability to inhibit the IgG-FcyRI interaction (Lund et al., 1991). The most marked effect is seen with substitution at position 235, with replacement of the Leu with Glu giving a >lOO-fold decrease in affinity. Substitution of Ala for residues Leu 234, Gly 236, and Gly 237 generated antibodies with affinities reduced -4-fold, -4-fold, and -30-fold, respectively. Rosetting studies with U937 cells confirmed these binding inhibition measurements. Taken together, these findings confirm the importance of the lower hinge region in the interaction of IgG with FcyRI (Fig. 2). Evidently, only a single site (i.e., one heavy chain) is required for interaction with a single FcyRI molecule (O’Grady et ul., 1986; Koolwijk et al., 1989). Studies using an array of further IgG mutants appear to be in broad agreement with the above-cited work (Canfield and Morrison, 1991). A mutant human IgGz molecule, in which the Cy3 domains of human IgG3 were substituted for its own, did not bind to Fc,RI. A human IgG3 bearing the Cy3 domains of hunian IgG2, however, did bind FcyRI, with an affinity comparable to that of wild-type IgG3. Thus, the critical importance of the Cy2 domains in the interaction is reaffirmed. Substitution of Glu for Leu at position 235 in human IgG3 again resulted in a >lOO-fold reduction in affinity for FcyRI. Conversion in human IgG3 of residue 234 to Phe, as found in human IgG4, generated a mutant molecule with affinity equivalent to that of wild-type human IgG4. The reciprocal experiment, in which Phe 234 in human IgG4 was converted to Leu, generated an IgG4 molecule with increased affinity for FcyRI, about threefold lower than that of human IgG3. This slight shortfall in affinity compared to human IgG3 prompted the authors to search elsewhere in the C,2 domain for amino acid differences that might serve as an explanation (Canfield and Morrison, 1991). Thus it was proposed that residue 331 (Pro in all tight-binding IgGs but Ser in human IgG,), located on a loop lying close to the lower hinge site, might make some contribution to the FcyRI interaction. The substitution experiments, however, yielded somewhat inconclusive results and further investigation into the possible role of this loop would be interesting. A second bend, lying close to the lower hinge region in molecular models, may also be of importance. Proton NMR revealed that on aglycosylation of Fc the minor structural changes in the Cy2 domain were sensed predominantly by His 268, which lies in this second bend (Matsuda et aZ., 1990). The marked reduction in affinity for Fc,RI

34


accompanying aglycosylation may be explained in terms of perturbation of this lower hinge-proximal loop (Lund et al., 1990). Turning to the interaction site on the receptor, presently rather little information is available. The third domain of Fc,RI shares less homology with the two extracellular domains of the low-affinity receptors, Fc,RII and Fc,RIII, than do the first two domains. This perhaps suggests a role of this third domain in conferring high affinity on Fc,RI. Indeed, preliminary studies introducing point mutations into the third domain are reported to confirm this (Allen and Seed, 1989). Further, studies on the mouse homologue, mouse Fc,RI, which also possesses three extracellular domains, suggest that domain 3 confers high affinity and specificity on domains 1 and 2, which form a low-affinity IgGbinding motif (Hulett et al., 1991). Figure 16 shows a model of the interaction of human Fc,RI (modeled as three IgG domains) with an IgGl molecule simultaneously binding to antigen (hemagglutinin on the surface of a virally infected cell). Although admittedly speculative, such models serve to demonstrate the relative sizes of the molecules involved and highlight conformation constraints imposed by the location of interaction sites on the molecules. 3. Fc,RZl Human Fc,RII (CD32), a 40-kDa glycoprotein, has the most widespread distribution of the three human receptors for IgG, being present on monocytes, macrophages, eosinophils, platelets, neutrophils, basophils, and B cells. Several cDNA clones coding for this receptor predict a transmembrane molecule with two immunoglobulin-like extracellular domains (Stuart et al., 1987, 1989; Hibbs et al., 1988; Stengelin et al., 1988). Fc,RII is encoded by a minimum of three genes (IIA, IIB, and IIC) that yield six distinct transcripts (Brooks et al., 1989). Gene IIB gives rise to three transcripts, termed Fc,RIIbl, Fc,RIIb2, and Fc,RIIb3, by alternative slicing of cytoplasmic exons or of signal sequence exons. Transcripts of IIB have been demonstrated by a combination of Northern blotting and polymerase chain reaction (PCR)amplification after reverse transcription in macrophages, monocytes, and B lymphocytes, with lower levels found in neutrophils (Brooks et al., 1989). Fc,RIIb expression has also been demonstrated in placental trophoblasts (Stuart et al., 1989). The three other Fc,RII transcripts (two Fc,RIIa transcripts and one Fc,RIIc) are derived from two further genes referred to as Fc,RIIA and Fc,RIIC (Brooks et al., 1989; Stuart et al., 1989). Alternative polyadenylation in Fc,RIIA gives rise to two transcrints. All three molecules are highly homologous to Fc,RIIb, with differences being restricted to

FIG. 16. Antibody as adaptor linking target and effector cells. In the lower part ofthe picture, the target cell infected by influenza virus is shown expressing two hemagglutinin molecules at its surface. Hemagglutinin (Wilson et ul., 1981) is recognized by the Fab arms of the IgG molecule. In turn, the IgG molecule is recognized at the site seen in Fig. 2 by an Fc receptor molecule anchored to the cell membrane ofthe effector cell. The IgG molecule as shown is dislocated in the sense that the Fc part is in a plane roughly at right angles to that containing the Fab arms. The Fc receptor, FcRI, is known from the work of Allen and Seed (1989)to consist extracellularly of three immunoglobulin-like domains. The structure as shown and antibody binding to the outermost domain are speculative. The effect of opening of the Fab arms (opening the Y) and of dislocation is that the antibody molecule tends to draw the target and effector cells closer together than many other possible arrangements. The use of one of the inner domains of FcRI for antibody binding would enhance this effect still further.

36


the signal sequence and part of the cytoplasmic domain in Fc,RIIa, and to a portion of the cytoplasmic domain in Fc,RIIc. Ignoring differential polyadenylation, the Fc,RIIa and Fc,RIIc transcripts are almost identical apart from their signal sequences. The leader sequence of Fc,RIIc is homologous to that of Fc,RIIb, whereas that of Fc,RIIa shows homology to that of human Fc,RIII (see later). Thus, IIC may have evolved initially by gene duplication and mutation from an ancestral IIB gene, with a later recombination event between IIC and Fc,RIII genes giving rise to IIA. Transcripts of IIA have been found readily in monocytes and neutrophils but less prominently in lymphoid cell lines, whereas sensitive PCR amplification techniques have detected Fc,RIIc transcripts in neutrophils, monocytes, and B cells (Brooks et al., 1989). Thus, there appears to be preferential expression of Fc,RIIA in neutrophils and Fc,RIIB in lymphocytes, with both forms being expressed in monocytes (Ravetch and Anderson, 1990). The mature protein products predicted from the cDNAs of Fc,RIIa, Fc,RIIc, and Fc,RIIb are closely related. The predicted Fc,RIIa and Fc,RIIc proteins are virtually identical (>95%). The extracellular and transmembrane domains of the Fc,RIIb receptors share a high degree of homology with the predicted FcyRIIa and FcyRIIc receptors. However, beyond the first 10-12 residues of the intracellular regions the similarity between Fc,RIIb and Fc,RIIa/IIc ends and highly divergent cytoplasmic tails result (Fig. 15).The conserved nature of the extracellular domains together with the diversity of the intracytoplasmic regions suggests that this family of receptors may mediate an array of different functions in response to recognition of the same ligand. 4 . Fc,RII-lgG Interaction at the Molecular Level The affinity of human Fc,RII for monomeric IgG is too low to be easily experimentally determined ( K , < 1 x lo7 M - ’ ) (Karas et al., 1982; Kurlander et al., 1984) but binding to complexed IgG is much tighter [for dimers of human IgC1, K , (2-5) X lo7 M-’1 (Karas et al., 1982; Van d e Winkel and Anderson, 1991). Recently, the affinity of Fc,RII for complexed IgG has been shown to increase after treatment with proteolytic enzymes such as pronase and elastase (Van de Winkel et al., 1989b).This process may serve to “activate” Fc,RII at inflammatory sites (Tax and Van de Winkel, 1990). Human Fc,RII is specific for complexes of human IgGl and IgG3 and appears not to bind IgG4 (Karas et al., 1982; Walker et al., 1989b; Van de Winkel and Anderson, 1991).Reports on the affinity of Fc,RII for complexed human IgGz appear to give conflicting results, with some

-


37

workers reporting binding (Van de Winkel and Anderson, 1991) and others not (Walker et al., 1989b). An explanation, now emerging, is that the ability to bind human IgG2 appears to reside with only one allelic form of Fc,RII, termed low responder (see below) (Warmerdam et al., 1991). Fc,RII also has affinity for mouse IgG2b but not for mouse IgG2,. A functional polymorphism of Fc,RIIa has been demonstrated by different assays on monocytes (Tax et al., 1983;Leeuwenberget al., 1987; Van d e Winkel et al., 1987, 1989a; Debets et al., 1990), neutrophils (Leeuwenberg et al., 1990; Gosselin et al., 1990), and alveolar macrophages (Kindt et al., 1991).The two allelic forms of Fc,RIIa involved differ in their ability to bind complexed mouse IgG1-the highresponder (HR) form binds whereas the low-responder form (LR) does not (Anderson et al., 1987). A monoclonal antibody, 41H16, has been shown to be specific for the HR form (Gosselin et al., 1990). Recently, cDNA clones coding for HR and LR forms of Fc,RIIa have been isolated and sequenced (Clark et al., 198913; Warmerdam et al., 1990).The alleles appear likely to differ in just two residues; residue 27 in domain 1 is Gln in HR and Trp in LR forms, and residue 131 in domain 2 is Arg in HR and His in LR forms. Domain swap experiments between the two forms indicate that Arg 131 is a critical requirement for the binding of mouse IgG, to the HR form (Warmerdam et al., 1991). Arg 131 also appears to form part of the epitope recognized by 41H16. Further, these studies indicate that His at residue 131 in the LR receptor is essential for the binding of this form to human IgG2. Loss of the N-linked carbohydrate moieties from the C,2 domain of human IgG, and IgG3 molecules results in a marked reduction in their ability to interact with Fc,RII (Walker et al., 1989b). This effect is reminiscent of that seen with Fc,RI and therefore suggests the importance of the C,2 domain in the Fc,RII-IgG interaction. The localized perturbation on aglycosylation particularly in the vicinity of His 268, as assessed by the proton NMR methods mentioned earlier (Matsuda et al., 1990), perhaps further points toward a similarity between Fc,RI and Fc,RII interaction sites. In order to assess this possibility, the earlier described panel of mutant IgG3 molecules with point substitutions in residues 234 to 237 was utilized. The ability of these antibodies to form rosettes with the Daudi cell line, which expresses Fc,RII only, was determined and compared to that of wild-type human IgG3 (Lund et al., 1991).Substitutions of Leu + Ala 234, Leu + Ala 235, Leu + Glu 235, Gly + Ala 236, and Gly --$ Ala 237 in each case reduced the number of rosettes formed with Daudi cells. The most marked effect was seen with Leu+ Ala substitution at residue 234, where rosette formation was reduced to IgGz, > IgGzb >> IgGl (Kipps et a[., 1985; Anasetti et al., 1987). Fc,RIII is able to interact with certain lectin molecules, probably via its high-mannose-type oligosaccharides. Thus, the phagocytosis of Con A-treated erythrocytes by neutrophils is inhibitable by aggregated


41

IgG, 3G8, and monosaccharides such as D-mannose (Salmon et al., 1987; Kimberly et al., 1989).Thus, the oligosaccharide moieties recognized by Con A have been suggested to contribute to the integrity of the IgG-binding site on Fc,RIII. Turning to the interaction site on IgG, early experiments indicated that the integrity of the Fc region was required for binding to neutrophi1 Fc,RIII. Hence, neither a pFc’ fragment nor a tryptic Cy2 domain fragment could inhibit the interaction (Barnett-Foster et al., 1978).The importance of the C,2/C,3 interface was suggested by the reported ability of protein A, which binds in this region, to block the binding of IgG to the receptor (Barnett-Foster et al., 1982). However, other workers found an anti-IgG monoclonal antibody, which recognizes an epitope in the Cy2/C,3 interface region, unable to inhibit the binding of IgG to FcyRIII on monocyte-depleted peripheral blood mononuclear cells (Sarmay et al., 1985). This same study, using a panel of anti-IgG monoclonal antibodies and the same effector cells, proposed that the Fc, receptor involved, assumed to be Fc,RIII, binds to a region in the C,3 domain of IgG and that a second region in the C,2 domain is critical for triggering of ADCC via this receptor. More recently, aglycosylation of Fc of human IgG3 has been reported to render it incapable of mediating ADCC via FcyRIII on K cells (T plus Null cells) (Lund et al., 1990). Currently, however, our understanding of how IgG and Fc,RIIII interact at the molecular level seems to require further clarification. B. Fc, RECEPTORS Two classes of receptor specific for the Fc region of IgE, termed Fc,RI and Fc,RII, have been described and will be discussed here in turn.

1 . Fc,RI Human Fc,RI is found exclusively on the surface of mast cells and basophils, where it binds IgE with high affinity. Upon aggregation of Fc,RI-IgE complexes by interaction with multivalent antigen the cell degranulates, releasing mediators of the allergic response (Metzger et al., 1986; Metzger, 1988). Fc,RI is a multichain glycoprotein consisting of one a subunit, one p subunit, and two y subunits (Metzger et al., 1983; Alcarez et al., 1987) (Fig. 17).The a subunit, which contains the binding site for IgE, is an integral membrane protein with a single membrane-spanning region and two immunoglobulin-like extracellular domains. cDNA clones for the human a subunit have been isolated (Kochan et al., 1988; Shimizu et al., 1988). The predicted product

42


FIG.17. Schematic representation of Fc,RI. The receptor is a tetramer of one Q subunit, one /3 subunit, and two y subunits.

shares a considerable degree of homology to human Fc,RIII-A. Between species, the most highly conserved region is the transmembrane portion, with the cytoplasmic tail showing most divergence (Kinet, 1989).These observations may suggest that the Q subunit intracellular region is not involved in a key function, whereas the membranespanning portion serves some specific role, perhaps involving interaction with the p or y subunits (Kinet and Metzger, 1990).Indeed, recent results from experiments expressing wild-type and mutant forms of the rat Fc,RI in COS cells support these suggestions and will b e discussed a little later. No sequence data is currently available for the human /3 subunit. However, cDNA clones of both the rat and mouse p chains have been isolated (Kinet et al., 1988; Ra et al., 1989b). The predicted protein products share an extremely high degree of identity (83%), suggesting that the human counterpart may also do so. The predicted topology of the p subunit comprises four membrane-spanning regions with both Nand C-termini in the cytoplasm. Both cDNA and genomic clones of the human y subunit have been isolated (Kiister et al., 1990). The y subunit appears to b e well conserved, with 86% identity between the predicted polypeptides of hu-


43

man, mouse, and rat y chains. This small third Fc,RI subunit spans the membrane once to give a short N-terminal extracellular region ofjust 5 amino acids and a longer cytoplasmic tail (42 amino acids) (Kuster et at., 1990).As mentioned earlier, t h e y subunit displays homology to the 5 chain of CD3 and both are members of a family of proteins that associate as disulfide-linked dimers (Orloff et al., 1990).Recently it has been shown that human CD35 can substitute for the y subunit in assembly and functional expression of rat Fc, RI in a Xenopus oocyte expression system (Howard et al., 1990). Similarly, both y and 5 dimers may associate with human Fc,RIII-A (see earlier). A series of experiments in which component subunits of the receptor were cotransfected into COS cells to reconstitute rodent, human, or chimeric Fc,RI molecules have yielded interesting information on requirements for efficient cell surface expression. In both the rat and mouse systems, cotransfection of the a , p, and y subunits are necessary for expression of Fc,RI at the cell surface (Blank et al., 1989; Ra et al., 1989b). However, cotransfection of human a and y subunits is sufficient to give efficient expression of a molecule capable of binding IgE (Kuster et al., 1990).Addition of rodent p chain did not increase expression efficiency. Chimeric receptors of human a subunit plus rat p and y of either rat or mouse have also been expressed and shown to bind IgE with affinity comparable to that of Fc,RI on normal cells (Miller et al., 1989; Ra et al., 1989b). Site-specific mutagenesis of the a, p, and y subunits of a rat Fc,RI expressed in COS cells revealed that removal of cytoplasmic tails from any or all of the subunits had little effect on surface expression. However, even minor changes within the transmembrane regions led to reduced expression (Varin-Blank and Metzger, 1990). In the rat receptor, therefore, the membrane-spanning regions of the subunits appear critical for optimal expression. By contrast, cotransfection of human a subunit and a truncated rat y subunit lacking a cytoplasmic domain resulted in no expression of human a, suggesting that the human and rodent receptors may assemble differently (Varin-Blank and Metzger, 1990). Models to describe the molecular interaction between transmembrane segments of the subunits are now emerging (Varin-Blank and Metzger, 1990; Farber and Sears, 1991).

2 . Fc,RZ-lgE lnteraction at the Molecular Level Human Fc,RI binds human IgE with high affinity ( K , - 1 x 10" and interacts somewhat more weakly with rat and mouse IgE (Conrad et al., 1983).A chimeric a subunit consisting of the extracellular portion of the human a subunit fused to the transmembrane and M-l)

44


cytoplasmic domains of the p55 IL-2 receptor has recently been shown to bind IgE with high affinity. Thus, the interaction site on the receptor appears to lie in the extracellular portion of the a subunit, with no apparent contribution from the other subunits (Hakimi et al., 1990). Using the same chimeric receptor, it has been further demonstrated that monoclonal antibodies recognizing epitopes in the second domain ofthe a subunit (adjacent to the cell membrane) can inhibit the interaction with IgE (Riske et al., 1991). This may suggest that IgE also binds directly to the second a-subunit domain. Alternatively, the inhibiting antibodies may exert their effect by steric hindrance of a site distal to their point of recognition or by induction of conformational changes in the IgE-binding region. The carbohydrate moieties of the a subunit appear not to be involved in the IgE interaction because rat Fc,RI on basophilic leukemia cells cultured in the presence of tunicamycin, an inhibitor of N-linked glycosylation, are still able to bind IgE (Hempstead et al., 1981). Turning to the site of interaction on IgE, an excellent review of the various studies attempting to localize the binding site appeared some 3 years ago (Metzger, 1988). Here, therefore, we will concentrate on advances made since then. The binding site for Fc,RI is known to lie in the Fc portion of IgE, which is composed of the paired C,2, C,3, and the C,4 domains (Fig. 8) (Ishizakaand Ishizaka, 1975).The importance of the C,2/C,3 interface was earlier suggested by the finding that interaction of rat IgE with rat Fc,RI protected that region of the antibody, in particular from tryptic proteolysis (Perez-Montfort and Metzger, 1982). Further mapping of the human Fc,RI site has made use of recombinant peptides of human IgE expressed in Escherichia coli. The lack of glycosylation of Fc, fragments generated in this manner does not impede their ability to interact with Fc,RI (Ishizaka et al., 1986). This is consistent with an earlier report that nonglycosylated intact rat IgE still bound to rat Fc,RI (Kulczycki and Vallina, 1981). A monomeric recombinant peptide of 76 amino acids, comprising residues 301-376 spanning the C,2/C,3 interface, was reported to interact with human Fc,RI with an affinity similar to that of intact IgE (Helm et al., 1988). Use of a series of peptides subsequently demonstrated that residues 363-376 in the above peptide are not essential for Fc,RI binding (Helm et al., 1989). Further, an epitope lying within this peptide recognized by an anti-IgE monoclonal antibody demonstrates a sensitivity to heat and alkylation similar to that displayed by IgE when interacting with Fc,RI (Del Prado et al., 1991). A recent report suggests that an octapeptide lying at the C,3/C,4 interface within the peptide above (residues 345-352) is capable of


45

inhibiting histamine release by human basophils (Nio et al., 1990). However, this result should perhaps be questioned, as the octapeptide only appears to inhibit at molar concentrations several orders of magnitude greater than those required with intact human myeloma IgE. A series of resonance energy transfer studies provide further evidence for the possible involvement of the Ce2/C,3 interface region in Fc,RI binding (Holowka and Baird, 1983; Holowka et al., 1985; Zheng et al., 1991).Distances between fluorescent donor probes, placed at specific sites on IgE and anti-IgE antibodies, and acceptor probes at the cell membrane surface were determined. These measurements indicate IgE bound to Fc,RI has a bent conformation, with the C,2 and C,3 domains lying closest to the cell membrane and therefore presumably to the receptor (see later). A number of recent studies have utilized a panel of chimeric IgE molecules to assess the relative contribution of each domain. One group found that mutant mouse IgE molecules lacking either 45 amino acids from the carboxy end of C,3 or almost the entire C,4 domain no longer bound to rat Fc,RI (Schwarzbaum et al., 1989). Further, the sites of recognition of two anti-IgE antibodies earlier shown to inhibit the IgE-Fc,RI interaction (Baniyash and Eshhar, 1984; Baniyash et al., 1988)were localized to the C,3 domain. A third anti-IgE antibody, known not to inhibit receptor binding, was shown to bind the C,4 domain. These observations were interpreted as indicating that the C,3 domain plays a key role in Fc,RI recognition, whereas the C,4 domain, although not directly involved in the interaction, serves to stabilize the conformation of Fc, necessary for Fc,RI binding. A further study made use of the ability of human IgE to bind to human Fc,RI but not to rat Fc,RI, in contrast to the reactivity of mouse IgE with both receptors (Nissim et al., 1991). Chimeric human/mouse IgE molecules, in which single or multiple mouse domains substituted for human ones, were assessed for binding to either rat Fc,RI or a reconstituted human Fc,RI expressed in COS cells. When the C,2 domains of human IgE were replaced by those of mouse IgE, the resultant molecule bound human but not rat Fc,RI. In contrast, a chimeric human IgE containing mouse C,3 domains (CHM3) bound both receptors. Furthermore, deletion of C,2 from CHM3 produced no impairment of binding to rat Fc,RI. Again, these results suggest that the C,3 domain is the principal region involved in interaction with Fc,RI (Nissim et al., 1991). A second group of investigators has generated chimeric mouse IgE in which one or more IgE domains are substituted by homologous regions from human IgG, (Weetall et al., 1990).An IgE molecule in

46


which the C,4 domains were replaced by IgG Cy3 domains bound rat Fc,RI with affinity comparable to that of wild-type mouse IgE. All other chimeric molecules tested did not contain both C,2 and C,3 and were unable to bind rat Fc,RI. The favored interpretation was that both C,2 and C,3 domains are necessary for the binding interaction (Weetall et al., 1990). However, the observations mentioned earlier (Nissim et al., 1991) suggest that the role of C,2 here may be in stabilizing the conformation of C,3 necessary for optimal Fc,RI interaction. A substituted human IgGl hinge may not be able to perform this role adequately. This possibility may explain the inability of a chimeric human IgG molecule, in which mouse C,3 domains substitute for Cy2 domains, to interact with rat Fc,RI (Weetall et al., 1990). The current weight of evidence, therefore, would seem to support the idea that the recognition site for Fc,RI lies (1)within the region encoded by the C,3 exon and (2) within the peptide Gln 301-Leu 363. The region of Fc, fulfilling both these criteria lies between Asp 330 and Leu 363. Of particular pertinence is the emergence of a new molecular model for the Fc region of IgE (Helm et al., 1991). This revised model incorporates the finding that the inter-C,2 disulfide bonds involving Cys 238 and Cys 241 are parallel, rather than crossed as in the earlier models (Padlan and Davies, 1986; Pumphrey, 1986). The most pronounced consequence of modeling parallel disulfide bridges is the appearance of an exposed segment of approximately 2.4 nm in length, comprising residues 329-335 lying between C,2 and C,3 (Fig. 3).This region may constitute the structural equivalent of the lower hinge region in IgG and hence it is tempting to speculate that the Fc,RI interaction site may lie here. Should further mutagenesis experiments verify this possibility, a common theme of Ig-like Fc receptor domains interacting with a flexible lower hinge region (or its equivalent) in immunoglobulins may emerge. 3. Fc,RZI The second class of receptor for the Fc region of IgE (Fc,RII or CD23) has lower affinity for its ligand than does Fc,RI and hence is sometimes referred to as the low-affinity receptor. Fc,RII is present on inflammatory cells, including monocytes, eosinophils, and platelets, and on B lymphocytes. cDNA clones coding for human Fc,RII have been isolated and shown to bind IgE when expressed in mammalian cell systems (Kikutani et al., 1986; Ikuta et al., 1987; Ludin et al., 1987). Unlike all other leukocyte Fc receptors described here, Fc,RII is not related to the immunoglobulin gene family of proteins. Rather it displays homology to a family of animal lectins, which includes the human and rat asialoglycoprotein receptors.


47

On SDS gels human Fc,RII has a molecular weight of about 43,000 and is composed of a 321-amino acid polypeptide core of -36,000 molecular weight and both 0- and N-linked oligosaccharides (Delespesse et al., 1989). The receptor spans the membrane once in a rather unusual orientation, because the short 23-amino acid N-terminal domain lies inside the cell and the much longer C-terminal region is found to the exterior. The domain exhibiting homology to animal lectins spans about 120 residues of the extracellular portion and includes three cysteine pairs. Close to the C-terminus lies an Arg-Gly-Asp (RGD)sequence in reverse, i.e., DGR (Kikutani et al., 1986). A number of molecules that bind to the integrin family of receptors contain this RGD motif, suggesting that FceRII may be able to interact with adhesion molecules (Gordon et al., 1989). More recently, a second species of human Fc,RII, termed Fc,RIIb, has been identified and differs from the earlier described receptor (Fc,RIIa) only in the first few amino acids at the N-terminus (Yokota et al., 1988). mRNA for Fc,RIIa is constitutively expressed in B cells alone. In contrast, Fc,RIIb is expressed in monocytes, eosinophils, and B cells only after stimulation with IL-4 (Yokota et al., 1988). Functionally, Fc,RIIa appears to be involved in the regulation of €? cell development (Gordon et al., 1989), whereas Fc,RIIb plays a role in IgEdependent cytotoxity against parasites such as schistosomes (Capron and Dessaint, 1985). 4. Fc,RZI-IgE Interaction at the Molecular Level Monomeric human IgE binds to Fc,RIIb with a K , of about 3 x 107 M-' (Anderson and Spiegelberg, 1981; Joseph et al., 1986). Dimeric IgE may have a slightly higher affinity for the receptor than monomers, at least in the rat system (Finbloom and Metzger, 1982). The natural occurrence of soluble proteolytic cleavage products of Fc,RII, termed sFc,RII, soluble CD23, or IgE-BF, which are capable of binding IgE, indicates that the site of interaction lies in the Cterminal part of the extracellular portion (Letellier et al., 1989). Indeed, expression of recombinant soluble Fc,RII has localized the binding site within a 172-amino acid stretch at the C-terminus (Uchibayashi et al., 1989). This C-terminal portion also includes the region of homology with animal lectins. Despite the implication that Fc,RII may therefore interact with the oligosaccharide chains of IgE, this is not the case. Certain recombinant human echain fragments synthesized in E . C a l i and therefore devoid of carbohydrate are still able to bind to human Fc,RII (Vercelli et al., 1989).Further, enzymatically deglycosylated myeloma IgE interacts with the receptor. In fact, it appears to bind slightly more tightly to the receptor than does the

48


parent IgE. Moreover, high concentrations of mono- and disaccharides do not inhibit the interaction between mouse IgE and Fc,RII. The interaction does, however, share a dependence on calcium and p H with other lectin proteins (Richards and Katz, 1990). Use of recombinant Fc, fragments mentioned above has allowed further localization of the Fc,RII interaction site on IgE (Vercelli et d., 1989). A fragment (rE2-4) corresponding to the whole of the Fc, (paired C,2, C,3, and C,4 domains) bound to Fc,RII on a human B cell line with comparable affinity to intact IgE. A shorter fragment, comprising paired C,3 and C,4 domains, also bound but with much lower affinity. An intermediate peptide (rE2'-4) with the 30 C-terminal residues of C,2 plus C,3 and C,4 was almost as active as the whole Fc,. In contrast, removal of C,4 (leaving C,2 plus C,3) generated an inactive peptide. Thus, receptor recognition appears to require the presence of all three domains. The C,4 domain, however, may be substituted for by the C,3 domain of mouse IgGzb without considerable loss of reactivity with Fc,RII. The indirect role of the C,4 domain would thus appear to be to promote the dimerization of the two L chains, necessary for receptor binding. This is borne out by the demonstration that replacement of Phe 506, lying at the interface between the paired C,4 domains, with Arg generates a monomeric form of the Fc, chain that is unable to bind to Fc,RII (Vercelli et al., 1989).Unlike Fc,RI, binding of Fc,RII therefore has a clear requirement for both heavy chains in Fc,. One possibility is that two Fc,RII molecules may interact simultaneously with one IgE molecule. The observation that dimers of mouse Fc,RII may preexist adds weight to this idea (Peterson and Conrad, 1985). Anti-IgE monoclonal antibodies against epitopes lying between residues 307-315 in C,2 and residues 367-370 in C,3 exhibit marked abilities to inhibit the IgE-Fc,RII interaction (Chrdtien et al., 1988). These two regions may lie close to one another in three-dimensional space (Fig. 3). An attractive hypothesis, incorporating the findings with recombinant peptides and monoclonal antibodies alike, is that the Fc,RII site may lie close to residues 367-370 in the C,3 domain (Vercelli et al., 1989). The anti-C,2 antibodies might then exert their effect by steric hindrance. C. Fc, RECEPTORS The presence of receptors for IgA has been reported on human monocytes, macrophages, and neutrophils (Fanger et al., 1980; Maliszewski et al., 1985; Chevailler et al., 1989), T cells (Briere et al., 1988; Millet et al., 1988), B cells (Gupta et al., 1979; Millet et al., 1989),


49

eosinophils (Abu-Ghazaleh et al., 1989), and NK cells (Kimata and Saxon, 1988). Purification of the receptor from neutrophils and monocytes has revealed a heavily glycosylated protein of about 60 kDa (Albrechtsen et ul., 1988; Monteiro et al., 1990). Fc, receptors on monocytes and neutrophils are capable of mediating phagocytosis of IgA-coated target cells (Fanger et al., 1983; Gorter et al., 1987; Yeaman and Kerr, 1987). They may also serve to promote ADCC by synergism with Fc, receptors (Shen and Fanger, 1981). Interaction of IgA, aggregated either artificially or at a cell surface, with monocyte and polymorphonuclear (PMN) Fc, receptors can trigger both the release of inflammatory mediators such as leukotrienes and prostaglandins and the generation of superoxide (Ferreri et al., 1986; Gorter et aZ., 1987; Stewart and Kerr, 1990; Padeh et al., 1991). Cross-linking of Fc, receptors in this way may also result in neutrophil degranulation (Albrechtsen et al., 1988). Recently, a cDNA clone coding for a human Fc, receptor has been isolated from a monocyte-like cell line cDNA library (Maliszewski et al., 1990). COS cells transfected with the cDNA clone readily bind IgA-coated erythrocytes. The deduced amino acid sequence indicates an integral membrane protein with a peptide core of about 30 kDa. The remainder of the mass is contributed by carbohydrate moieties attached at up to six potential extracellular N-glycosylation sites and perhaps further O-glycosylation sites. The N-terminal206 amino acids, lying outside the cel1, comprise two immunoglobulin-like domains that display homology to the extracellular regions of Fc,RI, Fc,RII, Fc,RIII, and the a-subunit of Fc,RI. A single transmembrane segment of 19 hydrophobic amino acids is then followed by a C-terminal cytoplasmic domain of 41 amino acids. Northern blot analysis revealed that mRNA coding for the Fc, receptor was present in peripheral blood monocytes and neutrophils. No message was detected in tonsillar B or T cells, suggesting either that the IgA receptor reported in these cell types is structurally distinct, or that receptor expression was not induced under the particular conditions used. Fc,R-lgA Interaction at the Molecular Level The human monocyte/PMN Fc, receptor appears to bind human serum IgAl and IgAz with similar affinity (Chevailler et al., 1989; Stewart and Kerr, 1990). Estimates using solubilized receptor indicate that both subclasses give half-maximal inhibition of the receptor-IgA M , suggesting that the afinteraction at concentrations of 4.8 x finity constant lies around 5 x lo7 M-' (Mazengera and Kerr, 1990). Human secretory IgA of both subclasses also exhibits very similar

50


inhibitory abilities. Thus, the receptor interacts equally well with monomeric serum IgA and the dimeric IgA, in complex with J chain and secretory component, of secretory IgA. The site of interaction lies entirely in the Fc region of IgA, because Fc, can inhibit the binding of polymeric IgA to the receptor, as shown in an indirect immunofluorescence assay (Monteiro et al., 1990). Further, the rebinding of purified receptor to IgA-Sepharose beads is inhibited by Fc, (Mazengera and Kerr, 1990). The above results with secretory IgA therefore suggest that the presence of either J chain or secretory component does not impede interaction of Fc,R with the Fc region of secretory IgA. A pepsin digestion product of IgA, lacking the C,3 domains, retains a somewhat reduced ability to interact with the neutrophil Fc, receptor (Mazengera and Kerr, 1990). This may suggest that the presence of both the C,2 and C,3 domains is necessary for full reactivity with the receptor. It is possible, however, that the C,3 domain serves the indirect role of maintaining the conformation of C,2 as we have seen with homologous Ig domains in the Fc,RI-IgG and Fc,RI-IgE interactions earlier. Site-directed mutagenesis experiments on human IgA should, in the future, help to elucidate the precise molecular requirements for binding to Fc, receptors (Woof et al., 1992).

D. Fc, RECEPTORS Subpopulations of human and mice B and T lymphocytes are reported to express functional Fc, receptors (Moretta et al., 1975; Ferranini et al., 1977; Mathur et al., 1988a,b). Biochemical analysis has revealed an IgM-binding protein of about 60 kDa on activated human B cells (Sanders et al., 1987). This molecule was not, however, detected on T cells, monocytes, or granulocytes. More recently, binding inhibition experiments have confirmed that this protein binds to the Fc portion of IgM and hence is a true Fc, receptor (Ohno et al., 1990). It appears to be expressed throughout the various stages of B cell differentiation and, in this respect, differs from the Fc, receptor on murine B cells (Mathur et al., 1988b). The human receptor is anchored to the membrane via a phosphatidylinositol-glycan linkage and possesses O-linked but not N-linked oligosaccharides. Human B cell Fc, receptor interacts with human and mouse IgM and, as mentioned previously, Fc5, fragments generated by hot trypsin digestion (Ohno et al., 1990). As the major trypsin cleavage site lies in the C,2 domain, the Fc5, fragments consist primarily of paired C,3 and C,4 domains. The same report details experiments using mouse IgM domain deletion mutants, which help to further localize the Fc,R binding site on IgM. Loss of the CJ domain did not impair binding to


51

the receptor. Deletion of both domains C,1 and C,2 resulted in reduced but still significant binding. However, a mutant lacking domains C,1, C,2, and C,3 and another lacking C,4 no longer bound Fc,R. Hence, Fc,R seems to require the presence of both C,3 and C,4 domains for recognition. Further study will be necessary to determine whether each domain contributes directly or indirectly to binding. It is perhaps of interest to note that the C,3 domain of mouse IgM appears to play the major role in binding to Fc,R on murine T and B cells (Mathur et al., 1988a,b). E. Fc8 RECEPTORS Receptors for human IgD have been detected on human B and T cells (Sjoberg, 1980; Rudders and Andersen, 1982; Tamma and Coico, 1991). The receptors on B cells, at least, interact with the Fc portion of human IgD. No significant degree of information on the molecular basis of the interaction, in the human system, is currently available.

F. EFFECTORCELL-TARGET CELLINTERACTION MEDIATED BY LEUKOCYTE Fc RECEPTORS We will now attempt to consider the interaction of antibodies and Fc receptors in a more physiological situation. First, we will discuss how antibody array formation at cell surfaces may facilitate interaction with Fc receptors. Second, we will consider the multiple factors that influence the “linkage” of effector cell and target cell by antibody molecules. The possibility of antibody array formation on target cell surfaces would appear an attractive hypothesis. Earlier, we mentioned the potential interaction of arrays of antibody molecules, possibly stabilized by Fc-Fc interactions, with C l q . Arrays of dislocated IgG molecules, for example, might also be expected to interact advantageously with Fc receptors for two main reasons. First, multiple FC receptors could bind simultaneously to an array of antibody molecules. The receptors would thus be effectively cross-linked, constituting a trigger for subsequent effector function. Second, dislocation of IgG molecules would serve to maximize access to the Fc,R interaction sites(s) lying in the lower hinge region by rotation of Fc regions perpendicular to Fab arms (Burton, 1986). Similar arguments may also apply to IgE. Indeed, experimental evidence suggests that IgE bound to Fc,RI has a bent conformation in which the C,2/C,3 interface, interacting with the receptor, lies about 45 A away from the cell surface. The remainder of the Fc lies about 55 away from the membrane whereas the tips of the Fab arms extend some 100 away from the cell

A

52


surface (Holowka and Baird, 1983; Holowka et al., 1985; Zheng et al., 1991). Further experimental evidence for antibody arrays was discussed earlier. We will now turn to the factors influencing recognition of antibodycoated target cells b y FcR-bearing effector cells. Because many FcR+ cells express more than one class of receptor, and an opsonized target cell may be coated with multiple antibody isotypes, the in v i m situation is presumably rather complex. It seems likely, for example, that different types of Fc receptor may act synergistically to trigger effector mechanisms (Crockett-Torabi and Fantone, 1990; Kimberly et al., 1990; Koolwijk et al., 1991). In order to simplify matters here we will discuss a model system involving the interaction of erythrocyte targets, coated with a single antibody isotype, with Fc, receptor-bearing effector cells. This simplification allows assessment of the cell-cell interaction by rosette formation, the microscopically visible binding of several erythrocytes to an effector cell. The interaction of the opsonized target with the effector cell in the rosette, in energetic terms, can be seen as the result of two opposing contributions, described by the following equation (Walker et al., 1989b): AG(rosette formation) = AG(Ab-FcR)interaction AG(nonspecific cell-cell interaction)

+

The first term [AG(Ab-FcR)interaction], promoting rosette formation, is associated with the free energy of occupation of Fc receptors by antibody molecules. The second opposing term [AG(nonspecific cellcell interaction)], militating against rosette formation, is associated with the repulsive forces of bringing together two cell surfaces close enough to allow bridging by antibody molecules. Hence, a sufficient input of free energy from antibody-receptor interactions to overcome cell-cell repulsion is necessary to allow rosette formation (see experimental examples in Fig. 18).These two terms appear to depend on a number of factors, each of which may influence rosette formation. Thus, the rosette-promoting term is governed by both the antibody isotype and the Fc receptor involved and by the number of antibodyreceptor interactions. The opposing term depends on the net surface charge of the two cells and the geometries of the antigen epitope and the Fc receptor. Both antigen and receptor must be “accessible” and suitably orientated to give optimal interaction with the antibody molecule. Finally, the relative structure of the antibody is important. For example, human IgG3 molecules, with their extended hinge regions,

HUMAN ANTIBODY EFFECTOR FUNCTION RED

BLOOD

53

CELL

BBOMELIW TREATED EFFECTOR EFFECTOR

OELL

@ELL

FIG. 18. Schematic representation of effector cell-target cell interactions. Human IgC, antirhesus D oranti-NIP antibodies (open Y shapes) and mouse IgC, antiglycophorin A antibodies (striped Y shapes) interact with their respective antigens on the surface of a human rhesus Df erythrocyte. Under favorable conditions, these antibodies may also interact simultaneously with the IgC-binding sites (dark areas) on Fc, receptors on a neighboring effector cell. The antibodies then serve to “bridge” between the two cell types, a situation that, under the microscope, would appear as a rosette, with several erythrocytes bound per effector cell. Five different combinations of antigen, antibody, and Fc, receptor are shown here to illustrate the theoretical considerations discussed in the text. In the first case, on the left, a human IgC3 anti-D antibody bridges between Fc,RI and the somewhat buried D antigen in the erythrocyte membrane. Thus, this illustrates the fact that an effector cell bearing Fc,RI (e.g., U937) can form rosettes with IgCG anti-D-sensitized erythrocytes. In the second case, however, human IgC, is unable to bridge between the relatively inaccessible D antigen and Fc,RII. This represents the finding that an effector cell bearing Fc,HII alone (e.g., Daudi) cannot form rosettes with IgC3 anti-D-sensitized erythrocytes. In the third combination, the antigen, NIP, is at a more accessible location on the red blood cell surface. This allows the IgG3 anti-NIP antibody to bind both to NIP and to Fc,RII. Hence, Daudi cells are able to form rosettes with NIP-derivatized erythrocytes coated with IgG3 anti-NIP antibodies. In the fourth example, as in the second, the antibody, in this case a mouse IgC, molecule, is unable to bridge between the antigenic epitope on the red cell and Fc,RII on the effector cell. Hence, no rosettes are seen with this combination. However, when the Fc,RII+ effector cell is pretreated with the enzyme bromelin, as in the fifth example, this same combination of antigen, antibody, and Fc receptor may simultaneously interact. Bromelin treatment results in cleavage of surface glycoproteins, thereby reducing the net surface charge of the cell. The two cell surfaces may therefore approach each other more closely as mutual repulsion is reduced. Thus, Daudi cells, after bromelin treatment, become able to form rosettes with erythrocytes coated with the mouse IgC, antiglycophorin A antibody.

appear to mediate rosette formation more readily than IgGl molecules do, when all other parameters are the same (Walker et al., 1988,1989b). However, this interaction process is merely the first stage of target cell destruction and some evidence suggests that optimal interaction may not always lead to optimal destruction. Hence, although human IgG,

54


molecules more readily promote target cell-effector cell interaction than do human IgG1 molecules, it is the latter antibody that mediates most effective ADCC against target cells once interaction has occurred (Bruggemann et al., 1987). In terms of the “design” of antibodies for therapy, a consideration of the multiple parameters influencing interaction of target cell and effector cell would seem pertinent. Thus, in order to generate a monoclonal antibody effective at Fc receptor-mediated target-cell killing, we must take into account not only the more obvious factors of antibody isotype, class of Fc receptor, and type of effector cell(s) involved, but also antigen density, epitope accessibility, and surface charge on the target cell. We must ensure that we have an antibody capable of both specifically recognizing a target cell, and of “bridging” efficiently between that cell and an effector cell, prior to triggering target cell destruction.

V. Catabolism of Antibodies

Although not an effector function, the survival of antibodies in blood has important consequences in understanding the biochemistry of these molecules and for their potential therapeutic and diagnostic uses. The subject has recently been reviewed by Zuckier et al. (1989) and we shall concentrate on points relating to molecular aspects. In humans, antibodies of the IgG class have the longest half-lives of any of the serum proteins (average tllz = 21 days), with the other classes having shorter half-lives: IgA, 6 days; IgM, 5 days; IgD and IgE, 3 days. IgG3 demonstrates a significantly shorter t l / z (“days) than the other IgG subclasses whereas the two IgA subclasses have similar t l l z . The unique feature of IgG catabolism in humans and in other species studied is that the catabolic rate is proportional to the serum concentration. Thus serum IgG is degraded much more rapidly in hypergammaglobulinemic individuals and much more slowly in hypogammaglobulinemics. This has led Brambell et a2. (1964) to propose that there are a limited number of receptors that complex IgG and protect it from degradation in blood. When levels of IgG are high, the receptors will be saturated, making more IgG available for degradation. Conversely, when levels are low, most of the IgG will be protected, prolonging serum survival. There is no direct evidence for this mechanism but it has been considered the best available (Waldmann and Strober, 1969). An alternative occurring to us is one based on an equilibrium between monomeric and associated IgG species. Higher serum concen-


55

trations of IgG would favor aggregate formation that might be catabolized more rapidly than monomer. Such an aggregated population might form only a small proportion of the total IgG, and therefore be difficult to observe, but be catabolized very rapidly. The site(s) regulating the catabolism of IgG (whatever the molecular mechanism involved) is believed to be in the Fc part of the molecule because Fc fragments are catabolized at the slow rates characteristic of IgG, in contrast to Fab fragments, which are rapidly cleared (Spiegelberg and Weigle, 1965a,b; Wochner et al., 1967; Zuckier et al., 1989).Furthermore, infusion of Fc fragments can mimic IgG in accelerating the catabolism of circulating IgG in mice (Fahey and Robinson, 1963). Interestingly, the Fc fragments of all four human IgG subclasses appear to have identical fractional catabolic rates, implying that structures outside the Fc region are responsible for the accelerated catabolism of IgG3 (Spiegelberg and Fishkin, 1972). A number of studies have sought to localize the site more precisely. Yasmeen et al. (1976)reported that a C,2 fragment from a human IgGl protein was cleared from rabbit circulation with a tll2 similar to intact IgG and Fc and much longer than that for Fab, pFc', or C,3 fragments. Arend and Webster (1977) reported that rat pFc' was rapidly catabolized in rats compared to Fc. These studies imply a crucial role for the C,2 domain. In contrast, Pollock et al. (1990), using mutant mouse IgG molecules, have obtained evidence that deletion of any of the constant domains has an effect on clearance in the mouse. Further, using IgG2b/za hybrid molecules, they suggest that sequences at the C terminal end of C,2 or within the C,3 domain, or conformations controlled by these sequences, are important in catabolism. Recently Wawrzyncak et al. (1992a) looked at the rates of clearance, from the circulation of mice, of mutant mouse IgGzbs used in the Clq-binding and Fc receptor studies described earlier. They found no significant differences between t1/2 values for mutant and wild-type IgGs, implying that clearance is independent of the ability to bind C l q or mouse FcRI (mutation of Glu to Leu at position 235 was shown to generate IgGzb binding to mouse as well as human FcRI). Though there has been debate about the role of carbohydrate, the present consensus seems to imply that the carbohydrate moieties have only a limited effect on serum half-life (Waldmann and Strober, 1969; Tao and Morrison, 1989; Zuckier et al. 1989; Wawrzynak et al. 1992b). However, the terminal galactose residues on IgA molecules, and in some cases on other Ig classes, do target them to the hepatic galactose receptor and thus can play a decisive role in their catabolism (Zuckier et al., 1989).

56


VI. Bacterial Fc Receptors

Certain bacteria express on their surface proteins capable of binding specifically to the Fc region of mammalian immunoglobulins. These so-called bacterial Fc receptors have been demonstrated on many staphylococcal and streptococcal strains (Boyle and Reis, 1987). Those receptors specific for the Fc of IgG have currently been classified, according to their functional reactivity with different species and subclasses of IgG, into six groups or types. The nature of the interaction of each oftype with human IgG will be dealt with in turn. Relatively little is currently understood, in precise molecular terms, about the interaction of bacterial Fc receptors with other classes of immunoglobulin. These types of receptors will not, therefore, be discussed here. A. TYPEI Fc RECEPTOR

Type I Fc receptor, frequently termed protein A, is found on the surface of the majority of strains of StaphyZococcus aureus (Forsgren and Sjoquist, 1966).This very extended protein has been cloned and found to consist of five homology units, each capable of binding the Fc of IgG and a sixth region that does not bind Fc but binds to cell walls (Moks et al., 1986). Active 7-kDa fragments, each corresponding to a homology domain, can be generated by trypsin digestion. It is possible for two such fragments to bind simultaneously to one Fc molecule. A typical binding affinity ( K , - 3 x lo6 M - l ) is demonstrated by fragment B binding to rabbit Fc (Lancet et al., 1978). Earlier crystallographic data (Deisenhofer et al., 1976, 1978) have been refined to produce a model of the complex of human Fc and fragment B at 2.8 A resolution (Deisenhofer, 1981).Fragment B forms two contacts with Fc molecules in the crystals but one of these is argued to be merely a crystal contact. In the other, thought to exist in solution, fragment B binds at the interface between the C,2 and Cy3 domains of IgG (Fig. 19).The residues involved comprise parts of two hydrophobic patches on Fc, one on the Cy2 domain (Met 252, Ile 253, Ser 254, Leu 309, His 310, and Glu 311) and the other on the C,3 domain (His 433, His 435, Tyr 436, and Asn 434).. Protein A binds the human subclasses IgG,, IgG2, and IgG4, and also IgG3 molecules of the allotype IgGsm(l5, 16), characteristic of mongoloid populations (Recht et al., 1981). Each of these proteins has a histidine residue at position 435 involved in the protein A interaction. By contrast, in IgG3 molecules of the Caucasian allotypes IgGam(5) and IgG,m(21), this histidine is replaced by arginine. Model building (Deisenhofer, 1981)reveals that the inability of such proteins


57

FIG.19. Structure ofthe complex of fragment B of protein A with human Fc. ApproxiThe a-carbon trace of only one mate centers of carbohydrate hexose units are shown (0). Fc heavy chain is shown on the right. The interaction involves the contact of two a-helices of fragment B (left) with a hydrophobic patch in the C,2/C,3 interface (after Deisenhofer, 1981).

to bind protein A probably results from the prevention of favorable IgG-protein A contact formation by the lengthy side chain of arginine. Loss of the N-linked carbohydrate moieties from the C,2 domains of IgG appears not to affect the interaction, as aglycosylated forms of both mouse IgGz, and IgGzb still bind to protein A (Leatherbarrow and

58

DENNIS R. BURTON A N D J E N N Y M. WOOF

Dwek, 1983; Nose and Wigzell, 1983). It has been demonstrated recently, by use of 'H NMR, that aglycosylation results in only a small and localized structural change in the vicinity of the His 268 reporter group at the N-terminal end of the Cy2 domain (Lund et al., 1990). Hence it is hardly surprising that loss of the carbohydrate residues from Fc does not affect the interaction with protein A. Similarly, neither reduction and alkylation nor hinge deletion of human IgG1 perturbs the Cy2-Cy3 interface region, as demonstrated by the continued reactivity of these modified IgGs with protein A.

B. TYPE11 Fc RECEPTOR This second class of receptor is associated with certain strains of group A streptococci. There is, however, a considerable degree of heterogeneity in the IgG subclass binding profiles among different group A isolates. Further, the binding profile of a particular isolate may vary with passage and even individual colonies within that isolate may express a variety of IgG-binding capacities (Yarnall et al., 1984). Thus binding of IgG to these bacteria is complex and appears to be a function of several different type I1 Fc receptors. A recent study has defined five subtypes of the receptor based on their IgG species and subclass specificities (Raeder et al., 1991a). Type IIa receptor binds human IgG1, IgG2, and IgG4, but not IgG3. It also binds rabbit, pig, and horse IgGs. Immunoblotting techniques with the strain 64/14 have demonstrated a molecule of -50 kDa with these binding properties (Yarnall and Boyle, 1986a). More recently, an IgG-binding protein, cloned from the group A strain CSllO (Heath and Cleary, 1987, 1989), displayed the IgG-binding profile of type IIa receptor (Cleary and Heath, 1990). A study assessing the ability of various dipeptides to inhibit the binding of radiolabeled IgG to strain 64/14 provided information on the localization of the type IIa receptor site on IgG (Yarnall and Boyle, 1986b). The dipeptides glycyltyrosine and glycylhistidine the binding of human IgG1, IgG2, and IgG4 and rabbit and pig IgG to 64/14. This result, taken with the human IgG subclass specificity profile of this receptor, suggests that the IgGbinding site for the type IIa receptor may be very similar to that of the type I receptor. Histidine and tyrosine residues would appear to be important in the interaction. Interestingly, an IgGS molecule capable of binding protein A, and therefore presumably of an allotype with histidine at position 435, was found to bind the type IIa receptor on 64/14 (Yarnall and Boyle, 1986a). Reactivity of only certain IgG3 allotypes with the cloned receptor was also noted (Cleary and Heath, 1990).


59

The type IIb Fc receptor has been demonstrated on the group A strains 64/14, A992S, and 11434 and substrain A928 A2 (Raeder et al., 1991a) with a molecular weight of -35,000 in the two former strains and -47,000 in the two latter. It binds solely to human IgG3 (Raeder et al., 1991a; Yarnall and Boyle, 1986a). The largest differences between IgG3 and the other human subclasses reside in its very extended hinge region, suggesting that a potential interaction site for type IIb receptor may lie in this part of the Fc. Attempts to inhibit the interaction with monoclonal antibodies specific for the hinge of human IgG3 (Lowe et al., 1982) might prove informative. Type IIc receptor, a 116-kDa protein demonstrated on the strain A992S, has been defined as displaying specificity for human IgGl and rabbit, pig, and horse IgGs, and weak reactivity with human IgG,. Sequence comparison of the C,2 and C,3 domains between the human IgG subclasses reveals that all residues conserved in both human IgG, and IgG, are also conserved in IgGz and IgG,. Thus the specificity of type IIc receptor is difficult to explain by sequence differences in the ligands. Rather, the idea that distinct IgG-binding reactivities may be the sum of a number of independent binding units coming together in different combinations (Raeder et al., 1991a) may be more appropriate. Type 110 receptor, present on the substrain A928 Al, binds all human IgG subclasses as well as rabbit, pig, and horse IgGs (Raeder et al., 1991a). This specificity is associated with a 47-kDa protein. The fifth receptor subgroup, type 11’0,is a variant of the above and binds only human IgG (all four subclasses) and rabbit IgG. A protein displaying this specificity has been cloned from strain AP1 and given the alternative name of protein H (Akesson et al., 1990; Gomi et al., 1990). It has a molecular weight of -40,000 and shows homology to protein Arp, an IgA-binding streptococcal protein, but not to protein A, protein G (see later), or type IIa Fc receptor. Protein H is reported to block the binding of both protein A and protein G to IgG, suggesting it may also interact with the Cy2/Cy3interface (Akesson et al., 1990). C. TYPE111 Fc RECEPTOR Type 111 Fc receptor is expressed on the surface of most human group C and group G streptococci. This receptor, more frequently termed protein G, exhibits a remarkably wide reactivity, interacting with all human IgG subclasses and IgG from rabbit, goat, cow, sheep, rat, mouse, guinea pig, horse, and pig. It displays a considerably higher affinity ( K , 1 x 10’ M-’ for human IgG subclasses) than does protein A (Reis et al., 1984; Akerstrom and Bjorck, 1986). Type I11 receptor is reported to have some affinity for Fab fragments of IgG mediated at a

-

60


site distinct from that binding to Fc (Erntell et al., 1988). In addition, the receptor has affinity for human serum albumin. The type I11 receptor has been successfully cloned from two different group G sources (Fahnestock et al., 1986; CUSS et al., 1986; Olsson et al., 1987). The deduced amino acid sequence shares features in common with that of protein A. Starting at the C-terminus, there is a region responsible for cell wall anchorage, followed by six repeated elements predicted to adopt a conformation of linearly arranged domains. The capacity to bind serum albumin is mediated by the Nterminal half of the domain structure (Guss et al., 1986), whereas the site for IgG lies in the C-terminal half ofthe molecule (Wkerstromet al., 1987; Sjobring et al., 1988). Despite the similarity in overall organization of type 111 receptor and protein A, there is no homology between their IgG-binding regions, suggesting that they arose by convergent evolution. In order to localize the type I11 receptor site on Fc, two different approaches have been used. First, the ability of enzymatically derived fragments of IgG to interact with the receptor was assessed in direct binding and inhibition studies (Schroder et al., 1986; Stone et al., 1989). The pFc’ fragment (Cy3dimer) of both human and rabbit IgGs did not bind to the receptor. Both human and rabbit F (ab’)zfragments failed to inhibit binding of radiolabeled Fc to group C and G streptococcal strains. Further, the C,3-lacking rabbit Facb fragment showed little inhibitory ability. In the second approach, chemical modification of IgG implicated the involvement of IgG tyrosines in the interaction (Stone et al., 1989).These results suggest that type 111receptor binds to an interaction site very similar to that of protein A, especiaIly since the monovalent fragment D of protein A inhibits the binding oftype I11 protein to IgG. Indeed, type I11 receptor was also shown to be capable of inhibiting the binding of IgM rheumatoid factors, which generally bind to the Cy2-C,3 domain interface on IgG. However, a marked difference between protein A and type 111receptor is that the former is sensitive to the His 435 --* Arg interchange in the IgG3 molecules of most Caucasians, but the latter is not. This suggests that this residue, clearly central in protein A binding, is on the periphery of the type I11 receptor site. Subtle differences in the recognition processes of the two receptors may also explain the differences in pH optimum for IgG binding to protein A (-pH 8) and type I11 receptor (pH 4-5) (Akerstrom and Bjorck, 1986). X-Ray crystallographic analysis of the complex of Fc and type I11 receptor (or a fragment of it) would perhaps now allow definite identification of the interaction on both ligand and receptor.


61

D. TYPEIV Fc RECEPTOR Type IV Fc receptors are associated with some P-hemolytic bovine group G streptococci. They are capable of binding human IgG1, IgG,, and IgG4 and IgG from several other species, particularly rabbit (Reis et al., 1990; Raeder et al., 1991b). Sequence analysis ofthe Cy2 and C y 3 domains of rabbit IgG and the human IgG isotypes reveals only three residues, which are conserved in rabbit IgG, human IgGl and IgG4, but not in human IgG2. These are residues Leu 235, Gly 236, and Ala 339, all in the Cy2 domain, the latter residue being located on the interior surface of the domain. Hence, one might speculate that a potential interaction site may encompass residues 235 and 236, which, as mentioned earlier, form part of the lower hinge site for the highaffinity human FcyRI. Attempts to inhibit the binding to one receptor with soluble fragments of the other might be an illuminating future experiment.

E. TYPEV Fc RECEPTOR This fifth bacterial receptor for the Fc of IgG is found on certain strains of Streptococcus zooepidemicus. It has a binding specificity very similar to that of protein A, displaying affinity for human IgG,, IgG2, and IgG4 and pig, guinea pig, and rabbit IgGs. Only weak reactivity is reported with cow, sheep, goat, horse, rat, dog, and cat IgGs (Myhre and Kronvall, 1980).More recently, a subtype of this receptor has been described (Yarnall and Widders, 1990). The subtype protein has a molecular weight of -45,000 and similar specificity to the original type V receptor except that it has strong reactivity with cat and horse IgGs. This subtype V receptor may recognize a site on Fc distinct from the protein A site because protein A apparently still bind to IgG after binding of subtype V receptor in blotting experiments. This is somewhat surprising considering the similar binding profiles of the two receptors, especially with the human IgG subclasses. Further binding inhibition studies with native receptors might aid clarification of the situation.

F. TYPEVI RECEPTOR An IgG-binding protein of -46 kDa, present on the surface of S. zooepidemicus strain S212, has been designated type VI Fc receptor (Reis et al., 1988).It has specificity for rabbit, pig, sheep, goat, and COW IgGs and weaker reactivity with human, mouse, and rat IgGs. This affinity for rat IgG is, however, the greatest displayed by any bacterial Fc receptor thus far described. An explanation of this binding profile in terms of IgG site localization is, with present knowledge, difficult.

62


VII. Conclusions

Considerable advances have been made in the past few years in delineating the sites on antibodies interacting with effector molecules. Although far from complete, such information allows much to be done in terms of rational design of effector function. For instance, single point mutations can be used to eliminate C l q and Fc receptor (FcRI and FcRII) binding from an IgG molecule. More subtly, point mutations can be used to eliminate complement lysis while leaving C l q binding unaffected or to eliminate FcRI binding with small effects on FcRII. In the opposite experiment, a single point mutatian can be used to convert an IgG displaying no measurable affinity for FcRI into one with fully functional affinity. Generally, the introduction of function to an inactive antibody is likely to be more demanding of the level of our understanding, and here we have further to go. A recurring theme in this review is that binding of the appropriate antibody isotype to antigen is a necessary but not sufficient criterion for effector triggering. Examples abound of antibodies, for instance, that bind C l q but do not sustain a later step in the complement cascade. Similarly, there are antibodies that link effector and target cells effectively but do not lead to target damage. There would appear to be extra requirements associated with the antigen. Further, this does not appear to be simply a question of antigen density. One explanation might be that certain antigens are able to trigger allosteric changes whereas others are not. We have discussed why we think this unlikely. The explanation we favor is that antibodies linking arrays of antigen and effector molecules have preferred arrangements for optimal effector function. Some antigens, or their local environment, would preclude the formation of such arrangements and so they would bind antibody but be unable to activate the effector function. This hypothesis could accommodate the independent sensitivities of many of the steps of the complement cascade. For instance, a particular antibody arrangement could be sufficient for Clq binding and C l activation but inappropriate for C4 activation. Accepting for a moment the notion of preferred arrangements, how might they look? We have described some circumstantial reasons for favoring the formation of IgG hexamers involving dislocated IgG molecules with Fc-Fc interactions, but have no hard data to support this as yet. Certainly IgM appears to function best with respect to complement activation in a “preferred arrangement” (hexamer) with Fc-Fc interactions. Dislocation of antibody molecules is a feature that is difficult not to embrace given the localization of Fc receptor binding


63

regions to the middle part of the antibody IgG and IgE molecules. In simple terms it is difficult to envisage the topology of the bridging of two cells by antibody without invoking movement of Fc out of the plane of the two Fab arms. The available physical data support the notion that such movements should be allowed. Indeed, flexibility of the antibody molecule could be useful in a number of ways when linking antigen and effector. For instance, dislocation and Fab arm opening will both serve to bring effector molecules closer to the antigenic surface. Thus, for example, the surface of an effector cell will be brought closer to that of the target cell (Fig. 16),facilitating target cell destruction, and the generation of activated C4b and C3b will occur closer to the target cell membrane, facilitating complement lysis. Again, flexibility is expected to be advantageous given the diversity of antigens that must be linked to the common effector systems. Clearly, much of this discussion is conjecture. One would like to have experimental measurements on the interacting triumvirate of antigen, antibody, and effector molecule. As discussed earlier this is a tall order, but it is to be hoped that appropriate methodologies will b e developed. An interesting feature of antibody effector function is the occurrence of the human IgG subclasses. Typically one wonders about the role of the IgGz and IgGl subclasses given their very poor reputations in this area. We would suggest the following propositions: IgGl and IgG3 both appear to mediate effector functions, although the relative efficacy of the two may vary according to the conditions of effector triggering. IgG3, with its long hinge, would seem to have an advantage in bringing antigen and effector closer together in that repulsive forces, e.g., between cell surfaces, would be minimized. IgGl, on the other hand, by bringing antigen and effector closer together, may facilitate target damage. It could be that, in vivo, the two function cooperatively. IgGz does activate complement under conditions of high epitope density and this provides a “rationale” for the preponderance of IgGz anticarbohydrate antibodies (Michaelsen et al., 1991) because carbohydrates are often presented at high density on microbial surfaces and at lower densities more ubiquitously. IgG2 does not generally interact with Fc receptors except for one form of FcRII as discussed. IgG, does not appear to activate complement under any circumstances. It does interact with FcRI, albeit more weakly than IgG1 and IgG3, but not with FcRII or FcRIII. It may be that there are situations, e.g., blocking of the function of certain viruses, wherein it is desirable to have antibody binding without, e.g., cellular uptake. Many effector studies are now carried out with monoclonal antibod-

64


ies, although in uiuo, of course, the response is polyclonal. Because the effector systems tend to recognize arrays of antibodies that could differ depending on the composition of the antibodies involved, the distinction may be important. An example is provided by the synergistic effect of two monoclonal antibodies on complement activation described previously. As more human monoclonal antibodies become available (Burton, 1991; Burton et al., 1991; Persson et al., 1991) this is an area which should be explored. It is our guess that we shall find that antibodies work best as mixtures, both in terms of epitope specificity and subclass.

ACKNOWLEDGMENTS We are most grateful to Tim Clackson, Geoff Hale, T e j e Michaelsen, Craig Morton, Inger Sandlie, Verne Schumaker, Bob Sim, and Jan van d e Winkel for helpful comments on the manuscript. We thank Peter Artymiuk, Geoff Ford, Neil Cooper, Ken Davis, Birgit Helm, Nevin Hughes-Jones, and Greg Winter for use of diagrams. We acknowledge the financial support of the Medical Research Council, the Lister Institute of Preventive Medicine, and Johnson & Johnson.

REFERENCES Abu-Ghazaleh, R. I., Fujisawa, T., Mestecky, J., Kyle, R. A., and Gleich, G. J. (1989). IgA-induced eosinophil degranulation. J . lmmunol. 142,2393-2400. Akerstrom, B., and Bjorck, L. (1986). A physiochemical study of protein G, a molecule with unique immunoglobulin G-binding properties.]. Biol. Chem. 261,10240-10247. Akerstrom, B., Nielsen, E., and Bjorck, L. (1987). Definition of IgG- and albuminbinding regions of streptococcal protein G. J. B i d . Chem. 262,13388-13391. Akesson, P., Cooney, J., Kishimoto, F., and Bjorck, L. (1990). Protein H-A novel IgG binding bacterial protein. Mol. lmmunol. 27,523-531. Albrechtsen, M., Yeaman, G. R., and Kerr, M. A. (1988). Characterization of the IgA receptor from human polymorphonuclear leucocytes. Immunology 64,201-205. Alcarez, G . , Kinet, J.-P., Liu, T. Y., and Metzger, H. (1987). Further characterization of the subunits ofthe receptor with high affinity for immunoglobulin E. Biochemistry 26, 2569-2575. Alcolea, J. M., Anton, L. C., Marques, G., Sanchez-Corral, P., and Vivanco, F. (1987). Formation of covaIent complexes between the fourth component of human complement and IgG immune aggregates. Complement 4,21-32. Alexander, R. J., and Steiner, L. A. (1980). The first component of human complement from the bullfrog Rana catesbeinana: Functional properties of C1 and isolation of subcomponent C1q.J. lmmunol. 124,1418-1426. Allen, J. M., and Seed, B. (1988). Nucleotide sequence of three cDNAs for the human high affinity Fc receptor (FcRI). Nucleic Acids Res. 16,11824. Allen, J. M., and Seed, B. (1989). Isolation and expression of functional high-affinity Fc receptor complementary DNAs. Science 243,378-380. Anasetti, C., Martin, P. J., Morishita, Y., Badger, C., Bernstein, I. D., and Hansen, J. A. (1987).Human large granular lymphocytes express high affinity receptors for murine monoclonal antibodies of the IgC3 subclass. J. lmmunol. 138,2979-2981. Anderson, C. L., and Spiegelberg, H. L. (1981). Macrophage receptors for IgE: Binding


65

of IgE to specific IgE Fc receptors on a human macrophage cell line, U937. J. Zmmunol. 126,2470-2473. Anderson, C. L., Guyre, P. M., Whitin, J. C., Ryan, D. H., Looney, R. J., and Fanger, M. W. (1986). Monoclonal antibodies to Fc receptors for IgG on human mononuclear phagocytes. Antibody characterization and induction of superoxide production in a nionocyte cell. J . Biol. Chem. 261, 12856-12864. Anderson, C. L., Ryan, D. H., Looney, R. J., and Leary, P. C. (1987). Structural polymorphism of the human monocyte 40 kilodalton Fc receptor for IgC. J. Immunol. 138, 2254-2256. Anderson, C. L., Eicher, D. M., Wewers, M. D., and Gill, J. K. (1990a). Phagocytosis mediated by three distinct Fcy receptor classes on human leukocytes. J . E x p . Med. 171,1333-1345. Anderson, C. L., Looney, R. J., Culp, D. J., Ryan, D. H., Fleit, H. B., Utell, M. J., Franipton, M. W., Manganiello, P. D., and Guyre, P. M . (1990b). Human alveolar and peritoneal macrophages bear three distinct classes of Fc receptors for 1gG.J. Immunol. 145, 196-201. Anton, L. C., Alcolea, J. M., Sanchez-Corral, P., Marques, G., Sanchez, A., and Vivanco, F. (1989). C 3 binds covalently to the C,3 domain of IgG immune aggregates during complement activation by the alternative pathway. Biochem. J. 257,831-838. Arend, W. P., and Webster, D. E. (1977). Catabolism and biologic properties of two species of rat IgG2a Fc fragments. J. Immunol. 118,395-400. Arlaud, G. J., Colomb, M. G., and Gagnon, J. (1987).A functional model ofthe human C1 complex. Zmmunol. Today 8,106-111. Baniyash, M., and Eshhar, Z. (1984). Inhibition of IgE binding to mast cells and basophils by monoclonal antibodies to niurine IgE. Eur.J. Immunol. 14,799-807. Baniyash, M., Kehry, M., and Eshhar, Z. (1988). Anti-IgE monoclonal antibodies directed at the Fc. receptor binding site. Mol. Immunol. 25,705-711. Barnett-Foster, D. E., Dorrington, K. J., and Painter, R. H. (1978). Structure and function of immunoglobulin domains. VII. Studies on the structural requirements of human immunoglobulin G for granulocyte binding. J. Immunol. 120, 1952-1956. Barnett-Foster, D. E., Sjoquist, J., and Painter, R. H. (1982). The effect of fragment B of staphylococcal protein A on the binding of rabbit IgG to human granulocytes and monocytes. Mol. Immunol. 19,407-412. Beale, D., and Feinstein, A. (1976). Structure and function of the constant regions of immunoglobulins. Q. Reu. Biophys. 9, 135-180. Bianchino, A. C., Poon, P. H., and Schumaker, V. N. (1988). A mechanism for the spontaneous activation of the first component of complement C1 and its regulation by C1-inhibitor. J . Immunol. 141,3930-3936. Bindon, C. I., Hale, G., Hughes-Jones, N., Gorick, B., and Waldmann, H. (1987). Synergistic complement lysis by monoclonal antibodies to the human leucocyte common antigen requires both the classical and the alternative pathways. Mol. Immunol. 24, 587-594. Bindon, C. I., Hale, G., Bruggemann, M., and Waldmann, H. (1988a). Human monoclonal IgG isotypes differ in complement activating function at the level of C4 as well as C1q.J. E x p . Med. 168,127-142. Bindon, C . I., Hale, G., and Waldmann, H. (1988b). Importance ofantigen specificity for complement-mediated lysis by monoclonal antibodies. Eur. J. Immunol. 18, 15071514. Bindon, C . I., Hale, G., and Waldmann, H. (1990).Complement activation by immunoglobulin does not depend solely on C l q binding. Eur. J. Immunol. 20,277-281.

66


Blank, U., Ra, C., Miller, L., White, K., Metzger, H., and Kinet, J.-P, (1989). Complete structure and expression in transfected cells of high affinity IgE receptor. Nature (London) 337,187-189. Boackle, R. J., Johnson, B. J., and Caughman, G. B. (1979). An IgG primary sequence exposure theory for complement activation using synthetic peptides. Nature (London) 282,742-743. Borsos, T., and Rapp, H. J. (1965a). Hemolysin titration based on fixation ofthe activated first component of complement. Evidence that one molecule of hemolysin suffices to sensitise an erythrocyte. J . Immunol. 95,559-566. Borsos, T., and Rapp, H. J. (196513).Complement fixation on cell surfaces by 19s and 7s antibodies. Science 150,505-506. Borsos, T., Chapuis, R. M., and Langone, J. L. (1981). Activation of complement by natural IgM anti-hapten antibody: Effect of cell surface hapten density. Mot. Immunol. 18,869-872. Boyle, M. D. P., and Reis, K. J. (1987). Bacterial Fc receptors. BiolTechnology 5,697703. Brambell, F. W. R., Hemmings, W. A., and Morris, I. G. (1964). A theoretical model of y-globulin metabolism. Nature (London)203,1352-1355. Briere, F., Paliard, X., and De Vries, J. E. (1988). Induction of the receptor for the Fc portion of IgA by secretory IgA on human T cell lines and T cell clones. Eur. J . Immunol. 18,445-451. Brooks, D. G., Qui, W. Q., Luster, A. D., and Ravetch, J. V. (1989). Structure and expression of a human IgG FcRII (CD32): Functional heterogeneity is encoded by the alternatively spliced products of multiple genes. J . E x p . Med. 170, 1369-1386. Brown, E. J., Berger, M., Joiner, K. A., and Frank, M. M. (1983). Classical complement pathway activation by antipneumococcal antibodies leads to covalent binding of C3b to antibody molecules. Infect. Immun. 42,594. Bruggemann, M., Williams, G. T., Bindon, C. I., Clark, M. R., Walker, M. R., Jefferis, R., Waldmann, H., and Neuberger, M. S. (1987). Comparison of the effector functions of human immunoglobulins using a matched set ofchimeric antibodies. J . E x p . Med. 161, 1351-1361. Brunhouse, R., and Cebra, J. J. (1979). Isotypes of IgG-comparison of the primary structures of three pairs of isotypes which differ in their ability to activate complement. Mol. Immunol. 16,907-917. Burritt, M. F., Calvanico, J. J., Mehta, S., and Tomasi, T. B. (1977). Activation of the classical complement pathway by Fc fragment of human IgA. J . Immunol. 118,723725. Burton, D. R. (1985). Immunoglobulin G: Functional sites. Mol. Immunol. 22,161-206. Burton, D. R. (1986). Is IgM-like dislocation a common feature of antibody function? Immunol. Today 7,165-167. Burton, D. R. (1987). Structure and function of antibodies. In “Molecular Genetics of Immunoglobulin” (F. Calabi and M. S. Neuberger, eds.), pp. 1-50. Elsevier, Amsterdam. Burton, D. R. (1990a).Antibody: The flexible adaptor molecule. Trends Biochem. Sci. 15, 65-69. Burton, D. R. (1990b). The conformation of antibodies. In “Fc Receptors and the Action of Antibodies” (H. Metzger, ed.), pp. 31-54. Am. Soc. Microbiol., Washington, D.C. Burton, D. R. (1991). Human and mouse monoclonal antibodies by repertoire cloning. Trends Biotechnol. 9,169-175. Burton, D. R., Boyd, J., Brampton, A., Easterbrook-Smith, S., Emmanuel, E. J., Novotny,


67

J., Rademacher, T. W., van Schravendijk, M.-R., Sternberg, M. J. E., and Dwek, R. A. (1980). The C l q receptor site on immunoglobulin G. Nature (London)288,338-344. Burton, D. R., Artymiuk, P. J., and Ford, G. C. (1989). Death by antibody. New Sci. 122, 42-45. Burton, D. R., Barbas, C. F., Persson, M. A. A., Koenig, S., Chanock, R. M., and Lerner, R. A. (1991). A large array of human monoclonal antibodies to HIV-1 from combinatorial libraries of asymptomatic seropositive individuals. Proc. Natl. Acad. Sci. U.S.A.88, 10134- 10138. Byrn,R. A., Mordenti, J., Lucas,C., Smith, D., Marsters, S.A., Johnson, J. S.,Cossum,P., Chamow, S. M., Wurm, F. M., Gregory, T., Groopnian, J. E., and Capon, D. J. (1990). Biological properties of a CD4 immunoadhesin. Nature (London)344,667-670. Campbell, R. D., Dodds, A. W., and Porter, R. R. (1980). The binding of human complement component C4 to immune aggregates. Biochem. J . 189,67-80. Canfield, S. M., and Morrison, S. L. (1991).The binding affinity ofhuman IgG for its high affinity Fc receptor is determined by multiple amino acids in the CH2 domain and is modulated by the hinge region. J. E x p . Med. 173, 1483-1491. Capon, D. J., Chamow, S. M., Mordenti, J., Marsters, S., Gregory, T., Mitsuya, H., Byrn, R. A., Lucas, C., Wurm, F. M., Groopnian, J. E., Broder, S., and Smith, D. H. (1989). Designing CD4 immunoadhesins for AIDS therapy. Nature (London) 337,525-531. Capron, A., and Dessaint, J. P. (1985). Effector and regulatory mechanisms in immunity to schistosomes: A heuristic view. Annu. Reu. Iinmunol. 3,455-476. Carter, S. D., Leslie, R. G. Q., and Reeves, W. G. (1982). Human monocyte binding of homologous monomer and complexed IgG. Zininuiiology 46,793-800. Cattaneo, A,, and Neuberger, M. S. (1987). Polynieric immunoglobulin M is secreted by transfectants of non-lymphoid cells in the absence of imniunoglobulin J chain. E M B O J. 6,2753-2758. Chevailler, A., Monteiro, R., Kubagawa, H., and Cooper, M. D. (1989). Immunofluorescence analysis of IgA binding by human mononuclear cells in blood and lymphoid tissue. J . Zmmunol. 142,2244-2249. Chretien, I., Helm, B. A., Marsh, P. J., Padlan, E. A., Wijdenes, J., and Banchereau, J. (1988). A monoclonal anti-IgE antibody against an epitope (amino acids 367-370) in the CH3 domain inhibits binding to the low affinity IgE receptor (CD23).J. Zmmunol. 141,3128-3134. Ciccimarra, F., Rosen, F. S., and Merler, E. (1975). Localization of the IgG effector site for monocyte receptors. Proc. Natl. Acad. Sci. U.S.A.72,2081-2083. Circolo, A., and Borsos, T. (1982). Lysis of hapten-labelled cells by anti-hapten IgG and complement: Effect of cell surface hapten density. J . Zmmunol. 128,1118-1024. Circolo, A., and Borsos, T. (1984). Lack of binding of C3 to IgG antibodies during the activation ofthe classical complement pathway on the red cell. Mol. Zmmunol. 21,191. Clackson, T., and Winter, G. (1989). “Sticky feet” directed mutagenesis and its application for mapping antibody domains. Nucleic Acids Res. 17,10163-10170. Clark, M., Bindon, C., Dyer, M., Friend, P., Hale, G., Cobbold, S., Caine, R., and Waldmann, H. (1989a).The improved lytic function and in vivo efficacy ofmonovalent monoclonal CD3 antibodies. Eur. J. Immunol. 19,381-388. Clark, M. R., Clarkson, S . B., Ory, P. A., Stollman, N., and Goldstein, I. M. (1989b). Molecular basis for a polymorphism involving Fc receptor I1 on human monocytes. J . Zmmunol. 143,1731-1734. Cleary, P. P., and Heath, D. G. (1990).Type I1 immunoglobulin receptor and its gene. In “Bacterial Immunoglobulin Binding Proteins” Vol. l., Chapter 7, pp. 83-99. (M. D. P. Boyle, ed.), Academic Press, San Diego.

68


Colten, H. R., Borsos, T., and Rapp, H. J . (1969). Titration of the first component of complement on a molar basis: Suitability of IgM and unsuitability of IgG hemolysis as sensitiser. lnamunochemistry 6,461-467. Conrad, D. H., Wingard, J. R., and Ishizaka, T. (1983). The interaction of human and rodent IgE with the human basophil IgE receptor. J . lmmunol. 130,327-333. Cooper, N. R. (1985). The classical complement pathway: Activation and regulation of the first complement component. Ado. lmmunol. 37,151-216. Cosio, F. G., Ackerman, S . K., Douglas, S. D., Friend, P. S., and Michael, A. F. (1981). Soluble immune complexes binding to human monocytes and polymorphonuclear leucocytes. Immunology 44,773-780. Crockett-Torabi, E., and Fantone, J. C. (1990).Soluble and insoluble immune complexes activate human neutrophil NADPH oxidase by distinct Fcy receptor-specific mechanisms. J . lmmunol. 145,3026-3032. Dangl, J. L., Wensel, T. G., Morrison, S. L., Stryer, L., Herzenberg, L. A., and Oi, V. T. (1988). Segmental flexibility and complement fixation of genetically engineered chimaeric human, rabbit and mouse antibodies. EMBO J . 7,1989-1994. Davies, D. R., Padlan, E. A,, and Sheriff, S. (1990).Antibody-antigen complexes. Annu. Reu. Biochem. 59,439-473. Davis, A. C., and Shulman, M. J . (1989). IgM-Molecular requirements for its assembly and function. lmmunol. Today 10, 118-128. Davis, A. C., Roux, K. H., and Shulman, M. J. (1988).On the structure ofpolymeric IgM. Eur. J . lmmunol. 18,1001-1008. Debets, J. M. H., Van de Winkel, J. G . J., Ceuppens, J. L., Dieteren, I. E. M., and Buurnian, W. A. (1990). Cross-linking of both FcyRI and FcyRII induces secretion of tumor necrosis factor by human nionocytes, requiring high-affinity Fc-FcyR interactions. J . lmmunol. 114, 1304-1310. Deisenhofer, J. (1981). Crystallographic refinement and atomic models of a human Fc fragment and its complex with fragment B of protein A from Staphylococcus aureus at 2.9 and 2.8 resolution. Biochemistry 20,2361-2370. Deisenhofer, J., Colman, P. M., Huber, R., Haupt, H., and Schwick, G . (1976). Crystallographic structural studies of a human Fc-fragment. I. An electron density map at 4 i resolution and partial model. Hoppe-Seyler’s Z. Physiol. Chem. 357,435-445. Deisenhofer, J., Jones, T. A., Huber, R., Sjodahl, J., and Sjoquist, J. (1978). Crystallization, crystal structure analysis and atomic model ofthe complex formed by a human Fc fragment and fragment B of protein A from Staphylococcus aureus. Hoppe-Seyler’s Z. Physiol. Chem. 359,975-979. Delespesse, G . , Sarfati, M., and Hofstetter, H. (1989). Human IgE-binding factors. lmmunol. Today 10,159-164. Del Prado, J. M., Jimeno, L., Obispo, T., and Carriera, J. (1991). Monoclonal antibodies against human IgE. Identification of an epitope sharing properties with the highaffinity receptor binding site. Mol. lmmunol. 28,839-844. Dower, S. K., and Segal, D. M. (1981).Clq binding to antibody-coated cells: Predictions from a simple multivalent model. M o l . lmmunol. 18,823-829. Duncan, A. R., and Winter, G . (1988). The binding site for C l q on IgG. Nature (London) 332,738-740. Duncan, A. R., Woof, J. M., Partridge, L. J., Burton, D. R., and Winter, G . (1988). Localization ofthe binding site for the human high-affinity Fc receptor on IgG. Nature (London)332,563-564. Easterbrook-Smith, S. B., Vandenberg, R. J., and Alden, J. R. (1988). The role of Fc-Fc interactions in insoluble immune complex formation and complement activation. Mol. lmmunol. 25,1331-1337.


69

Elliot, E. V., Pinder, A., Stevenson, F. K., and Stevenson, G. T. (1978). Synergistic cytotoxic effects of antibodies directed against different cell surface determinants. Immunology 34,405-409. Ely, K. R., Colman, P. M., Abola, E. E., Hess, A. C., Peabody, D. S., Parr, D. M., Connell, G . E., Laschinger, C., A., and Edmundson, A. B. (1978). Mobile Fc region in the Zie IgG2 cryoglobulin: Comparison ofcrystals ofthe F(ab’)zfragment and intact immunoglobulin. Biochemistry 17,820-823. Emanuel, E. J., Brampton, A. D., Burton, D. R., and Dwek, R. A. (1982). Formation of complement subcomponent Clq-immunoglobulin G complexes. Thermodynamic and chemical modification studies. Biochem. J. 205,361-372. Endo, S., and Arata, Y. (1985).Proton NMR study of human immunoglobulins G1 and their proteolytic fragments: Structure of the hinge region and effects of hinge deletion on internal flexibility. Biochemisty 24, 1561-1568. Emtell, M., Myhre, E. G., Sjobring U . , and Bjorck, L. (1988). Streptococcal protein G has affinity for both Fab- and Fc-fragments of human IgG. Mol. Immunol. 25, 121-126. Fahey, J. L., and Robinson, A. G. (1963).Factors controlling serum y-globulin concentration. J. E x p . Med. 118,845-868. Fahnestock, S. R., Alexander, P., Nagle, J., and Filpula, D. (1986). Gene from an immunoglobulin-binding protein from a group G streptococcus. J . Bacteriol. 167,870-880. Fanger, M. W., Shen, L., Pugh, J., and Bernier, G. M. (1980).Subpopulations of human peripheral granulocytes and monocytes express receptors for IgA. Proc. Natl. Acad. Sci. U.S.A.77,3640-3644. Fanger, M. W., Goldstine, S. N., and Shen, L. (1983). Cytofluorographic analysis of receptors for IgA on human polyinorphonuclear cells and monocytes and the correlation of receptor expression with phagocytosis. M o l . Immunol. 20, 1019-1027. Fanger, M. W., Shen, L., Graziano, R. F., and Guyre, P. M. (1989).Cytotoxicity mediated by human Fc receptors for IgG. Immunol. Today 10,92-99. Farber, D. L., and Sears, D. W. (1991). Rat CD16 is defined by a family of class 111 Fcy receptors requiring co-expression of heteroprotein subunits. J . Immunol. 146,43524361. Feinstein, A., Munn, E. A., and Richardson, N. E. (1974). The three dimensional conformation of yM and yA globulin molecules. Ann. N . Y . Acad. Sci. 190, 104107. Feinstein, A,, and Richardson, N. E. (1981).Tertiary structure of the constant regions of immunoglobulins in relation to their functions. Monogr. Allergy 17,28-47. Feinstein, A., Richardson, N. E., Gorick, B. D., and Hughes-Jones, N. C. (1983).Immunoglobulin M conformational change is a signal for complement activation. In “Protein Conformation as an Immunological Signal” (F. Celada, V. N. Schuhmaker, and E. Sercarz, eds.), pp. 47-57. Plenum, New York. Feinstein, A., Richardson, N. E., and Taussig, M. J. (1986).Immunoglobulin flexibility in complement activation. Immunal. Today 7, 169-174. Ferranini, M., Hoffman, T., Fu, S. M., Winchester, R., and Kunkel, H. G. (1977). Receptors for IgM on certain human B 1ymphocytes.J. Imrnunol. 119,1525-1529. Ferreri, N. R., Howland, W. C., and Spiegelberg, H. L. (1986).Releaseofleukotrienes C4 and Bq and prostaglandin Ez from human monocytes stimulated with aggregated IgG, IgA and IgE. J . Immunol. 136,4188-4193. Finbloom, D. S., and Metzger, H. (1982).Binding of immunoglobulin E to the receptor on rat peritoneal macrophages. J . Immunol. 129,2004-2008. Folkerd, E. J., Gardner, B., and Hughes-Jones, N. C. (1980). The relationship between the binding ability and the rate of activation of the complement component C1. Immunology 41,179-185.

70


Forsgren, A., and Sjoquist, J. (1966). Protein A from S. aureus. I. Pseudoimmune reaction with human gammaglobulin. J . Immunol. 97,822-827. Fries, L. F., Hall, R. P., Lawley, T. J., Crabtree, G. R., and Frank, M. M. (1982). Monocyte receptors for the Fc portion of IgG studied with monomeric human IgG1: Normal in vitro expression of Fc receptors in HLA-BWDrw 3 subjects with defective FCmediated in vivo clearance. J . Immunol. 129, 1041-1049. Gadd, K. J., and Reid, K. B. M. (1981a). Importance of the integrity of the inter-heavy disulphide bond of rabbit IgG in the activation pathway of complement by the F(ab’)z regions of rabbit IgG antibody in immune aggregates. Immunology 4 3 , 7 5 4 2 . Gadd, K. J., and Reid, K. B. M. (1981b). Binding of C3 to antibody-antigen aggregates after activation of the alternative pathway in human serum. Biochem. J . 189,471. Garred, P., Michaelsen, T. E., and Aase, A. (1989). The IgG subclass pattern of complement activation depends on epitope density and antibody and complement activation. Scand. J. Immunol. 30,379-382. Garred, P., Michaelsen, T. E., Aase, A., and Mollnes, T. E. (1990). C3, C4 and the terminal complement complex differ from C l q by binding predominantly to the antigenic part of solid phase immune complexes. J . Immunol. 144,198-203. Gomi, H., Hozumi, T., Hattori, S., Tagawa, C., Kishimoto, F., and Bjorck, L. (1990). The gene sequence and some properties of protein H. A novel IgG-binding protein. J . Immunol. 144,4046-4052. Gordon, J., Flores-Romo, L., Cairns, J. A., Millsum, M. J., Lane, P. J., Johnson, G. D., and MacLennan, I. C. M . (1989). CD23: A multi-functional receptorilymphokine? Immunol. Today 10,153-157. Gorter, A., Hiemstra, P. S., Leijh, P. C. J., Van der Sluys, M. E., Van den Barselaar, M. T., Van Es, L. A., and Daha, M. R. (1987). IgA- and secretory IgA-opsonized S. aureus induce a respiratory burst and phagocytosis by polymorphonuclear leucocytes. Immunology 61,303-309. Gosselin, E. J., Brown, M. F., Anderson, C. L., Zipf, T. F., and Guyre, P. M. (1990). The monoclonal antibody 41H16 detects the Leu 4 responder form of human FcyRII. J . Immunol. 144,1817-1822. Greenspan, N. S., Monafo, W. J., and Davie, J. M. (1987). Interaction of IgG3 antistreptococcal group A carbohydrate (GAC) antibody with streptococcal group A vaccine: Enhancing and inhibiting effects of anti-GAC, anti-isotypic and anti-idiotypic antibodies. J. Immunol. 138,285-292. Gregory, L., Davis, K. G., Sheth, B., Boyd, J., Jefferis, R., Nave, C., and Burton, D. R. (1987). The solution conformations of the subclasses of human IgG deduced from sedimentation and small angle X-ray scattering studies. Mol. Immunol. 24,821-829. Gupta, S., Platsoucas, C. D., and Good, R. A. (1979). Receptors for IgA on a subpopulation of human B lymphocytes. Proc. Natl. Acad. Sci. U.S.A. 76,4025-4028. Cuss, B., Eliasson, M., Olsson, A., Uhlen, M., Frej, A.-K., Jornvall, H., Flock, J.-I., and Lindberg, M. (1986). Structure of the IgG-binding regions of streptococcal protein G. E M B O J . 5,1567-1575. Guyre, P. M., Morganelli, P. M., and Miller, R. (1983). Recombinant immune interferon increases immunoglobulin G Fc receptors on cultured human mononuclear phagocytes. J . C h . Invest. 72,393-397. Hakimi, J., Seals, C., Kondas, J. A,, Pettine, L., Danho, W., and Kochan, J. (1990). The (Y subunit of the human IgE receptor (Fc,RI) is sufficient for high affinity IgE binding. J. Biol. Chem. 265,22079-22081. Heath, D. G., and Cleary, P. P. (1987). Cloning and expression of the gene for an IgG FC receptor protein from group A streptococcus. Infect. Immun. 55,1233-1238.


71

Heath, D. G., and Cleary, P. P. (1989). Fc-receptor and M-protein genes of group A streptococci are products of gene duplication. Proc. N u t l . Acad. Sci. U.S.A. 86,47414745. Helm, B. A., Marsh, P., Vercelli, D., Padlan, E., Could, H., and Geha, R. (1988).The mast cell binding site on human immunoglobulin E. Nuture (London)331, 180-183. Helm, B. A., Kebo, D., Vercelli, D., Glovsky, M. M., Could, H., Ishizaka, K., Geha, R., and Ishizaka, T. (1989). Blocking of‘ passive sensitization of human mast cells and basophil granulocytes with IgE antibodies by a recombinant human &-chainfragment of 76 amino acids. Proc. Natl. Acad. Sci. U.S.A.86,9465-9469. Helm, B. A., Ling, Y.,Teale, C., Padlan, E. A., and Bruggemann, M. (1991). The nature and importance of the inter-s chain disulphide bonds in human IgE. Eur.J.Immunol. 21,1543-1548. Hempstead, B. L., Parker, C. W., and Kulczycki, A., Jr. (1981). The cell surface receptor for immunoglobulin E. Effect of tunicamycin on molecular properties of receptor from rat basophilic leukemia cells. J . Biol. Chem. 256, 10717-10723. Hibbs, M. L., Bonadonna, L., Scott, B. M., McKenzie, I. F. C., and Hogarth, P. M. (1988). Molecular cloning of a human immunoglobulin G Fc receptor. Proc. Natl. Acud. Sci. U.S.A. 85,2240-2244. Hibbs, M. L., Selvaraj, P., Carpen, O., Springer, T. A., Kuster, H., Jouvin M.-H. E., and Kinet, J.-P. (1989). Mechanisms for regulating expression of membrane isoforms of FcyRIII (CD16).Science 246,1608-1611. Hienistra, P. S., Gorter, A,, Stuurman, M. E., van Es, L. A,, and Daha, M. R. (1987). Activation of the alternative pathway of complement by human serum IgA. Eur. J . Immunol. 17,321-326. Hiemstra, P. S., Biewenga, J., Gorter, A,, Stuurman, M. E., Faber, A., van Es, L. A., and Daha, M. R. (1988).Activation of complement by human serum IgA, secretory IgA and IgAl fragments. Mol. Immunol. 25,527-533. Hiemstra, P. S., Rits, M., Gorter,A., Stuurman, M. E., Hoekzema, R., Bazin, H.,Vaerman, J. P., Van Es, L. A,, and Daha, M. R. (1990). Rat polymeric IgA binds C l q , but does not activate C1. Mol. Immunol. 27,867-874. Hoekzema, R., Martens, M., Brouwer, M. C., and Hack, C. E. (1988). The distortive mechanism for the activation of complement component C1 supported by studies with a monoclonal antibody against the “arms” of C l q . Mol. Immunol. 25,485-494. Holowka, D., and Baird, B. (1983). Structural studies on the membrane-bound immunoglobulin E-receptor complex. 2. Mapping of distances between sites on IgE and the membrane surface. Biochemistry 22,3475-3484. Holowka, D., Conrad, D. H., and Baird, B. (1985). Structural mapping of membranebound immunoglobulin E-receptor complexes: Use of monoclonal anti-IgE antibodies to probe the conformation of receptor-bound IgE. Biochemistry 24,6260-6267. Horgan, C., Brown, K., and Pincus, S. (1990). Alternation in H chain V region affects complement activation by chimeric antibodies. J . Immunol. 145,2527-2532. Hosoi, S., Circolo, A., and Borsos, T. (1987). Activation of human C1: Analysis with Western blotting reveals slow self-activation. J . Immunol. 139, 1602-1608. Howard, F. D., Rodewald, H.-R., Kinet, J.-P., and Reinherz, E. L. (1990).CD35 subunit can substitute for the y subunit of F ~ receptor E type I in assembly and functional expression of the high-affinity IgE receptor: Evidence for interreceptor complementation. Proc. Natl. Acad. Sci. U.S.A.87,7015-7019. Howard, J. C., and Hughes-Jones, N. (1988).Complement mediated lysis with monoclonal antibodies. Prog. Allergy 45, 1-15. Howard, J. C., Butcher, G. W., Galfre, G., Milstein, C., and Milstein, C. P. (1979).

72


Monoclonal antibodies as tools to analyse the serological and genetic complexities of major transplantation antigens. Immunol. Reu. 47, 139-174. Huber, R., Deisenhofer, J., Colman, P. M., Masaak, M., and Palm, W. (1976). Crystallographic structure studies of an IgG molecule and an Fc fragment. Nature (London) 264,415-420. Hughes-Jones, N. C. (1977). Functional affinity constants of the reaction between ‘‘’I labelled C l q and C l q binders and their use in the measurement of plasma C l q concentrations. Immunology 32, 191-198. Hughes-Jones, N. C., and Gardner, B. (1978). The reaction between the complement subcomponent Clq, IgG complexes and polyionic molecules. Immunology 34,459463. Hughes-Jones, N. C., and Gardner, B. (1979). Reaction between the isolated globular subunits of the complement component C l q and IgG-complexes. Mol. Zmmunol. 16, 697-701. Hughes-Jones, N. C., Gorick, B. D., and Howard, J. (1983). The mechanism ofsynergistic complement mediated lysis of red cells by monoclonal IgG antibodies. Eur. J. Zmmunol. 13,635-641. Hughes-Jones, N. C., Gorick, B. D., Howard, J. C., and Feinstein, A. (1985). Antibody density on rat red cells determines the rate of activation ofthe complement component C1. Eur. J. Immunol. 15,976-980. Huizinga, T. W. J., Van der Schoot, C . E., Jost, C., Klaasen, R., Kleijer, M., von dem Borne, A. E. G . K., Roos, D., and Tetteroo, P. A. T. (1988). The PI-linked receptor FcRIII is released on stimulation of neutrophils. Nature (London) 333,667-669. Huizinga, T. W. J., Kerst, M., Nuyens, J. H., Vlug, A., von dem Borne, A. E. G. K., Roos, D., and Tetteroo, P. A. T. (1989a). Binding characteristics of dimeric IgG subclass complexes to human neutrophils. J. Immunol. 142,2359-2364. Huizinga, T. W. J., Kleijer, M., Roos, D., and von dem Borne, A. E. G . K. (1989b). Differences between FcRIII of human neutrophils and human K/NK lymphocytes in relation to the NA antigen system. In “Leucocyte Typing IV” (W. Knapp et al., eds.), p. 582-585. Oxford Univ. Press, London and New York. Huizinga, T. W. J., Van Kemanade, F., Koenderman, L., Dolman, K. M., von dem Borne, A. E. G . K., Tetteroo, P. A. T., and Roos, D. (1989~). The 40-kDa Fcy receptor (FcRII) on human neutrophils is essential for the IgG-induced respiratory burst and IgGinduced phagocytosis. f . Zmmunol. 142,2365-2369. Hulett, M. D., Osman, N., McKenzie, I. F. C., and Hogarth, P. M. (1991). Chimeric Fc receptors identify functional domains of murine FcyRI. J. Zmmunol. 147, 1863-1868. Iida, K., Fujita, T., Inai, S., Sasaki, M., Kato, T., and Kobayashi, K. (1976). Complement fixing properties of IgA myeloma proteins and their fragments: The activation of complement through the classical pathway. Zmmunochemistry 13,747-752. Ikuta, K., Takami, M., Kim, C. W., Honjo, T., Miyoshi, T., Tagaya, Y., Kawabe, T., and Yodoi, J. (1987). Human lymphocyte Fc receptor for IgE: Sequence homology of its cloned cDNA with animal lectins. Proc. Natl. Acad. Sci. U.S.A.84,819-823. Imai, H., Chen, R. J., Wyatt, R. J., and Rifai, A. (1988). Lack ofcomplement activation by human IgA immune complexes. Clin. E x p . Zmmunol. 73,479-483. Isenman, 0 .E., Dorrington, K. J., and Painter, R. H. (1975).The structure and function of immunoglobulin domains. 11. The importance of interchain disulphide bonds and the possible role of molecular flexibility in the interaction between immunoglobulin G and complement. J . Zmmunol. 114,1726-1729. Ishizaka, T., and Ishizaka, K. (1975). Biology of immunoglobulin E. Prog. Allergy 19, 60- 121.


73

Ishizaka, T., Tada, T., and Ishizaka, K. (1968).Fixation of C‘ and C’la by rabbit yG and yM antibodies with particulate and soluble antigens.). Immunol. 100, 1145-1 153. Ishizaka. T., Helm, B., Hakimi, J., Niebly, J., Ishizaka, K., and Gould H. (1986). Biological properties o f a recombinant human immunoglobulin &-chainfragment. Proc. Natl. Acad. Sci. U.S.A. 83,8323-8327. Ito, W., and Arata, Y. (1985). Proton NMR studies on the dynamics ofthe conformation of the hinge segment of human G 1 immunoglobulin. Biochemistry 24,6260-6267. Jarvis, G. A., and McLeod Griffiss, J. (1989). Human IgAl initiates complementmediated killing of Neisseria meningitidis. J . Immunol. 143, 1703-1709. Jefferis, R. (1986). Polyclonal and monoclonal antibody reagents specific for IgG subclasses. Monogr. Allergy 19,71-95. Joseph, M., Capron, A., Ameisen, M.-C., Capron, M., Vorng, H., Pancre, V., Kusnierz, J.-P., and Auriault C. (1986).The receptor for IgE on blood platelets. Eur.J.Immunol. 16,306-312. Kabat, E. A., Wu, T. T., Reid-Miller, M., Perry, H. M., and Gottesman, K. S. (1987). . Sequences of Proteins of lmmunological Interest.” U S . Department of Health and Human Services, Public Health Service, National Institutes of Health, Washington, D.C. Karas, S. P., Rosse, W. F., and Kurlander, R. J. (1982).Characterization of the IgG-Fc receptor on human platelets. Blood 60, 1277-1282. Kikutani, H., Inui, S., Sato, R., Barsumian, E. L., Owaki, H., Yamasaki, K., Kaisho, T., Uchibayashi, N., Hardy, R. R., Hirano, T., Tsunasawa S., Sakiyama, F., Suemura, M., and Kishimoto, T. (1986). Molecular structure of human lymphocyte receptor for immunoglobulin E. Cell (Cambridge,Mass.) 47,657-665. Kilchherr, E., Hofniann, H., Steigemann, W., and Engle, J. (1985). A structural model of the collagen-like region of C l q comprising the kink region and the fiber-like packing of the six triple helices. J. Mol. B i d . 186,403-415. Kilian, M., Mestecky, J., and Russel, M. W. (1988). Defense mechanisms involving Fc-dependent functions of IgA and their subversion by IgA proteases. Microbiol. Rev. 52,296-303. Kimata, H., and Saxon, A. (1988). Subset of natural killer cells is induced by immune complexes to display Fc receptors for IgE and IgA and demonstrates isotype regulatory functi0n.J. Clin. Znoest. 82, 160-167. Kimberly, R. P., Tappe, N. J., Merriam, L. T., Redecha, P. B., Edberg, J. C., Schwartzman, S., and Valinsky, J. E. (1989). Carbohydrates on human Fcy receptors. Interdependence of the classical IgG and nonclassical lectin-binding sites on human FcyRIII expressed on neutrophils. J. Immunol. 142,3923-3930. Kimberly, R. P., Ahlstrom, J. W., Click, M. E., and Edberg, J. C. (1990).The glycosyl phosphatidylinositol-linked FcyRIIIpMNmediates transmembrane signaling events distinct from FcyR1I.J. Exp. Med. 171, 1239-1255. Kimberly, R. P., Edberg, J. C., Redecha, P. B., and Barinsky, M. (1991). FcyRIII-A glycoforms have distinct affinities for ligand. FASEB /. 5,5658. Kindt, G. C., Van d e Winkel, J. G. J., Moore, S. A,, and Anderson, C. L. (1991).Identification and structural characterization of Fcy receptors on pulmonary alveolar macrophages. Am. ]. Physiol. 260,6. Kinet, J.-P. (1989).Antibody-cell interactions: Fc receptors. Cell (Cambridge,Mass.) 57, 35 1-354. Kinet, J,-P., and Metzger, H. (1990).Genes, structure and actions of the high-affinity Fc receptor for immunoglobulin E. In “Fc Receptors and the Action of Antibodies” (H. Metzger, ed.), pp. 239-259. Am. Soc. Microbiol., Washington, D.C. ‘1

74


Kinet, J.-P., Blank, U., Ra C., White, K., Metzger, H., and Kochan, J. (1988). Isolation and characterization of cDNAs coding for the beta subunit of the high-affinity receptor for immunoglobulin E. Proc. Natl. Acad. Sci. U.S.A. 85,6483-6487. Kipps, T. J., Parham, P., Punt, J., and Herzenberg, L. (1985). Importance ofimmunoglobulin isotype in human antibody-dependent, cell-mediated cytotoxicity directed by murine monoclonal antibodies. J. E x p . Med. 161, 1-17. Klein, M., Haeffner-Cavaillon, N., Isenman, D. E., Rivat, C., Navia, M., Davies, D. R., and Dorrington, K. J. (1981). Expression of biological effector functions by IgG molecules lacking the hinge region. Proc. Natl. Acad. Sci. U.S.A.78,524-528. Kochan, J., Pettine, L. F., Hakimi, J., Kishi, K., and Kinet, J.-P. (1988) Isolation of the gene coding for the a subunit of the human high affinity IgE receptor. Nucleic Acids Res. 16,3584. Koolwijk, P., Spierenburg, G . T., Frasa, H., Boot, J. H. A., Van de Winkel, J. G. J.. and Bast, B. J. E. G . (1989). Interaction between hybrid mouse monoclonal antibodies and the human high-affinity IgG FcR, huFcyRI, on U937: Involvement of only one of the mIgG heavy chains in receptor binding. J. Zmmunol. 143,1656-1662. Koolwijk, P., Van de Winkel, J. G. J.. Pfefferkorn, L. C., Jacobs, C. W. M., Otten, I., Spierenburg, G. T., and Bast, B. J. E. G. (1991). Induction ofintracellular Ca2+mobilization and cytotoxicity by hybrid mouse monoclonal antibodies. FcyRII regulation of FcyRI-triggered functions or signalling?J. Zmmunol. 147,595-602. Kulczycki, A., and Vallina, V. L. (1981). Specific binding of non-glycosylated IgE to FCE receptor. Mol. Zmmunol. 18,723-731. Kurlander, R. J., and Batker, J. (1982). The binding of human immunoglobulin G1 monomer and small, covalently cross-linked polymers of immunoglobulin G1 to human peripheral blood monocytes and polymorphonuclear leukocytes. J . Clin. Znoest. 69, 1-8. Kurlander, R. J., Haney, A. F., and Gartrell, J. (1984). Human peritoneal macrophages possess two populations of IgG Fc receptors. Cell. Zmrnunol. 86,479-490. Kurosaki, T., and Ravetch, J. V. (1989).A single amino acid in the glycosyl phosphatidylinositol attachment domain determines the membrane topology of FcyRIII. Nature (London)342,805-807. Kuster, H., Thompson, H., and Kinet, J.-P. (1990). The gene for the human high affinity IgE receptor y subunit. Characterization and expression: Definition of a new gene family. J. Biol. Chem. 265,6448-6452. Lancet, D., Isenman, D., Sjodahl, J. I., Sjoquist, J., and Pecht, I. (1978). Interaction between staphylococcal protein A and immunoglobulin domains. Biochem. Biophys. Res. Commun. 85,608-614. Lanier, L. L., Cwirla, S., Yu, G., Testi, R., and Phillips, J. H. (1989a). Membrane anchoring of a human IgG Fc receptor (CD16) determined by a single amino acid. Science 246,1611-1613. Lanier, L. L., Yu, G . ,and Phillips, J. H. (1989b). Co-association of CD35 with a receptor ( 0 1 6 )for IgG Fc on human natural killer cells. Nature (London) 342,803-805. Lanier, L. L., Yu, G., and Phillips, J . H. (1991). Analysis of FcyRIII (CD16) membrane expression and association with CD31;and FceRI-y by site-directed mutati0n.J. Zmmunol. 146,1571-1576. Leatherbarrow, R. J., and Dwek, R. A. (1983).The effect of aglycosylation on the binding of niouse IgG to staphylococcal protein A. FEBS Lett. 164,227-230. Leatherbarrow, R. J., Rademacher, T. W., Dwek, R. A., Woof, J. M., Clark, A., Burton, D. R., Richardson, N., and Feinstein, A. (1985). Effector functions of a monoclonal aglycosylated mouse IgC2a: Binding and activation of complement component C1 and interaction with human monocyte Fc receptor. Mol. Zmmunol. 22,407-415.


75

Leeuwenberg, J. F. M., Lems, S. P. M., and Capel, P. J. A. (1987). Anti-T3 induced cytotoxicity: The role of target cell Fc-receptors in the lysis of autologous monocytes and the Fc-independent lysis of T3-positive target cells. Transplant. Proc. 19, 428431. Leeuwenberg, J. F. M., Van d e Winkel, J . G. J., Jeunhomme, T. M. A. A., and Buurman, W. A. (1990). Functional polymorphism of IgG FcRII (CD32) on human neutrophils. Immunology 71,301-304. Letellier, M., Sarfati, M., and Delespesse, G. (1989). Mechanisms of formation of IgEbinding factors (soluble CD23). I. FceRII bearing B cells generate IgE-binding factors of different molecular weights. M o l . Zmmunol. 26, 1105-11 12. Liu, A. Y., Robinson, R. R., Hellstrom, K. E., Murray, E. D., Chang, C. P., and Hellstrom, I. (1987). Chimeric mouse-human IgGl antibody that can mediate lysis of cancer cells. Proc. Natl. Acud. Sci. U.S.A. 84,3439-3443. Lowe, J., Bird, P., Hardie, D., Jefferis, R., and Ling, N . R. (1982). Monoclonal antibodies to determinants on human gamma chains: Properties of antibodies showing subclass restriction or subclass specificity. Immunology 47, 329-335. Ludin, C . , Hofstetter, H., Sarfati, M., Levy, C. A., Suter, U., Alainio, D., Kilchherr, K., Frost, H., and Delespesse, G. (1987).Cloning and expression ofthe cDNA coding for a human lymphocyte IgE receptor. E M B O J .6, 109-114. Lukas, T. J., Mufioz, H., and Erickson, B. W. (1981).Inhibition of C1-mediated immune hemolysis by monomeric and dimeric peptides from the second constant domain of human immunoglobulin G. J. Immunol. 127,2555-2560. Lund, J., Tanaka, T., Takahashi, N., Sarmay, C . , Arata, Y., and Jefferis, R. (1990). A protein structural change in aglycosylated IgG3 correlates with loss of huFcyRI and huFcyRIII binding and/or activation. Mol. Immunol. 27, 1145-1153. Lund, J., Winter, G., Jones, P. T., Pound, J. D.,Tanaka, T., Walker, M. R., Artymiuk, P. J., Arata, Y., Burton, D. R., Jefferis, R., and Woof, J. M. (1991). Human FcyRI and FcyRII interact with distinct but overlapping sites on human IgG. J. Immunol. 147, 26572662. Maliszewski, C . R., Shen, L., and Fanger, M. W. (1985). The expression ofreceptors for IgA on human monocytes and calcitriol-treated HL-60 cells. J. Immunol. 135, 38783881. Maliszewski, C. R., March, C. J., Shoenborn, M. A., Ginipel, S., and Shen. L. (1990). Expression cloning of a human Fc receptor for IgA. J. E x p . Med. 172,1665-1672. Marquart, M., Deisenhofer, J,, Huber, R., and Palm, W. (1980). Crystallographic refinement and atomic models of the intact -immunoglobulin molecule Kol and its antigen binding fragment at 3.0 and 1.9 A resolution. J. Mol. Biol. 141, 369391. Mathur, A,, Lynch, R. G., and Kohler, G. (1988a). The contribution of constant region domains to the binding of murine IgM to Fc, receptors on T cells. J . lmmunol. 140, 143-147. Mathur, A., Lynch, R. G., and Kohler, G. (1988b). Expression, distribution and specificity of Fc receptors for IgM on murine B ce1ls.J. Immunol. 141,1855-1862. Matsuda, H., Nakamura, S., Ichikawa, Y., Kozai, K., Takano, R., Nose, M., Endo, S., Nishimura, Y., and Arata, Y. (1990). Protein NMR studies of the structure of the Fc fragment of human immunoglobulin G1: Comparisons of native and recombinant proteins. Mol. Immunol. 27,571-579. Mazengera, R. L., and Kerr, M. A. (1990). The specificity of the IRA receptor purified from human neutrophils. Biochem. J. 272,159-165. Metzger, H. (1978). The effect ofantigen on antibodies: Recent studies. Contemp. Top. Mol. lmmunol. 7, 119-152.

76


Metzger, H. (1988). Molecular aspects of receptors and binding factors for IgE. Ado. Immunol. 43,277-312. Metzger, H., Kinet, J.-P., Perez-Montfort, R., Rivnay, B., and Wank, S. A. (1983). A tetrameric model for the structure of the mast cell receptor with high affinity for IgE. Prog. Immunol. 5,493-501. Metzger, H., Alcaraz, G., Hohman, R., Kinet, J.-P., Pribluda, V., and Quarto, R. (1986). The receptor with high affinity for immunoglobulin E. Annu. Reo. Immunol. 4,419470. Michaelsen, T. E., Aase, A., Westby, C., and Sandlie, I. (1990). Enhancement ofcomplement activation and cytolysis of human IgG3 by deletion of hinge exons. Scand. J . Immunol. 32,517-528. Michaelsen, T. E., Garred, P., and Aase, A. (1991). Human IgG subclass pattern of inducing complement-mediated cytolysis depends on antigen concentration and to a lesser extent on epitope patchiness, antibody affinity and complement concentration. Eur.J.Immunol. 21, 11-16. Miller, L., Blank, U., Metzger, H., and Kinet, J.-P. (1989).Expression of high-affinity binding of immunoglobulin E by transfected cells. Science 244,334-337. Millet, I., Panaye, C., and Revillard, J.-P. (1988). Expression of receptors for IgA on mitogen-stimulated human T cells. Eur. J . Immunol. 18,621-626. Millet, I., Briere, F., Vincent, C., Rousset, F., Andreoni, C., De Vries, J. E., and Revillard, J. P. (1989). Spontaneous expression of a low affinity Fc receptor for IgA (Fc,R) on human B cells. Clin. E x p . Immunol. 76,268-273. Moks, T., Abrahmsen, L., Nilsson, B., Hellman, U., Sjoquist, J., and Uhlen, M. (1986). Staphylococcal protein A consists of five IgG-binding domains. Eur. J. Biochem. 165, 637-643. Moller, N. P. H. (1979). Fc-mediated immune precipitation. I. A new role for the Fc-portion of IgG. Immunology 38,631-640. Monteiro, R. C., Kubagawa, H., and Cooper, M. D. (1990).Cellular distribution, regulation and biochemical nature of an Fc, receptor in humans.J. E x p . Med. 171,597-613. Moretta, L., Ferranini, M., Durante, M. L., and Mingari, M. C. (1975). Expression o f a receptor of IgM by human T cells in oitro. Eur. J . Immunol. 5,565-569. Morrison, S. L., Canfield, S., Tan, L. K., and Tao, M.-H. (1989). Constant region mutations and their influence on effector functions. Adu. Gene Technol. p. 100-101. Muller-Eberhard, H. (1975). Complement. Annu. Rev. Biochem. 44,697. Muroaka, S., and Shulman, M. J. (1989). Structural requirements for IgM assembly and cytolytic activity: Effects of mutations in the oligosaccharide acceptor site at Asn402. J . Immunol. 142,695-701. Myhre, E. B., and Kronvall, G. (1980). Demonstration of a new type of immunoglobulin G receptor in Streptococcus zooepidemicus strains. Infect. Immun. 27,808-816. Nezlin, R. (1990). Internal movements in immunoglobulin molecules. Ado. Immunol. 48, 1-40. Nio, N., Seguro, K., Ariyoshi, Y., Nakanishi, K., Kita A., Ishiii, K., and Nakarnura, H. (1990). Inhibition of histamine release by synthetic human IgE peptide fragments: Structure-activity studies. In “Peptide Chemistry 1989” (N. Yanaihara, ed.), pp. 204208. Protein Res. Found., Osaka. Nissim, A., Jouvin, M.-H. E., and Eshhar, Z. (1991). Mapping of the high affinity FCE receptor binding site to the third constant region domain of IgE. EMBO J . 10,101-107. Norderhaug, L., Brekke, O.H., Bremnes, B., Sandin, R., Aase, A., Michaelsen, T. E., and Sandlie, I. (1991). Chimeric mouse human IgG3 antibodies with an IgC4-like hinge region induce complement-mediated lysis more efficiently than IgG3 with a normal hinge. Eur. J . Immunol. 21,2379-2384.


77

Nose, M., and Leanderson, T. (1989).Substitution ofasparagine 324 with aspartic acid in the Fc portion of mouse antibodies reduces their capacity fo C l y binding. Eur. J . Zmmunol. 19,2179-2181. Nose, M., and Wigzell, H. (1983). Biological significance of carbohydrate chains on monoclonal antibodies. Proc. Natl. Acad. Sci. U.S.A.80,6632-6636. Nose, M., Okuda, T., Gidlund, M., Ramstedt, U., Okada, N., Okada, H., Heyman, B., Kyogoku, M., and Wigzell, H. (1988).Mutant monoclonal antibodies with select alteration in complement activation ability. J . Immunol. 141,2367-2373. Odermatt, E., Berger, H., and Sano, Y. (1981). Size and shape of human C1 inhibitor. F E B S Lett. 131,283-285. O’Grady, J. H., Looney, R. J., and Anderson, C. L. (1986). The valence for ligand of the human mononuclear phagocyte 72kD high-affinity IgC Fc receptor is 0ne.J. Immunol. 137,2307-2310. Ohno, T., Kubagawa, H., Sanders, S. K., and Cooper, M. D. (1990).Biochemical nature of an Fc, receptor on human B-lineage cel1s.J. E r p . Med. 172,1165-1175. Oi, V. T., Vuong, T. M., Hardy, R., Reidler, J., Dangl, J., Herzenberg, L. A., and Stryer, L. (1984).Correlation between segmental flexibility and effector function ofantibodies. Nature (London)307,136-140. Okada, M., and Utsumi, S. (1989).Role for the third constant domain ofthe IgC H chain in activation ofconiplement in the presence o f C l inhibitor../. Immunol. 142,195-201. Okada, M., Udaka, M., and Utsumi, S. (1985).Cooperative interaction of subcomponents of the first component of complement with IgG: A functional defect ofdimeric Facb from rabbit IgC. Mol. Zmmunol. 22, 1399-1406. Okafor, G. 0..Turner, M. W., and Hay, F. C. (1974). Localisation of monocyte binding site of human immunoglobulin C. Nature (London)248,228-230. Olsson, A,, Eliasson, M., Guss, B., Nilsson, B., Hellmann, U., Lindberg, M., and Uhlen, M. (1987). Structure and evolution of the repetitive gene encoding streptococcal protein G. E u r . J . Biochem. 168,319-324. Orloff, D. G., Ra, C., Frank, S. J., Klausner, R. D., and Kinet, J.-P. (1990). Family of disulphide-linked dimers containing the 6 and r ) chains of the T-cell receptor and the y chain of Fc receptors. Nature (London)347, 189-191. Ory, P. A., Clark, M. R., Kwoh, E. E., Clarkson, S. B., and Goldstein, I. M. (1989). Sequences of complementary DNAs that encode the NA1 and NA2 forms of Fc receptor I l l on human neutrophi1s.J. Clin. Znaest. 84, 1688-1691. O’Shea, J . J., Weissman, A. M., Kennedy, I . C. S., and Ortaldo, J. R. (1991).Engagement of the natural killer cell IgC Fc receptor results in tyrosine phosphorylation ofthe 6 chain. Proc. Natl. Acad. Sci. U.S.A.88,350-354. Padeh, S., Jaffe, C. L., and Passell, J. H. (1991).Activation ofhuman monocytes via their sIgA receptors. Immunology 72, 188-193. Padlan, E. A., and Davies, D. R. 91986). A model of the Fc of IgE. Mol. Immunol. 23, 1063- 1075. Partridge, L. J., Woof, J. M., Jefferis, R., and Burton, D. R. (1988). The use of anti-IgG monoclonal antibodies in mapping the monocyte receptor site on IgC. Mol. Zmmunol. 23, 1365-1372. Passlick, B., Flieger, D., and Zeigler-Heitbrook. H. W. L. (1989). Identification and characterization of a novel monocyte subpopulation in human peripheral blood. Blood 74,2527-2534. Perez-Montfort, R., and Metzger, H. (1982).Proteolysis of soluble IgE-receptor complexes: Localization of sites on IgE which interact with the Fc receptor. Mol. Zmmutlol. 19, 1113-1125. Perkins, S. J. (1985). Molecular modelling of human complement subcomponent C l q

78


and its complex with Clr2Cls2 derived from neutron-scattering curves and hydrodynamic properties. Biochem. J . 228,13-26. Perkins, S. J., Nealis, A. S., and Sim, R. B. (1990a). Molecular modeling of human complement component C4 and its fragments by X-ray and neutron solution scattering. Biochemistry 29, 1167-1 175. Perkins, S. J., Smith, K. F., Nealis, A. S., Lachmann, P. J., and Harrison, R. A. (199Ob). Structural homologies of component C5 of human complement with components C3 and C4 by neutron scattering. Biochemistry 29, 1175-1180. Persson, M. A. A,, Caothien, R. H., and Burton, D. R. (1991). Generation of diverse high-affinity human monoclonal antibodies by repertoire cloning. Proc. Natl. Acad. Sci. U.S.A.88,2432-2436. Perussia, B., Dayton, E. T., Lazarus, R., Fanning, V., and Trinchieri, G. (1983). Immune interferon induces the receptor for monomeric IgGl on human monocytic and myeloid cells. J . E x p . Med. 158, 1092-1113. Peterson, L. H., and Conrad, D. H. (1985).Fine specificity, structure and proteolytic susceptibility ofthe human lymphocyte receptor for IgE. J . lmmunol. 135,2654-2660. Pinteric, L., Painter, R. H., and Connell, G. E. (1971).Ultrastructure ofthe Fc fragment of human immunoglobulin G . lmmunochemistry 8,1041-1045. Pollock, R. R., French, D. L., Metlay, J. P., Birshtein, B. K., and Scharff, M. D. (1990). Intravascular metabolism of normal and mutant mouse immunoglobulin molecules. Eur. J. lmmunol. 20,2021-2027. Poon, P. H., and Schumaker, V. N . (1991). Measurement of macromolecular interactions between complement subcomponents Clq, Clr, C l s and immunoglobulin M by sedimentation analysis using the analytical ultracentrifuge. J . Biol. Chem. 266,57235727. Poon, P. H., Schumaker, V. N., Phillips, M. L., and Strang, C. J. (1983).Conformation and restricted segmental flexibility of C1, the first component of human complement. J . Mol. B i d . 168, 563-577. Poon, P. H., Phillips, M. L., and Schumaker, V. N. (1985). Immunoglobulin M possesses two binding sites for complement subcomponent C l q and soluble 1:l and 2:l complexes are formed in solution at reduced ionic strength. J . B i d . Chem. 260,9357-9365. Pound, J. D., and Walker, M. R. (1990). Membrane Fc receptors for IgG subclasses. I n “The Human IgG Subclasses. Molecular Analysis of Structure, Function and Regulation” (F. Shakib, ed.), Chapter 6, pp. 111-133. Pergamon, Oxford. Pumphrey, R. (1986). Computer models of the human immunoglobulins. Shape and segmental flexibility. fmmunol. Toduy 7 , 174-178. Ra, C., Jouvin, M.-H. E., Blank, U., and Kinet, J.-P. (1989a). A macrophage Fcy receptor and the mast cell receptor for IgE share an identical subunit. Nature (London)341, 752-754. Ra, C., Jouvin, M.-H. E., and Kinet, J.-P. (1989b). Complete structure ofthe mouse mast cell receptor for IgE (FceRI) and surface expression of chimeric receptors (ratmouse-human) on transfected cells. J . Biol. Chem. 264, 15323-15327. Raeder, R., Faulmann, E. L., and Boyle, M. D. P. (1991a). Evidence for functional heterogeneity in IgG Fc-binding proteins associated with group A streptococci. J . lmmunol. 146,1247-1253. Raeder, R., Otten, R. A., and Boyle, M. D. P. (1991b). Isolation and partial characterization of a type IV bacterial immunoglobulin binding protein. Mol. Immunol. 28, 661-67 1. Rajan, S. S., Ely, K. R . , Abola, E. E., Wood, M. K., Colman, P. M., Athay, R. J., and Edmundson, A. B. (1983).Three-dimensional structure of the Mcg IgGl imniunoglobulin. Mol. Immunol. 20,6349-6356.


79

Randall, T. D., King, L. B., and Corley, R. B. (1990). The biological effects of IgM hexamer formation. Eur. J . Immunol. 20, 1971-1979. Ratnoff, W. D., Fearon, D. T., and Austen, K. F. (1983). The role of antibody in the activation of the alternative complement pathway. Springer Semin. Immunopathol. 6, 361-371. Ravetch, J. V., and Anderson, C. L. (1990). Fcy family: Proteins, transcripts and genes. In “Fc Receptors and the Action ofAntibodies” (H. Metzger, ed.), pp. 211-235. Am Soc. Microhiol., Washington, D.C. Ravetch, J. V., and Perussia, B. (1989). Alternative membrane forms of FcyRIII (CD16) on human NK cells and neutrophils: Cell type specific expression of two genes which differ in single nucleotide substitutions. J . E x p . Med. 170,481-497. Recht, B., Frangione, B., Franklin, E., and Van Loghein, E. (1981). Structural studies ofa human y3 myeloma protein (Goe) that binds staph protein A.J. Immunol. 127,917923. Reid, K. B. M. (1983). Proteins involved in the activation and control of the two pathways of human complement. Biochem. SOC. Trans. 11, 1-12. Reid, K. B. M., and Porter, R. R. (1981). The proteolytic activation systems of complement. Annu. Reo. Biochem. 50,433-464. Reid, K. B. M., Sim, R. B., and Raiers, A. P. (1977). Inhibition ofthe reconstitution of the hemolytic activity of the final component of human complement by a pepsin-derived fragment of subcomponent C l q . Biochem. J. 161,239-241. Reid, K. B. M., Gagnon, J., and Frampton, J . (1982). Completion of the amino acid sequences ofthe A and B chains ofsubcomponent C l q ofthe first component ofhuman complement. Biochem.J.203,559-569. Reidler, J. Uzgiris, E. E., and Kornberg, R. D. (1986). Two-dimensional crystals of immunoglobulins. In “Handbook of Experimental Immunology” (D. M. Weir, ed.), 4th ed., Chapter 17. Blackwell, Oxford. Reis, K. J.. Ayoub, E. M., and Boyle, M. D. P. (1984). Streptococcal Fc receptors. 11. Comparison ofthe reactivity of a receptor from a group C streptococcus with staphylococcal protein A. J. Immuiiol. 132,3098-3102. Reis, K. J., Siden, E. J., and Boyle, M. D. P. (1988). Selective colony blotting to expand bacterial surface receptors: Applications to receptors for rat immunoglobulins. BioTechniques 6, 130-136. Reis, K. J., Salpeter, J., and Boyle, M. D. P. (1990). Type IV bacterial immunoglobulinbinding proteins. In “Bacterial Immunoglobulin Binding Proteins” (M. D. P. Boyle, ed.), Vol. 1, Chapter 12, pp. 149-154. Academic Press, San Diego. Richards, M. L., and Katz, D. H. (1990). The binding of IgE to murine FcsRII is calcium-dependent but not inhibited by carbohydrate. /. Immzrnol. 144,2638-2646. Riechmann, L., Clark, M., Waldmann, H., and Winter, G . (1988). Reshaping human antibodies for therapy. N a t u r e (London)332,323-327. Riske, F., Hakimi, J., Mallamaci, M., Griffin, M., Pilson, B., Tobkes, N., Lin, P., Danho, W., Kochan, J., and Chizzonite, R. (1991). High affinity human IgE receptor (FceRI). Analysis of functional domains of the a-subunit with monoclonal antibodies. J . B i d . Cheni. 266, 11245-11251. Rits, M., Hiemstra, P. S., Bazin, H., van Es, L. A., Vaerman, J.-P., and Daha, M. R. (1988). Activation of rat complement by soluble and insoluble rat IgA immune complexes. E u r . J. Immunol. 18, 1873-1880. Rodwell, J . R., Tang, L.-H., and Schuniaker, V. N. (1980). Antigen valence and Fclocalised secondary forces in antibody precipitation. Mol. Immunol. 17,1591-1597. Romer, W., Rother, U.,and Roelcke, D. (1980). Failure of IgA cold agglutinin to activate C. Immunobiology 157,41-46.

80


Rudders, R. A,, and Andersen, J . (1982). IgD-Fc receptors on normal and neoplastic human B lymphocytes. Clin. Erp. Immunol. 50,579-586. Russell, M. W., and Mansa, B. (1989). Complement-fixing properties of hunian IgA antibodies. Scund. ]. Immunol. 30, 175-183. Ryazantsev, S. N., Vasiliev, V. D., Abramov, V. M., Franek, F., and Zav’yalov, V. P. (1989). Electron microscopy study of non-precipitating anti-dinitrophenyl pig antibodies. F E B S Lett. 244,291-295. Ryazantsev, S. N., Tishchenko, V., Vasiliev, V. D., Zav’yalov, V. P., and Abramov, V. M. (1990).Structure ofhuman myeloma IgG3 Kuc. Eur. J . Biochem. 190,393-399. Salmon, J. E., Kapur, S., and Kimberly, R. P. (1987). Opsonin-independent ligation of Fcy receptors. The 3G8-bearing receptors on neutrophils mediate the phagocytosis of concanavalin A-treated erythrocytes and nonopsonized Escherichia coli. J . E x p . Med. 166,1798-1813. Salmon, J. E., Edberg, J. C., Kimberly, R. P., Mensa, E., and Ryan, R. (1990).Fcy receptor 111 on human neutrophils. Allelic variants have functionally distinct capacities.]. Clin. Inuest. 85,1287-1295. Salmon, J . E., Brogle, N . L., Edberg, J . C., and Kimberly, R. P. (1991). Fcy receptor 111 induces actin polymerization in human neutrophils and primes phagocytosis mediated by Fcy receptor II.]. Immunol. 146,997-1004. Sanders, S. K., Kubagawa, H., Suzuki, T., Butler, J. L., and Cooper, M. D. (1987). IgM binding protein expressed on activated B cells. J . Immunol. 139, 188-193. Sandlie, I., Aase, A., Westby, C., and Michaelsen, T. E. (1989).C l q binding to chimeric monoclonal IgG3 antibodies consisting of mouse variable regions and human constant regions with shortened hinge containing 15 to 47 amino acids. E u r . ] . lmmunol. 19, 1599-1603. Sarma, R., and Laudin, A. G. (1982). A three dimensional structure of a human IgGl immunoglobulin at 4,& resolution: A computer fit ofvarious structural domains on the electron density map. J . A p p l . Crystallogr. 15,476-481. Sarmay, C., Jefferis, R., Klein, E., Benczur, M., and Gergely, J. (1985). Mapping the functional topography of IgG Fc with monoclonal antibodies: Localization of epitopes interacting with the binding sites of Fc receptor on human Kcells. Eur. J . Immunol. 15, 1037-1042. Scallon, B. J., Scigliano, E., Freedman, V. H., Miedel, M. C., Pan, Y.-C., Unkeless, J. C., and Kochan, J. P. (1989). A human immunoglobulin G receptor exists in both polypeptide-anchored and phosphatidylinositol-glycan-anchored forms. Immunology 86,5079-5083. Schneider, W. P., Wensel, T. G., Stryer, L., and Oi, V. T. (1988).Genetically engineered immunoglobulins reveal structural features controlling segmental flexibility. Proc. N a t l . Acad. Sci. U.S.A.85,2509-2513. Schneidernian, R. D., Lint, T. L., and Knight, K. L. (1990). Activation ofthe alternative pathway of complement by 12 different rabbit-mouse chimeric transfectoma IgA isotypes. J. Immunol. 145,233-237. Schroder, A. K., Nardella, F. A., Mannik, M., Svensson, M.-L., and Christensen, P. (1986). Interaction between streptococcal IgG Fc receptors and human and rabbit IgG domains. lmmunology 57,305-309. Schnmaker, V. N., Calcott, M. A., Spiegelberg, H. L., and Miiller-Eberhard, H. J. (1976). Ultracentrifuge studies of the binding of IgG of different subclasses to the Clq subunit of the first component of complement. Biochemistry 16,5175-5181. Schumaker, V. N., Zavodsky, P., and Poon, P. H. (1987).Activation ofthe first component of complement. Annu. Rev. Immunol. 5,21-42.


81

Schumaker, V. N., Phillips, M. L., and Hanson, D. C. (1991).Dynamic aspects of antibody structure. Mol. lmmunol. 28, 1347-1360. Schwarzbaum, S., Nissim, A., Alkalay, I., Ghozi, M. C., Schindler, D. G., Bergman, Y., and Eshhar, Z. (1989).Mapping of murine IgE epitopes involved in IgE Fcs receptor interactions. Eur. J . Immunol. 19, 1015-1023. Sedmak, D. D., Singh, U. N., Cosio, F. G., and Anderson, C. L. (1990). Immune coniplexes (IC) induce Fc IgG receptors (FcyR) on cultured human mesangial cells (MC). J . Am. SOC. Nephrol. 1,537. Sedmak, D. D., Davis, D. H., Singh, U., Van de Winkel, J. G. J., and Anderson, C. L. (1991).Distribution of IgG Fc receptor antigens in placenta and on endothelial cells in man: An immunohistochemical study. Am. J . Pathol. 138, 175-181. Selvaraj, P., Rosse, W. F., Silber, R., and Springer, T. A. (1988).The major Fc receptor in blood has a phosphatidylinositol anchor and is deficient in paroxysmal nocturnal haenioglobinuria. Nature (London)333,565467. Selvaraj, P., Carpen, O., Hibbs, M. L., and Springer, T. A. (1989). Natural killer cell and granulocyte Fcy receptor Ill (CD16) differ in membrane anchor and signal transduction. J . Immunol. 143,3283-3288. Shaw, D. R., Khazaeli, M. B., Sun, L. K., Ghrayeb, J., Daddona, P. E., McKinney, S., and LoBuglio, A. F. (1987). Characterisation of a niouse/human chimeric monoclonal antibody (17-1A)to a colon cancer tumor-associated antigen.J . Immunol. 138,4534-4538. Shen, L., and Fanger, M. W. (1981). Secretory IgA antibodies synergize with IgG in promoting ADCC by human polymorphonuclear cells, monocytes and lymphocytes. Cell. Immunol. 59,75-81. Shimizu, A., Tepler, I., Benfey, P. N., Berenstein, E. H., Siraganian, R. P., and Leder, P. (1988). Human and rat mast cell high-affinity immunoglobulin E receptors: Characterization of putative a-chain gene products. Proc. N at l . Acad. Sci. U.S.A.85, 19071911. Shopes, B., Weetall, M., Holowka, D., and Baird, B. (1990). Recombinant human IgG1murine IgE chimeric Ig. Construction, expression and binding to human Fcy receptors. J . Immunol. 145,3842-3848. Silverton, E. W., Navia, M. A., and Davies, D. R. (1977).Three-dimensional structure of an intact human immunoglobulin. Proc. Natl. Acad. Sci. U.S.A. 74, S140-5144. Sjoberg, 0. (1980). Presence of receptors for IgD on human T and non-T lymphocytes. Scand.J. Immunol. 11,377-382. Sjobring, U., Falkenberg, C., Nielsen, E., Akerstrom, B., and Bjorck, L. (1988).Isolation and characterization of a 14-kDa albumin-binding fragment of streptococcal protein G. J . Immunol. 140, 1595-1599. Sledge, C. R., and Bing, D. H. (1973). Binding properties ofhuman complement protein C l q . J . Biol. Chem. 248,2818-2823. Snow, M . E., and Amzel, L. M (1988).A molecular mechanics study ofthe conformation of the interchain disulphide of human IgG4. Mol. Immunol. 25, 1019-1024. Spiegelberg, H. L., and Fishkin, B. G. (1972).The catabolism ofhuman yG immunoglobulin of different heavy chain subclasses. 111. The catabolism of heavy chain disease proteins and of Fc fragments of myeloma proteins. Clin. E x p . lmmunol. 10,599-607. Spiegelberg, H. L., and Weigle, W. 0. (1965a). The catabolism of homologous and heterologous 7 s gamma globulin fragments. J . E x p . Med. 121,323-338. Spiegelberg, H. L., and Weigle, W. 0. (1965b). Studies on the catabolism of gamma G subunits and chains. J . Immunol. 95, 1034-1040. Stengelin, S., Stamenkovic, I., and Seed, B. (1988). Isolation of cDNAs for two distinct human Fc receptors by ligand affinity cloning. EMBO J . 7 , 1053-1059.

82


Stewart, W. W., and Kerr, M. A. (1990). The specificity of the human neutrophil IgA receptor (Fc,R) determined by measurement of chemiluminescence induced by serum or secretory IgAl or IgA2. Zmmunology 71,328-334. Stone, G. C., Sjobring, U., Bjorck, L., Sjoquist, J., Barber, C. V., and Nardella, F. A. (1989). The Fc binding site for streptococcal protein G is in the Cy2-Cy3 interface region of IgG and is related to the sites that bind staphylococcal protein A and human rheumatoid factors. J . Zmmunol. 143,565-570. Stuart, S. G., Trounstine, M. L., Vaux, D. J. T., Koch, T., Martens, C. L., Mellman, I., and Moore, K. W. (1987). Isolation and expression of cDNA clones encoding a human receptor for IgG (FcyRII).]. E x p . Med. 166,1668-1684. Stuart, S. G., Simister, N . E., Clarkson, S. B., Kacinski, B. M., Shapiro, M., and Mellman, I. (1989). Human IgG Fc receptor (hFcRI1; CD32) exists as multiple isoforms in macrophages, lymphocytes, and IgG-transporting placental epithelium. EMBO J. 8, 3657-3666. Sutton, B. J., and Phillips, D. C. (1983). The three dimensional structure ofthe carbohydrate within the Fc fragment of rabbit immunoglobulin G. Biochem. SOC.Trans. 11, 130- 132. Takata, Y.,Tamura, N., and Fugita, T. (1984). Interaction of C3 with antigen-antibody complexes in the process of solubilization of immune precipitates. 1. Zmmunol. 132, 2531. Tamma, S. M. L., and Coico, R. F. (1991). Release of human IgD-R+T cells. FASEB j . 5, A1718. Tan, L. K., Shopes, R. J., Oi, V. T., and Morrison, S. L. (1990). Influence of the hinge region on complement activation, C l q binding and segmental flexibility in chimeric human immunoglobulins. Proc. Natl. Acad. Sci. U.S.A.87, 162-166. Tao, M.-H., and Morrison, S. L. (1989). Studies of aglycosylated chimeric mouse-human IgG. Role of carbohydrate in the structure and effector functions mediated by the human IgG constant regi0n.J. Zmmunol. 143,2595-2601. Tao, M.-H., Canfield, S. M., and Morrison, S. L. (1991).The differential ability ofhuman IgCl and IgG4 to activate complement is determined by the COOH-terminal sequence of the CH2 domain. j . E x p . Med. 173,1025-1028. Tax,W. J. M., and Van de Winkel, J. G. J. (1990). Human Fcy receptor 11: A standby receptor activated by proteolysis? Immunol. Today 11,308-310. Tax, W. J . M., Willems, H. W., Reekers, P. P. M., Capel, P. J. A., and Koene, R. A. P. (1983). Polymorphism in mitogenic effect of IgCl monoclonal antibodies against T3 antigen on human T cells. Nature (London) 304,445-447. Tetteroo, P. A. T., Van der Schoot, C. E., Visser, F. J., Bos, M. J. E., and von dem Borne, A. E. G . K. (1988).Three different types of Fcy receptors on human leukocytes defined by workshop antibodies: FcyRl,, of neutrophils, F C ~ Rof~K/NK , ~ lymphocytes, and FcyRII. I n “Leucocyte Typing 111” (A. J. McMichael, ed.), pp. 702-706. Oxford Univ. Press, London and New York. Traunecker, A., Schneider, J., Kiefer, H., and Karjalainen, K. (1989). Highly efficient neutralisation of HIV with recombinant CD4-immunoglobulin molecules. Nature (London)339,68-70. Uchibayashi, N., Kikutani, H., Barsumian, E. L., Hauptmann, R. Schneider, F.-J., Schwendenwein, R., Sommergruber, W., Spevak, W., Maurer-Fogy, I., Suemura, M., and Kishimoto, T. (1989). Recombinant soluble FCEreceptor (FcsRIIICD23) has IgE binding activity but no B cell growth promoting activity. j . Zmmunol. 142,3901-3908. Unkeless, J.C., Scigliano, E., and Freedman, V. H. (1988). Structure and function of human and murine receptors for IgG. Annu. Reu. Zmmunol. 6,251-281. Valim, Y.M. L., and Lachmann, P. J . (1991). The effect of antibody isotype and antigenic


83

epitope density on the complement-fixing activity of immune complexes: A systematic study using chimeric anti-NIP antibodies with human Fc regions. Clin. E x p . Immunol. 84, 1-8. Van d e Winkel, J. G. J., and Anderson, C. L. (1991). Biology ofhuman immunoglobulin G Fc receptors. J . Leuk. Biol. 49, 511-524. Van d e Winkel, J. G. J., Tax, W. J. M., Van Bruggen, M. C. J., Van Roozendaal, C. E. P., Willems, H. W., Vlug, A., Capel, P. J. A., and Koene, R. A. P. (1987). Characterization of two Fc receptors for mouse immunoglobulins on human nionocytes and cell lines. Scand. J . Immunol. 26,663. Van d e Winkel, J. G. J., Boonen, G. J. J. C., Janssen, P. L. W., Vlug, A., Hogg, N., and Tax, W. J. M. (1989a). Activity of two types of Fc receptors, FcyRI and FcyRII, in human monocyte toxicity to sensitized erythrocytes. Scond. J. Zmmunol. 29,23. Van d e Winkel, J. G. J., Van Ommen, R., Huizinga, T. W. J., De Raad, M. A. H. V. M., Tuijnman, W. B., Groenen, P. J . T. A., Capel, P. J. A,, Koene, R. A. P., and Tax, W. J. M. (1989b). Proteolysis induces increased binding affinity ofthe monocyte type I1 FcR for human I&. J. Inununol. 143,571-578. Varin-Blank, N., and Metzger, H. (1990). Surface expression of mutated subunits of the high affinity mast cell receptor for IgE. J. Biol. Cheni. 265, 15685-15694. Vercelli, D., Helm, B., Marsh, P., Padlan, E., Geha, R., and Could, H. (1989).The B-cell binding site on human immunoglobulin E. Nature (London)338,649-651. Vivier, E., Morin, P., O’Brien,C., Druker, B., Schlossman, S. F., and Anderson, P. (1991). Tyrosine phosphorylation of the FcyRIII(CD16):( complex in human killer cells. J . Immunol. 146,206-210. Waldmann, T. A., and Strober, W. (1969).Metabolism ofimmunoglobulins. Prog. Allergy 13,l-110. Walker, B. A. M., Hagenlocker, B. E., Stubbs, E. B., Jr., Sandborg, R. R., Agranoff, B. W., and Ward, P. A. (1991). Signal transduction events and FcyR engagement in human neutrophils stimulated with immune complexes. J . Zmmunol. 146,735-741. Walker, M. R., Kumpel, B. M., Thompson, K., Woof, J. M., Burton, D. R., and Jefferis, R. (1988). Immunogenic and antigenic epitopes of immunoglobulins. Binding of human monoclonal anti-D antibodies to FcRI on the monocyte-like U937cell line. Vox Sung. 55,222-228. Walker, M. R., Lund, J., Thompson, K. M., and Jefferis, R. (1989a). Aglycosylation of human IgCl and IgG3 monoclonal antibodies can eliminate recognition by human cells expressing FcyRI and/or FcyRII receptors. Biochem. 1.259,1356-1372. Walker, M. R., Woof, J. M., Briiggeniann, M., Jefferis, R., and Burton, D. R. (198917). Interaction of human IgG chimeric antibodies with the human FcRI and FcRII receptors: Requirements for antibody-mediated host cell-target cell interaction. Mol. Immunol. 26,403-41 1. Warmerdam, P. A. M., Van d e Winkel, J. G. J . , Gosselin, E. J., and Capel, P. J. A. (1990). Molecular basis for a polymorphism of human Fcy receptor I1 (CD32). J . E x p . Med. 172,19-25. Warmerdam, P. A. M., Van de Winkel, J. G. J., Vlug, A,, Westerdaal, N. A. C., and Capel, P. J. A. (1991). A single amino acid in the second Ig-like domain of the human Fcy receptor I1 plays a critical role in human IgC2 binding. J. Zmmunol. 147, 13381343. Wawrzyncak, E. J., Denham, S., Parnell, G. D., Cumber, A. J., Jones, P. T., and Winter, G . (1992a). Recombinant mouse monoclonal antibodies with single amino acid substitutions affecting C l q and high affinity Fc receptor binding have identical serum half-lives in the BALB/c mouse. Mol. Immunol. 29,221-227. Wawrzyncak, E. J., Cumber, A. J., Parnell, G. D., Jones, P. T., and Winter, G. (1992%).

84


Blood clearance in the rat of a recombinant mouse monoclonal antibody lacking the N-linked oligosaccharide side chains of the CH2 domains. Mot. fmmunol. 29,213-220. Weetall, M., Shopes, B., Holowka, D., and Baird, B. (1990). Mapping the site of interaction between murine IgE and its high affinity receptor with chimeric 1g.j.lmmunol. 145,3849-3854. Weiss, V., Fauser, C., and Engel, J. (1986). Functional model of subcomponent C1 of human comp0nent.J. Mol. Biol. 189,573-581. Wilson, I. A., Skehol, J. J, and Wiley, D. C. (1981). Structure of the haemagglutinin membrane glycoprotein of influenza virus at 3A resolution. Nature (London) 298, 366-373. Wochner, R. D., Strober, W., and Waldmann, T. A. (1967). The role of the kidney in the catabolism of Bence- Jones proteins and immunoglobulin fragments.]. E x p . Med. 126, 207-221. Woof, J. M., Nik Jaafar, M. I., Jefferis, R., and Burton, D. R. (1984). The monocyte binding domain(s) on human immunoglobulin G . Mol. lmmunol. 21,523-527. Woof, J. M., Partridge, L. J., Jefferis, R., and Burton, D. R. (1986). Localisation of the monocyte-binding region on human immunoglobulin G. Mol. lmmunol. 23,319-330. Woof, J. M. et al. (1992). In preparation. Wright, J. F., Shulman, M. J.. Isenman, D. E., and Painter, R. H. (1988). C l q binding by murine IgM. The effect of a pro-to-ser exchange at residue 436 of the p chain. j . B i d . Chem. 263,11221-11226. Wright, J. K., Tschopp, J., Jaton, J.-C., and Engel, J. (1980). Dimeric, trimeric and terameric complexes of immunoglobulin G fix complement. Biochcm.]. 187,775-780. Yarnall, M., and Boyle, M. D. P. (1986a). Identification of a unique receptor on group A streptococcus for the Fc region of human IgG3. J. lmmunol. 136,2670-2673. Yarnall, M., and Boyle, M. D. P. (1986b). Influence of dipeptides on the interaction of immunoglobulins with bacterial Fc receptors. Biochem. Biophys. Res. Commun. 135, 1105-1 111. Yarnall, M., and Widders, P. R. (1990). Type V Fc receptor from Streptococcus zooepidemicus. In “Bacterial Immunoglobulin Binding Proteins” (M. D. P. Boyle, ed.), Vol. 1, Chapter 13, pp. 155-164. Academic Press, San Diego. Yarnall, M., Reis, K. J., Ayoub, E. M., and Boyle, M. D. P. (1984). An immunoblotting technique for the detection of bound and secreted bacterial Fc receptors.]. Microbiol. Methods 3,83-93. Yasmeen, D., Ellerson, J. R., Dorrington, K. J., and Painter, R. H. (1976). The structure and function of immunoglobulin domains. IV. The distribution of some effector functions among the C,2 and C,3 homology regions of human IgG. ]. lmmunol. 116, 518-526. Yeaman, G . R., and Kerr, M. A. (1987). Opsonization of yeast by human serum IgA anti-mannan antibodies and phagocytosis by human polymorphonuclear leucocytes. Clin. E x p . lmmunol. 68,200-208. Yokota, A., Kikutani, H., Tanaka, T., Sato, R., Barsumian, E. L., Suemura, M., and Kishimoto, T. (1988). Two species of human Fce receptor I1 (FceRII/CD23): Tissuespecific and IL-4 specific regulation of gene expression. Cell (Cambridge,Mass.) 55, 611-618. Zheng, Y., Shopes, B., Holowka, D., and Baird, B. (1991). Conformations of IgE bound to its receptor FcaRI and in solution. Biochemistry 30,9125-9132. Zuckier, L. S., Rodriguez, L. D., and Scharff, M. D. (1989). Immunological and pharmacological concepts of monoclonal antibodies. Semin. Nucl. Med. 19, 166-186. This article was accepted for publication on 6 January 1992.

ADVANCES IN IMMUNOLOGY. VOL 51

The Development of Functionally Responsive T Cells ELLEN V. ROTHENBERG Division of Biology, California Institute of Technology,Pasadena, Colifornia 91 125

1. Properties of Mature T Cells: The Destination of the Process

T (thymus-derived) lymphocytes play central roles in immune responses and in the regulation of cell growth for several hematopoietic lineages. The importance of these cells is derived from their unique integration of specific antigen recognition with a n impressively versatile, highly regulated array of possible effector functions. The developmental processes that endow T cells with their functional responsiveness and couple this to an appropriate repertoire of recognition specificities have been of great interest for many years to immunologists and to cell and molecular biologists alike. In the past decade, these cells have become one of the best understood cell types in vertebrates. The enormous progress toward understanding immunological recognition on both the cell and the population level has dominated the field of T cell development in recent years. Antigen-specific recognition and antigen-specific tolerance are, after all, the classic features of immunocytes that both T and B cells share. There is an important distinction between the roles of B and T cells in the immune system, however, that goes beyond their different target-structure specificities. B cell responses consist of secreting immunoglobulin (Ig) molecules that are simply variants of the recognition structures through which the cells are triggered. Thus, both specificity and effector function can be understood through the regulation of Ig gene expression. By contrast, T cell effector function is based on the expression of genes that encode growth factors, differentiation factors, or directly cytolytic structures, a different set of genes from those that encode the molecular components of the T cell receptor (TCR) for antigen. Thus, a wellrounded view of T cell development encompasses the regulation of T cell signaling and response genes, as well as the assembly and selection of the TCR. At one level, throughout this review, we will use a checklist o f mature T cell characteristics as a way to stage the progress of pre-T cells toward maturity. For example, the transition from a TCR- to a TCR+ state may be seen as the outcome of a developmental process. At 85

Copyright 0 1992 b v A d e 1 1 1 1 Prras. ~ Inc All right\ of reploduction 111 dnv fiirni reserved

86

ELLEN V. ROTHENBERG

a second level, however, the very nature of these T cell markers makes them candidates for vital components of the inductive mechanisms that drive development. This is because most characteristic T cell gene products are either components of a signal transduction chain, cell surface triggering receptors, or polypeptide hormones that are potent regulators of cell growth and differentiation. Any T cell-specific signaling receptor that a pre-T cell already possesses must at least be considered as a possible conduit for the signal that drives the cell to the next state. This review therefore will trace T cell differentiation with explicit attention to the dual significance of the phenotypic transitions that are discussed: (1)what they tell us has occurred and (2)what they suggest as possible mechanisms for what will occur next. To orient the discussion, I will begin with an overview of the properties of mature T cells, including both the molecules they use for recognition and the molecules they use in responses. Appreciating the diversity of characteristics that are loosely considered “T cell specific” will then allow us to tackle critically the issue of how they are coordinated in development.

A. RECOGNITION STRUCTURES

The T cell receptors for antigen are structures designed to recognize cell-associated fragments of foreign antigens. The cell “presenting” the antigen is usually a macrophage, dendritic cell, or B cell that has processed the nominal antigen by proteolysis to generate fragments of about 10 amino acids (Parham, 1991; Rothbard and Gefter, 1991). These fragments are then associated noncovalently with major histocompatibility complex (MHC) glycoproteins of either class I or class I1 type and cycled to the surface of the antigen-presenting cell (APC) as a complex. It is this complex, of foreign peptide plus autologous MHC glycoprotein, that is the actual ligand for the TCR. The mechanics of antigen presentation have been reviewed in detail recently (Brodsky and Guagliardi, 1991), and we need only touch on several points here. The first point is that peptides generated from different antigens tend to be associated preferentially with either class I or class I1 MHC molecules, depending on their origin. Endogenously synthesized foreign molecules, such as viral nucleoproteins in infected cells, are predominantly presented with class I MHC. Proteins of extracellular origin, including many experimental immunogens, are scavenged and presented mainly in association with class I1 MHC. As we shall see, the T cells that recognize antigen with class I accordingly include a high frequency of killers that destroy their APC, whereas the T cells that recognize antigen with class I1 tend to be regulatory cells, or “helpers.”

FUNCTIONALLY RESPONSIVE T CELL DEVELOPMENT

87

The second point is that all T cell “antigen” recognition is really cell-cell recognition, because free, native antigenic proteins are generally invisible to T cells. Mature T cells can only recognize foreign peptide-MHC complexes on cell surfaces, and thus their avidity for antigen” is strongly influenced by whatever other receptor-ligand interactions can affect the duration of contact between a T cell and an APC. Because T cells are migratory cells, they have no stable cell-cell contacts, and the likelihood that they will be triggered by an antigen on an APC is greatly enhanced by any “nonspecific” adhesive interactions that encourage them to linger in contact with an APC long enough to sample the peptides it displays. Thus, molecules other than the T cell antigen receptor proper may b e expected to influence T cell function. T cells recognize antigens by means of T cell receptor glycoprotein complexes, in which the antigen-binding chains are encoded by rearranging gene families. Several reviews have provided excellent discussions of these gene families (Strominger, 1989; Raulet, 1989)and of the nature of antigen recognition by T cells (Fowlkes and Pardoll, 1989; Ashwell and Klausner, 1990; Davis and Bjorkman, 1988),and the major points will be summarized here only briefly. The TCR complex consists of a heterodimer of antigen-binding chains (either a and j3 or y and 6, depending on the cell) in combination with five or six polypeptide chains involved in membrane insertion and signal transduction. The latter group of proteins is referred to collectively as CD3 chains. All of the polypeptides in the complex are integral membrane proteins of the immunoglobulin gene superfamily. The C D 3 chains are invariant in sequence and assembled in the same combination by all T cells. The antigen-binding chains, on the other hand, are different in primary sequence for each T cell clone. These differences can be found at two levels. First, there are four different rearranging gene families that can encode these TCR chains (a,p, y , and a), but each T cell uses only one of two possible kinds of heterodimers: an crp heterodimer or a y6 heterodimer. Thus, all T cells can be classified either as “ap cells” or as “ y 6 cells.” We will later discuss the possible mechanisms responsible for enforcing these exclusive pairings. Second, as for immunoglobulin genes, the sequences encoding the N-terminal domains of each chain are assembled by somatic rearrangement in T cell precursors. As reviewed elsewhere (Davis and Bjorkman, 1988) and discussed further below, the random joining of different V, J, and in some cases D gene segments creates combinatorial diversity. The rearrangement process is limited to stages before the T cells become mature, and is arrested, in general, when the cell is able to express a full TCR/CD3 complex. The general result is that “

88

ELLEN V. ROTHENBERG

each virgin T cell is committed to a single recognition specificity which is different from that of essentially all other T cells. The rate-limiting events for acquisition of antigen-binding specificity in T cell development are the rearrangements of the appropriate combination of TCR genes, either a and /3 or y and 6 . As for immunoglobulin genes, the rearrangement process is imprecise with respect to the borders of the gene segments. Nucleotides are lost at the junctions by exonucleolytic digestion of the ends to be joined, and commonly random nucleotides are added to these ends by the action of terminal deoxynucleotidyl transferase (Desiderio e t al., 1984; Landau et al., 1987), before they are religated. The mutagenized junctions encode complementarity-determining region 3 (CDRS) in the TCR polypeptide, a region of particularly high diversity that has been suggested to be the key sequence for interaction with the target peptide proper (Davis and Bjorkman, 1988). However, another result is that many rearrangement events yield genes with insertions or deletions that disrupt the reading frame, so that no protein can be expressed. Rearrangement of the TCR loci on both sets of chromosomes, and occasionally successive rearrangements at the same locus, may be required before a particular developing T cell acquires a pair of compatible rearranged TCR genes. Once an in-frame rearrangement is produced and a full-length TCR polypeptide is synthesized, rearrangement generally stops both at that locus and at its homologue on the other chromosome. Rearrangements at nonhomologous loci, however, may continue. This phenomenon, which implies the existence of some locus-specific feedback mechanism, is termed “allelic exclusion.” VDJ rearrangement of all the TCR genes, and all Ig genes as well, appears to be carried out by a common set of recombinase proteins. Thus, pre-B cell lines immortalized by Abelson virus are fully capable of rearranging TCR genes, as well as Ig genes, if the TCR genes are introduced by transfection of naked DNA (Yancopoulos et al., 1986). One component of the rearrangement process is the product of the gene that is mutated in C.B-17 scidlscid (severe combined immunodeficiency; SCID) mice (Bosma and Carroll, 1991).This product is not yet biochemically characterized, but it appears to be involved in several DNA repair processes besides VDJ rearrangement (Fulop and Phillips, 1990).In SCID mice, both TCR and Ig rearrangements fail, apparently due to inefficient ligation. Two other products required for TCR and Ig gene rearrangement are RAG-1 and RAG-2 (Schatz et al., 1989; Oettinger et al., 1990), which are encoded by tightly linked genes at a locus distinct from the scid locus. There is no proof yet that either RAG-1 or RAG-2 is an enzymatically active recombinase, but together


89

they are sufficient to endow transfected fibroblasts with the competence to carry out VDJ joining. These genes are expressed together in immature mammalian B and T cells, but not in mature cells of either lineage. This suggests that they participate selectively in VDJ-type recombination and not in the homologous recombination process used in Ig class switching or in general DNA repair. However, RAG-1 and RAG-2 can be expressed independently in sites where they may serve other functions. RAG-2 alone seems to be utilized during the Ig gene conversion that takes place in the chicken bursa of Fabricius (Carlson et al., 1991), and RAG-1 alone seems to be expressed, interestingly enough, in the mammalian brain (Chun et al., 1991). In the thymus, where T cells are generated, both transcripts are present at extremely high levels throughout the region where differentiation takes place, i.e., the thymic cortex. The thymic medulla, where T cells with completed TCR rearrangement await export, is devoid of RAG-1 and RAG-2 expression (Boehm et al., 1991; Turka et al., 1991a). The pattern of expression of these two transcripts thus defines the unique developmental zone in which VDJ rearrangements of TCR and Ig loci are permitted. The use of common enzymatic machinery to carry out VDJ rearrangement at a variety of loci raises the question of how the process is confined to the appropriate loci for the T or B cell lineage. It is believed that gene segments are selectively rendered accessible to the reconibinase by developmentally regulated transcription while they are still in their germ-line configuration (Blackwell et al., 1986).Thus, it is likely that lineage-specific and/or stage-specific promoters, enhancers, and/or silencers flank the various TCR gene segments and maintain control of the rearrangement process. Some regulatory elements have been described that might turn out to serve such a function, as will be described in Section II1,B. TCRa and TCRP chains cannot be displayed as a cell surface receptor without coexpression of the various CD3 chains, y , 6, E , 5, and 7 (Ashwell and Klausner, 1990; Clevers et al., 1988).The relationships between the TCR and CD3 chains are both structurally and functionally intimate. First, TCR chains proper possess unusual basic amino acid residues interrupting the hydrophobic properties of their membrane-spanning regions. These charged residues appear to participate in salt bridges, within the plane of the membrane, to corresponding acidic residues in the transmembrane domains of the CD3y, CD36, and/or C D ~ chains. E These three chains are encoded by linked genes on mouse chromosome 9 and reach the cell surface only when TCRa and TCRP or TCRy and TCRS chains are coexpressed. The last pair of

90

ELLEN V. ROTHENBERG

components, either a CD355 homodimer or a 57 heterodimer, assemble with this pentamer and appear to be rate limiting for transport to the cell surface. The CD35 and CD3q chains are alternatively spliced products of a single gene on murine chromosome 1 (Jin et al., 1990; Clayton et al., 1991).The 5 polypeptide is expressed in -20 to 50-fold excess over q, and the two polypeptides are assembled into disulfidelinked dimers at the cell surface in a ratio of 10 : 1 55 : 57.Though the CD35 and CD3q chains are normally TCR associated in T cells, they appear to be the only members of the TCR/CD3 complex that can also be expressed independently of the others. In addition to this coassembly role, CD3 components are critical for the signaling function of the TCR complex. The TCR chains a,/3, y, and 6 have very short cytoplasmic regions with no sign of a signaling domain. The CD3y and C D ~ chains E become phosphorylated on serine residues duringT cell activation, whereas the 5 chain is phosphorylated on tyrosine residues. Although antibodies directed against the CD3.9 component can trigger strong polyclonal T cell responses, C D ~ E is not likely to be required for directly transducing signals after TCR cross-linking. Deletion of essentially all of the CD3.9 cytoplasmic domain does not impair its ability to assemble with the TCR/CD3 complex and participate in signaling (Transy et al., 1989). This does not rule out a role of some kind (see Burgess et al., 1991), but suggests that it is dispensable. By contrast, the 55 and 5q subunits do appear to have a direct function in signaling. In the absence of 66 or 5q expression, T cells cannot be triggered through the TCR (Sussman et al., 1988; Bauer et al., 1991). Deletion of the CD35 cytoplasmic domain does impair signaling, although it does not eliminate it (Frank et al., 1990). Conversely, fusion proteins that combine the cytoplasmic domains of CD35 with unrelated extracellular domains can serve efficiently as triggering receptors to activate a typical T cell response even in a model T lineage tumor line, without involving the rest of the TCR (Irving and Weiss, 1991). Expression of CD35 and CD3q is thus not only implicated in efficient cell surface display of TCR recognition structures, but also fundamentally in coupling T cell recognition structures to powerful signal transduction pathways. The various components of the TCR complex are differentially related to triggering structures used in other hematopoietic cells. Expression of the TCR is essentially the definition of a T cell, but it is noteworthy that the CD3y and CD36 components also appear to be restricted to T lineage cells (Biron et al., 1987; Biassoni et al., 1988; Anderson e t at., 1989).Note, however, that the only chains in the TCR complex that are directly implicated in signal transduction are not

-


91

strictly T cell specific. The CD34' chain is also expressed by natural killers (NK) cells, as a triggering subunit associated with the IgG-Fc, receptors (CDlS) on these quasi-lymphoid hematopoietic cells (Lanier et al., 1989; Anderson et al., 1989). Structural relatives of CD35/77 also appear to mediate triggering in yet another cell type, namely mast cells. The major triggering receptor of mast cells, a high-affinity receptor for the Fc portion of immunoglobulin E (IgE) chains (Fc,RI), includes a y subunit that is structurally homologous to CD34'/77. This subunit is encoded by a gene linked to the 6/77 locus, within only 6000 kb (Kuster et al., 1990; Orloff et al., 1990; Clayton et al., 1991), and might conceivably be related to it by gene duplication. Thus it is likely that some of the triggering events activated by antigen in T cells have close counterparts in other hematopoietic cells. B. RECOGNITION/AUXILIARY CORECEPTORS For many T cells, the TCR/CD3 complex is necessary but not sufficient to mediate recognition of antigen/MHC complexes. Another critical set of molecules is required, namely the CD4 or CD8 glycoproteins. These two types of glycoproteins, like TCR antigen-binding chains and CD3 chains alike, are members of the immunoglobulin gene superfamily. Their structure and function have been reviewed recently (Parnes, 1989;Janeway, 1989; Bierer et al., 1989; Rudd et al., 1989).Three of their features are important here: (1)the specific binding of MHC class I1 and class I molecules by CD4 and CD8, respectively, (2) their specific association with the T cell-specific tyrosine kinase, Lck, and (3) their mutually exclusive pattern of expression correlated with T cell sublineage. T cell receptor structures generically can bind either class I or class I1 MHC structures (with their associated peptides), with their specificity depending not on the identities of individual gene segments used, but on the particular combinations of V, D, J, and junctional sequences that become fixed in their gene rearrangements. In some cases, a TCR complex may be capable of dual interactions with both class I- and class II-associated ligands (Pircher et al., 1989; Fowlkes et at., 1988; MacDonald et al., 1 9 8 8 ~ )As . a rule, however, mature T cells are not dual-reactive cells, and their use of class I1 versus class I as a triggering ligand is strongly correlated with their own expression of either CD4 or CD8 as a coreceptor (Swain, 1980). The overwhelming majority of mature T cells express either CD4 or CD8, but not both, thus defining the two main subsets of T cells as CD4+ (recognizing class I1 MHC) and CD8+ cells (recognizing class I MHC). CD8 has been shown to be a specific receptor for the nonpolymorphic a3 region of

92

ELLEN V. ROTHENBERG

class I MHC (Norment et al., 1988; Salter et al., 1990), whereas CD4 binds similarly to class I1 MHC (Doyle and Strominger, 1987; Fleury et al., 1991). There is now a considerable body of data to indicate that coordinated binding or cross-linking of CD4 or CD8 together with the TCR complex strongly enhances T cell activation, particularly in primary T cell responses (see reviews cited above). A structural explanation for this effect has come from the demonstration that the cross-linking of the TCR with CD4 or CD8 can be responsible for introducing a potent tyrosine kinase activity into the vicinity of the TCR complex (Rudd et al., 1988; Veillette et al., 1988, 1989a). The cytoplasmic domains of CD4 and CD8 lack kinase activity in themselves, but both include a specific sequence that can be bound noncovalently by a sequence in the N-terminal region of the tyrosine kinase pp56Ick(Lck) and both appear to be associated with Lck in v i m . In effect, this gives CD4 and CD8 the potential to act as receptor tyrosine kinases. Though Lck is not the only kinase presently implicated in T cell triggering, it is very likely to participate in the effects of TcR x CD4 or TcR x CD8 cross-linking, which may affect not only the intensity of mature T cell triggering but also the determination of immature T cell fate. We will consider the pivotal developmental roles of these molecules, and their own regulation during T cell differentiation, in later sections. THROUGH TCR/CD3 AND COFLECEPTORS C. SIGNALING

The TCR/CD3 complex cooperates with the CD4 or CD8 coreceptor in generating a characteristic set of intracellular signals in response to antigen recognition. In mature T cells, over a period of hours or days, these signals lead to changes in gene expression. Over a period of minutes, the main second messengers for T cell activation are a sharp increase in cytoplasmic free Ca2+ concentration ( [Ca2+li), from -1OOnM to over 500 nM, and the activation of the Ca2+- and phospholipid-dependent kinase, protein kinase C (PK-C). Both elevated [Ca2'Ii and activated PK-C are necessary for induction of response gene expression (Weiss and Imboden, 1987; Altman et al., 1990). In certain model systems, moreover, these mediators appear to be sufficient, because the combination of Ca2+ionophore plus phorbol ester (a promiscuous PK-C activator) can powerfully activate T cells without any need for TCR engagement. Both the Ca2+ and PK-C second messengers are mobilized as effects of the inositol phospholipid breakdown pathway, in a sequence of reactions that T cells share with many other cell types. Following TCR engagement, phospholipase C (PLC) hydrolyzes phosphatidylinositol-


93

4,s-bisphosphate (PIP2) to yield inositol-1,4,5-trisphosphate (IP3) and diacylglycerol (DAG). IP3 opens a voltage-independent Ca2+channel (Kuno and Gardner, 19871, whereas DAG acts as a necessary cofactor for PK-C activation. Though the stoichiometric relationships between IP3, DAG, Ca2+ flux intensity, and duration of PK-C activation remain somewhat uncertain (Ledbetter et al., 1985, 1986a; Mercep et al., 1988; O’Rourke et al., 1990), this general sequence of events is well established for T cells and the evidence for it has been reviewed in detail (Weiss and Imboden, 1987; Weiss, 1989; Altman et al., 1990). The use by T cells of an essentially ubiquitous response pathway raises two questions, nevertheless. One is how Ca2+ and PK-C signals-which are common to many other cell types-can be used by T cells only to activate T cell-specific response genes. The other question is how the TCR/CD3 complex is coupled to the PIP2 breakdown pathway and how the specificity of T cell activation is maintained. Many cell surface receptors are coupled to PLC by means of G proteins (Neer and Clapham, 1988).G proteins constitute a large and diverse family of signaling mediators, many of which are expressed in a tissue-specific way, and until recently they seemed highly attractive candidates for the link between PLC and the TCR complex. Initial results with G protein activators and inhibitors appeared to favor such a role for G proteins in T cell activation (O’Shea et al., 1987; Imboden et al., 1986). To begin testing this model, Weiss and co-workers transfected the Jurkat leukemic T cell line with a type 1muscarinic acetylcholine receptor, which is known to use a G protein to activate PLC (Peralta et al., 1988). Indeed, in these cells a muscarinic receptor agonist could be used to trigger interleukin-2 (IL-2) production, a typical T cell response (Desai et al., 1990).This showed that T cells possess G proteins capable of activating PLC, and that PLC activation is sufficient to induce downstream response genes, at least in this cellular context. Recent experiments have shown, however, that the coupling of the TCR to PLC normally depends upon a different mediator, namely a tyrosine kinase (Klausner and Samelson, 1991). It has been noted for some time that phosphorylation on tyrosine residues of a number of proteins, including CD33, accompanied T cell activation along with the serine/threonine phosphorylation events catalyzed by protein kinase C. Elegant experiments conducted by several groups have now shown that the tyrosine kinase activity precedes the other responses and is necessary for them (June et al., 1990; Mustelin et al., 1990; Stanley et al., 1990;Trevillyan et al., 1990).The tyrosine kinase inhibitors, herbimycin and genistein, inhibit even the early Ca2+ and PK-C

94

ELLEN V. ROTHENBERG

responses to TCR/CD3 ligation. Conversely, inhibition of PK-C does not block the very rapid increases in tyrosine phosphorylation, even though it does block gene expression changes downstream. Detailed analysis of the Jurkat T leukemic cells transfected with muscarinic receptors showed that the TCR in fact uses a different mechanism to activate PLC than does the muscarinic receptor, as follows. In these cells, which possess two PLC-activating receptor systems, tyrosine kinase inhibitors that have no effect on activation via the G proteincoupled receptor completely block activation of the same cells via the TCR/CD3 complex (Desai et al., 1990).The demonstration of a critical role of a tyrosine kinase in the earliest stage of TCR signaling has naturally focused intense interest on the tyrosine kinases and tyrosine kinase substrates that are physically associated with the TCR. The TCR/CD3 complex is not intrinsically a tyrosine kinase, but, as we noted above, at least the CD35 chain is a stimulation-dependent tyrosine kinase substrate. One tyrosine kinase that probably phosphorylates CD35 during activation is Lck (Barber et al., 1989), the lymphoid-specific tyrosine kinase that is strongly, even if noncovalently, associated with the CD4 and CD8 coreceptors. Though Lck is not normally bound to the TCRICD3 complex, the TCR/CD3 complex colocalizes with the Lck-linked coreceptor during antigen recognition. This brings the Lck kinase activity into contact with a new set of substrates, apparently including CD35. If CD35 phosphorylation is in some way rate limiting for activation, this may be the basis for the considerable enhancement of T cell activation when the TCR and coreceptors are coengaged, as compared with activation through the TCR alone. Yet another tyrosine kinase, ~ ~ 5 9 (Fyn), ' ~ " appears to be even more intimately associated with the TCR/CD3 complex (Samelson et al., 1990). Fyn, like Lck, is a member of the src-encoded family of kinases. Though it is not absolutely T cell specific in its distribution (a form of Fyn is also expressed in the brain), it is the only tyrosine kinase that is reproducibly coprecipitated with the TCRlCD3 complex by anti-TCR/ CD3 antibodies even in resting T cells (Samelson et al., 1990). Thus, Fyn may also be the kinase that is responsible for phosphorylating CD35 in response to TCR ligation. More than Lck, it seems directly implicated in controlling the magnitude of Ca2+flux that follows TCR ligation (Cooke et al., 1991). The full range of substrates that Fyn and/or Lck phosphorylate to generate an activation cascade is not clearly identified as yet. There is at least apossibility that these kinases activate PLC directly via phosphorylation of the PLC-71 subunit (Weiss et al., 1991). It is important to note, however, that the use of


95

potent kinases as coupling mediators allows TCR/CD3 plus coreceptor ligation to activate additional signaling pathways in parallel to the well-characterized one emanating from PLC. Both Fyn and Lck need to be activated to express their full kinase activity, and this is the role of another transmembrane receptor, CD45 (Thomas, 1989). Both Fyn and Lck can be phosphorylated on a conserved tyrosine residue near the C terminus, which strongly inhibits their kinase activity. Dephosphorylation (or engineered deletion of this phosphorylation site) enhances their activity. CD45, a large transmembrane protein expressed in essentially all nucleated hematopoietic cells, contains a pair of protein-tyrosine phosphatase domains in its cytoplasmic region (Charbonneau et al., 1988).When this molecule is coclustered with the TCR/coreceptor complex, it dephosphorylates and activates the receptor-associated tyrosine kinases (Mustelin et al., 1989). In principle, CD45 phosphatase may limit TCR signaling, by dephosphorylating Fyn and Lck substrates as well as Fyn and Lck themselves, but in practice its net effect appears to be stimulatory (Koretzky et al., 1990; Pingel and Thomas, 1989;Weaver et al., 1991). There may be different ligands that bind CD45 expressed at different developmental stages, for the extracellular domain of CD45 is complex, with alternative exons utilized in the coding sequence in cells of different lineages and activation states (Thomas, 1989; Bottomly et al., 1989; Pilarski et al., 1989; H.-L. Chang et al., 1991). For our purposes here, it is sufficient to note that the ability of the TCR/CD3 complex to initiate a signal transduction cascade often depends not only on its own binding of antigen, but also on additional ligand-receptor interactions that bring it into proximity with at least two other surface receptors, i.e., CD45 and CD4 or CD8. In addition to the activation “module” defined by the TCR/CD3 complex, the coreceptor, and CD45, there are additional cell surface receptors that participate in regulating T cell responses in independent and distinct ways. Among these are the Ig superfamily members Thy-1, CD2, CD5, and CD28, and the integrin LFA-1 (Springer, 1990; DeFranco, 1991). Though we cannot discuss all of these receptors and their ligands, four will be mentioned later, at the appropriate points: Thy-1 and CD5, as early and midstage T cell differentiation markers, respectively, in Section 111,A; CD2 in Section IV,D; and CD28, which may help dictate the type of response made to activation, in Section 1,D. To summarize (Fig. l ) , when the TCR/CD3 complex binds a target antigen on an APC, it appears to activate a tyrosine kinase, probably Fyn, which phosphorylates PLC and thereby activates it. This initial

96

ELLEN V. ROTHENBERG


97

event seems to be strongly enhanced if both CD45 and the coreceptor CD4 or CD8, with its associated Lck activity, are coclustered in the vicinity with the same ligand (Fig. 1A). Once PLC is activated it begins to hydrolyze PIP2 in a process that may continue as long as the TCR is engaged (Fig. 1B). The cleavage products signal the T cell to increase [Ca2+Iiand to mobilize the serine-threonine kinase PK-C. These latter changes in the cell can also persist for hours, as long as the TCR remains engaged. Probably due to IP3-mediated feedback, the [Ca2+Ii oscillates (Berridge and Irvine, 1989; Lewis and Cahalan, 1989), in a long train of discrete peaks and valleys, while the PK-C translocates to the plasma membrane and remains bound there, presumably to phosphorylate an activation-specific set of substrates. The precise nature of the next targets is unclear; however, Ca2+ plus PK-C activation of T cells rapidly results in the mobilization of key DNA-binding proteins involved in gene regulation (Fig. 1C) (Ullman et al., 1990). Of these, at least AP-1, NF-KB and/or AP-3, SRF, Oct-2, and NF-AT are directly implicated in the induction of the response genes, IL-2 and IL-2Ra (Ullman et al., 1990). Before leaving this signaling cascade, it is worth noting that even in mature T cells, TCR engagement does not inevitably proceed toward response gene induction at the slightest encounter. Binding of the FIG. 1. T cell signaling during antigen recognition. The figure presents a simplified current view of the sequence of events whereby antigen recognition leads to induction of nuclear responses. At the top of each panel, the antigen-presenting cell is shown schematically, with its MHC class I1 protein and associated peptide (curled line), its ligand for CD28 (BBI/B7), and its ligand for CD4S (Y). For simplicity, neither the CD2, CD5, nor LFA-1 molecules on the T cells nor their ligands on the APC are depicted. On the T cell surface, CD28, CD4, and the tyrosine phosphatase CD45 are labeled accordingly. The antigen-binding chains of the TCR (aand /?) are depicted in association with the five chains of the CD3 complex, of which only E , 6, and r) (molecules most likely to be implicated in signal transduction) are individually labeled. In all three panels, thick black arrows are used to denote protein kiiiase activities. A spiky halo indicates a conspicuous effect or putative activation event. Thus, @ (phosphate) indicates activation via phosphorylation. (A) CD4S dephosphorylates Lck and Fyn (release of P), allowing these kinases to phosphorylate CD36, PLC, and other substrates (?). The precise identity of the kinase responsible for each of these catalytic events is not yet clear, so both Lck and Fyn are shown as possible participants. (B) The phosphorylated PLC is now activated to cleave PIPz into IP3 and DAG. IPS induces release of Ca”+ from the endoplasmic reticulum into the cytosol, where it cooperates with DAG (dashed arrows) to activate PK-C. A major portion of the PK-C then translocates physically to the plasma membrane (open arrow). (C) Mediators activated by PK-C-catalyzed phosphorylation cooperate with mediators activated in an unknown way by CD28 engagement to induce nuclear gene activation. Additional cooperative signals (P) may be derived from Ca’+, presumably via the Ca”/calmodulin (Ca’+/CaM)-dependent kinase, and from unknown substrates phosphorylated by the Lck and Fyn kinases.

98

ELLEN V. ROTHENBERG

TCR/CD3 complex with soluble ligands promotes internalization of the ligand-receptor complex and yields an abortive response (Ledbetter et al., 1986a; Manger et a!., 1987),particularly as measured by PK-C translocation. Such abortive responses often leave the T ceIl temporarily or permanently handicapped in its ability to respond to a more appropriate signal. Another physiological damper on T cell activation is the CAMPsecond messenger working through the CAMP-dependent kinase. Inducing elevated CAMP concentrations in T cells by any of a number of methods, including exposure to the macrophage products prostaglandin El and Ez, sharply inhibits T cell signaling. Not only does high CAMPappear to uncouple the TCR/CD3 complex from PLC (O’Shea et al., 1987; Mary et d., 1987; Kim et al., 1988; Muiioz et d., 1990; Gajewski et al., 1990),but it even interferes with the induction of particular response genes in response to pharmacological agents (Novak and Rothenberg, 1990; Muiioz et d . , 1990). The integration of TCR/CD3 and coreceptor signals with signals from other pathways, such as those coupling to CAMP, means that the outcome of T cell triggering depends on the quantitative balancing of different signals (Graber et al., 1991). The balance may be at least partly dependent on the duration of T cell-ligand contact, allowing a T cell to distinguish between a rapidly dissociating receptor interaction and a sustained interaction. Indeed, it appears to take 3-6 hours of receptor-ligand interaction for a normal T cell to commit to IL-2 expression (Kumagai et al., 1987; Weiss et al., 1987a). These kinetic and combinatorial effects on mature T cell signaling provide models for mechanisms that may cause the sharp developmental changes in T cell responsiveness; these will be described in Sections IV and V. D. RESPONSE GENES

1 . T Cell Subsets The goal of T cell activation is to induce the expression of a distinct set of T cell genes involved in effector function. We will call these response genes.” T cells are not homogeneous in their responses to “

stimulation, but instead fall into several subsets. All effector subsets can proliferate following antigen stimulation and, to a greater or lesser extent, synthesize and secrete the polypeptide hormones known as lymphokines. However, they differ in exactly which lymphokines they secrete and which, if any, additional functions they perform. Detailed comparisons of these subsets have been made with respect to a wide range of properties, and many of these features have been reviewed recently (Gajewski et al., 1989; Mosmann and Coffman,


99

1989; Tschopp and Nabholz, 1990). A summary of subset characteristics is offered in Table I. The various “helper” subsets are defined by their secretion of at least one growth factor among their arsenal of lymphokines, namely, IL-2 and/or IL-4. The killer cells (cytotoxic T lymphocytes, or CTLs) are defined by their abilities to kill target cells through direct cell-cell conjugate formation, often with supplemental killing options involving toxic lymphokines such as tumor necrosis factor-a (TNF-a), TNF-P, and interferon-y (IFN-y). Some CTLs are able to make IL-2 as well, but this property is easily lost with continued stimulation (von Boehmer et al., 1984). Note that many TCRyG cells appear to fall into the CTL class (Koizumi et al., 1991; refs. cited in Allison and Havran, 1991; Ferrick et al., 1991). Because of their different products, various T cell subsets induce different kinds of immune responses (Mosmann and Coffman, 1989).T Hcells ~ help activate complement-fixing antibody responses and stimulate CTLs, particularly in antiviral responses. T Hcells, ~ on the other hand, promote anaphylactic-type responses by inducing IgE production and activating mast cells and eosinophils. The role of CTLs is much simpler, namely, to kill, but they usually depend on T H cells for their supply of growth factors. The most dramatic distinctions are between different long-term, antigen-dependent, cloned T cell lines, and these provide the basis for Table I. In these lines it is clear that the combination of properties is not due to the presence of a mixture of different cell types. Of course, such lines may be altered or artificially selected by their passage in uitro. The TH1 and T H 2 cloned-cell categories probably do not define real subsets in vivo so much as they illustrate extreme end states for different memory T cell types (Gajewski et al., 1989; Firestein et al., 1989). Nevertheless, immune responses in vivo to certain pathogens are reproducibly polarized to be of ‘‘TH~’’ or ‘‘TH~” type (Mosmann and Coffman, 1989), and there is evidence that THo-like, THl-like, and T ~ z - l i k ecells in uiuo may be separated on the basis of differences in their expression of certain surface markers (Hayakawa and Hardy, 1988; Bottomly et al., 1989). While CTL and T H cloned ~ cell lines can share a large number of properties (Table I), substantial evidence indicates that their analogs in vivo exercise very distinct roles. A striking feature of the helper and killer subsets is that their functions generally appear to segregate according to expression of the coreceptors CD4 and CD8. As shown in Table I, most efficient helper lines are CD4+ whereas most efficient killer lines are CD8+. Even TcRyG cells can express CD8, but usually not CD4 (Goodman and LeFranqois, 1988). The CD4/CD8 distinction

TABLE I CHARACTERISTICS OF T CELLSUBSETS AS DEFINED BY CLONED LINES=

Cell T typeb

Coreceptor

T C R ~ P T H O CD4

Lymphokines Produced

Cytotoxicity

Lymphokines Stimulating Growth

Lymphokines Inhibiting Growth

TCRaPTH1

CD4

IL-2, IL-4, IL-10, IFN-y, TNF-a IL-2, IFN-y, TNF-a, TNF-P

None?

IL-2, IL-4

No?

Yes

IL-2>IL4

NO

TCRaPTHP

CD4

IL-4, IL-5, IL-6, IL-10, IL-1

NO

TCRaPCTL

CD8

Yes

TCRy8

None; CD8

IFN-7, TNF-a, TNF-/3 ( * IL-2) TNF-a, TNF-/3, IL-2, IFN-y

IL-2 + IL-1, IFN-y I L 4 + IL-1 IL-2, IL-4 No

Yes

IL-2

?

Inhibitors of Activation' Anergy-vs. IL-2, CAMPvs. IL-2 Anergy-vs. IL-2, CAMPvs. IL-2, IL-lo-vs. IL-2, Anergy-vs. proliferation, CAMPvs. TCR signal None No anergy, IL-10-vs.

IL-2

?

a Data in this table taken from Street and Mosmann (1991), Romagnani (1991), Koizumi et al. (1991), Betz and Fox (1991), Allison and Havran (1991), Ferrick et al. (1991),Munoz et al. (1990). Gilbert et 02. (1990), Ju et al. (1990), Firestein et al. (1989),Patel et al. (1989), Mosmann and Coffman (1989), Gajewski et al. (1989), Torbett and Clasebrook (1989). Greenbaum et al. (1988), Colamonici et al. (1988),and Fernandez-Botran et al. (1988). Most of the data in the table refer to murine T cells established as cloned, antigen-dependent lines. An exception is some ofthe evidence on TCRy6 cell cytotoxicity, which is derived from human cells and lines. Stimuli that inhibit activation may block IL-2 production selectively (vs. IL-2), block TCR-PLC coupling (vs. TCR signal), or block proliferation on restimulation with antigen + APC. Anergy is induced under conditions described in text.


101

in function is particularly pronounced in zjivo. Freshly isolated T cells from spleen or peripheral blood can be sorted preparatively into CD4+ and CD8+ fractions, in which the provision of growth factors and B cell help is found concentrated in the CD4+ fraction, whereas activatable killing activity is found almost exclusively in the CD8+ population. This correlation implies a similar correlation of the coreceptor gene expression with the expression of what I have termed response genes, i.e., the inducible loci that become transcriptionally activated in T cells as a result of antigen recognition. Yet at a literal, molecular level, CD4 clearly is not coordinately regulated with, for example, IL-4, because CD4 is transcribed constitutiveiy in helper T cells and IL-4 is transcribed only upon induction, and only in a subset of helper T cells. The mechanistic basis of the correlation between coreceptor/ recognition specificity and functional response type is one of the outstanding unresolved questions in T cell development. On inspection it is clear that most, or all, individual response genes can be utilized in more than one T cell subset, i.e.,in combination with more than one set of other response genes. For example, IL-2 can be expressed in cells that also make IL-4 (THO cells) or, occasionally, in cells that are killers. Conversely, though CD4+ cells in uiuo lack killing activity, T H 1 long-term lines can acquire the ability to kill. IFN-y is not expressed in IL-4-expressing T H 2 cells, but it can be expressed with IL-4 in THO cells, with IL-2 in TI31 cells, and without IL-2 or IL-4 in conventional CTLs. Leaving aside for the moment the lineage relationships among these various cell types, such mosaic expression patterns imply that response genes can be programmed for inducibility independently of one another. In other words, combinations of response properties that we do not normally see together may still be mechanistically possible. To explain why we see certain characteristics most commonly clustered with certain others, it is necessary to ask how particular T cells manage to meet the induction requirements of different individual response genes in the course of the same response. One clue is that different response genes probably are not activated b y exactly identical signaling biochemistry. This conclusion comes from several kinds of evidence. First, when lymphokine production is compared with CTL killing activity, the kinetics of the responses are very different. The mitogenic lectin concanavalin A (ConA) can trigger IL-2 production within hours (Ullman et al., 1990), whereas CTL function is not detectable for 3-4 days. Killing of target cells is a process that differs from lymphokine expression; though the killing is almost instantaneous, it depends on the prior assembly, in the killer T

102

ELLEN V. ROTHENBERG

cell, of lytic granules that have been stocked with an array of serine proteases (granzymes) and membrane-perforating proteins (perforin, or pore-forming protein). Such an assembly process would be expected to take time. However, even the mRNAs encoding the granzymes are not expressed at maximal levels unti-l after several days of antigen activation, in the presence of lymphokines, including IL-2 (Hooton et al., 1990; Liu et al., 1989).This raises the possibilities either (1)that T cells can only express CTL genes after differentiating to a new “antigen-experienced” state, or (2) that some compound generated in the course of the incubation (e.g., a lymphokine) is needed as a supplementary stimulus after TCR ligation to induce the cell to express CTL genes. Indeed, IL-2 appears to induce some of these genes more rapidly than restimulation via the TCR in preactivated human T lymphoblasts (Liu et al., 1989). There are other cases in which lymphokines are required as adjuncts to TCR ligation to induce particular response genes. For example, IL-4 expression can normally be induced in murine splenic T cells only after several days of prestimulation and culture (Powers et al., 1988; Swain et al., 1988a). However, W. Paul and co-workers have found that the delay can be eliminated if exogenous IL-2 is supplied to the cells at the time of their initial contact with antigen (Ben-Sasson et al., 1990; LeGros et al., 1990).This suggests that the signal for induction of IL-4 in primary murine T cells is the sum of TCR-ligand interaction plus IL-2/IL-2 receptor interaction. Third, Bohjanen et al. (1990)have shown that T Hclones ~ actually induce IL-4 and IL-5 in response to different cytokine signals, using different biochemical pathways, even though both are also expressed in response to TCRICDS ligation. A fourth case, in which different combinations of signals probably activate different response genes in the same cells, is the effect of cAMP elevation on IL-2 versus IL-4 production. This is likely to be of more than pharmacological significance, for T cells encounter many CAMP-elevating stimuli, from P-adrenergic agonists to prostaglandin E l and Ez, in their normal courses of circulation as well as during intrathymic development (Papiernik and Homo-Delarche, 1983; Kammer, 1988; McCormack et al., 1991). cAMP sharply inhibits the induction of IL-2, but has little if any effect on the induction of IL-4 (Kim et al., 1988; Gajewski et al., 1990; Muiioz et d.,1990; Novak and Rothenberg, 1990; Betz and Fox, 1991). Part of this effect is cell subset specific: cAMP appears to uncouple the TCR from PLC in T H but ~ not in T Hcells ~ (Lerner et al., 1988; Gajewski et al., 1990; Muiioz et al., 1990; Betz and Fox, 1991). However, part of the effect is to modulate expression of the two genes differentially at the RNA synthesis level even in the same cells (Novak


103

and Rothenberg, 1990; Betz and Fox, 1991). Thus the combination of a stimulus that elevates CAMP levels with the signals derived from TCR-antigen recognition can result in the expression of a different balance of response gene products than activation with TcR ligands alone. This implies that some of the selectivity of response gene expression in different T cell subsets may be related to the differential expression of receptors for non-TCR-mediated signals. A particularly interesting example of such a receptor is likely to be the CD28 molecule, a receptor for a cell surface molecule, called BBl/B7, of activated B cells and monocytes (Linsley et al., 1990).CD28 is expressed on the great majority of human CD4+ cells and on about half of the CD8.+cells (Yamada et al., 1985). Antibodies against CD28 have been known for years to provide a unique signal to T cells, not duplicating TCR signals but greatly potentiating their effects (Hara et al., 1985; Moretta et al., 1985). CD28 ligation induces a dramatic increase in the half-lives of labile lymphokine mRNAs and enhances their rates of transcription (Lindsten et al., 1989; Fraser et al., 1991).The products affected are the characteristic products of T H cells: ~ IL-2, IFN-.)I, and TNF-a (Thompson et al., 1989; June et al., 1989). Yet another way CD28BBl/B7 interaction may facilitate T H cell ~ responses particularly is that it can apparently overcome the THI-Specific inhibitory effects of CAMP (Ledbetter et al., 1986b). It now seems increasingly likely that purified T cells may not be able to activate IL-2 gene expression at all unless the TCR-antigen interaction is accompanied by CD28-BBlI B7 interaction (June et al., 1989; Gimmi et al., 1991; Linsley et al., 1991). As Mueller et al. (1990) point out, even the pharmacological stimuli of Ca2+ ionophore + phorbol ester appear to provide an insufficient signal for IL-2 induction in cases in which cell-cell interaction is disfavored. 2 . IL-2 Gene Regulation: A Case Study The mechanisms of combinatorial effects such as those just described can operate at several levels. One level, at least, seems to be the interaction of separately regulated transcription factors in the 5' flanking regulatory regions of the response genes. The promoters for IL-4 and for the CTL-specific products perforin and various granzymes are only now becoming defined. However, the IL-2 regulatory region has been extensively dissected and reveals a rich complexity of interactions. The IL-2 5' flanking sequence possesses binding sites for a large

104

ELLEN V. ROTHENBERG

array of known and novel transcription factors. Many of the better characterized ones are shown schematically in Fig. 2. There are at least two octamer-binding sites (Kamps et al., 1990); two sites for distinct factors, both with AP-l-like specificity (Serfling et al., 1989; Muegge et al., 1989; Novak et al., 1990; Randak et al., 1990); a site for an NF-KBlike factor (Hoyos et al., 1989; Radler-Pohl et al., 1990); a site for a factor with a specificity resembling AP-3 (Serfling et al., 1989; Fraser et al., 1991); at least one likely site for Spl (Chen and Rothenberg, 1992); and a long, complex site for a factor called NF-AT (Shaw et al., 1988). The two AP-l-like factors, NF-AT, and NF-KB are all differentially regulated by various combinations of pharmacological stimuli (Emmel et al., 1989; Muegge et al., 1989; Novak et al., 1990; GranelliPiperno et al., 1990).One AP-1 factor and NF-KBappear to be activated posttranslationally, whereas NF-AT is synthesized de nouo. One AP-3like factor, CD28RC, in fact is not inducible at all in Jurkat cells in response to simple TCR ligands; it is only activated when the cell is costimulated with ligands for CD28 (Fraser et al., 1991). If IL-2 gene induction depends on the presence of all its positive factors (and on the absence of any putative negative ones), then its transcription can only be initiated when all the disparate activation conditions for all the separate positive regulators are met. Note also that many of the factors listed are not T cell specific. Thus, even the presence of the candidate T cell-specific factor NF-AT would not guarantee IL-2 induction if, for example, a posttranslational modification failure prevented AP-1 from binding as well. Many other genes expressed with precise tissuespecific and temporal regulation have been found to have similarly complex regulatory sequences, which have been described as acting like a “logic chip” (Davidson, 1990). Thus, each of the different response genes may be activated by overlapping but nonidentical sets of transcription factors. We can understand synchronous induction of two response genes in the same cell in terms of the activation of a common, rate-limiting transcription factor, provided that all other required factors for both genes are already available. In other cases, only the requirements for activating one gene may be fulfilled at a given time. In summary, the ability of a T cell to express a particular set of response genes is an indication both that the cell expresses all transcription factors necessary for their induction, and that it mobilizes the right signaling pathways, on contact with antigen and APC, to activate these factors in the right combinations. It is in these terms that we must ultimately explain why the preference for certain response pathways is correlated with CD4 versus CD8 expression.

105


324

298

285

265

-195

178 -161

-143

93

63

I-= -1

300 -278

260

240-211 -192 - 1 7 4

156

A L,

t

*

IL-I

4 * CAMP FIG.2. Combinatorial regulation of the IL-2 gene. The figure diagrams the 325-bp enhancer/promoter region of the IL-2 gene, with some of the binding sites for characterized factors. Shaded rectangles indicate sites for inducible binding activities, whereas striped ovals indicate sites apparently constitutive activities. Data are from the work of Crabtree and co-workers (Ullman et al., l990), Emmel et al. (1989), Fraser et al. (1991), Randak et al. (1990), Radler-Pohl et al. (1990), Hoyos et al. (1989), Novak et al. (1990), Granelli-Piperno et al. 1990), Serfling et al. (1989), Kamps et al. (1990), and Chen and Rothenberg (1992). Arrows indicate the pathways for the induction of each inducible binding activity, with the most important stimuli boxed. Thus, NF-AT binding activity is dependent on the full stimulus of TCR engagement, and requires new protein synthesis. Cyclosporin A (CSA) blocks the induction of this activity. Protein kinase C (PK-C) activation alone, however, is sufficient to induce the NF-KB-like activity, and normally also suffices to induce the factor binding at the AP-lp site. In Jurkat cells, only a signal derived from CD28 ligation is able to induce the binding activity of CD28RC. Outside of the boxes indicating the principal stimuli for each binding activity are listed the effects of additional stimuli. Thus, for example, the addition of IL-1 specifically enhances AP-1 and NF-KB binding activity, whereas forskolin also enhances the former but not the latter. CSA has no effect on these binding activities. Although Oct-1 binding activity is constitutive, several lines of evidence indicate that an inducible activity with activation requirements similar to NF-AT binds at or near the proximal Oct binding site (major stimulus TCR??). It is not clear whether this factor cooperates with or competes with the constitutive Oct-1 binding activity. Two constitutive binding factors interact with the region from -300 to -320 (320) and the purine-rich region distal to the NF-AT site (AGF). The latter appears to be S p l (D. Chen and E. V. Rothenberg unpublished).

106

ELLEN V. ROTHENBERG

E. THEPROLIFERATIVE RESPONSE 1 . Growth The ability of mature T cells to proliferate when stimulated by antigen deserves separate consideration because of its importance for the system and the two-step nature of its regulation. Though proliferation is the most commonly used indicator of q T cell’s functional maturity, it is not a direct response to antigen contact. Instead, it is mediated by the joint inducibility both of growth factors such as IL-2 and of their high-affinity receptors. TCR- and coreceptor-derived signals induce a “stage 1” activated state, in which IL-2 is produced and high-affinity IL-2R are expressed. It is then the IL-2/IL-2R interaction that induces “stage 2,” i.e., the commitment to enter the cell cycle. These two stages are diagrammed in Fig. 3. As shown in the figure, not only are the stimuli of “stage 1” activation and “stage 2” cycling different, but also their immediate effects are quite distinct. Moreover, the two stages can be shown to be independent by using pharmacological inhibitors to block one but not the other [cyclosporin A (CsA), FK506 versus stage 1; rapamycin versus stage 2, in Fig. 31. Thus, to proliferate, a T cell must undergo two distinct inductive interactions in proper sequence. For most T cells, the rate-limiting factor in the ability to proliferate is their expression of high-affinity IL-2R. The IL-2R comprises at least two ligand-binding chains, a and 0, both of which are integral membrane proteins with single bilayerspanning regions. Both are needed to form a high-affinity, internalizable receptor ( K d - 1OpM).The IL-2Ra chain alone cannot be internalized. The IL-2Rp chain can, however, serve as an IL-2 receptor in the absence of the a chain: it is internalizable and transduces signals. Even so, it only binds IL-2 with a & of -1 nM. The significance of these differences in binding affinity emerges from the differential regulation of the a and p chains of the IL-2R. The P chains appear to be expressed constitutively on resting T cells (Siege1et al., 1987; Ohashi et al., 1989). The a chains, however, are only expressed after activation, responding to signals either from the TCR and its coreceptors or from the IL-2/IL-2R interaction. Resting T cells prior to antigen recognition can be stimulated by IL-2, via the IL-2RP chain, but very high local concentrations of IL-2 are needed (Lbthi Bich-Thuy et al., 1987; Yagita et al., 1989a; Ben Aribia et al., 1989).IL-2 is a labile protein, and it is produced only transiently after stimulation, so that such high levels of hormone are rarely encountered in viuo. After antigen stimulation, however, the expression of IL-2Ra chains in association with the p chains makes the T cell 100-fold more sensitive to IL-2, in a


107

Rapamycin

FK506

\ Y

CsA

fi Cycling

v :s+ &/ 1

.la //

tw

1

tc-fos

J A

I

APC

tc-myc tPH i t1UR

[

.T I

tc-myc t& ttransferrin-R t1ER tDNA svnth

tz

t u FIG.3. Two stages o f T cell activation. The figure diagrams the two discrete stages of activation needed to induce a resting T cell to enter cycle. The first stage, induced by TCR-coreceptor-Ag interaction, leads to an activated state. The effects associated with this process conspicuously include the induction of Ca2+ fluxes and PK-C activation, with resulting induction of genes such as the IL-2 gene and those encoding transcription factor components myc and fos. The second stage, induced by an IL-2/IL-2R interaction, leads to a cycling state. Note that this does not replicate the Ca2+and PK-C effects of the first stage, but instead leads to a series of effects associated with commitment to cycle. Induction of transferrin receptors (transferrin-R) is one such effect, as is the induction of myb expression and the onset of DNA synthesis. In the list ofeffects associated with each transition, those likely to be unique to the first or second stage are marked on the figure by underlining. Certain genes are induced in both stages of activation, e.g., myc and the gene for IL-2Ra. This appears to reflect dual regulation of the genes rather than a duplication of activation signals, for the two stages can be shown to be mechanistically independent by the specific, reciprocal effects of inhibitory drugs. Cyclosporin A (CSA) and FK506 block the first events but not the second, whereas rapamycin blocks the second stage but not the first (Bierer et al., 1990; Dumont et al., 1990). The target proteins that mediate the effects of CSA and FK506 are indicated on the figure [cyclophilin (CyP) and FKBP, respectively]. FKBP also appears to mediate the very different effects of rapamycin, suggesting that it interacts differently with signaling mediators for both stages of activation. As noted in the text, the ability of a cell to progress from stage 1 activation into stage 2 depends in part on the quantitative balance of TCR stimuli relative to IL-2/IL-2R stimuli and the time interval between them. IPS, Inositol phosphates; pHi, internal pH. Induction of protooncogenes is described in Moore et al. (1986), Shipp and Reinherz (1987), and Churilla et (11. (1989). Evidence that the promoters used b y the c-myc transcript are different in the first and second stages of activation is discussed by Broome et al. (1987).

108

ELLEN V. ROTHENBERG

practical sense making the difference between responsiveness and nonresponsiveness. At this level of sensitivity, cells become capable of being driven into cycle by amounts of IL-2 that may be synthesized by even relatively rare cells in the vicinity. We have already noted that induction of IL-2 is stringently controlled, limited to certain T-cell types, and dependent on adequate Ca2+, PK-C, and CD28-derived signals and the maintenance of low levels of CAMP. By contrast, IL-2Ra chain induction is much more permissive. Inducibility of IL-2Ra and thereby of responsiveness to IL-2 is essentially universal among mature effector T cells. In some cases, phorbol ester alone can induce IL-2Ra expression and cellular responsiveness to exogenous IL-2, even though it fails to induce IL-2 (Mills et al., 1985a; Yamamoto et al., 1985). IL-2Ra induction is also resistant to inhibition by CAMP, glucocorticoids, or the immunosuppressive drug cyclosporin A under conditions in which IL-2 induction is blocked (Hess et al., 1982; Chouaib et al., 1985; Isakov and Altman, 1985; Granelli-Piperno et al., 1986, 1990; Reed et al., 1986). Under conditions of insufficient stimulation, a cell of an IL-2 producer subset may fail to make IL-2, but may become responsive, even so, to IL-2 made by other cells. Thus, the mitogenic response in a given tissue site is limited by two thresholds: the threshold of triggering to activate IL-2 expression by a small number of cells, and the much lower threshold of triggering to confer high-affinity IL-2 receptors on a larger, more diverse set of T cells. The IL-2/IL-2R interaction induces cell cycle progression through a signaling pathway completely distinct from that of the TCRcoreceptor-antigen stimulus (Albert et al., 1985; Mills et al., 198513; Weiss and Imboden, 1987). IL-2 can act on cells without inducing detectable phosphoinositide hydrolysis, Ca2+ elevation, or cyclic nucleotide elevation (Kozumbo et al., 1987; Dunn et al., 1987; Tigges et al., 1989), and the IL-2/IL-2R signal does not even appear to require the presence of the major isoforms of PK-C in the cell (Mills et aZ., 1988; Valge et RZ., 1988).Some kinase noncovalently associated with the IL-2R complex appears to become activated in response to IL-2 exposure (Gaulton and Eardley, 1986; Ferris et al., 1989; Turner et al., 1991), and Fung et al. (1991) suggest that it is a tyrosine kinase. In certain cell lines, there is evidence that Lck can interact with IL-2RP (Hatakeyama et al., 1991).Whether this is the rule and whether IL-2RP must compete for Lck with CD4 or CD8 in a conventional T cell remain to be determined. Some insight into the IL-2R signaling cascade should ultimately come from identification of the regulatory factors that control response genes that are inducible by IL-2. The IL-2/IL-2R interaction can in-


109

duce several lymphokines, including IFN-y in T H or CTL cells (Farrar et al., 1981; Heckford et al., 1986; Dunn et al., 1987),IL-la in T Hcells ~ (Zubiaga et al., 1991), and the IL-2Ra chain, as mentioned above. Of these, the regulatory region of the IL-2Ra gene is the best characterized. It utilizes several unique transcription factors for induction as well as NF-KB and a “serum response factor” (CArG box binding factor) (Ballard et al., 1989), and is repressed by S p l (Roman et al., 1990).Certain regions of the IL-2Ra 5’ flanking sequence are more important for expression in IL-2-dependent YT cells than in IL-2independent Jurkat cells (Lowenthal et al., 1988). They may therefore define regulatory elements responsive to IL-2/IL-2R signaling. IL-2 is not the only growth factor that can drive T cell proliferation: once primed, at least some T cells are also responsive to IL-4, especially T H cells. ~ IL-4 is by no means a universal T cell growth factor, however. Whereas its receptor (IL-4R) currently appears to be a single, constitutively expressed chain with a single binding affinity (Mosley et al., 1989), T Hcells, ~ like other T cell types, require antigen priming to boost their responsiveness to IL-4. The activity of the IL-4R appears to depend on a costimulating signal, delivered by the hormone IL-1 (Greenbaum et al., 1988; Kurt-Jones et al., 1987). IL-1 is normally a product of non-T cells, but it can be made by T Hcells ~ when they are stimulated (Tartakovski et al., 1988; Zubiaga et al., 1991). Thus, although the biochemical details differ, growth stimulation via the IL-4R requires at least two gene induction events (IL-4 and IL-1), as does growth stimulation via the IL-2R (IL-2 and IL-2Ra). Once under way, a successful proliferative response can involve many rounds of cell division over a period of days, as long as IL-2 or another growth factor is available. Through an unknown mechanism, however, the response is self-limited. Even T cells that retain strong expression of high-affinity IL-2R become refractory to the growthstimulatory effects of IL-2 in several experimental systems (Gullberg and Smith, 1986; Churilla and Braciale, 1987). After the cells cease to cycle, the sensitivity of their TCR to stimulation returns to maximal levels. 2 . Anergy The various ways T cell mitogenesis can fail have received broad attention recently with the description of the phenomenon of clonal anergy. Though long discussed as a hypothetical possibility (Nossal, 1983),defined experimental protocols for inducing T cell anergy were only recently developed (Jenkins and Schwartz, 1987; Quill and Schwartz, 1987; Mueller et al., 1989). When a T H 1 clone was activated b y chemically fixed APC or by purified antigen-MHC complexes in a

110

ELLEN V. ROTHENBERG

lipid bilayer, the T cells underwent an abortive response. Not only did the cells fail to proliferate or to make IL-2, but after several hours of exposure to the inappropriate stimulus they also became deeply refractory to restimulation through the TCR. At the same time, they expressed high-affinity IL-2R and could still be induced to proliferate well by addition of exogenous IL-2. Recent results suggest that anergy may be the consequence of TCR engagement without CD28 engagement (June et d., 1989; Mueller et d., 1990). As we shall see below, anergy appears to be very important for the fine tuning of the effective T celI repertoire. Thus, it is particularly important to distinguish whether anergy is the same as simple insufficiency of stimulation. Both result in IL-2Ra induction without IL-2 expression or proliferation, but anergy is implicated in the lasting paralysis of inappropriately activated clones. By contrast, the relatively lax requirements for IL-2Ra induction can be viewed as a mechanisms for amplifying T cell proliferative responses, i.e., recruiting a broad spectrum of T cells into a response once a pathogen meets the stringent activation criteria needed to induce IL-2 production from even a few T cells. A priori, it would seem that the system-wide utility of anergy as a mechanism for maintaining tolerance would be seriously compromised if “anergy” included all cases in which IL-2Ra induction ~ studied exceeds IL-2 induction. For example, several cloned T Hlines by Fitch and co-workers exhibit their most prolific autocrine growth, i.e., their most potent IL-2 responsiveness, at doses of TCR ligand that induce barely detectable IL-2 secretion, where presumably only a minority of the cells secrete IL-2 (Gajewski et al., 1989).CD8+ CTLs, which apparently do not become anergic, obligatorily respond to induction by expressing IL-2Ra but not IL-2. It remains to be determined whether the difference between priming to exogenous IL-2 dependence and induction of anergy are distinguished qualitatively, or only kinetically. Of all the properties of mature T cells, the control of proliferation has the most significant systemic impact. By regulating clone size and the balance of subsets of different types, differential proliferative behavior shapes both the effective T cell recognition repertoire and the nature of an immune response. 11. The Thymus and Its Seeding

A. THETHYMIC ENVIRONMENT In the previous section the case was made that differentiation into a mature T cell is as much a process of creating response potential as of restricting developmental potential. A mature T cell may be terminally


111

differentiated in one (traditional) sense, but it retains considerable gene expression plasticity and a virtually unlimited proliferative capacity. Its ability to respond explosively to peptide-MHC complexes that might be found anywhere in the body make it particularly dangerous. It is perhaps understandable that the signals that control T cell development need not only to make the right genes accessible for expression, but also to ensure that they are appropriately regulated. Accordingly, most T cells, unlike all other known hematopoietic cells, are produced in the thymus, an organ specialized for the expansion and stringent selection of T cell precursors. The role of the nonlymphoid component of the thymus, the thymic epithelium, is to stimulate extensive proliferation in early precursor cells, control their differentiation, and sequester the products until all ineffective or autoaggressive cells have been culled. To accomplish these various functions, the thymus is organized into distinct domains with separate classes of epithelial cells and auxiliary colonization by separate classes of nonlymphoid hematopoietic cells. The structure of the thymus has recently been reviewed in depth (Lobach and Haynes, 1987; van Ewijk, 1991). Here those features that will be most pertinent to T cell development are briefly summarized. In mammals, the thymic epithelium is initially formed from the interaction of the third and fourth branchial pouch endoderm with the overlying ectoderm of the pharyngeal cleft. When this occurs, at days 10-11 of gestation in the mouse and in weeks 7-8 in the human (Lobach and Haynes, 1987; Janossy et al., 1980), no hematopoietic cells are present in the tissue, although they begin to immigrate soon afterward. The epithelial tissues sort themselves into two distinct compartments: a cortex surrounding a medulla. These epithelial domains become distinguishable at an early stage, when there are still very few lymphocytes present. Ultimately, the cortex becomes a lacy meshwork of epithelial cells with long, branching processes, which is heavily packed with lymphocytes. The cortex also contains scattered macrophages. The medullary region has a distinct epithelial cell type of more conventional form, and it includes bone marrow-derived dendritic cells, particularly in the region near the cortical-medullary border. One may consider the outer cortical region, just under the fibrous capsule of the organ, to be yet another distinct microenvironment, for this is observed to be a unique region in which the majority of lymphocytes are actively dividing. Thus, it is likely that the stroma here delivers a particularly powerful mitogenic signal. The distinctive epithelial cell type in the subcapsular region is the thymic nurse cell, a large, MHC class II+ epithelial cell type that binds thymic lymphocytes so tightly that, when isolated from the thymus by collagenase

112

ELLEN V. ROTHENBERG

treatment, the nurse cells are found to have engulfed many lymphocytes (Wekerle and Ketelsen, 1980; Kyewski, 1986). It is noteworthy that the engulfed thymocytes can remain viable inside the nurse cells and subsequently may be released in culture. This raises the fascinating possibility that some of the inductive interactions that guide T cell development in vivo may occur while the immature thymocytes are physically enclosed in a nurse cell, and sequestered from the outer hormonal or antigenic environment. A detailed summary of the characteristics of different thymic stromal cell types is provided by van Ewijk (1991). T cell development in the thymus is normally organized by a programmed intrathymic migration route. Precursors from the bone marrow probably enter near the cortical-medullary border, from which they slowly migrate to the outer region of the cortex. During this period they begin dividing. Proliferation reaches its maximum intensity in the subcapsular zone. As described below, this phase culminates with the expression of TCR/CD3 on the surface. The cells thus generated stop proliferating, shrink, and begin to migrate back into the thymic cortex. One-third of all cortical thymocytes, or about one-quarter of all thymocytes, are newly generated each day in the mouse. These cells retain a distinctively nonmature phenotype, described in Section 111, as long as they reside in the cortex. The cortical epithelial cells appear to be necessary and sufficient to convert lymphoid precursors into cells with a “mature” pattern of cell surface markers (Jenkinson et al., 1982; Benoist and Mathis, 1989;Bill and Palmer, 1989; Kingston et al., 1985). Only lymphocytes that have acquired a mature phenotype enter the medulla. The role of the medulla is more subtle. It may be involved in functional maturation as well as in certain selection processes (see Sections IV and V). In fact, the medulla is also open to recirculation of activated peripheral T cells (Naparstek et al., 1982),and may serve as a point of entry for antigens and antigen-reactive cells from the body generally. In the postnatal thymus, the stromal architecture remains essentially stable until puberty, when sex hormones begin to induce a gradual involution process. During late fetal life, however, the thymic microenvironment undergoes reorganization and differentiation even while the first cohorts of T cells are maturing. The first wave of T cells is produced by about day 15 of gestation in the mouse (we will discuss some distinctive features of these cells in Section III,B,C). Whereas the stroma first expresses detectable class I1 MHC expression by day 13, it does not express class I MHC detectably before day 16 (Jenkinson et al., 1981).Thus, the first wave of thymocytes may mature in a


113

class I-deficient environment. Dendritic cells, which may be implicated in crucial negative selection processes, are not reported to exist in the fetal thymus until even later in ontogeny (Sminia et al., 1986). Cortical stromal cells marked by the distinctive 6C3 marker do not appear until 3 weeks after birth (Adkins et al., 1988).Thus, the successive waves of thymic lymphocytes that differentiate in the fetal and early postnatal thymus must each receive their inductive signals from significantly changing microenvironments. This process is diagrammed schematically in Fig. 4. We do not yet know the extent to which these microenvironments deliver functionally different signals, but such a process could participate in generating T cell cohorts with considerably different properties (see Section 111,C). In the postnatal animal with a fully formed thymic stroma, the most

T-cell maturation

-

FIG.4. Overlapping waves ofT cell differentiation in a changing microenvironment. The figure schematically depicts the simultaneous changes in sources of stem cells (pre-A, -B, and -C) and maturation states of the thymic microenvironment (1-4) upon which are superimposed the sequential differentiation events affecting individual cohorts of T cells (thick arrows). The initial inductive signals acting on T cell precursors (curves in thick arrows) may differ according to the developmental state ofthe stroma at the time, as may the continuing microenvironmental influences exerted by the stroma until T cell differentiation is complete (dashed arrows). Thus not only the cell source, but also the sum of influences acting on cohort TAmay be considerably different from the sum of influences acting on cohort Tc.

114

ELLEN V. ROTHENBERG

conspicuous cell-cell interactions are influences of the thymic stroma, i.e., epithelium and nonlymphoid hematopoietic cells, upon lymphoid precursors. This is clearly demonstrated by the ability of radioresistant thymic elements to direct the differentiation of donor-type lymphoid immigrants during repopulation of a radiation chimera. In fetal life, however, and in the thymus of mutant SCID mice, the interaction is revealed to be bidirectional. First, thymic epithelial cells are demonstrably sensitive in vitro to lymphokines of the kinds made by developing T cells (Kisielow et al., 1985). Papiernik and colleagues (Rocha et al., 1988)have shown that a certain class of thymic stromal cell expresses IL-2 receptors and grows in vitro in response to IL-2. Nonmacrophage thymic stromal cell lines, presumably of epithelial origin, can also acquire the capacity to present antigen to T cell hybridomas when treated in vitro with the T cell product IFN-y (Ransom et al., 1987). The in vivo relevance of such observations is shown by two cases in which direct perturbations of thymic lymphocyte populations have dramatic effects on the organization of the surrounding thymic epithelium. In SCID mice, T cell maturation is blocked because the cells usually die at or just after the stage during which they would normally rearrange their TCR loci (Bosma and Carroll, 1991). Staining of the SCID thymus epithelium with specific antibodies reveals that these thymi are profoundly deficient in epithelial cells of the medullary type (Shores et al., 1990, 1991). The introduction into the SCID thymus of any cells capable of differentiating into TCR+ cells, either via making a bone marrow chimera or in a naturally “leaky” SCID animal, results in the appearance of a normal, well-organized medullary epithelium. Thus the presence of TCR+ cells is necessary to induce the thymic epithelial cells to form or maintain a normal medulla. The second case of T cell-stromal interaction suggests the identity of a T cell lineage that may have medullary induction as one of its roles. Transgenic mice with a previously rearranged V,l.l JC,4 transgene are effectively forced to use this chain in all their y6 TCRs, although normally it is used by very few y8 T cells. We will discuss below the fact that y6 cells dominate the first wave of T cells to arise in the fetal thymus (Sections III,B and 111,C);the effect in this case is that cells with the transgenic TCRs mature very early in ontogeny. Surprisingly, in these animals, the development of TCRaP’ cells is not blocked but even slightly accelerated, and there is a major expansion of the thymic medullary epithelium (Ferrick et al., 1989, 1990). Thus, either directly or indirectly, the expression of this particular TCRy chain allows lymphocytes to stimulate the development of the thymic medulla.


115

B. ORIGINSAND ALTERNATIVES Since 1977 (Abramson et al., 1977) it has been clear that the same stem cells that populate myeloid hematopoietic lineages can also give rise to T cells. For many years it has been assumed that differentiation into a prothymocyte necessarily preceded entry into the thymus. Recently, it has become possible to address the question of hematopoietic lineage commitment in a direct way as a result of the identification and preparative isolation of self-renewing, pluripotent hematopoietic stem cells (Spangrude et al., 1988). Using injection of nearly pure pluripotent hematopoietic stem cells at limiting dilution into the thymus, Spangrude and Scollay ( 1990) have provided spectacular confirmation that even without prior differentation, such cells can respond directly to the thymic microenvironment and develop as thymocytes. A question remains whether the cells that normally migrate to the thymus retain the full range of hematopoietic developmental potential, or if they are already lymphoid restricted. Given the powerful inductive effect of the thymic microenvironment, the acquisition of a thymushoming receptor may become defucto a commitment event, even if thymic immigrants remain capable in principle of responding to myeloid differentiation-inducing factors. There are several reports that pluripotent myeloid cells or cells capable of nonlymphoid differentiation can indeed be isolated from the thymus (Kurtzberg et al., 1989; Ezine et al., 1991). However, it is not certain whether these are cells that would otherwise have continued along a T cell developmental pathway or cells that would simply have remained undifferentiated in the thymus. Thus the extent of true developmental plasticity among cells migrating to the thymus remains unresolved. Though the thymus is certainly a potent inductive microenvironment, T cell development is strongly, but not absolutely, dependent on the thymus, Congenitally athymic “nude” mice, which fail to form a normal thymic epithelial anlage due to a defect in embryogenesis (Cordier and Haumont, 1980; Jenkinson et al., 1981; van Vliet et al., 1985), are severely deficient in most T cell populations. The T cells they possess throughout puberty are almost exclusively TCRy8 cells. Yet if they can be kept alive long enough, they ultimately accumulate an adequate population of TCRaP cells, presumably by slow production in a nonthymic site coupled with peripheral expansion. Though the mechanism of peripheral expansion is not yet well-defined, it is clear from a number of systems that even a small number of mature T cells can expand to significant levels in the absence of new thymic output (Miller and Stutman, 1984; Rocha et al.. 1989).It is not yet clear

116

ELLEN V. ROTHENBERG

whether this peripheral expansion effect is completely dependent on proliferation of preexisting cells; it may also reflect activation of extrathymic differentiation sites under the influence of small numbers of peripheral T cells. The groups of Hurwitz and Miller have independently reported in vitro conditions in which bone marrow cells can initiate TCRaP gene rearrangement, expand, and differentiate into T cells (Hurwitz et al., 1988; Benveniste et al., 1990).These culture systems may be a clue that relatively simple stimuli can make supporting cells in the bone marrow capable of replacing the purely differentiative thymic functions. Extrathymic differentiation environments are likely to be quite significant for two classes of cells: a subset of CD8+ TCRaP+ cells, and a subset of TCRy8 cells. For both of these subsets it appears to be the rule, rather than the exception, to develop extrathymically. We will discuss the extrathymic TCRys lineage in Section II1,B (TCR V,5+ cells homing to the intestinal epithelium). The existence of an extrathymically derived subset of CD8’ TCRaP+ cells was suspected a decade ago, when it was found that many of the CD8+ cells appearing in thymus-grafted nude mice did not appear to have been processed by the thymus (Kruisbeek et al., 1981).The argument was as follows. The hallmark of intrathymic processing is that the T cells that emerge have TCRs “restricted” by the particular MHC glycoproteins expressed within the thymus, i.e., their TCRs interact only with antigenic peptides that are associated with MHC allelic products matching, those that were present in the thymus (Zinkernagel et al., 1978). As we will describe below, this is the consequence ofthe positive selection mechanism that controls mature T cell production in the thymus (Sections III,A and V,B). In animals with an allogeneic thymus graft, the MHC restriction specificity of the CD4+ cells generally matched the thymus, but that of many CD8+ cells matched the rest of the body instead. Nongrafted, aging nude mice also have been shown to accumulate CD8+ cells in numbers well in excess of the CD4+ cells (MacDonald et al., 1986). The case for a discrete population of extrathymically derived CD8+ cells has recently been enhanced by the identification of a surface antigen, encoded by the Ly-6 gene complex, which is expressed by essentially all CD8+ cells that develop in the absence of a thymus, but by none of the CD8+ cells within a normal thymus (Leo et al., 1988; Kung, 1988). If, indeed, this marker is restricted to the extrathymic lineage, its expression on 30-40% of peripheral CD8+ cells in normal mice suggests that the extrathymic CD8+ lineage is quite prolific. By contrast, CD4+ TCRaP+ cells are produced very inefficiently in the absence of the thymus. Furthermore, a distinct form


117

of CD8 expression (CD8a homodimers instead of the usual CD8aP heterodimers) has been found to mark a gut-associated population of CD8' TCRa/3+ cells that appears to be completely thymus independent (Guy-Grand et al., 1991; Rocha et al., 1991). There is some evidence that the gut epithelium may be the site for their differentiation (Guy-Grand et al., 1991).We will return to the implications of these alternative pathways after reviewing the transformations that are normally induced by the thymus in cohorts of developing T cells. 111. lntrathymic Transitions in Outline

Over a decade of intensive work b y a large number of laboratories has resulted in a detailed picture of the stages of intrathymic development. As we shall see, appearance of a typical TCR+ CD4+ or CD8+ phenotype is one of the last events in this pathway. The critical changes that progressively distinguish T cells from other hematopoietic cells take place predominantly before such molecules are expressed. Thus it has been particularly useful to analyze immature T cells for their patterns of reactivity with numerous combinations of monoclonal antibodies, including many that bind structures that are not particularly associated with T-cell function. The picture that has emerged is dauntingly intricate, but it allows us to identify specific transitions that will become attractive targets for molecular dissection. Most of the data we will discuss come from work in rodents, where the developmental potential of cells in individual subsets can be tested directly by injection intrathymically into congenic recipients (Shimonkevitz et al., 1987; Crispe et al., 1987; Scollay et al., 1988; Guidos et al., 1989). Genetically marked, preselected immature cells can thus expand and differentiate in a relatively normal environment, without having been required to express homing receptors to find the thymus in the first place. This allows cells to be assayed for differentiative potential both at early and at relatively late stages when homing receptors would normally have been lost. By adjusting the input dose, significant percentages of donor cells can be recovered from precursors with greater or lesser proliferative capacity (Goldschneider et al., 1986). Thus, the increasingly advanced state of different precursors can be verified by the decreasing time their progeny need to reach maturity after intrathymic injection. The following discussion will first deal with the major pathways for producing TCRaP and then TCRy8 cells; then the variants of this scheme for prenatal cohorts of T cells will be discussed, followed by a comparison of the rodent pathways with those that generate thymocytes in different species.

118

ELLEN V. ROTHENBERG

A. THETCRaP LINEAGE First we will consider the transitions and likely developmental branch points for the main TCRaP lineage of T cells that is generated postnatally from bone marrow precursors. In abbreviated form, cells of the TCRaP lineage progress from the CD4-8-TCRICD3- “immature” state (comprising 2-4% of thymocytes) to a CD4+8+ TCRICD3’”” “major cortical thymocyte” state (comprising 80%of thymocytes). Maturation beyond this intermediate condition is restricted by positive and negative selection events, which yield TCR/CD3high“medullary” or presumptively “mature” thymocytes of both CD4+8- (about 10% of thymocytes) and CD4-8+ (about 4% of thymocytes) types. The cells that are exported from the thymus to the periphery are mainly of these last two types (Scollay et al., 1984). This simple scheme provides useful landmarks, which we will return to as points of reference. However, it describes only the penultimate events in T cell development and gives no hint of underlying mechanisms. Therefore, to appreciate the full process, we must turn to the broader scheme provided in Fig. 5 [for other versions and additional literature citations, see Petrie et al. (1990a), Shortman et al. (1990), and Boyd and Hugo (1991).] We will consider the stages indicated in Fig. 5 in order. The most immature cell type yet identified in the postnatal thymus (Wu et al., 1991a) is a cell resembling the primitive pluripotent hematopoietic precursor in the bone marrow (CD4’” in Fig. 5) (Spangrude et al., 1988; Frederickson and Basch, 1989). It has all its TCR genes in germ-line configuration and expresses the hyaluronic acid-binding/ homing receptor CD44 (Lesley et al. 1990), with very low levels of

FIG.5. Outline of T cell development stages in the mouse. The figure summarizes the discussion of Section III,A. Stages in the differentiation of postnatal precursors are shown; these differ in detail from those seen in the fetus. Transitions accompanied by strong proliferation are denoted here by thick arrows. On the left of the figure, an approximate time scale is given in days, leading to the appearance of the first CD4’8and CD4-8+ TCR+++cells, and/or to the disappearance of cortical cells, mostly via “death by default.” The figure does not include the approximately 2 weeks of residence in the medulla that appear to follow positive selection (see Section IV,D and V,B). On the right of the figure, the rearrangement status of TCR genes is indicated, very approximately, for the corresponding stages. Two possible points of divergence between TCRaP and TCRy8 precursors are indicated by dashed arrows (?). In each case, the marker phenotype given is that of the cell surface. This may not always reflect the expression status of components present but incompletely assembled in the cytoplasm. The figure does indicate, however, the distinction between immature IL-2Ra+ cells containing no complete p TCR transcripts ( p - ) and those containing complete p transcripts, and possibly p protein (p’?). For discussions, see text.

119


IMajor cell types I

I

0 cD4io

all germline

CD4108-CD44hi HSA" MHC-I"

0

-9 days

CD4-8-CWhiHSA' MHC-I"

/

S rearr? y rearr 3

?,/

D J rearr ~ 1 a germline I

c

MHC-l'+ VDJp rearr

1

CD43'or CD4'8'CD4'8' IL2R- HSA'" JMHC-I C W TCR/CD3-(pi?)

4

CD4'8' TCRaOlCD3'

VJm rearr Major Cortical thymocyte (Postmitotic)

DEATH SUICIDE tMHC-I rHSA DEFAULT CD4W TCRCIE/CD~'~~ MHC-I" HSA-

4

CD8' mature T

CD4% TCRaO/CD3'" MHC-lt' HSA-

I

CD4' mature T

/

/

120

ELLEN V. ROTHENBERG

CD4 and Thy-1. It also expresses intermediate levels of a useful marker called heat-stable antigen (HSA), the function of which is unknown. Its major distinction from hematopoietic stem cells (Spangrude et al. 1988)is its expression of Sca-2 (Wu et al., 1991b).In the steadystate murine thymus, such cells are exceptionally rare. The earliest changes in this cell appear to be the loss of CD4 and an increase in Thy-1 expression, accompanied by the initiation of a period of slow exponential growth. This may ultimately prove to be the most important period with respect to developmental commitment events, for it is likely to be during this period that expression of the CD37, CD36, and CD3c genes begins and the first TCR gene rearrangement events occur (Furley et al., 1986; van Dongen et al., 1987; Campana et al., 1987, 1989; Carroll and Bosma, 1991). In terms of surface phenotype, however, the cells in this 7- to 9-day period of growth remain unremarkably CD4-8- and surface TCR/CD3-, with only HSA and Thy-1 expression as positive markers and CD44 expressed at declining levels. Toward the end of the first period of expansion, the cells acquire a higher level of HSA, lose CD44, and begin to express a recognizable T cell “response” gene, namely IL-2Ra (Penit et al., 1988; Spangrude and Scollay, 1990; Lesley et al., 1990). By this point, the cells have expanded to 1-2% of all thymocytes [i.e., (2-4) x lo6cells in a 4-weekold mouse]. They are still surface TCR negative, but appear to be partially synchronized at an early stage of TCRP rearrangement (Takei, 1988; Boyer et al., 1989; Pearse et al., 1989). The IL-2Ra+ HSA” stage marks the end of the early phase of T cell development. The cells apparently arrest to await a particular triggering signal from the thymic microenvironment (Nakano et al., 1987; Howe and MacDonald, 1988; Ewing et al., 1988; Petrie et al., 1990a). The next series of events, once initiated, proceeds rapidly and autonomously. After receiving a triggering signal, the cells undergo a transition to a very rapid cycling state, shed their IL-2Ra receptors (Boyer et al., 1989; Pearse et al., 1989; Penit and Vasseur, 1989; Petrie et al., 1990b), and begin an intense program of TCR gene rearrangement that will leave them with fully rearranged P and a genes within 72 hours. The first event in this program is the completion of their TCRP rearrangements, which-even if productive-do not enable the cells to express convincingly detectable levels of cell surface TCRP-CD3 complex. Immediately afterward, the cells begin to express CD8 for the first time (Paterson and Williams, 1987; MacDonald et al., 1988b; Nikolik-Zugi6 and Bevan, 1988), followed 12-24 hours later by reexpression, at a high level, of CD4, giving a CD4+8+ phenotype (Penit and Vasseur, 1988,1989). The TCRa gene rearrangement is completed

FUNCTlONALLY RESPONSIVE T CELL DEVELOPMENT

121

shortly thereafter, a TCRaPICD3 complex is expressed for the first time on the cell surface (Guidos et al., 1990),and proliferation comes to an abrupt halt. The resulting cell type, a small, CD4+8+ TCRaP+, nonproliferating cell (major cortical thymocyte, Fig. 5) is the most numerous cell type in the thymus. It is the major population filling the thymic cortex, and it is the critical population for selection of the immunological repertoire. The details of the events in this rapid proliferative transition are somewhat variable from species to species, and even among cells within a species. In rats, as a rule the TCR/CD3 complex is assembled before CD4 is reexpressed. Coordinate with surface TCR expression, the proliferation stops, so that the transition from CD8'4- to CD8+4+ occurs postmitotically (Hunig, 1988; Hiinig et al., 1989). Of course, ethical considerations prevent detailed population dynamics from being defined in humans, and unfortunately few markers are available to stage these events in most other species. However, even in mice it appears that a small fraction of cells actually acquires CD4 and CD8 in reverse order (Hugo and Petrie, 1991). Furthermore, some workers find that low levels of TCRaP appear on the surface as early as the first CD4 and CD8 (Nikolik-Zugi6 and Moore, 1989).Thus it seems that the biologically significant point is that the cells are brought rapidly to a CD4+8+ TCRapICD3+ state, with only the order of the TCR rearrangement events strictly constrained. Note that rapidly proliferating CD4-8- blasts, once they have lost IL-2Ra expression, can differentiate all the way into small cortical-type thymocytes even when removed from the thymus, simply by culture in conventional tissue culture medium overnight (Fowlkes et al., 1985; Ceredig et al., 1983a; Nakano et al., 1987; Wilson et al., 1989). Thus it seems likely that the actual developmental programming for CD4 and CD8 expression and ordered TCRP and TCRa rearrangements has been completed prior to this stage, i.e., during the long phase of slow expansion that ends with IL-2Ra expression. The triggering mechanism that activates this terminal program is unknown. The many rounds of exponential growth coupled to the earlier developmental transitions build up a penultimate blast cell population in the mouse thymus that produces small cortical cells at the staggering rate of about 5 x lo7per day. This is about 20-50 times higher than the number of mature T cells produced per day by the same thymus (Scollay et al., 1980).Thus, it is clear that the great majority of each crop of cortical thymocytes will not differentiate into mature T cells. This unambiguous mathematical inference, along with other peculiar features of small cortical cells (discussed in detail in later sections),

122

ELLEN V. ROTHENBERG

convinced many investigators for years that cortical thymocytes could not give rise to mature cells. Over the last 4 years, however, insights into intrathymic repertoire selection processes have given circumstantial evidence (MacDonald et al., 1988a; Fowlkes et al., 1988;Teh et al., 1988; Sha et al., 1988a,b), and sophisticated cell transfer experiments have given direct evidence (Guidos et al., 1989), that mature cells do differentiate from a CD4+8+ TCRaP+ cortical-type precursor. Thus, among the cortical thymocytes are a minority that give rise to competent, CD4+ or CD8+, TCRaP peripheral cells. What remains controversial is whether the bulk of cortical thymocytes that fail to mature are cells that have already lost the competence to mature (thus, are not only destined but committed to die), or whether they simply are not presented with the opportunity. We will return to this question in Section V,B as we discuss maturation via the positive selection process. One population of cortical thymocytes that does not mature is the population that has happened to assemble TCRs with autoreactive specificity. These cells are killed by induction of an active suicide process termed negative selection. Such cells may be deleted soon after their first expression of TCR/CD3 (Kisielow et al., 1988; Sha et al., 1988a), or at any time before the cell acquires a medullary phenotype (Kappler et al., 1987,1988; MacDonald et al., 1988a). As indicated in Fig. 5, susceptibility to suicide induction does not end precisely when a cell matures beyond the CD4+8+state. Cells with certain TCR specificities can be eliminated even from the population with superficially mature phenotypes (MacDonald and Lees, 1990; Nieto et al., 1990). We will return to discuss this process in detail in Section V,A. The most unusual feature of the developmental pathway is the mechanism by which competent T cells are derived from cortical thymocytes. Maturation is the one developmental transition, other than death, that depends on a TCR-coreceptor-ligand interaction. None of the events leading to the production of cortical cells is affected by the presence or absence on the developing cell of TCR/CD3 complexes, and even TCR-SCID thymocytes in some animals acquire CD4 and CD8 (Shores et al., 1990; D. Chen and E. V. Rothenberg, unpublished results). In vivo and in fetal thymic organ cultures, administration of saturating amounts of blocking antibodies against the TCRs, against CD4 or CD8, or against the MHC ligands of these receptors has minimal gross effect on the cortical thymocyte population (Kruisbeek et al., 1985; McDuffie et al., 1986; Born et al., 1987; MaruSi6-GaleSiC et al., 1988; Ramsdell and Fowlkes, 1989).However,


123

these treatments all drastically affect the number and distribution of cells in the CD4+ and CD8’ mature subsets. Whether a cortical thymocyte becomes a CD4+ cell or a CD8+ cell appears to depend completely on whether it can form a higher avidity TCR-coreceptor interaction with a class I or with a class I1 self-MHC molecule (Robey et al., 1991; Teh et al., 1991) on the cortical thymic epithelium (Benoist and Mathis, 1989; Berg et at., 1989a; Bill and Palmer, 1989). Thus, positive selection is a singular process in which cell fate is determined in two senses: not only providing access to maturation instead of death, but also dictating the lineage along which the selected cells will mature. We will discuss the evidence bearing on a possible mechanism in Section V,B, and here merely describe the outcome. The visible signs of this process are (1)the increased surface density of TCRaPICD3 complexes, (2) the loss of either CD4 or CD8 from the “doublepositive” cortical thymocyte precursor, yielding “single-positive” cells with mutually exclusive expression of CD4 and CD8, resembling mature T cells, (3)the strongly increased expression of MHC class I molecules, which cortical thymocytes express at abnormally low levels, and (4) the gradual down-regulation of HSA expression. Upregulation of TCR/CD3 expression on the surface is the earliest of these changes (Blue et al., 1987b; Penit, 1990; Shortman et al., 1991), and is probably due to a posttranslational stabilization ofthe TCR/CD3 complexes (Bonifacino et al., 1990) (see Section IV,A). Loss of one of the two coreceptors follows closely (Guidos et al., 1990).The result is a cell that resembles in most respects the conventional peripheral T cell described in Section I. Because T cells can only receive positive selection signals by interaction with the allelic MHC products present in the thymus, only “self-restricted” TCRs are selected. The nature of the interaction that leads to positive selection remains mysterious. Though there is an interaction with self-MHC, it is not the same as recognition of selfMHC in a conventional sense, because cells with TCRs that are fully activated by self-MHC are condemned to negative selection. Instead, the resultant T cell repertoire is positively biased to favor restriction by self-MHC, i.e., to recognize foreign antigens presented by selfMHC. Due to the extreme polymorphism of MHC products, such cells are unlikely to cross-react with other MHC products in the same way. This is the basis of the term “positive selection,” because it creates a powerful bias in the emerging TCR repertoire (Zinkernagel et al., 1978), as was noted in Section II,B. Because the emerging T cells can recognize any of a wide spectrum of peptides in association with

124

ELLEN V. ROTHENBERG

self-MHC, it is not clear what peptide or peptides have participated in the process of positive selection. This and other intriguing aspects of the positive selection process will be discussed in Section V,B.

B. THETCRyG LINEAGES Some of the most provocative phylogenetic insights into the possible origin of T cells come from consideration of the minority class of T cells that utilize non-ap TCRs, namely, the TCRyG lineages. In contrast to TCRaP cells, which generate highly diverse TCRs through combinatorial V(D)J joining events, TCRyG cells utilize a highly restricted set of gene segments for their rearrangements. Junctional mutagenesis, which simply adds to the diversity in ap TCRs, is the major source of potential diversity of yG TCRs, and in some yG cell lineages even this mechanism appears to go unused. Furthermore, there is a separate C, exon cluster for each 1-4 V, genes in the mouse instead of a common C region for all. Each C region possesses its own dedicated J, segment, so that the J segment variability is no greater than that of the C regions. In a sense, this blurs the distinction between “constant” and “variable” portions of the TCRy chain. As we shall see, some TcRyG cell types appear to have TCRs that are completely invariant. The ontogeny and function of TCRy6 cells have drawn the interest of many groups, and several excellent reviews have recently treated these cells in detail (Raulet, 1989; Ferrick e t al., 1991;Tonegawa et al., 1991; Raulet et al., 1991; Allison and Havran, 1991). Here we will only highlight some of the key developmental issues raised by these cells without attempting to be comprehensive. For orientation, Fig. 6 diagrams the organization of the TCRy and TCRG genes in the mouse, the organism in which most of the experimental manipulation of TCRy8 cell developmental pathways has been carried out. For conformity to the majority of the reviews just cited, we will use the nomenclature of Garman et al. (1986) in the text. The legend of Fig. 6 provides the corresponding gene segment designations according to the nomenclature of Heilig and Tonegawa (1986). The whole TCRG locus is nested within the TCRa locus, such that any VaJa rearrangement event deletes the whole D-J-C 6 cluster. Thus, TCRG genes can only b e expressed in cells that retain at least one TCRa locus in unrearranged configuration. The TCRy locus in the mouse is actually three expressible loci,’ any of which could in principle be used to contribute y chains to a TCR complex. There is strikingly little combinatorial diversity A fourth y locus, found in some mouse strains, is a pseudogene and will not be discussed here. It is indicated in Fig. 6 as the V,J,C,S cluster.


"JCp"'

125

"Vp'" "V"2+"

"vr5+" w

\r

Sequences joined by DNA rearrangement

Sequences joined by RNA splicing

FIG.6. Organization of TCRy and TCRG genes in the mouse. The figure is adapted from Raulet (1989) with additional data from B. Vernooy and L. Hood (personal communication). The maps are not drawn to scale. Directions of transcription (after rearrangement) are indicated by arrows. Under the diagram ofthe TCRy gene cluster are indicated the segments used in specific transcripts whose products are referred to in the text. The nomenclature given in the text and the figure is that of Garman et ctl. (1986). In the nomenclature of Heilig and Tonegawa (1986),the designations of J, and C, segments are the same, but the V, segments are numbered individually as follows: V,1.1 in the figure is numbered V y l ; V,1.2 in the figure is numbered Vy2;V,3 in the figure is numberedV,5; Vy4 in the figure (reproductive epithelial lineage) is numbered Vy6; V$ in the figure is numbered Vy4; and V,5 in the figure (intestinal intraepithelial lineage, extrathymic) is numbered V,7. Both nomenclatures are used extensively in the literature, but we have chosen one arbitrarily for the text to simplify the presentation.

possible in TCRy chains, as each C, is associated with only a single J, segment. The lack of diversity gives more prominence to directed gene rearrangement mechanisms in controlling the TCRyG repertoire than in controlling the TCRaP repertoire. In the mouse, because the TCRy genes are clustered in separate subloci, the allelic exclusion problem for TCRy is also a locus exclusion problem. Prior rearrangement of any expressible TCRy locus does not block subsequent TCRaP rearrangement (see below), but does block other y rearrangements. The exception that proves the rule is the JC,2 locus, which is apparently ex-

126

ELLEN V. ROTHENBERG

pressed promiscuously at the RNA level in both TCRyd and TCRaP cells, because for unknown reasons it does not contribute to cell surface receptor expression. The other two loci, JC,1 and JC,4, are stringently regulated in their rearrangements with respect to each other. Furthermore, as we shall see, there is tight developmental regulation even of the particular V segment used by the JC,l locus, the only TCRy locus with more than one V segment choice. TCRyG cells are present in chickens as well as in a variety of mammalian species besides mice (Sowder et al., 1988; Triebel and Hercend, 1989; Haas et al., 1990; Hein and Mackay, 1991).As will be discussed below, there are surprisingly marked differences among all these organisms in the utilization and tissue distribution of their TCRyG cells. Unfortunately, the organization of the TCRy and TCRG genes has not yet been determined in chickens or ruminants, animals that develop and utilize TCRyG cells in ways distinct from either mice or humans. It is interesting, however, to note that the organization of the TCRy genes in the human differs in several ways from that in mice. There are two C, segments in the human, but they are organized in tandem, like Cp segments, both oriented so as to be able to service the same, rather large array of V, genes. Furthermore, each C, in the human is also associated with two or three J, segments, permitting considerably more combinatorial diversity than in the mouse. The overall organization of the TCRG locus is similar in mouse and human. We will return at the end of this section to consider the possible relationships between ontogeny, tissue assignment, lineage, and gene organization in the TCRyG class. The cell transfer experiments that have been so instrumental in clarifying precursor-product relationships among TCRaP precursors have rarely been utilized for TCRyG progeny cells. As a rule, TCRy8 cells do not acquire CD4 or CD8 in the course of their maturation. Thus, whereas CD4 or CD8 is sometimes expressed on these cells (Goodman and LeFrancois, 1988; Cron et al., 1989; Itohara et al., 1989), a CD4-8- phenotype is not indicative of immaturity. However, CD4'" immature precursors can give rise to both TCRaP and TCRyd cells after intrathymic transfer (Wu et aZ., 1991b). Some cells with yd TCRs can be found in CD4-8- populations that otherwise have an immature phenotype: i.e., they are HSA++ and fail to express the T cell marker, CD5, which is normally abundant on TCR+ cells. On the other hand, IL-2Ra' thymocytes and the rapidly proliferating blasts derived from them express no TCRs, neither afi nor yG TCR complexes (Miescher et al., 1988; Boyer et al., 1989).Thus, early expression of TCRyG may prevent thymocytes from entering the IL-2Raf stage.


127

Alternatively, expression of TCRy6 may not occur until after loss of IL-2Ra. The presence of surface TCRy6 might then abort or block entry of the cells into the phase of rapid cycling that would normally culminate with acquisition of CD4 and CD8. In contrast to the rarity of y6 subsets in the adult rodent, in prenatal life yS lineages are much more prevalent. In fact, the first wave of surface TCR/CD3+ cells that appears in the mouse fetus (at day 14) or in the developing chick (at day 12) is entirely T C R y P . In both organisms, the first cell surface ap TCRs are not expressed until 2 days later (Bucy et al., 1990; Allison and Havran, 1991). The nature of the first TCRy6 cells appearing in the mouse fetus is indicative of the highly stringent regulation of y6 gene rearrangement in development, for all ofthe cells in this first wave use the same V, rearrangement (Vy3-JC,1) and the same V,DaJa rearrangement (ValDa2J82Ca). There is minimal diversity in these receptors. At this stage of fetal life, little if any terminal transferase is expressed (Gregoire et al., 1979; Rothenberg and Triglia, 1983), so the absence of N regions is not surprising; but even the recombinase breakpoints are found to be identical in multiple independently isolated clones (reviewed in Allison and Havran, 1991). These rearrangements are specific to fetal life, for cells with Vy3-JCyl rearrangements disappear from the thymus before birth and are never detectable there again. This is due both to the uniqueness of the precursors in fetal life and to some characteristic of the fetal thymic microenvironment, for while these Vy3+ cells can be generated from fetal thymic stroma repopulated with fetal cells, neither adult bone marrow precursors nor adult stroma support their production (Ikuta et al., 1990).A second class of invariant TCRy6 marks a subsequent wave of fetal thymocytes, cells utilizing the identical ValDs2J&C, rearrangement but this time combining it with a V,4-JCY1 TCRy chain. As indicated in Table 11, the y6 cells that are found in the thymus later in gestation and after birth utilize different receptor rearrangements. These later waves of y6 cells also differ from the early waves in that they have extensive junctional diversity in their TCR gene rearrangements. Those that are present around the time of birth are also the only intrathymic TCRy6 cells that have been observed to express CD4+8+ and CD4’8- phenotypes (Itohara et al., 1989). Not only the order of appearance of cells using different TCRy8 rearrangements, but also their destinations in the body, appear highly defined in a way quite unlike the behavior of TCRab cells (Table 11). Tissue assignment is correlated with V, segment utilization even when the C, segment involved in the rearrangement is the same. Thus, the “Vy3/V~1”cells can be found after birth specifically targeted to

128

ELLEN V. ROTHENBERG

TABLE II PROPERTIES OF DIFFERENT CLASSES OF MURINE y8 T CELLS“ Diagnostic TCR Rearrangement Vy3-JC,1 V,4- JC, 1 V,2- JC, 1 V, 1.1- JC,4 VY-JCyl

Diversity

Target Tissue

No No Yes Yes Yes

Skin Vagina, uterus, tongue Lymphoid tissue Spleen Intestinal epithelium

Thymus Dependence Yes Yes Yes Yes?

NO

This simplified table is based on more detailed presentations in Allison and Havran (19911, Ferrick et 01. (1991), and Tonegawa et al. (1991).

skin, as dendritic epidermal cells (Havran and Allison, 1988; Ito et al., 1989). However, “V,4/V81” cells, which utilize the same C, (and C,) segment, appear to localize to the female reproductive tract and the tongue. Cells using V,2- JC,1 TCRy rearrangements behave as conventional T cells. However, a particularly interesting case is that of cells utilizing the V,5 segment, which also rearranges to the same JC,1 sequence as V,2, V,3, and V,4. These cells have a homing preference for the intraepithelial region of the gut, where they are the major TCRy8 cell class. Though RNAs encoding V,5+ y chains can be found in the thymus from the earliest fetal stages, they also appear extrathymically before the first wave of TCR+ cells is exported from the thymus (Carding et al., 1990).Unlike certain other yS cells, those utilizing the V,5- JC,l receptor rearrangement can also be found in undiminished numbers in congenitally athymic nude mice. The implication is that for cells using this V, segment, unlike those using V,3 with the same JC,, it is not necessary to migrate to the thymus to undergo precursor expansion and TCR gene rearrangement. A striking confirmation of this interpretation comes from the ability of bone marrow cells to reconstitute the Vy5+ intestinal intraepithelial population almost as rapidly in thymectomized radiation chimeras as in sham-operated recipients (Bandeira et al., 1991).Thus, some extrathymic site, perhaps either the marrow or conceivably the gut (Guy-Grand et al., 1991), fulfills both the qualitative and quantitative requirements for the production of this cell subset. These differences in phenotype, regulation, and fate lead to serious questions of whether TCRy8 cells can be regarded as a homogeneous lineage. Yet in spite of their differences, their common utilization of TCRy8 receptors links these cells at a regulatory level. There appears to be strong allelic exclusion of TCRy gene rearrangement extending


129

across loci, so that a rearranged TCRy transgene will dominate the whole T C R y P population whether it involves a Vy2-JCyl rearrangement, a Vy3-JCyl rearrangement, or a Vyl.l-JCy4 rearrangement (Bonneville et al., 1989; Ferrick et al., 1989; Ishidaet al., 1990;Dent et al., 1990). Furthermore, various lineages of TCRy8 cells appear to home appropriately to their target tissues even when an inappropriate y6 receptor is imposed on them. Ferrick et al. (1989)report that expression of a Vyl.l-JCy4 transgene replaces that of endogenous Vy3-JCyl in dendritic epidermal cells, whereas Bonneville et al. (1990a) report that even the intestinal epithelial lymphocytes can use transgenic Vy2-JCyl+ or Vy3-JCy1+ TCRs. Thus, the ordered succession of homing specificities is correlated with but functionally independent of the ordered succession of TCRy gene rearrangements in the y6 cell lineage. Little is known about the maturation of TCRy6 cells beyond their acquisition of a receptor structure. Studies in the chicken, wherein TCRy8 cells represent a rather large minority of the circulating T cell population, indicate that TCRyi3 thymocytes make their transit through the thymic cortex and into the medulla more quickly than do TCRaP cells (Coltey et al., 1989; Bucy et al., 1990). This has led to the suggestion that TCRy6 cells do not undergo positive or negative selection. Because at most times they generally lack CD4 and CD8, and because CD4 and CD8 are strongly implicated in both of these selection processes for TCRaP cells (see Section V), this supposition appears reasonable. However, several independent reports using transgenic TCRs of known specificity indicate that TCRy6 cells do need to be positively selected and can be subject to negative selection (Itohara and Tonegawa, 1990; Lafaille et al., 1990; Dent et al., 1990; Bonneville et al., 1990b;Wells et al., 1991; Pereira et al., 1991). Ifthis is the case, it remains to be established whether the operative mechanisms are the same as in TCRaP cells [cf. differential effects of cyclosporin A in Kosugi et al. (1989); see Section V]. There are also indications that postthymic selection continues to shape the tissue-specific TCRy6 repertoire, at least in some antigenically complex tissue sites (Kyes et al., 1991).As discussed by Born et al. (1991), the target antigens of subsets of TCRy6 cells may correspond to heat-shock proteins and other “distress” antigens that are displayed in response to cell trauma. The Vy3-JCy1/Vs1Ds2J82Cs receptors on dendritic epidermal cells appear to recognize a keratinocyte-specific distress antigen (Havran et al., 1991). A controversial question is when and how the precursors of TCRy6 and TCRaP lineages diverge. Thus far, the evidence is indirect. In part

130

ELLEN V. ROTHENBERG

it is based on the rearrangement status of TCR genes in the inappropriate lineage. It is also based on the ability of rearranged TCR transgenes, either TCRaP or TCRyG, to block production of cells with rearrangements of their endogenous TCR genes by means of allelic (or locus) exclusion (Section 1,A). In this argument, if production of cells using a particular class of endogenous TCR is blocked, it is concluded that their precursors were within the lineage expressing the transgene at the time when endogenous rearrangements should take place. Conversely, if cells utilizing an endogenous TCR are found, it is concluded that those cells had already diverged from the lineage that would use the transgene. There are caveats to all these interpretations. First, even TCRa gene rearrangement in TCR@ transgenic mice clearly fails to show allelic exclusion, in agreement with reports showing that certain functional T cell clones from normal mice may have productive TCRa rearrangements on both chromosomes (Malissen et al., 1988; Kisielow et al., 1988).When development of cells with a particular type of TCR is found to be blocked, moreover, allelic exclusion may not be the only explanation. We have already noted some evidence that different T cell lineages can cooperate in conditioning the microenvironment for each other’s differentiation. Cells with certain TCRs might conversely have a deleterious effect. Thus, appearance of cells in the TCRaP or TCRyG lineage may be blocked not by TCR locus exclusion but by an indirect effect on thymic microenvironment. Accepting these caveats, the results suggest an early divergence of TCRaP and TCRyG lineages, even prior to the completion of either set of gene rearrangement events. Among the key pieces of evidence are the following. First, the excision products of TCRa gene rearrangement frequently are found to contain TCRG genes in germ-line configuration (Ohashi et al., 1990a;Winoto and Baltimore, 1989a).This is not invariable: that is, differentiation as a TCRaP cell does not appear to depend on prior suppression of TCRG rearrangement events (Ohashi et al., 1989; Takeshita et al., 1989; Thompson et al., 1990). However, rearrangement of the TCRG locus normally takes place before any other TCR locus, and TCRa rearrangement is normally last, so that the rarity of successive TCRG and TCRa rearrangements on the same chromosome is due to some mechanism other than temporal competition. Second, some though not all of the regulatory sequences controlling TCRy and TCRa transcriptional activity appear to have different effects on expression in TCRaP and TCRyG cell lines (Winoto and Baltimore, 1989b; Ishida et d.,1990; Spencer et al., 1991). Thus, lineage-specific transcription factors may help control mRNA expression or even rearrangement. Third, transgenic mice with rearranged


131

TCRy and TCRG genes often, if not always, exhibit continued development of TCRaP cells, although most or all TCRyG cells in the transgenics utilize the transgenic receptors (Ferrick et al., 1989; Ishida et al., 1990; Bonneville et al., 1989; Dent et al., 1990).This suggests that the cells have already segregated into two types prior to TCR gene rearrangement, in one of which, the TCRyG cells, allelic exclusion blocks further rearrangement and in the other of which, the TCRaP cells, it does not. There is evidence, however, for the existence of stages at which the two lineages are fused. Rearranged TCRaP transgenes, which lead to precocious TCRaP expression, can apparently block TCRyG gene rearrangements (Fenton et al., 1988; von Boehmer et al., 1988). Close inspection of transgenic phenotypes suggests that the precise timing of the first TCR expression in the fetus may play a role in dictating which lineages are established. In another TCRaP transgenic model, Ohashi et al. (1990a) found most TCRyG cells to be absent, but the earlyappearing V,3+VSl+ cells to be present at almost normal levels. There are reported cases of poor TCRaP development in the presence of TCRy8 transgenes, but these may require further analysis. As most TCRy8 target antigens are undefined, it is difficult to exclude the possibility that cells with some transgenic yG TCRs react against some intrathymic structure to block development indirectly. In any case, while both early- and late-appearing lineages of TCRyG cells can be engineered to utilize each other’s characteristic TCRy rearrangements, they may diverge from the precursors of TCRaP cells at different developmental stages. Comparison of the TCRyG populations in different species reveals surprisingly variable systemwide utilization of these cells even within a narrow phylogenetic range. A forceful indication that TCRyG cells might have an essential physiological role was provided by the example of the chicken, wherein TCRyG cells constitute about 25% of all circulating T cells, instead of 1-5% as in humans and mice (Sowder et al., 1988). However, this does not represent a simple expansion of the TCRy8 role in chickens relative to mice, for the dendritic epidermal population that is dominated by TCRyG cells in mice is absent in chickens (Bucy et al., 1988).As a rule, TCRyG cells are found in the gut epithelium in all species, but their representation in the skin and peripheral blood appears to be highly variable and independently regulated. Furthermore, the prevalence of TCRyG cells in skin and in the circulation differs just as markedly among different mammalian species as it does between mice and chickens. Humans not only have rather few TCRyG cells in peripheral blood, but also lack any spe-

132

ELLEN V. ROTHENBERG

cialized TCRyG cell type in skin (Bucy et al., 1989; Groh et al., 1989). Conversely, ruminants such as sheep and cattle have abundant TCRyG cells in skin as well as a high representation in the peripheral blood (Hein and Mackay, 1991). This diversity begs the question of what developmental mechanism is used to regulate the production of different TCRy8 lineages that are targeted to different sites. Unfortunately, there is not enough information yet to link this either with specific TCRy rearrangements or with order of appearance in ontogeny. In the human, for example, it is still equivocal whether there is a first wave of fetal thymocytes that is dominated by TCRyS cells, as in the mouse and chicken (Haynes et al., 1988; Campana e t al., 1989). Human TCRy6 cells can be divided into subsets on the basis of the gene segments used for rearrangement (primarily VJ versus Vs2), but these appear to correlate better with postthymic expansion than with order of intrathymic differentiation (Parker et al., 1990). In animals other than humans and mice, the structures of the TCRy and TCRG loci are not yet known. Therefore, it is still uncertain whether the prominent tissue populations that are subject to such pronounced phylogenetic variation are each defined by distinctive TCR rearrangements, or not. Furthermore, without such rearrangements to distinguish the cohorts of TCRyG cells that arise at different times, it would be difficult to prove, e.g., whether or not the epidermal population in the sheep arises early in ontogeny, and thus whether or not it bears any developmental relationship to the V,3+ dendritic epidermal cells in the mouse. Nevertheless, though the nature of the underlying mechanisms is still unresolved, the existence of so much phylogenetic variation is an important result. All the properties of the various TCRyG populations appear to be readily variable from species to species: the balance of thymus-dependent versus thymus-independent lineages, the tissues that are targets for homing, the timing of development, the extent of combinatorial diversity, and the extent of junctional diversity. This suggests that the designation of TCRy and TCRS genes for rearrangement, and subordination to TCRyG allelic exclusion, may be properties that can be induced rather easily in T cell precursors in any of a variety of physiological states. The relationship between TCRyG and TCRaP cells also has fascinating evolutionary overtones. Whereas some of the murine TCRyG sublineages are clearly T cells in their properties, other sublineageseven rearranging the same JC, gene cluster-appear almost not to be T cells at all. In some sublineages they mature to develop effector function without actually generating diversity in the receptor repertoire. In other lineages, they violate the general rule of T cell dependence on


133

the thymus for development. The precisely regulated rearrangement of the y and 6 genes appears to be part of a developmentally rigid scheme to generate effector cells with generically useful receptor types for each of a small number of vulnerable body surfaces. One might compare the range of yG recognition specificities in certain tissues with the recognition specificities of NK cells, or even of macrophages. Even the cytolytic activity reported for yG cells and their occasional expression of CD8 (Goodman and LeFranqois, 1988; Cron et al., 1989; Saito et al., 1990; Guy-Grand et al., 1991) link them as closely to NK cells as to the extrathymic lineage of CD8+ TCRaP cells. The functions of yG cells have not been analyzed in nearly sufficient detail to compare them with aP cells, but it would be very instructive to do so. Less diversity of receptors, coupled with highly tissuespecific homing, should result in a far higher concentration of the appropriate yG receptor type in its target site than of any given aP TCR in a conventional lymph node. Thus, for TCRyG cells antigen recognition may not need to trigger clonal amplification to as great an extent as in aP cells; activation of key effector functions, such as killing, may be largely sufficient for a high-impact response. By this line of reasoning, it would be valuable to determine whether TCRyG cells as a rule segregate a specialized IL-2-producing subset. If not, the elaborate regulatory circuits that restrict inappropriate TCRaP cell activation may not need to be used by TCRyG cells at all. C. ONTOGENY OF MAJORTHYMIC LINEAGES After this review of the major lineages produced by the thymus, we can now turn to the ontogeny of thymic function in the developing animal. Here again most of the work available is drawn from the mouse, with valuable comparative insights provided by the frog, chicken, and human systems. We will review the stages in the fetal mouse initially. Immediately after the epithelial anlage of the murine thymus is formed, it is invaded by large hematopoietic blast cells. The blasts resemble the most immature cells in the postnatal thymus, with low Thy-1 and high expression of CD44 (Husmann et al., 1988). These hematopoietic cells begin to proliferate exponentially, expanding in number about 1000-fold in the 8-9 days between the time of their entry and birth. Germ-line transcripts of the TCRP and TCRy loci can be detected as early as day 12, before the cells express any T cell surface markers (Pardoll et al., 1987a). It is likely that the TCRaP and TCRyG lineage precursors separate by days 12-13, for the subsequent schedules of rearrangement events and functional differentiation di-

134

ELLEN V. ROTHENBERG

verge markedly in the two lineages. By days 13-14, the fetal thymus harbors T cells engaged in at least two separate developmental programs. The most rapid of these programs is the one that gives rise to Vy3+Vsl+cells. These cells proliferate, undergo their y and 6 gene rearrangements, but never acquire CD4 or CD8. Their maturation to full TCR expression is complete by day 15, when they briefly account for essentially all the TCR/CD3+ cells. They leave the thymus soon afterward, dwindling to undetectable numbers by the time of birth (Havran and Allison, 1988).Some of the unique functional characteristics of fetal thymocytes, often assumed to be associated with immaturity, may actually be lineage-specific characteristics of the Vy3+ cells. For example, from days 13-15, IL-2Ra chains appear on a fraction of fetal thymocytes, reaching 80-90% before declining again to low levels by the time of birth. Over the same time course, IL-4 and IL-2 mRNA appears in a smaller, but still substantial fraction, of the thymocytes (Carding et al., 1989; Fowlkes and Pardoll, 1989; Sideras et al., 1988), and IL-2 protein is strongly detected in the day 15 fetal thymus (Waanders, 1991).This has led to a widespread assumption that IL-2 is a major growth factor for immature thymocytes in general. However, there are several clear differences between 14- to 15-day old fetal thymocytes and postnatal immature CD4-8- cells. The IL-2Rat cells in the fetal thymus are distinct phenotypically from those in the postnatal thymus because the former express CD44 at high levels and the latter do not (Husmann et al., 1988). In uitro, the fetal thymocytes grow well in response to IL-4 (Pelkonen et al., 1987) under conditions in which postnatal CD4-8- cells do not (Chen et al., 1989). In fact, IL-2 stimulation of postnatal and fetal thymocytes leads to a highly preferential expansion of Vy3+ cells relative to other TCR+ lineages (LeClercq et al., 1990). By contrast, IL-2Ra+ cells in the postnatal thymus respond poorly to IL-2 (Raulet, 1985; von Boehmer et al., 1985) and cannot internalize bound IL-2 (Lowenthal et al., 1986). Thus, whereas IL-2 and IL-4 may be growth factors for Vy3+cells in the fetal thymus, this is less likely to be due to the immaturity of the fetal thymocytes than it is to be due to a lineage-specific trait, or even to the precocious arrival of these cells at a mature, lymphokine-responsive state. We will discuss IL-2 responsiveness further in Section IV,D. The day 14 fetal thymus certainly contains the precursors of TCRaP cells as well, for thymic lobes explanted at this stage can be cultured in vitro, in the absence of further cell input, to give rise to abundant mature TCRaP cells (Mandel and Kennedy, 1978). However, TCRaP cell differentiation is considerably slower than that of the first cohort of


135

y6 cells. By days 15-16, the TCRp gene can be fully rearranged in these cells, but they only acquire CD8 and then CD4 by days 16-17. Transcripts of fully rearranged a genes are found beginning on day 17, and the first cells with mature CD4+8- or CD8+4- phenotypes emerge by days 18 and 19, respectively. These single-positive cells are mature by some criteria; the appearance of CD4'8- thymocytes is correlated with detectable in uitro helper activity (Ceredig et al., 1983b; Marrack et al., 1988a) and likewise the appearance of CD4-8+ cells is correlated with detectable in uitro killer activity (Ceredig et al., 1983b). However, these cells are not completely typical of postnatal TCRap lineage T cells. For one thing, the level of terminal transferase expression in the fetal thymus is considerably lower throughout gestation than it is in the postnatal thymus (Grkgoire et al., 1979; Janossy et al., 1980; Rothenberg and Triglia, 1983). Thus it is likely that the repertoire of specificities in these cells is less diversified by N-region addition than the repertoire of ap cells developing after birth. Furthermore, for unknown reasons, differentiation to a single-positive phenotype in fetal thymocytes is not correlated with up-regulation of surface TCR expression, in marked contrast to postnatal thymocytes (Roehm et al., 1984). Cross-linking of the TCR complexes on neonatal thymocytes leads to a poor Ca2+response characteristic of pure cortical thymocytes (to be discussed later), in spite of the presence of large minorities of CD4+ and CD8+ single-positive cells (Finkel et al., 1987, 1989a). Finally, the cohort of cells that matures around the time of birth is found to be proliferating after positive selection, in marked contrast to the corresponding set of thymocytes in later life (Ceredig, 1990; Lawetzky et al., 1992). This suggests that the mechanism of positive selection may in fact be slightly different in the fetal and perinatal thymus than in the postnatal thymus. The staging of the appearance of mature Vy3+ cells versus a/3 cells may be partially accomplished by the programmed time course of gene rearrangement events at different TCR loci. Different classes of TCR loci begin the rearrangement process at different stages, presumably in response to different environmental signals. Though the 6 rearrangement process involves both D-J and V-DJ rearrangements, there is evidence that rearrangements of this locus begin considerably earlier than any p or y rearrangement events, for example, within the fetal liver before the cells immigrate to the thymus (Chien et al., 1987; also see Bosma and Carroll, 1991). Next, the single V-J y rearrangement event appears to occur in the fetal thymus synchronously with the first of the two p rearrangements (Born et al., 1986).The result is that y6 cells are able to complete a full receptor before the a@cells can express

136

ELLEN V. ROTHENBERG

a single chain. Rearrangements of the a locus, both in fetal and in postnatal life, begin only after fi rearrangements are complete (Snodgrass et al., 1985a,b; Haars et al., 1986). Locus-specific regulation of rearrangement of V, other than V,3, including those rearranging to the same JC,1 locus as V,3, is developmentally ordered and temporally delayed as we have already described. Similarly, Vs segments other than Val are only expressed later in fetal development (Chien et at., 1987; Elliott et aE., 1988).We have already noted the multiple differences between T cells using different individual V, segments for their TCRs, and it is conceivable that there is an early divergence among the lineages of different y6 cell types, as well as between cup and y6 cells. This might be accomplished by mechanisms that could, for example, target different V, segments for rearrangement by germ-line transcription from lineage-specific promoters, depending on stimuli from a slowly maturing microenvironment (states 1-4, in Fig. 4). It is possible, however, that some of the diversity among T cell lineages is also a result of their derivation from different precursor populations altogether (precursors A-C, in Fig. 4). In fact, there is evidence from a variety of animals that lymphoid precursors enter the thymus not once, but in waves during gestation. Thus, postnatal T lineage cells could as a rule be derived from different precursors than the lineages that dominate the thymus in fetal life. The best evidence for waves of immigration comes from animals other than mice, particularly avian embryos. The chick thymus becomes populated by the first lymphocytes at a very early stage in the 20-day incubation period, i.e., at 5 days. By making reciprocal chickquail chimeras with donor thymuses and recipients of different stages, LeDouarin and co-workers have elegantly demonstrated that precursors enter the thymus in three separate waves before hatching (Jotereau and LeDouarin, 1982; Coltey et al., 1987). Each cohort of precursors enters and remains quiescent during the efflorescence of the cells from a previous wave, then becomes active as the earlier cell types emigrate or die. This mechanism appears to be general, and not unique to the chicken, for a similar phenomenon occurs in the frog, where at least three waves are seen, the last one bridging metamorphosis (Turpen and Smith, 1989). Even in the much less tractable experimental system of the mouse, Jotereau et al. (1987)succeeded in showing that there are likely to be at least two waves or precursors, one arriving late in gestation and activated only after birth. These immigration waves are diagrammed in Fig. 7. The functional significance of the waves is still to be proved, although intriguing hints have already been cited. In the chicken and the

137


Felt. A I 4

I I0

16

A

22

28

22

I 28

B

1 4

10

16

t

Hch A

4

16

10

28

22

A

4

12

20

28

-

36 days

44

52

60

melamorph

Period of stem cell entry

FIG. 7. Waves of precursors populating the thymus in development. The figure summarizes data for four different species, indicating the times of stem cell entry and the delays usually seen after immigration before the expansion of each set of progeny. The x axis denotes time after fertilization, in days. Note the change of scale for Xenopus. Hch, Hatching; metamorph, metamorphosis. The y axis impressionistically depicts the fraction of thymocytes at a given time that are derived from a particular set ofprecursors. The curves plotted are arbitrarily offset from the baseline to allow the connection between each set of stem cells and the progeny derived from them to be visualized. In fact, progeny of wave B cells are essentially undetectable during the period of dominance of wave A progeny, and the progeny of wave C cells are likewise eclipsed during the dominance ofwave B progeny. Data for each curve are from the following references: Chicken-Jotereau and LeDouarin (1982), Coltey et al. (1987), and LeDouarin (1991); quail-Jotereau and LeDouarin (1982); mouse-Jotereau et al. (1987); XenopusTurpen and Smith (1989).

mouse, yS cells predominate in the first cohort of developing cells, but never do so again (Bucy et al., 1990; Pardoll et al., 1987b). It is possible, therefore, that the unknown features that favor early commitment to a y8 lineage are preferentially associated with the initial-wave precursors. We have already cited the evidence that this is the case in the

138

ELLEN V. ROTHENBERG

mouse for the V,3 lineage of yS cells (Ikuta et al., 1990). It will be of interest to determine for the chicken and the human whether the earliest appearing TCRs are also, as a rule, invariant ones. It is not yet clear whether any properties of TCRaP cells in mice distinguish those derived from fetal precursors and those deriving from postnatal precursors, One possibility could be that fetal and perinatal precursors differ in their intrinsic ability to express terminal transferase. If this is not simply due to environmental conditions, then the source of the precursors could have a significant effect not so much on individual T cells but on the diversity of the fetal versus postnatal T cell repertoire. In other regards, the later waves of precursors may not have any additional developmental capabilities, but perhaps only a progressive narrowing of developmental potential to favor the prevalent TCRaP differentiative fate. D. ONTOGENY OF MINORLINEAGES There is one more type of thymocyte with a controversial relationship to the major lineages. Though their significance is presently arguable, these cells may later be found to be important products of the development and/or selection process. Within the murine CD4-8- population is a variable percentage of cells with cell surface TCRaP expression (Fowlkes et al., 1987; Miescher et al., 1988; Scollay et al., 1988). This unconventional cell type possesses a series of mature characteristics: lack of expression of HSA and high expression of CD5 (Fowlkes et al., 1987; MacDonald et al., 1988d; Scollay et al., 1988; Takahama et al., 1991). These cells also express CD44, which is not only a marker of the earliest thymic immigrants but also a marker frequently up-regulated on peripheral memory T cells (Budd et al., 1987a,b). Certain mouse strains, such as CBA, accumulate these cells to much higher levels than others, such as C57BL/6 (Scollay et al., 1988). There are similar CD3'4-8- TCRaP cells in the human, where their relationship to conventional CD4+8and CD4-8+ cells has been debated. At least in the mouse, in which direct cell transfers are possible, these cells do not appear to be developmental intermediates. Most available data show that these cells are devoid of precursor activity (Scollay et al., 1988; Fowlkes et al., 1987) when assayed by direct intrathymic transfer into irradiated recipients. They do not appear in ontogeny until weeks after birth (Fowlkes et al., 1987; Takahama et al., 1991). Thus, it seems very likely that the TCRaP+ CD4-8- cells are either end products of a separate branch of the TCRaP lineage, or a population derived late from relatively mature cells.


139

A proposal that this population represents a sink of discarded autoreactive cells has been widely discussed, based on the study of TCRap-transgenic mice with autoreactive TCR specificities imposed on their T cells. Those of the transgenic cells that survive clonal deletions in these circumstances often down-regulate their surface levels of TCRs and coreceptors, yielding a TCRaP+ CD4-8- phenotype (Teh et al., 1989). In normal mice, however, when the TCR repertoire of TCRa/3+ CD4-8- cells is analyzed, there is no evidence for accumulation of autoreactive receptors (see Section V) (Takahama et al., 1991).Thus it is unlikely that the simple form of this hypothesis is true. Recently, another possible role has been proposed: namely, that these cells may represent a distinct tissue-homing lineage analogous to the different V, lineages of TCRyti cells. Some of the TCRa@+ CD4-8- cells express a distinctive marker, NK1.l, which is normally found on NK cells but not on T cells (Ballas and Rasmussen, 1990; Levitsky et al., 1991). This suggests a link between the natural TCRa@+CD4-8- cells and some peculiar members of the murine and human CD4-8- thymocyte populations, which acquire a collection of NK-like markers and activities along with CD3 expression upon growth in culture with IL-2 (Mingari et al., 1991; Ramsdell et al., 1988; Denning et al., 1991; Michon et al., 1988; Budd et al., 1986). In the mouse, TCRap+ cells with the NK1.l marker are found in the periphery, in the spleen at low levels, and in the bone marrow at high levels relative to conventional T cells (Levitsky et al., 1991), suggesting marrow-specific homing. The function of such cells in the bone marrow, however, remains unknown.

E. SUMMARY I have attempted to describe the major lineages into which developing T cells diverge, and to sketch the main phenotypically defined transitions with the TCRaP lineage. This section has provided phenomenology: however, as I have tried to indicate, a careful consideration of the phenomena provides telling hints as to possible underlying mechanisms. In the next section, I will describe how the functional reactivity of the cells is observed to change throughout differentiation. IV. Functional Maturation of Thymocytes

A. MATURE VERSUS IMMATURE CHARACTERISTICS-HISTORY Long before diverse markers were available to dissect thymocyte developmental stages, it was noted that thymocytes are poorly responsive in immunological assays when compared to T cells from periph-

140

ELLEN V. ROTHENBERG

era1 sites, such as spleen and lymph node. Workers in the 1970s found that a prime contributing factor to the weak performance of thymocytes was that they were functionally heterogeneous, with a substantial majority of inert cells. The responsive fraction could be enriched in mice by in vivo glucocorticoid treatment, which shrank the thymus to a few percent of its normal cellularity and left only functionally competent cells alive (Blomgren and Anderson, 1971; Droege et al., 1974; Boersma et al., 1979).Alternatively, less destructive methods could be used to recover both fractions, such as separating the cells on the basis of their binding to the galactose-specific lectin peanut agglutinin (Reisner et al., 1976).Through the work of Scollay, Shortman, Ceredig, and MacDonald and their respective co-workers, it became clear that essentially all cells in the major cortical thymocyte population are functionally inert. The responsive cells were fully accounted for within the CD4'8- and CD4-8+ populations (Ceredig et al., 1982, 1983b; Chen et al., 1983a,b). The assays used in these early studies measured responses to mitogenic lectins such as Con A or, more frequently, responses to allogeneic target cells. The thymocytes were scored for the ability to proliferate, to secrete IL-2, or to kill labeled targets in a cytoplasmic isotope release assay. Thus, each assay required that the cells fulfill multiple criteria of maturation in order to score: expression of a TCR/ CD3 complex, adequate signaling responses to TCR ( + coreceptor?) engagement, inducibility of appropriate genes, viability in culture (sometimes for as long as a week), and ability to carry out the desired effector activity. With better definition of the molecular basis of T cell responses and the nature of signaling mediators, it has become possible to dissect the components of functional maturity and define separately the stages at which they are acquired. The results have shown that different components of maturity are in fact acquired noncoordinately, and are subject to differential positive and negative regulation in intrathymic development. The following discussion is not historical, therefore, but rather organized according to individual criteria for responsiveness. Unfortunately, there is not yet suacient information on the TCRys lineages to define the times at which they normally become functionally responsive, so I will focus on TCRaP cells. B. TCR/CD3 COMPLEXES AND SIGNALING MEDIATORS 1. The TCR/CD3 Complex In the TCRaP lineage, as I have noted, cell surface expression of the TCR/CD3 complex is limited by the late rearrangement and expression of the TCRa locus. All other components, including CD35, are


141

expressed cytoplasmically or at the RNA level prior to this time (Takashi et d.,1991).There is some evidence that transgenic p chains can be expressed on the cell surface in immature TCR/CD3- CD4-8- thymocytes (von Boehmer et al., 1988),in a unique structure with neither C D 3 nor TCRa chains (Kishi et al., 1991).However, even if this occurs in normal animals, there is no evidence that this variant p structure could act as a triggering receptor. As we noted above, the first appearance of TCRaPICD3 surface complexes is closely correlated in time with the transition to a CD4+8+phenotype and/or the end of proliferation. In a normal mouse, the surface density of TCR/CD3 complexes on the majority of CD4+8+ blasts, presumably in their last round of proliferation, is extremely low (Guidos et al., 1989; Shortman et al., 1991; Penit, 1990). It is primarily the postmitotic CD4+8+ cells, i.e., small cortical thymocytes, that include a high frequency of surface TCRa/3+ cells and express the highest level of mature TCRa mRNA (Kinnon et al., 1986). Therefore it seems reasonable to assume that the TCR proper cannot influence development until the last round(s) of proliferation of cortical CD4+8+ blasts. The existence of the positive and negative selection processes illustrates that once expressed, the TCRICD3 complexes on cortical thymocytes are capable of transducing a signal. Thus the question is not whether cortical cells fail to use their surface TCRs, but rather why cortical cell responses to TCR engagement are different from those in peripheral T cells (Finkel et al., 1991).There are two levels at which responses could be qualitatively determined. One is the signaling level: the cortical cell complement of coreceptors, accessory molecules, and linked signal transducers may yield a combinatorial signal that alters the impact of TCR/CD3 triggering. The other is the nuclear response level: that the cortical cell complement of activationdependent gene regulatory proteins may act on different target genes than those of peripheral T cells. There are hints of effects at both levels, although we still do not know which of the multiple differences are crucial for explaining how cortical cells respond. Use of the TCR/CD3 complex by cortical thymocytes appears to be quantitatively constrained by the odd assembly and turnover kinetics of the complex in cells of this stage. Despite continuously rapid synthesis of all TCR/CD3 component proteins, small cortical thymocytes degrade an unusually large proportion of the newly synthesized chains before they reach the cell surface (Bonifacino et al., 1990).This results in a characteristically low surface density of TCR/CD3 (Roehm et al., 1984), in spite of levels of TCR mRNA that exceed those in “mature” thymocyte subsets (Kinnon et al., 1986; Maguire et al., 1990). The rapid degradation of the majority of complexes is reminiscent of

142

ELLEN V. ROTHENBERG

TCR/CD3 assembly kinetics in mutant T cells with low CD35 expression as described in Section 1,A. Essentially all cortical thymocytes do appear to express both CD35 and its splice variant CD3q, so that there is no qualitative lesion in this component of the signaling complex (Nakayama et al., 1989; Vivier et al., 1991a; Clayton et al., 1991; but see Finkel et al., 1991, for another view). During the transition to the “mature” phenotype, i.e., during positive selection, there is a marked increase in surface density of assembled TCR/CD3 complexes (Roehm et al., 1984; Kappler et al., 1987). The steady-state levels of TCR mRNAs do not increase at this time, but even decrease slightly (Kinnon et al., 1986; Maguire e t al., 1990; Clayton et al., 1991). Instead, the increased TCR expression is due to the posttranslational stabilization of TCR components to promote more efficient assembly (Bonifacino et al., 1990),which may be correlated with increased expression of CD35/q at the protein level (Clayton et al., 1991; Vivier et al., 1991a; Finkel et al., 1991). Though the fraction of newly synthesized TCRaP chains appearing in TCRICD3 surface complexes is probably lower in cortical cells than in “mature” cells, most data thus far suggest that the composition of the surface complexes that cortical cells manage to assemble is normal [for a dissenting view, see Finkel et al. (1991)l. This raises an interesting question with respect to the behavior of the TCR complexes on cortical thymocytes, which is different from that of mature T cells even at a proximal signaling level. The TCRs expressed by cortical thymocytes can be triggered to induce a calcium flux, at least in response to antibodies against TCRI CD3 components. However, in general, in both mouse and human, the magnitude of the calcium flux is substantially less in cortical thymocytes than in TCRhighcells (Weiss et al., 1987b; Havran et al., 1987; Finkel et al., 1987) due to a cortical cell defect in the import of extracellular Ca2+ (Finkel et al., 1987, 1989a). In principle, this might be due either to a TCR signaling defect (see Finkel et al., 1991)or to a possible deficiency of IP,-gated Ca2+ channels. However, Finkel et al. (1989a,b) have noted that TCR/CD3 signaling exhibits a kind of variability in cortical cells that is not evident in medullary cells. Antibodies against the TCRaP chains proper triggered Ca2+ fluxes in only 30-50% as many cortical cells as were triggered in the same population by antibodies against C D ~ Esuggesting , that the various TCRlCD3 components might not be uniformly integrated into a signaling unit in the remaining 50-70% of cortical thymocytes. The results show that murine cortical cells are quite heterogeneous in the coupling of their TCRaP chains to the CD3 signaling unit. Finkel et al. (1989b) argue that there are two subsets, on the grounds that the cortical thymocyte


143

population that does flux Ca2+ in response to TCRaP ligation can be selectively deleted in uitro (see Section V,A), leaving behind a reciprocal population that does not repsond to anti-TCRaO. It remains to be determined if these two populations are really discrete or if they are arbitrarily separated from a continuum, e.g., from the higher CD35 and lower CD35 ends of the smoothly unimodal distributions shown by Vivier et al. (1991a). The structural basis for the signaling heterogeneity in the population could reside in the TCR/CD3 complex (Punt et al., 1991; Kappes and Tonegawa, 1991; Vivier et al., 1991b), or in the response mediators that normally transduce TCR signals (Hengel et al., 1990). The phosphorylation status of CD35 in cortical thymocytes raises the possibility that cortical thymocytes may be undergoing too much signaling rather than too little. In freshly isolated CD4+8+ cells from either mice or humans, the CD35 is found constitutively phosphorylated on Tyr (Nakayama et al., 1989; Vivier et al., 1991a). Culture at 37°C in uitro results in specific dephosphorylation of these sites, but they can be quickly rephosphorylated if the cells are stimulated in uitro with anti-TCR/CD3 ligands (Nakayama et al., 1989). By contrast, cross-linking of the CD4 or CD8 coreceptors on these cells does not appear to result in CD35 phosphorylation (Veillette et al., 198913).This suggests that most CD35+ cortical thymocytes in uico have already received a TCR-mediated signal from the microenvironment, to which they are still, in some sense, responding. We have already discussed evidence that inappropriate TCR-mediated signaling can lead to various kinds of nonresponsiveness in peripheral T cells (Section I,E), and it is tempting to speculate that the condition of cortical thymocytes may share some common elements with such blocked states.

2 . Mediators f o r Immediate TCR Responses

The signaling mediators used by TCR/CD3 to activate T cells are present, in general, in a more gradual developmental distribution than the cell surface TCR/CD3 complex. The molecules thus far examined show quantitative differences among thymocyte subsets, but no evidence yet of a sharp discontinuity that would clearly explain the basis for TCR uncoupling in cortical cells. However, a property distinguishing 30% of cortical cells from the other 70% could appear as a mere quantitative effect on the bulk population level, so these signaling mediators deserve consideration. All TCR-dependent gene expression responses appear to require some form of PK-C, although, as we have noted, lymphokinedependent growth responses may not (Mills et al., 1988; Valge et al.,

144

ELLEN V. ROTHENBERG

1988). It has been clear for some time that cortical thymocytes must express at least some PK-C isoform because they respond to phorbol esters by modulating CD4 from the cell surface (Wang et al., 1987). In combination with Ca2+ ionophores, phorbol esters further induce the modulation of CD8 and the increased expression of class I MHC molecules on murine cortical thymocytes (Havran et al., 1987; Boyer and Rothenberg, 1988). However, over six different isoforms of PK-C are known (Kikkawa et aE., 1989; Ono et al., 1989), of which at least three are atypical. Isoforms 6 , E , and 5 do not absolutely depend on Ca2+ or phospholipid, and 5 even fails to be activated by phorbol esters. Recently, isoform-specific primers have been used in a polymerase chain reaction to compare the levels of PK-C mRNAs in peripheral T cells and thymocyte subsets (Freire-Moar et al., 1991; J. Ransom, personal communication). Though both mature T cells and cortical cells express PK-C a,p, E , and E , and neither expresses PK-Cy, mRNA encoding the 6 isoform is expressed constitutively in peripheral T cells but not in cortical thymocytes ( J . Ransom, personal communication). Thus, if there are specific substrates and/or triggering receptors that require the 6 isoform, they may specifically fail to participate in early cortical thymocyte responses to stimulation. Data are not yet available to resolve whether the other components of the PIP2 breakdown pathway, e.g., the PLC, phosphatidylinositol kinases, or even the IPS-gated Ca2+ channels, are represented by unusual isoforms or expressed at distinctive levels at different stages of thymocyte development. T cells also have voltage-dependent K+ channels, which are thought to control the amplitude of Ca2+ (Gray et al., 1987). These turn out to be expressed in a developmentally regulated way (McKinnon and Ceredig, 1986; Lewis and Cahalan, 1988a,b). CD4-8- thymocytes and cortical thymocytes both apparently express very high levels of “n”-type K+ channels, but resting CD4+8- mature cells express reduced levels. Mature CD4-8+ cells also express only low numbers of different types of K+ channels. High numbers of n-type channels comparable to those in cortical thymocytes can be expressed in both mature subsets, but only after 1-2 days of stimulation via the TCR. Thus, the level of n channels found in small cortical thymocytes suggests a postmitotic residue from their previous activated state. The large number of K+ channels on cortical thymocytes may also contribute to the low amplitude of their Ca2+ fluxes. Expression of Fyn, the leading present candidate for the CD35associated kinase, parallels TCR surface expression in the later stages of thymocyte development. Levels of f g n mRNA are considerably higher in CD4-8+ cells and CD4+8- cells than in CD4+8+ cells


145

(Cooke et al., 1991). It is less clear howfyn mRNA is regulated before TCR expression, for example, in TCR- CD4-8- cells. However, in transgenic SCID mice engineered to express a known TCR, active Fyn kinase is clearly found among the CD4-8- cells (Carrera et al., 1992a), which might indicate that Fyn can be expressed in normal mice during a state that is normally TCR-. Because normal CD4-8- cells express CD35 RNA considerably before cell surface TCR expression (Takashi et al., 1991), some Fyn might be associated with CD35 at this stage. At any rate, in normal mice Fyn activity is clearly discernible in the cortical CD4+8+population and is then further elevated in the mature CD4+8- and CD4-8+ populations. Perlmutter and colleagues have provided strong evidence that the low expression of this kinase in cortical cells is at least partially responsible for their poor Ca2+ fluxes upon TCR ligation. They have derived transgenic mouse lines expressing Fyn from a cortical thymocyte-specific promoter (i.e., the proximal Lck promoter; see below), and examined the resulting mice for the magnitude of Ca2+ fluxes inducible by anti-CD3 in cortical versus medullary thymocytes. In animals overexpressing Fyn activity, cortical thymocytes acquire the high Ca2+ flux response characteristic of medullary thymocytes (Cooke et al., 1991). Because the TCR/CD3 surface density on cortical cells is not increased in these animals, it suggests that cortical thymocytes normally have less Fyn per cell surface TCR/CD3 complex than is needed to saturate the signaling capacity ofthese receptors. A Fyn deficiency could then account for the TCR “uncoupling” seen in many or most cortical cells. We will return to the developmental consequences of these dramatic changes in signaling in Sections IV,E and V. Lck, the kinase that associates with CD4 and CD8, is not strictly coregulated with the coreceptor molecules. Instead, it is expressed earlier, in CD4-8- immature thymocytes, as we11 as in CD4+8+ cortical cells and “mature” medullary cells (Wildin et al., 1991; Carrera et al., 1992a). Perhaps related to its expression in immature thymocytes, this kinase is also prominently expressed in non-T, CD4-8- NK cells, but otherwise in few if any other cell types (Perlmutter et al., 1988).The Zck gene has two promoters that are differentially utilized in immature and mature thymocytes (Reynolds et al., 1990; Wildin et al., 1991). Each initiates a different 5’ exon, which is then spliced to a common second exon and the remainder of the transcript. Though delineation of these two promoters has been useful to engineer gene expression in transgenic mice, the alternative 5’ exons do not appear to include any of the coding sequence for the Lck protein. Thus, it appears that the Lck proteins of immature, cortical, and “mature” thymocytes are all of

146

ELLEN V. ROTHENBERG

identical sequence. Available data suggest that the overall level of lck transcripts and Lck protein are at least as high in the cortical cells as in the medullary cells (Veillette et al., 1989b). Thus, high expression of Lck with CD4 and CD8 may endow cortical thymocytes with a potential for unusually strong coreceptor signals. Of all these signaling molecules, only Lck is clearly expressed in CD4-8- immature thymocytes, the precursors of cortical cells. Thus, Lck might function during early differentiation events that lead to the cortical cell state, as well as in the selection processes that determine how cells will emerge from it. Overexpression of Lck or expression of a nonregulatable Lck from its own proximal promoter has striking effects on thymocytes (Cooke et al., 1991; Abraham et al., 1991), blocking their development beyond the immature CD4-8- and CD4-8’ transitional blast cell stages. Thus, Lck can act in the absence of its “normal” partners CD4 and/or CD8, and strong Lck signaling appears to antagonize the final events required both for TCR surface expression and for normal progression to the CD4+8+ resting cell state. It is not clear yet whether these developmental effects of overexpressed Lck result from its partitioning to an inappropriate receptor and phosphorylating the wrong substrates, or whether they are simply consequences of excessive kinase activity. One developmentally “inappropriate” association site for Lck might be the IL-2RP chain, as we have noted above (Hatakeyama et al., 1991).At least in the rat and human systems, there is evidence that certain immature thymocytes express these IL-2 receptor chains (Toribio et al., 1989; Hunig and Mitnacht, 1991) and can respond to IL-2. In principle, Lck might be associated with any other receptor structure that was coexpressed with it in CD4-8- immature cells. However, note that the ability of Lck to associate with triggering receptors of both the stage 1 (coreceptor) and stage 2 (IL-2Rp) pathways (see Fig. 3) raises the possibility that competition for Lck could act as a switch between developmental or activation states. Thus, it is possible that the onset of CD4 and CD8 expression is needed to sequester Lck away from its previous receptor pathway, in order to arrest proliferation in the thymic cortex. C. RESPONSEGENES

I. Proliferation Proliferation is one of the major responses of mature peripheral T cells to their target antigens. As was described in Section I,E, many of the prominent response genes induced by TCR/CD3 stimulation encode the growth factors and receptors that act to drive this prolifera-


147

tion. The fact that in vivo many of the cells in the thymus are also proliferating has led to some confusion, over the years, between intrathymic proliferation and antigen responses. In fact, proliferation in the thymic cortex is antigen independent: it is analogous to the formation of colonies by other hematopoietic precursors as a part of their differentiation program. Both in the fetal and in the postnatal thymocyte lineages, appearance of cell surface TCR/CD3 complexes is coupled with the end of proliferation (Owen et al., 1986; Hunig et al., 1989; Shortman et al., 1991; Huesmann et al., 1991). Conversely, the very real responses of cortical thymocytes to antigen are nonproliferative. Contrary to early expectations, positive selection does not involve clonal expansion of thymocytes with appropriately self-restricted TCRs. Instead, the transformations of positive selection are generally postmitotic, with the initial changes appearing about 1 day after the last cell cycle and the appearance of the “mature” single-positive phenotype about 2-3 days later (Penit, 1986; Penit and Vasseur, 1988; Shortman et al., 1991; Egerton et al., 1990; Huesmann et al., 1991). Cortical cells can be transformed into single-positive cells even under conditions in which all DNA synthesis is pharmacologically blocked (Penit and Vasseur, 1988). Thus, we consider the proliferation in the thymus to be ontogenic, involved in differentiation but not in selection. We will consider selection mechanisms separately in Section V. Even if TCR-ligand engagement does not trigger thymocyte proliferation, there is still the possibility that thymocytes use the same growth factors and receptors as peripheral T cells, induced by an alternative receptor. This would indicate that all the components needed for the mature T cell proliferative response except the TCR were already acquired during thymocyte proliferation. Alternatively, if the thymocytes proliferate in response to a completely different stimulus, the inducibilities of IL-2 and IL-2R genes might be acquired in their own developmental progressions, parallel to the acquisition of TCRs. Two kinds of approaches have been used to address this question, which address separately the early phase of proliferation before IL-2Ra expression, and the terminal phase associated with TCR gene rearrangement and CD4 and CD8 expression. Because of the exponential nature of thymocyte growth, at steady state the overwhelming majority of poliferating thymocytes are in the second, terminal phase. The properties and requirements of the earlier proliferative phase have been inferred, therefore, from thymus repopulation kinetics after adoptive transfer, or from fetal thymic development in vivo or in organ culture, or from the manipulation of T cell-specific genes in transgenic

148

ELLEN V. ROTHENBERG

mice. Taken together, the results indicate that the mitogens for intrathymic proliferation are distinct from the cytokines produced by mature T cells. The ligand-receptor interaction that drives the terminal proliferative phase has not yet been identified, but it is clear that it is not IL-2/IL-2R binding. First, the IL-2Ra expression found in the mouse subcapsular region is not phylogenetically conserved: little if any IL-2Ra expression can be detected in the cortical proliferative zones of either humans or most strains of rats (TakBcs et al., 1985,1988; Hofman et al., 1985). Even in mice, the IL-2Ra expression in CD4-8- cells (Raulet, 1985; Ceredig et al., 1985) does not account for the rapid proliferation in the cortex. Lug0 et al. (1985) preparatively purified thymic lymphoblasts and showed that the great majority of the proliferating cortical blasts were completely devoid of IL-2Ra expression. Whereas this did not rule out the presence of lower affinity IL-2RP, subsequent reports have shown that even when IL-2 stimulates proliferation through IL-BRP, IL-2Ra is also generally induced (Le thi BichThuy et al., 1987; Ben Aribia et al., 1989; Jankovic et al., 1990). The levels of IL-2 in the thymic cortex of postnatal mice are also extremely low (McGuire and Rothenberg, 1987; Rothenberg et al., 1988, 1990b; Carding et al., 1989),although not completely absent. Waanders et al. have found small patches of staining with anti-IL-2 antibodies (Waanders, 1991),and J. A. Yang-Snyder and E. V. Rothenberg (unpublished data) have observed very rare, isolated thymocytes with IL-2 RNA in thymus sections. Very rare cells in the human thymus also appear to be expressing IL-2 RNA sequences spontaneously (Toribio et al., 1989). However, it seems unlikely that such rare reservoirs of IL-2 could act on a relatively weak receptor (IL-2Rp) to produce the confluent lymphoid proliferation seen throughout the subcapsular region of cortex. Perhaps the best case against an IL-2-dependent proliferation pathway is that Horak and co-workers have recently generated transgenic mice homozygous for an IL-2 gene disruption. These animals cannot make IL-2 at all, and yet they make cortical thymocytes (and other subsets) in normal numbers (Schorle et al., 1991). Thus IL-2 cannot be essential. When normal CD4+8+ cells, including most terminal blasts, are incubated in vitro with any of a variety of other recombinant-derived purified interleukins, they are also completely inert (Lug0 et al., 1986; Zlotnik et at., 1987; Murray et aE., 1989; Okazaki et al., 1989; Suda et al., 1990). Thus neither IL-2, IL-4, IL-7, nor any combination of these with other cytokines can drive them through another cell cycle. It seems most likely that direct contact with subcapsular thymic epithelial cells is needed to induce, sustain, and regulate this phase of proliferation.


149

There have been several indications that IL-2 might be involved in the earlier phase of proliferation, long before TCR rearrangement. IL-2RP mRNA is detectable in human CD3-4-8- cells and in murine fetal thymocytes throughout days 14-20 of gestation (Toribio et al., 1989; Montgomery and Dallman, 1991). Between days 14 and 15 of gestation in the mouse, 80-90% of the lymphoid cells in the thymus transiently express IL-2Ra, as described in Section III,C. These appear to participate in the formation of high-affinity IL-2-binding sites (Zufiiga-Pflucker et al., 1990a). A functional role for these IL-2R in development was strongly indicated by the reported ability of anti-IL2Ra antibody to block development of thymocytes, including TCRaP+ cells, in fetal thymic organ culture (Jenkinson et al., 1987). Tentori et al. (1988), Zuiiiga-Pflucker and Kruisbeek (1990), and Zuniga-Pflucker et al. (1990a) showed that in vivo treatment with anti-1L-2Ra could also retard both fetal thymocyte development and the reconstitution of adult thymic radiation chimeras, indicating a postnatal role as well. There are several problems with interpreting these results as evidence for an intrathymic role of IL-2, however. First, treatment of IL-2Ra+ cells with an anti-IL-2Ra antibody may have effects on the cells other than simply blocking their access to IL-2. Second, Rocha et al. (1988) have reported that nonlymphoid cells of the fetal thymic stroma can also express IL-2Ra and respond to IL-2. If the function of such cells is essential to establish a competent thymic microenvironment, then T cell development might be blocked pleiotropically by anti-IL-2Ra through an inhibition of the responses of the stromal cells. Deliberate addition of excess IL-2 to fetal thymic organ cultures does not provide a simple stimulus to TCRaP+ T cell development. Instead, it results in lymphoid-depleted lobes with thickened stroma (Skinner et al., 1987) and enrichment of early precursors, NK-like cells and Vy3+ TCRy6 cells (Plum and De Smedt, 1988; Plum et al., 1990; Waanders and Boyd, 1990; Leclercq et al., 1990), as we have noted before. The strongest argument against a necessary role for IL-2 in any phase of TCRaP T cell development is, again, the phenotype of the homozygous, IL-2-disrupted mice. The normal thymocyte numbers and distribution of cell surface phenotypes in these mice, at least with respect to TCRaPICD3, CD4, and CD8, indicate that IL-2 is most unlikely to be necessary or rate limiting for the proliferation of thymocytes at any state (Schorle et al., 1991). It will be very interesting when further analysis elucidates whether this conclusion really holds for all stages of development and all T cell lineages: for example, whether the IL-2-disrupted mice generate IL-2Ra+ developmental intermediates, or whether they have normal numbers of V,3+ dendritic epidermal cells.

150

ELLEN V. ROTHENBERG

Transgenic technology has not yet identified another growth factor of mature T cells that might be responsible for the early proliferation. Mice that overexpress IL-4 in the thymus (Lewis et al., 1991; Tepper et al., 1990) exhibit reduced production of mature T cells and a shift to CD8+ cells. Mice in which the IL-4 gene is deleted have no obvious defect in T cell development (Kuhn et al., 1991). One potentially interesting factor is IL-7, originally characterized as a pre-B cell growth factor, which can replace IL-2 in mature T cell stimulation, although it is not a product of T cells (Watson et al., 1989; Morrissey et al., 1989; Chazen et al., 1989; Welch et al., 1989). This cytokine is produced at high levels by thymic epithelium in vivo (Murray et al., 1989) and can maintain the viability, though not the proliferation, of thymocytes with precursor activity in vitro (Suda and Zlotnik, 1991). Surprisingly, transgenic mice that overexpress IL-7 reportedly have no obvious perturbation in their thymocyte numbers or distribution (Samaridis et al., 1991). It remains to be seen, however, whether disruption of the IL-7 gene will reveal a necessary, if not rate-determining, role. A separate issue from the question of which cytokines might drive proliferation in vivo is the question of which cytokines can be used to induce proliferation in vitro. As indicated in the discussion above, in vitro tests of cytokine activity on isolated thymocyte subsets are a vital tool to narrow the spectrum of factors that must be tested in vivo. However, hematopoietic cells routinely express receptors for a wider variety of cytokines than those in the immediate microenvironment. The ability of thymocyte subsets to respond to a given hormone in vitro provides valuable insights into their differentiative or proliferative potential, even when it does not prove that they were in the process of responding to that hormone in vivo at the time of their isolation. Cases in point are the TCRy8 and NK cells, which can be expanded dramatically in response to IL-2 (Budd et al., 1986; Phillips and Lanier, 1987; Denning et al., 1991;Mingari e t at., 1991;Groh et al., 1990;Plum et al., 1990); the “mature” TCRaP single-positive cells and TCR/CD3+ CD4-8- cells, which grow extensively in response to IL-7 (Okazaki et al., 1989; Watson et al., 1989; Groh et al., 1990; Suda et al., 1990);and the CD4-8+ single-positive cells, which are specifically and synergistically stimulated by a combination of IL-2 with IL-4 (Zlotnik et al., 1987; Carding and Bottomly, 1988).In the human system, a subset of CD3-4-8- cells has been reported to grow and differentiate all the way to mature TCRaP+ or CD8+ cells via culture with IL-2 (Toribio et al., 1988a,b).This is in accord with their reported expression of IL-2RP chains (Toribio et al., 1989). IL-4, when supplied with a costimulus of phorbol ester, is a potent growth factor for immature IL-2Ra+ CD4-8postnatal cells (Takei, 1988) and for early fetal thymocytes, which may


151

further differentiate into CTLs (Palacios et al., 1987; Pelkonen et al., 1987). Unfortunately, a full review of these intriguing approaches is beyond the scope of this work [for a current survey of results, see Carding et al. (1991)l. In general, it should be noted that all these proliferative treatments lead to further differentiation and loss of precursor activity (Suda and Zlotnik, 1991). Whether the cytokines are permissive or instructive for the differentiation paths taken is still unresolved.

2 . Response Gene Inducibility The question of when thymocytes first acquire responses characteristic of mature T cells can be posed more precisely by studying the inducibility of individual response genes, such as IL-2Ra, IL-2, IL-4, and other cytokine genes. All these genes share the property of being activatable by the cascade of biochemical events that follow TCR/ coreceptor ligation. As we have just described, the assembly and display of a fully competent TCR/CD3 complex is one of the latest events in thymocyte differentiation. Therefore, to analyze fairly the inducibility of response genes throughout thymocyte development, it is essential to bypass the TCR/CD3 complex pharmacologically. Once it was demonstrated that the TCR used Ca2+ and PK-C as signaling mediators, several laboratories accordingly began to dissect thymocyte responses using Ca2+ ionophores (A23187 or ionomycin) plus phorbol esters [phorbol myristate acetate (PMA) or phorbol dibutyrate (PDB)] as a proxy for TCR/CD3 stimulation (Rothenberg and Lugo, 1985; Palacios and von Boehmer, 1986; Lug0 et al., 1986; Ceredig, 1986; Howe et al., 1986; Ceredig et at., 1987; Boyer and Rothenberg, 1988; Howe and MacDonald, 1988; Rothenberg et al., l988,1990a,b; Fischer et al., 1991; Bendelac and Schwartz, 1991). The results of these experiments are summarized in Fig. 8 and are discussed in the following sections. a. Early Inducibility of Response Genes in Immature Thymocytes. First, it is clear that response genes become inducible in cells long before positive selection, and in fact prior to cell surface TCR/ CD3 expression. Inducible IL-2 producers are found in the IL-2Ra+ CD4-8- subset (Howe and MacDonald, 1988; Fischer et al., 1991; Rothenberg, 1992) and in the thymocytes of mutant SCID mice, which die during the terminal proliferation phase and rarely complete their TCR gene rearrangements (Rothenberg and Chen, 1992). There are some differences between the induction requirements of IL-2 in mature and immature cells: the immature cells rquire IL-1 as a costimulus with PMA + A23187 or PMA + ionomycin, whereas the mature cells do not, and their kinetics of IL-2 RNA induction are steeper and more

152

ELLEN V. ROTHENBERG

Immature

Peripheral (virgin)

"Mature" I

iFN-j Perf? TNFa IL2R

I

I

--

--

"S LOSS of responsiveness Y

Gfanzyrne B eta/. psrforin

Granzyme B sf a/. Derlorin

--

*iL-i dependent response development possible development Induced by growth In IL-2 POSSlbly induced by growth in IL-2

FIG.8. Response gene inducibility patters in developing thymocytes. This is a hypothetical scheme summarizing the data described in the text. The figure is adapted from Fig. 2 in Fischer et al. (1991), modified by the inclusion of data from the following references: Howe and MacDonald (1988), Boyer and Rothenberg (1988), Boyer et ul. (1989), Gajewski et al. (1989), Held et al. (1990), Rothenberg et al. (1990a,b), Bendelac and Schwartz (1991), A. Zlotnik, personal communication; and M. Dohadwala. R. J. Hill, and E. V. Rothenberg (unpublished). The ability of IL-2 to induce CTLs or NK-like killing activity is equated here with granzyme and perforin induction. An area of uncertainty in the published literature is whether the cytokine secretion responses observed in IL-2Ra- CD4-8- cells are actually effected by CD44+ precursors ofthe IL-2Ra+ cells (Le., the left-most population in the figure) or by the CD44- progeny ofthe IL-2Ra+ cells (third population from left), i.e., the cells that are actively engaged in the terminal proliferation program. From data in Boyer et ul. (1989), we favor the possibility that the less mature population is responsible, and this interpretation is supported by new data showing that only the CD44+ fraction of IL-Ra-CD4-8- cells makes IL-2 and IFN-7 (A. Zlotnik, personal communication). For discussion, see text. Note that this scheme only applies to postnatal T cell precursors, and that the phenotypes indicated represent inducibility of response genes, not actual expression of these response genes in uiuo.

transient than those of mature cells (Howe and MacDonald, 1988; Rothenberg et al., 1990a,b; Fischer et al., 1991). However, with these requirements met, the immature cells are fully competent to make IL-2. In thymocytes of SCID mice, as many as 20% of all viable cells can score as IL-1-dependent IL-2 producers, and overnight these cells can secrete amounts of IL-2 (>500 U/2 X lo6 cells) similar to those secreted by splenic T cells under optimal conditions (Rothenberg, 1992). Certain response genes used predominantly by CTLs may also


153

be inducible in immature cells, although the responding cells are not as well characterized in this case as in the case of IL-2. CD4-8- cells could be induced to express granzyme B but not perforin RNA during 18-20 hours of stimulation (M. Dohadwala, R. J. Hill, and E. V. Rothenberg, unpublished results). Some cells with this pattern of response gene inducibility may also be activated in uiuo, for Held et al. (1990) report that rare cells expressing granzyme B but not perforin occur naturally in the thymus. It is likely, though as yet unproven, that the cells responding to these short-term stimulation conditions include those that have the potential to develop fully into CTLs or NK cells during prolonged culture. These results suggest that programming for response gene inducibility occurs during the initial phase of intrathymic expansion, if it requires thymic influence at all. Response gene inducibility in immature cells is also significant because of its reflection on the mechanism leading to subset restrictions on mature T cell function. Because the inducibility of these genes is acquired before TCR/CD3 expression, it is acquired before there is a basis for positive selection to either a CD4’ or CD8’ lineage. This implies that response genes are already programmed for inducibility before it is determined whether they will be “appropriate” to the subset assignment of the developing cell. Does the thymus, then, produce omnifunctional T cells? Additional mechanisms can limit the expression of individual response genes in inappropriate mature T cell types, as we have discussed elsewhere (McGuire et al., 1988; Rothenberg et al., 1991). Critical to the resolution of this point would be a determination of whether individual immature cells that are IL-2 inducible are also inducible for CTL-associated genes (and IL-4 inducible). Unfortunately, due to the different inductive signal requirements and induction kinetics of these genes, this issue has remained difficult to resolve. b. Loss of Responsiveness in Cortical Cells. The second point that emerges from the studies summarized in Fig. 8 is that at some point between the immature, TCR/CD3- thymocyte stage and the CD4+8+ TCR’” cortical thymocyte stage, there is dramatic loss of functional responsiveness, blocking inducibility of IL-2 and, at least in mice, of IL-2Ra as well (Lug0 et al., 1986; Havran et al., 1987; Vives et al., 1987; Boyer and Rothenberg, 1988; McGuire and Rothenberg, 1987; Rothenberg et al., 1988, 1990b; Boyer et al., 1989; GonzalezFernandez et al., 1991;Yang et al., 1988a; Riegel et al., 1990; Cooke et al., 1991). Note that this must reflect an additional lesion beyond the nonresponsiveness to TCR/CD3 ligands discussed above, for in these studies Ca2+ ionophores and phorbol esters were used to bypass the

154

ELLEN V. ROTHENBERG

defects in TCR coupling to signaling pathways. In the transgenic mice described by Cooke et al. (1991), overexpression of a Fyn transgene could render cortical cells highly competent to flux Ca" in response to TCR ligands, but could not endow them with IL-2 inducibility. There are two ways to consider the sharp loss of responsiveness at this stage. One is that it may be a result of having failed selection, i.e., that it is an early functional manifestation of cells condemned to negative selection or death by default. The other is that it represents an adjustment of cell physiology to provide the preconditions for positive and negative selection. Whereas the first view has the virtue of simplicity, the second is supported better by a variety of circumstantial evidence. One kind of evidence is in the timing of the loss of responsiveness relative to TCR expression. Loss of responsiveness appears too early and persists too late to be explained by selection. Monitoring IL-2Ra inducibility of murine thymocytes, Boyer et al. (1989)found that this response is first shut off in the robust CD4-8- blasts ofthe late proliferative phase, prior to any detectable surface expression of TCRs. Fischer et al. (1991) similarly found that the IL-l-dependent IL-2 inducibility of IL-ZR+ CD4-8- immature cells has already been lost by their cycling CD4-8+ TCR/CD3- descendants (see Fig. 5) and recent data suggest that IL-2 inducibility is lost even earlier, along with IL-2Ra inducibility, at the CD44-IL-2Ra-CD4-8- blast cell stage (A. Zlotnik, personal communication). Moreoever, even after positive selection most murine thymocytes continue to appear functionally compromised. This is most pronounced in responses to TCR/CD3 ligands (Yang et al., 1988a; Pierres et al., 1990; Ramsdell et al., 1991), but it can also be discerned in the relatively poor expresion of IL-2 that these cells display upon stimulation with Ca2+ionophore and phorbol ester, as compared with peripheral splenic T cells, even though they are significantly better than cortical cells (McCuire and Rothenberg, 1987; Rothenberg et al., 1988,1990b; J.-F.Chang et al., 1991a,b). The continued functional depression of TCR/CD3h'gh medullary thymocytes even after selection is clearer upon comparison between the murine and human systems. Murine cortical thymocytes are exceptionally fragile (Hopper and Shortman, 1976) and tend to die upon stimulation (Rothenberg et al., 1988) (see Section V,A), as well as being incapable of inducing any response gene expression. This makes murine medullary thymocytes appear highly competent in comparison. By contrast, however, human and rat cortical thymocytes retain considerable viability in culture, and in many laboratories they have been found to make limited responses to stimulation at the level of gene


155

expression, e.g., induction of IL-2Ra although not IL-2 (Blue et al., 1986, 1987a; Pierres et al., 1990; Turka et al., 199la; Hiinig and Mitnacht, 1991). Yet, as we shall see below and in Section IV,D, positively selected medullary thymocytes in humans fall considerably short of mature peripheral T cells in responsiveness. In effect, this brings the response characteristics of human medullary thymocytes closer to those of human cortical thymocytes, stressing the similarity of their response handicaps as compared with peripheral T cells. These observations strongly suggest that some loss of responsiveness occurs as cells enter the selectable stage, and is apparently relieved only some time after positive selection. To explain the loss of conventional responsiveness as thymocytes enter the cortical stage we must postulate a reversible change in the signal transduction pathways that lead from Ca2+ and PK-C activation to the induction of genes such as IL-2 and IL-2Ra. The extensive work done to characterize the transcription factors for these two genes makes it possible to address the nature of the changes through their effects on the mobilization of specific DNA-binding proteins. These studies are just beginning, but they have already yielded several interesting results. Cortical thymocytes as a whole are deficient in the inducibility of NF-AT, the complex regulatory factor needed for induction of IL-2 expression (Riegel et al., 1990). By contrast, immature SCID thymocytes (IL-2Ra+ TCR- CD4-8- cells) can be induced to express an NF-AT-like DNA-binding factor (Chen and Rothenberg, 1992).Thus, NF-AT is one candidate for a discrete activity that is lost at the transition from the immature to the selectable state^. The failure to express NF-AT does not indicate whether this is a single lesion or a pleiotropic blockade of signaling, because the activity of the NF-AT factor depends on de novo synthesis and the cooperation of multiple components (Flanagan et al., 1991; Fiering et al., 1990). However, other results suggest that the cortical cell defect is rather specific. In preliminary work, we (D. Chen and E. V. R.) have found that cortical thymocytes as a whole respond readily to A23187 + PMA stimulation by induction of CD28RC and NF-KB-like DNA-binding activities (see Fig. 2). Normally these factors could be mobilized by signaling via CD28 and the TCR, respectively (Fraser et al., 1991; Jamieson et al., 1991).Thus, in principle, such factors might participate in the responses these cells do make to TCR ligands and APCs in their environment, including the response of negative selection. Interaction with other factors found in the thymus, such S p l and Jun-B, both of which are expressed at unusually high levels (Saffer et

156

ELLEN V. ROTHENBERG

al., 1991; Ryder and Nathans, 1988), would further be expected to modulate the choice of genes actually induced in activated cortical cells, as a result of the combinatorial nature of gene regulation. c. Emergence from Paralysis to Functional Maturity. Very little is known about the functional maturation processes that transform a fresh cohort of positively selected thymocytes with a “mature” phenotype into T cells that are truly mature, Ramsdell et al. (1991) have argued persuasively that only a discrete minority of the single-positive medullary thymocytes is functionally mature. This minority is distinguished by expression of a marker, Qa-2, which is also expressed on peripheral T cells but not on any of the immature or cortical thymocyte subsets (Vernachio et al., 1989). The question remains whether the transition from Qa-2- to Qa-2’, and by implication from functionally immature to mature, takes place within the thymus at all. An alternative possibility is that Qa-2+ cells are mature T cells that were activated in the periphery. Such cells are known to be able to recirculate to the thymic medulla (Naparstek et al., 1982).The kinetics of Qa-2+ cell appearance in ontogeny or during repopulation of radiation chimeras do not fully resolve this; positively selected thymocytes are known to require 10-20 days to complete their transit through the medulla, and the first emigrants to the periphery also appear during this time (Hirokawa et al., 1985). J.-F. Chang et al. (1991a) suggest that murine thymocytes acquire full IL-2 inducibility only in the periphery. In the human system, numerous reports indicate that virgin peripheral T cells may be just as poorly responsive as medullary thymocytes (Sanders et al., 1988; Byrne et al., 1988; Koulova et al., 1990). There is some circumstantial evidence, however, to suggest that conversion of CD4’8medullary cells to functionally competent helpers does depend specifically on intrathymic cell-cell contacts. Amagai et al. (1987) have noted that recovery of TH function in whole-body irradiated radiation chimeras is delayed for weeks after the appearance of CD4+8- cells, whereas in thymus-shielded radiation chimeras the TH activity, as measured per CD4+8- cell, is restored earlier. The implication is that there is a radiation-sensitive cell type that is rate limiting for CD4+8cell functional maturation, and that this cell type is also located in the thymus. Thus, at this time it seems most likely that full functional responsiveness is restored slowly to positively selected thymocytes during their long residence in the medulla. Nothing is yet known about the molecular basis of these changes, or about the ligand-receptor interactions that are critical to bring them about. It is likely nonetheless that this functional recovery process is one of the most interesting events in the thymus from a standpoint of developmental biology, for it


157

may well be this process that forges the linkage between TCR/ coreceptor specificity and assembly of the signaling pathways and transcription factor ensembles that enforce commitment to a particular functional lineage. D. AUXILIARY SIGNALING RECEFTORS Late developmental changes in the coupling of receptors to signaling pathways are brought into sharp focus by comparison of the effects of triggering via the TcR with those of triggering via different receptors. The two that have provoked the most interest are CD28, discussed in Section I above, and the alternative T cell activation receptor CD2. CD2 is a receptor that may replace the signals delivered through the TCR/CD3 complex; CD28 is a receptor that gives signals essential to complement the TCRICD3 complex. Both were defined in the human years before they were found in the mouse, and most of the data presently available are from the human system. 1 . CD2 In mature T cells, appropriate combinations of anti-CD2 antibodies trigger responses highly similar to those induced by TCR/CD3 ligands, even including activation of the same tyrosine phosphorylation pathways (Meuer et al., 1984; Bierer et al., 1989; Koretzky et al., 1991). Activation of these cells by anti-CD2 antibodies is, however, dependent on coexpression of the TCR/CD3 complex; in mature cell mutants lacking one or another TCR chain, or in cells from which TCR/CD3 complexes have selectively been down-modulated, CD2 ligands do not trigger (Bockenstedt et al., 1988). Thus, it seems likely that these receptors transiently cluster with TCR/CD3, with the CD3 components probably acting as signal transducers (Bierer et al., 1989; Altman et al., 1990). In the immature subsets of human thymocytes, however, CD2 has a different significance. Human thymocytes acquire surface expression of CD2 at a distinct stage before they first express TCR/CD3 complexes (Lanier et al., 1986; Haynes et al., 1989), and in these CD2+ TcR/CD3- cells, the CD2 molecule is a competent receptor for triggering by itself (Fox et al., 1985; de la Hera et al., 1989).The difference between mature and immature cells can be explained if CD2 and TCR/CD3 compete for a common limiting signal transduction mediator that is usually quantitatively sequestered by TCR/CD3. If the signaling molecule were, like CD35 or Lck, expressed earlier in development than the TCR/CD3 complex, then it could be available to service CD2 signaling without competition. In fact, in NK cells, which also express CD2 without TCR/CD3, anti-CD2 antibodies lead to

158

ELLEN V. ROTHENBERG

activation through association with CD35 (Vivier et d.,1991~). Alternatively, stage-specific signaling molecules in immature cells could confer signaling activity on CD2 at an early stage but then disappear in mature T cells. In any case, using CD2 as a triggering receptor, immature TCR-negative human thymocytes can not only flux Ca2+,but also turn on the expression of the response gene IL-2Ra. When IL-2Ra is induced by CD2 ligation of human thyniocytes, it confers full competence to proliferate in response to IL-2 (Fox et al., 1985; Ramarli et al., 1987). It is quite plausible that CD2 is engaged in vivo in the thymus, because thymic epithelial cells appear to bind human thymocytes through a mechanism involving CD2 and its ligand LFA-3 (Vollger et al., 1987). This binding has recently been implicated in the bidirectional thymocyte-epithelial inductive interactions discussed above (Le et al., 1990), and may help condition the thymic microenvironment by inducing epithelial cell functions. In principle, CD2 might well provide signals to the developing T cells. There has been great interest in the possibility that CD2-ligand interactions trigger IL-2Ra expression in the course of normal thyrnocyte development, as they can in uitro. However, the early expression of CD2 in T cell development does not appear to be phylogenetically conserved. CD2 mRNA as well as CD2 surface protein appear at a stage in fetal mouse development after the peak of IL-2Ra expression (Owen et aZ., 1988; Yagita et al., 1989b) (see Section 111). Thus CD2 is unlikely to be the receptor that triggers IL-2Ra expression in immature murine thymocytes in uiuo. Nor is there direct proof that the human immature cells activatable via CD2 are developmentally equivalent to murine IL-2Ra’ CD4-8- cells or their immediate precursors. Duplay et al. (1989) have used multiparameter flow cytometry to show that in murine thymocyte subsets, CD2 is only expressed on cells that also express TCR/CD3. Indeed, it is not found on any immature CD4-8- cells, including IL-2Ra’ cells, but is expressed on TCR+ CD4-8- cells and on all cortical and “mature” medullary thymocytes alike. Ironically, in the human thymus, where CD2 does appear to be expressed before TCRKD3, only a small minority of CD4-8- cells are IL-2Ra+ in uiuo (Aspinall et al., 1991). Thus, the case for CD2 as the main receptor triggering IL-2Ra expression is unsatisfying. CD2 has functional significance beyond the early phases of thymocyte differentiation, however, and it is possible that its later effects may be the ones that are better conserved. One of the first functional distinctions recognized between “mature” thymocytes and peripheral T cells in humans was that while anti-CD2 antibodies could trigger the


159

proliferation of both, anti-CD3, surprisingly, could not. The singlepositive thymocyte subset, although “mature” by surface phenotype and expressing high levels of TCR/CD3, responded well to anti-CD2 by IL-2 induction but responded only poorly to anti-CD3. Furthermore, the simultaneous binding of both CD2 and C D 3 by triggering antibodies enhances the activation of mature peripheral T cells, but yields poorer responses than anti-CD2 alone in the “mature” thymocyte subset (Ramarli et al., 1987). At the very least, this effect provides graphic evidence that the TCRICD3 complex does not simply fail to activate thymocytes, but actually may transduce a negative signal at certain stages. We will discuss such negative signals in Section V.

2. CD28 CD28 engagement does not lead to a duplication of TCR signals, but may be essential to convert the TCR-PLC pathway signal into an unequivocally activating one in mature T cells, as was described in Section 1,D. Thus, a lack of CD28 would be an attractive explanation for the nonactivation responses of cortical cells to TCR/CD3 triggering. High-density CD28 expression is confined to the TCRhigh,singlepositive medullary class of thymocytes in humans (Yang et d.,1988a; Turka et al., 1990).However, some CD28 is also detectable on cortical CD4+8+thymocytes, at least in the human (Turka et al., 1991b). Upon stimulation with phosphoinositide pathway agonists, the cells further up-regulate their levels of CD28 (Turka et al., 1990). A strong interpretation of CD28 function in this subset is handicapped by the asymmetry of information available for the mouse and human systems. As we have noted previously, the functional status of murine and human cortical thymocytes is not the same: in general, human CD4+8+ cells manifest more conventional T ceIl activation responses and are less likely to die in uitro than murine CD4+8+cells. Thus, the murine system offers a more dramatic example than does the human system of developmentally stage-specific responses to TCRI CD3 ligation, and potentially a better test case for differential CD28 involvement. However, virtually all published data on CD28 effects, as of this writing, are from the human system. It is therefore still possible that the essentially complete lack of activation-type responses in murine cortical cells will be correlated with a complete absence of CD28, but other interesting possibilities are discussed in Section V. Even in the human, CD28 expression is coupled to responsiveness in a developmentally stage-specific way. Though CD28 ligation alone is ineffective, anti-CD28 plus phorbol ester or the combination of anti-CD2 antibodies with anti-CD28 can induce IL-2 and IL-2Ra ex-

160

ELLEN V. ROTHENBERG

pression and proliferation much more potently than anti-CD2 antibodies alone (Pierres et al., 1988; Van Lier et at., 1988; Yang et al., 1988a). Yet in spite of the broad expression of CD2 in the human system and the possibility that CD28 is expressed in cortical CD4+8+ thymocytes or even earlier, this response is seen in only a minority of thymocytes. Immature human CD4-8- thymocytes, which can be enriched to high purity, are completely unable to proliferate in response to combined ligation of CD2 and CD28 (Yang et al., 1988a). Cortical thymocytes also fail to proliferate comparably to medullary TCR/CD3highcells under these conditions (Yang et al., 1988a; Pierres et al., 1990). Only the TCR/CD3highmedullary fraction can make the full proliferative response, involving anti-CD2 plus anti-CD28 induction of both IL-2 and IL-2Ra chains. It is noteworthy, however, that some TCR/CD3'" cells can respond to anti-CD28 plus anti-CD2 to the extent of becoming responsive to exogenous IL-2 (Pierres et d., 1990; cf. Ramarli et at., 1987). Thus at least some of these cells, presumably at a CD4+8+ stage, must express CD28 and be capable of responding to signals that it transduces. Thus, these studies suggest a developmental gradient in the responsiveness to triggering via CD2 and CD28. Ligands of CD2 alone can trigger some IL-2Ra expression, even in cells of immature or cortical type, but cannot trigger IL-2 production. Anti-CD2 plus anti-CD28 are significantly more potent in inducing IL-2Ra, but still cannot trigger IL-2 induction in CD4-8- or in cortical cells. In medullary cells, however, this combination of stimuli activates both response genes and drives proliferation efficiently. Thus by this stage both CD2 and CD28 are engaged with their signaling pathways, and the cells are fully competent to activate IL-2 and IL-2Ra expression. It is even more striking, therefore, that the same TCRhighcells fail to respond to the stimuli of anti-CD2 or phorbol ester when they are combined with anti-CD3. There is a specific uncoupling of TCR/CD3 from effective activation pathways in thymocytes, and an alternative coupling of TCR/CDS with dominant growth arrest pathways, that persist to some extent even after the cells have passed through positive selection and reacquired response gene inducibility.

E. POSTTHYMIC MATURATION Functional maturation, unlike TCR/CD3 assembly, continues after cells leave the thymus. The encounter with antigen triggers substantial changes in T cells, including both long-lasting alterations in surface phenotype and alterations in function (Akbar et al., 1991). The existence of these changes blurs the distinction between activation re-


161

sponses and differentiation responses. The arming of CTL precursors with lytic granules and other toxic components is only the most conspicuous form of this process; CD4+ helpers are also transformed. In general, the cells express more adhesion molecules after stimulation and become easier to trigger to proliferate (Sanders et al., 1988; Budd et al., 1987a,b; Springer, 1990). It is not clear whether the weak proliferative responses of virgin cells are due to a difficulty in achieving sufficient stimulation or to an excessive tendency to become anergized (see Section 1,E). The first round of stimulation also generates a clone of cells that can apparently express a wider range of cytokines than their clonal founder could express in its virgin state (e.g., Salmon et at., 1989). Subsequent rounds of stimulation, in an environment conditioned by both autocrine and paracrine cytokines, appears to select for ~ T Hsub~ cells with more restricted cytokine repertoires. The T H and sets described in Table I are likely to be end points of this process of induction coupled with selection. Several excellent reviews have recently summarized our current understanding of these processes (Mosmann and Coffman, 1989; Swain et al.,1988b, 1991; Gajewski et al., 1989; Vitetta et al., 1991). There remain some gaps in the linkage of intrathymic functional maturation events to the response phenotype of the presumed product, the virgin peripheral T cell. In the mouse, virgin peripheral T cells probably secrete only IL-2 upon Stimulation, yet CD4-8+ medullary thymocytes can also secrete IFN-.)Iand TNF-a, and CD4+8- thymocytes secrete all of these and IL-4 as well (Bendelac and Schwartz, 1991; Fischer et al., 1991). In these responses, medullary thymocytes appear more like activated memory/effector T cells than like virgin T cells. One possible resolution to this problem would be the interpretation that all the functionally responsive cells in the thymic medulla were, in fact, mature preactivated cells returning from the periphery, as described above. There are other resolutions, however. We need not consider the spectrum of lymphokines produced by a cell to represent a full disclosure of that cell’s functional potential, but only its response to a particular set of stimuli that impinge on it at a particular time. As discussed in Section I,D, IL-2 induction appears to require the synchronized integration of signals from a variety of pathways. By extension, the ability to express cytokines other than IL-2 may depend not on maturation per se but on the activated physiological context in which prestimulated cells frequently receive their antigenic challenge. The background states into which TCR-derived (or PMA A23187) stimuli must be integrated may therefore differ between medullary thymocytes and virgin peripheral T cells. Support for

+

162

ELLEN V. ROTHENBERG

this general view comes from recent work from several laboratories, which have reported that the presence or absence of particular cytokines at the time of initial stimulation can decisively influence whether virgin T cells will be observed to express IL-4 (Ben-Sasson et al., 1990; Daynes et aE., 199l;.LeGros et al., 1990).Elsewhere, we have argued that changes in signaling pathways, many of them reversible, may play a major role in determining the responses available to developing T cells in different lineages (Rothenberg et al., 1991). The pattern of acquisition and loss of function during intrathymic development vividly illustrates the dominant effect of physiology over developmental history in determining the responses of thymocyte subsets.

F. FUNCTIONAL MATURATION VERSUS COMMITMENT TO THE T CELLLINEAGE The data already summarized make it clear that acquisition of response gene inducibility is not, as a rule, temporally linked to TCR/ CD3 expression or selection. This raises the question of whether acquisition of T cell functional responses depends on commitment to the T cell lineage at all. The two interesting cases in point are mast cells and NK cells. As we have described above, both NK cells and mast cells share one feature with T cells, namely the use of CD35 or a close relative as a major component of the triggering receptor (Anderson et al., 1989; Lanier et al., 1989; Orloff et al., 1990).NK cells, at least, also clearly express Lck kinase (Perlmutter et al., 1988; Hatakeyama et al., 1991). It remains to be determined whether these triggering molecules can instructively specify the inducibility of what we think of as T cell-specific responses, for both NK cells and mast cells can express complex batteries of genes that precisely duplicate the batteries of response genes utilized by particular mature, memory T cell subsets. Some mast cells, upon stimulation, synthesize and secrete the same set of lymphokines that are distinctively expressed by T H cells: ~ IL-4, IL-5, IL-6, and 1L-10 (Plaut et al., 1989; Burd et al., 1989; Wodnar-Filipowicz et al., 1989; Moore et al., 1990).Also like T Hcells, ~ mast cells can use IL-4 as a growth factor and are inhibited in their growth by IFN-y (Brown et al., 1987; Takagi et at., 1990). Mast cells elaborate numerous cellular structures that distinguish them from T cells; they do not require the thymus for their differentiation; they also appear not to express IL-2. Nevertheless, their inducible response gene programming is fully convergent with that of a subset of postthymic, antigen-experienced T cells. At least, this suggests that the accessibility O f T Hresponse ~ genes to induction does not depend on having previously undergone commitment to the T cell lineage.


163

NK cells are an even more interesting parallel. Their profile of response genes, including granzymes and perforin, is indistinguishable from that of mature CD8+ CTLs (Tschopp and Nabholz, 1990). Their ability to grow and differentiate in response to IL-2 further links them to CTLs. In the human, they clearly express CD2, which is functionally coupled to CD35. By definition, NK cells do not rearrange their TCR or Ig genes, and as we have noted above they never express CD3y or CD36. They share with TCRa@+CD8+ cells and TCRy6+ cells the high constitutive level of their expression of IL-2RP chains (Hattori et al., 1990), and only differ in that their rapid IL-2Ra chain inducibility gives them greater sensitivity to IL-2 (Ben Aribia et al., 1989). The NK cell phenotype shows that CTL response genes, like T Hresponse ~ genes, are accessible to cells that have never entered the thymus. More than in the case of mast cells, NK cell differentiation suggests a fundamental uncoupling of response gene programming for inducibility from the events that regulate TCR expression and selection. Even CD3-4-8- cells within the thymus can respond to culture with IL-2 by differentiating into cells indistinguishable from NK cells, as we have seen. It is not certain yet whether the outgrowth of such NK cells really reflects the diversion of a pre-T cell from the conventional TCRaP lineage to an NK developmental fate. However, these experiments raise the possibility that under conditions of strong IL-2 stimulation, the CTL effector program can be accelerated independently of other differentiation events, to the extent that cells bypass TCR gene rearrangement altogether. The complex relationship between functional maturation and maturation of the recognition structures of T cells can be explained if the two processes are normally parallel but mechanistically independent. Thus, we can postulate that pre-T cells gain access to killing function inducibility or lymphokine production inducibility through developmental changes that can be shared with other hematopoietic lineages, even if these changes happen to occur in the thymus. Such events naturally have their own inductive triggers and restrictions, but by saying that these conditions are satisfied for a particular subset of thymocytes we need not also mean that they include a dependence on all the characteristics of the subset. For example, under certain conditions, SCID thymocytes can go on to turn on CD4 and CD8 expression in the absence of TCR rearrangements (Shores et al., 1990). I wish to propose that the gene expression subroutines involved in programming for function can be similarly uncoupled from the canonical sequence of TCR and coreceptor expression. In NK cells, as indeed in extrathymically derived CD8+ cells, they are uncoupled even from residence in the thymus.

164

ELLEN V. ROTHENBERG

Of course, TCR-linked changes in cell physiology can impinge on the exercise of functional responsiveness. This is how one can interpret the loss of function that appears to accompany the transition to the cortical thymocyte state. In part, these cells may preempt their own activation responses by superimposing upon them a coinducible suicide pathway, as I shall describe below (see Section V,A). In part, simple alterations in signal transduction mediators or transcription factors such as NF-AT can result in pleiotropic changes in the choice of gene batteries that are actually induced upon activation. It is interesting to speculate how these kinds of changes may participate in negative and positive selection. In both cases, however, I suggest that the cortical thymocyte paralysis simply obscures the prior programming of these cells for function. The unresolved question is whether, beneath the cover of nonresponsiveness, all thymocytes are programmed for multiple functions, or whether mutually exclusive classes of thymocytes are programmed for NK-like functions or for lymphokine-producing (mast cell-like?) functions. In the first case, CD4' and CD8+ cells would have to become matched with appropriate functions through a narrowing of competence, perhaps of modification of signaling pathways. In the second case, cells with inappropriate combinations of function and recognition specificity would have to be purged from the repertoire after positive selection. The second possibility seems unlikely, but at this time it cannot be excluded. Both the nature of the response block and the problematic matching of recognition and response type bring us to the question of how negative and positive selection may work.

V. Selection: Questions of Mechanism

Of enormous significance for the understanding of immunological tolerance has been the elucidation of the selection processes that shape the T cell repertoire. A breakthrough in this area was the discovery that the use of particular Vp segments in their TCRs strongly predisposes T cells to respond to certain tissue antigens or bacterial toxins (Matis, 1990; Herman et al., 1991). This specificity can be demonstrated irrespective of any other target-antigen specificity the T cell possesses. Thus, a cell can be functionally bispecific, with one of its specificities defined combinatorially by TCRa and TCRp V, J, and junctional regions, and the other specificity defined by the Vp region alone. Because the same Vp segment is shared by many other T cell


165

clones, the antigens and toxins that react with Vp alone are referred to as “superantigens” (White et al., 1989; Herman et al., 1991). Recent data have shown that at least some superantigens, such as the endogenous mouse mammary tumor provirus (MMTV) products (called Mls) (Janeway, 1991) and the staphylococcal enterotoxins, bind to TCRP chains at a site completely distinct from any of the complementaritydetermining sequence loops that normally contact the target MHCpeptide complex (Pullen et al., 1990; Choi et al., 1990). For cellular immunologists, the discovery of superantigens and the identification of their cognate V, segments have been extraordinarily valuable. It has meant that if one can identify the Vp segment used in any cell’s TCR, it is possible to know at least one target antigen against which that cell can respond. In practice, the effects of introducing a particular superantigen into the environment can now be monitored easily by tracing the fate of all T cells that use the appropriate V, in their TCRs. Some superantigens, such as staphylococcal enterotoxins, can be administered acutely. Others, such as the Mls-encoded products, are inherited by vertical transmission of stably integrated MMTV genomes. By comparing mouse strains with and without particular MMTV proviruses, it has been possible to study the fate of the same Vp segment in cases where it does and cases where it does not react with a congenital “self” antigen. This approach has provided the incontrovertible evidence that negative selection can occur, and in a variety of ways.

A. NEGATIVE SELECTION 1 . lntrathymic Clonal Deletion

a . Cell Biology of Negative Selection. T cells with potentially superantigen-reactive receptors mature efficiently into medullary thymocytes and peripheral T cells in animals that do not express that superantigen. When a superantigen such as Mls is expressed congenitally, however, essentially none of the T cells with superantigenreactive V, are allowed to mature (Kappler et al., 1987, 1988; MacDonald et al., 1988a). These initial observations showed that autoreactive cells were either arrested in their development or selectively killed by a clonal deletion mechanism. Dramatic evidence for deletion was provided when von Boehmer and Loh and their colleagues constructed transgenic mice with already rearranged TCRa and TCRP transgenes that encoded receptors of known specificity. Through allelic exclusion, the receptors the transgenes encode dominated the entire thymocyte and peripheral T cell populations. When the transgenes

166

ELLEN V. ROTHENBERG

were bred onto a genetic background where the target antigen was a self-antigen (H-Y plus H-2Db in one case, H-2Ld in another), the resulting animals had thymi of sharply reduced size in which cortical CD4+8+and medullary thymocytes alike were severely depleted (Kisielow et al., 1988; Teh et al., 1988; Sha et al., 1988a,b).The reduction in overall cellularity showed that the cells were not simply being diverted or delayed in their maturation pathway. Soon thereafter, similar observations were made in mice transgenic for other TCRs, including several of class I1 MHC-restricted specificity (Berg et al., 1989a,b; Kaye et al., 1989). In each case, when the target antigen of the transgenic TCR was present as a self-antigen, the normally dominant cells bearing transgenic TCRs disappeared. When bone marrow cells from TCR transgenics were used to reconstitute radiation chimeras, they expanded and differentiated in a target antigen-free environment, but gave poor repopulation and failed to differentiate in a host that expressed the target antigen (Kisielow et d.,1988). Thus, TCR engagement with its cognate antigen in the thymus microenvironment can curtail thymocyte expansion and abort thymocyte development. In vitro systems have substantiated the interpretation that the autoreactive cells undergo lysis and not just a proliferative or differentiative arrest. Smith et al. (1989) showed that administration of any TCR ligand, including anti-Vpor anti-CD3 antibody, stimulates classical apoptosis in fetal thymic organ culture, i.e., nuclear pyknosis, early DNA degradation even in the context of chromatin, and subsequent fragmentation of the cells. Apoptosis can also be induced in cortical thymocytes isolated in suspension (Kizaki et al., 1989; McConkey et nl., 1989a)or by injection of anti-CD3 in vivo (Shi et al., 1989).Though acute treatment with anti-TCWCD3 antibodies does not delete 100% of TCR-bearing cortical thymocytes (Finkel et al., 1989b; and see below), the number killed significantly exceeds the number of these cells that are in cycle (Penit, 1986).It thus represents the elimination of postmitotic cells, not just a way of limiting cortical cell expansion. When TCR-transgenic CD4+8+ thymocytes are incubated in the presence of cells expressing their target antigen, massive cell death ensues (Swat et al., 1991).We will discuss the biochemistry of the killing process below. The efficacy of intrathymic deletion as a means for establishing tolerance reflects the range of self-antigens available in the thymus to engage the TCR on developing cells. Operationally, some form of tolerance can be established to self-antigens expressed either by thymic epithelial cells or by bone marrow-derived medullary dendritic


167

cells and macrophages (von Boehmer and Schubiger, 1984; Ready et al., 1984; Jenkinson et al., 1985). Both cortical and medullary stromal cells expressing class I1 MHC can induce deletion of thymocytes reactive with class II-restricted superantigens (van Ewijk et al., 1988). It is notable, however, that the deletion process appears to be triggered most efficiently when thymocytes interact with bone marrow-derived antigen-presenting cells (Marrack et al., 1988b; Roberts et al., 1990; Ramsdell et al., 1989; Lo and Sprent, 1986) at least in cases where the target antigen is a class II-associated superantigen. Thymic epithelial cells are much less effective. This is not due to a lack of class I1 MHC antigens on epithelial cells, for the whole cortical epithelium is class II+, albeit not expressing levels as high as those on medullary dendritic cells. Rather, the inefficiency of thymic epithelial cells in triggering deletion of class II-restricted receptors may be linked to the poor ability of these cells to deliver needed costimulatory signals, as assayed in conventional presentation of class II-associated antigens to functional T cell clones (Lorenz and Allen, 1989). When THI cells encounter antigen as presented by cortical epithelial cells, they tend to be anergized rather than activated. This may be because cortical epithelial cells appear to lack expression of BBl/B7, the ligand for the costimulatory receptor CD28 (Turka et al., 1990). Thus, the results imply that the signals needed to induce deletion efficiently in cortical thymocytes go beyond the signals needed to induce anergy in mature T cells. In fact, the most potent cells in the thymus for triggering deletion to class II-restricted superantigens appear identical to the cells that are the best APCs for activating mature T cells, namely the dendritic cells of the cortico-medullary region (Marrack et al., 1988b). Splenic accessory cells can also induce antigen-dependent suicide in vitro (Swat et al., 1991). These results establish that the signals that activate mature cells are highly similar to the signals that induce deletion of thymocytes. Whether an interaction is read as a signal for activation or a signal for deletion probably depends less on the antigen-presenting cell than on the stage of development of the responding thymocyte. The developmental window within which deletion can occur has been controversial. Initial experiments tracing the fate of cells with TCRs directed against class II-restricted superantigens showed that those TCRs were represented at full, unselected levels within the cortical thymocyte population, but were deleted from the medullary TCRh'ghsubsets (Kappler et al., 1987).Examples were Vp17, against an I-E-restricted self-antigen on B cells (Marrack and Kappler, 1988),and

168

ELLEN V. ROTHENBERG

Vp8.1 and Vp6 against Mls-la (Kappler et al., 1988; MacDonald et al., 1988a).Circumstantial evidence suggested nevertheless that the decision to undergo deletion occurred in the CD4+8+ subset, because these class 11-restricted TCRs were deleted not only from the CD4+8but also from the CD4-8+ medullary thymocyte subsets (Kappler et al., 1988; MacDonald et al., 1988a). Fowlkes et al. (1988) and MacDonald et al. (1988~) further showed that the fate ofVp6+cells that can differentiate into the CD4-8+ lineage must be decided at a time when those cells also express CD4. Zn v i m treatment with anti-CD4 antibody, under conditions in which it blocks CD4 but does not kill CD4+8+thymocytes, enables some cells with Vp6+ receptors to escape deletion and mature into CD4-8+ medullary thymocytes. More recently, Guidos et al. (1989) and Hugo et al. (1991) have traced the timing of Vp6 deletion in Mls-la mice, relative to TCR up-regulation and CD4 or CD8 down-regulation. Their results show that the earliest signs of deletion occur during the transition to a TCRhighphenotype when the cell first begins to lose expression of CD4 or CD8. As we shall see below, strong circumstantial evidence indicates that cells acquire such phenotypes only as a result of positive selection, suggesting that negative selection of Vp6 by Mls-1” follows positive selection. This case is not sufficient, however, to define the full window in which deletion takes place. All of these results define the narrowest, not the widest, extent of the negative-selection window, because the signal to undergo negative selection can be received prior to positive selection, even though its manifestation at the population level is not clear until later. Yet to understand the molecular mechanisms that trigger negative selection, it is essential to know the true complement of triggering receptors present in the cells at the time they become committed to die. Other considerations are likely, a priori, to affect the stage at which negative selection is seen to occur: the location of the deleting selfantigen in the thymus and the affinity of the selected TCR. We have already addressed the relative inability of most cortical epithelial cells to promote deletion via class 11-associated superantigens, yet these are the cells with which most cortical thymocytes are in contact. For many antigens, efficient deletion may only occur as thymocytes migrate to regions of the thymus with better antigen presentation. Thus, cortical cells may have the competence to be deleted before positive selection, even when they do not have the opportunity. Receptor-ligand affinity may also play a role. Although deletion appears to be triggered by even lower affinity TCR-target interactions than mature T cell activation, there is an avidity threshold for thymocyte deletion (Murphy et al.,


169

1990; Sha et al., 1990; Pircher et al., 1991). Both antigen density and TCR density (and probably CD4 or CD8 coreceptor density) contribute to the avidity of cell-cell interaction for a given TCR-ligand affinity value. Thus, different TCRs become capable of triggering deletion at different stages (e.g., see Matsuzaki et al., 1990). Little is known about the range of affinities of TCR-superantigen binding, and in at least one case it is likely that TCRa chains can influence the affinity of binding of Mls-la to TCR V,8.1 (Blackman et al., 1990). In other words, even T cells that share the same superantigen-responsive V, region are still likely to be heterogeneous in their TCRsuperantigen affinities (Berg et al., 1989b).The rather late stage when deletion is usually noted could represent the stage when TCR surface density has increased to the point where the average superantigenreactive cell can be engaged. These considerations help in the interpretation of the dramatic deletion phenotypes observed in TCR-transgenic mice. Here, as noted above, not only medullary cells but also the great majority of cortical cells can be eliminated. This occurs whether the imposed TCR is class I or class I1 restricted. Reconciling these data with the phenotypes observed in nontransgenic mice depends upon recognition of three artifacts” of the transgenic systems. In fact, all of these are illuminating. First, the TCRs on transgenic thymocytes are homogeneous in affinity. Thus the entire transgenic receptor-positive population will meet the avidity requirements for negative selection at the same stage. Second, the transgenic animals are generally selected for expression of higher than normal TCR surface densities. Therefore their thymocytes exhibit amplified sensitivity, relative to normal cells at the same stage, to stimuli that might trigger deletion. Finally, because sequential stages of gene rearrangement are unnecessary to confer TCR expression on the transgenic thymocytes, they can express TCRaP precociously (Berg et al., 1989b; von Boehmer, 1990). This enables them to be assayed for susceptibility to deletion while still in microenvironments and at stages of development in which the triggering criteria for deletion normally would not be met. The results of these analyses indeed show that the machinery for a deletion response is assembled in thymocytes considerably earlier than the transitional stage after positive selection. Only the CD4-8- cells, and cells expressing abnormally low levels of the TCRs or the appropriate coreceptor molecule, may escape deletion (Teh et al., 1989; Berg et al., 198913). A transgenic mouse model has been used elegantly to show that whether the same cortical thymocytes are deleted early or late depends on the negatively selecting antigen. Pircher et al. (1989) constructed trans‘I

170

ELLEN V. ROTHENBERG

genic mice with dual-reactive TCRs: the combinatorial ( a @specificity ) directed against a class I-restricted antigen of lymphocytic choriomeningitis virus (LCMV), and the Vp8.1 segment contibuting weak reactivity with Mls-la. When this TCR was bred into an Mls-la background, the thymocytes were deleted as expected at the cortical to medullary transition, but when the same TCR was bred into a host with the appropriate MHC and a chronic LCMV infection, the cells were deleted early in the cortical stage. These experiments indicate the basic compatibility of the transgenic and normal experimental systems, and sharply distinguish between competence to be deleted, which appears to extend at least throughout the CD4+8+ phase, and opportunity. b. The Search for Mechanism. A likely mechanism for clonal deletion has been suggested by the proclivity of cortical thymocytes, in particular, to undergo apoptosis upon TCR stimulation (Smith et al., 1989; Shi et al., 1989, 1991; Tadakuma et al., 1990; McConkey et al., 1989a). The biochemical events involved in this process further strengthen the link between thymocyte deletion and mature T cell activation. Apoptosis is an active suicide response involving protein synthesis, RNA synthesis, Ca2+-dependent signaling, and protein phosphorylation, which culminates in the internucleosomal fragmentation of chromatin (Kizaki et al., 1989; Wyllie et al., 1984; Shi et al., 1989; Cohen and Duke, 1984). Apoptosis is not uniquely provoked by TCR ligation: cortical thymocytes make the same response to elevated levels of systemic glucocorticoids, irradiation, and agents that elevate intracellular CAMP(Wyllie, 1980; Cohen and Duke, 1984; Huiskamp et al., 1985; Sellins and Cohen, 1987; Screpanti et al., 1989; McConkey et al., 1990a). Though the diversity of these inducing agents might seem to lessen the relevance of apoptosis to clonal deletion, most, but not all, of these stimuli trigger a sharp Ca2+ flux in thymocytes, like stimulation of the TCR (Kaiser and Edelman, 1977; Wyllie et al., 1984; McConkey et al., 1989a,b,c; Kizaki et al., 1989; Smith et al., 1'389; but see McConkey et al., 1990a).These overlapping biochemical features suggest multiple triggering pathways converging to a single suicide pathway. The contribution these acute models of negative selection may make is that they may ultimately provide clues to the mechanisms that distinguish negative selection from positive selection. Protein kinase C activation or IL-1 treatment appears to counteract the induction of suicide both in response to TCR ligands and in response to CAMP elevation; however, at least in murine thymocytes, protein kinase C activation by itself also induces suicide (McConkey et al., 1989b, 1990a,b; Kizaki et al., 1989). Certain immortal T lineage hybridoma


171

lines can also undergo apoptosis in response to TCR ligands or glucocorticoids (Ucker et al., 1989; Mercep et al., 1989),and these cells have provided another intriguing lead. In hybridomas, under certain conditions, two stimuli that induce apoptosis by different pathways can cancel each other out, leading to cell survival. TCR ligands and glucocorticoids, for example, do not induce but prevent suicide when they are added at the same time (Zacharchuk et al., 1990). Thus, these two kinds of evidence indicate that suicide may not be the response to one unique death signal, but rather a response to an unbalanced signal. Susceptibility to apoptosis does not disappear immediately upon emergence from the TCR'" CD4+8+ subset. Several groups have reported that in vitro treatment with TCR ligands can kill certain medullary thymocytes, particularly CD4'8- cells, as well as cortical cells (MacDonald and Lees, 1990; Nieto et al., 1990). Resistance to apoptosis induction by TCR ligands may develop slowly in medullary cells, for the quasi-synchronous cohort of cells differentiating in the neonatal thymus progresses from susceptibility to essentially full resistance between day 0 and day 7 (Zacharchuk et al., 1991). It is not yet clear whether this process is linked to the slow acquisition of functional maturity discussed in Section IV (Ramarli et al., 1987; Yang et al., 1988a; Rothenberg et al., 1988; Ramsdell et al., 1991; J.-F. Chang et al., 1991a,b); the fully functional cells expressing the Qa-2 marker are not detected in the medulla until considerably later than day 7 after birth (Ramsdell et al., 1991).However, the response defects of medullary thymocytes as a whole were not routinely noted in early studies using cortisone-resistant thymocytes as a source of presumptively mature cells (Ceredig et ul., 1982,1983b; see above, Section 1V.A.).Thus it is possible that the acquisition of complete resistance to apoptosis, whether by antigen or by glucocorticoid, is one of the final events in thymocyte maturation. The effects of a natural inhibitor of apoptosis suggest that thymocytes are subject to more than one negative-selection mechanism. The bcl-2-encoded protein appears to prolong survival in a variety of cell types that normally exhibit regulated growth, and protects many (although not all) cytokine-dependent cell lines from dying for many days after their requisite growth factors are removed (Nufiez et al., 1990). Notably, it fails to protect mature IL-2-dependent lines from apoptosis when IL-2 is exhausted. However, in the normal thymus, bcl-2 is expressed at much higher levels in medullary cells than in cortical cells (Hockenbery et al., 1991), suggesting that its absence may be responsible for rendering cortical thymocytes susceptible to negative selection. Thus, two groups (Sentman et al., 1991; Strasser et al., 1991)

172

ELLEN V. ROTHENBERG

have tested the effect of bcl-2 on cortical thymocytes when it is expressed in transgenic mice under the control of the cortical thymocytespecific lck promoter or enhanced in the T lineage generally by the E p enhancer. In these animals, cortical thymocytes exhibit dramatically improved viability, resisting apoptosis in response to irradiation, antiCD3 injections, or other acute stimuli. The cortical cells also remain viable for unprecedentedly extended periods of time in uitro. However, as examined in steady-state in uiuo, it is clear that autoreactive TCRs continue to be deleted from the medullary TCRhighpopulation in Lck-bcl-2 animals almost as well as in normal animals (Sentman et al., 1991). In mice expressing bcl-2 from the E p enhancer, a larger fraction of auto-reactive T cells appear to reach a TCR-intermediate stage, but they still do not mature to a TCRhighstate nor accumulate in the periphery (Strasser et al., 1991). Thus at least some mechanism for inducing clonal deletion intrathymically is able to overcome the protective effects of bcl-2. Upon further analysis, the phenotypes of these animals may reveal the existence of two separate clonal deletion mechanisms: one acting early and counteracted by bcl-2, and the other acting late and potentially bcl-2 resistant. The precise nature of the triggering signal or signals for suicide are still somewhat elusive. As noted in Section IV,B, it is surprising that cortical cells are especially susceptible to a presumably Ca2+dependent suicide pathway when 50-70% of them are so poor at inducing Ca2+ fluxes in response to TCR ligands (Finkel et al., 1989a,b).Only 30-50% of cortical thymocytes are deleted in response to antibodies against the TCRaP chains or in response to superantigens such as staphylococcal enterotoxin B (Finkel et al., 1989b). However, real self-antigens presented in the thymic context are not simply ligands for the TCRa and TCRP chains, as already discussed in Section I. Another possibility, therefore, seems more attractive: namely, that the physiological ligand for induction of suicide is not a TCR ligand alone, but a multivalent ensemble of cell surface molecules. Several reports show that specific co-cross-linking of the TCR with CD2, CD4, CD8, or even with Thy-1 can markedly enhance the ability of cortical thymocytes to flux Ca2+ (Turka et al., 1991b; Deusch et al., 1990; Gilliland et al., 1991; Nakashima et al., 1991). In the case of Thy-1, the mixed stimulus efficiently potentiates suicide induction (Nakashima et al., 1991). Such results may also provide a key to the ability of soluble anti-CD4 to block deletion of Vp6+ thymocytes in Mls-1' mice, if antibody treatment prevents CD4 from associating with the TCR/CD3 complex. Through another pathway, CD28 engagement might itself complete the signal for deletion. While these experiments do not deci-


173

sively identify any specific receptor association that must be used to induce apoptosis in aim, they offer strong evidence for a possible mechanism. In summary, CD4+8+ TCR’” cortical thymocytes appear to be a preferential target for most mechanisms inducing clonal deletion, but susceptibility to deletion may extend throughout the cortical thymocyte stage and beyond. It is possible that the early and late mechanisms are not the same. However, the ligand-receptor pairs that trigger thymocyte suicide at any stage appear indistinguishable from those mediating activation in mature T cells. The proximal molecular events, such as Ca2+ flux, inositol phospholipid metabolism, protein tyrosine phosphorylation, and (at least in murine thymocytes) some level of protein kinase C activation, also appear similar. Why, then, does the interaction result in death? CD4+8+ thymocytes appear to be unusually easily triggered to commit themselves to apoptosis in response to a variety of stimuli. Their diversion of signals to a suicide pathway is consistent with their idiosyncratic failure to use those signals to activate expression of conventional “response” genes. This presumably reflects their harboring a unique complement of transcription factors, and/or signal transducers that fail to provide signals in the proper quantitative balance for activation. The continued sensitivity to clonal deletion in medullary thymocytes may similarly be consistent with their own specific inability to use TCR/CD3 to trigger activation, even when they are competent to respond well to other stimuli (Section IV,D). The persistence of susceptibility to apoptosis well into the TCRhighmedullary thymocyte phase can ensure that even TCRs with relatively weak affinity for important self-antigens can achieve sufficient binding avidity to be purged from the repertoire.

2 . Other Mechanisms Though cortical thymocyte apoptosis is the most spectacular tool for imposing tolerance on the T cell repertoire, it is not the only one. I will close this section by briefly noting two other mechanisms that play important roles. One important additional mechanism for establishing tolerance is the intrathymic imposition of clonal anergy. As we described above, thymic cortical epithelial cells are relatively poor at inducing deletion, but they can program developing cells for nonresponsiveness, at least by the criteria of permanently depressed responsiveness to TCR stimuli and extremely poor inducibility of IL-2 (Roberts et al., 1990; Ramsdell et al., 1989; Ramsdell and Fowlkes, 1990; Blackman et al., 1990, 1991). The evidence presently available suggests a gradient of re-

174

ELLEN V. ROTHENBERG

sponses based on the affinities of TCR-ligand interactions leading to deletion, in the case of strong interactions, or to anergy, in the case of weaker interactions. For example, superantigens are presented more effectively by certain MHC class I1 products than by others. In a genetic background where the superantigen is presented well, thymocytes are deleted; but in a background where it is presented poorly, cells using the relevant Vp appear and become phenotypically mature, but remain poorly responsive. It is not clear whether anergy and deletion are truly related, in a continuum of responses to a single negative-selection mechanism, or whether they represent two distinct tolerogenic mechanisms with overlapping affinity thresholds. This question remains to be resolved. The observed response phenotype also raises the question of whether intrathymically imposed anergy may be related to a failure to relieve the response paralysis initially imposed on thymocytes as they enter the major cortical population (see Section IV). The other important mechanism for maintaining tolerance is the ability of elements in the periphery to impose either anergy or even deletion on postthymic cells that have inappropriate reactivity. Anergy induction in the periphery is correlated with encountering the target antigen at an inappropriate time in development (Jones et al., 1990a), by an inappropriate immunization route (Rammensee et al., 1989), or in certain class 11+ epithelial contexts, such as pancreatic islet cells (Burkly et al., 1990), which may lack the ability to provide costimulatory signals. Deletion can also occur in the periphery (Jones et al., 1990b; Webb et al., 1990), although it is not as well defined at the responding cell level as it is in the thymus. Unlike the response in the thymus, there is evidence that disappearance of the reactive T cells in the periphery is preceded by a transient burst of proliferation (Webb et al., 1990; Rocha and von Boehmer, 1991). This provides an evocative link between peripheral deletion and the regulation of clonal proliferation discussed in Section I,E, in particular the delicately controlled switching between the TCR-dependent state, the interleukindependent state, and the refractory state. It is not yet clear whether peripheral deletion is a variant of apoptosis, a novel phenomenon, or simply an inappropriately modulated response to a strong, temporally abrupt, systemic antigenic challenge. While the appreciation of peripheral tolerance mechanisms is rather new, these must be considered absolutely essential to the maintenance of tolerance. Not all tissue-specific self-antigens can be present at a suprathreshold concentration in the thymus. Some types of T cells never develop in the thymus in the first place. Furthermore, normal


175

developmental changes, such as puberty and aging, imply that old cohorts of postthymic cells may be faced with newly expressed selfantigens, long after they have left the thymus (Jones et al., 1990a). Postthymic mechanisms are thus the means by which the T cell repertoire can keep pace with the changing definition of self.

B. POSITIVESELECTION Positive selection is the key developmental process controlling the production of mature T cells, and its mechanism remains one of the most challenging questions in T cell development. Whereas most of the developmental transitions that thymocytes undergo may be controlled by simple signals to stop or to go, positive selection also determines the direction of differentiation, to a CD4+ helper or a CD8+ killer lineage. With the advent of TCR transgenic mice, transgenic mice forced to express CD4 or CD8, and “knockout” mice in which coreceptor, class 11, or class I MHC expression has been abolished, the phenomenology of the process is becoming clear. We shall therefore consider the problem of positive selection in three parts: (1) rescue from death-by-default, (2)acquisition of an appropriate single-positive phenotype, and (3)acquisition of appropriate function.

1 . Positive Selection versus Two Kinds of Death Maturation into a single-positive state protects cortical thymocytes from the common lot of disappearance and death within 3-4 days of the last cell cycle. Though it is still unclear exactly where and how so many cortical cells disappear (see Rothenberg, 1990), careful analysis of population dynamics has established that medullary CD4+8- and CD4-8+ thymocytes as a whole have much longer intrathymic life spans than do cortical cells, on the order of 13days in mice (Hirokawa et al., 1985; Shortman et al., 1991).Work with several TCR-transgenic models has shown that rescue is absolutely dependent on the ability of the TCR to interact with either class I or class I1 molecules in the thymic microenvironment (von Boehmer, 1990).In the absence of such an interaction, the cells remain CD4+8+ TCR’” and remain in the cortex until they die on a normal schedule (Teh et al., 1988; Sha et al., 1988a; Scott et al., 1989; Huesmann et al., 1991). This interaction must not, however, be equivalent to allo-MHC recognition, which leads to clonal deletion as we have just described (Section V,A). Instead, it is equivalent to self-MHC restriction for a mature T cell. As we shall discuss below, what this distinction must mean on a molecular level is an interesting and controversial question. Because of the overwhelming likelihood of death for the average

176

ELLEN V. ROTHENBERG

cortical thymocyte, there has been a protracted debate as to whether most of these cells are irrevocably committed to die or are simply awaiting a rare opportunity to be rescued. The contrast between the extreme fragility of most cortical cells and the much better viability of their immediate CD4+8+blast cell precursors has made it seem attractive to propose that all postmitotic CD4+8+ cells are beyond rescue (Guidos et al., 1989). At the other extreme, Penit and co-workers (Penit, 1986; Penit and Vasseur, 1988) showed that single-positive thymocytes can be generated in a fully postmitotic process, and that the first CD4+8- cells do not arise until 3 days after the last cell division in their CD4+8+ precursors. As described in Section III,A, an apparent resolution to this disparity has recently been provided by a renewed focus on the minor population of TCR/CD3highCD4+8+ cells initially described by Blue et al. (1987b), which now appear to be the earliest distinct form of positively selected cells. This conclusion is based on the expansion of this subset in TCR-transgenic mice whose receptors are strongly positively selected, combined with the superb kinetic fit between the rate of production of TCRhighCD4+8+cells and the rate of production of medullary single-positive cells in steady state (Borgulya et al., 1991; Shortman et al., 1991). Although most workers find that the first CD4+8+ TCRhigh cells appear postmitotically, this occurs much sooner than the first appearance of cells with a clear single-positive phenotype, i.e., only 1 day as opposed to 3 days after the last cycle. Thus, although it may take a day to up-regulate TCR expression, and an additional 2 days or more to lose expression of the inappropriate coreceptor from the surface, cortical thymocytes may indeed undergo the interaction that leads to positive selection during or very soon after their last cell division. It is noteworthy that throughout cell division, most cortical cells express very low levels of surface TCRaP (Guidos et al., 1990), and if the decisive interaction actually occurs before the last mitosis, the high amounts of CD4 and CD8 present on the large blast cells would provide an exceptionally skewed coreceptor :TCR ratio for the likely molecular participants in the selection process. Negative selection can be induced by a variety of ligands, including superantigens that contact only the TCRO chain; as we have seen, these ligands can be presented on any of a variety of cell types, from the thymic cortex to the spleen, and still trigger cortical thymocyte suicide. By contrast, positive selection appears to have much more precise requirements. There are no reports of superantigens inducing recognizable positive selection, and the only anatomical domain in which this process appears to take place is the thymic cortex (Benoist


177

and Mathis, 1989; Berg et al., 1989a; Bill and Palmer, 1989).Radiationresistant, 2-deoxyguanosine-resistantepithelial elements efficiently carry out positive selection, although they are ineffective for negative selection (Ready et at., 1984; Lo and Sprent, 1986; Ron et at., 1986) (see Section V,A). Thus, the inability of these stromal cells to provide a costimulatory signal to mature T cells (Lorenz and Allen, 1989) at least does not block positive selection; it may even be an essential component of positive selection. The available sites where positive selection can occur are also limiting, even within the thymus. Positive selection of transgenic TCR+ thymocytes occurs distinctly more efficiently in animals homozygous for the proper restricting MHC antigen than in heterozygotes (Berg et al., 1990).Huesmann et al. (1991)have used graded transfers of transgenic pre-T cells with a selectable TCRs to demonstrate directly the saturation of positive selection niches. It is not yet clear whether this reflects the existence of a specialized, limited epithelial subdomain, or a low-efficiency interaction with epithelial cells throughout the cortex. Understanding of positive selection at both the cellular and molecular levels will ultimately depend on resolving why the commitment to positive selection appears to occur so soon after first cell surface TCR expression. One possibility is that there is an irreversible, intrinsic loss of competence to be positively selected about 1 day after the last cortical thymocyte mitosis. In this case, the properties of at least twothirds of the cortical thymocyte population are irrelevant to the mechanism of the positive-selection process, as these cells would be effectively committed to die, by default if not by negative selection. A version of this model has been proposed by Finkel et al. (1991). An alternative possibility, however, given the limited number of selection niches, is that those niches are anatomically restricted to the outer cortex. The centripetal pressure of intrathymic migration may then sweep cortical cells away from the positive-selection niches even while they still retain the capability to respond. In this case, most or even all cortical cells might be intrinsically selectable. Unlike negative selection, positive selection has not yet been modeled successfully in vitro with purified cell populations. This makes it difficult or impossible to resolve how long after their last mitosis cortical thymocytes retain the intrinsic competence to be positively selected. Therefore, as we consider the state of the evidence on the mechanism of positive selection, it will be important to bear in mind the critical uncertainty over whether typical cortical thymocyte properties are or are not representative of the cells that actually undergo positive selection.

178

ELLEN V. ROTHENBERG

2 . Ligands and Receptors i n Positive Selection

Those molecular participants in the positive-selection-determining interaction that have been defined to date are qualitatively indistinguishable from those that trigger negative selection. TCR specificity is of decisive importance, for the introduction of a class I-restricted transgenic TCR can direct an overwhelming majority of thymocytes to a CD8+ fate, whereas a class II-restricted transgenic TCR correspondingly shifts the balance to the almost exclusive production of CD4+ cells (Teh et al., 1988; Sha et al., 1988a,b; Scott et al., 1989; Kaye et al., 1989; Berg et al., 1989a,b). Other critical participants are CD4, CD8, and the MHC class I and class I1 ligands in the microenvironment. TCR specificity appears to be the dominant influence in determining whether a positively selected cell will become a CD4+8- cell or a CD4-8+ cell. The relevant specificity is the canonical one determined by TCRaP V regions and junctional regions, not the class II-associated superantigen reactivity that may be conferred by the use of a particular V, segment. It seems likely that the cell can distinguish between cross-linking of the TCR in cis with CD4, when it is bound to class 11, and in cis with CD8, when it is bound to class I. This co-cross-linking appears to dictate the result even though the other coreceptor can engage in independent interactions with the MHC. Only the coreceptor coengaged with the TCR is retained, whereas expression of the other coreceptor is turned off. The best evidence for this mechanism is provided by studies of transgenic mice in which a class I-restricted TCR transgene is present as well as a CD8 transgene constitutively expressed on all T lineage cells (CD8T)(Borgulya et al., 1991; Robey et al., 1991). In these cases, the TCR and CD8T would be sufficient to induce positive selection regardless of the expression status of CD4. If TCR-CD8-MHC class I engagement did not deliver any direct “instructive” signal to turn off expression of CD$ then some class Irestricted TCRs should be found in the c D 4 + 8 population ~ that had randomly turned off endogenous CD8 instead. However, all positively selected cells with the transgenic TCR do turn off their endogenous CD4 expression. Thus TCR specificity can control coreceptor downregulation whether or not a choice between the coreceptors is neces sary. The MHC antigens in the microenvironment are also required for positive selection. Treatment of mice with high doses of anti-class I1 antibodies blocks the development of CD4+8- cells (Kruisbeek et al., 1985), and treatment with anti-class I blocks the development of CD4-8+ cells (Marugi6-GaleSi6et al., 1988). An elegant confirmation


179

that these effects are due primarily to preventing interactions with MHC molecules, and not to inappropriate signaling, has come from mice homozygous for gene disruptions that destroy either class I1 or class I MHC expression, respectively (Zijlstra et al., 1990; Cosgrove et al., 1991; Grusby et al., 1991). In each case, functional absence of one class of MHC molecule has no effect on the production of TCR'" cortical thymocytes or on the development of thymocytes restricted by the other MHC class: CD8+ T cell production is normal or expanded in class I1 MHC-deficient mice (Cosgrove et al., 1991; Grusby et al., 1991), whereas CD4+ cells are also produced normally in 02microglobulin-deficient mice, which cannot express any class I products (Zijlstra et al., 1990). Thus, neither successful positive selection nor the ongoing accumulation of CD4+8+ cells depends on maintaining a balance of signals from the different MHC ligands ofthe different coreceptors. Note that the minimal effects of MHC antigen deletion intensify the mystery about the nature of the signal responsible for constitutive CD35 chain phosphorylation in cortical thymocytes in vivo (Section IV,B). Very similar phenotypes are observed when the coreceptors, rather than the ligands, are blocked or deleted. In initial antibody-blocking studies (discussed in Section V,A), the same anti-CD4 treatments that allowed superantigen-reactive TCRs to mature normally into CD4-8' cells resulted in the complete disappearance of the CD4+8- subset This disappearance (Fowlkes et al., 1988; MacDonald et al., 1988~). was due to a blockade of positive selection to the CD4' lineage, rather than to lysis, because under these conditions the anti-CD4 antibody did not eliminate preexisting peripheral CD4+ cells nor even fragile CD4+8+ thymocytes (Ramsdell and Fowlkes, 1989). Corresponding results were obtained for treatment with anti-CD8 (Zuiiiga-Pflucker et al., 1990b; Ramsdell and Fowlkes, 1989). In these cases, the binding of antibodies to thymocytes might alter their behavior artifactually. However, more recently, gene disruption experiments have fully confirmed the conclusion that coreceptors are essential for the positive selection of the lineage in which they are expressed, but not for the lineage in which they will be down-regulated. Homozygous CD8disrupted mice produce normal percentages of TCR'" CD4+ cortical cells and normal numbers of TCRhighCD4+ positively selected cells, exporting functionally competent CD4+ helper cells to the periphery (Fung-Leung et al,, 1991). Disruption of CD4 produces the reciprocal phenotype, with strikingly undisturbed production of cortical cells and TCRhighCD8+ cells (Rahemtulla et al., 1991). Thus, in spite of the finely programmed expression of CD4 at low levels on early intrathy-

180

ELLEN V. ROTHENBERG

mic precursors (see Section 111) and CD8 expression on proliferating blasts just before TCR appearance, neither molecule is indispensable until the positive selection event. The surface densities of CD4 and CD8 can influence the rate and efficiency of positive selection. In the double TCR/CD8 transgenics discussed above, the elevated total CD8 levels allow cells with the class I-restricted transgenic TCRs to be selected more efficiently to the CD8' lineage (Borgulya et al., 1991; Robey et al., 1991). On the other hand, signals derived from CD4-ligand engagement can compete against signals derived from CD8-ligand engagement, because forced expression of the wrong coreceptor can reduce the efficiency of positive selection. When mice with the class I-restricted transgenic TCRs are crossed with animals that express a CD4 transgene in all T cells, positive selection to a CD8+ lineage (CD8+,CD4P) is impaired (Teh et al., 1991).' These results demonstrate the interdependence of signals derived from the TCRs and coreceptors in controlling the positive-selection process. 3. Molecular Basis of a Four-Way Developmental Choice a. Positive versus Negative Selection. The mechanism of positive selection must account for its distinction from negative selection as well as for the divergence of its two types of successful products, CD4+8- and CD4-8+ cells. The first question is how both positive and negative selection use similar ligand-receptor complexes to effect such different outcomes. We will consider the evidence for the roles of at least five elements: (1) TCR affinity, (2) an epithelial-specific peptide-MHC complex, (3) Ca'+ flux generation, (4) receptorassociated tyrosine kinases, and (5)the CD28 costimulatory receptor. Unfortunately, the question is still unanswered, but the clues are interesting. At least part of the difference between negative and positive selection appears to be related to the affinity of the TCR-MHC interaction. In two cases of TCRs restricted by the class I MHC antigen H-2Kb, mutations in the Kb molecule have striking effects on whether selection will be positive or negative, or will fail altogether (Sha et al., 1990; In these mice, the CD4 transgene appears to allow a few cells with apparently class 11-restrictedendogenous TCRs to mature with retention of CD8 expression. However, it is not clear whether this is evidence for an asymmetry between the effects ofTCR-CD4 coengagement and TCR-CD8 coengagement. Unlike the transgenic TCRs, the endogenous TCRs on these unusual cells are of uncharacterized specificity and affinity, and very likely to be heterogeneous. Thus, it is not certain what kinds ofinteractions actually guided their maturation.


181

Nikolib-Zugib and Bevan, 1990).In a thymus expressing wild-type Kb molecules, these TCRs are positively selected to the CD8+ lineage. In general, the mutations that are likely to increase TCR-ligand affinity, making the “restricting element” into an “alloantigen,” also induce negative selection instead of positive selection. Other mutations that reduce affinity, as monitored by a failure even to serve as a restricting element, generally fail to promote selection at all. Thus, TCR affinity or the ratio of TCR affinity to coreceptor affinity appears to be important in determining whether and what kinds of selection will occur. The results of these experiments also suggest a role for the selfpeptides that are normally associated with MHC glycoproteins. This is because the mutations with the most dramatic effects on thymocyte selection probably do not affect the face of the MHC antigen that contacts the TCR directly, but rather alter the walls of the cleft into which a peptide would fit. Not only overall affinity for the TCR, but the specific peptides bound by the mutant MHC antigens, may be the real basis for these effects. Thus, it is possible that positive selection is caused by interaction with MHC antigens that bear a unique class of self-peptides. Some evidence for a class of peptides unique to the thymic cortical epithelium was reported by Marrack et al. (1989). These authors showed that even in the absence of nominal antigen, autologous thymic epithelial cells or nurse cells could stimulate certain T cell hybridomas, whereas the autologous splenic accessory cells could not. CD4, the TCRs, and their common class I1 MHC ligands were implicated in this activation event. The defect in antigen presentation by the cortical epithelial cells (Lorenz and Allen, 1989) (Section V,A) would not affect hybridoma activation, because hybridomas do not require costimulatory signals (cf. Shimonkevitz et al., 1983; Watts et al., 1984; Quill and Schwartz, 1987; Jenkins and Schwartz, 1987). Thus, these results reveal the existence of a unique MHC-associated target structure on thymic epithelial cells, presumably an unusual MHC-epithelial peptide complex. To date, none af the chemical properties of this complex is known. It is possible, therefore, to suggest that such a complex could be recognized in a unique way by TCRs, selecting for MHC restricting-element specificity but not for peptide-binding specificity. It is further possible that such a complex could transduce a unique type of signal through the TCR, resulting not in conventional activation, incomplete activation, or suicide, but in positive selection. It is not clear how the basic signaling biochemistry of TCR/CD3 signaling would be altered by the identity of the MHC-associated peptide, yet a new report suggests that such modulations of T cell response

182

ELLEN V. ROTHENBERG

may be possible in other cases (Evavold and Allen, 1991). Until more is known about the epithelial peptide-MHC complex (or complexes), it will not be possible to test this suggestion. Whether there need be a unique class of positive-selectiondetermining ligands depends on whether positive and negative selection are alternative or sequential events. Logically, a sequential model in which negative selection follows positive selection is quite attractive. In this case, the entire universe of self-peptides could be presented to thymocytes in the outer cortex, resulting in positive selection of all cells with any detectable reactivity, Those from this set that met a criterion of severe self-reactivity could then be purged through negative selection. This could work provided that cells are only susceptible to positive selection early, and only susceptible to negative selection later. Finkel et al. (1991) have interpreted the heterogeneity in the coupling of TCR-CD3 to activation mediators in cortical thymocytes (see Section IV,A) as evidence for this switching in selection responses. In the sequential model, all negatively selecting ligands would be a subset of the positively selecting ligands. Furthermore, there would be no need for a separate process to select MHC “restriction” specificity as opposed to MHC-peptide “recognition” specificity. Positively selected cells would react with different MHC-peptide complexes in the periphery because they had been positively selected by different MHC-peptide complexes in the thymus. Positive selection as an alternative to negative selection, on the other hand, requires an alternative triggering signal. In spite of the attractiveness of the sequential model, an alternative pathway model is necessitated by the evidence that negative selection can precede positive selection, as discussed above, and the evidence from transgenic animals that negative selection can be considerably more efficient than positive selection (von Boehmer, 1990; Rothenberg, 1990). These inconsistencies are the basis for the presentation of negative and positive selection as alternatives in Fig. 5. However, the sequential model may not be ruled out at this time. Whereas negative selection can precede the completion of positive selection, or even the increase in surface TCR/CD3 expression (Ohashi et al., 1990b) it might not precede some early positive-selection event for which there is no cell surface marker. Furthermore, complicating the analysis of positive-selection versus negative-selection efficiencies is the lack of a system in which the rates of appearance of positively selected cells can be measured. The data on steady-state populations in vivo simply give the cumulative yields of positively selected cells, which may be strongly affected by regulation at the level of export (Berg et al., 1989a; Chaffin and Perlmutter,


183

1991)and/or other intrathymic homeostatic mechanisms. Once again, the question must ultimately be resolved at the level of the cortical thymocyte population: whether there is only a unique subset that can be positively selected, and whether that subset can only be selected positively. b. Signaling Pathways that Control Positive Selection. Positive and negative selection are likely to share some of the same signal transduction mediators inside the cell. Circumstantial evidence suggests that one may be Ca2+. Cyclosporin A, a potent inhibitor that preferentially blocks Ca2+-dependent activation processes, not only prevents experimentally induced apoptosis (Shi et al., 1989) but also completely eliminates positive selection (Jenkins et al., 1988; Gao et al., 1988; Kosugi et al., 1989). Finkel et ul. (1991) have proposed that positively selectable thymocytes may be those protected from suicide by not fluxing Ca2+ in response to TCR ligands, but co-cross-linking of TCRs with certain other molecules can increase Ca2+ fluxes without promoting suicide (Nakashima et al., 1991; see below). Thus, Ca2+dependent activation event may be involved in positive selection as well as in negative selection. Furthermore, in the lck/bcl-2-transgenic mice described in Section V,A, a striking effect of the transgene is the increase in TCR/CD3 expression in CD4+8+ cells, giving a TCRhiRh phenotype to about 30-40% of the cortical population. Though much remains to be learned about the basis for this effect, it raises the possibility that an early form of negative selection might be convertible directly into a partial signal for positive selection, provided that death is blocked. If the affinity of the TCR-ligand interaction controls selection, it is also reasonable to suppose that the availability of Fyn and Lck activity would participate in setting the affinity thresholds for positive, as well as for negative, selection. The positive-selection process indeed seems to mobilize these kinases: Carrera et al. (1992a) have found both Fyn and Lck to be activated in thymocytes undergoing positive selection as compared to cells failing to be selected. Furthermore, they note that Lck redistributes in cells undergoing positive selection through a class I-restricted receptor. They find that this kinase is now more tightly associated with CD8 than with CD4, its usual preferred partner (Hurley et al., 1989), and it now coprecipitates with CD3 chains (Carrera et al., 199213). This suggests that during class I-mediated positive selection there is an unusually intimate association between the CD8- and TCR/CD3-associated kinases and kinase substrates, of a kind previously found only in certain lymphoma cells (Burgess e t al.,

1991).

184

ELLEN V. ROTHENBERG

Though these kinases are likely to be involved, their activity does not appear to be rate limiting for negative or positive selection. Perlmutter and colleagues have constructed a series of transgenic mice that overexpress Fyn, underexpress Fyn, or overexpress Lck in thymocytes (Abraham et al., 1991; Cooke e t al., 1991). These provide the material for testing the hypothesis that the ability of cortical thymocytes to mount strong Ca2+ fluxes controls the outcome of selection. As discussed previously, Lck overexpression leads to a sharp developmental arrest, so that any effects on positive selection per se are difficult to measure. On the other hand, Fyn overexpression leads to a strong increase in the triggering capacity of the TCRs, leading to markedly enhanced Ca2+ flux responses in Fyn-overexpressing CD4+8' TCR'" thymocytes. Correspondingly, Fyn inhibition leads to lower TCR signaling efficiency. If the rates of positive or negative selection depended mainly on the magnitude of Ca2+ flux, whether because of developmental changes in signaling capacity or because of the affinity of TCR-ligand interaction, these transgenic mice should show highly altered selection patterns. It is striking, therefore, that these quantitative signaling changes have very little effect on either the distribution of CD4/CD8 subsets or the pattern of TCR/CD3 expression (Cooke et al., 1991; R. Perlmutter, personal communication). At least in steady state, only subtle effects can be seen with increased Fyn activity leading to slightly enhanced levels of putative positive-selection intermediates (R. Perlmutter, personal communication). A strong interpretation of this surprising lack of impact must await further experiments based on breeding the Lck and Fyn transgenics with TCR transgenics of known specificity. However, a straightforward interpretation of the results would be that the overall rates of negative and positive selection are controlled more stringently by some other factor than by the quantitative intensity of TCR signaling. Other receptors besides the TCR and the CD4/CD8 coreceptors could very well participate in changing the outcome of TCR triggering, due to the combinatorial effects of their signals. At present, all possibilities are purely speculative, but I raise the following for illustration. First, the triggering receptor CD2 may be capable of modifying TCR signals at this stage, for it is clearly associated with an activation pathway (Section IV,D). Even at a later stage, in human medullary thymocytes, simultaneous engagement with CD2 and CD3 ligands cancels out stimulation (Ramarli et al., 1987). Thus, in a manner analogous to the antagonism of glucocorticoids and anti-TCRs in inducing apoptosis, the CD2-LFA-3 interaction could block the toxicity of TCR engagement. Also worth noting for a different role is CD28, which can deliver unique activation signals to medullary thymocytes to comple-


185

ment CD2- or TCR-coreceptor coligation (Section IV,D). The population distribution of CD28 in the thymus suggests that it is increased primarily as a result of positive selection (Yang et al., 1988a; Turka et al., 1990). Thus, it might be involved in preventing the induction of anergy in medullary thymocytes (Ramsdell et al., 1989; Ramsdell and Fowlkes, 1990).In the cortex, there is as yet no evidence on whether CD28 is present on cells at the stages when they are susceptible to positive selection. However, the intrathymic location of the CD28 ligand (Turka et al., 1991b) appears to be interlobar rather than on the cortical epithelial cells that probably induce positive selection. At a phenomenological level, a pattern emerges: stromal cells that provide typical costimulatory signals seem to induce negative selection, whereas those that fail to provide costimulatory signals induce positive selection. Thus, if CD28-BBl/B7 interaction is the costimulator, it is possible that an essential aspect of positive selection is to avoid integration of TCR triggering with CD28-derived signals. c. CD4 uersus CD8 Down-Regulation. Different fates await cells as a function of whether their TCRs are co-cross-linked with CD4 and CD8. In experimental efforts to duplicate these signals in uitro with antibodies, co-cross-linking with either CD4 or CD8 induces a higher Ca2+ flux than does TCR cross-linking alone. The differences in the signals transduced, however, and even in the proximal cellular responses, remain uncertain. There are reports that on CD4+8+ human cells, with similar surface densities of both coreceptors, TCR-CD4 cross-linking elicits a stronger Ca2+ flux than does TCR-CD8 crosslinking (Gilliland et al., 1991).In part, this may be due to the fact that many of the CD8 molecules on cortical thymocytes are variant molecules that cannot bind to Lck (Zamoyska and Parnes, 1988; Zamoyska et al., 1989). The significance of these CD8 variants is intriguing but unknown. Whereas superficially an enhanced Ca2+ flux might seem more likely to induce negative selection, the effects of TCR-coreceptor coengagement seem to be specifically protective (Nakashima et al., 1991):TCR-CD4 cross-linking does not potentiate apoptosis, whereas TCR-Thy-1 cross-linking, which yields a similar Ca2+ signal, does. The protective effect might be mobilized by some specific substrate of Lck. Thus, assuming that suicide is prevented, it is possible that differences in the intensity of Ca2+ signals may be used to discriminate between the signals to undergo maturation along the CD4+ or CD8+ pathway. Other differences remain to be discovered. There is another distinction between CD4 and CD8 engagement, although its significance is not yet understood. This is the specific, striking effect that the disruption of CD4-class I1 MHC contacts has in

186

ELLEN V. ROTHENBERG

the thymus. There is no evidence, as yet, that CD8-class I interactions within the thymic microenvironment have any effect on the properties of CD4+8+ thymocytes. But CD4-class I1 in vivo interactions seem to be involved in the maintenance of the TCR'" phenotype of the cortical cells. When these contacts are inhibited by any of a variety of means, in vivo or in uitro, CD4+8+thymocytes quickly increase their functional surface TCR expression to levels approaching those of medullary thymocytes (McCarthy et al., 1988; Finkel et al., 1989a; Bonifacino et d., 1990; Nakayama et al., 1990; Cosgrove et al., 1991). The phenotype is reminiscent of cells beginning to be positively selected, or cortical cells in ZcklbcE-%transgenic mice. These effects are not simple to interpret, for it is uncertain whether elevated TCR levels per se are either unique or instrumental to the positive-selection process. Nevertheless, these specific effects of CD4 engagement when it is not coligated with the TCR suggest that linked and unlinked coreceptors may activate separate signaling pathways in the cortical thymocyte population. This provides a potential example of signaling competition, a means for the cell to set affinity thresholds for TCR-CD4-MHC ternary associations using the binary CD4-MHC association as a standard. These results therefore have two kinds of significance: one, a strong indication that coreceptor-Lck complexes can act on signaling pathways other than the familiar one of T cell activation, and two, that CD4-Lck and CD8-Lck must differ in at least some of their intracellular targets, In summary, positive selection may be influenced by quantitative changes in TCR-ligand affinity. Yet the process itself is surprisingly resistant to perturbation in the face of alterations either in the balance of MHC-coreceptor signals or even in the amplitude of Ca2+ fluxes triggered through the TCRs. It appears, if anything, to be negatively correlated with CD28-ligand interactions. Identification of the auxiliary interactions that do participate is the critical next step, and to do this, it will be essential to focus only on cells that are capable of being positively selected. Only recently has it become clear how early positive selection normally occurs. If commitment to selection must occur early, further progress will depend on isolating the minority of stillselectable cells away from the majority of cortical thymocytes, which may well be beyond the possibility of rescue.

4. Functional Divergence: The Open Frontier Ultimately, positive selection is linked with functional specialization. One of the most striking results, reproducibly observed in mice with manipulated positive-selection yields, is the faithful concordance


187

between CD4-CD8 phenotype and effector function. Even cells maturing via artificially unbalanced selection processes show the appropriate subset characteristics, when assayed in the periphery with conventional antigens as stimuli. Both CD8-disrupted animals and animals lacking Pz-microglobulin (and thus surface class I MHC as well) are deficient in CTL activity, but not in TH activity, and both CD4-disrupted animals and class II-negative animals are selectively deficient in help (Zijlstra et al., 1990; Fung-Leung et al., 1991; Rahemtulla et al., 1991; Cosgrove et al., 1991). The durability of the association argues for a fundamental link between receptor choice and nuclear programming for function. How this link is created is a mystery and a challenge. The stages in the process are not yet well studied, but one can envision three different kinds of linkages. First is the possibility that the same signal that selectively shuts off either CD4 or CD8 expression in the thymic cortex also selectively targets a subset of response genes for derepression, presumably from among those that were inducible prior to TCR rearrangement (see Section IV). Second, functional specialization might occur in a process temporally distinct from positive selection, during the long period of residence in the medulla. In this case the signals and selective criteria might be totally unrelated to positive selection per se. Third, functional specialization could occur as a direct consequence of the utilization of CD4 versus CD8 as coreceptors, if these two receptor-Lck complexes actually generate different costimulatory signals in a response to antigen. This last mechanism could operate even in cells that had already been exported from the thymus, for-at least in the human-many peripheral T cells are only poorly responsive prior to their first antigen stimulation (Byrne et al., 1988; Sanders et al., 1988; Koulova et al., 1990).Thus, it is difficult to know exactly which responses, if any, they may yet be committed to make. We have discussed some mechanistic aspects of such models elsewhere (Rothenberg et al., 1991). The evidence reviewed in Section IV clearly disfavors any model in which acquisition of a single-positive TCRh’gh surface phenotype is coincident with full functional maturation. I have already discussed the temporal lag in the appearance of effector competence (J.-F. Chang et al., 1991a,b) and conditions under which positive selection can be seen to take place without functional maturation, e.g., in radiation chimeras (Amagai et al., 1987; R. A. Diamond and E. V. Rothenberg, unpublished results). We do not know, however, when functional commitment takes place as distinct from manifestation of competence. An obstacle to testing for commitment is the quantitative,

188

ELLEN V. ROTHENBERG

rather than qualitative, difference seen between the levels of expression of any given “response” gene in CD4+ and CD8+ cells, even when they are fully mature peripheral T cells. There is ample evidence for inducibility of subset-“specific” response genes in mature T cells of the “inappropriate” lineage, although usually at lower levels than in the appropriate subset (Miller and Stutman, 1982; von Boehmer et al., 1984; Podack et al., 1988; Tsuchida and Sakane, 1988; Gajewski et al., 1989; Rothenberg et al.,1991). Our own studies of the inducibility of the IL-2 gene in mature CD4+ and CD8+ peripheral T cells indicated that signal transduction details can dictate which of the response genes inducible in a cell will actually be expressed in any particular encounter with antigen, affecting whether CD8+ cells will make IL-2 at all, and whether THO-like cells will make both IL-2 and IL-4 or IL-4 alone (McGuire et al., 1988; Novak and Rothenberg, 1990). Thus a real possibility exists that the signals caused by engagement of CD4 and CD8 may be sufficiently different that they themselves trigger the preferential activation of different sets of response gene upon encounter with antigen and APC. Because among the response gene products of T cells are lymphokine receptors, differential induction of such receptors in the initial stimulation might result in differential sensitivity to a some lymphokine-mediated signal that had in turn qualitative effects on a secondary stimulation. This influence could be repeated in every individual round of stimulation or might eventually alter the cell’s physiology in a cumulative, irreversible way. Further work will be needed to test whether a discrete nuclear programming event or the effects of different cytoplasmic signaling molecules account for the divergences seen between CD4+8- and CD4-8+ cell responses. Enormous progress has been made in understanding repertoire selection, but the very nature of functional maturation, as well its regulation, remains largely unknown. We can quantitatively monitor the proliferation and differentiation of cells with different TCRs, but we know nothing of whether the underlying programming for particular functional responses is induced in a recognition-dependent way, or rather determined and selected independently. The weak effector responses of medullary thymocytes and recent thymic emigrants may obscure any selection processes that may sort cells on the basis of functional commitment. Indeed, after tracing so many stages of differentiation that CD4+ and CD8+ mature T cells presumably share, and after acknowledging the many genes that both can express under certain conditions, it is striking once again to confront the fact that in their ensemble of properties


189

they are really quite different cell types. After several rounds of stimulation, their response gene programs become as distinct as the gene expression programs of mast cells and NKcells. They express different populations of K+ channels (Lewis and Cahalan, 1988a,b).They regulate their proliferation behavior and IL-2R subunit expression differently (Crispe et al., 1985; Gullberg and Smith, 1986; Tsuchida and Sakane, 1988; Yang et al., 1988b; Hattori et al., 1990). CD8+ CTLs can lose antigen specificity easily, drifting into NK-like killing and/or pure IL-2 responsiveness (Brooks, 1983; Nabholz and MacDonald, 1983; Shortman et al., 1984; Havele et al., 1986), whereas CD4+ cells are generally rigidly dependent on a precise density of TCR-cross-linking interactions and a regular costimulus from a well-chosen source of accessory cells (Erard et al., 1985; Dupuy d'Angeac et al., 1986).As we have also noted, many CD8+ cells can develop extrathymically, sharing this and other features with NK cells and many TCRyG cells. By contrast, CD4+ T H cells are almost completely thymus dependent. This raises the question of whether the developmental separation of CD4-type cells from CD8-type cells really occurs as late as the positive-selection-mediated adoption of a CD4+ or CD8+ surface phenotype. If the developmental choice of commitment to a CD4' or CD8+ lineage is only made at the time of positive selection in the thymus, long after most response gene expression capabilities have been acquired and, by definition, after arrival in the thymus, it is difficult to explain the extent of these other developmental and regulatory asymmetries. CD4+ and CD8+ T cells share a TCR structure and many features of a selection history. However, the divergent richness of their cellular responses remains an inadequately explored terrain with many secrets. VI. Summary

The work reviewed in this article separates T cell development into four phases. First is an expansion phase prior to TCR rearrangement, which appears to be correlated with programming of at least some response genes for inducibility. This phase can occur to some extent outside of the thymus. However, the profound T cell deficit of nude mice indicates that the thymus is by far the most potent site for inducing the expansion per se, even if other sites can induce some response acquisition. Second is a controlled phase of TCR gene rearrangement. The details of the regulatory mechanism that selects particular loci for rearrangement are still not known. It seems that the rearrangement of the

190

ELLEN V. ROTHENBERG

TCRy loci in the y6 lineage may not always take place at a developmental stage strictly equivalent to the rearrangement of TCRP in the ap lineage, and it is not clear just how early the two lineages diverge. In the TCRaP lineage, however, the final gene rearrangement events are accompanied by rapid proliferation and an interruption in cellular response gene inducibility. The loss of conventional responsiveness is probably caused by alterations at the level of signaling, and may be a manifestation of the physiological state that is a precondition for selection. Third is the complex process of selection. Whereas peripheral T cells can undergo forms of positive selection (by antigen-driven clonal expansion) and negative selection (by abortive stimulation leading to anergy or death), neither is exactly the same phenomenon that occurs in the thymic cortex. Negative selection in the cortex appears to be a suicidal inversion of antigen responsiveness: instead of turning on IL-2 expression, the activated cell destroys its own chromatin. The genes that need to be induced for this response are not yet identified, but it is unquestionably a form of activation. It is interesting that in humans and rats, cortical thymocytes undergoing negative selection can still induce IL-2Ra expression and even be rescued in vitro, if exogenous IL-2 is provided. Perhaps murine thymocytes are denied this form of rescue because they shut off IL-2RP chain expression at an earlier stage or because they may be uncommonly Bcl-2 deficient (cf. Sentman et al., 1991; Strasser et al., 1991). Even so, medullary thymocytes remain at least partially susceptible to negative selection even as they continue to mature. As we have just discussed in the previous section, positive selection is an exceptionally important developmental transition, but its mechanism has no clear equivalent in mature T cells. None of the alterations induced by positive selection are defined yet in terms of changes in gene expression, with the possible exception of CD4 or CD8 down-regulation, Understanding this profound mechanism for the determination of cell fate will depend on identifying its molecular targets, not to mention identifying more of the molecular indices of its effects. As we already noted, a critical area of uncertainty is the range of developmental stages in which a thymocyte remains positively selectable. This question will need to be resolved in order to perceive accurately the relative regulation of the opposing selection processes. The fourth stage is the reconversion of the positively selected TCRaP cell into a responsive effector cell. The nonresponsiveness imposed during the final stages of TCR rearrangement must be reversed. In addition, the CD4+ and CD8+ medullary cells must acquire


191

the appropriate patterns of responsiveness to fit their recognition specificities. The average residence time of thymocytes in the medulla after positive selection is comparable to the entire length of time that has elapsed before this to differentiate from pluripotent stem cells through positive selection. Thus, there is time for many further transformations to occur. As we noted above, it is not at all clear whether the divergence between helper and killer types takes place at the same time as the divergence between CD4+8- and CD4-8+ coreceptor utilization patterns. It may take place much earlier, necessitating a functional selection process after receptor selection. Alternatively, it may not be right to consider functional response programming in terms of “commitment” or “divergence” at all. If TCR-CD4 coengagement transmits a different biochemical signal than TCR-CD8 coengagement, then different response genes may b e activated directly. The whole cascade of resulting effects could drive CD4+ and CD8+ cells to exercise widely different aspects of what may be their common lymphocyte response potential. To resolve these basic issues of developmental biology, much more will need to be learned about the molecular mechanisms that dictate the choice of genes utilized in lymphocyte responses. Beyond providing this detailed list of transitional stages, the work reviewed here raises several questions of mechanism and regulation that are not yet readily answered. One is the lineage relationship between CD4+ TCRaP cells, CD8+ TCRaP cells, and TCRyG cells. Another related question is the regulatory basis for allocation of different functions to T cell subsets. The third is the relationship between T and non-T hematopoietic cells. Finally, there is the question of the developmental significance of the thymus. The use of the TCR structure to define cells as T cells has Ied to the general view that all TCRaP cells are more closely related to each other than they are to TCRyG cells, and both TCR classes more closely related to each other than to any TCR- cell type. It is worth noting nonetheless that the dauntingly prolific, wasteful thymic population dynamics have thus far made it impossible to prove that CD4+ and CD8+ cells arise from a common precursor at any later stage than that of the most primitive repopulating stem cells (Kingston et al., 1985; Spangrude and Scollay, 1990). The imposition of a transgenic TCR of known specificity has dramatic effects on the survival of different classes of progeny, but cannot be proved to change the commitment of any individual immature or cortical thymocyte to a CD4+ or CD8+ lineage. The issue of relatedness is problematic from a developmental point of view because, when characteristics other than TCRs are con-

192

ELLEN V. ROTHENBERG

sidered, the bulk of CD8+ TCRaP+ cells have more in common with many TCRy6’ cells than they do with many CD4+ TCRa/3+ cells. It is conceivable that a “CD8”-like lineage diverges from a “CD4”-like lineage before TCR gene rearrangement even occurs. In this case, to account for the observed TCR activation, we would simply have to postulate that the “CD8”-like lineage expresses transcription factors allowing germ-line transcription of all four TCR loci, whereas the “CD4”-like lineage expresses silencing factors for y and/or 6 and enhancing factors for a and P only. Such a model would be favored by any cladistic tree that gave as much weight to each of the differences in signaling, growth control, and thymus dependence that distinguish CD8+ from CD4+ cells as to the difference between TCRaP and TCRy6 structures. It may not be entirely fair to weight these response differences so heavily, however, because we still do not know the extent to which they are all interdependent. After all, even within the CD4’ TCRa/3+ lineage, it appears that altered activation conditions and stimulation histories can make different sets of response genes expressible at different times (Gajewski et al., 1989; Swain et al., 1991; Vitetta et al., 1991). As already noted above, CD4-linked Lck and CD8-linked Lck may very well give different costimulatory signals in spite of the fact that the differences are as yet uncharacterized at a biochemical level. Thus, we cannot yet rule out the possibility that all the response differences between these cell types are direct downstream consequences of the activation of a CD8-dependent transcription factor as opposed to a CD4-dependent one. Great progress is likely to be made to resolve this issue in the near future as the components of the regulatory machinery for various response genes-IL-2R/3, IL-2Ra, IL-4, various granzymes, IL-5, and perforin-become defined. The extent to which antigen recognition really does lead to different signals in CD4+ and CD8+ cells can be determined with further insight into the substrates of the initial “layer” of kinases that mediate T cell activation. By simply considering the logical relationships of T cells on the basis of properties other than TCR expression, we approach the question of whether T cells are distinguished from the continuum of other hematopoietic cell types by anything other than their TCR structures. This need not be considered in terms of historical relatedness within a real cell lineage: I mean primarily to address how deep a hierarchy of developmental choices a hematopoietic cell must descend in becoming a T or a non-T cell. Throughout this article, I have noted the very close correlation between the properties of CD8+ T cells and NK cells. The similarity extends beyond the accessible set of response genes in


193

these two lineages to include the constitutive expression of IL-2RP chains and even the use of CD2, CD35, and Lck as triggering molecules. A grouping of TCRy6+ cells, NK cells, and TCRaP' CD8+ cells as variations within a basic hematopoietic cell class, separate from the TCRaP+ CD4+ cells, would not be unreasonable. The exceptions among the TCR+ members of this class further support such a division. Both CD8+ TCRaP' and TCRy6+ cells include lineages or sublineages that can develop extrathymically. Also, within the TCRy6 class are subsets that span the range from cells with highly diverse, selectable TCRs to subsets with essentially invariant receptors. The latter are presumably used very much like organ-specific NK cells. There is no simple way to reclassify these cell types to improve their hierarchy of relatedness: TCRaP+ CD8+ cells, NK cells, and TCRy6+ cells all have extrathymic subsets, and NK cells and TCRy6+ cells both have subsets with invariant receptors, but the TCRy&+cells with the most invariant receptors are also thymus dependent, not derived extrathymically like NK cells. This raises the possibility that the production of various TCRyG subsets, TCRaP cells, and NK cells may not be regulated in a hierarchical way. For example, as found in tissue culture experiments, a pre-T cell may be competent to turn on its proliferative and cytolytic machinery even if it has not yet finished, or even begun, using its VDJ recombinases. A relatively plastic differentiation scheme like this can also serve well to produce the rather surprisingly diverse patterns of TCRaP and TCRy6 cell utilization that are observed in different avian and mammalian species. The existence of extrathymic T cells with apparently full response capabilities begs the question of the role of the thymus in development per se. In a formal sense, passage through the thymus is not strictly required for most lymphokine production capability, expression of a CD3g-based triggering receptor, targeting of TCR a,P, y , and 6 genes for rearrangement, activation of VDJ recombinases and terminaltransferase-dependent mutagenesis, or acquisition of killing function. However, there is an enormous quantitative difference between the yields of T cells in normal and thymus-deprived animals, indicating that the thymus is a particularly efficient inducing microenvironment for pre-T cell mitogenesis. Its potent diversion of pluripotent hematopoietic stem cells to T cell developmental fates (Spangrude and Scollay, 1990) is further evidence for the dominant instructive signals that the thymic epithelium provides. Almost nothing is known yet about the molecular nature of these signals. Thus, the attempts in a number of laboratories to develop defined in uitro systems to reproduce thymic cell-cell interactions will be critically important for

194

ELLEN V. ROTHENBERG

the ultimate progress of the field. I have not attempted to cover these studies in the present review, for their results are still difficult to fit into a simple consensus picture, Nevertheless, from the ultimate success of in vitro reconstruction studies would come the essential tools with which thymocyte development could be dissected into a clear sequence of definable, accessible molecular events. Not only the immunology of T cell repertoire selection, but also the cellular, developmental, and molecular biology of T cell function are likely to remain fertile, challenging, and rewarding fields of study for many years to come. ACKNOWLEDGMENTS I wish first of all to express my gratitude to the many colleagues who shared with me their unpublished and published work. I was overwhelmed by their generosity and regret that the scope of this work could not be large enough to do them all justice. I hope to be forgiven for my omissions, In particular, I would like to thank Alfred Bendelac, Melvin Bosma, Rhodri Ceredig, Ray Daynes, B. J. Fowlkes, Douglas Green, Barton Haynes, Steven Hedrick, Lee Hood, Thomas Hiinig, Eric Jenkinson, Stanley Korsmeyer, Tak Mak, Carlos Martinez-A, Diane Mathis, Werner Miiller, John Ransom, Ronald Palacios, Martine Papiernik, Drew Pardoll, Roger Perlmutter, Jean Plum, David Raulet, Stuart Schlossman, Roland Scollay, Ken Shortman, Albert Singer, Susan Swain, Susumu Tonegawa, Roger Tsien, Harald von Boehmer, Gary Waanders, and Albert Zlotnik for their help. The interpretations throughout this work are my own, and I apologize for cases where I have misstated or misunderstood the results of the original authors. The manuscript was greatly improved by the intensive reading, critical comments, and encouragement of my colleague Eric Davidson. A later, painstaking reading by my associate Rochelle Diamond helped to clarify the language and the logic throughout the text. At moments of need, Patricia White of my laboratory also provided valuable assistance. The artwork was created expertly by Dorothy Lloyd. I thank all of them for their much appreciated efforts to help in this large undertaking. I am especially indebted to Cathy Blagg, who has worked tirelessly, swiftly, and skillfully on this manuscript day after day throughout its composition. Her editorial skill, high standards, and spirit of enterprise have made her partnership integral to the project. She has my deep gratitude. Work in the author’s laboratory was supported by two grants from the USPHS-CA 39605 and A1 19752-and also by funds from the Lucille P. Markey Charitable Trust. The cell sorting and microchemical facilities used in work cited here were supported by a USPHS Cancer Center Core Grant CA 32911 and by additional funds from the Markey Trust.

REFERENCES Abraham K. M., Levin, S. D., Marth, J. D., Forbush, K. A., and Perlmutter, R. M. (1991). J . E x p . Med. 173, 1421. Abramson, S., Miller, R. G., and Phillips R. A. (1977).J. E x p . Med. 145, 1567. Adkins, B., Tidmarsh, G. F., and Weissman, I. L. (1988). Zmmunogenetics 27, 180. Akbar, A. N., Salmon, M., and Janossy, G . (1991). Immunol. Today 12,184.


195

Albert, F., Hua, C., Truneh, A., Pierres, M., and Schmitt-Verhufst, A.-M. (1985). J . Immunol. 134,3649. Allison, J. P., and Havran, W. L. (1991).Annu. Rev. Immunol. 9,679. Altman, A., Coggeshall, K. M., and Mustelin, T. (1990).Ado. Immunol. 48,227. Amagai, T., Kina, T., Hirokawa, K.-I., Nishikawa, S.-I., Imanishi, J., and Katsura, Y. (1987).J . Immunol. 139,358. Anderson, P., Caligiuri, M., Ritz, J., and Schlossman, S. F. (1989).Nature (London)341, 159. Ashwefl, J. D., and Klausner, R. D. (1990).Annu. Reu. Immunol. 8, 139. Aspinall, R., Kampinga, J., and van den Bogaerde, J. (1991). Immunol. Today 12,7. Ballard, D. W., Bohnlein, E., Hoffman, J. A., Bogerd, H. P., Dixon, E. P., Franza, B. R., and Greene, W. C. (1989). New Biol. 1,83. Ballas, Z. K., and Rasmussen, W. (1990).J. Immunol. 145, 1039. Bandeira, A., Itohara, S., Bonneville, M., Burlen-Defranoux, O., Mota-Santos, T., Coutinho, A., and Tonegawa, S. (1991). Proc. Natl. Acad. Sci. U.S.A.88,43. Barber, E. K., Dasgupta, J. D., Schlossman, S. F., Trevillyan, J. M., and Rudd, C. E. (1989).Proc. Natl. Acad. Sci. U.S.A.86,3277. Bauer, A., McConkey, D. J., Howard, F. D., Clayton, L. K., Novick, D., Koyasu, S., and Reinherz, E. L. (1991).Proc. Natl. Acad. Sci. U.S.A.88,3842. Ben Aribia, M.-H., MoirC, N., MCtivier, D., Vaquero, C., Lantz, O., Olive, D., Charpentier, B., and Senik, A. (1989).J. Immunol. 142,490. Bendelac, A., and Schwartz, R. H. (1991). Nature (London)353,68. Benoist, C., and Mathis, D. (1989). Cell (Cambridge,Mass.) 58, 1027. Ben-Sasson, S. Z., LeGros, G., Conrad, D. H., Finkelman, F. D., and Paul, W. E. (1990). J . Immunol. 145, 1127. Benveniste, P., Chadwick, B. S., Miller, R. G., and Reimann, J. (1990).J . Immunol. 144, 411. Berg, L. J . , Pullen, A. M., Fazekas de St. Groth, B., Mathis, D., Benoist, C., and Davis, M. M. (1989a). Cell (Cambridge,Mass.) 58, 1035. Berg, L. J., Fazekas de St. Groth, B., Pullen, A. M., and Davis, M. M. (1989b). Nature (London)340,559. Berg, L. J., Frank, G. D., and Davis, M. M. (1990).Cell (Cambridge,Mass.)60, 1043. Berridge, M. J., and Irvine, R. F. (1989).Nature (London)341, 197. Betz, M., and Fox, B. A. (1991).J.Immunol. 146, 108. Biassoni, R., Ferrini, S., Prigione, I., Moretta, A,, and Long, E. 0. (1988).J. Immunol. 140,1685. Bierer, B. E., Sleckman, B. P., Ratnofsky, S. E., and Burakoff, S. J. (1989). Annu. Reu. Immunol. 7,579. Bierer, B. E., Mattila, P. S., Standaert, R. F., Herzenberg, L. A,, Burakoff, S. J., Crabtree, G., and Schreiber, S. L. (1990). Proc. Natl. Acad. Sci. U.S.A.87,9231. BiII, J., and Palmer, E. (1989) Nature (London)341,649. Biron, C. A., Van Den Elsen, P., Tutt, M. M., and Terhorst, C. (1987)J . Immunol. 139, 1704. Blackman, M. A,, Gerhard-Burgert, H., Woodland, D. L., Palmer, E., Kappler, J. W., and Marrack, P. (1990). Nature (London)345,540. Blackman, M. A., Finkel, T. I%, Kappler, J., Canibier, J., and Marrack, P. (1991) Proc. Natl. Acad. Sci. U.S.A.88,6682. Blackwell, T. K., Moore, M. W., Yancopoulos, G. D., Suh, H., Lutzker, S., and Alt, F. W. (1986).Nature (London)324,585. Blomgren, H., and Andersson, B. (1971).Cell Immunol. 1,545.

196

ELLEN V. ROTHENBERG

Blue, M.-L., Daley, J. F., Levine, H., and Schlossman, S. F. (1986).Eur.1. Imniunol. 16, 771. Blue, M.-L., Daley, J. F., Levine, H., Craig, K. A., and Schlossman, S . F. (1987a). J. Immunol. 138,3108. Blue, M.-L., Daley, J. F., Levine, H., Craig, K. A, and Schlossman, S. F. (1987b). J. Immunol. 139, 1065. Bockenstedt, L. K., Goldsmith, M. A,, Dustin, M., Olive, D., Springer, T. A,, and Weiss, A. (1988).J.Immunol. 141,1904. Boehm, T., Gonzalez-Sarmiento, R., Kennedy, M., and Rabbitts, T. H. (1991). Proc. Natl. Acad. Sci. U.S.A.88,3927. Boersma, W., Betel, I., and van der Westen, G. (1979).Eur. J. Immunol. 9,45. Bohjanen, P. R., Okajima, M., and Hodes, R. J. (1990). Proc. Natl. Acad. Sci. U.S.A.87, 5283. Bonifacino, J. S., McCarthy, S. A., Maguire, J. E., Nakayama, T., Singer, D. S., Klausner, R. D., and Singer, A. (1990).Nature (London) 344,247. Bonneville, M., Ishida, I., Mombaerts, P., Katsuki, M., Verbeek, S., Berns, A,, and Tonegawa, S. (1989). Nature (London)342,931. Bonneville, M., Itohara, S., Krecko, E. G., Mombaerts, P., Ishida, I., Katsuki, M., Berns, A., Farr, A. G., Janeway C. A., Jr., and Tonegawa, S. (1990a). J . E x p . M e d . 171, 1015. Bonneville, M., Ishida, I., Itohara, S., Verbeek, S., Berns, A., Kanagawa, O., Haas, W., and Tonegawa, S. (1990b).Nature (London)344,163. Borgulya, P., Kishi, H., Miiller, U., Kirberg, J., and von Boehmer, H. (1991).E M B O J . 10, 913. Born, W., Rathbun, G., Tucker, P., Marrack, P., and Kappler, J. (1986).Science 234,479. Born, W. K., McDuffie, M., Roehm, N., Kushnir, E., White, J., Thorpe, D., and Marrack, P. (1987).J.Zmmunol. 138,999. Born, W. K., O’Brien, R. L., and Modlin, R. L. (1991). FASEBJ. 5,2699. Bosma, M. J., and Carroll, A. M. (1991).Annu. Rev. lmmunol. 9,323. Bottomly, K., Luqman, M., Greenbaum, L., Carding, S., West, J., Pasqualini, T., and Murphy, D. B. (1989). Eur.J.Immunol. 19,617. Boyd, R. L., and Hugo, P. (1991). Immunol. Today 12,71. Boyer, P. D., and Rothenberg, E. V. (1988).J.Immunol. 140,2886. Boyer, P. D., Diamond, R. A., and Rothenberg, E. V. (1989).J.Zmmunol. 142,4121. Brodsky, F. M., and Guagliardi, L. E. (1991).Annu. Reu. Immunol. 9,707. Brooks, C. G. (1983) Nature (London)305,155. Broome, H. E, Reed, J. C., Godillot, E. P., and Hoover, R. G. (1987). M o l . Cell. Biol. 7, 2988. Brown, M. A., Pierce, J. H., Watson, C. J., Falco, J. Ihle, J. N., and Paul, W. E. (1987).Cell (Cambridge,Mass.) 50,809. Bucy, R. P., Chen, C.-L. H., Cihak, J., Losch, J., and Cooper, M. D. (1988).J. Zmmunol. 141,2200. Bucy, R. P., Chen, C.-L., and Cooper, M. D. (1989).J.Immunol. 142,3045. Bucy, R. P., Chen, C.-L. H., and Cooper, M. D. (199O).J.Immunol. 144,1161. Budd, R. C., Cerottini, J.-C., and MacDonald, H. R. (1986).J.Zmmunol. 137,3734. Budd, R. C., Cerottini, J.-C., and MacDonald, H. R. (1987a).J.Immunol. 138, 1009. Budd, R. C., Cerottini, J.-C., Horvath, C., Bron, C., Pedrazzini, T., Howe, R. C., and MacDonald, H. R. (1987b).J.Immunol. 138,3120. Burd, P. R., Rogers, H. W., Gordon, J. R., Martin, C. A., Jayaraman, S., Wilson, S. D., Dvorak, A. M., Galli, S. J., and Dorf, M. E. (1989).J. Erp. M e d . 170,245.


197

Burgess, K. E., Odysseos, A. D., Zalvan, C., Druker, B. J., Anderson, P., Schlossman, S. F., and Rudd, C. E. (1991) Eur.1. Zmmunol. 21, 1663. Burkly, L. C., Lo, D., and Flavell, R. A. (1990). Science 248, 1364. Byrne, J . A., Butler, J. L., and Cooper, M. D. (1988).J. Zmmunol. 141,3249. Campana, D., Thompson, J. S., Amlot, P., Brown, S., and Janossy, G. (1987).J. Zmmunol. 138,648-655. Campana, D., Janossy, G., Coustan-Smith, E., Amlot, P. L., Tian, W.-T., Ip, S., and Wong, L. (1989).J. Immunol. 142,57. Carding, S. R., and Bottomly, K. (1988).J. Zmmunol. 140, 1519. Carding, S. R., Jenkinson, E. J., Kingston, R., Hayday, A. C., Bottomly, K., and Owen, J. J. T. (1989). Proc. Natl. Acad. Sci. U.S.A.86,3342. Carding, S. R., Kyes, S., Jenkinson, E. J., Kingston, R., Bottomly, K., Owen, J. J. T., and Hayday, A. C. (1990). Genes Deu. 4, 1304. Carding, S. R., Hayday, A. C., and Bottomly, K. (1991). Immunol. Today 12,239. Carlson, L. M., Oettinger, M. A., Schatz, D. G., Masteller, E. L., Hurley, E. A., McCormack, W. T., Baltimore, D., and Thompson, C. B. (1991). Cell (Cambridge,Mass.)64, 201. Carrera, A. C., Baker, C. L., Roberts, T. M., and Pardoll, D. M. (1992a). Submitted for publication. Carrera, A. C., Roberts, T. M., and Pardoll, D. M. (199213). Submitted for publication. Carroll, A. M., and Bosma, M. J. (1991). Genes Deu. 5, 1357. Ceredig, R. (1986).J. Zmmunol. 137,2260. Ceredig, R. (1990). Int. Zmmunol. 2,859. Ceredig, R., Glasebrook, A. L., and MacDonald, H. R. (1982).J. Exp. Med. 155,358. Ceredig, R., Sekaly, R. P., and MacDonald, H. R. (1983a). Nature (London)303,248. Ceredig, R., Dialynas, D. P., Fitch, F. W., and MacDonald, H. R. (1983b).J. Exp. Med. 158,1654. Ceredig, R., Lowenthal, J. W., Nabholz, M., and MacDonald, H. R. (1985). Nature (London)314,98. Ceredig, R., Lynch, F., and Newman, P. (1987). Proc. Natl. Acad. Sci. U.S.A.84,8578. Chaffin, K. E., and Perlmutter, R. M. (1991). Eur.1. Zmmunol. 21,2565. Chang, H.-L., Lefrancois, L., Zaroukian, M. H., and Esselman, W. J. (1991).]. Zmmunol. 147,1687. Chang, J.-F., Thomas, C. A., 111, and Kung, J. T. (1991a).J. Zmmunol. 147,851. Chang, J.-F., Thomas, C. A., 111, and Kung, J. T. (1991b).J. Immunol. 147,860. Charbonneau, H., Tonks, N. K., Walsh, K. A,, and Fischer, E. H. (1988).Proc. NatLAcad. Sci. U.S.A. 85,7182. Chazen, G. D., Pereira, G. M. B., LeGros, G., Gillis, S., and Shevach, E. M. (1989). Proc. Natl. Acad. Sci. U.S.A.86,5923. Chen, D., and Rothenberg, E. V. (1992). In preparation. Chen, W.-F., Scollay, R., Clark-Lewis, I., and Shortman, K. (1983a). Thymus 5, 179. Chen, W.-F., Scollay, R., and Shortman, K. (1983b). Thymus 5,197. Chen, W.-F., Fischer, M., Frank, G., and Zlotnik, A. (1989).J. Zmmunol. 143,1598. Iwashima, M., Wettstein, D. A,, and Davis, M. M. (1987). Nature (London) Chien, Y.-H., 330,722. Choi, Y ., Herman, A., DiGiusto, D., Wade, T., Marrack, P., and Kappler, J. (1990).Nature (London)346,471. Chouaib, S., Welte, K., Mertelsmann, R., and Dupont, B. (1985).]. Immunol. 135, 1172. Chun, J. J. M., Schatz, D. G., Oettinger, M. A., Jaenisch, R., and Baltimore, D. (1991). Cell (Cambridge,Mass.)64, 189.

198

ELLEN V. ROTHENBERG

Churilla, A. M., and Braciale, V. L. (1987).J . Immunol. 138,1338. Churilla, A. M., Braciale, T. J., and Braciale, V. L. (1989).]. Exp. Med. 170, 105. Clayton, L. K., D’Adamio, L., Howard, F. D., Sieh, M., Hussey, R. E., Koyasu, S., and Reinherz, E. L. (1991). Proc. Natl. Acad. Sci. U.S.A. 88,5202. Clevers, H., Alarcon, B., Wileman, T., and Terhorst, C. (1988).Annu. Reo. Immunol. 6, 629. Cohen, J. J., and Duke, R. C . (1984).J. Immunol. 132,38. Colamonici, 0. R., Ang, S., Quinones, R., Henkart, P., Heikkila, R., Cress, R.,Felix, C., Kirsch, I., Longo, D., Marti, G., Seidman, J. G., and Neckers, L. M. (1988).J . Immunol. 140,2527. Coltey, M., Jotereau, F. V., and LeDouarin, N. M. (1987). Cell D$feer. 22,71. Coltey, M., Bucy, R. p., Chen, C. H., Cihak, J., Losch, U., Char, D., LeDouarin, N. M., and Cooper, M. D. (1989).J . E x p . Med. 170,543. Cooke, M. P., Abraham, K. M., Forbush, K. A,, and Perlmutter, R. M. (1991) Cell (Cambridge,Mass.) 65,281. Cordier, A. C., and Haumont, S. M. (1980).Am.]. Anat. 157,227. Cosgrove, D., Gray, D., Dierich, A., Kaufman, J., Lemeur, M., Benoist, C., and Mathis, D. (1991). Cell (Cambridge,Mass.) 66, 1051. Crispe, I. N., Bevan, M. J., and Staerz, U. D. (1985) Nature (London)317,627. Crispe, I. N., Moore, M. W., Husmann, L. A., Smith, L., Bevan, M. J., and Shimonkevitz, R.P. (1987) Nature (London)329,336. Cron, R. Q., Gajewski, T. F., Sharrow, S. O., Fitch, F. W., Matis, L. A., and Bluestone, J. A. (1989).J.Immunol. 142,3754. Davidson, E. H. (1990). Deoelopment (Cambridge, U K ) 108,365. Davis, M. M., and Bjorkman, P. J. (1988). Nature (London)334,395. Daynes, R. A., Dowell, T., and Araneo, B. A. (1991).J.E x p . Med. 174,1323. DeFranco, A. L. (1991). Nature (London)351,603. De la Hera, A,, Toribio, M. L., and Martinez-A., C. (1989). Intl. Immunol. 1,496. Denning, S. M., Jones, D. M., Ware, R. E., Weinhold, K. J., Brenner, M. B., and Haynes, B. F. (1991). l n f .Immunol. 3, 1015. Dent, A. L., Matis, L. A,, Hooshmand, F., Widacki, S. M., Bluestone, J. A., and Hedrick, S. M. (1990). Nature (London)343,714. Desai, D. M., Newton, M. E., Kadlecek, T., and Weiss, A. (1990). Nature (London) 348,66. Desiderio, S . , Yancopoulos, G., Paskind, M., Thomas, E., Boss, M., Landau, N., Ah, F., and Baltimore, D. (1984). Nature (London)311,752. Deusch, K., Daley, J. F., Levine, H., Languet, A. J., 111, Anderson, P., Schlossman, S. F., and Blue, M.-L. (199O).J.Immunol. 144,2851. Doyle, C., and Strominger, J. L. (1987). Nature (London)330,256. Droege, W., Zucker, R.,and Jauker, U. (1974). Cell. Immunol. 12,173. Dumont, F. J., Staruch, M. J., Koprak, S. L., Melino, M. R.,and Sigal, N. H. (1990). J . Immunol. 144,251. Dunn, D. E., Herold, K. C., Otten, G. R., Lancki, D. W., Gajewski, T., Vogel, S . N., and Fitch, F. W. (1987).J. Immunol. 139,3942. Duplay, P., Lancki, D., and Allison, J. P. (1989).J . Immunol. 142,2998. Dupuy-DAngeac, A. D., Monis, M., and RBme, T. (1986).J.Immunol. 137,3501. Egerton, M., Scollay, R.,and Shortman, K. (1990). Proc. Natl. Acad. Sci. U.S.A. 87, 2579. Elliott, J. F., Rock, E. P., Patten, P.A., Davis, M. M., and Y.-H. Chien. (1988). Nature (London)331,627.


199

Emmel, E. A., Verweij, C. L., Durand, D. B., Higgins, K. M., Lacy, E., and Crabtree, G. R. (1989). Science 246, 1617. Erard, F., Nabholz, M., Dupuy-D’Angeac, A., and MacDonald, H: R. (1985)J E x p . Med. 162,1738. Evavold, B. D., and Allen, P. M. (1991). Science 252, 1308. Ewing, T., Egerton, M., Wilson, A,, Scollay, and Shortman, K. (1988).Eur. J . Zmmunol. 18,261. Ezine, S., Papiernik, M., and Lepault, F. (1991). Znt. Zmmunol. 3,237. Farrar, W. L., Johnson, H. M., and Farrar, J. J. (1981).J.Zmmunol. 126, 1120. Fenton, R. G., Marrack, P., Kappler, J. W., Kanagawa, O., and Seidman, J. G. (1988). Science 241, 1089. Fernandez-Botran, R., Sanders, U. M., Mosmann, T. R., and Vitetta, E. S. (1988).]. E x p . Med. 168,543. Ferrick, D. A., Sambhara, S. R., Ballhausen, W., Iwamoto, A., Pircher, H., Walker, C. L., Yokoyama, W. M., Miller, R. G., and Mak, T. W. (1989). Cell (Cambridge, Mass.) 57, 483. Ferrick, D. A., Chan, A,, Rahemtulla, A., Widacki, S. M., Xia, M., Broughton, H., Gajewski, D. A., Ballhausen, W., Allison, J. p., Bluestone, J. A., Burki, K., van Ewijk, W., and Mak, T. W. (1990).J . Zmmunol. 145,20. Ferrick, D. A., Sydora, B., Wallace, V., Gemmell-Hori, L., Kronenberg, M., and Mak, T. W. (1991). Zmmunol. Reu. 120,51. Ferris, D. K., Willette-Brown, J., Ortaldo, J. R., and Farrar, W. L. (1989).J.Immunol. 143, 870. Fiering, S., Northrop, J. P., Nolan, G. P., Mattila, P. S., Crabtree, G. R., and Herzenberg, L. A. (1990). Genes Dew. 4, 1823. Finkel, T. H., McDuffie, M., Kappler, J. W., Marrack, P., and Cambier, J. C. (1987). Nature (London)330,179. Finkel, T. H., Marrack, P., Kappler, J. W., Kubo, R. T., and Cambier, J. C. (1989a). J . Zmmunol. 142,3006. Finkel, T. H., Cambier, J. C., Kubo, R. T., Born, W. K., Marrack, P., and Kappler, J. W. (198913).Cell (Cambridge,Mass.) 58, 1047. Finkel, T. H., Kubo, R. T., and Cambier, J. C. (1991). Zmmunol. Today 12,79. Firestein, G. S., Roeder, W. D., Laxer, J. A., Townsend, K. S., Weaver, C. T., Hom, J. T., Linton, J., Torbett, B. E., and Glasebrook, A. L. (1989).J.Zmmunol. 143,518. Fischer, M., MacNeil, I., Suda, T., Cupp, J. E., Shortman, K., and Zlotnik, A. (1991). J . Zmmunol. 146,3452. Flanagan, W. M., Corthksy, B., Bram, R. J., and Crabtree, G. R. (1991).Nature (London) 352,803. Fleury, S., Lamarre, D., Meloche, S., Ryu, S.-E., Cantin, C., Hendrickson, W. A., and Sekaly, R.-P. (1991). Cell (Cambridge,Mass.) 66, 1037. Fowlkes, B. J., and Pardoll, D. M. (1989).Adu. Zmmunol. 44,207. Fowlkes, B. J., Edison, L., Mathieson, B. J., and Chused, T. M. (1985).J.E x p . Med. 162, 802. Fowlkes, B. J., Kruisbeek, A. M., Ton-That, H., Weston, M. A., Coligan, J. E., Schwartz, R. H., and Pardoll, D. M. (1987). Nature (London) 329,251. Fowlkes, B. J., Schwartz, R. H., and Pardoll, D. M. (1988). Nature (London) 334,620. Fox, D. A., Hussey, R. E., Fitzgerald, K. A., Bensussan, A., Daley, J. F., Schlossman, S. F., and Reinherz, E. L. (1985).J . Immunol. 134,330. Frank, S. J., Niklinska, B. B., Orloff, D. G., Mercep, M., Ashwell, J. D., and Klausner, R. D. (1990). Science 249, 174.

200

ELLEN V. ROTHENBERG

Fraser, J. D., Irving, B. A., Crabtree, G. R.,and Weiss, A. (1991). Science 251,313. Fredrickson, G. G., and Basch, R. S. (1989).J.E x p . Med. 169,1473. Freire-Moar, J., Cherwinski, H., Hwang, F., Ransom, J., and Webb, D. (1991).J.Zmmunol. 147,405. Fulop, G. M., and Phillips, R.A. (1990). Nature (London)347,479. Fung, M. R.,Scearce, R.M., Hoffman, J. A,, Peffer, N. J., Hammes, S. R.,Hosking, J. B., Schmandt, R.,Kuziel, W. A,, Haynes, B. F., Mills, G. B., and Greene, W. C. (1991). J . Zmmunol. 147, 1253. Fung-Leung, W.-P., Schilham, M. W., Rahemtulla, A,, Kundig, T. M., Vollenweider, M., Potter, J., van Ewijk, W., and Mak, T. W. (1991). Cell (Cambridge,Mass.) 65,443. Furley, A. J., Mizutani, S., Weilbaecher, K., Dhaliwal, H. S., Ford, A. M., and Greaves, M. F. (1986). Cell (Cambridge, Mass.) 46,75. Gajewski, R. F., Schell, S. R.,Nau, G., and Fitch, F. W. (1989). Zmmunol. Rev. 111,79. Gajewski, R. F., Schell, S. R.,and Fitch, F. W. (199O).J.Zmmunol. 144,4110. Gao, E.-K., Lo, D., Cheney, R., Kanagawa, O., and Sprent, J. (1988). Nature (London) 336, 176. Garman, R.D., Doherty, P. J., and Raulet, D. H. (1986).Cell (Cambridge, Mass.)45,733. Gaulton, G. N., and Eardley, D. D. (1986).J. Zmmunol. 136,2470. Gilbert, K. M., Hoang, K. D., and Weigle, W. 0. (199O).J.Zmmunol. 144,2063. Gilliland, L. K., Teh, H. S., Uckun, F. M., Norris, N. A., Teh, S.-J., Schieven, G. L., and Ledbetter, J. A. (1991).J. Zmmunol. 146, 1759. Gimmi, C. D., Freeman, G. J., Gribben, J. G., Sugita, K., Freedman, A. S., Morimoto, C., and Nadler, L. M. (1991). Proc. Natl. Acad. Sci. U.S.A.88,6575. Goldschneider, I., Komschlies, K. L., and Greiner, D. L. (1986).J.E x p . Med. 163,l. GonzBlez-FernBndez, A., Diaz-Espada, F., Kreisler, M., and Deza, F. G. (1991). Eur. J. Zmmunol. 21, 115. Goodman, T., and LeFranqois, L. (1988).Nature (London)333,855. Graber, M., Bockenstedt, L. K., and Weiss, A. (1991).J.Zmmunol. 146,2935. Granelli-Piperno, A., Andrus, L., and Steinman, R. M. (1986).J. E x p . Med. 163,922. Granelli-Piperno, A., Nolan, P., Inaba, K., and Steinman, R. M. (1990). J. E r p . Med. 172, 1869. Gray, L. S., Gnarra, J. R., Russell, J. H., and Engelhard, V. H. (1987). Cell (Cambridge, Mass.)50, 119. Greenbaum, L. A., Horowitz, J. B., Woods, A., Pasqualini, T., Reich, E. P., and Bottomly, K. (1988).J . Immunol. 140,1555. Gregoire, K. E., Goldschneider, I., Barton, R. W., and Bollum, F. J. (1979).J. Zmmunol. 123,1347. Groh, V., Porcelli, S., Fabbi, M., Lanier, L. L., Picker, L. J., Anderson, T., Warnke, R. A., Bhan, A. K., and Brenner, M. B. (1989).J. E x p . Med. 169,1277. Groh, V., Fabbi, M., and Strominger, J. L. (1990). Proc. Natl. Acad. Sci. U.S.A.87,5973. Grusby, M. J., Johnson, R. S., Papaioannou, V. E., and Glimcher, L. H. (1991). Science 253,1417. Guidos, C. J., Weissman, I. L., and Adkins, B. (1989). Proc. Natl. Acad. Sci. U.S.A.86, 7542. Guidos, C. J.,Danska, J. S., Fathman, C. G., and Weissman, I. L. (199O).J.E x p . Med. 172, 835. Gullberg, M., and Smith, K. A. (1986).J.Erp. Med. 163,270. Guy-Grand, D., Cerf-Bensussan, N., Malissen, B., Malassis-Seris, M., Briottet, C., and Vassalli, G. (1991).J.E x p . Med. 173, 471. Haars, R.,Kronenberg, M., Gallatin, W. M., Weissman, I. L., Owen, F. L., and Hood, L. (1986).J.E x p . Med. 164,l.


20 1

Haas, W., Kaufrnan, S., and Martinez-A, C. (1990). Zmmunol. Today 11,340. Hara, T., Fu, S. M., and Hansen, J. A. (1985).J.E x p . Med. 161, 1513. Hatakeyama, M., Kono, T., Kobayashi, N., Kawahara, A., Levin, S. D., Perlrnutter, R. M., and Taniguchi, T. (1991). Science 252,1523. Hattori, M., Okazaki, H., Ishida, Y.,Onurna, M., Kano, S., Honjo, T., and Minato, N. (1990).J . Immunol. 144,3809. Havele, C., Bleackley, R. C., and Paetkau, V. (1986).J. Immunol. 137, 1448. Havran, W. L., and Allison, J. P. (1988). Nature (London)335,443. Havran, W. L., Poenie, M., Kirnura, J., Tsien, R., Weiss, A., and Allison, J. P. (1987). Nature (London)330,170. Havran, W. L., Chien, Y.-H., and Allison, J. P. (1991). Science 252,1430. Hayakawa, K., and Hardy, R. R. (1988).J . E r p . Med. 168,1825. Haynes, B. F., Singer, K. H., Denning, S. M., and Martin, M. E. (1988).]. Zmmunol. 141, 3776. Haynes, B. F., Denning, S. M., Singer, K. H.,and Kurtzberg, J. (1989). Zmmunol. Today 10,87. Heckford, S. E., Gelrnann, E. P., Agnor, C. L., Jacobson, S., Zinn, S., and Matis, L. A. (1986).J.Zmmunol. 137,3652. Heilig, J. S., and Tonegawa, S. (1986). Nature (London)322,836. Hein, W. R., and Mackay, C. R. (1991). Zmmunol. Today 12,30. Held, W., MacDonald, H. R., and Mueller, C. (1990). Int. Zmmunol. 2,57. Hengel, H., Wagner, H., and Heeg, K. (1991).J.Immunol. 147,1115. Herman, A,, Kappler, J. W., Marrack, P., and Pullen, A. M. (1991).Annu. Reu. Zmmunol.. 9,745. He& A. D., Tutschka, P. J., and Santos, G . W. (1982).J.Immunol. 128,355. Hirokawa, K., Sado, T., Kubo, S., Karnisaku, H., Hitomi, K., and Utsuyama, M. (1985). J . Immunol. 134,3615. Hockenbery, D. M., Zutter, M., Hickey, W., Nahm, M., and Korsrneyer, S. J. (1991).Proc. Natl. Acad. Sci. U.S.A.88,6961. Hofrnan, F. M., Modlin, R. L., Bhoopat, L., and Taylor, C. R. (1985).J . Zmmunol. 134, 375 1. Hooton, J. W. L., Miller, C. L., Helgason, C. D., Bleakley, R. S., Gillis, S., and Paetkau, V. (199O).J.Zmmunol. 144,816. Hopper, K., and Shortrnan, K. (1976). Cell. Zmmunol. 27,256. Howe, R. C., and MacDonald, H. R. (1988).J.Zmmunol. 140, 1047. Howe, R. C., Lowenthal, J. W., and MacDonald, H. R. (1986).]. Zmmunol. 137,3195. Hoyos, B., Ballard, D. W., Bohnlein, E., Siekevitz, M., and Greene, W. C. (1989). Science 244,457. Huesrnann, M., Scott, B., Kisielow, P., and von Boehmer, H. (1991). Cell (Cambridge, Mass.) 66,533. Hugo, P., and Petrie, H. T. (1991).Adu. Cell Biol. (in press). Hugo, P., Boyd, R. L., Waanders, G. A,, Petrie, H. T., and Scollay, R. (1991). Znt. Immunol. 3,265. Huiskarnp, R., van W e t , E., and van Ewijk, W. (1985).J.Zmmunol. 134,2170. Hiinig, T. (1988). E u r . 1 . Zmmunol. 18,2089. Hiinig, T., and Mitnacht, R. (1991).1.E x p . Med. 173,561. Hiinig, T., Wallny, H.-J., Hartley, J. K., Lawetzky, A,, andTiefenthaler, G. (1989).J.E x p . Med. 169,73. Hurley, T. R., Luo, K., and Sefton, B. M. (1989). Science 245,407. Hurwitz, J. L., Sarnaridis, J., and Pelkonen, J. (1988). Cell (Cambridge, Mass.) 52, 821.

202

ELLEN V. ROTHENBERG

Husmann, L. A., Shimonkevitz, R. P., Crispe, I. N., and Bevan, M. J. (1988).J.Zmmunol. 141,736. Ikuta, K., Kina, T., MacNeil, I., Uchida, N., Peault, B., Chien, Y.-H., and Weissman, I. L. (1990). Cell (Cambridge,Mass.) 62,863. Imboden, J. B., Shoback, D. M., Pattison, G., and Stobo, J. D. (1986). Proc. Natl. Acad. Sci. U.S.A.83,5673. Irving, B. A., and Weiss, A. (1991). Cell (Cambridge, Mass.) 64,891. Isakov, N., and Altman, A. (1985).J . Imrnunol. 135,3674. Ishida, I., Verbeek, S., Bonneville, M., Itohara, S., Berns, A,, and Tonegawa, S. (1990). Proc. Natl. Acad. Sci. U.S.A.87,3067. Ito, K., Bonneville, M., Takagaki, Y.,and Tonegawa, S. (1989). Proc. Natl. Acad. Sci. U.S.A.86,631. Itohara, S., and Tonegawa, S. (1990). Proc. Natl. Acad. Sci. U.S.A.87,7935. Itohara, S., Nakanishi, N., Kanagawa, O., Kubo, R., and Tonegawa, S. (1989). Proc. Natl. Acad. Sci. U.S.A.86,5094. Jamieson, C., McCaffrey, P. G., Rao, A., and Sen, R. (1991).J.Zmmunol. 147,416. Janeway, C. (1991).Nature (London)349,459. Janeway, C. A., Jr. (1989). Zmmunol. Today 10,234. Jankovic, D. L., Rebollo, A., Kumar, A., Gibert, M., and ThBze, J. (199O).J.Immunol. 145, 4136. Janossy, G., Thomas, J. A,, Bollum, F. T., Granger, S., Pizzolo, G., Bradstock, K. F., and Wong, L. (1990).J . Immunol. 125,202. Jenkins, M. K., and Schwartz, R. H. (1987).J.E x p . Med. 165,302. Jenkins, M. K., Schwartz, R. H., and Pardoll, D. M. (1988).Science 241, 1655. Jenkinson, E. J., van Ewijk, W., and Owen, J. J. T. (1981).J.E x p . Med. 153,280. Jenkinson, E. J., Franchi, L. L., Kingston, R., and Owen, J. J. T. (1982).Eur.1.Immunol. 12,583. Jenkinson, E. J., Jhittay, P., Kingston, R., and Owen, J. J. T. (1985).Transplantation,39, 331. Jenkinson, E. J., Kingston, R., and Owen, J. J. T. (1987). Nature (London)329,160. Jin, Y.-J., Clayton, L. K., Howard, F. D., Koyasu, S., Sieh, M., Steinbrich, R., Tarr, G. E., and Reinherz, E. L. (1990).Proc. Natl. Acad. Sci. U.S.A.87,3319. Jones, L. A., Chin, L. T., Merriam, G. R., Nelson, L. M., and Kruisbeek, A. M. (1990a). J . E r p . Med., 172, 1277. Jones, L. A., Chin, L. T., Longo, D. L., and Kruisbeek, A. M. (1990b). Science 250, 1726. Jotereau, F. V., and LeDouarin, N. M. (1982).J.Zmmunol. 129, 1869. Jotereau, F., Heuze, F., Salomon-Vie, V., and Gascan, H. (1987)J. Immunol. 138,1026. Ju, S.-T., Ruddle, N. H., Strack, P., Dorf, M. E., and DeKruyff, R. H. (199O).J.Zmmunol. 144,23. June, C. H., Ledbetter, J. A., Lindsten, T., and Thompson, C. B. (1989)J. Zmmunol. 143, 153. June, C. H., Fletcher, M. C., Ledbetter, J. A., and Samelson, L. E. (1990).J . Zmmunol. 144,1591. Kaiser, N., and Edelman, I. S. (1977). Proc. Natl. Acad. Sci. U.S.A.74,638. Kammer, G. M. (1988). Zmmunol. Today 9,222. Kamps, M. P., Corcoran, L., LeBowitz, J. H., and Baltimore, D. (1990).Mol. Cell Biol. 10, 5464. Kappes, D. J., and Tonegawa, S. (1991). Proc. Natl. Acad. Sci. U.S.A.88, 10619. Kappler, J. W., Wade, T., White, J., Kushnir, E., Blackman, M., Bill, J., and Marrack, P. (1987).Cell (Cambridge, Mass.) 49,263.


203

Kappler, J. W., Staerz, U.,White, J., and Marrack, P. C . (1988).Nature (London)332,35. Kaye, J., Hsu, M.-L., Sauron, M.-L, Jameson, S. C., Gascoigne, N. R. J., and Hedrick, S. M. (1989). Nature (London)341,746. Kikkawa, U., Kishimoto, A., and Nishizuka, Y. (1989).Annu. Rev. Biochem. 58,31. Kim, D.-K., Nau, G. J., Lancki, D. W., Dawson, G., and Fitch, F. W. (1988).J.Immunol. 141,3429. Kingston, R., Jenkinson, E. J., and Owen, J. J. T. (1985). Nature (London)317,811. Kinnon, C., Diamond, R. A., and Rothenberg, E. V. (1986).J.lmmunol. 137,4010. Kishi, H., Borgulya, P., Scott, B., Karjalainen, K., Traunecker, A., Kaufman, J., and von Boehmer, H. (1991). EMBOJ. 10,93. Kisielow, P., Krammer, P. H., Hultner, L., and von Boehmer, H. (1985). Thymus 7, 189. Kisielow, P., Bliithmann, H., Staerz, U. D., Steinmetz, M., and von Boehmer, H. (1988). Nature (London)333,742. Kizaki, H., Tadakuma, T., Odaka, C., Muramatsu, J., and Ishimura, Y. (1989).J.Immunol. 143,1790. Klausner, R. D., and Samelson, L. E. (1991). Cell (Cambridge,Mass.)64,875. Koizumi, H., Liu, C.-C., Zheng, L.-M., Joag, S. V., Bayne, N. K., Holoshitz, J., and Young, J. D.-E. (1991).J.E x p . Med. 173,499. Koretzky, G. A., Picus, J.. Thomas, M. L., and Weiss, A. (1990).Nature (London)346,66. Koretzky, G. A., Picus, J., Schultz, T., and Weiss, A. (1991). Proc. Natl. Acad. Sci. U.S.A. 88,2037. Kosugi, A., Zuniga-Pflucker, J. C., Sharrow, S. O., Kruisbeek, A. M., and Shearer, G. M. (1989).J. lmmunol. 143,3134. Koulova, L., Yang, S. Y., and Dupont, B. (199O).J.lmmunol. 145,2035. Kozumbo, W. J., Harris, D. T., Gromkowski, S., Cerottini, J.-C., and Cerutti, P. A. (1987). J . Immunol. 138,606. Kruisbeek, A. M., Sharrow, S. O., Mathieson, B. J., and Singer, A. (1981).J.lmmunol. 127, 2168. Kruisbeek, A. M., Mond, J. J., Fowlkes, B. J., Carmen, J. A., Bridges, S., and Longo, D. L. (1985).J . E x p . Med. 161, 1029. Kuhn, R., Rajewsky, K., and Muller, W. (1991). Science 254, 707. Kumagai, N., Benedict, S. H., Mills, G. B., and Gelfand, E. W. (1987).J.lmmunol. 139, 1393. Kung, J. T. (1988).J.lmmunol. 140,3727. Kuno, M., and Gardner, P. (1987).Nature (London)326,301. Kurt-Jones, E. A., Hamberg, S., Ohara, J., Paul, W. E., and Abbas, A. K. (1987).]. E x p . Med. 166,1774. Kurtzberg, J., Denning, S. M., Nycum, L. M., Singer, K. H., and Haynes, B. F. (1989). Proc. Natl. Acad. Sci. U.S.A. 86,7575. Kuster, H., Thompson, H., and Kinet, J. P. (1990).J. Biol. Chem. 265,6448. Kyes, S., Pao, W., and Hayday, A. (1991). Proc. Natl. Acad. Sci. U.S.A.88,7830. Kyewski, B. A. (1986).Immunol. Today 7,374. Lafaille, J. J., Haas, W., Coutinho, A., and Tonegawa, S. (1990).Immunol. Today 11,75. Landau, N. R., Schatz, D. G., Rosa, M., and Baltimore, D. (1987).Mol. Cell. Biol. 7,3237. Lanier, L. L., Allison, J. P., and Phillips, J. H. (1986).J. Immunol. 137,2501. Lanier, L. L., Yu, G., and Phillips, J. H. (1989). Nature (London)342,803. Lawetzky, A., Kubbies, M., and Hunig, T. (1991). Eur.J.Immunol. 21,2599. Le, P. T., Vollger, L. W., Haynes, B. F., and Singer, K. H. (199O).J.Immunol. 144,4541. Leclercq, G., De Smedt, M., Tison, B., and Plum, J. (1990).J. Immunol. 145,3992. Ledbetter, J. A., Martin, P. J.,Spooner, C. E., Wofsy, D., Tsu, T. T., Beatty, P. G., and Gladstone, P. (1985).J.Immunol. 135,2331.

204

ELLEN V. ROTHENBERG

Ledbetter, J. A., June, C. H., Martin, P. J., Spooner, C. E., Hansen, J. A., and Meier, K. E. (1986a).]. Zmmunol. 136,3945. Ledbetter, J. A,, Parsons, M., Martin, P. J., Hansen, J. A., Rabinovitch, P. S., and June, C. H. (1986b)J. Zmmunol. 137,3299. LeDouarin, N. (1991).Adu. E x p . Med. Biol. 292, 19. LeGros, G., Ben-Sasson, S. Z., Seder, R., Finkelman, F. D., and Paul, W. E. (199O).J.E x p . Med., 172,921. Leo, O., Foo, M., Forman, J., Shivakumar, S., Rabinowitz, R., and Bluestone, J. A. (1988). J . Zmmunol. 141,37. Lerner, A., Jacobson, B., and Miller, R. A. (1988).J. Zmmunol. 140,936. Lesley, J., Trotter, J., Schulte, R., and Hyman, R. (1990).Cell. Zmmunol. 128,63. L6 thi Bich-Thuy, Dukovich, M., Peffer, N. J., Fauci, A. S., Kehrl, J. H., and Greene, W. C. (1987).]. Zmmunol. 139, 1550. Levitsky, H. I., Golumbek, P. T., and Pardoll, D. M. (1991).J.Zmmunol. 146,1113. Lewis, D. B., Yu, C. C., Forbush, K. A., Carpenter, J,, Sato, T. A., Grossman, A., Liggitt, D. H., and Perlmutter, R. M. (1991).J. Exp. Med. 173,89. Lewis, R. S., and Cahalan, M. D. (1988a). Science 239,771. Lewis, R. S., and Cahalan, M. D. (1988b). Trends Neurosci. 11,214. Lewis, R. S., and Cahalan, M. D. (1989). Cell Regul. 1,99. Lindsten, T., June, C. H., Ledbetter, J. A,, Stella, G., and Thompson, C. B. (1989). Science 244,339. Linsley, P. S., Clark, E. A., and Ledbetter, J. A. (1990).Proc. Natl. Acad. Sci. U.S.A.87, 5031. Linsley, P. S., Brady, W., Grosmaire, L., Aruffo, A., Damle, N. K., and Ledbetter, J. A. (1991).J. E x p . Med., 173,721. Liu, C.-C., Rafii, S. D., Granelli-Piperno, A., Trapani, J. A., and Young, J. D.-E. (1989). J . E x p . Med. 170,2105. Lo, D., and Sprent, J. (1986).Nature (London)319,672. Lobach, D. F., and Haynes, B. F. (1987).J. Clin. Zmmunol. 7,81. Lorenz, R. G., and Allen, P. M. (1989).Nature (London)340,557. Lowenthal, J. W., Howe, R. C., Ceredig, R., and MacDonald, H. R. (1986).J. Zmmunol.. 137,. 2579. Lowenthal, J. W., Bohnlein, E., Ballard, D. W., and Greene, W. C. (1988). Proc. Natl. Acad. Sci. U.S.A.85,4468. Lugo, J. P., Krishnan, S. N., Sailor, R. D., Koen, P., Malek, T., and Rothenberg, E. (1985). J . E x p . Med. 161, 1048. Lugo, J. P., Krishnan, S. N., Diamond Sailor, R., and Rothenberg, E. V. (1986).Proc. Natl. Acad. Sci. U.S.A.83, 1862. MacDonald, H. R., and Lees, R. K. (1990).Nature (London)343,642. MacDonald, H. R., Blanc, C., Lees, R. K., and Sordat, B. (1986).J. Zmmunol. 136,4337. MacDonald, H. R., Schneider, R., Lees, R. K., Howe, R. C., Acha-Orbea, H., Festenstein, H., Zinkernagel, R. M., and Hengartner, H. (1988a).Nature (London)332,40. MacDonald, H. R., Budd, R. C., and Howe, R. C. (1988b).Eur.1. Zmmunol. 18,519. Nature (London) 335, MacDonald, H. R., Hengartner, H., and Pedrazzini, T. (1988~). 174. MacDonald, H. R., Howe, R.C., Pedrazzini, T., Lees, R. K., Budd, R. C., Schneider, R., Liao, N. S., Zinkernagel, R. M., Louis, J. A., Raulet, D. H., Hengartner, H., and Miescher, G. (1988d). Zmmunol. Reu. 104,157. Maguire, J. E., McCarthy, S. A., Singer, A., and Singer, D. S. (1990).FASEBJ. 4,3131. Malissen, M., Trucy, J., Letourneur, F., and Malissen, B. (1988).Cell (Cambridge,Mass.) 55,317.


205

Mandel, T. E., and Kennedy, M. (1978).Zmmunology 35,317. Manger, B., Weiss, A., Irnboden, J., Laing, T., and Stobo, J. D. (1987).J. lmmunol. 139, 2755. Marrack, P., and Kappler, J. (1988).Nature (London)332,840. Marrack, P., Kushnir, E., Born, W., McDuffie, M., and Kappler, J. (1988a).J. lmrnunol. 140,2508. Marrack, P., Lo, D., Brinster, R., Palmiter, R., Burkly, L., Flavell, R. H., and Kappler, J. (1988b). Cell (Cambridge, Mass.) 53,627. Marrack, P., McCormack, J., and Kappler, J. (1989). Nature (London)338,503. MaruSiC-GaleSiC, S., Stephany, D. A,, Longo, D. L., and Kruisbeek, A. M. (1988).Nature (London)333,180. Mary, D., Aussel, C., Ferrua, B., and Fehlmann, M. (1987).J. Immunol. 139, 1179. Matis, L. A. (1990).Annu. Rew. Zmmunol. 8,65. Matsuzaki, C., Yoshikai, Y., Ogimoto, M., Kishihara, K., and Nomoto, K. (199O).J.Zmmunol. 145,46. McCarthy, S. A., Kruisbeek, A. M., Uppenkamp, I. K., Sharrow, S. O., and Singer, A. (1988).Nature (London)336,76. McConkey, D. J., Hartzell, P., Amador-Perez, J. F., Orrenius, S., and Jondal, M. (1989a). J . lmmunol. 143, 1801. McConkey, D. J., Hartzell, P., Jondal, M., and Orrenius, S. (1989b).J. Biol. Chem. 264, 13399. McConkey, D. J., Nicotera, P., Hartzell, P., Bellorno, G., Wyllie, A. H., and Orrenius, S. Arch. Biochem. Biophys. 269,365. (1989~). McConkey, D. J., Orrenius, S., and Jondal, M. (1990a).J . Zmmunol. 145,1227. McConkey, D. J., Hartzell, P., Chow, S. C., Orrenius, S., and Jondal, M. (1990b).J. Biol. Chem. 265,3009. McCormick, J. E., Kappler, J., Marrack, P., and Westcott, J. Y. (1991).]. Zmmunol.. 146, 239. McDuffie, M., Born, W., Marrack, P., and Kappler, J. (1986).Proc. NatLAcad. Sci. U.S.A. 83,8728. McGuire, K. L., and Rothenberg, E. V. (1987).EMBOJ.6,939. McCuire, K. L., Yang, J. A., and Rothenberg, E. V. (1988). Proc. Natl. Acad. Sci. U.S.A. 85,6503. McKinnon, D., and Ceredig, R. (1986).J. E x p . Med. 164,1846. Mercep, M., Bonifacino, J. S., and Ashwell, J. D. (1988).Science 242,571. Mercep, M., Noguchi, P. D., and Ashwell, J. D. (1989).J.Zmmunol. 142,4085. Meuer, S. C., Hussey, R. E., Fabbi, M., Fox, D., Acuto, O., Fitzgerald, K. A., Hodgdon, J. C., Protentis, J. P., Schlossman, S. F., and Reinherz, E. L. (1984).Cell (Cambridge, Mass.) 36,897. Michon, J. M., Calgiuri, M. A., Hazanow, S. M., Levine, H., Schlossman, S. F., and Ritz, J. (1988).J. Zmmunol. 140,3660. Miescher, G. C., Howe, R. C., Lees, R. K., and MacDonald, H. R. (1988).J. Immunol. 140, 1779. Miller, R. A., and Stutman, 0. (1982).J. Immunol. 68, 114. Miller, R. A., and Stutman, 0. (1984).J. Zmmunol. 133,2925. Mills, G . B., Cheung, R. K., Grinstein, S., and Gelfand, E. W. (1985a).J. Zmmunol. 134, 1640. Mills, G . B., Cheung, R. K., Grinstein, S., and Gelfand, E. W. (1985b).J. Zmmunol. 134, 2431. Mills, G. B., Girard, P., Grinstein, S., and Gelfand, E. W. (1988).Cell (Cambridge,MUSS.) 55,91.

206

ELLEN V. ROTHENBERG

Mingari, M. C., Poggi, A,, Biassoni, R., Bellomo, R., Ciccone, E., Pella, N., Morelli, L., Verdiani, S., Moretta, A., and Moretta, L. (1991).J.E x p . Med. 174,21. Montgomery, R. A., and Dallman, M. J. (1991).J. Immunol. 147,554. Moore, J. P., Todd, J. A., Hesketh, T. R., and Metcalfe, J. C. (1986).J.Biol. Chem. 261, 8158. Moore, K. W., Vieira, P., Fiorentino, D. F., Trounstine, M. L., Khan, T. A., and Mosmann, T. R. (1990). Science 248,1230. Moretta, A., Pantaleo, G., Lopez-Botet, M., and Moretta, L. (1985).]. E x p . Med. 162,823. Morrissey, P. J,, Goodwin, R. G., Nordan, R. P., Anderson, D., Grabstein, K. H., Cosman, D., Sims, J,, Lupton, S., Acres, B., Reed., S. G., Mochizuki, D., Eisenman, J., Conlon, P. J., and Namen, A. E. (1989).J.E x p . Med. 169,707. Mosley, B., Beckmann, M. P., March, C. J,, Idzerda, R. L., Gimpel, S. D.,VandenBos,T., Friend, D., Alpert, A., Anderson, D., Jackson, J., Wignall, J. M., Smith, C., Callis, B., Sims, J. E., Urdal, D., Widmer, M. B., Cosman, D., and Park, L. S. (1989). Cell (Cambridge,Mass.) 59,335. Mosmann, T. R., and Coffman, R. L. (1989).Annu. Reu. Immunol. 7,145. Muegge, K., Williams, T. M., Kant, J., Karin, M., Chiu, R., Schmidt, A,, Siebenlist, U., Young, H. A., and Durum, S. K. (1989). Science 246,249. Mueller, D. L., Jenkins, M. K., and Schwartz, R. H. (1989).Annu. Reu. Immunol. 7,445. Mueller, D. L., Jenkins, M. K., Chiodetti, L., and Schwartz, R. H. (199O).J.Immunol. 144 3701. Mufioz, E., Zubiaga, A. M., Merrow, M., Sauter, N. P., and Huber, B. T. (1990).J . E x p . Med. 172,95. Murphy, K. M., Heimberger, A. B., and Loh, D. Y. (1990). Science 250, 1720. Murray, R., Suda, T., Wrighton, N., Lee, F., and Zlotnik, A. (1989).Int. Immunol. 1,526. Mustelin, T., Coggeshall, K. M., and Altman, A. (1989).Proc. Natl. Acad. Sci. U.S.A.86, 6302. Mustelin, T., Coggeshall, K. M., Isakov, N., and Altman, A. (1990). Science 247,1584. Nabholz, M., and MacDonald, H. R. (1983).Annu. Reu. Immunol. 1,273. Nakano, N., Hardy, R. R., and Kishimoto, T. (1987). Eur. J. Immunol. 17, 1567. Nakashima, I., Zhang, Y.-H., Rahman, S. M. J., Yoshida, T., Isobe, K.-I., Ding, L.-N., Iwamoto, T., Hamaguchi, M., Ikezawa, H., and Taguchi, R. (1991).J. Immunol. 147, 1153. Nakayama, T., Singer, A., Hsi, E. D., and Samelson, L. E. (1989).Nature (London)341, 651. Nakayama, T., June, C. H., Munitz, T. I., Sheard, M., McCarthy, S. A., Sharrow, S. O., Samelson, L. E., and Singer, A. (1990). Science 249, 1558. Naparstek, Y., Holoshitz, J., Eisenstein, S., Reshef, T., Rappaport, S., Chemke, J., BenNun, A., and Cohen, I. R. (1982). Nature (London)300,262. Neer, E. J., and Clapham, D. E. (1988).Nature (London)333,129. Nieto, M. A., Gonzalez, A., L6pez-Rivas, A., Diaz-Espada, F., and Gambon, F. (1990). J . Immunol. 145,1364. NikoliC-hgie, J., and Bevan, M. J. (1988). Proc. Natl. Acad. Sci. U.S.A.85,8633. NikoliC-hgiC, J., and Bevan, M. J. (1990). Nature (London)344,65. NikoliC-ZugiC, J., and Moore, M. W. (1989). Eur. J . Immunol. 19, 1957. Norment, A. M., Salter, R. D., Parham, P., Engelhard, V. H., and Littman, D. R. (1988). Nature (London)336,79. Nossal, G. J. V. (1983).Annu. Reo. Immunol. 1,33. Novak, T. J., and Rothenberg, E. V. (1990). Proc. Natl. Acad. Sci. U.S.A.87,9353. Novak, T. J., Chen, D., and Rothenberg, E. V. (1990). Mol. Cell. B i d . 10,6325.


207

NuAez, G., London, L., Hockenbery, D., Alexander, M., McKearn, J. P., and Korsmeyer, S. J. (1990).J. Zmmunol. 144,3602. Oettinger, M. A., Schatz, D. G., Gorka, C., and Baltimore, D. (1990). Science 248,1517. Ohashi, P. S., Pircher, H., Biirki, K., Zinkernagel, R. M., and Hengartner, H. (1990a). Nature (London)346,861. Ohashi, P. S., Wallace, V. A., Broughton, H., Ohashi, C. T., Ferrick, D. A., Jost, V., Mak, T. W., Hengartner, H., and Pircher, H. (1990b).Eur.J.Zmmunol. 20,517. Ohashi, Y., Takeshita, T., Nagata, K., Mori, S., and Sugamura, K. (1989).;. Zmmunol. 143, 3548. Okazaki, H., Ito, M., Sudo, T., Hattori, M., Kano, S., Katsura, Y., and Minato, N. (1989). J. Zmmunol. 143,2917. Ono, Y., Fujii, T., Ogita, K., Kikkawa, U., Igarashi, K., and Nishizuka, Y. (1989). Proc. Natl. Acad. Sci. U.S.A. 86,3099. Orloff, D. G., Ra, C., Frank, S. J., Klausner, R. D., and Kinet, J:P. (1990). Nature (London)347,189. O’Rourke, A. M., Mescher, M. F., and Webb, S. R. (1990). Science 249,171. O’Shea, J. J., Urdahl, K. B., Luong, H. T., Chused, T. M., Samelson, L. E., and Klausner, R. D. (1987).J. Zmmunol. 139,3463. Owen, J. J. T., Kingston, R., and Jenkinson, E. J. (1986). Zmmunology 59,23. Owen, M. J., Jenkinson, E. J., Brown, M. H., Sewell, W.A.,Krissansen,G. W.,Crumpton, M. J., and Owen, J. J. T. (1988). Eur.J.Zmmunol. 18, 187. Palacios, R., and von Boehmer, H. (1986). Eur. J. Immunol. 16, 12. Palacios, R., Sideras, P., and von Boehmer, H. (1987). E M B O J . 6,91. Papiernik, M., and Homo-Delarche, F. (1983). Eur.J. Zmmunol. 13,689. Pardoll, D. M., Fowlkes, B. J., Lechler, R. I., Germain, R. N., and Schwartz, R. H. (1987a). /. E r p . Med. 165, 1624. Pardoll, D. M., Fowlkes, B. J., Bluestone, J. A., Kruisbeek, A., Maloy, W. L., Coligan, J. E., and Schwartz, R. H. (1987b). Nature (London)326,79. Parham, P. (1991). Nature (London)351,271. Parker, C. M., Groh, V., Band, H., Porcelli, S. A., Morita, C., Fabbi, M., Glass, D., Strominger, J. L., and Brenner, M. B. (1990).]. E x p . Med. 171,1597. Parnes, J. R. (1989). Ado. Zmmunol. 44,265. Patel, S. S., Wacholtz, M. C., Duby, A. D., Thiele, D. L., and Lipsky, P. E. (1989). /, Immunol. 143,1108. Paterson, D. J., and Williams, A. F. (1987).J.E r p . Med. 166, 1603. Pearse, M., Wu, L., Egerton, M., Wilson, A., Shortman, K., and Scollay, R. (1989). Proc. Natl. Acad. Sci. U.S.A. 86, 1614-1618. Pelkonen, J., Sideras, P., Rammensee, H.-G., Kaqalainen, K., and Palacios, R. (1987). J. E x p . Med. 166, 1245. Penit, C. (1986).J.Zmmunol. 137,2115. Penit, C. (1990).Znt. Immunol. 2,629. Penit, C . , and Vasseur, F. (1988).J.Zmmunol. 140,3315. Penit, C., and Vasseur, F. (1989).J. Zmmunol. 142,3369. Penit, C., Vasseur, F., and Papiernik, M. (1988). Eur.J.Zmmunol. 18, 1343. Peralta, E. G., Ashkenazi, A., Winslow, J. W., Ramachandran, J., and Capon, D. J. (1988). Nature (London)334,434. Pereira, P., Zijlstra, M., McMaster, J., Loring, J. M., Jaenisch, R., and Tonegawa, S. (1992). EMBOJ. 11,25. Perlmutter, R. M., Marth, J. D., Ziegler, S. F., Garvin, A. M., Pawar, S., Cooke, M. P., and Abraham, K. M. (1988). Biochim. Biophys. Acta 948,245.

208

ELLEN V. ROTHENBERG

Petrie, H. T., Hugo, P., Scollay, R., and Shortman, K. (1990a).J. Exp. Med. 172,1583. Petrie, H. T., Pearse, M., Scollay, R., and Shortman, K. (1990b). Eur. J. Immunol. 20, 2813. Phillips, J. H., and Lanier, L. L. (1987).J. Immunol. 139,683. Pierres, A., Lopez, M., Cerdan, C., Nunes, J., Olive, D., and Mawas, C. (1988). E u r . J. Immunol. 18,685. Pierres, A., Cerdan, C., Lopez, M., Mawas, C., and Olive, D. (1990).J. Immunol. 144, 1202. Pilarski, L. M., Gillitzer, R., Zola, H., Shortman, K., and Scollay, R. (1989). E u r . J . Immunol. 19,589. Pingel, J. T., and Thomas, M. L. (1989). Cell (Cambridge,Mass.) 58,1055. Pircher, H., Biirki, K., Lang, R., Hengartner, H., and Zinkernagel, R. M. (1989). Nature (London)342,559. Pircher, H., Rohrer, U. H., Moskophidis, D., Zinkernagel, R. M., and Hengartner, H. (1991). Nature (London)351,482. Plaut, M., Pierce, J. H., Watson, C. J., Hanley-Hyde, J., and Nordan, R. P. (1989).Nature (London)339,64. Plum, J., and De Smedt, M. (1988).Eur. J. Immunol. 18,795. Plum, J., Koning, F., Leclercq, G., Tison, B., and De Smedt, M. (1990).J.Immunol. 144, 3710. Podack, E. R., Lowrey, D. M., Lichtenheld, M., Olson, K. J., Aebischer, T., Binder, D., Pupp, F., and Hengartner, H. (1988). Immunol. Reu. 103,203. Powers, G. D., Abbas, A. K., and Miller, R. A. (1988).J. Immunol. 140,3352. Pullen, A. M., Wade, T., Marrack, P., and Kappler, J. W. (1990).Cell (Cambridge,Mass.) 60, 1365. Punt, J. A., Kubo, R. T., Saito, T., Finkel, T. H., Kathierson, S., Blank, K. J., and Hashimoto, Y. (1991).J.E x p . Med. 174,775. Quill, H., and Schwartz, R. H. (1987).J.Immunol. 138,3704. Radler-Pohl, A., Pfeuffer, I., Karin, M., and Serfling, E. (1990). New Biol. 2,566. Rahemtulla, A., Fung-Leung, W. P., Schilham, M. W., Kiindig, T. M., Sambhara, S. R., Narendran, A., Arabian, A., Wakeham, A., Paige, C. J., Zinkernagel, R. M., Miller, R. G., and Mak, T. W. (1991). Nature (London)353,180. Ramarli, D., Fox, D. A., and Reinherz, E. L. (1987).Proc. Natl. Acad. Sci. U.S.A.84,8598. Rammensee, H.G., Kroschewski, R., and Frangoulis, B. (1989) Nature (London)339,541. Ramsdell, F., and Fowlkes, B. J. (1989).J. Immunol. 143, 1467. Ramsdell, F., and Fowlkes, B. J. (1990). Science 248, 1342. Ramsdell, F. J., Gray, J. D., and Golub, S. H. (1988). Cell. Immunol. 114,209. Ramsdell, F., Lantz, T., and Fowlkes, B. J. (1989). Science 246, 1038. Ramsdell, F., Jenkins, M., Dinh, Q., and Fowlkes, B. J. (1991).J.Immunol. 147, 1779. Randak, C., Brabletz, T., Hergenrother, M., Sobotta, I., and Serfling, E. (1990).E M B O J . 9,2529. Ransom, J., Fischer, M., Mosmann, T., Yokota, T., DeLuca, D., Schumacher, J., and Zlotnik, A. (1987).J. Immunol. 139,4102. Raulet, D. H. (1985). Nature (London)314, 101. Raulet, D. H. (1989).Annu. Reo. Immunol. 7,175. Raulet, D. H., Spencer, D. M., Hsiang, Y.-H., Goldman, J. P., Bix, M., Liao, N.-S., Zijlstra, M., Jaenisch, R., and Correa, I. (1991). lmmunol Reo. 120,185. Ready, A. R., Jenkinson, E. J., Kingston, R., and Owen, J. J. T. (1984).Nature (London) 310,231. Reed, J. C., Abidi, A. H., Alpers, J. D., Hoover, R. G., Robb, R. J., and Nowell, P. C. (1986).J.Immunol. 137, 150.


209

Reisner, Y., Linker-Israeli, M., and Sharon, N. (1976).Cell. Zmmunol. 25, 129. Reynolds, P. J., Lesley, J., Trotter, J., Schulte, R., Hyman, R., and Sefton, B. M. (1990). Mol. Cell. Biol. 10,4266. Riegel, J. S., Richie, E. R., and Allison, J. P. (199O).J.lmmunol. 144,3611. Roberts, J. L., Sharrow, S. O., and Singer, A. (1990).J . E x p . Med. 171,935. Robey, E. A., Fowlkes, B. J., Gordon, J. W., Kioussis, D., von Boehmer, H., Ramsdell, F., and Axel, R. (1991). Cell (Cambridge,Mass.) 64,99. Rocha, B., and von Boehmer, H. (1991). Science 251,1225. Rocha, B., Lehuen, A., and Papiernik, M. (1988).J. lmmunol. 140, 1076. Rocha, B., Dautigny, N., and Pereira, P. (1989). Eur.1. Zmmunol. 19,905. Rocha, B., Vassalli, P., and Guy-Grand, D. (1991).J.E x p . Med. 173,483. Roehm, N. Herron, L., Cambier, J., DiGuisto, D., Haskins, K., Kappler, J., and Marrack, P. (1984). Cell (Cambridge, Mass.) 38,577. Romagnani, S. (1991). Zmmunol. Today 12,256. Roman, D. G., Toledano, M. B., and Leonard, W. J. (1990). New Biol. 2,642. Ron, Y., Lo, D., and Sprent, J. (1986).J.lmmunol. 137, 1764. Rothbard, J. B., and Gefter, M. L. (1991).Annu. Reu. Zmmunol. 9,527. Rothenberg, E. V. (1990).Zmmunol. Today 11,116. Rothenberg, E. V. (1992).In preparation. Rothenberg, E. V., and Chen, D. (1992). In preparation. Rothenberg, E. V., and Lugo, J. P. (1985). Deu. B i d . 112, 1. Rothenberg, E., and Triglia, D. (1983).J.Zmmunol. 130, 1627. Rothenberg, E. V., McGuire, K. L., and Boyer, P. D. (1988). Zmmunol. Reo. 104,29. Rothenberg, E. V., Diamond, R. A., Novak, T. J., Pepper, K. A., and Yang, J. A. (1990a) UCLA Symp. Mol. Cell Biol. [N.S.] 125,225. Rothenberg, E. V., Diamond, R. A., Pepper, K. A., and Yang, J. A. (1990b).J.Zmmunol. 144,1614. Rothenberg, E. V., Chen, D., Diamond, R. A., Dohadwala, M., Novak, T., White, P. M., and Yang-Snyder, J. (1991).Adu. E x p . Med. B i d . 292,71 Rudd, C. E., Trevillyan, J. M., Dasgupta, J. D., Wong, L. L., and Schlossman, S. F. (1988). Proc. Natl. Acad. Sci. U.S.A.85,5190. Rudd, C. E., Anderson, P., Morimoto, C., Streuli, M., and Schlossman, S. F. (1989). lmmunol. Reu. 111,225. Ryder, K., and Nathans, D. (1988). Proc. Natl. Acad. Sci. U.S.A. 85,8464. Saffer, J. D., Jackson, S. P., and Annarella, M. B. (1991).Mol. Cell. Biol. 11,2189. Saito, T., Pardoll, D. M., Fowlkes, B. J., and Ohno, H. (1990). Cell. Immunol. 131,284. Salmon, M., Kitas, G. D., and Bacon, P. A. (1989).J.lmmunol. 143,907. Salter, R. D., Benjamin, R. J., Wesley, P. K., Buxton, S. E., Garrett, T. P. J., Clayberger, C., Krensky, A. M., Norment, A. M., Littman, D. R., and Parham, P. (1990). Nature (London)345,41. Samaridis, J., Casorati, G., Traunecker, A,, Iglesias, A., Gutierrez, J. C., Miiller, U., and Palacios, R. (1991). Eur. J . lmmunol. 21,453. Samelson, L. E., Phillips, A. F., Luong, E. T., and Klausner, R. D. (1990). Proc. Natl. Acad. Sci. U.S.A. 87,4358. Sanders, M. E., Makgoba, M. W., Sharrow, S. O., Stephany, D., Springer, T. A., Young, H. A., and Shaw, S. (1988).J.Zmmunol. 140,1401. Schatz, D. G., Oettinger, M. A., and Baltimore, D. (1989). Cell (Cambridge, Mass.) 59, 1035. Schorle, H., Holtschke, T., Hiinig, T., Schimpl, A., and Horak, I. (1991)Nature (London) 352,621. Scollay, R., Butcher, E. C., and Weissman, I. L. (1980). Eur.J.Zmmunol. 10,210.

210

ELLEN V. ROTHENBERG

Scollay, R., Chen, W.-F., and Shortman, K. (1984).j . Immunol. 132,25. Scollay, R., Wilson, A., D’Amico, A,, Kelly, K., Egerton, M., Pearse, M., Wu, L., and Shortman, K. (1988). lmmunol. Reu. 104,82. Scott, B., Bliithmann, H., Teh, H. S., and von Boehmer, H. (1989).Nature (London)338, 591. Screpanti, I., Morrone, S., Meco, D., Santoni, A., Gulino, A., Paolini, R., Crisanti, A., Mathieson, B. J., and Frati, L. (1989).J.lmmunol. 142,3378. Sellins, K. S., and Cohen, J. J. (1987).J . Immunol. 139,3199. Sentman, C. L., Shutter, J. R., Hockenbery, D., Kanagawa, O., and Korsmeyer, S. J. (1991). Cell (Cambridge Mass.) 67,879. Serfling, E., Barthelmas, R., Pfeuffer, I., Schenk, B., Zarius, S., Swoboda, R., Mercurio, F., and Karin, M. (1989). EMBO J . 8,465, Sha, W. C., Nelson, C. A., Newberry, R. D., Kranz, D. M., Russell, J. H., and Loh, D. Y. (1988a).Nature (London)335,271. Sha, W. C., Nelson, C. A., Newberry, R. D., Kranz, D. M., Russell, J. H., and Loh, D. Y. (1988b).Nature (London)336,73. Sha, W. C., Nelson, C. A., Newberry, R. D., Pullen, J. K., Pease, L. R., Russell, J. H., and Loh, D. Y. (1990). Proc. Natl. Acad. Sci. U S A . 87,6186. Shaw, J.-P., Utz, P.-J., Durand, D. B., Toole, J. J., Emmel, J. A., and Crabtree, G . R. (1988).Science 241,202. Shi, Y.,Sahai, B. M., and Green, D. R. (1989). Nature (London)339,625. Shi, Y., Bissonnette, R. P., Parfrey, N., Szalay, M., Kubo, R. T., and Green, D. R. (1991). J . Immunol. 146,3340. Shimonkevitz, R. P., Kappler, J., Marrack, P., and Grey, H. ( 1 9 8 3 ) ~E.x p . Med. 158,303. Shimonkevitz, R. P., Husmann, L. A,, Bevan, M. J., and Crispe, I. N. (1987). Nature (London)329,157. Shipp, M. A., and Reinherz, E. L. (1987).J.Immunol. 139,2143. Shores, E. W., Sharrow, S. O., Uppenkamp, I., and Singer, A. (1990). Eur. J . Immunol. 20.69. Shores, E. W., van Ewijk, W., and Singer, A. (1991).Eur.]. lmmunol. 21,1657 Shortman, K., Wilson, A., and Scollay, R. (1984).]. lmmunol. 132,584. Shortman, K., Egerton, M., Spangrude, G. J., and Scollay, R. (1990). Semin. Immunol. 2,3. Shortman, K., Vremec, D., and Egerton, M. (1991).J.E x p . Med. 173,323. Sideras, P., Funa, K., Zalcberg-Quintana, I., Xanthopoulos, K. G., Kisielow, P., and Palacios, R. (1988). Proc. Natl. Acad. Sci. U.S.A.85,218. Siege], J. P., Sharon, M., Smith, P. L., and Leonard, W. J. (1987). Science 238,75. Skinner, M., LeGros, G., Marbrook, J., and Watson, J. D. (1987).]. Exp. Med. 165,1481. Sminia, T., Van Asselt, A. A., Van De Ende, M. B., and Dijkstra, C. D. (1986). Thymus 8, 141. Smith, C. A., Williams, G. T., Kingston, R., Jenkinson, E. J., and Owen, J. J. T. (1989). Nature (London)337,181. Snodgrass, H. R., Kisielow, P., Kiefer, M., Steinmetz, M., and von Boehmer, H. (1985a). Nature (London)313,592. Snodgrass, H. R., DembiC, Z., Steinmetz, M., and von Boehmer, H. (1985b). Nature (London)315,232. Sowder, J. T., Chen, C.-L. H., Ager, L. L., Chan, M. M., and Cooper, M. D. (1988).j.Erp. Med. 167,315. Spangrude, G. J., and Scollay, R. (199O).J.Immunol. 145,3661. Spangrude, G. J., Heimfeld, S., and Weissman, I. L. (1988). Science 241,58.


21 1

Spencer, D. M., Hsiang, Y.-H., Goldman, J. P., and Raulet, D. H. (1991).Proc. Natl. Acad. Sci. U.S.A.88,800. Springer, T. A. (1990). Nature (London)346,425. Stanley, J. B., Gorczynski, R., Huang, C.-K., Love, J., and Mills, G. B. (199O).J.Zmmunol. 145,2189. Strasser, A., Harris, A. W., and Cory, S. (1991). Cell (Cambridge,Mass.) 67,889. Street, N. E., and Mosmann, T. R. (1991).FASEBJ. 5,171. Strominger, J. L. (1989). Science 244,943. Suda, T., and Zlotnik, A. (1991).J.Zmmunol. 146,3068. Suda, T., Murray, R., Guidos, C., and Zlotnik, A. (199O).J.Zmmunol. 144,3039. Sussman, J. J.. Mercep, M., Saito, T., Germain, R. N., Bonvini, E., and Ashwell, J. D. (1988).Nature (London)334,625. Swain, S. L., (1980). Fed Proc. Fed. Am. S O C . E r p . Biol. 39,3110. Swain, S. L., McKenzie, D. T., Weinberg, A. D., and Hancock, W. (1988a).J.Zmmunol. 141,3445. Swain, S. L., McKenzie, D. T., Dutton, R. W., Tonkonogy, S. L., and English, M. (1988b). Zmmunol. Rev. 102,77. Swain, S. L., Bradley, L. M., Croft, M., Tonkonogy, S., Atkins, G., Weinberg, A. D., Duncan, D. D., Hedrick, S. M., Dutton, R. W., and Huston, G. (1991). Immunol. Rev. 123, 115. Swat, W., Ignatowicz, L., von Boehmer, H., and Kisielow, P. (1991). Nature (London) 351, 150. Tadakuma, T., Kizaki, H., Odaka, C., Kubota, R., Ishimura, Y., Yagita, H., and Okumura, K. (1990). Eur. J. Zmmunol. 20,779. Takacs, L., Osawa, H., Toro, I., and Diamantstein, T. (1985).Clin. Exp. Zmmunol. 59,37. Takacs, L., Ruscetti, F. W., Kovacs, E. J., Rocha, B., Brocke, S., Diamantstein, T., and Mathieson, B. J. (1988).J. Zmmunol. 141,3810. Takagi, M., Koike, K., and Nakahata, T. (199O).J.Zmmunol. 145, 1880. Takahama, Y., Kosugi, A., and Singer, A. (1991).J.Zmmunol. 146,1134. Takashi, T., Cause, W. C., Wilkinson, M., MacLeod, C. L., and Steinberg, A. D. (1991). Eur. J . Zmmunol. 21, 1385. Takei, F. (1988).J.Zmmunol. 141, 1114. Takeshita, S., Toda, M., and Yamagishi, H. (1989). EMBOJ. 8,3261. Tartakovski, B., Finnegan, A., Muegge, K., Brody, D. T., Kovacs, E. J., Smith, M. R., Berzofsky, J. A., Yang, H. A., and Durum, S. K. (1988).J . Zmmunol. 141,3863. Teh, H.-S., Kisielow, P., Scott, B., Kishi, H., Uematsu, Y., Bliithmann, H., and von Boehmer, H. (1988). Nature (London)335,229. Teh, H.-S., Kishi, H., Scott, B., and von Boehmer, H. (1989).J.Exp. Med. 169,795. Teh, H.-S., Garvin, A. M., Forbush, K. A,, Carlow, D. A., Davis, C. B., Littman, D. R.,and Perlmutter, R. M. (1991) Nature (London)349,241. Tentori, L., Longo, D. L., Zufiiga-Pflucker, J. C., Wing, C., and Kruisbeek, A. M. (1988). J . E r p . Med. 168,1741. Tepper, R. I., Levinson, D. A., Stanger, B. Z., Campos-Torres, J., Abbas, A. K., and Leder, P. (1990).Cell (Cambridge Mass.) 62,457. Thomas, M. L. (1989).Annu. Reo. Zmmunol. 7,339. Thompson, C. B., Lindsten, T., Ledbetter, J. A., Kunkel, S. L., Young, H. A., Emerson, S. G., Leiden, J. M., and June, C. H. (1989). Proc. Natl. Acad. Sci. U.S.A.86, 1333. Thompson, S. D., Pelkonen, J., and Hunvitz, J. L. (1990).Proc. Natl. Acad. Sci. U.S.A.87, 5583. Tigges, M. A., Casey, L. S., and Koshland, M. E. (1989).Science 243,781.

212

ELLEN V. ROTHENBERG

Tonegawa, S., Berns, A., Bonneville, M., Farr, A. G., Ishida, I., Ito, K., Itohara, S., Janeway, C. A., Jr., Kanagawa, O., Kubo, R., Lafaille, J. J.. Murphy, D. B., Nakanishi, N., Takagaki, Y.,and Veebeek, S. (1991).Adu. E x p . Med. Biol. 292,53. Torbett, B. E., and Glasebrook, A. L. (1989). FASEB J. 3, A1269. Toribio, M. L., de la Hera, A., Borst, J., Marcos, M. A. R,, Marquez, C., Alonso, J. M., Barcena, A., and Martinez-A, C. (1988a).J. E x p . Med. 168,2231. Toribio, M. L., Alonso, J. M., BBrcena, A., Gutierrez, J. C., de la Hera, A., Marcos, M. A. R., Mkrquez, C., and Martinez-A, C. (1988b). Immunol. Rev. 104,55. Toribio, M. L., Gutierrez-Ramos, J. C., Pezzi, L., Marcos, M. A. R., and Martinez-A., C. (1989).Nature (London)342,82. Transy, C., Moingeon, P., Stebbins, C., andReinherz, E. L. (1989).Proc. NatLAcad. Sci. U.S.A. 86,7108. Trevillyan, J. M., Lu, Y., Atluru, D., Phillips, C. A., and Bjorndahl, J. M. (1990). J , lmmunol. 145,3223. Triebel, F., and Hercend, T. (1989). Immunol. Today 10, 186. Tschopp, J., and Nabholz, M. (1990).Annu. Rev. Immunol. 8,279. Tsuchida, T., and Sakane, T. (1988).J. Immunol. 140,3446. Turka, L. A., Ledbetter, J. A., Lee, K., June, C. H., and Thompson, C. B. (1990). J. Immunol. 144, 1646. Turka, L. A., Schatz, D. G., Oettinger, M. A., Chun, J. J. M., Gorka, C., Lee, K., McCormack, W. T., and Thompson, C. B. (1991a).Science 253,778. Turka, L. A., Linsley, P. S., Paine, R., 111, Schieven, G. L., Thompson, C. B., and Ledbetter, J. A., (1991b).J. Immunol. 146,1428. Turner, B., Papp, U., App, H., Greene, M., Dobashi, K., and Reed, J. (1991).Proc. Natl. Acad. Sci. U.S.A.88, 1227. Turpen, J. B., and Smith, P. B. (1989).]. Immunol. 142,41. Ucker, D. S., Ashwell, J. D., and Nickas, G. (1989).J . Immunol. 143,3461. Ullman, K. S., Northrup, J. P., Verweij, C. L., and Crabtree, G. R. (1990). Annu. Reu. lmmunol. 8,421. Valge, V. E., Wong, J. G. P., Datlof, B. M., Sinskey, A. J., and Rao, A. (1988). Cell (Cambridge,Mass.) 55, 101. van Dongen, J. J. M., Quertermous, T., Bartram, C. R., Gold, D. P., Wolvers-Terrero, I. L. M., Comans-Bitter, W. M., Hooijkaas, H., Adriaanson, H. J., deKlein, A., Raghavachar, A,, Ganser, A., Duby, A. D., Seidman, J. G., van den Elsen, P., and Terhorst, C. (1987).J. lmmunol. 138, 1260. van Ewijk, W. (1991).Annu. Rev. lmmunol. 9,591. van Ewijk, W., Ron, Y.,Monaco, J., Kappler, J., Marrack, P., LeMeur, M., Gerlinger, P., Durand, B., Benoist, C., and Mathis, D. (1988).Cell (Cambridge,Mass.)53,357. Van Lier, R. A. W., Brouwer, M., and Aarden, L. A. (1988).Eur.]. lmmunol. 18,167. van Vliet, E., Jenkinson, E. J., Kingston, R., Owen, J. J. T., andvan Ewijk, W. (1985).Eur. J . Immunol. 15,675. Veillette, A., Bookman, M. A., Horak, E. M., and Bolen, J. B. (1988). Cell (Cambridge, Mass.) 55,301. Veillette, A., Bookman, M. A., Horak, E. M., Samelson, L. E., and Bolen, J. B. (1989a). Nature (London)338,257. Veillette, A., Zufiiga-Pflucker,J.-C., Bolen, J. B., and Kruisbeek, A. M. (1989b).J . E x p . Med. 170,1671. Vernachio, J., Li, M., Donnenberg, A. D., and Soloski, M. J. (1989).J. Immunol. 142,48. Vitetta, E. S., Berton, M. T., Burger, C., Kepron, M., Lee, W. T., and Yin, X.-M. (1991). Annu. Rev. lmmunol. 9, 193. Vives, J., Sole, J.. and Suarez, B. (1987). Proc. Natl. Acad. Sci. U.S.A.84,8593. ~


213

Vivier, E.,Morin, P., Tian, Q., Daley, J., Blue, M.-L., Schlossman, S. F., and Anderson, P. (1991a).J. Zmmunol. 146, 1142. Vivier, E., Rochet, N., Kochan, J. P., Presky, D. H., Schlossman, S. F., and Anderson, P. (1991b).J. Zmmunol. 147,4263. Vivier, E., Morin, P., O’Brien, C., Schlossman, S. F., and Anderson, P. (1991~).Eur. J. Immunol. 21, 1077. Vollger, L. W., Tuck, D. T., Springer, T. A., Haynes, B. F., and Singer, K. H. (1987). I. Zmmunol. 138,358. von Boehmer, H. (1990).Annu. Rev. Immunol. 8,531. von Boehmer, H., and Schubiger, K. (1984). Eur. J. lmmunol. 14, 1048. von Boehmer, H., Kisielow, P., Leiserson, W., and Haas, W. (1984).J. lmmunol. 133,59. von Boehmer, H., Crisanti, A., Kisielow, P., and Haas, W. (1985). Nature (London) 314, 539. von Boehmer, H., Bonneville, M., Ishida, I., Ryser, S., Lincoln, G., Smith, R. T., Kishi, H., Scott, B., KisieIow, P., and Tonegawa, S. (1988).Proc. Natl. Acad. Sci. U.S.A. 85, 9729. Waanders, G. A. (1991). Ph.D. Thesis, Monash University, Clayton, Victoria, Australia. Waanders, G. A., and Boyd, R. L. (1990). Znt. Zmmunol. 2,461. Wang, P. T. H., Bigby, M., and Sy, M . 3 . (1987).J. lmmunol. 139,2157. Watson, J. D., Morrissey, P. J., Namen, A. E., Conlon, P. J., and Widmer, M. B. (1989). J. lnimunol. 143, 1215. Watts, T. H., Brian, A. A., Kappler, J. W., Marrack, P., and McConnell, H. M. (1984).Proc. Natl. Acad. Sci. U.S.A.81,7564. Weaver, C. T., Pingel, J. T., Nelson, J . O., and Thomas, M. L. (1991). Mol. Cell. Biol. 11, 4415. Webb, S., Morris, C., and Sprent, J. (1990). Cell (Cambridge,Mass.) 63, 1249. Weiss, A. (1989). In “Fundamental Immunology” (W. E. Paul, ed.), 2nd ed., p. 359. Raven Press, New York. Weiss, A., and Imboden, J. B. (1987). Adu. Zmmunol. 41, 1. Weiss, A., Shields, R., Newton, M., Manger, B., and Imboden, J. (1987a).J. Zmmunol. 138,2169. Weiss, A., Dazin, P. F., Shields, R., Fu, S. M., and Lanier, L. L. (1987b).J. lmmunol. 139, 3245. Weiss, A,, Koretzky, G., Schatzman, R. C., and Kadlecek, T. (1991). Proc. Natl. Acad. Sci. U.S.A.88,5484. Wekerle, H., and Ketelsen, U.-P. (1980). Nature (London)283,402. Welch, P. A,, Namen, A. E., Goodwin, R. G., Armitage, R., and Cooper, M. D. (1989). J. Zmmunol. 143,3562. Wells, F. B., Gahm, S.-J., Hedrick, S. M., Bluestone, J. A., Dent, A., and Matis, L. A. (1991). Science 253,903. White, J., Herman, A., Pullen, A. M., Kubo, R., Kappler, J. W., and Marrack, P. (1989). Cell (Cambridge,Mass.) 56,27. Wildin, R. S., Gamin, A. M., Pawar, S., Lewis, D. B., Abraham, K. M., Forbush, K. A., Ziegler, S. F., Allen, J . M., and Perlmutter, R. M. (1991).J. E x p . Med. 173,383. Wilson, A., Petrie, H. T., Scollay, R., and Shortman, K. (1989). lnt. Zmmunol. 1,605. Winoto, A,, and Baltimore, D. (1989a). Nature (London)338,430. Winoto, A., and Baltimore, D. (1989b). Cell (Cambridge, Mass.) 59,649. Wodnar-Filipowicz, C., Heusser, H., and Moroni, C. (1989). Nature (London) 339, 150.

Wu, L., Scollay, R., Egerton, M., Pearse, M., Spangrude, G. J., and Shortman, K. (l99la). Nature (London)349,71.

214

ELLEN V. ROTHENBERG

Wu, L., Antica, M., Johnson, G. R., Scollay, R., and Shortman, K. (1991b).J. E x p . Med. 174,1617. Wyllie, A. H. (1980).Nature (London)284,555. Wyllie, A. H., Morris, R. G., Smith, A. L., and Dunlop, D. (1984).J. Pathd. 142,67. Yagita, H., Nakata, M., Azuma, A,, Nitta, T., Takeshita, T., Sugamura, K., and Okumura, K. (1989a).J. E x p . Med. 170,1445. Yagita, H., Asakawa, J.-I., Tansyo, S., Nakamura, T., Habu, S. and Okumura, K. (1989b). Eur.J.Immunol. 19,2211. Yamada, H., Martin, P. J., Bean, M. A., Braun, M. P., Beatty, P. G., Sadamoto, K., and Hansen, J. A. (1985).Eur. J. Zmmunol. 15, 1164. Yamarnoto, Y., Ohmura, T., Fujimoto, K., and Onoue, K. (1985). Eur. J. Immunol. 15, 1204. Yancopoulos, G . D., Blackwell, T. K., Suh, H., Hood, L., and Alt, F. W. (1986). Cell (Cambridge,Mass.) 44,251. Yang, S. Y., Denning, S., Mizuno, S., Dupont, B., and Haynes, B. F. (1988a).J.E x p . Med. 168,1457. Yang, S. Y., Rhee, S., Welte, K., and Dupont, B. (1988b).J. Zmmunol. 140,2115. Zacharchuk, C. M., Mercep, M., Chakraborti, P. K., Simons, S. S., Jr., and Ashwell, J. D. (1990).J. Zmrnunol. 145,4037. Zacharchuk, C. M., Mercep, M., June, C. H., Weissrnan, A. M., and Ashwell, J. D. (1991). J. Immunol. 147,460. Zamoyska, R., and Parnes, J. R. (1988). EMBOJ. 7,2359. Zamoyska, R., Derham, P., Gorman, S. D., von Hoegen, P., Bolen, J. B., Veillette, A., and Parnes, J. R. (1989).Nature (London)342,278. Zijlstra, M., Bix, M., Simister, N. E., Loring, J. M., Raulet, D. H., and Jaenisch, R. (1990). Nature (London)344,742. Zinkernagel, R. M., Callahan, G. N., Althage, A., Cooper, S., Klein, P. A., and Klein, J. (1978).J. E x p . Med. 147,882. Zlotnik, A., Ransom, J., Frank, G., Fischer, M., and Howard, M. (1987).Proc. Natl. Acad. Sci. U.S.A.84,3856. Zubiaga, A. M. Mufioz, E., and Huber, B. T. (1991).J. Immunol. 146,3849. Zufiiga-Pflucker,J. C., and Kruisbeek, A. M. (199O).J.Zmmunol. 144,3736. Zufiiga-Pflucker, J. C., Smith, K. A., Tentori, L., Pardoll, D. M., Longo, D. M., and Kruisbeek, A. M. (1990a).Deu. Immunol. 1,59. Zufiiga-Pflucker,J. C., Jones, L. A., Longo, D. L., and Kruisbeek, A. M. (1990b).J. E x p . Med. 171,427. This article was accepted for publication on 6 January 1992.

ADVANCES IN IMMUNOLOGY, VOL. 51

Role of Perforin in Lymphocyte-MediatedCytolysis HIDE0 YAGITA, MOTOMI NAKATA, AKEMI KAWASAKI, YOlCHl SHINKAI, AND KO OKUMURA Department of Immunology, Juntendo University School of Medicine, Tokyo 113, Japan

1. Introduction

Perforin, also termed pore-forming protein (PFP) or cytolysin, has been the central subject of extensive studies during the past 10 years and is a primary candidate as mediator of the cellular cytotoxicity exhibited by cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells (1-5). The granule exocytosis model of lymphocyte-mediated cytolysis (Fig. l),wherein perforin is expected to play a central role, has been proposed on the basis of a great deal of circumstantial evidence as follows:

1. Ring-like tubular lesions that look similar to the lesions formed by complement-mediated cytolysis were found on target cell membranes after NK- or CTL-mediated killing (Fig. 2) (6-8). 2. Cytoplasmic granules isolated from NK and CTL cell lines were directly cytotoxic against various target cells and formed pores on the target cell membrane in a manner similar to that exhibited by intact killer cells (9-14). 3. A pore-forming protein (perforin) was purified from the cytoplasmic granules (15-19). 4. Cytoplasmic granules in CTLs and NK cells were reoriented toward target cells following CTL and N K binding to susceptible target cells (20-24). In addition, exocytosis of the granule contents, including perforin, into the cleft formed between killer and target cell membranes, and their binding to target cell membrane, were observed by cinemicrography and electron microscopy (24-26). 5. Release of granule contents, such as serine esterases and proteoglycans, was observed upon killer-target cell interaction (27-35). However, the central role of perforin and universality of the granule exocytosis model have been challenged based on evidence that some CTLs apparently lacking perforin expression were found (36-40) and that some CTLs and lymphokine-activated killer (LAK) cells lysed certain target cells without apparent granule exocytosis (41-45). Re215 Copyright B 1993 b y Academic Presa. Inc All light\ ol repinduction in dnv form ic3erved

216

HIDE0 YAGITA ET AL.

FIG. 1. Granule exocytosis model of lymphocyte-mediated cytolysis. (1)Release of granule contents by exocytosis of the cytoplasmic granule. (2) Binding of perforin to target cell membrane. (3)Insertion of perforin into target cell membrane. (4)Polymerization of perforin. (5)Channel formation, which leads to influx of other granule contents, such as serine esterases (SE),tumor necrosis factor (TNF)-like molecules, and TIA-1 (a cytolytic granule-associated protein), and osmotic lysis. (6) Induction of nuclear disintegration.

cent isolation of perforin cDNA (46-51) and generation of monoclonal antibodies (mAbs) against perforin (52) have enabled us to reexamine the contribution of this potent cytotoxic molecule to CTL and NK cell-mediated cytolysis. We discuss here the physiological relevance of the perforin-dependent and -independent pathways of lymphocytemediated cytolysis, mainly based on our recent observations.

II. Structure of Perforin

Perforin purified from cytoplasmic granules of murine NK and CTL clones is a glycoprotein exhibiting an apparent molecular weight of

PERFORIN IN LYMPHOCYTE-MEDIATED CYTOLYSIS

217

FIG.2. Pores detected on target cell membranes after CTL-mediated cytolysis by electron microscopy.

65,000 to 70,000 on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Recently, we and others have cloned cDNA encoding murine, human, and rat perforins and have revealed their primary structures (46-51). Mature perforins of all three species consisted of 534 amino acids with a calculated molecular weight of about 60,000 for the core polypeptides. Two potential N-glycosylation sites were conserved among these species and murine and rat perforins contained an additional glycosylation site, which resulted in the molecular weight heterogeneity observed on SDS-PAGE. Murine and human perforins exhibited 68% overall amino acid homology; a similar homology was found for human and rat porforins (69%),and mouse and rat porforins exhibited 85% homology. All 20 cysteine residues were completely conserved among these species. Human perforin had been called C9-related protein (CSRP), based on functional and structural similarities between perforin and the complement components consisting of the membrane attack complex (MAC) (53-58). Determination of primary amino acid sequences of these molecules revealed a low overall but high regional homology between perforin and the complement components. The N-terminal 100 amino acids and the C-terminal 100 amino acids were quite unique to perforin. It has recently been reported that the first 34 amino acids of this N-terminal region could spontaneously insert into membranes and

218

HIDE0 YAGITA ET AL.

had lytic activity, indicating that this region might be important for pore formation (59). The perforin central portion covering approximately 300 amino acids showed about 20% overall homology to the complement components C6,C7, C8a, C8$, and C9 (consisting of the MAC) (60-67). The highest homology was detected between perforin residues 190-220 and the lipid-binding domains of the complement components, and between perforin residues 355-388 and the epidermal growth factor (EGF)-type cysteine-rich domains of the complement components (Fig. 3). In particular, the lipid-binding domains of mouse, rat, and human perforins and C9 form a similar amphipathic a-helix structure (Fig. 4), which is supposed to be important in inserting and polymerizing in the lipid bilayer by exposing the hydrophobic side toward the lipid outside and the hydrophilic side toward the inside, to line up the inner surface. Based on such structural similarities, pore formation b y perforin is supposed to proceed in a manner similar to that proposed for the MAC construction (68-72), except for the initial step. Although the C5b-8 complex acts as an acceptor of C9 in the complement system, perforin directly binds to phosphorylcholine moieties of the target cell membrane in the presence of Ca" (73). This process is not temperature dependent and proceeds even at 4°C. At 37"C, perforin is supposed to

lipid binding domain

EGF-type cysteine-richdomain

I1 L-UL

mouse perforin 355

N C S R P C R S G Q H

D S K VTNQDW-CP

human perforin 355

D C S R P C P P G R Q

G S A V

human

c9

487

human

C8a

467

-

human

C8fl

450

- VGS

human

c7

434

- P Y T FGAACEQG

- F K F L G S

Q!LMv

FIG. 3. Homology between perforin and complement components consisting of membrane attack complex. EGF, Extracellular growth factor.


mouse Berforin

219

mouse C9

Hydrophilic Hydrophobic rat cyfolysin

human Derforin

FIG.4. Amphipathic a-helix structures of the lipid-binding domains of'perforin and C9 (Edmundson wheel plot analysis).

undergo a conformational change similar to that found in C9, and to insert into the lipid bilayer of the target cell membrane, and polymerization proceeds in a temperature-dependent manner (74). Finally, approximately 20 perforin molecules form a tubular complex with an inner diameter of 16 nm (Fig. 5), which was detectable as ring-like membrane lesions on electron microscopy (Fig. 2). The transmembrane channel formed in this way leads to the efflux of cytoplasmic molecules from the target cells and the influx of ions, water, and other cytoplasmic granule contents exocytosed from the killer cells, resulting in osmotic lysis and nuclear disintegration (Fig. 1). 111. Expression of Perforin

One major claim arguing against the central role of perforin in lymphocyte-mediated cytolysis was the failure to detect perforin expression in CTLs elicited in zjizjo; these CTLs exhibited potent cytotoxic

220

HIDE0 YAGITA ET AL. PFP

-I

1Onm -4

complement-mediated membrane pore formation

e

l 6nm-4

lymphocyte-mediated membrane pore formation

FIG.5. Schematic structure of transmembrane channels formed by membrane attack complex of complement and perforin. PFP, Pore-formingprotein.

activity (36-40). This led to a strong suspicion that perforin expression in CTLs and NK cells might be an in vitro artifact and perforin might have no role in physiological conditions. However, such observations were obtained by estimating perforin expression by virtue of hemolytic activity or Western blotting using polyclonal antisera with a very low sensitivity. Recently available perforin cDNA probes and mAbs raised against recombinant perforin have enabled us to reexamine perforin expression by more sensitive means, including Northern blotting, in situ hybridization, and immunohistochemical staining (Fig. 6). Definitive proof indicating that perforin expression in vivo in physiological and pathological conditions has been accumulating as follows, supporting the role of perforin in vivo.

A. PERFORIN EXPRESSION IN MURINELYMPHOCYTES By immunohistochemical staining with mAbs raised against murine recombinant perforin, we previously detected perforin expression in 12-15% of asialo-GM: NK cells and 7-21% of CD8+ T cells, but not in CD4+ T cells in normal spleen cells (52).In further experiments (75), the cells expressing perforin among asialo-GM: cells were highly enriched in the NK1.1+ subpopulation, where all the N K activity resided (Table I). We also examined perforin expression in primary CTLs in peritoneal exudate lymphocytes (PELS)after intraperitoneal immunization with allogeneic spleen cells (Table I). Perforin was detected in 48% of CD8+ T cells and was significantly enriched in the CD8+ asialo-GMT subpopulation that exhibited the highest CTL activity (75). Perforin expression in PEL-CTL has been also demonstrated by Northern blot analysis (76). In other physiological


22 1

FIG.6. Immunohistochemical staining of murine CTL clone with antiperforin mAb.

conditions, granulated metrial gland cells exhibiting an NK-like phenotype (77,78),skin y / 6 T cells (Thy-l+dendritic epidermal cells) (79), and both alp and y / 6 T cells in intestinal intraepithelial lymphocytes (80,81) have been demonstrated to express perforin by Northern blotting, in situ hybridization, or immunohistochemical staining. TABLE I PERFORIN EXPRESSION IN MURINE SPLENIC NK CELLS AND PEL-CTLs Perforin+ Cells Cell

Phenotype ~~

Splenic NK ceIfs"

PEL-CTLS~

(%)

~

AsiaIo-GMlt NK1.1+ CD8+ CD8+ asialo-GMI+ CD8+ asialo-GM1-

40.8 83.0 48.0 66.1 22.4

Asialo-CMl' or NK1.l' NK cells were isolated from nylon woolnonadherent C57BL16 spleen cells by fluorescence-activated cell sorting. bCD8' T cell subpopulations were isolated from nylon woolnonadherent C57BL16 peritoneal exudate lymphocytes (PELS)5 days after an i.p. injection of BALB/c spleen cells by fluorescence-activated cell sorting.

222

H I D E 0 YAGITA ET AL.

Un-der pathological conditions, perforin expression has been demonstrated in infiltrating lymphocytes in damaged tissue specimens in lymphocytic choriomeningitis virus infection (82,83),viral myocarditis (84), allograft rejection (85), and autoimmune diseases observed in nonobese diabetic (NOD) mice (86) and NZB/W F1 mice (lupus) (87). In all cases, in vivo expression of perforin appears to be in a good correlation with cytotoxic activity and tissue injury.

EXPRESSION IN HUMAN LYMPHOCYTES B. PERFORIN Human peripheral blood lymphocytes (PBLs) expressed low but substantial levels of perforin mRNA (49,50), which mainly accounted for the constitutive expression in NK and y/6 T cells (88,89). We recently demonstrated perforin expression in unstimulated and stimulated PBL subpopulations by immunohistochemical staining with an antimurine perforin mAb cross-reacting with human perforin (89)or an antihuman perforin mAb raised against recombinant human perforin (90) (Table 11).In unstimulated PBLs, perforin was highly expressed in almost all CD56' CD3- NK cells and y / 6 T cells (89). Perforin was moderately expressed by 20-30% of CD8+ T cells, with some individual differences, but not in CD4+ T cells. CD8+ T cells could be subdivided into two functionally distinct subsets by their expression of C D l l b . Almost all CD8+ T cells expressing perforin resided in the TABLE I1 IN UNSTIMULATED AND STIMULATED CYTOLYTIC POTENTIAL AND PERFORIN EXPRESSION SUBPOPULATIONS OF HUMAN PERIPHERAL BLOODLYMPHOCYTES ~~~

~~

~~

Unstimulated Cell

Phenotype"

Cytol ytic potential" ~

NK cells T cells

B cells

CD56+ CD3CD4+ CD8+ CD8+ C D l l b + CD8+ CD llb yIGTCR+ CD20+

High

No Low Medium

No High

No

Cytolytic potential

Perforin+ cells (%) ~

95-100 0 20-30 95-100 0-5 95-100 0

~

Stimulated'

~~

Perforin cells (%

~~

ND Medium High High High High

ND

ND 5-20 60-80 100 40-60 100 ND

a Subpopulations of the indicated phenotype were isolated from human peripheral blood mononuclear cel by fluorescence-activated cell sorting. Cytolytic potential was tested against P815 target cells in the presence of anti-CD3 or anti-CD16 mAbs for cells or NK cells, respectively. ' Isolated T cell subpopulations were stimulated in anti-CD3-coated plates for 2 days and then cultured wit 100 U/ml of IL-2 for 5 days; ND, not determined.

*


223

C D l l b + subset, suggesting that this subset represents in vivo effector CTLs. However, perforin was highly inducible in 40-60% of the CD8+ C D l l b - T cells after stimulation with anti-CD3 mAb and IL-2, suggesting that this subset contains CTL precursors. Interestingly, 5-20% of CD4+ T cells also acquired perforin after similar stimulation. In all instances, perforin expression was always in good correlation with the cytolytic potential exhibited by the unstimulated and stimulated human PBL subpopulations (Table 11). A very similar distribution of perforin and its correlation with the cytolytic potential of the in uitro-stimulated PBLs were demonstrated in viuo in the PBLs of infectious mononucleosis patients infected with Epstein-Barr virus (EBV) (90). In other pathological conditions, perforin expression by the lymphocytes infiltrating damaged tissues has been demonstrated: in acute rejection of cardiac (85,91)and renal (92)allografts, in rheumatoid arthritis (85), and in myocarditis (93). In the latter case, tubular lesions on damaged myocytes were demonstrated by electron microscopy, directly indicating the contribution of perforin. C. REGULATION OF PERFORIN EXPRESSION

The putative promoter and enhancer sequences in the 5’ flanking regions of human and murine perforin genes have been characterized (94,95). The 5’ flanking sequence closest to the transcription initiation site contains a GC box both in humans and in mice, a CCAAT box in humans, and two TATA boxes in mice. Further upstream, several other enhancer elements are found, including TPA-, CAMP-, and interferon-y-responsive elements. AP-2-like and NF-KB-like elements are also found. Both human and murine 5’ flanking regions contain a highly repetitious GCCCTG consensus sequence of unknown significance. In human PBLs, both perforin protein and mRNA are constitutively expressed in NK and y16 T cells and are not significantly up-regulated by IL-2 exposure (88,89). In CD8+ (YIPT cells, in contrast, IL-2 alone induced a rapid and transient increase of perforin mRNA (88,96). IL-6 alone was unable to induce perforin expression, but acted synergistically with IL-2 (97). TGF-P acted inhibitory in IL-2 (and IL-6) induction of perforin mRNA (98). In all systems tested in these studies, the cytolytic potential induced in CD8+ T cells was always in good correlation with perforin expression. Taken together, all of the findings indicate that IL-2 is the primary cytokine responsible for regulating perforin expression in CD8+ T cells. As was the case for other stimuli, OKT3 mAb rapidly up-regulated perforin expression by peripheral blood T cells in the presence of

224

H I D E 0 YAGITA ET AL

monocytes (88).This appeared to result from a direct signal from CD3 independently of IL-2, since anti-IL-2 or anti-IL-2 receptor mAbs did not inhibit up-regulation (our unpublished observation). Although the tetraphorbol acetate (TPA)-responsiveelement has been found in the 5' flanking sequence, phorbol myristate acetate (PMA)and/or calcium ionophore failed to induce perforin mRNA in purified peripheral blood T cells (88).At present, the mechanisms responsible for the differential expression in N K and y/6 T cells (constitutive) and CD8+ T cells (inducible), and the mechanisms responsible for the restricted expression in NK and T cells, remain unclear. IV. Role of Perforin in Cytolysis

The propriety of the perforidgranule exocytosis model has been challenged by extensive studies using various T cell clones, including classical CD8+ CTLs and recently characterized CD4+ helper/killer T cells. It has been reported that most CD4+ helper T cell clones, especially those of the T H type, ~ exhibit cytotoxic activity against antigenpresenting cells or when redirected by anti-CD3 mAbs (99-104). Through such studies, it has become evident that both perforindependent and -independent pathways exist. The following discussions of this issue are mainly based on recent studies by us and others wherein the contribution of perforin was verified at the molecular level. A. VERIFICATIONIN CELLLINES All murine CD8+ CTL clones so far tested express perforin, as estimated by Northern blotting or immunohistochemical staining (our unpublished observation). In addition, the majority of CD4+ T H ~ clones and some T H 1 clones have been demonstrated to express perforin and lyse anti-CD3-coated erythrocytes as well as nucleated target cells (102). In contrast, perforin could not be detected in several T H ~ and some T Hclones ~ by Northern blotting, although these cells efficiently lysed nucleated target cells but not erythrocytes (102). We also have observed that some CD4+ T cell clones, apparently lacking perforin as estimated by Northern blotting, reverse transcription polymerase chain reaction (RT-PCR),immunohistochemical staining, and enzyme-linked immunoassay (ELISA), could lyse nucleated target cells as efficiently as CD8+ CTL clones expressing perforin (103).This clearly indicates the presence of a perforin-independent cytolytic mechanism. The consistent correlation between perforin expression and ability to lyse anti-CD3-coated erythrocytes observed in all CD8'


225

and some CD4+ CTL clones suggests that the perforin-dependent pathway also operates against nucleated target cells. However, cyclosporin A (CsA),which can efficiently inhibit exocytosis as estimated by BLT esterase release, profoundly inhibited the lysis of erythrocytes but did not significantly inhibit the lysis of nucleated target cells (102). This suggests that the perforiniexocytosis-independentmechanism is also operative in target cell lysis even by CTLs expressing perforin. We represent here our recent study using a murine CTL hybridoma. KSH 4.13.6 is an H-2d-specific CTL hybridoma with a CTL activity that is reversibly inducible by depletion and supplementation of rat concanavalin A (ConA) supernatant (RCS) (105).As indicated in Table 111, the CTL activity of KSH 4.13.6 was inhibited by anti-LFA-1 mAb, EGTA, and CsA, indicating that it is contact and exocytosis dependent. CTL activity was lost after RCS depletion, but no changes in phenotype and in ability to form conjugates with the target cells were observed (Table IV). Electron-dense cytoplasmic granules were also retained and could be reoriented toward the contact site upon binding to target cells, as estimated by electron microscopy. The only change detected was the loss of perforin and BLT estrase, as estimated by Northern blotting, immunohistochemical staining, and ELISA. This suggests that these molecules were critical for exhibiting CTL activity in KSH 4.13.6. In this system, however, it is not clear whether perforin and/or BLT esterase alone or any other undefined molecules were important. We are now trying to express perforin and/or granzyme A Constitutively in this hybridoma by cDNA transfection; the goal is to test whether perforin alone or in combination with granzyme A is sufficient to reconstitute CTL activity.

TABLE 111 CHARACTERIZATION OF CTL ACTIVITY OF KSH 4.13.6 Experiment 1

Treatment

2 mM EGTA 10 pg/ml CsA

2

Anti-LFA-1 (10 pg/ml)

Cytotoxicity (LU/106 cells)" 22.1

23-fold increase in mRNA for TNF-a and IL-6 0.5 to 1 hour after activation (55).Increased secretion of biologically active TNF-a was maximal 4 hours after activation. BMMCs had a low level of basal transcription of IL-6 detected by Northern analysis. Interestingly, only about 50% of BMMCs demonstrated increased mRNA for IL-6 1hour after IgE-mediated activation, raising the possibility that only a subpopulation of BMMCs transcribe cytokines after activation. The amount of granule mediators released was only 30 to 35%. Therefore, it is possible that only a proportion of BMMCs were activated by cross-linking of Fc,RI, or, alternatively, most were activated to release a proportion of their granule constituents and there was heterogeneity in the capability of BMMCs to generate cytokines. There are as yet limited data on the generation of cytokines by human mast cells. Steffen et al. reported the presence ofTNF-a in mast cellslbasophils derived from culture of human bone marrow (56). mRNA for TNF-a was detected and localized to metachromatically staining cells by in situ hybridization. Immunoreactive TNF-a was also demonstrated in the granules of these cells. Benyon has demonstrated cytotoxicity of purified human skin mast cells toward WEHI164 fibrosarcoma cells, which was inhibited by antibody against TNF-a (57). Mast cell cytotoxicity was enhanced by cross-linking of Fc,RI. Preliminary data suggest that purified human pulmonary mast cells contain TNF-a bioactivity (58), and there are preliminary data on the localization of IL-4 to human pulmonary mast cells in bronchial biopsies from asthmatic individuals (M. K. Church, personal communication). Thus, rodent mast cells produce a wide range of cytokines in response to IgE-mediated stimuli. There are limited data on human

PATHOBIOLOGY OF BRONCHIAL ASTHMA

331

mast cells, but it is likely that they will also be found to produce a range of cytokines. Nevertheless, these studies do not address the question of the relative importance of mast cell-derived cytokines compared to those derived from other cells in allergic diseases. It is tempting to postulate that the mast cell may play a pivotal role in the recruitment and priming of inflammatory cells following allergen challenge in sensitized subjects. This would be of clear relevance to seasonal and other types of allergic asthma. It is also possible that mast cells may play an important role in the recruitment and activation of T cells and eosinophils, and in the regulation of local IgE production within the airways in chronic asthma.

C. EOSINOPHILS It has been known for a long time that bronchial asthma is associated with eosinophilia of the blood and lung (59). With the finding that these cells were able to metabolize histamine, inactivate leukotrienes and platelet-activating factor (PAF), and suppress histamine release, they were attributed a protective role in allergic responses (60). Subsequent research has suggested that they may serve a proinflammatory function since they are capable of secreting preformed and newly generated mediators capable of eliciting tissue damage (61,62).Their presence in the airway and ability to express a low-affinity receptor for IgE (63)provide a mechanism whereby these cells can be activated via IgE-dependent mechanisms. With allergen challenge, there is a transient blood eosinopenia at 6 hours postchallenge that is followed by a progressive eosinophilia occurring up to 24 hours postchallenge (64).The postchallenge eosinophilia appears only in those asthmatics in whom a late asthmatic response develops and correlates not only with the magnitude of this reaction, but also with the basal airway responsiveness (64). The recent finding that circulating eosinophil precursors increase during the late asthmatic reaction (65),and that their numbers fluctuate in relation to seasonal exposure in atopic subjects (66), suggests a role for allergen in stimulating eosinophil production by the bone marrow. Factors that may be responsible for stimulating eosinophilopoiesis are granulocyte-macrophage colony-stimulating factor, IL-3, and IL-5 (67-70). In addition to supporting colony growth and maturation (71-74), these cytokines may also liberate cells from the bone marrow (67) and prime them for augmented proinflammatory functions (71-74). The observation that human eosinophils produce IL-3 and GM-CSF when stimulated by ionophore or IFN--y suggests additional mechanisms by which eosinophils may augment the local inflamma-

332

JONATHAN P. ARM A N D TAK H . LEE

tory response, including autocrine effects of IL-3 and GM-CSF upon eosinophil survival and function (75,76). An increased number of eosinophils in asthmatic airways is an almost constant finding. Several centers have identified increased numbers of eosinophils in the bronchial mucosae of asthmatic individuals (12,13,15,77). Electron microscopic and immunohistochemical studies demonstrate activation of the infiltrating eosinophils as evidenced by electron lucency of the granule matrix, loss of the central core of the granules, and positive staining with EG2, a monoclonal antibody that recognizes the cleaved form of eosinophil cationic protein (ECP) (3,12,13,15) (Fig. 2). Peripheral blood eosinophilia, levels of eosinophils and ECP in BAL fluid, and numbers of intraepithelial eosinophils correlate with the severity of asthma (12). The number of EG2-positive cells was also higher in individuals with airway hyperresponsiveness than in those without (15). Bronchial provocation with allergen elicits eosinophil influx into the airway lumen (78,79). Metzger and co-workers have shown that there is a prominent BAL eosinophilia at 4 hours and persisting for 24 hours after allergen challenge (78). The infiltrating eosinophils demonstrate

FIG.2. EG2-positive cells in the submucosa of asthmatic airway (arrows).Note the loss of epithelium (E) and subepithelial collagen deposition (C).


333

appearances consistent with degranulation. DeMonchy and associates have shown that after allergen challenge there occurs an increase in the ECP/albumin ratio, thereby providing further evidence for eosinophi1 degranulation during the allergen-provoked asthmatic response

(79). Four highly charged arginine-rich proteins have been located in the granules of the human eosinophils, namely, major basic protein, eosinophil cationic protein, eosinophil peroxidase (EPO), and the eosinophil-derived neurotoxin (EDN) (80). There is compelling evidence that MBP may be related to the inflammatory changes and tissue damage seen in the bronchial mucosa. MBP is toxic to tracheal epitheM (81).This is well below lial cells in concentrations as low as 9 x the to lop5 M concentrations found in asthmatic sputum (61). Using an immunofluorescent technique, MBP deposition is seen in the bronchial wall and in the mucus in asthmatic lung tissue obtained at necropsy (62).The sites of MBP deposition coincide with the widespread epithelial damage characteristic of the airways in severe bronchial asthma. MBP levels in sputum correlate with disease activity (61) and the levels decrease after appropriate treatment. In one study, the concentration of MBP correlated with indices of airway responsiveness and the number of ciliated epithelial cells recovered by bronchoalveolar lavage (6). ECP is also located in the eosinophil granule matrix. The levels of ECP in serum rise after allergen-provoked asthma and during the pollen season in atopic individuals (82,83).ECP is neurotoxic and is toxic to guinea pig trachea (84).It is present in the submucosae of patients who have died from asthma (85). Circulating eosinophils of subjects with blood eosinophilia display a range of densities upon separation by discontinuous density centrifugation. In asthma, there is a predominance of hypodense cells. These cells are believed to be in an activated state, as shown by the increased oxygen consumption, phagocytic and cytotoxic capacity, increased spontaneous release of granule contents, and augmented generation of lipid mediators. Thus, the capacity of these activated eosinophils to effect tissue damage by releasing cytotoxic granule contents and by the generation of the newly formed mediators suggests that they are important proinflammatory cells in the asthmatic process.

Eosinophils and Adhesion Mechanisms The mechanisms responsible for the selective recruitment of eosinophils into the airway mucosa is not known. The endothelial cell is a major regulatory step in the passage of leukocytes into tissues. The signals generated at the site of allergic inflammation can activate both

334

JONATHAN P. ARM AND TAK H. LEE

endothelial cells and circulating leukocytes to become adhesive for one another. Thus, it is pertinent to reflect on the adhesion molecules involved in leukocyte/endothelial cell interactions. Endothelial cells express adhesion molecules upon activation with appropriate stimuli, such as cytokines and endotoxin (86-89). The time course of expression depends upon whether the molecules are contained within intracellular stores or whether new protein synthesis is required. It is likely that surface expression of GMP-140 on endothelium may be involved in the initial phase of neutrophil migration. The mechanism does not require protein synthesis and is at least partially attributable to the redistribution of GMP-140 from the granule membrane (90).GMP-140 has been cloned and is a member of the selectin cellular adhesion molecule (LEC-CAM) family, having similarities to endothelial leukocyte adhesion molecule-1 and MEL-14 (91). Later phases of leukocyte infiltration are likely to be related to the induction or up-regulation over a matter of hours of cytokine-inducible endothelial adhesion molecules by a process involving de nouo protein synthesis. Several of these molecules have been identified; endothelial leukocyte adhesion molecule-1 (ELAM-1) (86), intercellular adhesion molecule-1 (ICAM- 1) (92), and vascular cell adhesion molecule-1 (VCAM-1) (89). ELAM-1 is a 110- to 115-kDa single-chain glycoprotein expressed on endothelial cells after stimulation with IL-1, TNF-a, or lipopolysaccharide (LPS) (86).The ligand for ELAM-1 has been identified as the sialylated Lewis x carbohydrate (93).The expression of ELAM-1 is dependent upon new protein synthesis, peaks at 4 to 6 hours, and decreases to near basal levels in cultured cells by 24 hours. ELAM-1 is not constitutively expressed in normal tissues but has been noted on the endothelium at sites of allergic inflammation and experimental delayed hypersensitivity reactions in uiuo (88,94). Incubation of skin biopsies from atopic individuals with allergen for 5 hours i n uitro also led to expression of ELAM-1 (94). Preincubation with antibodies to TNF-a alone and antibodies to IL-1 alone did not alter the allergeninduced expression of ELAM-1. However, preincubation with both anti-IL-1 and anti-TNF-a completely suppressed ELAM-1 expression. ICAM-1 is a 90-kDa single-chain glycoprotein that is basally expressed on endothelial cells and is markedly up-regulated by IL-1, TNF-a, lymphotoxin, LPS, or IFN-y (87,95). ICAM-1 has been cloned and is a member of the immunoglobulin supergene family with five immunoglobulin-like domains of C2 type (92). The increase in ICAM-1 expression on the cultured endothelial cells occurs well after ELAM-1 and plateaus at 24 hours, with expression being sustained as


335

long as the cytokine is in the medium (96).Intensified ICAM-1 expression in vivo by dermal endothelial cells has been observed in skin biopsy tissue following intradermal injection of antigen that elicited a cutaneous late-phase response (88). ICAM-2 is a truncated form of ICAM-1, containing only two immunoglobulin-like domains. It is constitutively expressed in endothelial cells but is not subject to regulation by IL-1, TNF-a, or LPS (97). VCAM-1 is another member of the immunoglobulin supergene family, having six immunoglobulin-like domains of C2 type (89). Its receptor is VLA-4, a p-1 integrin, which binds VCAM at a site distinct from its fibronectin-binding domain (98). Recent work has shown that ICAM-1 and ELAM-1 bind both neutrophils and eosinophils, whereas VCAM-1 binds eosinophils preferentially (99,100). On leukocytes, the adhesion molecules that act as counterligands for ICAM-1 consist of the LEU-CAM family of p-2 integrins (101).These comprise three heterodimers with a common p chain (CD18) and different a chains designated LFA-1 (CDlla), MAC-1 (CDllb), and P150,95 (CDllc). Endothelial Iigands for C D l lalCD18 are ICAM-1 and ICAM-2 and for CDllb/CD18 may be ICAM-1. No adhesion ligand on endothelial cells for C D l l c has yet been identified. LFA-1 is found on all leukocytes, whereas MAC-1 and P150,95 are found only on phagocytes and large granular lymphocytes. LFA-1 is involved in lymphocyte adhesion and all three heterodimers are involved to varying degrees in neutrophil, eosinophil, and mononuclear cell adhesion. Leukocyte p-2 integrins are expressed on the surface of leukocytes in an inactive form (102). Stimulation of the cells with chemotaxins leads to activation of the receptors (102,103) and an increase in their cell surface expression (104-106). It has recently been shown that ELAM-1 may activate MAC-1 on the surface of neutrophils (107); recombinant soluble ELAM-1 was also shown to be a neutrophil chemoattractant. Whether ELAM-1 up-regulates MAC-1 on eosinophils was not studied; however, these experiments serve to show one level of complexity in the interaction between different endothelial adhesion molecules. The interest in the role of adhesion molecules in asthma was heightened recently when Wegner and colleagues (108)identified a relationship between eosinophilic infiltration and airway responsiveness in a primate model of asthma. They demonstrated that the presence of eosinophils in the airways following repeated administration of aerosolized ascaris extract to sensitized animals appeared to be predictive of the intensity of the early asthmatic response to inhaled antigen. In addition, antigen challenge induced a prolonged and specific eosinophil influx into the airways, which is evident at 6 hours after

336

JONATHAN P. ARM A N D TAK H . LEE

challenge and persisted for 7 days. This was associated, both in terms of magnitude and time, with increased airway responsiveness to methacholine. The administration of anti-ICAM-1 monoclonal antibody in vivo reduced eosinophil infiltration and airway hyperresponsiveness following antigen inhalation challenge. The type of response in the airways was dependent upon the model employed. When a subset of intrinsically hypereosinophilic animals was given a single inhaled allergen challenge, there was both an immediate and a delayed asthmatic response. The latter was associated with a neutrophil influx into the airways and an associated airway hyperresponsiveness, and was inhibited by anti-ELAM-1 but not anti-ICAM-1 (109). The apparently different roles of ELAM-1 and ICAM-1 in these different models of asthma are consistent with the time course of their induction. Clearly, by successfully determining the key molecular interactions that might be responsible for selective eosinophil influx in asthma and the development of bronchial hyperresponsiveness, novel targets for the development of antiinflammatory therapies in allergic disease may be elucidated.

D. NEUTROPHILS Neutrophils have been reported to be associated with the transient bronchial hyperresponsiveness induced by ozone (110).However, the evidence for the involvement of these cells in the mechanisms of bronchial asthma is much less certain. The presence of a few neutrophils in epithelium is probably a normal phenomenon because neutrophils may be found even in the lavage fluid of normal subjects (6). Neutrophil numbers were even more numerous in the control specimens in a recent biopsy study of asthmatic patients (77). A number of studies have not shown any significant difference in the neutrophil infiltration between asthma and normal individuals. Following allergen and excercise challenge, a heat-stable high-molecular-weight neutrophil chemotactic activity (NCA) has been detected in the peripheral circulation (31-35). The release ofthis molecule is associated with the increased expression of cell surface markers of neutrophil activation, namely, complement receptors (111)and IgG Fc receptors (112), and with enhanced neutrophil cytotoxicity toward opsonized schistosomulae (113).The extent of neutrophil cytotoxicity and the release of NCA was related to the magnitude of the provoked decrease in FEVl (forced expiratory volume in one second). Premedication with cromolyn sodium prevented both NCA release and the increase in neutrophi1 cytotoxicity. The significance of these findings in relation to the occurrence of airway obstruction and the increase in bronchial hyperresponsiveness has not been elucidated.


337

E. MACROPHAGES AND MONOCYTES

The classification of lung macrophages has traditionally divided them into two categories based upon their anatomic distribution: alveolar or interstitial. The term “alveolar” referred to the macrophages retrieved by bronchoalveolar lavage. However, considerable proof now exists that macrophages also reside at the air-surface interface of conducting airways in the lower respiratory tract of humans. The presence of macrophages in asthmatic lungs has been evaluated in uivo using bronchoalveolar lavage from patients with mild disease, and bronchial biopsies taken from patients with moderate to severe disease. Most studies of BAL fluid have found that the numbers of mononuclear phagocytes present were not increased, compared to the control subjects (6,36,39,78,79,114,115).One exception to this was the study by Metzger and colleagues (116). In contrast, immunohistochemistry of the bronchial biopsy specimens showed that the submucosa had a significantly increased macrophage population in asthmatic patients (77). The macrophage population had phenotypic characteristics of peripheral blood monocytes, suggesting that they had migrated recently into the lung. HLA class I1 antigen was expressed on the infiltrating cells of the airway mucosa to a greater extent in asthmatic subjects than in normal individuals (77). Metzger has shown that the number of monocytes in the airways increases at 48 hours after antigen provocation in asthmatic patients (116).The mononuclear cells of patients with bronchial asthma demonstrate increased complement receptor expression, as compared with those of normal subjects, and also greater enhancement of receptor expression following stimulation with casein. Capron and co-workers were the first to discover that rat macrophages could be activated by IgE-dependent mechanisms and that they possess a low-affinity IgE receptor (Fc,RII), as opposed to the IgE receptor on basophils and mast cells, which is high affinity and is referred to as Fc,RI. From sequence analysis ofthe cloned cDNAs, it is now clear that the FcR,II on B cells (Fc,RII,) and on U937 cells, a human monocytic cell line (Fc,RIIb), differ at the N-terminal cytoplasmic region but share C-terminal extracellular regions (117). In normal subjects, the percentage of Fc,RII-positive lung macrophages and peripheral blood monocytes is low, approximately 5-10 and 10-15%, respectively (118-120). In atopic individuals, the numbers of IgE Fc,RII-positive lung macrophages and peripheral blood monocytes are increased (120).Patients with severe asthma and atopic dermatitis treated with corticosteroids had the !owest percentage of IgE Fc,RII-positive peripheral blood monocytes, suggesting that the

338

JONATHAN P. ARM A N D TAK H. LEE

expression of this receptor can be modulated by therapy (120). Lung macrophages retrieved by bronchoalveolar lavage from mild atopic asthmatic subjects showed that the percentage of Fc,RII-positive lung macrophages was increased in these individuals to approximately 20%, compared with nonatopic controls (119). Speigelberg and colleagues found that peripheral blood monocytes from severely atopic individuals induced significantly more 51Cr release from IgE-coated red blood cells than did monocytes from nonatopic or mildly atopic asthmatic subjects (118). The number of Fc,RII present on mononuclear phagocytes can be reegulated by a variety of cytokines. Using the U937 cell, it has been shown that IL-4 or IFN-7 enhance Fc,RII gene expression and the production of protein (121). IgE is also able to augment the number of Fc,RII on these cells (122).These observations suggest that cytokines and other molecules, such as IgE, may be important regulators of mononuclear phagocyte Fc,RII in uiuo. Dessaint et al. demonstrated that IgE-antigen complexes activated rat peritoneal macrophages to release lysosomal enzymes and superoxide anion (123). Macrophage products of oxygen metabolism have proinflammatory effects and may contribute to the inflammatory airway reaction observed in patients with asthma. Bach et at. (124) observed that rat peritoneal macrophages were capable of releasing leukotriene C4. Rankin and co-workers (125,126) demonstrated that normal rat macrophages could be activated by monoclonal IgE and its specific antigen to release both leukotriene B4 (LTB4) and LTC4. Several laboratories have now demonstrated that mononuclear phagocytes could have an important role in IgE-mediated diseases. Ferreri et al. (127) challenged peripheral blood mononuclear cells in uitro with chemically aggregated IgE and found that these cells release small quantities of LTB4, LTC4, and PGE2. Fuller et al. (128) observed the release of LTB4, PGF2,, thromboxane Bz, and P-glucuronidase from macrophages obtained from patients with a variety of lung disease when these cells were challenged with anti-IgE. Analysis of bronchoalveolar lavage fluid of patients with asthma after antigen challenge revealed increased amounts of p-glucuronidase, whereas macrophage intracellular levels were decreased (129). These results suggest that macrophage secretory processes can be activated by allergen, acting through Fc,RII. In view of the evidence that alveolar macrophages (AMs) secrete molecules that can influence the functions of inflammatory granulocytes, and the compelling evidence for the participation of eosinophils in airway inflammation, Howell and colleagues have studied the interactions of AMs and eosinophils in asthmatic subjects (130). Eosino-


339

phils incubated with AM supernatants from asthmatic patients, followed by stimulation with A23187, demonstrated an enhanced capacity to secrete LTC4. AM supernatants derived from normal individuals had no enhancing effects when compared with culture medium alone. The activity derived from asthmatic AMs could be neutralized by incubation with specific antibodies to human GM-CSF, suggesting that the major active component is identical or closely related to GM-CSF. This is supported by the observation that pretreatment of eosinophils with recombinant GM-CSF primed the cells for enhanced LTC4 generation following stimulation with A23187. GMCSF is an acidic glycoprotein with a PI of 4.5 and a molecular weight of 22,000. It elutes from size exclusion columns with an apparent molecular weight of between 15,000and 40,000 due to variations in its glycosylation. It stimulates the proliferation and differentiation of normal granulocytes and monocytic stem cells. It also modifies the function of mature granulocytes, leading to enhancement of expression of granulocyte functional antigens 1 and 2, Mol, Leu-M5, and C3bi. GM-CSF induces histamine release from basophils and enhances eosinophil survival in culture. Thus, the presence of GM-CSF in the lung may precondition eosinophils for enhanced proinflammatory functions upon subsequent stimulation, and either alone or in concert with other cytokines, lead to eosinophil colony formation from bone marrow progenitors. GM-CSF may therefore play an important role in the amplification of the eosinophilic inflammation, which is characteristic of asthmatic airways. F. LYMPHOCYTES An area of considerable current interest is the role of the T lymphocyte in the regulation of the inflammation associated with allergy and asthma. T cell-derived lymphokines, IL-4 and IFN-y, are involved in the regulation of IgE production (131,132). Other lymphokines (IL-5, GM-CSF, and 1L-3) control eosinophil production and function (see above) and regulate mast cell differentiation. Necropsy examination of the airways of asthmatic patients showed large numbers of lymphocytes (1).Increased natural killer cell activity has been described in the peripheral blood of asthmatic patients (133). Since natural killer cell activity is an inducible property of T cells and of non-T, non-B lymphocytes, this is a nonspecific indicator of lymphocyte activation. Lymphocytes from the peripheral blood of patients with status asthmaticus demonstrate significant elevations of the expression of T

340


lymphocyte activation markers, namely interleukin-2 receptor (IL-BR), class I1 HLA-DR, and “very late activation” antigen (VLA-l), compared with control subjects (134). Phenotypic analysis of the IL-2Rpositive T cells showed that these cells were exclusively of the CD4 helper-inducer phenotype. Percentages of IL-2R-positive and HLADR-positive but not VLA-l-positive lymphocytes tended to decrease as the patients were treated and improved clinically. The serum concentrations of IFN-.)I and soluble IL-2R were also significantly elevated in patients with acute severe asthma (135).Concentrations decreased as the patients improved clinically during the first 7-day period of hospital treatment. A significant correlation was observed between the degree of airway obstruction as measured by the peak expiratory flow rate and the percentages of peripheral blood T cells expressing IL-2R and the serum concentrations of soluble IL-2R. These observations provide evidence that CD4 T cell activation is associated with acute severe asthma. Biopsies of asthmatic airways reveal a tendency for increased numbers of T cells compared to biopsies from normal controls (15,77),with an increase in the number of cells expressing receptors for IL-2R (CD25),reflecting lymphocyte activation (15,136).CD25-positive cells were greater in airways of asthmatic subjects with bronchial hyperresponsiveness than in those without (15).Hamid et al. used the technique of in situ hybridization to examine the expression of IL-5 in bronchial biopsies from normal and asthmatic subjects (136).Using an antisense cRNA probe for IL-5, they found IL-5 mRNA in the bronchial mucosa of 6 of 10 asthmatic subjects, and in none ofthe 9 controls. No signal was obtained with sense cRNA probes, nor in tissue pretreated with RNase A, demonstrating the specificity of the hybridization. Although the number of patients studied was small, there was a trend for the six IL-5-positive asthmatics to have more severe asthma than those in whom no signal for IL-5 was observed. Further, biopsies positive for IL-5 mRNA also had a greater number of CD25-positive cells, a greater number of eosinophils, and a greater number of EG2positive cells. These data are consistent with the suggestion that IL-5, secreted by activated T lymphocytes, contributes to the recruitment and activation of eosinophils in the bronchial mucosa in asthma. 111. Eicosanoids and the Pathophysiology of Asthma

Several observations suggest that leukotrienes, prostaglandins, and thromboxane are important mediators in bronchial asthma. They are potent proinflammatory and spasmogenic mediators; they are present


34 1

TABLE I1 PEPTIDOLEUKOTR~ENES AND ASTHMA"

Properties Potent bronchoconstrictor agonists Increase mucus secretion Increase vascular permeability Selective hyperresponsiveness to LTE4 in asthma Increase airway hyperresponsiveness Presence in the airways and release in asthma In asthmatic airways at rest Release following allergen challenge, aspirin-induced asthma, isocapnic hyperventilation, and acute severe asthma Leukotriene antagonists Decrease resting airway tone in asthma Attenuate early asthmatic response to exercise, allergen, and isocapnic hyperventilation Attenuate late asthmatic response to allergen and accompanying hyperresponsiveness Efficacy in chronic asthma Evidence for the role of leukotrienes in bronchial asthma. LTE4, Leukotriene E+

in asthmatic airways at rest and during an acute attack of asthma; and the acute asthmatic response to various stimiili is attenuated b y potent and selective receptor antagonists and inhibitors of the relevant enzymes (Table 11). A. LEUKOTRIENES

1. Synthesis and Metabolism The synthesis and metabolism of eicosanoids has been reviewed extensively (137). Metabolism of arachidonic acid by 5-lipoxygenase generates the unstable intermediate 5-hydroperoxyeicosatetraenoic acid (5-HPETE) (138), which is reduced to 5-hydroxyeicosatetraenoic acid (5-HETE)or is converted to an epoxide, leukotriene A4 (139-141) (Fig. 3 ) . LTAQis processed by an epoxide hydrolase to LTB4 (142) or, by a glutathione-S-transferase, to LTC4 (124,143,144).LTCl is cleaved by y-glutamyl-transpeptidase to LTD4, which is cleaved by a dipeptidase to LTE4 (124,143,145-148). LTA4 also undergoes nonenzymatic diastereoihydrolysis to 5S,12R- and 5S,12S-dihydroxy-6-trans-LTB4 somers and to minor products, 5,6-dihydroxyeicosatetraenoicacid diastereoisomers (149). LTC4, LTD4 and LTE4 comprise the activity previously recognized as slow-reacting substance of anaphylaxis (SRS-A), and are collectively known as the sulfidopeptide leukotrienes.

342

JONATHAN P. ARM AND TAK H . LEE Ether Phospholipids

I

I

Acetyl transferase

2-lyso-PAF

1

PAF

Acetyl hydro lase

ARACHl DON IC

5-HETE

+5-HPETE

PGG2

+

PGH l

1

PGD2

I

2

PGE 2

PGF

2a

Hydrolase

PG'2

LTB4

LTC4

-+

LTD4 +LTE

T X A2

4

FIG.3. Generation of platelet-activating factor (PAF) from ether phospholipids and metabolism of arachidonic acid by 5-lipoxygenase and cyclooxygenase pathways to leukotrienes (LT) and prostaglandins (PG), respectively; 5-HETE, 5-hydroxyeicosatetraenoic acid; 5-HPETE, 5-hydroperoxyeicosatetraenoicacid.

With the molecular cloning of 5-lipoxygenase (5-LO) it became apparent that cellular 5-LO activity was dependent upon an additional factor. Osteosarcoma cells transfected with the cDNA for 5-LO were unable to generate leukotrienes upon stimulation with the calcium ionophore A23187, although cell lysates expressed active enzyme (150). Furthermore, a class of compounds, of which MK-886 is an example, inhibit the generation of leukotrienes by intact cells but have no inhibitory effect on soluble 5-LO (151).The target of MK-886 was identified as a membrane protein of M,18,000, termed 5-LO-activating protein (FLAP) (152). Osteosarcoma cells transfected with 5-LO or FLAP alone did not generate leukotrienes upon activation with A23187. Transfection with cDNAs for both 5-LO and FLAP was required for significant generation of leukotrienes.


343

LTB4 is converted intracellularly by a hydroxylase to 20-hydroxy LTB4, and by further oxidation to a biologically inactive molecule, 20-aldehyde LTB4 (153- 155). The sulfidopeptide leukotrienes may be metabolized by granulocytes, upon the triggering of the respiratory burst (156), through an extracellular hydrogen peroxide-peroxidase chloride-dependent reaction. In addition to their generation by inflammatory cells, leukotrienes may be synthesized and metabolized b y lung tissue. Thus, the conversion of LTA4 to LTB4, LTC4, LTD4, and LTE4 has been demonstrated in guinea pig lung (157),and of LTC4 to LTD4 and LTE4 in human lung parenchyma (158).

2 . Biological Activities

LTB4 is a potent proinflammatory mediator. Its in uitro activities are apparent at concentrations as low as lo-” M and include chemokinesis and chemotaxis of human neutrophils and eosinophils (104,159),chemokinesis of monocytes (16O), aggregation of neutrophils (159), enhanced expression of complement receptors on granulocytes (105), release of lysosomal enzymes from neutrophils (161), and augmentation of neutrophil adherence to endothelial cell monolayers (162).In uiuo, intradermal injection of LTB4 promotes a prolonged neutrophil infiltration into human skin, with induration and tenderness 4-6 hours after injection (163). LTB4 also contracts smooth muscle through the biosynthesis of cyclooxygenase products (164). Sulfidopeptide leukotrienes constrict nonvascular smooth muscle, enhance mucus secretion, constrict arterioles, and enhance venopermeability (163,165-167). The activity and binding of the sulfidopeptide leukotrienes in various tissues and cells have been characterized. Stereospecific, reversible, and saturable binding of LTC4, LTD4, and LTE4 have been demonstrated in guinea pig and human lung. The existence of receptor heterogeneity for these agonists in guinea pig lung is suggested by differences in the contractile properties and kinetics of action of the separate leukotrienes, the effects of leukotriene receptor antagonists, and radioligand binding studies (145,168-177). In contrast to the results in guinea pig tissues, a study conducted in the presence of bioconversion inhibitors on intralobar airways isolated from human subjects undergoing surgery for carcinoma of the bronchus did not reveal evidence for multiple leukotriene receptors (178). However, it should be emphasized that data from human tissue are limited; the effects of underlying disease on the expression of the different leukotriene receptors have not been studied and data are not available for asthmatic lung.

344


3. Potency LTC4 and LTD4 are potent constrictors of human airways both in vitro and in vivo (168,179,180). LTC4 and LTD4 are approximately 1000-fold more potent than histamine, on a molar basis, in contracting isolated human bronchi in vitro (180). In normal subjects the concentrations of LTC4 and histamine required to produce a 30% decrease in V30 were 2-20 pg/ml and 2-10 mg/ml, respectively (181).LTC4 was 600- to 9500-fold more potent than histamine and LTD4 was 6000-fold more potent than histamine on a molar basis (182). By comparison, in asthmatic subjects, LTD4 was 140-fold more potent than histamine in eliciting a 30% decrease in V3o (183). Asthmatic subjects were only one-third more responsive to LTD4 than the normal subjects, despite an approximate 100-fold hyperresponsiveness to inhaled histamine. The relative lack of hyperresponsiveness to LTC4 and LTD4 in asthmatic subjects was confirmed in other studies (184,185).In addition, correlation was observed between airway responsiveness to methacholine and the relative responsiveness to LTC4 and LTD4; subjects with the most responsive airways demonstrated the lowest relative responsiveness to LTC4 and LTD4 as compared to methacholine (185).In contrast, the relative potency of LTE4 compared with histamine and methacholine was two to three times greater in asthmatics than in normal subjects (186). Because of the inherent difficulties in comparing studies performed in different subjects using different methodologies, the potencies of LTC4, LTD4, and LTE4 relative to one another and to both histamine and methacholine were compared in normal and asthmatic subjects (187). The airways of asthmatic subjects were 14-fold, 15-fold, 6-fold, 9-fold, and 219-fold more responsive than the airways of normal subjects to histamine, methacholine, LTC4, LTD4,and LTE4, respectively. The cumulative data therefore suggest that the airways of asthmatic subjects are relatively unresponsive to LTC4 and LTD4, but have a marked hyperresponsiveness to LTE4. 4 . Leukotrienes and Airway Hyperresponsiveness Brocklehurst demonstrated that slow-reacting substance of anaphylaxis enhanced the contractile response of guinea pig ileum to histamine in vitro (188). It was subsequently shown that pretreatment of guinea pig tracheal spirals with 10-23 nM LTE4, but not LTC4, or LTD4, enhanced the subsequent contractile response to histamine (168). This effect was not observed when parenchymal strips were contracted with LTE4. A detailed in vitro investigation of LTE4-


345

induced hyperresponsiveness suggested that LTE4 augments the contractile response of guinea pig tracheal spirals to histamine by facilitating cholinergic neurotransmission, and is mediated via the secondary generation of cyclooxygenase products acting at the thromboxane A2 (TP) receptor (189). A similar mechanism may operate in human airways (189). In vivo studies support a role for the sulfidopeptide leukotrienes in enhancing airway hyperresponsiveness in asthma. Inhalation of a bronchoconstricting dose of LTD4 in normal subjects produced an approximate twofold increase in airway methacholine responsiveness (190),which was maximal at day 7 and persisted for up to 2 to 3 weeks (191). In normal subjects inhalation of LTD4 did not significantly enhance the airway response to exercise (192) or histamine (184,193), although it increased the sensitivity of the airways to inhaled PGF2, by approximately sevenfold (184). Normal and asthmatic airway responses in vivo differ not only in their sensitivity to a wide range of pharmacological and nonpharmacological stimuli (194),but also by the presence of maximal airway narrowing to histamine and methacholine in nonasthmatic subjects. Asthmatic subjects show a leftward shift of the dose-response curve and progressive airway narrowing with increasing dose of agonist, whereas the airway response in normal subjects reaches a plateau at mild degrees of airway narrowing (195,196). The degree of maximal airway narrowing is greater in response to LTD4 than to methacholine. Prior inhalation of LTD4 did not change the position of the methacholine dose-response curve, although the maximal airway response to methacholine increased (197).The maximal airway narrowing response to LTD4 was diminished and the LTD4-induced augmentation of maximal airway narrowing in response to methacholine was prevented by pretreatment with inhaled budesonide for 1 week (198). Studies in asthmatic subjects have been more limited. The inhalation of bronchoconstricting doses of LTC4 did not enhance the airway response to ultrasonically nebulized distilled water (199). In contrast, preinhalation of a bronchoconstricting dose of LTC4, LTD4, or LTE4 in asthmatic subjects increased histamine responsiveness by approximately threefold, 4 to 7 hours after inhalation of the leukotriene (193,200). Neither LTC4, LTD4, nor LTE4 elicited any change in airway responses to histamine in normal subjects, although each mediator was administered in a dose that elicited a mean 35% fall in SGaw (airways specific conductance) (193,200).The lack of effect in normal individuals is in contrast to the studies of Kaye (191) and Kern (190) (see above), and may be due to a selective effect of LTD4 on normal

346

JONATHAN P. ARM AND TAK H . LEE

airway responses to methacholine (as opposed to histamine), to the timing of measurements of airway responsiveness, or to individual variability. In addition to the capacity of inhaled leukotrienes to enhance subsequent airway responses to histamine in subjects with asthma, LTC4 may interact synergistically with histamine and PGDz in the acute bronchoconstrictor response (201).

5. Release of Leukotrienes in Asthma Using various physicochemical techniques, leukotrienes have been detected in bronchoalveolar lavage fluid of asthmatic subjects, both at rest and following bronchial challenge. Lam found LTE4 in the BAL fluid of 15 out of 17 asthmatic subjects (9); LTD4 was detected in 2 subjects and 20-hydroxy-LTB4 was found in 12 subjects. Other studies have confirmed the presence of significant quantities of LTC4 and LTB4 in the BAL fluid of asthmatic subjects compared to normal controls (202-204). Following allergen challenge, mean LTC4 levels rose from 64 pg/ml of lavage fluid to 616 pg/ml (203). Following asthma provoked by isocapnic hyperventilation, BAL concentrations of LTB4 and immunoreactive sulfidopeptide leukotrienes rose from baseline levels of 10 and 46 pg/ml, respectively, to 121 and 251 pglml, respectively

(204). Measurement of urinary LTE4 has been used as a marker of sulfidopeptide leukotriene generation (205-207). Increased urinary LTE4 levels have been reported during acute severe asthma and at 3 hours following antigen challenge of asthmatic subjects (41-43), but not following exercise-induced asthma (42).

6 . Leukotriene Antagonists and Inhibitors If leukotrienes play a significant role in the pathogenesis of asthma, then attempts to inhibit their generation or to antagonize their action at specific receptors should be of some benefit. Several studies have found that administration of peptidoleukotriene antagonists leads to bronchodilatation in asthmatic but not in normal individuals (208210), suggesting that leukotrienes may contribute to the resting airway tone in asthma. This effect is additive to that of albuterol (208-210). These agents have also been shown to inhibit the acute asthmatic response to exercise (211,212), allergen (213-215), and isocapnic hyperventilation (216). In addition ICI 204,219 inhibited the allergeninduced late asthmatic response and the increased airway hyperresponsiveness that followed allergen challenge (213).Preliminary data also suggest that administration of leukotriene antagonists leads to an improvement in the severity of clinical asthma (218,219).Compared to placebo, treatment with MK-571 led to a mean 8-14% improvement in


347

FEVI, 30% decrease in morning and evening symptom scores, and an approximate 30% decrease in usage of albuterol (218). There are few data yet on the effects of 5-LO inhibitors in asthma. A significant inhibition of the asthmatic response to cold, dry air was demonstrated (217), and a small inhibition of the early asthmatic response to allergen (220).Although the latter effect was not statistically significant, there was a correlation between the inhibition of urinary LTE4 excretion and the attenuation of the early asthmatic response, suggesting that the lack of clinical effect may have been related to insufficient inhibition of 5-LO in the lung.

B. PROSTAGLANDINS AND THROMBOXANE

1 . Synthesis und Metabolism Arachidonic acid may be metabolized by cyclooxygenase to the cyclic endoperoxides, PGGz and PGHZ, which are then converted b y specific synthesis to thromboxane (TX) Az, or to various prostaglandins, PGDZ, PGFZ,, PGE2, and PGIz (221,222) (Fig. 3). PGEz is the predominant cyclooxygenase product of a number of different types of cells, including epithelial cells and macrophages. PGD2 is the major cyclooxygenase product of the mast cell (223,224)and is metabolized to 9m,ll/3-PGFz, which contracts airway smooth muscle both in vitro and in vivo (225).

2 . Biological Activities

The cyclic endoperoxides, PGGz and PGHz, and TXAz are labile molecules with short half-lives and appear to act at a common receptor (226).They constrict vascular and bronchial smooth muscle and aggregate platelets (227,228). PGIz is active in many tissues, producing vasodilatation, inhibiting platelet aggregation, and relaxing bronchial smooth muscle (229,230).PGEZ has diverse properties, including inhibition of platelet aggregation and contraction or relaxation of vascular and nonvascular smooth muscle (231,232).In human airways it acts as a bronchodilator (233). PGFZ,, PGDz, and its stable metabolite 9a, 11P-PGFZ are potent bronchoconstrictors (225). In addition, PGDz stimulates neutrophil chemokinesis (234), causes vasodilatation, and increases postcapillary venular permeability (163),and may act synergistically with LTB4 in promoting neutrophil infiltration (163). 3 . Potency Initial studies of the bronchoconstrictor effects of prostaglandins were directed to the properties of PGFz,. PGFz, was shown to contract human airways in vivo, and asthmatic subjects were shown to be more

348


sensitive to PGF2, than were normal controls. The airway responsiveness to PGFz, correlates with that to methacholine (235).Furthermore, the airway hyperresponsiveness that is characteristic of asthma is markedly greater toward PGF2, than toward histamine (233,235,236). However, subsequent studies suggested that the airway response to PGF2, in uiuo might not be as simple as originally described. Both Fish (237)and Beasley (238)have reported complex biphasic or triphasic responses to inhaled PGFz,, possibly due to the action of PGFz, on separate receptors mediating bronchodilatation and bronchoconstriction. Studies of the effects of cholinergic blockade on airway responses to PGFz, have yielded conflicting results (236-241), but the cumulative data suggest that the contribution of cholinergic pathways to PGFz,-induced bronchoconstriction is small. Neither a-adrenergic blockade (239) nor pretreatment with cromolyn (239,240) inhibited airway responses to PGFz,. In normal and asthmatic subjects PGDz is a potent contractile agent when inhaled (225,238,242), being approximately 3.5 and 10 times more potent than PGFz, and histamine, respectively. The major metabolite of PGD2, 9a,llP-PGF2, is a potent contractile agonist for human airways both in uitro and in uiuo (225). It is approximately 4 times more potent than PGD2 in contracting human bronchial smooth muscle in uitro, but is equipotent in eliciting bronchoconstriction in uiuo, suggesting that some of the contractile activity of PGD2 may be mediated through its metabolite. Prostaglandins may constrict human airways directly via TP receptors and indirectly through cholinergic pathways (238). The inhalation of 55 pg of PGEl and PGE2 in normal human subjects led to a mean increase in SGaw of 10 and 18%, respectively (233). In asthmatic subjects the same doses of these agonists led to a mean increase of 41 and 39%, respectively. These increases in airway caliber were comparable to those induced by 550 pg of inhaled isoprenaline. PGEz was also noted to speed the recovery from PGFz,-induced bronchoconstriction. However, both PGEl and PGE2 were highly irritating when inhaled, making them unsuitable for therapeutic use. PGIz has complex effects on the airways in humans. Precontracted human bronchus relaxes in response to PGI2 in uitro (229). However, in uiuo, inhaled PGIz had no consistent effect on airway caliber as measured by changes in SGaw in normal and asthmatic subjects (243). In contrast, concentration-related decreases in both FEVl and Vmax30 were observed in allergic asthmatic subjects. Inhaled PGIz protected the airways against the bronchoconstrictor effects of PGD2 and methacholine. The paradoxical effects of PGI2 on the airways of asthmatic


349

subjects might be explained by its effect on the vasculature within the airways. Increased mucosal blood flow might lead to engorgement of the mucosa with a significant reduction in caliber of the small airways. Increased mucosal blood flow might also lead to a more rapid clearance of inhaled bronchoconstrictor agonists from the airways, providing a degree of functional antagonism (243). 4 . Prostanoids and Airway Hyperresponsiveness

Inhalation of a subthreshold dose of PGF2, enhanced airway responsiveness to histamine by approximately fourfold in asthmatic subjects but had no effect on the airway response to methacholine (244). Inhalation of a noncontractile dose of PGDZ, but not saline, bradykinin, or histamine, enhanced airway responses to subsequent histamine and methacholine in asthmatic individuals by approximately twofold (245). Hardy et al. (246) confirmed the potentiating effect of PGDz on airway histamine responsiveness in three asthmatic subjects, but suggested that the results may represent a physiological rather than a pharmacological effect of PGDZ. They showed that histamine and PGDz were additive and not synergistic in their bronchoconstrictor effects on the airways of asthmatic subjects.

5 . Measurements of Prostanoids in Biological Fluids There have been various attempts to measure prostanoids in the lungs, blood, and urine of asthmatic subjects in both stable asthma and asthma provoked by a number of stimuli. Liu and colleagues found that levels of PGDz, 9a,l lP-PGF2, and PGF2, in bronchoalveolar lavage fluid were elevated 10- to 20-fold in subjects with atopic asthma, compared to levels in the controls (8).There was an inverse correlation between levels of these mediators and the responsiveness of the airways to methacholine. PGD2 is the major cyclooxygenase product released by activated mast cells (38,223), and has therefore been measured in the BAL fluid of asthmatic subjects after allergen challenge. PGDz levels rose from basal levels of

Advances in Immunology Volume 51

Advances in Parasitology Volume 51

Advances in Agronomy, Volume 51

Advances in Agronomy, Volume 51

Advances in Immunology Volume 45

Advances in Immunology Volume 34

Advances in Immunology Volume 1

Advances in Immunology Volume 10

Advances in Immunology Volume 30

Advances in Immunology Volume 82

Advances in Immunology Volume 31

Advances in Immunology Volume 52

Advances in Immunology Volume 48

Advances in Immunology Volume 56

Advances in Immunology Volume 12

Advances in Immunology Volume 21

Advances in Immunology Volume 41

Advances in Immunology Volume 71

Advances in Immunology Volume 84

Advances in Immunology Volume 43

Advances in Immunology Volume 73

Advances in Immunology, Volume 61

Advances in Immunology Volume 14

Advances in Immunology Volume 46

Advances in Immunology, Volume 106

Advances in Immunology Volume 19

Advances in Immunology Volume 99

Advances in Immunology Volume 81

Advances in Immunology Volume 4

Advances in Immunology Volume 80

Advances in Immunology Volume 51

Advances in Parasitology Volume 51

Advances in Agronomy, Volume 51

Advances in Agronomy, Volume 51

Advances in Immunology Volume 45

Advances in Immunology Volume 34

Advances in Immunology Volume 1

Advances in Immunology Volume 10

Advances in Immunology Volume 30

Advances in Immunology Volume 82

Advances in Immunology Volume 31

Advances in Immunology Volume 52

Advances in Immunology Volume 48

Advances in Immunology Volume 56

Advances in Immunology Volume 12

Advances in Immunology Volume 21

Advances in Immunology Volume 41

Advances in Immunology Volume 71

Advances in Immunology Volume 84

Advances in Immunology Volume 43

Advances in Immunology Volume 73

Advances in Immunology, Volume 61

Advances in Immunology Volume 14

Advances in Immunology Volume 46

Advances in Immunology, Volume 106

Advances in Immunology Volume 19

Advances in Immunology Volume 99

Advances in Immunology Volume 81

Advances in Immunology Volume 4

Advances in Immunology Volume 80

Recommend Documents