ADVANCES IN
Immunology VOLUME 1 1
CONTRIBUTORS TO THIS VOLUME BARUJBENACERRAF IVORN. BROWN ROBERTR. DOURMASHKIN
N. ...
10 downloads
1246 Views
21MB 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
ADVANCES IN
Immunology VOLUME 1 1
CONTRIBUTORS TO THIS VOLUME BARUJBENACERRAF IVORN. BROWN ROBERTR. DOURMASHKIN
N. MICHAELGREEN G ~ R A HOLM N JOHN
H. HUMPHREY
H. S. LAWRENCE HUGH0. MCDEVITT PETERPERLMANN
ADVANCES IN
Immunology EDITED BY
F. J. DIXON, JR.
H E N R Y G. K U N K E L
Division of Experirnenfol Pathology Scripps Clinic ond Research Foundofion l a lolla, California
The Rockefeller University New York, New York
V O L U M E
1 1
1969
ACADEMIC PRESS
New York and London
COPYRICHTO 1969,
BY
ACADEMIC PRESS,INC.
ALL RIGHTS RESERVED
NO PART OF THIS BOOK MAY BE REPRODUCED IN ANY FORM,
13Y PHOTOSTAT, MICROFILM, RETRlGVAL S l S T E M , OR ANY OTHER MEANS, WITHOUT WRITTEN PERhlISSlON FROM THE PUBLISHERS.
ACADEMIC PRESS, INC. 121 Fifth Avenue, New York, New York 10003
United Kingdom Edition published by ACADEMIC PRESS, INC. (LONDON) LTD. Berkeley Square House, London WlX6BA
LIBRARYOF CONGRESS CATALOG C A ~ NUMBER: D 61-17057
PRINTED IN THE UNITED STATES OF AMERICA
LIST OF CONTRIBUTORS Numbers in parentheses indicate the pages on which the authors' contributions begin.
BARUJ BENACERRAF, Division of Immunology, Department of Medicine, Stanford University School of Medicine, Stanford, California ( 31) IVORN. BROWN, Division of Parasitology, National Institute for Medical Research, London, England (267) ROBERTR. DOURMASHKIN, National Institute for Medical Research and Imperial Cancer Research Fund Laboratories, Mill Hill, London, England ( 7 5 )
N. MICHAELGREEN,National Institute for Medical Research, Mill Hill, London, England (1) GORANHOLM,Department of Immunology, The Wenner-Gren Institute, University of Stockholm, Stockholm, Sweden ( 117)
H. HUMPHREY, National Institute for Medical Research and Imperial Cancer Research Fund Laboratories, Mill Hill, London, England (75)
JOHN
H. S. LAWRENCE, Infectious Disease and Immunology Division, Department of Medicine, New York University School of Medicine, New York, New York (195)
HUGH0. MCDEVITT, * Laboratory of Immunology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesdu, Maryland ( 31) PETERPERLMANN, Department of Immunology, The Wenner-Gren Institute, University of Stockholm, Stockholm, Sweden ( 117) Present address: Division of Immunology, School of Medicine, Stanford University Medical Center, Stanford, California.
V
This Page Intentionally Left Blank
PREFACE The diversity of fields in which exponents of immunology may be found continues to surprise even its most enthusiastic adherents. It is not just the exquisite sensitivity and precision of immunological techniques that have served as the impetus for this diffusion. More significant appears to be the recognition of the broad biological impact of the relevant events that precede and follow an immune stimulus. Volume 11 exhibits such diversity unusually well, with physicist to parasitologist represented. The volume also includes three chapters by English scientists illustrating once again the continuing strong contribution of this group to immunology. The first chapter deals with the exciting contributions of electron microscopy to the analysis of antibody moIecuIes. Dr. Green initiated the use of small bivalent haptenes to link antibody inolecules which has proven so successful in their characterization. The concept of a threearmed molecule with n flexible hinge region is well documented for yG globulin through the surprisingly clear electron micrographs obtained by this procedure. The ameboid appearance of yM molecules is most esthetically satisfying. Dr. McDevitt and Dr. Renacerraf review the recent important findings concerning “immune response genes” in the second chapter. The fact that genetic factors are involved in the response to antigenic stimulus has Iong been known. However, the credit for establishing this on a firm scientific basis in terms of modern genetics must be given to the authors of this chapter. The use of synthetic polypeptide antigens played a major role in elucidating the multiple genes which are described. The intriguing question of at what level in the immune response these genes act remains to be determined. It appears clear that they do not represent the structural genes for the antibody molecule. The third chapter has been contributed by Dr. Humphrey and Dr. Dourniashkin and deals with that most important of all complement questions, the terminal phase of cellular injury. Their most elegant electron microscope pictures of the holes in the cell membrane produced by complement have intrigued all immunologists. Considerable progress in the understanding of the underlying mechanism involved has been gained although the final answer is not yet in. Is an enzyme attacking lipid Vi i
viii
PREFACE
moieties in the membrane primarily involved? Many unpublished studies of the authors relating to these questions are included in this fascinating review. The fourth chapter by Dr. Perlmann and Dr. Holm deals with the complex problem of different types of cytotoxic effects of lymphoid cells. These outstanding workers in the field have managed to present a cohesive picture of the various effects on target cells. The role of “nonspecific” factors is particularly wcll clarified. The interrelationships among contact lysis, release of pharmacologically active substances, and the terminal components of the complement system are given special consideration. There is little question that significant developments conccrning in vim events will stem from these in vitro findings. In another chapter Dr. H. S. Lawrence reviews the extensive and confusing literature on various factors involved in cellular immunity. Transfer factor, which h e first described, is placed in perspective with the various substances under active current investigation in the guinea pig. This is a very enlightening review of an area of immunology from which much will be heard in the future. The methodology has been partially worked out for obtaining transfer factor, as well as some of the other materials, in sufficient purity for chemical analysis, and further results in this area are awaited with great interest. The assay systems remain difficult but the shift to in vitro systems has been a major achievement. The last chapter by Dr. Ivor Brown deals with immunity in malaria, an old subject that has suddenly become of considerable current interest. New methods for the study of the relevant antibodies and a new appreciation for a role for cell-mediated immunity are responsible for this development. The very diverse contributions to this subject present unusual difficulties for a reviewer. However, a clear and interesting summary of the subject has emerged which should prove of considerable value as a reference for all immunologists. The complete cooperation of the publishers in all aspcbcts of the work involved in the production of Volume 11 is gratefully acknowledged.
H. G. KUNKEL F. J. DIXON
CONTENTS
. PREFACE . . . . . . CONTENTS OF PREVIOUS VOLUMES. LIST OF CONTFXBUrORS
.
.
.
.
.
.
. .
. .
.
. . .
. . .
. . . . . . .
,
.
. . . . . .
.
.
.
V
vii
xi
Electron Microscopy of the Immunoglobulins
N. MICHAELGREEN
I. Introduction . . . . . . . 11. Electron Microscopy at the Molecular Level 111. Electron Microscopy of IgG . . . . IV. Electron Microscopy of IgM . . . . V. Comments and Conclusions . . . . References . . . . . . .
. . . . .
.
. . . . .
. . . .
.
. . . .
. .
.
.
. .
.
1 2 G 17 26 28
Genetic Control of Specific Immune Responses
HUGH0. MCDEVITTAND BARUJBENACEHRAF I. Introduction . . . . . . , . . . . . 11. Constitutional Differences in Individual Responses to Complex Multi. . . . . . . . . , determinant Antigens . 111. Analysis of the Mechanism of Gene Action . . . . . . IV. Genetic Differences in Imiriune Response to Defincd Protein Antigens . V. Genetic Differences in Immune Response.; to Synthetic Polypeptide Antigens . . . . . . . . . . . . . . VI. Conclusions . . . . . . . . . . . . . References . . . . . . . . . . . .
31 33 37 38 39 69 71
The lesions in Cell Membranes Caused by Complement JOHN
H. HUMPHREY AND ROBERT R. DOURMASHKIN
I. Introduction . . . . . . . . . . . . . 11. Description of Holes Produced by the Action of C‘ . . . . . 111. Holes Produced by C‘ in Substrates Other than Erythrocyte Membranes . IV. The Relationship of Holes t o Sites of Damage on the Cell Surface . V. Occurrence of Multiple Holes (Clusters) at Single Sites of Damage . VI. The Number of Antibody Molecules Required to Produce a Lesion VII. The Stage of C’ Action at Which Holes Are Formed . . . . VIII. The Nature of C’ Holes . . . . . . . . . . IX. Arti6ciaI Membrane ModeIs . . . . . . . . . . X. Biological Significance of the Terminal C’ Lesion . . . . . References . . . . . . . . . . . . . ix
75 77 85 88 92 95 98 101 108 110 113
X
CONTENTS
Cytotoxic Effects of Lymphoid Cells in Vifro
PETERPERLMANN AND GORANHoLhr
.
I Introduction . . . . . . . . I1. Methods . . . . . . . . . I11. Different in Vitro Models . . . . . 1V Some in Vivo Implications of the in Vitro Models V Summary . . . . . . . . References . . . . . . . .
. .
. .
. .
.
.
. .
. . .
117 119 127 172 183 185
I. Introduction . . . . . . . . . . . . . I1. Definitions and General Principles . . . . . . . . I11. Transfer of Delayed Hypersensitivity with Viable Blood Leukocytes . 1V. Transfer Factor-Characterization and Mechanism of Action . . . V Nature and Properties of Dialyzable Transfer Factor . . . . VI . Transfer Factor and in Vitro Correlates of Cellular Immunity . . VII . Mechanism of Action of Transfer Factor in Viuo and in Vitro . . VIII . Transfer Factor and Mechanisms of Cellular Immune Deficiency Diseases IX . Transfer Factor and Reconstitution of Cellular Immune Deficiency . X . Transfer Factor. Immunological Surveillance. and Tumor Immunity . XI . Conclusion . . . . . . . . . . . . . References . . . . . . . . . . . . .
196 199 202 217 229 234 245 248 252 258 259 261
. .
. . .
.
. . .
.
. . .
.
.
.
.
Transfer Factor
.
13 S. LAWRENCE
.
Immunological Aspects of Malaria Infection
IVOR N . BROWN
I . Introduction . . . . . . . I1. The Life Cycle of Malaria Parasites . . I11. Innate and Nonspecific Immunity to Malaria IV . Immunity Acquired through Infection . V . Relapses and Antigenic Variation . . VI . Cellular Factors in Malaria Infection . . VII . Antigens of Malaria Parasites . . . VIII . Humoral Factors in Malarial Immunity . IX. Active Immunization to Malaria . . . . X . Experimental Modification of Immunity XI . Immunopathology . . . . . . XI1. Discussion . . . . . . . XI11. Summary . . . . . . . References . . . . . . . AUTHORINDEX.
SUBJEW INDEX .
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
. .
. .
. .
. .
. .
.
.
.
.
.
.
. .
. .
. .
. .
. .
. .
. . .
. . .
. . .
. . .
. . .
. . .
. .
. .
. .
. .
. .
. .
.
.
.
.
.
.
268 269 275 278 284 288 296 303 323 329 331 338 339 340
.
.
.
.
.
.
351 368
Contents of Previous Volumes Volume 1 Transplantation Immunity and Tolerance
M. HA~EK, A. LENCEROV~, AND T. HRABA Immunological Tolerance of Nonliving Antigens
RICHARDT. SMITH Functions of the Complement System
ABRAHAM G. OSLER In Vitro Studies o f the Antibody Response
ABRAMB. STAVITSKY Duration of Immunity in Virus Diseases
J. H. HALE Fate and Biological Action of Antigen-Antibody Complexes
WILLIAM0. WEIGLE Delayed Hypersensitivity to Simple Protein Antigens
P. G. H. GELLAND B. BENACERRAF The Antigenic Structure of Tumors
P. A. GORER AUTHORINDEX-SUBJ
E
INDEX ~
Volume 2 Immunologic Specificity and Molecular Structure
FREDKARUSH Heterogeneity of y-Globulins JOHN
L. FAHEY
The Immunological Significance o f the Thymus
J. F. A. P. MILLER,A. H. E. MARSHALL, AND R. G. WHITE Cellular Genetics of Immune Responses
G. J. V. NOSSAL Antibody Production by Transferred Cells
CHARLES G. COCHRANE AND FRANK J. DIXON Phagocytosis
DERRICK ROWLEY xi
CONTENTS OF PREVIOUS VOLUMES
Xii
Antigen-Antibody Reactions in Helminth Infections
E. J. L. SOULSBY Embryological Development of Antigens
REEDA. FLICKINGER AUTHOR INDEX-SUB JECT INDEX Volume 3 In Vifro Studies o f the Mechanism of Anaphylaxis
K. FRANK AUSTENAND JOHN H. HUMPHREY The Role of Humoral Antibody in the Homograft Reaction
CHANDLER A. STETSON Immune Adherence
D. S. NELSON Reaginic Antibodies
D. R. STANWORTH Nature of Retained Antigen and Its Role in Immune Mechanisms
DAN€1. CAMPBELL AND JUSTINE S. GARVEY Blood Groups in Animals Other Than Man
W. H. STONEAND M. R. IRWIN Heterophile Antigens and Their Significance in the Host-Parasite Relationship
C. R.
JENKIN
AUTHOR INDEX-SUB JECX INDEX Volume 4 Ontogeny and Phylogeny o f Adoptive Immunity
ROBERTA. GOODAND BENW. PAPERMASTER Cellular Reactions in Infection
EMANUEL SUTERAND HANSRUEDY RAMSEIER Ultrastructure of Immunologic Processes JOSEPH
D. FELDMAN
Cell W a l l Antigens of Gram-Positive Bacteria
MACLYNMCCARTY AND STEPHENI. MORSE Structure and Biological Activity of Immunoglobulins
SYDNEY COHENAND RODNEY R. PORTER
COXTENTS OF PREVIOUS VOLUMES
Autoantibodies and Disease
H. G. KUNKELAND E. M. TAN Effect of Bacteria and Bacterial Products on Antibody Response
J. MUNOZ AUTHORINDEX-SUBJECT IXDEX Volume 5 Natural Antibodies and the Immune Response
STEPHENV. BOYDEN Immunological Studies with Synthetic Polypeptides
MICHAELSELA Experimental Allergic Encephalomyelitis and Autoimmune Disease
PHILIPY. PATERSON The Immunology of Insulin
C. G . POPE Tissue-Specific Antigens
D.
c. D U h I O N D E
AUTHORINDEX-SUBJECT INDEX Volume 6 Experimental Glomerulonephritis: Immunological Events and Pathogenetic Mechanisms
EMILR. UNANUE AND FRANK J. DIXOX Chemical Suppression of Adaptive Immunity
ANN E. GABRIELSON AND ROBERTA . GOOD Nucleic Acids as Antigens OTTO
J. PLESCIA
AND If’ERiXER
BRAVN
In Vifro Studies of Immunological Responses of Lymphoid Cells
RICHARDW. DUTTON Developmental Aspects of Immunity J A R O S L ~ VSTERZL AND
ARTHURM. SILVERSTEIN
Anti-antibodies
PHILIPG. H. GELLAXD ANDREWS. KELUS Conglutinin and lmmunoconglutinins P . J. LACHMANN AIJTHOR IKDEX-SITB JECT INDEX
...
Xlll
xiv
CONTENTS OF PREVIOUS VOLUMES
Volume 7 Structure and Biological Properties of Immunoglobulins
SYDNEY COHENAND CESAR MILSTEIN Genetics of Immunoglobulins i n the Mouse
MICHAELPOITERAND ROSELIEBERMAN Mimetic Relationships between Group A Streptococci and Mammalian Tissues
JOHNB. ZABRISKIE lymphocytes and Transplantation Immunity
DARCY B. WILSON AND R. E. BILLINGHAM Human Tissue Transplantation
JOHNP. MERRILL AUTHORINDEX-SUBJECT INDEX Volume 8 Chemistry and Reaction Mechanisms of Complement
HANSJ. M~~LLER-EBERHARD Regulatory Effect of Antibody on the Immune Response JONATHAN
W. UHRAND GORANMOLLER
The Mechanism of Immunological Paralysis
D. W. DRESSER AND N. A. MITCHISON In Vifro Studies of Human Reaginic Allergy
ABRAHAM G. OSLER,LAWRENCE M. LICHTENSTEIN, AND DAVID A. LEVY AUTHOR INDEX-SUBJECT INDEX Volume 9 Secretory Immunoglobulins
THOMAS B. TOMASI, JR., AND JOHNBIENENSTOCK Immunologic Tissue Injury Mediated by Neutrophilic leukocytes
CHARLESG. COCHRANE The Structure and Function of Monocytes and Macrophages
ZANVILA. COHEN The Immunology and Pathology of NZB Mice
J. B. HOWIEAND B. J. HELYER
AUTHOR INDEX-SUBJECTINDEX
CONTENTS OF PREVIOUS VOLUMES
Volume 10 Cell Selection by Antigen in the Immune Response
GREGORY W. SISKIND AND B A R U J BENACERRAF Phylogeny of Immunoglobulins
M. GREY HOWARD Slow Reacting Substance of Anaphylaxis
ROBERTP. ORANGE A N D K. FRANK AUSTEN Some Relationships among Hemostasis, Fibrinolytic Phenomena, Immunity, and the Inflammatory Response
OSCARD. RATNOFF Antigens of Virus-Induced Tumors
KARLHABEL Genetic and Antigenetic Aspects of Human Histocompatibility Systems
D. BERNARD AMOS
AUTHOR INDEX-SUBJECT INDEX
xv
This Page Intentionally Left Blank
Electron Microscopy of the Immunoglobulins
N. MICHAEL
GREEN
Nafionol Insfifufe for Medical Research, Mill Hill, London, England
I. Intioduction . . . . . . . 11. Electron hlicroscopy at the Molecular Level 111. Electron Microscopy of IgG . . . . A. Results Obtained by Shadowing . . B. Results Obtained by Negative Staining C. The Question of Conformational Change IV. Electron Microscopy of IgM . . . . V. Comments and Corclusions . . . . References . . . . . . .
I.
.
. .
. .
.
.
.
. .
. .
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
1 2 6 6 7 14 17 26 28
Introduction
Chemical techniques have provided extensive information about the structure of the constituent pcptide chains of immunogIobulin molecules and the way in which they are linked to each other (reviewed by Cohen and Porter, 1964; Cohen and Milstein, 1967). They have also shown how the chains interact to give the compact Fab and Fc fragments which are produced by splitting of a few peptide and disulfide bonds. A different approach is required to determine the overall layout of the molecule and the spatial relationships of the fragments to each other. Before electron microscopy and X-ray crystallography had reached their present level of development the only approach to this problem was through hydrodynamics. Sedimentation, diffusion, and viscosity measurements consistently showed that the IgG molecule was either highly hydrated or asymmetric (Neurath, 1939; Oncley et d.,1947). A plausible hydration of 0.2 ml./gm. was usually assumed, from which the axial ratio of about 6: 1 was calculated. IgM has an even higher frictional ratio ( f / f o = 1.9) (Miller and Metzger, 196:3a), and in these terms would have an axial ratio of about 1O:l. The asymmetry of IgG was supported by the early electron micrographs and a rod-shaped or ellipsoidal model was accepted for some time. However, in 1965, Noelken et aL pointed out that there were other possible interpretations of the high fractional ratio and intrinsic viscosity, which were more consistent with the chemical evidence. The Fab and Fc fragments showed normal fractional ratios and viscosities and so were not unusually hydrated or asymmetric, It was suggested that these three fragments were joined in Y formation by a relatively flexible 1
2
N. 'MICHAEL GREEN
region of peptide chain, of which the extensive hydration could explain both the high frictional ratio and the susceptibility to enzymatic attack. The contribution of electron microscopy to the solution of the problem provides the main subject for this review. Brief reviews of the subject have appeared elsewhere (Horne, 1965, 1968; Stanworth and Pardoe, 1967). The use of ferritin-labeled antibody as a specific marker for cellular constituents has been treated elsewhere (Andres et al., 1967) and is also the subject of a forthcoming article (Rifkind, 1969), so it will not be considered here. II.
Electron Microscopy at the Molecular Level
The recent application of electron microscopy to the study of structure of macromolecules followed the exploitation of negative contrast methods for the study of viruses (Brenner and Horne, 1959). The limiting factor both now and in the earlier work was not the resolving power of the microscope (about 5 A.) but the difficulty in obtaining sufficient contrast with specimens of molecular dimensions (Valentine, 1961). Unless the thickness of a protein molecule is greater than 70 A. (mol. wt. 150,000), it will not, if untreated, scatter a sufficient proportion of the incident electrons to render it visible against the usual background of carbon film. The first advance in technique was the use of metal shadowing (Williams and Wyckoff, 1945) which proved very successful with virus particles and was particularly useful for revealing detail of the surface and the height of the particle. The unavoidable granularity (20 A.) of the evaporated metal film limited the effective magnification to about 50,000, which was not quite sufficient to reveal subunit structure in proteins. The method was, however, used to determine the lengths of some highly asymmetric molecules (Hall and Doty, 1958) and provided one of the earliest pictures of unattached antibody molecules (Hall et al., 1959) . Two other general methods have been used for enhancing contrast. Positive staining, although satisfactory for sectioned material, is of little use at the molecular level because it is difficult to combine sufficient stain with the specimen (Valentine, 1961). Negative staining (more accurately, negative contrast), on the other hand, has proved to be both simple and effective for the study of viruses and of a variety of protein molecules. In its simplest form a droplet of the dilute (0.01%)protein solution, mixed with 2% sodium phosphotungstate or other suitable salt, is applied to a carbon (or nitroccllulose-supported carbon) film on the grid. After removing thc exccss fluid the film is allowed to dly. The molecules appear as low-density footprints in the thin layer of surrounding phosphotung-
ELECTRON hfEROSCOPY OF THE IMMUNOGLOBULINS
3
state. Success depends on obtaining a faithful replica of thc molecule in a uniform a~norphouslayer of a stain of high weight density. The properties and uses of various heavy metal salts havc been discussed by Valentine and Horn(. ( 1962) and by IIornc ( 1968). Much of the work with antibodies has employed phosphotungstatc, but recently silicotungstate, introduced by Wilcox et al. (1963) for use with vii-uses, has been found to give a slightly less granular background. It is more stable at neutral pH than phosphotungstate (Baker et al., 1955) and the surface activity of its solutions causes it to spread more evenly at low protein concentration. A variation of the negative staining technique, described by Valentine et al. (1968) for use with enzymes, is worth repeating here with some additional details, in view of the cxcellent results it has given with antibodies. The molecules were picked up on carbon film dcposited on freshly cleaved mica, by dipping the mica, film upward, into the protein solution (30-60 pg. antibodylml.). The solution penetrated between the hydrophilic mica and the floating hydrophobic film and the molecules were adsorbed on the carbon in a few seconds. The film was transferred on the mica to a dish of 2%sodium silicotungstate, where it was left floating for a minute or two. A 400-mesh copper grid, coated with a thin layer of adhesive, was placed on the film, followed by a square of adsorbent paper (e.g., newsprint). The paper was removed together with the adhering grid and film and placed on filter paper to drain. The grid dries in a short time and is ready for examination. Uranyl salts have been used by Hoglund (1967a,b) and give higher contrast and greater pcnetration than the tungstates, but this advantage is offset by the more granular background. Uranyl fonnate, in particular, penetrates further into molecules and between subunits (Leberinan, 1965; Finch and Holmes, 1968), and this has been turned to advantage by Svehag et al. (1969) who were sometimes able to resolve H chains from each other and from L chains in fragments of IgM (see below). Objections that negative staining and drying may disrupt labile protein molecules are difficult to refute and cannot be disregarded. In some cases it has been shown that the biological activities of enzymes (Valentine et al., 1968) and antibodies (Chesebro et al., 1968) are not affected b y drying down in the presence of the stain. A further check for artifacts can be made by comparing the volume estimated from the linear dimensions with the molecular weight of the macromolecule ( Rowe, 1966). The curves in Fig. 1 facilitate such estimations for spherical and cylindrical subunits. Unfortunately this can only be a very rough comparison since there are several possible sources of error in the measurcment of overall linear dimensions. The blurred out-
4
N. MICHAEL GREEN
0
20
40 Diameter
(A)
60
80
FIG. 1. Relation between linear dimensions and molecular weight for protein subunits of various shapes. ( A ) Cylindrical subunit of diameter d A. and height 10 A.; ( B ) cylindrical subunit of diameter d A. and height d A.; ( C ) spherical subunit of diameter d A. The molecular weights were calculated from the volume using (the partial specific the expression M = ( N / v ) * V .If a value of 0.73 is used for volume), then M = 0.82V, where V is the volume of the molecule in A.’. Curve A can be used to calculate the molecular weights of cylindrical subunits of any length.
v
lines of the molecules introduce a subjective element which can be only partially overcome by averaging. The apparent size of a spherical or cylindrical object is likely to decrease as the thickness of the layer of stain increases and this is a factor which is difficult to control. This is vividly illustrated by the appearance of the projecting “fiber antigens” of the adenovirus (Valentine and Pereira, 1965; Valentine, 1969), which appear as uniform rods (20 A. diameter) when separated from the virus, but of which the proximal ends vanish in the thick layer of stain surrounding the virus when they are attached. Similarly, IgG bound to ferritin
ELECTRON MICROSCOPY OF THE IMMUNOGLOBULINS
5
appears much narrower in the thicker stain surrounding the ferritin than it does when separated from it ( Fig. 4; and Feinstein and Rowe, 1965). A further source of error derives from the unknown third dimension of a molecule seen only in profilc in negative stain. Sometimes the strength of the contrast can provide qualitative information (see discussion of Fig. 7 ) but noiinally it is necessary to use shadowing to estimatc the third dimension of a molecule and this is not very accurate for small particles. However, in some favorable cases the presence of different, clearly assignable orientations of the molecule may permit all three dimensions to be determined accurately ( t 5 A.) from a negatively stained preparation (Valentine et al., 1968). Molecular weights of many proteins agree well with those calculated from the linear dimensions (Horne, 1965) which gives confidence in the general validity of the method. Several recent papers support this conclusion (Valentine et al., 1966, 1968; Pcnhoet et al., 1967) and show that it is often more accurate to measure center-to-center distances between subunits than to try to estimate the absolute size of a molecule. Unfortunately this approach has not been applicable to the immunoglobulins because of the variable relationships between positions of subunits. Although many protcins appear to survive the process of negative staining without damage, it must be admitted that some proteins do dissociate into subunits in the negative stain and others, such as serum albumin, may be completely disorganized so that they give no appreciable contrast. These undesirable effects may sometimes be eliminated by pretreatment of the protein with glutaraldehyde to produce stabilizing cross-links (e.g., Valentine et al., 1966, 1968). Broadly speaking, negative-contrast electron micrographs provide information about the number, size, and arrangement of subunits in a protein molecule. Attempts have been made to see structure within subunits, but a careful study (Mellema et al., 1968) suggests that such is largely artifact. Under favorable circumstances (thinly spread stain and homogeneous specimens) the smallest detectable protcin would be about 20 to 25 A. diameter (mol. wt. 5000-10,000), but even with a diameter of 35 A. (20,000-30,000 mol. wt.) the contrast is poor. When contrast is good, it is possible to resolve objects as little as 10 A. apart, since the ions of the dodecatungstate stains are about 10 A. in diameter (Baker et al., 1955) and can penetrate between them. Sometimes subunits or molecules which are actually in contact may appear to be separate because of partial penetration of stain between them. Finally, it is worth remembering that the negatively stained eIectron micrograph is a c01lcction of two-dimensional profiles and that apparent changes in shape
6
N. MICHAEL GREEN
under different experimental conditions may reflect changes in preferred orientation of the molecules on the carbon film superimposed upon any change in molecular structure. Ill.
Electron Microscopy of IgG
Early attempts to observe combination between antigen and antibody in the electron microscope (Anderson and Stanley, 1941) were made even before the introduction of shadowing techniques. They succeeded in showing an increase in diameter of particles of tobacco mosaic virus from 150 to 600 A. after combination with antibody. Little further was done until Easty and Mercer (1958) made an estimate of the length of the antibody link from the appearance of thin films or sections of ferritinantiferritin precipitates. They observed a halo of low density material 300-400 A. thick around the iron core of the ferritin, which was consistent with the length of about 300 A. for the antibody molecule suggested by hydrodynamic measurements ( Neurath, 1939; Oncley et al., 1947).
A. RESULTSOBTAINED BY SHADOWING The first pictures in which individual antibody molecules were resolved were obtained with shadowed preparations of rabbit antibody (Hall et al., 1959) unattached to antigen. These showed molecules the width of which was greater than their height and the dimensions of which corrected for the thickness of the shadowing metal, were 250 x 40 x 40 A. The length was calculated as a weight average and was, perhaps, unduly influenced by a prolonged heavy tail to the distribution, The number-average length was only 170 A. which was more consistent with the molecular weight (Fig. 1). It is interesting to note that Hall et al. attempted to polymerize their antibody using a divalent hapten, in order to locate the binding sites, a technique which has met recently with considerable success (Valentine and Green, 1967). Unfortunately the affinity of the antibody for the hapten was not high enough to give significant polymerization. From the few dimers that were present, Hall et al. concluded that the molecules probably aggregated end to end. Shortly before this work was published the advantages of neg at'ive contrast methods (for work on viruses and protein molecules) were emphasized by Brenner and Horne ( 1959). The technique was much simpler than that of shadowing and enabled resolution of much finer details, so that it rapidly displaced the earlier technique at least for the cxamination of small molecules. The improvement in detail is clearly illustrated in the article by Valentine (1961) on methods of contrast enhancement.
ELECTRON MICROSCOPY O F THE IMMUNOGLOBULINS
7
Recent investigations of antibocly molecules have employed negative staining methods almost exclusively, sometimes with shadowing used in parallel, to provide supplemcntary information ( Feinstcin and Rowe, 1965)- The clearest results obtained by shadowing (Hoglund, 1967a, 1968) are consistent with a structure for IgG containing three subunits arranged in a variety of configurations. Shadowing has also been used to reveal the surface structurc: of ferritin-antiferritin precipitates in antibody excess ( Robinson, 1966), since precipitates are in general too bulky to be studied by negative staining methods, but it is difficult to interpret the rcsults in molecular terms. B. RESULTSOBrAINED
BY
NEGATIVESTAINING
Most electron micrographs of free antibody molecules ( Valentine, 1959; Feinstein and Rowe, 1965; Valentine and Green, 1967) have shown a disappointing lack of cliaracteristic structure. Thcy appear usually as irregular globular particles with a maximum dimension of about 120 A., nowhere approaching the length of 250 A. deduced from hydrodynamic evidence. In contrast, other protein molecules of similar inolccular weight have shown charactcristic shape and sometimes a subunit structure ( Horne and Greville, 1963; Penlioct ct nl., 1967). The application of the method to antigcn-bound antibody was more rewarding. Hunimeler et al. (1962) observed threadlike molecules as a fuzz on the surface of poliovirus particles. Clearly resolved single antibody links were first obscrvcd by Almeida et al. (1963) in their study of antibodies to the eicosahcdral wart and polyoma viruses (550 and 450 A. diameters, respcctively ) . In antibody excess they confirmed the presence of a layer of threadlike molecules on the surface of the virus dirccted outward in an approximately radial fashion. In antigen excess, single antibody molecules could be seen as thin straight rods (width 34 A . ) linking the virus particles together. The mean length was 150 A. and the maximum length 270 A. It was argued that the full length of the antibody molecule would rarely be visible because of the curved surface of thc virus and, therefore, that the longest links gave the best approximation to the true length of the molecule. Alineida et nl. concluded that the most probable dimensions were 35-40 x 250-270 A. Although this implied a high molecular weight (240,000 t 40,000), it was consistcnt with the axial ratio obtained from sedimentation results. They also commented on the occasionally beadcd appearance of the molecules suggesting several subunits strung togethcr. But this could have been a consequence of the thick layer of stain which makes it difficult to obtain good contrast with the thin antibody molecule the diameter of which is less than one-tenth
8
N. MICHAEL GREEN
7
of that of the virion. It is likely that many of the attached antibody molecules remained undetected, which might account for the sparse distribution of clearly visible antibody molecules in the preparations. Elek et al. (1964) examined Salmonella flagella agglutinated by rabbit IgG antibody. The small diameter (120 A.) and linear geometry of the flagella provided favorable conditions for revealing the antibody, but few single molecules were observed. In antibody excess the flagella, like the virions of Hummeler et al. (1962), were covered with a continuous layer of radially projecting threadlike molecules about 140 to 160 A. thick. Smaller proportions of antibody led to agglutination, and from the interflagellar distance the length of the molecules linking the flagella was calculated ( 180 A , ) . More recently, Hoglund ( 1967a, 1968) used the small satellite tobacco necrosis virus (diameter 180 A.) as antigen and obtained pictures of rodlike links between virions rather shorter than those of Almeida et al. ( 100-150 A . ) , He also obtained similar results with antibodies to T2 bacteriophage ( Hoglund, 196%). These results all tended to confirm the conventional interpretation of the high frictional coefficient of IgG in terms of a molecule of high axial ratio. In addition, they provided evidence that the antigen binding sites were located at opposite ends of the long axis of such a molecule. This view was embodied in the model proposed by Edelman and Gally (1964) which also incorporated the chemical evidence on the arrangement of peptide chains and the size of the binding sites. There was already, however, some preliminary evidence of a more complex molecule. Lafferty and Oertelis ( 1963) had obtained pictures of influenza virus combined with excess of antibody in which single antibody molecules could be seen to be bent into a loop linking two neighboring surface antigens on the same virion. Similar effects had also been seen by Almeida et al. (1963) but were partially discounted because of the possibility of superimposition of two different antibody molecules and the consequcnt difficulty in arriving at a clear interpretation. At this stage the evidence from electron microscopy was very confusing. The antibody molecule, like the cloud of Polonius, was sometimes globular, sometimes elongated, sometimes looped. This polymorphism was emphasized by Feinstein and Rowe (1965) using antibodies to ferritin, the smaller diameter (105 A . ) of which allowed a clearer picture of the attached IgG to be obtained. They observed a maximum dimension for uncombined IgG (rabbit and human) of 105 A. either in negative stain or after metal shadowing. Combination with the ferritin led to a marked change in appearance of the antibody molecules. Thin rods (15 A. across), frequently bent in the middle, were
ELECTRON iWC13OSCOPY O F THE IbfMUNOGLOBULINS
9
observed joining the ferritin molecules together. The maximum separation of the ferritin molecules was about 140 A., but Feinsteill and Rowe assumed that part of the antibody was obscured by the ferritin and concluded that its fully extcnded length was about 200 A. A central swelling, often seen on the thin strands was rcmovcd by pcpsin and somewhat straighter links wcre them observcd. Reduction of the pepsin product gave pictures of Fab fragments projecting 60-70 A. from the suiface of the ferritin. Feinstein and Rowe stressed the variable angle between the arms of the molecule and proposed a globular model for IgG, which could open up about a hinge in the Fc region when combined with antigen, to give an elongated structure of about double the length of the uncombined molecule. They did not attempt to reconcile the globular shape of the uncombined antibody with the hydrodynamic results. The Fab fragments were also observed by Almeida et al. (1965), when bound to polyoma virus. They appcared as short rods (25 x 70 A. ) projecting radially from the viral surface, and, thus, resembling closely those observed on ferritin by Feinstein and Rowe (1965). Valentine and Green (1967) avoided the problem of visualizing and measuring molecules of antibody bound to a large antigen by employing a small bivalent hapten. Bisdinitrophenyl( DNP)-octamethylenediamine reacted with an equivalent amount of high affinity rabbit anti-DNP antibody to give soluble polymers the shape of which revealed the arrangement of the three fragments of the IgG molecule (Fiq. 2). Polygonal rings containing any number from three to ten or more distinct IgG molecules could be seen. Dimers and linear polymers were also present. [A number of different fields have been published elsewhere (Valentine, 1967, 1969; Cohen and Milstein, 1967; Kabat, 1968), and together these provide an adequate impression of the variety of structures that were observed.] The short projections present at cach corner of the polygons could be removed by digestion with pepsin, leaving the rings intact. Each projection was, therefore, an Fc fragment and each corner of a polygon was the center of a Y-shaped IgG molecule. The edges of the polygons were, therefore, dimers of two rod-shaped Fab fragments linked end to end by a molecule of bivalent hapten. This was confirmed by reduction of the products of peptic digestion, which broke the disulfide bonds and liberated the hapten-linked dimers (35 x 130 A ) of Fab. These results confirmed some of the conclusions of Feinstein and Rowe (1965), whereas the greater resolution provided evidence for a more precise molecular model (Fig. 3 ) , similar in many respects to that put forward by Noelken et al. (1965). The angle between the two Fab
FIG.2. Polymers produced by reaction of rabbit anti-DNP IgG with an equiv-
ELECTRON hIICRObCOPY OF THE IMMUNOGLOBULINS
11
-25
FIG.3. Scale diagram of a molecule of rabbit IgG based on measurements of the dimensions of cyclic trimers and on chemical evidence ( Cohen and Porter, 1964). The lengths of the Fab and Fe fragments are 10% greater than those published previously (Valentine and Green, 1967) following a more extended set of nieasurements. The mean distance between the extremes of F c fragments in twenty cyclic trimers was 245 A. (range 215-270 A ) . The variable orientation of the Fc fragment probably accounts for the rather wide range of the measurements. The molecular weight of each fragment (calculated from Fig. 1A) would be 52,000, assuming a cylindrical cross section and making no allowance for the rounded corners illustrated. The relative positions of the L and H chains in the Fab fragments and the orientation of the cleavage plane between them are unknown and, therefore, arbitrary. This has been emphasized by reversing the positions of the L and H chains relative to that shown in a previous diagram (Valentine and Green, 1967). The arrow between the two halves of the F c fragment indicates the position of the twofold symmetry axis observed both in crystals of Fc (Goldstein et al., 1968) and of human IgG myeloma protein (Terry et al., 1968). The location of the binding site in a cleft between L and H chains is consistent with the available evidence on the roles of the two chains (Cohen and Milstein, 1967) but cannot be regarded as firmly established. The smooth contour of the junction between hapten-linked Fab’s is consistent with the central location of the binding site.
d e n t amount of a divalent haptcn ( his-DNP-octamethylenediamiue ). The antibody molecules are centered at the corners of the polygonal shapes. The Fc fragments project from the corners and the Fab fragments form the edges of the polygons. Single molecules ( M ) and tlimers ( D ) can also be seen. Arrows indicate edgewise profiles of dimers. The electron micrographs sliown in Figs. 2, 4, and 7 were all made by the technique of Valentiile et a/. ( 1968) described in Section 11, using sodirtin silicotriiigstate as the negative stain. hlaguificatiorl: X400,OUO.
12
N. MICHAEL GREEN
arms varied between about 10' in dimers to 180" in some of the large rings and open chain polymers; there was no suggestion of a preferred angle. The Fc arm was not always symmetrically disposed between the Fab's, and its apparently variable size suggested that it was not confined to the plane defined by the two Fab arms and the carbon film. This was confirmed by the characteristic appearance (1) of some of the dimers seen edge on (Fig. 2; Valentine, 1967, Fig. 3a). This flexibility allows a single IgG molecule to link pairs of suitably orientated antigenic sites at any distance between about 40 and 140 A. and provides a simple explanation for the facility with which IgG can form cross-linked aggregates with a variety of antigens. Bridging distances longer than 140 A. can only be explained by postulating a considerable stretching or unravelling of the molecule for which there is no evidence in the electron micrographs of the cyclic polymers or of the anti-DNP antibody bound to DNP-ferritin (Fig. 4; Valentine, 1967). This bridging distance is consistent with recent results of Hoglund (1967a,b, 1968) but is significantly shorter than inany of the earlier estimates discussed above. The wide range of angles observed between the three arms suggests a flexible hinge region with low-energy barriers between the various alternative conformations. It could also be argued that the energy barriers are high and that the IgG exists as a population of stable isomers, but this is not supported by the uniform amino acid sequence of the hinge region of rabbit IgG (Cebra et al., 1968), since a unique sequence would be unlikely to foId into a large number of different stable conformations. The results of fluorescence polarization ( Weltman and Edelman, 1967; Wahl and Weber, 1967) do not provide much assistance on this question of flexibility. They suggest that the IgG molecule rotates as a single unit and that any rotations of the fragments relative to each other are too slow to affect the polarization. However, this requires only that the rotational relaxation time of the fragments relative to each other he increased about 20-fold over that of the free fragment (from about 0.05 psec. to a few microscconds) which implies only a small potential energy barrier ( 1-2 kcal. ) hindering rotation. Rotational relaxation times of the fragments greater than microseconds are not, therefore, excluded by these experiments ( Edclman, 1967). One observation that is difficult to reconcile with a very flexible hinge is the high density of this region in the electron micrographs. It is usually as well contrasted as the Fab and Fc fragments themselves, and there appears to be no systematic penetration of stain, which might have been expected if the region had consisted of rather open polypeptide chains. Occasionally there does appear to be a slight separation of the Fc from
ELECTROX 3lICROSCOPY OF THE IMMUNOGLOBULINS
13
FIG.4. Rabbit anti-DNP IgG coinl>inetl with DNP-ferritin ( 8 DNP groups per molecule ) . (Valentiiic and Grcen, iii~publisliedexperiments; Valcirtine, 1967). Many unbound Y-shaped anti1)ody inoloculcs can be scan as wrll as those linking the ferritin molecules. The apparcirt width of t h r fcrritin-boiuntl antibody is smaller than that of free antibody becniise of the greatrr t h i c h s s of staiti surmunding the fewitin. hfagirification: X400,OOO.
14
N. MICHAEL GREEN
the Fab fragments, and in the pepsin-treated polymer (Valentine and Green, 1967) a clear division can be seen between the two Fab fragments. This observation of a compact hinge conflicts slightly with the hydrodynamic model of Noelken et nl. (1965) which in most respects agrecs well with the model that is shown in Fig. 3. If the axial ratio of IgG is taken from the elcctron micrographs as approximately 3:1, then a frictional ratio of 1.47 implies 0.9 ml./gni. of hydrodynamically trapped solvent (Edsall, 1953). The fractional ratios and dimensions of the fragments show that they are not highly solvated, so that more than half of this solvent would have to be in the hinge region, implying a much more open structure than is seen in the electron micrographs. An alternative analysis of the hydrodynamic data (Charlwood and Utsumi, 1969) by the method of Bloomfield et al. (1967) led to rather similar conclusions. They showed that it was possible to account for the sedimentation constant of IgG in terms of those of its fragments, only if the centcr-to-center distances of the Fc from the Fab fragments was 78 & 10 A., some 20 A. greater than that iiiexured on Fig. 2. These discrepancies are probably significant and may indicate that the hinge region in solution is less compact than it appears to be in the electron micrographs. This would also be consistent with its susceptibility to proteolysis.
C. THEQUESTION OF CONFORMATIONAL CHANGE The electron microscopic evidence for the three-armed modcl was derived from polymers obtained with bivalent hapten. Electron micrographs of single antibody molecules often do not show this structure clearly, but it is not certain whether this is a result of technical difficulties or of a genuine change in structure. The idea that the IgG molecule undergoes a change in conformation when complexecl with polyvalent antigens has received support from many quarters. The main evidence comes from the new immunological activities of antigen-antibody complexes, such as their ability to fix coniplement, to bind to skin receptors, and to induce new antibody response (e.g. Kabat and Mayer, 1961; Gel1 and Kelus, 1967; Henney and Ishizaka, 1968). Complexes with simple haptcns do not show any of these properties, whereas nonspecific aggregates of IgG or of Fc fragments do. The new properties oftcn appear to be more a consequence of aggregation than of specific combination with antigen followed by specific change in Conformation. Elrctron microscopy is a w r y insensitive test for conformational change since it can only dctcct changes in stability or arrangemcnt of
ELECTROS hIICROSCOPY O F THE IMMUNOGLOBULINS
15
whole subunits. It may bc: argued conversely that any change seen h ~ this method is likely to reflect a fairly substantial altcration in the molecule. Feinstein and Rowe (1965) suggested a correlation between the appearance of the new immunological propcrties and the apparent opening up of the molecule suggested by their electron micrographs. The inore detailed picture obtained by Valentine and Gwen (1967) did not support such an extensive change though there was sonic evidence for a more clcarly defined subunit structurc whctn thc IgG molccules were polymerized. ‘The most serious obstacle to reaching a firm conclusion about the changes that take place on combination with antigen has been the difficulty exprienced in obtaining wcll-stained preparations of uncombined IgG molccules. Where uniform, well-contrasted specimens are obtained, thc tripartite nature of thc single molccule becomes clear. For example, occasional single Y-shaped molecules can be seen among thc polymers formed with bivalcnt hapten (Fig. 2 ) . The clearest singlc molecules that have been observed so far were in preparations of anti-DNP antibody combined with DNP-ferritin ( Fig. 3; Valcntine, 1967). Thesc elrctron micrographs were made for direct comparison with the results of Feinstcin and Rowe (1965) and they illustrate two further points of interest. First, the lengths of the antibody molecules linking ferritin molecules togethcr (aliout 140 A ) is not significantly greater than that observed in the cyclic polyincw; second, thc. appearance of the ncgatively contrasted molecule depends on the thickness of the stain. The connecting strands in thc thicker stain surrounding the ferritin appear to be little morc than half thc width of the molecules in the thinner areas away from the ferritin. There are two further factors which lead to enhancement of the regular appearance of polymerized molecules of IgG and, hence, to ; i n impression of a change in conforniation. First, the molecules lic on the grid in the same orientation (apart from variation in the anglc hetween the arms) ;md, sccond, the union of the Fa11 arms end to cncl emphasizcs their length. This sccond factor can hc c,liminatcd b y isolating the corners of the cyclic polymers with an appropriate mask so that constituent molecules are seen in isolation. Profiles traced in this way arc shown in Fig. 5 where t h y are compared with those of sing](. molecnlcs. Taking nll these factors into account. I would conclude that there is no firm electron-microscopic evidence for an extensive change in the molecular conformation accompanying comhination with antigcn. HOWTVCT, tlierc. clearly must he some change to account for the new immiinological properties. Sincc, tlwse are associated mainly with the Fc fragmcnt of the inolcculc, thcy could wcll be ;I rcsiilt of slight chances in the hing(’ rcxgion ( Fcinstcin and Rowc, 1965).
7
N. MICIIAEL GHEEN
FICA5. Tracings of IgG moleciiles at a magnification of X300,OOO. ( a ) Taken from corners of cyclic polymers; (11) single molecules from Fig. 4; ( c ) single molecules from Valentine and Green (1937). The single molecule approximates to an isosccIes triangle with the sides slightly longer than the base, of which the dimensions are 120 X 120 X 102 A. (iiienn of ineasurenients on twmty-five molecules). The corresponding dimensions for molecules taken from the corners of cyclic trimers was 121) x 120 x 115 A.
Although therc are no obvious differences between the populations of antibody molecules shown in Figs. 2 and 4, there could be a differcnce in the fraction of molecules with large inter-Fab angles, which could be critical for the binding of the complex to a third component. If this binding is multivalent as it appears to be for C', ( Muller-Eberhard, 1968j, then this effect could well be amplified, since it would be dependent on the proximity of several molecules in the appropriate open conformation. An inteiprctation of this type would also be consistent with the formation of weak complexes between free TgG or TgM and C , ( Muller-Eberhard and Calcott, 1966j, The suggestion of an increased angle between the Fab's is supported by the results of the measurements of single and polymerized molecules, summarized beneath Fig. 5. Whereas the molecules at the corners of cyclic trimers can be represented by a triangle that is approximately equilateral ( 120 A. x 120 >< 115 A.), the single molecules shown in Fig. 4 have one dinicnsion that is significantly shorter than the other two ( 120 x 120 X 102 A. j. This suggests that the angle between the Fab fragments in the most probable conformations of the nioiioiner is appreciably lcss than GO", so that angles greater than GO" may only be common in complexes with polyvalent antigens. This argument can only he tentative, first, since it assumes the IgG molccules shown in Fig. 4 provide a representative sample of the conformations present in solution and, second, since thr difierence between 102 and 115 A. is barely significant.
ELECTRON hIICROSCOPY OF THE Ih$MUKOGLOBULINS
17
The optical rotatory dispersion ( O R D ) spectrum is a more sensitive indicator of confoimational change than is electron microscopy, and it is worth considering briefly R few relevant observations. Changes in Q, have been observed in soluble complexes of antibody with scrum albumin, with ferritin, and with a synthetic trivalent hapten (Ishizaka and Campbell, 1959; Henney and Stanworth, 1966), but, except for the last example, it cannot be certainly concluded that they are due to changes in the antibody molecule, because of the laige contribution from the antigen to the total rotation. Coinplexcs with simplc haptens that have been examined do not fix complement, nor do they show any changes in ORD ( Steiner and Lowey, 1966; Cathou et nl., 1968). Howevcr, the correlation between complement fixation and change in optical rotation docs not always hold. The complexcs of anti-DNP with biq-DNP-octamcthylenediaminch do fix coinplcmeiit ( N. Hyslop, unpublished evpeiiments ) although their ORD spectrum is indistinguishable from that of the cornplcx with univalent DNP aminocapl-oate ( N. M. Green, unpublished experiments). Clearly more experimental work is rcquircd, preferably with simple antibody-hapten systems and purified components, to obtain a clcar mswcr to the problem. It will probably be more informative to look directly at the binding intcractions in such systems then to make ORD measurements on the antigen-antibody complexes, since the critical hinge region may have a somewhat Ioosc and variable structure which will make no characteristic contribution to the ORD curve. In support of this view, it has been found that proteolytic splitting of this part of the n~oleculeis without effect on the ORD or circular dichroism of antiDNP antibodies (Stc-iner and Lowey, 1966; Cathou et al., 1968) nor docs it affect the ORD of IgM (Dorrington and Tanford, 1968). IV.
Electron Microscopy of IgM
The chemical evidence for the structure of IgM is lccs extensive than for that of IgG. Much of it derives from the study of Waldenstrom macroglobuhs ( Miller and Metzgcr, 1965a,b), hut similar results havc been obtaincd with ralhit IgM antibodies (Lamni and Small, 1966). A fairIy consistent picture of the overall layout of the peptide chains in thc moleculc has bcen provided by this work which has shown many similarities between the 7 S subunit of Ighii and the IgG molecule. Each 19 S molecule (mol. wt. 900,000) contains five 7 S wbunits linked together 13y single disulfitle tionds, possibly nccir the C - t c m k n l ends of the FC fragments (Abel and Gray, 1967). Each 7 S subunit contains two H ( p ) chains and two L ( h or A ) chains joincd b y 21 p'ittein of disulfide bonds differing little froin tllclt in IgG ( hIi1lc.r and \Ictzqcr. 1965b). It has l)c.cn
18
N. MICHAEL GREEN
possible to obtain Fab and Fc fragments homologous with those derived from IgG by slight modification of the proteolytic procedures used so successfully on the smaller molecule. Digestion with pepsin ( Mihaesco and Seligmann, 196813; Kishimoto et al., 1968) or trypsin (Miller and Metzger, 1966) gave Fab p or F(ab), p fragments, depending on the extent of proteolysis. The Fc p fragments (3.2 S), similar to those from IgG, result from the usual digestion with papain in the presence of cysteine ( Mihaesco and Seligmann, 1968a). If the activating thiol was removed before the reaction, a much larger Fc (10.6 S ) was obtained in which five 3.2 S fragments were united by disulfide bonds to give a structure with a molecular weight of 320,000 (Onoue et al., 1968b). The task of the electron microscopist is to assemble these fragments to give a molecule consistent with the rather varied and unusual profiles shown in his pictures. Electron microscopy may also help to answer the question of the number and location of the binding sites, which is still under discussion. Scveral groups of workers (e.g., Onoue et ul., 1965; Frank and Humphrey, 19168) gave evidence for only five sites in the whole molecule (one for each subunit) which implicd that the two Fab p fragments in each subunit differed from each othcr. Recent work (Merler et al., 1968; Stone and Metzger, 1968; Ashman and Metzger, 1969) suggests that there are ten sites, a finding which is more consistent with the other chemical evidence. It has been suggested that these sites belong to two different classes ( Onoue et al., 1968a ) . The earliest electron micrographs of IgM ( Hoglund, 1965; Hoglund and Levin, 1965) were of shadowed preparations. They showed ellipsoidal molecules (300 x 200 A , ) , which appeared to be much too large for a molecular weight of 900,000 (Rowe, 1966). Later studies by Hoglund (1967a,b) using negative staining also showed molecules about 250 A. across, with a smooth approximately circular profile, but no indication of their thickness was obtained. Occasionally the stain (uranyl acetate) penetrated into the molecule and a suggestion of a more open structure of rodlike components was obtained. Other authors have consistently found two different characteristic appearances of IgM molecules, dependent on whether they were free or bound to the surface of an antigen. The bound antibody often appeared as a well contrasted bar (170 x 35 A.) about 80 to 100 A. from the surface of the antigen and linked to it by a number of thin strands which ;ire difficult to define and count. This structure has been seen on fragments of erythrocyte mcmbrane ( Humphrey and Dou~mashkin,1965) and on Salinonella flagella ( Feinstein and Munn, 1966; see, also, Cohen and hlilstein, 1967). When the antigenic sites were prcsent on virus particles rather than on an extended surface, a slightly different picture was
ELECTHOR; 1\IICHOSCOPY O F THE I2r4MUSOGLOBl.JLIh’S
19
FIG.6. Poliovirus aggregated by rabbit Ighl antilmtly. The antibody molecule appears as a well-contrasted bar linked to the viral surface by thin, soiiietiines invisible, strands. Occnsionally a central ring structure can be seen ( a ). Magnification: ( a ) X200,OOO (Chesebro imtl Svehag, 1969); (11) X400,OOO ( Svehag and Bloth, 1967).
obtained (Fig. 6 ) . The central bar was not always straight and arms protruded from it in all directions, sometimes linking virus particles together (Almeida et al., 1967; Svehag and Bloth, 1967). Sometimes the link appeared as a continuous straight bar up to 370 A. long, and it has in fact
20
N. MICHAEL GREEN
FIG.7a FIG.7. R l t anti-DNP IgM. The contrast is weak apart from bright spots, 3040 A. across, seen on the projecting arms or near their bases, but not in the center of the molecule, The mean span of the molecule is about 270 A,, the mean area 31,000 A?, and the arms are not more than 90 A. in length, Magnification: ( a ) X400,OOO; ( I ) ) X1,100,000 (unpublished work of R. C. Valentine and R. Binaghi).
ELECTRON MICROSCOPY O F THE IMMUNOGLOBULINS
21
FIG. 7b
been suggested that IgM may be a long flexible rod 50 x 370 A. ( Almeida et al., 1967). However, such a simple picture is difficult to reconcile both with the chemical evidence that the molecule contains five 7 S subunits and with the electron micrographs of uncombincd IgM. Svehag and Bloth (1967) described the IgM molecules attached to poliovirus as multiple (2-1)looped striicturcs like repeated “in.” The number of l o o p varied, and in antigen excess extended tails could be seen linking two or more virions together (Fig. 6 ) . The maximum length was 360 A. (mean 330 A ) , and the mean distance from the vertex of a loop to the virion was 85 A. It was suggested ( Svehag et ul., 1967b) that each loop (sometimes with a projection from the vcrtex) might represent a 7 S subunit of IgM and, thus, resemble the loops of IgG observed on other virus particles (Lafferty and Oertelis, 1963; Almeida et d.,1963; Chesebro and Svehag, 1969). Uncombined IgM appears usually as a multiarmed ameboid structure spread out on the carbon film. This was first observed for Waldenstrom niacroglobulins, for rabbit antipoliovirus IgM ( Svehag et aZ., 1967a,b), and for whole bovine IgM (Feinstein a i d M u m , 19167). This is illustrated in Fig. 7 which shows rat anti-DNP IgM. The appearance is very different from that of IgM bound to antigen, and this led Feinstein and Munn
22
N. MICHAEL GREEN
(1966) to suggest a marked change in structure when the antibody coinbined with membrane, flagellum, or virus. However, careful consideration of the geometrical factors involved show that the two contrasting appearances are by no means incompatible with each other. For the most part, the uncombined molecule shows very weak contrast suggesting that it is disclike rather than ellipsoidal. The central area of the moleculc appears particularly thin. The mean area of such molecules (taken from Fig. 7 and other similar fields), measured by superimposing tracings on graph paper, was 31,000 A.' (mean of 30, range 27,000-36,000 Az).From this area and from the molecular weight (900,000) and partial specific volume (0.73), an average thickncss of 35 A. was calculated, very close to the thickness of the strands seen when the molecule is bound to antigen. It therefore seems likely that the highly contrasted bar seen with bound antibody is an edgewise view of the central disc and that this is linked by hinged arms to the surface of the antigen. Since the bound molecules are seen most clearly when the antigen surface is perpendicular to the carbon film, thc aspect of the molecule obtained is totally different from that when it is adsorbed directly on the carbon film. The picture becomes slightly more complicated when the binding surface is provided by several virus particles-arms may project in any direction from the central disc (Fig. 6)-but the basic interpretation can still be maintained. If this is correct, then the arms represent Fab fragments attached to a central disc of Fc fragmcnts and the high contrast spots on the ameboid IgM molecules (Fig. 7) are probably Fab arms viewed end on. This interpretation can be supported and elaborated following further extensive work by Svehag and his colleagues. They took into account the chemical evidencc and concluded that the IgM molccule was basically a cyclic pentamer (Svehag et al., 1967a, 1968a,b; Chesebro et al., 1968), the units being held together by disulfide bonds bctwecn Fc p fragments, which would then form a central ring. This suggestion was supported by the characteristic appearance of a small proportion of the IgM molecules, which showed such a ring with a central hole (Fig. 8a), rather than the continuous thin disc suggested by Fig. 7. These were seen in about twenty preparations of IgM from different sources (S.-E. Svehag, personal communication), but only occasionally (e.g., Fig. 6a) when the IgM was bound to antigen, presumably because edgewise views of the molecule are favored in this situation. The arms of the pentamer were about 35 x 125 A., but it was not often that all five were clearly resolved. If there are only five arms and there are ten Fab p fragments to be accounted for, either each arm consists of two Fab's or onc of each pair of Fab's is somehow incorporated into the central disc (Steffen, 1968). The
ELECTHON hfICROSCOPY OF THE IMMUNOGLOBULINS
23
sccond hypothesis involves an improbable asymmetry, which would imply two different ,U chains in the one molecule, and would be difficult to reconcile with the evidence for ten binding sites mentioned above. The presence of two Fab p fragments in cach arm would avoid such difficulties and it coulcl account for the very variable appearance of the arms, sometimes short, sometimes fused with their neighbors, and sometimes more than five. This suggestion is supported by two papers on the submit s and protcolytic fragments of IgM (Svehag et al., 1969; Chesebro and Svehag, 1969) . Preliminary studies of alkylated IgM subunits, produced by reduction with 0.2 hl mercaptoethanol (Chesebro et aZ., 1968) showed very thin (15 A.) strands about 100 A. long with a knob of diameter 50 A. at one end. This would appear to be too sinall to represent a subunit of rnolecular weight 180,000 but would be consistcnt with the size of a half subunit containing one light and one heavy chain. It is known that reduction (0.2 A t iiiercaptocthanol) cleavcs the disulfide bond between the two halves of the subunit (Miller and Metzger, 1965b), and it is likcly that negativc staining in silicotungstate could cause separation of these two halves since they are held togcther only by noncovaleiit bonds. Such dissociation has been observed by Valentine and Green ( unpublished experiments ) with the polymcrs of anti-DNP IgG, discussed above. In silicotungstate, these dissociated after mild reduction ( 0.01 91 mercaptoethanol, pH 7.4) into thin strands which were probal-jly hapten-linked dimers of half molecules. A morc detailed study (Chesebro and Svehag, 1969; Svehag and Bloth, quoted in Svehag et al., 1969) using milder reduction conditions, which should give covalently linked subunits ( Millcr and Metzger, 1965b), showed units of two sizes 90 x 35 A. and 90 x 55 A. The appearance of the larger one, amounting to up to 30%of the total, was more consistent with the expected molecular weight, but it is difficult to see how it coulcl be assembled to give all the variable profilcs of IgM. Reoxidation of the reduced subunits led to regeneration of 30 to 50% of the antibody activity and to IgM molecules of characteristic appearance. Fragments obtained by proteolytic digcstion provided clearer information about the substructure of the IgM molecule ( Svehag et nl., 1969). The large (10.6 S ) Fc ,U fragment was obtained using papain in the absence of cystine (Onoue et al., 1968b) and w a s separated from smaller fragments on Sephadex. Electron microscopy showed two types of structure: ( 1 ) a ring with a central (40 A.) hole, an outer diameter of about 85 A,, and a number (nciirer to ten than to five) of radial projections, 20-25 A. long ( 8b, 8c); ( 2 ) “screwlike” structures, 40 x 200-500 A. long,
24
N. MICHAEL GREEN
FIG.8. IgM and its fragments. All scale lines are 200 A. ( a ) Intact molecule showing central ring structure. The central hole has a diameter of 40 A. and the five arms are 70-100 A. X 35 A. Magnification: X470,OOO (S.-E. Svehag, unpublished observation). ( b ) Fc p fragments from papain digest of human IgM. The outer diameter (taken froiii a number of difFerent niolecules) is 85 A. A number of sIiiall (20 A , ) projections can often be seen. Magnification: X460,000 (S.-E. Svehag, unpublished observations). ( c ) Aggregated Fc ,u fragments from papain digest of human IgM. Magnification: X230,OOO (Svehag et al., 1969). ( d ) F ( a b ) ? ,u from pepsin digest of human IgM. Club-shaped Fab ,u fragments are joined in pairs by their narrow ends, which cannot always be seen. Magnification: X330,OOO (Svehag et al., 1969).
consisting of chains of several subunits, each unit being a double bar orientated at right angles to the screw axis. The ring structures probably originated from the five 3.2 S Fc p fragments which form the center of the IgM molecule. The screwlike structures probably arise froin the rings by breakage of a disulfide bond followed by rotation of the subunits relative to each other and consequent opening up of the ring. The longer (500 A ) polymers could then be produced by a further disulfide interchange with another ring or oxidative dimerization with another chain. The initial ring opening could have been brought about by the -SH of the papain itself since the digestion mixture contained several moles of papain for each mole of IgM. Small (3.2 S ) Fc p fragments were sometimes seen as dimeric structures, the stain having penetrated between the two halves of the molecule.
ELECTRON MICROSCOPY O F T H E IMhfUNOCLOBULINS
25
In the s m i e p a p c ~ Svehag , ct nl., examined Fall’ and F( ah’)? l’c obtaincd by pcpsin digestion ( Mihacsco and Scligmann, l968n; Kishimoto et mZ., 1968). ’The monomcrs wcrc rod-shapcd i d oftell q p g a t c d sido to side. Sometimcs thcy appeared to consist of two parallel strands, one 80 x 15 A. and the other 50 and 15 A., possibly represcmting half the H chain and the L chain, respectively. If this interpretation is correct then in order to account for their niolecrilar wcights the folded H and L chains would appear to have a ribbonlike cross section about 15 x 35 A. Thc F(ab’), p fragments (Fig. 8 d ) consisted of two parallel Fah’ p,’s linked by their narrow ends, though it was not always possible to resolve the, linking region. The electron micrographs which showed the division between H and L chains and betwcen the two sections of H chain in the Fc fraginent were all obtained with uranyl fortnate as thc negative stain, of which the small size facilitates pcnetration. Thcse effects were not observed when the larger silicotungstate ion was used ( S.-E. Svehag, pcrsonal communication), and this may account for thc failure to sec any such divisions in the Fab and Fc fragments of IgG (Valentine and Grcvii, 1967). Most electron micrographs of IgM are consistent with its i‘or~nulation as a cyclic pcntainc~r,provided that due allowance is made for differences in orientation. The question of whethcr there are five or ten arms may still be regarded :is open. Svehag et al. (1969) concluded that the Fa11 fragments could not be seen as separate units in the whole IgM moleculc and that each arm represented a pair of fragments which were not in&pendently hinged. Stone and Mctzger (1968) used a similar model, with steric hindrance betwcen pairs of sites, to explain the inability of thcir anti-Fc, IYaldenstrom IgM ( YR,Lay) to bind morc than five molecules of antigen, whereas tlie separated Fab p fragments could bind tcn. Howevcr, a minority of IgM molecules show a clear five-armed structure, and it is much easier to nccount both for the molccular polymorphism and for the localized areas of high contrast within the molecule (Figs. 7, 8a; and Chesebro et al., 1968) if a model with ten independently hinged Fall arms is assumed. This would imply that each 7 S subunit had thc same basic Y structure as the molccule of IgG. Thcrc must, of course, be considerable differences of detail since the p chain is some 208 longer than the y chain. This extra length may be iiccountcd for by the narrow tail (30 x 15 A , ) of the Fab p fragment and by the 20-25 A. projections from the ring-shaped Fc fragment. The conformational changes in IgM which may occur when it binds to surface antigens are even less well characterizcd than are those in IgG. The complex certainly binds and activates complement very effectively, ,(,
26
N. MICHAEL GREEN
so it is likely that changes parallel to those in IgG occur near the junction region of the two Fab fragments of each subunit. This would provide further support for the idea that the two Fab arms have some freedom to move independently. It is possible that IgM from different sources may differ in this respect, but so far there is no good evidence for this. V.
Comments and Conclusions
It is unlikely that improvements in electron-microscopic technique will show finer details of structure than have already been obtained, although it may eventually be possible to resolve more clearly the individual H and L chains (perhaps better called H and L subunits, since they possess a definite tertiary structure). One of the next developments is likely to be the study of further antibody types and classes. Detailed morphology of IgG has been reportcd only for rabbit antibodies (Feinstein and Rowe, 1965; Valentine and Green, 1967), however, thc chemical and physical similarities between antibodies from different species suggest that there are likely to be few detectable interspecific differences detectable by electron microscopy. A preliminary examination of yl and y2 anti-DNP antibodies from guinea pigs (Valentine, Green, and Binaghi, unpublished experiments) showed some cyclic polymers with bifunction haptens, similar to those seen with rabbit antibody. Polymerization was less extensive and more open chain structures were present, possibly because of the lower affinity of the antibody ( K = 2 x M ) . It will be of interest to examine antibodies with known differcnces in the pattern of disulfide bonding such as the different classes of human IgG ( Frangione et al., 1969) and mouse IgA ( Abel and Grey, 1968) to see if this affects the angular relations between the fragments. IgM from different species shows no distinguishable differences, although the serum a,-macroglobulin, of unknown function, has a quite different shape, resembling the Russian xc (Bloth et al., 1968). Bifunctional haptens may be useful both for clarifying the subunit structure and for the formation of relatively simple antibody complcxes, of which the interaction with complement components may be observable by electron microscopy. It is, therefore, worth making a few general comments about their use. The bis-DNP-polymethylene diamines used by Valentine and Green (1967) were very insoluble in water, so that they had to be dissolved in dimethylformamide and added very slowly to the antibody. If the dissociation constant of antibody for DNP was M the reagent precipitated and did not react quantigreater than tatively. It is possible to apply the technique to lower affinity antibodies only by use of water-soluble bis-DNP compounds (e.g., derived from
ELECTI3OK hIICROSCOPY OF THE IhIhIUNOGLOBULINS
27
diaminosuccinic acid; N. M. Green, unpublished experiments). Further problcws aris? if K > lo-' R I , bccause of the low protein concentration employed for clectron microscopy. For example, N suitable concentration of antibody is about 50 pg./ml. ( 3 x lo-; AZ) which limits the hapten concentrations also to this order of magnitude. Use of higher concentrations of hapten to saturate the sites would give predominantly monofunctional binding, defeating thc purposc of the experiment. Attempts to produce specific polymerization of rat anti-DNP IgM using bis- ( DNPaminocaproy1)diaiiiinosuccinate gave ncgativc results, presumably because the dissociation constant ( 3 x lo-' M ; Binaghi and Oriol, 1968) was too high. I t may be possible to circumvent this limitation by mixing antibody and hapten at higher concentration and fixing with glutaraldehyde before dilution. In the case of IgM it may not even be necessary to fix thc polymers provided that there arc polyvalent links between molecules and thc molecules are taken up on the carbon film ininiediately after dilution. One furthcr application of bifunctional reagents in which the two functional groups are different may be mentioned. For example, a compound containing 110th a DNP group and a reactive group X, specific for say the catalytic center of a multisubunit enzyme, could be used to bind anti-DNP antibody (or a univalent Fab fragment) to this site. This could enable the arrangement of catalytic sites within the molecule to be determined b y electron microscopy. This would be a high resolution modification of the use of ferritin-labeled antibody, with the added advantage that it would be unnecessary to prepare antibody specific for the catalytic site. The specificity would bc provided by the appropriate reactivc group, X. Electron microscopy, like alniosc all physical techniques for the study of protclin structure, provides only a limited view of the protein molecule, so thdt the clc,ctron micrographs can rarely be interpretcd unequivocally without consideratioii of the results from other techniques. The results discusscd in this review do lcad to structures for both IgM and IgG reasoualdy consistent ~ i t hmost of the other physical and chemical evidence, although a few qiicstions remain which cannot yet be resolved. Oiic: of thest. is tlie cxtent of the eonformational changes accompanying tlie binding to polyvalent antigens, which lead to well-defined biological effects. In this revicw, I havc discounted most of the earlier electroniiiicroscopic evidence in favor of cxtensivc, changes since most of it can be sntisfactorily intciprcted without invoking anything more than selection of certain pre-existing conformations b y the antigen. There still reInnins the possi1)ility that the differcnccs olxervcd in cblectron micro-
28
N. MICHAEL GREEN
graphs bctwcen free and bound antibody reflect genuine large changes in molecular structure and are not merely the consequence of variation in orientation or in stain thickness. However, stronger evidence in support of this view is required before it can be accepted, for these differences imply rather extensive deformation of the Fab and Fc fragments, which is nt variance with their bchatior in solution as stable globular proteins. Active stretching of an antibody by a large antigen is an unlikely occurrence in the absence of a specific chemical mechanism. Since the mcan trmslntional kinetic energy of a molecule is independent of its size, large antigens are no more likely to produce such effects than are small ones, unless forces due to liquid flow or surface tension ale involved. Steric interference between large antigens may, however, tend to stabilize 3 more open conformation of an IgC antibody. ACKNOWLEDGMENTS This review was to have been written in collaboration with the late Dr. R. C. Valentine. His contribution can be seen clearly in the electron micrographs used for illustration as well as in the interpretations, which owe mnch to his influence. I would also like to thank Dr. R. R. Dourinashkin for much helpful advice and assistance in evaluating the evidence.
REFERENCES Abel, C . A,, ancl Grey, H. M. (1967). Science 156, 1609. Abel, C. A., and Grey, H . hl. (1968). Biochemistry 7, 2682. Almeicla, J. D., Cinncler, B., and Howatson, A. (1963). J. Exptl. Med. 118, 327. Alnieida, J. D., Cinnder, B., and Naylor, D. (1965). Inimzinochernisliy 2, 169. Almeida, J. D., Brown, F., and Waterson, A. P. (1967). J . Inimunol. 98, 186. Anderson, T. F., and Stanley, W. Ivl. (1941). J. Biol. Chem. 139, 339. Andres, G. A., Hsu, K. C., and Seegal, B. C. (1967). I n “Handbook of ExperimentaP Immunology” ( D . hl. Weir, ed.), p. 527. Blackwell, Oxford. Ashnian, R. F., and Metzger, H. (1969). J. Biol. Chcm. 244, 3405. Baker, M. C., Lyons, P. A., and Singer, S. J. (1955). J . Am. Cheni. Soc. 77, 2011. Binaghi, R., ancl Oriol, H. ( 1968). B d . Soc. Chim. Biol. 50, 1035. Bloomfield, V. A,, Van Holde, K. E., and Dalton, W. 0. (1967). Biopolyrners 5, 149. Bloth, R., Chesebro, B., antl Svehag, S.-E. (1968). J. Ezpfl. Aged. 127, 749. Brenner, S., and Horne, R. W. ( 1 ). Biochint. Biophys. Acfa 34, 103. Cuthou, R. E., Kiilczycki, A,, and Haber, E. (1968). Biochem. J. 7, 3958. Crbra, J. J,, Steiner, L. A., and Porter, R. R. (1968). Biochen~.J. 107, 79. Cliarlwood, P. A,, and Utsnmi, S. (1969). Biockem. J. 112, 357. Chesehro, B., and Svehag, S.-E. (1969). J. lniniunol. 102, 1064. Chcscbro, B., Bloth, B., antl Svehag, S.-E. (1088). J. E x p t f . Med. 127, 399. Cohen, S., and Milstein, C. ( 1967). Atloan. I n i t t i u n d . 7, 1. Cohen, S., antl Portcr, R. R. ( 1964). Adunn. ItnmtmoZ. 4, 287. Dorrington, K. J,, a n d Tanfortl, C. ( 1968). J. BioZ. C h e m 243, 4745. Easty, G . C., ant1 hlet.ccr, E. 13. ( 1958). lmmtmoZog!/ 1, 353.
ELECTHOS MICROSCOPY OF THE IhlXZUXOGLOBULINS
29
Etlelinan, G. hf. (1967). Nobel Symp. 3, 281. Wiley (Interscience), New York. Edelman, G. M . , and Cally, J. A. (1964). Proc. Natl. Acad. Sci. U.S. 51, 846. Edsall, J. T. (1953). In “The Proteins” (H. Neurath and K. Bailey, eds.), Vol. 1, p. 549. Acudemic Press, New York. Elck, S. D., Kingsley Smith, B. V., and Highman, IV. (1964). lmnirinology 1, 353. Feinstcin, A,, and M u m , E. (1966). J . Phy.sioZ. ( L o n d o n ) 186, 64P. Feinstein, A,, and Xlunn, E. ( 1967). 171 “Handhook of Experiinental Immunology” ( I ) . hl. Weir, etl.), p. 277. Blackwell, Oxford. Fcinstein, A,, and Rowe, A. J. (1965). Natirre 205, 147. Finch, J. T., and Holines, K. (7. (1968). I n “Methods in Virology” ( K . Maramorosch and H. Koprawski, eds.), Vol. 3, p. 352. Academic Press, New York. Frangione, R., Xlilstcin, C., ancl Pink, J. R. L. (1969). Nature 221, 147. Frank, M. %I., and Humphrcy, J. H. (1968). J. ExptZ. Med. 127, ‘367. Cell, P. C,. H., and Kelns, A . S. (1967). Adoan. Inimunol. 6, 461. Goldstein, D. J., Humphrey, R. L., and Poljak, R. J. (1968). J. Mol. B i d . 35, 247. Hall, C. E., and Doty, P. (1958). J. AVL.Chcm. Soc. 80, 1269. Hdl, C. E., Nisonoff, A., and Shyter, H. S. (1959). J. Biochem. Biopkys. Cytol. 6, 407. Henney, C. S., and Isliizaka, R. (1968). J. Inmitinol. 100, 718. Henney, C. S., and Stnnworth, D. R. (1966). Natirre 210, 1071. Hoglurid, S. ( 1965). Proc. European Rcgioiial Conf. Electron Microscopy, Pragtrc, 1964, Vol. B, p. 55. Hiiglund, S . (1967a). Nohel S y m p . 3, 259. Wiley (Interscience), New York. Hiigluiid, S. ( 196,713) . Virology 32, 662. Hoglund, S. (1968). J. Gen. Virol. 2, 427. Hiiglund, S., :ind Levin, 0. (1965). J. Mol. B i d . 12, 866. IIorne, R. W. (1965). I n “Qu;intitdtive Electron Microscopy” ( G . F. Bahr and E. H. Zeitler, eds. ), p. 316. Willianis & Wilkins, Baltiinore, Maryland. Horne, R. W. ( 1968). In “hlethods in Virnlogy” ( K. Maramorosch and H. Koprowski, ecls.), Vol. 3, p. 522. Acadeinic Press, New Yolk. Horne, R. W., and Creville, C . D. (19G3). J. Mol. Biol. 6, 506. Hummeler, K., Anderson, T. F., ancl Brown, R. A. (1962). Virolo,g!g lG, 84. Humphrey, J . H., and Ilourmashkin, R. R. (1965). In “Complement” ( G . E. W. Wolstcnholme and J. Knight, eds. ), p. 175. Churchill, London. Ishizakn, K., and Campl)rll, D. H. (1959). J. Inimunol 83, 318. Kalxit, E. A. ( 1968). “StructitraI Concepts in Imninnohgy and IminuI-lochemistry,” 13. 192. Holt, New York. Kalxtt, E. A., and Mayer, hl. hl. ( 1961) , “Experiinental Immunocliemistr~,”p. 133. Thomas, Springfieltl, Illinois. Kishiinoto, T,, Onoue, K,, and Yaniamura, Y. ( 1968). J. Imniurtol. 100, 1032. Lnfferty, K. J., and Oertelis, S. (1963). Virolog!/21, 91. Lamin, bl. E., and Small, P. A. (1966). Biochenti.ytry 5, 267. Lcberman, R. ( 1963). J. Mol. B i d . 13, 606. Mellenia, J . E., Van Britggen, E. F. J., and Gruber, M. (1968). J . Aid. B i d . 31, 75. hlerler, E., Karlin, L., and hlatsumoto, S. ( 1968). J. Biol. Chcnt. 243, 386. Mihaesca, C . , and Seligmann, M. (1968a). I . Exptl. Med. 127, 431. Mihaesco, C., and Seligmann, hl. ( 196813). Immunoclwmistry 5, 457. Miller, F., ancl Metzger, H. ( 1965a). J. Biol. Chern. 240, 3325. Miller, F., and hfetzgcr, H. ( 19651) ). J. Bid. Chctn. 240, 4740.
30
PIT, MICHAEL GREEN
Miller, F., and Metzger, H. (1966). J. Biol. Chem. 241, 1732. Miiller-Eberliard, H. J. (1968). Aduan. I~ninirnol.8, 1. hliiller-Eberhard, H. J., and Calcott, M. A. ( 1966). Immunoclaeniistry 3, 500. Neurath, H. (1939). J. Am. Chem. SOC. 61, 1841. Noelken, hl. E., Nelson, C. A., Bnckley, C. E., and Tanforcl, C. (1965). J. B i d . Chem. 240, 218. Oncley, L., Scatchard, G., and Brown, A. (1947), J. Phys. Chem. 51, 184. Onoae, K., Yagi, Y., Grossman, A., and Pressinan, D. ( 1965). Immunochemistry 2, 401. Onouca, K., Grossberg, A. L., Yagi, Y., and Pressman, D. (196%). Scietice 162, 574. Onouc, K., Kishimoto, T., and Yamamura, Y. ( 19681,) . .1. Inzmrittol. 100, 239. I’enhoet, E., Kochman, M., Valentine, R. C., and Rutter, W. J. (1967). Biocliernistry 6, 2940. Hifkintl, R. ( 1968 ) . I n “Methods in Immunology and I~nmunoclie~~iistry” ( C. A. Williams and M. W. Chase, eds.), Vol. 3. (in pi-ess). Academic Press, New York. Robinson, 1. 1’. ( 1966). J. Mol. B i d . 17, 456. R o w , A. J. (1966). J. M o l . Biol. 16, 553. Stanworth, D. R., and Pardoe, G. ( 1967 ). In “Handbook of Experimental Imniunology” ( D. XI. Weir, ecl.), p. 298. Blackwell, Oxford. Steffen, C. ( 1968). Z. lninirinilatsfnrsch. Allerg. Klim. lmrnunol. 135, 395. Steiner, L. A., and Lowey, S. (1966). J. B i d . Chem. 241, 231. Stone, A I . J., and hletzger, H. (1968). J. Biol. Cheiti. 243, 5977. Svrhag, S.-E., and Bloth, B. ( 1967). Virologil 31, 676. Svrliag, S.-E., Chesebro, B., and Bloth, B. (1967a). Science 158, 933. Svchag, S.-E., Chesebro, B., and Bloth, B. ( 1967b). Nobel S!/mp. 3, 269. Wiley (Interscierice), New York. Svehag, S.-E., Chesebro, B., and Bloth, B. (1968a). J. Exptl. Med. 127, 749. Svchag, S.-E., Chesebro, B., and Bloth, B. ( 1968b). Bull. Soc. Chim. Biol. 50, 1013. Svehag, S.-E., Bloth, B., and Seligmann, M. (1969). J. Exptl. Med. ( i n press). Terry, W. D., hlatthems, B. W., and Davies, D. R. ( 1968). Nutiire 220, 2.39. Valentine, R. C . (1959). Nature 184, 1838. Valentine, R. C. (1961). Arloan. Virirs Res. 8, 287. Valentine, I{, C. (1967). Nohel Symp. 3, 251. Wiley (Interscience), New York. Valentine, R. C. ( 1969). Proc. Europeaii Regional Cotif. Electron Microscopy, Rome, 1968. VOl. 2, p. 3. Valentine, R. C., and Green, N. M. (1967). J. Mol. Biol. 27, 615. Valentine, R. C., and Horne, R. \V. (1962). In “The Interpretation of Ultrzstructnre” ( R. J. C . Harris, ed. ), p. 263. Academic Press, New York. Valentine, R. C., a i d Pereira, H. G. (1965). J. Mol. B i d . 13, 13. Valentine, R. C., LVrigley, N. G., Scrutton, hl. C., Irias, J. J., and Utter, M. F. (1966). Biochcmistq 5, 3111. Valentine, R. C., Shapiro, B., and Stadtinan, E. R. (1968). Biochemistry 7, 2143. Wahl, P., and Weber, G. ( 1967). J. Mol. B i d . 30, 371. Wcltman, J. K., and Edelnian, C . M. (1967). Biochemistry 6, 1437. Wilcox, W. C., Ginsberg, H. S., and Anderson, T. F. ( 1963). 3. Expt!. Med. 118, 307. \\’illianis, R , C., and Wyckoff, R. W. G. (1945). Proc. Soc. Exptl. B i d . hled. 58, 265
Genetic Control of Specific I m m u n e Responses’ HUGH 0.McDEVITT’ A N D BARUJ BENACERRAF Loborafory o f Immunology, Nofional lnsfifufeo f Allergy and Infectious Diseases, Nafionol lnsfitutes o f Health, Bethesda, Maryland and Division of Immunology, Department o f Medicine, Stanford University School o f Medicine, S f a n f o r d , California
I.
Introdaction
.
.
.
.
.
.
.
.
.
.
.
.
.
11. Constitutional Differences in Individual Responses to Con~plrx Multi-
deterininant Antigens
.
.
.
.
.
111. Analysis of the hlechanism of Gene Action
I\’.
. .
. .
. ,
. .
. ,
Genetic Differences iii Immrine Response to Defined Protein Antigcms
V. Grnetic Differences in Immune Rrsponses to Synthetic Polypeptide . . . . . . . . . . . . Antigens .
. .
.
.
The Immune Response to Linear Random Copolymers of L-a-Amino Acids . . . . . . . . . . . . . B. The Iminiinc Rcaponsc~ of Guinea Pigs to Poly-L-Lysine and to . . . . . . . Hapteii-I’ol\.-~-Ly\ine Conjugates C. The Hesponse of Mice to Aranclie:l, hlnltichain Amiiio Acid Copolymers . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . References . . . . . . . . . . . . .
A.
VI.
I.
Introduction
The immune responsc to a specific antigen is a complex process which must involve genetic control at several levels. Following introduction into the animal, the antigen must interact with one, and probably several, cell types, thereby initiating a complex process of cell division and diffcrentiation which results in the appearance of plasma cells producing specific antibodies against the antigen and in the appearance of sensitized lymphoid cells capable of specific intcmction with the antigen. Not all of the steps in this process are known, but it is clear that genetic control could be exerted at many different points. Experimentally, the genetic control of specific immune responses can be approached in two ways: ( a ) by structural analysis of the products of the immune response-the antibodies; or ( b ) by scarching for and analyzing genetic control of the ability to recognize and respond to specific antigens. Genetic and structural analysis of normal immui~oglobuliiis,inyeloma proteins, and antibodies ( 1 ) has produced a great dcal of information ___ ‘ \Vork supported iii part b y U.S.P.11.S. grallt ilI 07757 Senior Iiivc’stig!;itor,Arthritis Fountlation.
31
32
HUGH 0. h l C DEVITT AND BARUJ BENACERRAF
about the genetic basis of antibody structure but has not yet given us a clear picture of the genetic basis of antibody specificity. The rabbit heavy-chain, allotypic markers controlled by the a locus are found on y , a, and p heavy chains (2, 3 ) . Since the a locus allotypic markers appear to be controlled by allelic genes, whereas the cistrons controlling the y-, a-, and p-immunoglobulin classes are nonallelic and present in all animals, this finding implies that the rabbit H-chain is controlled by at least two different structural genes. The evidence presently available indicates that the structural diff ercnces correlated with individual a allotypic markers are present on the N-terminal or variable half of the rabbit H-chain ( 4 ) . This finding further suggests that the segment of the variable region containing the a markers is coded for by a singlc gcrm line g e m since, if thcre were a large nuinbcr of such genes, crossing over would quickly lead to the loss of the genetic polymorphism represented by allotypes. The rabbit H-chain allotype data is, therefore, compatible with the existence of at least two separate cistrons-one coding for the variable region and one coding for the constant region-which combine in some way to code for the entire H-chain. Structural analysis of myeloma proteins has led to a similar conclusion for the human ( 5 ) (where three K subtypes have been described) and mouse light chain ( 6 ) . Amino acid sequence studies on mouse K-type Bencc-Jones myeloma proteins have shown that there are at least two different K-chain variable regions differing by four amino acid residues in length, both of which havc been found on K-chains with the same constant region amino acid sequence ( 6 ) . The implication of these results is that light chains are also coded for by two different structural genes, but that, in this casc, there are several variable region genes found in association with the same constant region gene. Thus, the results of genetic and structural analyses of immunoglobulins are in accord in indicating that both light and heavy chains have an N-terminal variable region and a C-terminal constant region, and that these regions are coded for by separate structural genes. The genetic control of antibody specificity is less well understood, and two nicchanisms have been postulated for generating the variability seen in the N-terminal regions of both light and heavy chains. One postulated mechanism is that there exist a large number of germ line genes, each of which codes for a particular variable region sequence ( 6 ) . The alternate postulate is that thcrc arc vcry few germ line, variable region genes, with a very high degree of somatic variability introduced by somatic recombination ( 7 ) or somatic mutation ( 8 ) . There are arguments for and against both of these hypothetical mechanisms, but there is no convincing
GENETIC CONTHOI, OF SI’ECIk’IC I \ I l I U S E
RESPOSSES
33
evidence for eithcr one. At the present time it is difficult to reconcilch the allotypic data on the rabbit I-I-cliain ( 2 - 4 ) with the existc>nccof a large number of germ line V genes coding for the entire variable segment of the chain. It is conceivable that enough sequence or allotype data will accumulate to permit a choice betwecm the two postulated mechanisms for the origin of antibody variability, and thus lead to greater understanding of the genctic mcchanisni by which nntibody structure has evolved and is controlled. It is possible, however, that a coinplet-e understanding of the genetic control of specific immune responscs will require genetic studies on the inheritance of the ability to rccognize antigenic determinants and to synthesize specific antibodies. This ~7ouldrequire the detection of individual animals which arc incapaldc of selected specific rcsponscs. Considering the enormous hctcrogeneity of the iinmune response, the likelihood of finding animals unabIe to respond to R particular antigen would be expected to be low, but studirs of this problem have, nevertheless, resulted in the detection of a wide variety of specific nonrespondcrs and have shown that the ability to respond to specific antigens and to specific :intigenic configurations is under direct genetic control. This revicw is concerned with a description of these studies and an evaluation of their implications. It is written in the expectation that the final rcxsults of these experiments will complement gcnetic and stnictural analyses of immunoglobulins to permit a more complete picture of the genetic basis of the imm LIIW response. II.
Constitutional Differences in Individual Responses to Complex Multideterminant Antigens
In an early study, Gorcr and Schutze were able to demonstrate a correlation between antibody formation and resistance to Snlnmrze&z infections in strains of mice which were geneticnlly resistant to this type of infection ( 9 ) . One of the most straightforward demonstrations that genetic factors play a role in the capacity to form specific antiliodics was presented by Scllcil~el( 1 0 ) in 1943. Random-1)red guinea pigs were iinmuiiizcd with diphtheria toxoid and were separated into good and poor responders to this antigen. There thcn followed several gencrations of selective brccding in which good rcq~onderswcrc brcd with good responders and poor respondcrs with poor. Thc results of s e ~ ~ geiiera~ a l tions of such sclectivc breeding are shown in Fig. 1. It is clerir that b y such a process of sclcsction it is possible to producc populations of guincn pigs that are unifonnly p c i d or uniforinly poor responders to diphthcria
34
HUGH 0. M C D E V I T T AND BARUJ BENACERRAF
a0
-
60 40 20 -
CURVE 2 100-
a0 -
-
60 40 20
-
-
51751 F,[1791 FJ2251 511621 51691
FIG.1. Curve 1 s h o w the percentage of good antitoxin pro-icers (dashed line) and poor antitoxin producers (solid line) in successive generations of inatings between good producers. Curve 2 is similar, and shows the results for matings between poor antitoxin producers with progressive selection for good or poor response. ( Froln I. F. Scheibel, Actn Pothol. Illicrobiol. Scnnd. 20, 464484, 1943. )
toxoid, and Scheibel suggests that the small number of generations required for such a selcctivc process indicates that there were relatively few genes segregating betwcen the good and poor responders. By using a similar approach, Biozzi et al. ( I 1 ) were able to select for thc capacity to procluce sheep rcd cell agglutiiiins in random-hred Swiss
GENETIC COSTROL O F SPECIFIC IMMUNE RESPONSES
35
mice. By the ninth generntion, thcre was a thirty-fold difference in mean heinagglutinin titer in the high- and low-responding populations. It is intcresting to notc that selection for response to sheep red cells also led to a similar segregation in the rcsponse to pigcon eiythrocytcs. In addition to these detailed genetic studies, nunierous observations with a wide variety of antigens and in several different species have shown that there is either reproducible strain variation or a heritable influence on ability to respond to particular antigens. Fink and Quinn ( 1 2 ) found marked diff ercnces in the quantitative ability of inbred strains of mice to respond to a battery of five different antigens. Ipsen ( 1 3 ) showed that there were large differences between inbred strains of mice in the dose of tetanus toxoid required to elicit the same antitoxin titer. Dineen ( 1 4 ) found a large quantitative variation in the ability of diffcrcnt inbred strains of mice to produce antisheep red ccll antibody. Carlinfanti ( 1 5 ) demonstrated in man a statistically significant correlation between thc isohcmagglutinin titer of parents and offspring. Stern, Brown, and Davidsohn ( 1 6 ) studied the production of natural antisheep red cell agglutinins in two inbred strains of mice, their F, offspring, and the two reciprocal backcrosses. The results demonstrated a quantitative, dominant genetic effect. Playfair ( 1 7 ) dctcctcd differences in the number of antisheep red cell plaque-forming cells produced in different inbred strains of mice after a single standard dose of s h c q red cells given on the seventh day of life. Analysis of the F, and backcross populations resulting from a cross between strains producing very diff erent numbers of plaque-forming cells indicated a definite genetic effect. Sobey and Adams ( 2 8 ) found a heritable factor controlling the response of mice to thc antigens of Rhixobizim meliloti and to two strains of influenza virus. Sobey, Magrath, and Reisner (19 ) have recently shown that randoin-bred mice can be selectively bred for inahility to respond to bovine serum albumin ( BSA ) . Thcir results are very similar to those of Scheibel ( l o ) ,since within five generations they were ablc to produce offspring which were over 90% unresponsive to BSA. However in a subsequent study, Hardy and Rowlcy ( 20) demonstrated that these genctically unresponsive mice were, in fact, able to respond quite vr7eIl to BSA when they were given n lower dose of antigen. Polak et ab. ( 2 1) studying hyperscmsitivity to reactive cliemicals have demonstrated a striking, antigen-specific, genetic control of contact sensitization in guiiiea pigs. Thcy showed that strain 2 guinea pigs could readily be sensitized to potassium dichromate and beryllium fluoride, but not to mercuric chloride, whereas strain 13 guinea pigs could not bc sensitized to potassium dichromate and beryllium fluoride but were
36
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
rpadily sensitized to mercuric chloride. By using Hartley guinea pigs, these workers showed that crosses bctwccn two strong reactors or two negative reactors gave rcwlts compatible with genctic control of contact sensitivity to these metals. The immune response of about 20 inbred mouse strains to influenza viruses: was: studied by Von Sengbush and Lcnnox (22) following n single injection of virus suspensions. Antibody was assayed by i t 5 inhibition of red cell agglutination. Thc mouse strains could be grouped roughly into good responders and poor responders to all influenza virus strains investigated. Crosses of several of the good respondcrs by one of the poor-responding str‘iins ( DBA/2) yielded in general good responders. However, a most interesting finding was that the mating of two lowresponder strains, DBA/2 N, yielded hybrids that were good rcsponders to FM1 and WSE viruses. (These were the only poor x poor hybrids tested.) These findings are of considerable interest as they indicate a multigenic control of the immune response of mice to influenza virus antigens. The immune responsc of domestic inbred mice to the erythrocyte antigens Ea-1“ and Ea-1” found only in wild type Mus mnsculus has been studied by Gasser (23). Domestic mice arc type Ea-lo. Gasser has shown that several domestic inbred strains differ in their ability to produce agglutinating antibodies against Ea-la and Ea-1”. Breeding studies have shown that this difference is due to a single Mendelian recessive factor linked to the agouti coat color locus in the Vth linkage group. This genetic control, designated Ir-2 by Gasser, maps very close to two minor histocompatibility loci, H-3 and H-6, in preliminary mapping studies. Thc possible significance of linkage between gcnetic controls of the immune response and loci determining histocompatibility antigens will be discussed below (Section V,C). However, it should be noted that the gene controlling the Ea-1 erythrocyte antigens is located in the XVIIIth linkage group ( 2 3 ) , and that anti-Ea-la or anti-Ea-1’’ antibody from a highresponding strain was not absorbcd by erythrocytes from a low-responding strain. Both of these findings suggest that low response to Ea-la or En-1” is not due to a sharing of antigenic determinants bctween the wild type erythrocytes and those of the low-responding domestic strain (see Section 111). All of the studies cited above have established the existence of constitutional differences in immune response to complex, niultideterminant antigens. Breeding experiments have confirmed the genetic origin of these differences, and have indicated that there are often multiple genes involved, as might be expected for the control of responses to complex
x
GENETIC CONTROL OF SPECIFIC IMMUNE RESPONSES
37
antigens. Thc operation of multiple genes is shown by the rcquircmicnt for several generations of selcctive hrecding to obtain uniform population$ of low nnd high responders. While the data do indicatc that thcre arc numerous instances of gcnetic control of specific immune responscs, they do not permit us to draw any conclusions conceiiiing the mechanism of gene action. Ill.
Analysis of the Mechanism of Gene Action
Studies with antigens of known structure or synthetic polypeptides of restricted heterogeneity have shown that, in sevcral systems, there are dominant, autosonial genes controlling specific immune responses. Many of these systems have been extensively investigated in an attempt to deteiniine the mechanism of gene action. Before describing in detail the observations and conclusions made in each of these system5 (Sections IV and V ) , we would like to present an analysis of the different levels at which genetic factors could be expected to affect the immune response to a specific antigen. In studying any genetically controlled immune response to a specific antigen, an attempt should first bc made to determine whether the ability to respond is a dominant or recessive trait. It is a general rule that an animal is immunologically tolerant to his own antigcns, and it is likely that there will be instances of genctic control in which an animal is unable to respond to a particular antigenic determinant because it c r o s reacts with his self-antigens. Since moat self-antigens are under dominant or codominant genetic control, inability to rcspond would be expectcd to be a dominant trait and ability to rcspond a recessive trait. For this rcasoii, genetic controls in which ability to respond is recessive should first be considered to be examples of “cross-tolerance.” This hypothesis can be testccl by showing that antibody to the antigenic determinant in question produced in another strain also reacts with cells or antigens of the strain which is unable to rcspond. This type of gcnctic control of response to cert,un antigens has been extensively discussed by Cinadcr (24). In the analysis of dominant genetic control of the immune response to a specific antigen, three major mechanisms of gene action must be considered. First, a gene may control antigen-specific processes that are completely unrelated to any aspect of the immune response. It is obvjous that this is the first possibility that must bc ruled out before a useful analysis of any experimental system can be carried farther. The second type of gene may control ‘I process which is ‘in integral pait of a specific immune response but is not responsible for the development of specific cells or
38
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
the structure of specific immunoglobulins. For example, a gene conccrncd with the binding of certain antigens to macrophages or macrophage components might express a certain degree of specificity for the antigen. The third and most important type of genetic factor to be considered is onc which would control the antigen receptor of a specific cell and/or the structure of specific immunoglobulins. These three types of gene action may be differentiated by the detailed study of ( i ) the specificity of thc sensitized cells and/or of the antibody produced; (ii) the results of attempts to transfer the capacity to respond with various cell populations; and (iii) the identification in such transfer experiments of the origin (donor versus recipient) of the cells responding and of the antibodies produced. IV.
Genetic Differences in Immune Response to Defined Protein Antigens
Arquilla and Finn presented some of the earliest evidence that the specificity of antibodies produced in different animals in response to the same antigen was under genetic control (25). Thcse authors used an insoluble insulin-cellulose conjugate saturated with antibodies from a reference rabbit antiserum, and then tcsted the ability of anti-insulin antibodies from strain 2 or strain 13 guinea pigs to add onto the antibody-insulin-cellulose complex. In several different experiments, strain 2 antisera contained antibodies which bound to portions of the insulin molecule not already covered by antibody from the reference rabbit antiserum. This was in marked contrast to strain 13 anti-insulin antisera, which were unable to bind to the insulin-cellulose conjugate after saturation with antibody from the reference rabbit antiserum. This result indicates that strain 2 guinea pigs produce antibodies to portions of the insulin moleculc to which strain 13 animals are unable to produce antibodies. Since it was impossible to demonstrate additional binding when either strain 2 or strain 13 antiinsulin antisera were used to saturate the insulin-cellulose conjugate prior to the addition of either strain 13 or strain 2 antisera, it was concluded that the two strains of guinea pigs produce antibodies to antigenic sites on the insulin molecule which are in close proximity. These studies can be criticized on the grounds that the assay system detects only hemolytic antibody and that a quantitative difference in the amounts of antibody of differing specificities might appear to be a qualitative difference. Despite these objections, the evidence supports the concept of genetic control of the ability to produce antibodies of different specificities against the same antigen (26). Further support for this concept was obtained through the use of modified insulin derivatives. Removal
GENETIC CONTROL OF SPECIF'IC IMMUNE RESPONSES
39
of the eight C-terminal ainino acids from the insulin B-chain resulted in a much greater loss of renctivity with strain 13 than with str,iin 2 anti-
insulin antisera. Conversely, reaction of thc N-terminal cu-miiiio groups on the A- and B-chains with fluorescein resulted in a prefercntl'il loss of reactivity with strain 2 anti-insulin antisera. These results indicate that strain 2 guinea pigs produce antibodies that react preferentially with the N-terminus of the insulin A- and R-chains, whereas strain 13 guinea pigs produce antiboclieq that react prefcrentially with the C-terminal portion of the insulin molecule (27). While these results are compatible with some heritable genetic control of the structure of antibody-combining sites, a much more detailed analysis of the mechnnism of gene action will be required before this conclusion can be supported with any force. Evidence supporting the concept that the recognition of an antigen is genetically controlled ( see Section V,B ) was obtained by Armerding and Rajewsky (28) who studied the immune responsc of rats to porcine lactic dchydrogenase ( LDH ) isoenzymes. The LDH isoenzymes are tetrameric molecules composed of two typcs ( A and B ) of iinmunologically non-cross-rencting subunits in all possiblc combinations. Outbred Sprague-Dnw ley m d Wistar rats respond equally well to LDH-B,, but differ markedly in their response to LDH-A,. Sprague-DawIcy rats respond well to LDH-A4,while Wistar rats respond poorly, only at higher doses of antigens, and fail to give a clear-cut secondary response. However, when nonresponder (Wistar) rats were primed and boosted with A-subunits coupled to B-subunits, definite secondai y responses were observed, suggesting that the gcnctic dcfect in Wistar rats is a inilure to recognize the A-subunit as antigenic, in the presence of a normal ability to produce anti-A antibody. The ability to respond to LDH-A, is inherited as a single dominant gene which is not linked to allotype marker\ very probably located on Iight chains of rat immunoglobulins (28) ( see Section V,C ) . V.
Genetic Differences in Immune Responses to Synthetic Polypeptide Antigens
A. THE IMMUNE RESPONSE TO LINEARRANDOM COPOLYMERS OF L-WAAIINO ACIDS When random copolymers of L-@-aminoacids became available to immunologists (29-32), it was soon realized that the immiinogenicity of these materials for experimental animals and man depended primarily upon their degree of complexity. Although the precise amino acid sequence of the random copolymers studied is not known, thc structural
40
HUGH 0. MCDEVITT AND BARUJ BENACERRAF
TABLE I RANDOMLINEARCOPOLYMERS L-PAMINO ACIDS“
IIESPONSE O F VARIOUS SPECIES TO
OF
Coinpositionh
Micec
Rahbit.s
0/58 0/38 0/20 20/20
8/17 60/101
20/20 17/35 10/10 -
5/12 30/4 1 9/12 7/11 4/6 59/59
Ciiinen pigs 68/181
6/21 0/18 3/7 7/24 lO/22 3/5 8/8
Man 0/30 -
0/4 :3/3
2/6 11/12
16/20
From P. Pinchuck and P. H. Maurer, in “ltegdatiori of the Antibody Respoilbe” (B. C i n d e r , ed.), Thomas, Springfield, Illinois, 1968. Sribscripts refer to molar percentage of amino acid in copolymer. Niimher of iespoiicleis/tiiin~l)er immunized.
heterogeneity of these compounds clearly depends upon the number and the relative proportions of the different amino acids which they contain. Thus polymers of a single amino acid or copolymers of two amino acids are considerably less complex than copolymers of three or four amino acids. In the evaluation of the results of immunization with these materials, which are gencrally wcak antigens even in responder animals, it should be stressed that their antigenicity has been evaluated in nearly all instances (except in man) after repeated immunization with complete Freunds adjuvant to ensure maximum responses (30, 3 2 ) . As shown by Pinchuck and Maurer ( 3 3 ) , the immunogeiiicity of thesc synthetic polypeptides for different species increases, with the number of amino acids they contain both with respect to the number of responding animals and the amounts of specific antibody produced. Thus, as shown in Table I, random copolymers of four amino acids-glutamic acid, lysine, alanine, and tyrosine (GLAT)-or of three amino acids-glutamic acid, lysine, and alanine (GLA)-arc immunogenic in the majority of the animals tested. In contrast, homopolymers of single L-amino acids are rarely antigenic. Poly-L-alanine, poly-L-glutamic acid, and poly-L-tyrosine do not induce specific immune responses (30, 32, 3 3 ) . Poly-L-lysine (PLL) and poIy-L-arginine are only immunogenic in guinea pigs possessing the PLL gene, as discussed in Section V,B. The nonantigenic homopolymers may, nevertheless, behave as excellent haptens and induce specific antibody synthesis when bound to immunogenic carriers (34, 3 5 ) . Co-
GENETIC CONTROL O F SPECIFIC IMMUNE RESPONSES
41
polymers of two L-amino acids (or of two L-amino acids with only a vcrv small proportion of a third amino acid), e.g, gliitamyl-alanine ( G A ) , glutamyl-lysine ( GL ) , glutamyl-tyrosinc ( GT), arc most intcresting compounds since they are able to induce a significant immune respoiisc only in certain individuals of a givm spccics and in some inbred strains but not in others (Table I ) (30, 31, 36, 37). Copolymers GA, GL, and GT are recognized as antigens by some but not all random-bred guinea pigs and rabbits (30, 33, 3 6 ) and by none of the random-bred or pure strain mice tested so far ( 3 8 ) . Similar observations were also made by Sinionian et al. ( 3 9 ) on the response of inbred rats to several copolymers of several amino acids. In this species, there also exist largc quantitative strain variations in immune responses to the same polypeptide. The ability of a particular strain to respond well to one antigen did not correlate with its ability to respond to a second sti-ucturally different copolymer. The pattern of response to these homopolymers and copolymers with relatively simple strrrctnrc, suggested that the individual variations observed could be explained by the presence or absence of specific genes controlling the capacity to recognize thcw respective structures as antigenic determinants ( 36, 40). Breeding experiincnts have lwcn performed with two of thesc systems: (1) the immune response of guinea pigs to PLL and to hapten conjugatcs of this homopolymer, which will be discussed in Section V,B; and ( 3 ) Pinchuck and Maurer investigated the immunogenicity for mice of a random copolymer of glutamic acid and lysine with only 5% alanine, Gj7Ll8Aj( 4 1 ) . This polypeptide was found to induce specific antibodies in only 47%of random-bred Swiss mice a s detemiined by passive hemagglutination following several courses of immunization with complete adjuvants. This copolymer was also immunogenic in the following mouse inbred strains: C3H/ HeJ, BALB/cJ, 1391J, but not in C57RL, A / J, or CBA/ J strains. (If, however, the relative content of alanine was raised from 5 to lo%, all mice including the nonresponding strains formed antibodies against GLAIo after standard himunization with this polypcptide. ) The pattern of transmission of the ability to respond to GLA, in the progeny of responder and nonresponder Swiss mice was then investigated (41) . ( 1) The mating of nonresponder parents produced consistently nonrcsponder offspring. ( 2 ) The progeny of the mating of responder mice consisted of 19 responder and 3 nonresponder offspring; when 2 of these nonresponders were, in turn, mated they produced only nonresponder offspring (Table 11). ( 3 ) Furthermore, the specific F, hybrids of responder C3H/HeJ nonresponder C57BL strains were all able to form antibodies to GLAj whcn immunized with
x
42
HUGH
0.
MC DEVITT AND BARU J BENACERRAF
TABLE I1 I M M U N E 1 t . E S P O N S E S TO COPOLYMER O F GLUTAMIC ACID,
(GL&)
BY O F F S P R I N G O F
R.ESPONDERA N D
LYSINE,A N D
~ 0 N R E S I ' O X I ) E R HIVISS
ALANINE
MICE
.4ND I N B R E D h f l C E "
1':trents
1'rogelI)b
Bot,h noiiresponder Swiss Both respoider Swiss C3H/HeJ C67BL/6J c:m x C57HL P I
0/19 19/22 6/6 O/X 29/29
From .'I I'iiichiick stid P. H. & h i r e r , .J. h'.rpL/. Alrrl. 122, 67.5, 1965. Ntimber of iespoiic~eis/ririni~ie~ immiinixed.
this antigen. These results indicate that the ability of mice to form an immune response to GLA, is governed by the presence of an autosomal dominant gene. Unfortunately no attempts have yet been made in this system to transfer the capacity to respond to GLA, from responder to nonresponder mice with cells known to be concerned with the immune response (spleen or bone marrow cells). All these experiments on the immune response to random copolymers of L-a-amino acids have concerned themselves with individual differences in the ability to respond to some of these polymers and with the genetic control of the capacity to respond to antigens with relatively restricted heterogeneity but not with the specificity of thc antibodies produced. However, even in those instances [as in the responses of mice to GLA, (41 ) or to branched copolymers, discussed in Section V,C, and of guinea pigs to hapten-PLL conjugates (35,36, 4 0 ) ] where the ability to form a significant immune response is shown to be controlled by single dominant autosomal genes, the antibodies produced are clearly heterogeneous with respect to antibody class, specificity, and affinity. It would, therefore, be of considerable importance for the interpretation of these findings if the specificity of the antibody populations produced b y individual responder animals or by responder inbred strains to some o€ these antigens could be shown to be characteristic of the individual or of the strain and if these specific properties of the antibody population could be shown to be inheritable. In an attempt to achieve this result, the specificity of the antibodies produced by inbred mice to selected copolymers was investigated by Pinchuck and Maurer in collaboration with Bcnacerraf ( 33, 4 2 ) . Thc patterns of cross-reactivity with related polypeptides displayed by antibodies produced by different mouse strains or diffcrent individual mice
GENETIC COSTROL OF SPECIFIC IhlhlUXE RESPONSES
43
to different copolymer were investigated. For instance, inbred mice were immunized with G,,,A,,T,, or GLA,, and the cross-reaction of the specific 'intibodies with GA, GT, and GL, respectively, were investigated. The pattrm of cross-rcactivity w a s found to be charactcristic of the strain (Table 111) which indicates that genetic factors are concerned with the T IRLTC 111 PRECIPITIV I:F4< T l O \ b OF I \ H R E U .1\TI-C;I.i
T411 IC
hfOLT\E s T R 4 I h IIY P E R IZ IR Il~\E
I( li)-.I1,4\1\ E-TYRO~IYE (( ;,ZT)
sER.\"
St rail1 of mice
I'ci~cciI t age -re:ictioii
C:S11/1 IeJ
(76)
aitli
t :.vr
loo~o'y4'21)'
G so.Ir0 c: ,oT L O
3)''
-1,
121
120/.J ( %)
I 00yoy802)" 40 9,5
n From P. Piiichiick :tiid P. H. ~ I a i i r e r In , "RegiiI:~tioii of the *kirtibo(Iy I ~ e s p o n ~ e " (B. Ciiinder, ctl.), Tliornas, Spiiirgfieltl, Illinois, 1968. h c
Percent of homcilogoris ant igm. hlicrograms A41)S / m l .
capacity to recognizc specific detcrminants on these molecules. These observations are analogous with the findings of Arquilla and Finn (25, 26) concerning the specificity of the antibodies produced by strain 2 and strain 13 guinea pigs to bovine insulin (discussed in Section I V ) which showed that these two strains of guinea pigs produced antibodies directed to differcnt deterniinants on the insulin molecule. If the genetic factors controlling the ability to respond, respectively, to GLA,, haptenPLL, or (T,G)-A--L and ( H,G)-A--L ( scc Section V,C) arc, indeed, concerned with the ability to recognize specific detcrminants on these niolecules, it is not surprising that unifactorial genetics have been observed only with such synthetic polypeptides of limited complexity. With more complex antigens the multiplicity of genes concerned and, therefore, of possible responses no longer permits genetic andysis.
B.
RESPONSE OF GUINEAPIGS TO POLY-I,-LYSINI: AND HAPTEN-POLY-L-LYSINE CONJUGATES
T H E Ih4h.fUNE
TO
1. Ncituie of the Response As originally olwrvcd by Kantor, Ojeda, and Benaccrraf ( 3 6 ) , approxiniately 30% of Hartley strain guinea pigs immunized with 0.001 to 1 mg. of 2,4-dinitrophenyl-pol!'-r.-lysinc ( DNP-PLL ) in complete
-kljuvatit and zaliiie
10-19
3
D S P-PLL
75-85
2
DSP-PLL RS i
2
DXP-(;L
1
"
13
13-19
1 'b (1 39-0 70) 9 1 (10 5-7 7) .5 69 (6 6-4 7) DSP-GL (100 pg./ml.) 7.71 (7 .6.i-6.78)
1 07
-
-
9 .i (20 9-3 4) 3 0,i (4 4-1 731
-
-
( 1 22-0 97)
6 84 (6 84-6 81)
DNP-G L
DSP-GL
( 1 pg.ln11.) 6.74
( 0 . 0 1 pg./ml.) 3.14 (3.72-2.36)
(7.30-3.46)
1 .?8 (1 6 1 .i6)
-
tlinitrophenyl; PLL = p ~ l y - i ~ l y s i BYA ~ ~ e ;= bovine seriim a1t)iimiii; GL = glritxniyl-Iysine. roiitrts in DN.1 f r o m experimental culttires with antigen . Tallies > 1 indicate stimrilatioti of DXA This number is the ratio, c'otiiit9 i l l D S - 1 from roiitrol cultures withorit antigen
DSP
=
ci P
3 6" cl
td
m
z
P 0
GENETIC COSTROI, OF SPECIFIC 1MMUhTE RESPONSES
45
Freund's adjuvant containing 0.5 mg./ml. Mycobacterium butyricurn ( Difco complete adjuvant ) produccd an immune response characterized by the development of delayed hypersensitivity to DNP-PLL and, at 2 to 3 weeks after immunization, b y the synthesis of high serum levels (1-2 mg./ml.) of anti-DNP-PLL antibodies, belonging to both ys- and y ,-immunoglobulin classes. Their lymph nodc cclls are able to respond to DNP-PLL in culture with increased DNA synthesis (43, 44) (Table I V ) . This reaction and delaycd hypcrscnsitivity rewtions to DNP-PLL ( $36) show specificity for the PLL carrier a s well as for the haptcn--a phenomenon which is known to cliaracterizc the response of sensitized cells to hapten-protein conjugates ( 4 5 ) . In contrast, nonresponding guinea pigs fail to bcconie delayed sensitive to DNP-PLL (35, 36) or to display in vitro evidence of cellular immunity (43, 4 4 ) (Table \') and do not produce anti-DNP antibodies detectable by double diffusion in agar gel or anaphylaxis with DNP-proteins (361, or by cquililx-ium dialysis with ,"H D~P-E-arninocaproic acid ( 3 5 ) . The two available inbred strains of guinea pigs also differ markedly in their rclsponse to this antigen; strain 2 guinea pigs show the characteristic rmpoiise to DNP-PLL, displaying both delayed sensitivity and high sei-um concentrations of specific antibodies, whereas strain 13 guinea pigs show no evidence of an immune rcsponse whcn injected with DNP-PLL in Difco compIcte adjuvant ( 4 0 ) (Table V I ) . The same guinea pigs (responders) with the ability to respond to DNP-PLL can also be sensitized by unconjugated PLL ( 3 5 ) and to conjugates of PLL with other haptens, immunologicidly unrelated to DNP, such as the benzylpenicilloyl and the p-toluenesulfonyl haptens ( 4 6 ) . Guinea pigs that do not recognize PLL as an antigen arc not able to form significant immunc responses to any hapten coupled to the homopolymer. Respondcr guinea pigs immunized with PLL in complete adjuvants develop clelayed hypersensitivity to this polymer but no detectable serum antibody ( 3 5 ) . In addition, only the identical randombred Hnrtley strain and strain 2 guinea pigs capablc of responding to PLL and to hapten conjugatcs of this polymer recognize as antigcms a copolymer of 1,-glutamic acid and ~-lysinc,G,,,,L,,,(the subscripts refer to the molar pcrcentage of the amino acids), and DNP conjugates of this copolymer, although no significant immunological cross-reaction can bc detectcd betwecn GL and DNP-PLL eithcr in the specificity of the antibodies produccd or in thc response of thc sciisitized cells i n viva or in oitro to these antigens ( 3 6 ) . Strain 13 guinea pigs do not rcspond to GL or to DNP-GL ( Tal~leVI ). Thc virtual idcntity olwrved in the response to PLL and GL, is pro1xtl)ly clxplaincd l ~ thc y c3xistcnce in CI, of c>xtc.nsivc
T.IBI,F. V E F F E C T OF h T I G E X OK THE i / l T’ifW
LY?,lPII X O n E C E L L S FROM DNp-pLL.B&k OR DKP-GL“B‘~
ISCo R I W R . 4 TlOS OF 3 H - T H Y U I ~ I S E 1.U
G E S E T I C S O S R E S P O S D E R G1’ISE.A PIGS IMMUSIZED \YITH
3
Test. aittigens
0
DSP-PLL.BS.1
so.d a y s
S o . of
animals
Tmmiuiizing aiitigen
after
DSP-PLL
DNP-PLL
or DSP-PI,L.OT’A
OVA
?
imrniiiiizatioii
(10 pg./ml.)
(1 pg./ml.j
( 1 pg./ml.)
( 1 pg./ml.1
5
4
DSP-PLL
-
DKI’-PLL.BSA
7
DTP-PI,L.OV.~
11-21
0.707 (1.47-0.438)
2
DSP-C;I,
1% 19
DSP-GL ( I 0 pg.lml.1 1.13 (1.17-1.09)
)
31
14-13 2”
0.712c (0.91-0.40) -
1.03 (1.4-0.77) 1.24 (1.26-1.22) 1.04 (1.41-0.34) DXP-GL (1 ~g./nil.‘l 1.13 (1.20-1.11)
-
-
7.2
-
4.8 (10.2-1.86)
DNP-GL (0.01 pg./ml.) 0.93 (1.37-0.50)
3 0 (5.0-2.3) -
Fium I. Green, B. B. Levine, If-. E. Paul, and B. Benacerraf, in, “Xiicleic Acids in Immruiology” (0.J. Plescia aiid W. Brawl, eds.), Springer, Berlin, 1968. D S P = diuitropheiij-1; I’LL = poly-I,-lysine; BS.1 = hovine serum a1’r)iimin; GI, = glutamyl-lysine; OVA = O V ~ L ~ ~ ) I I I I ~ ~ I I . coiiiits in DXL1from experimental cidtiires wit.h untigell . Values > 1 iiitlicate stimulntioti of D S . l This niimher is the ratio, counts in D S A from coiitrol ciilt.ures without. antigen synthesis.
cci 4
* 3
m
&C
LI
m
2!
P
n
1 ki
IMMUNE RESPONSES OF STRAINS 2 Guinea Pig strain (No. of animals)
Antigen
AND
TABLE VI 13 GUINEAPIGSTO DNP-PLL
Adjuvant (mg./ml.)
2 (4)
DPS-PLL
0 5 M. butyricirni
(4)
DNP-PLL
(4)
DNP-GL
10 hl. tubercitloszse 0 5
M . bulyrtciini (4)
DIP-GL
10
M. 13 (8)
DIP-PLL
(8)
DXP-PLL
DNP-GL
Stimulation of DNA synthesisb (10 pg. Ag/ml.)
++++ ++++ ++++ ++++
++++ ++++
hTeg.
Seg.
Keg.
Keg.
WITH
DIFFERENTADJUVANTS"
Serum anti-DNP antibodies a t 26 days" (mg./ml.)
Av. % bindingd 3H DNP-EACA
DSP-GL
0 3
8.5
2F
2 3-3 9
89
8
1 4-2 4
87
ATot nieasiirable
7
0-0 1
2
Xeg.
Not mea~rirable
0
Keg.
Sot measurable
3
M . biit?)riclrnl (8)
DNP-GL
10
M . t iiberculoszs a
D N P = dinitrophenyl; PLL = poly-zilysine; GL = glutamyl-lysine. Stimulation of incorporation of 3H-thymidine into DNA by lymph node cells in culture. Precipitin analysis with DSP-bovine fibrinogen. Binding of 0.1 ml. 10-8 M 3H DKP-eaminocaproic acid (EACA) by 0.1 ml. serum. M yobacterzitm tuberciilosis H37Rv strain.
8
0 82-0 94
M. tiibercirloszs (8)
m
88
Oittilrzcirm
10
cl
1 29-2 1
tic brrculosis
0 3
M.
Del. sensitivity to 10 pg. Ag
AND
48
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
PLL sequences. In contrast with the identical behavior of GL and PLL, in PLL responder guinea pigs, immunization of Hartley strain guinea pigs with a copolymer of L-glutamic acid and L-alanine, not containing lysine, and with DNP-PLL shows that these two compounds can be immunogenic in different individual animals ( 33, 43, 4 8 ) . In all the experiments described above, immunization was carried out in complete Freund’s adjuvant with low doses of Mycobacterium butyricum or of Mycobacteriurn tubercdosis H37Ra, 0.5 mg./ml. (commercial Difco adjuvant). Under these conditions the differences between responder guinea pigs (Hartley and strain 2 ) and nonresponder guinea pigs (Hartley and strain 13) immunized with DNP-PLL and especially with DNP-GL are striking, and the responses can be considered to be all or none (49). Nonresponder animals show no evidence of cellular immunity or of synthesis of specific antibodies (Table VI). If, however, the dose of mycobacteria in the adjuvant is increased to 10 mg./ml. and/or a more powerful adjuvant is used ( M . tuberczcZosis H37Rv strain), nonresponder guinea pigs, both Hartley and strain 13, at 3 to 4 weeks after immunization with DNP-PLL, produce very low levels of p anti-DNP antibodies, generally well below 0.1 mg./ml., but still do not show delayed sensitivity to DNP-PLL or other in vitro evidence of cellular hypcrsensitivity ( 50). The response to DNP-GL, a nearly neutral polypeptide, after immunization with the higher dose of mycobacteria is even weaker -strain 13 does not form significant levels of anti-DNP antibodies against DNP-GL, and HartIey strain nonresponder guinea pigs synthesize lower levels of anti-DNP antibodies than after immunization with DNP-PLL. In summary, if higher doses of mycobacteria or more powerful adjuvants are used to immunize with DNP-PLL the differences observed between responder and nonresponder guinea pigs remain qualitative with respect to their capacity to display cellular sensitivity but become quantitative ( although marked, generally over 20-fold) with respect to their ability to form anti-DNP-PLL antibodies ( 50). A study of the response of guinea pigs to the corresponding D-polypeptides with the two types of adjuvant is very informative in this respect. Strain 2 ( 5 1 ) and Hartley strain guinea pigs, irrespective of their PLL responder status, do not become sensitized to DNP-poly-D-lysine (PDL) (35) or to DNP-D-GL and do not form detectable levels of specific antibodies following immunization with 0.5 mg./ ml. M Y C O ~ U C terium butyricum as adjuvant. However, the production of low levels of anti-DNP antibodies, but no cellular immunity, is observed after immunization with adjuvant containing 10 mg./ml. Mycobacterium tuberculosis H37Rv ( 5 2 ) .Thus the response of random-bred guinea pigs to the
GENETIC CONTROL O F SPECIFIC I M M U N E RESPONSES
49
DNP conjugates of the D-polypeptides is analogous to that of the PLL nonresponder guinea pigs to DNP-PLL or DNP-GL. As shown below (Scction V,B,7), increasing the adjuvant may be interpreted as exerting a “schlepper” effect (35, 53) in nonresponder guinea pigs where the DNP-PLL or DNP-GL molecules may behave as haptens. 2. Breeding Experiments Mating experiments with random-bred Hartley strain guinea pigs have permitted an analysis to be made of the genetic control of the response to DNP-PLL. The study of the offspring from the matings of responders, of nonrespondcrs, and of the F, generation from strain 2 and strain 13 guinea pigs shows that the ability to respond to DNP-PLL and, therefore, to GL and DNP-GL is inherited as an autosomal dominant which has been designated as the PLL gene ( 4 7 ) . (These experiments were carried out by immunization with 0.5 mg./ ml. Mycobacterium butyricum as adjuvant, under conditions where nonresponder animals show no evidence of an iinniune response.) This conclusion is based on the following evidence from Levine and Benacerraf (40, 5 4 ) (Tables VII and VIII): (1) the mating of nonresponder parents produced nonresponder offspring; ( 2 ) the progeny from the mating of responder parents consisted of 82%responder guinea pigs; ( 3 ) crosses between nonresponders and guinea pigs heterozygous for the responder trait resulted in a progeny consisting of nearly 50% responder animals ( S 4 ) (in this experiment, heterozygosity was established by the capacity of responder guinea pigs to yield both responder and nonresponder offspring); (4) the F, generation of the mating of strain 2 and strain 13 guinea pigs all TABLE VII A\NTIGENICiTY OF D N P 2 r - P L L , , 6
IN OFFSPRING O F
RESPONUER PARENTS
A N D NONRESPONDER P A R E N T S “ , ‘
50
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
TABLE VIII OF OFFSPRING OF THE MATING(NONRESPONDERS X HETEROZYGOUS PERCENTAGE RESPONDERS) WHO ARE IMMUNERESPONDERS TO HAPTEN-POLYLYSINE CONJIJGATESG.~ OfTspritig Family SO.
I:espoiitlers
Nonresponders
I I1 I11 IV '1' 1'1
VII VIII IX X Total a
14 (45 3%)
17 (54.7%)
From I). B. Leviiie and B. Benacerraf, Science 147,617 (1965).
* The offspring were immiinized with 0.1 mg. of DNPZJ-PLL3la in complete adjuvant.
Itesponders gave positive allergic reactions to the iinmuriiziiig antigen and their serums showed antibodies to DNP. Nonresponders showed no evidence of an immune response t u DNP-PLL. (DSP = clinitrophenyl; PLL = poly-r.-lysine.)
rcspondcd to DNP-PLL as do strain 2 guinea pigs. No evidence of sex linkage was noted in any of these experiments. 3. Studies on the Metabolism of DNP-PLL in Responder and Nonresponder Guinea Pigs
Considering that DNP conjugates of polypeptides of D-lysine are not or only marginally immunogenic, even in guinea pigs capable of responding to DNP-PLL (51, 5 2 ) , and that these polypeptides are poorly degraded ( 5 5 ) , the possibility that nonresponder guinea pigs lack a proteolytic enzyme able to degrade PLL was investigated. Lymph node macrophages from both responder and nonresponder guinea pigs were shown to phagocytize DNP-PLL to the same degree (56). The L polymer, in contrast to DNP-PDL was degraded equally well an vivo into small fragments excreted in the urine of responder and nonresponder guinea pigs. In addition, spleen extracts from both respondcr and nonresponder guinea pigs degraded fluorescein conjugates of PLL in vitro (55).
51
GENETIC CONTROL O F SPECIFIC IMMUNE RESPONSES
4. hlinimum Size of PLL Antigen
Schlossman and associates investigated the ability of DNP conjugates of L-oligolysines of various sizes to induce characteristic immune responses in guinea pigs with the PLL gene (57). Their earlier experiments showed that the smallest molecule consistently immunogenic was a,N-DNP-octalysine. Strain 2 (Table IX) and Hartley responder guinea pigs immunized with a,N-DNP-octalysine or a,N-DNP-nonalysine developed delayed sensitivity to these antigens and synthesized significant levels of anti-DNP antibodies of which the binding energy was almost totally directed to the tetra- or hexalysine conjugate (57, 58). In spite of this, a,N-DNP-hexalysine and lower polymers were not immunogenic and were not able to elicit delayed hypersensitivity reactions in guinea pigs immunized with the higher polymers (59). Guinea pigs sensitized with a,N-DNP-octa- or nonalysine responded to these antigens in vitro with increased DNA synthesis, but the reaction could be neither elicited nor inhibited by larger concentrations of a,N-DNP-hexalysine (60). These experiments demonstrate that a molecule with a minimum of eight lysyl residues is required both to be recognized as an antigen by guinea pigs with the PLL gene and to elicit the response of sensitized TABLE I X RESPONSEOF STRAINS2 AND 13 GUINEAPIGSTO INJECTION OF CY.N-DNP-OLICOLYSINES~ Skin reaction (No. of animals)
Antigens Strain 2 ~~,N-DNP-oligo-r,lysitie (W = 8.4) or,N-DNP-riona-L-lysine CY,N-D N P-oc t a-L-1ysine LY, N-DNP-hept a-L-lysirie CY, N-D NP-hexa-L-lysine a,N-DNP-penta-L-lysirie a,N-DNP-tetrs-L-lysiiie ~~,l\i-DNP-poly-~-lysilie ( A = 80) Strain 13 a,N-DNP-oligo-I,-lysirie (fi = 8.4)
Tested
Iriimediate (3-6 hr. av.)
30 0 9
0 0
10 ) )
Delayed (24 hr. av.)
0 0 0 0 0 0 0 0
0
0 0 0 0
5
o
3
0
Passive cutaneous anaphylaxis
0
0
52
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
cells, in spite of the fact that the serum antibodies produced in response to these antigens bind equally well a,N-DNP-hexalysine. The studies of the immunogenicity of DNP-oligolysines of various sizes, with the DNP determinant on a specific c-amino group, and of the specificity of the immune responses observed could provide considerable information concerning the process controlled by the PLL gene. For this purpose c,N-DNP-oligolysines of various length with the DNP hapten conjugated either to the N-terminal or the C-terminal lysine were synthesized by Paul et al. (61) using the Merryfield technique. The observations of Schlossman et al. were confirmed. Only r-DNP-oligolysines that contained at least eight or nine lysyl residues induced delayed hypersensitivity and were able to elicit delayed reactions and to stimulate increased DNA synthesis by specific lymph node cells in &To. Extensive cross-reactions were observed in these experiments between 1 e,N-DNPnonalysine, 9 c,N-DNP-nonalysine, and also DNP-PLL ( 61 ) . 5. The Response of Guinea Pigs with the PLL Gene to Protamine, DNP-Protamine, and Poly-L-Arginine
The gene in guinea pigs controlling the response to PLL and GL was found to determine also their ability to respond to protarnine, DNPprotamine, and poly-L-arginine. Protamine, a small positively charged protein with a very high arginine content and no lysine, and DNPprotamine were shown by Green, Paul, and Benacerraf ( 6 2 ) to be able to induce specific immune responses only in guinea pigs with the PLL gene. Only strain 2 and Hartley strain guinea pigs whose responder genetic status had been verified by their ability to respond to immunologically unrelated GL were able to develop delayed sensitivity to DNP-protamine and to form anti-DNP antibodies (about 0.3 mg./ml.) following immunization with DNP-protamine in complete Freunds adjuvant containing 0.5 mg./ml. Mycobacterium butyricum. Strain 13 and PLL nonresponder Hartley guinea pigs did not show these responses. Similar results were obtained with unconjugated protarnine. In preliminary experiments strain 2 but not strain 13 developed delayed sensitivity to poly-L-arginine and to DNP-poly-L-arginine following immunization with these antigens in complete adjuvants. 6. Cell Transfer Experiments
As discussed in Section 111, cell transfer studies are essential for an adequate analysis of gene action in these systems. Attempts must therefore be made to transfer to nonresponder guinea pigs the capacity to respond to DNP-PLL or GL with lymphoid cells or macrophages from
GENETIC CONTROL OF SPECIFIC IMMSJNE RESPONSES
53
animals with the PLL gene. Transfer of the capacity to respond to DNPPLL to lethally irradiated nonresponder random-bred Hartley guinea pigs has been successfully carried out by Foerster et al. with bone marrow cells from responder Hartley guinea pigs ( 6 3 ) .The genetic responder status of the animals had been determined by previous immunization with GL. A major di5culty with these experiments was the high mortality (over 70%)from graft versus host reactions in the recipient animals. Nevertheless, of the 15 guinea pigs which survived the transfers and were immunized with DNP-PLL in complete adjuvants, 13 developed delayed sensitivity to DNP-PLL. The lymph node cells of 9 of these animals responded to this antigen in vitro with increased DNA synthesis, and 10 recipient guinea pigs also produced low concentrations of antiDNP antibodies. As control for these experiments, bone marrow cells from 5 nonresponder guinea pigs were transferred to 5 irradiated nonresponder animals. None of these animals developed delayed sensitivity, cellular immunity, or antibodies to DNP-PLL when immunized with this antigen in complete Freund's adjuvant. A more appropriate experimental system involves the transfer of immunocompetent cells from strain 2 X strain 13 F, animals into lethally irradiated stain 13 guinea pigs. Accordingly, the process controlled by the PLL gene has been successfully transferred by lymph node and spleen cells from strain 2 x strain 13 F, animals to lethally irradiated strain 13 animals whose bone marrow had been reconstituted by strain 13 bone marrow. Thirteen such transfer experiments were performed and 10 of the 13 recipient strain 13 guinea pigs gained the capacity to form an immune response to DNP-PLL or to DNP-GL following the transfer of F, lymph node and spleen cells. The cell populations used contained about 95%cells with the appearance of small or large lymphocytes and 5%cells staining vitally with neutral red (Green and Benacerraf, unpublished data). These experiments demonstrate that the capacity to recognize PLL conjugates is expressed in the cells of the bone marrow and of lymph nodes and spleens of responder guinea pigs, but the cell type involved has not yet been identified.
7. The Behavior of DNP-PLL As a Complete Antigen in Guinea Pigs with the PLL Gene and As a Hapten in Nonrespondm Guinea Pigs Nonresponder guinea pigs lacking the PLL gene do not develop cellular immunity (delayed sensitivity) to DNP-PLL and form very low concentrations of specific antibodies only when high levels of mycobacteria are used as adjuvant. However, PLL is a highly charged molecule and can form stable aggregates with negatively charged foreign albumins
54
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
IIESPONSE
TABLE X NORMAL ANI) BSA-TOLERANT NONRESPONUER GI11 N E A DNP-PLL.BSA I N COMPLETE FREUND'S ADJUVANT^^^
OF
Normal
P I G S TO
BSA-ToleraritC
Anti-DNP Coric. (mg./ml. serum)
No.
Anti-DNP Conc. (mg./ml. serum)
2 3 4 5 6 7 8 9 10 11
1.11 0.82 2.46 1.78 1.88 2.23 1.26 1.97 1.67 1.85 1.25
12 13 14 15 16 17 18 19 20 21 22
0.55 0.56 0.02 0.21 0.16 0.25 0.13 0.34 0.16 0.16 0.30
Av.
1.66
Av.
0.26
NO.
1
From G. A. Theis, I. Green, B. Benacerraf, and G. W. Siskind, J . Immunol. 102, 513 (1969). b Guinea pigs immunized with 100 pg. of DNP-PLL.BSA emulsified in complete Freund's adjuvant with 0.5 mg./ml. Myobactwium butyricum. Animals were skin tested a t 7 clays af!.er immunization with 10 pg. DNP-PLL and bled to death at 14 days after immuriixation. Antibody concentration was determined by precipitin reaction with DNP bovine fibrinogen. (DNP = dinitrophenyl; PLL = poly-lilysine; BSA = bovine serum albumin.) c Tolerance to BSA was induced by four injections of 1 mg. BSA dissolved in PBS (phosphate buffered saline) given over a 2-week period ending 7 days before immunization with DNP-PLLsBSA. Q
where the albumin can act as an immunogenic carrier. By this device, nonresponder guinea pigs immunized with complexes of DNP-PLL with ovalbumin or BSA, in complete adjuvants with 0.5 mg./ml. Mycobmterium butyricum, can be induced to form high concentrations of antiDNP-PLL antibodies (1-2 mg./ml., Table X ) in the absence of delayed sensitivity or in vitro evidence of cellular immunity to DNP-PLL ( 3 5 , 4 3 , 6 4 ) (Table V). An immune response to the carrier albumin is essential for the synthesis of high levels of anti-DNP-PLL antibodies in spite of the fact that anti-DNP-PLL antibodies and anticarrier albumin antibodies are synthesized in different plasma cells ( 6 5 ) . Thus, the establishment of tolerance to the carrier albumin markedly decreased the ability of anti-DNP-
GENETIC CONTROL O F SPECIFIC IhIMUNE RESPONSES
!5S
PLL albumin complexes to stimulate the formation of anti-DNP-PLL antibodies (43, 6 4 ) (Tablc X ) . The immunological specificity of the an ti-DNP-PLL antibodies synthesized by these nonresponder nnin~alswas coniparcd with that of the antibodies produced by responder guinea pigs immunized with DNPPLL alone, and was shown to be identical, with most of the binding affinity directed against DNP-lysine and partial specificity (less than 1 kcal./mole) directed against the PLL molecule ( 3 5 ) . The ability of an immunogenic molecule to act as a “carrier” for DNP-PLL in guinea pigs lacking the PLL gene may explain in part why nonresponder guinea pigs immunized with this antigen and high doses of mycobacteria were able to synthesize low serum levels of anti-DNY antibodies ( 5 0 ) . The inycobacteria may have acted as a carrier in a manner similar to the foreign albumin. The much weaker response observed with DNP-GL, a weakly charged molecule, is compatible with this interpretation (49, 50).
8. Analysis of the Process Controlled by the PLL Gene a. The PLL gene is concerned with a process which is an integral part of the immune response, sincc the capacity to respond to DNP-PLL can be transferred by bone marrow cells from responder guinea pigs, that is, by cells or precursors of cells directly involved in the immune response (63). b. The most challenging finding in this and in other genetic systems is the observation that a single dominant genetic factor controls the immune response to a family of antigens which have structural features in common (in this case sequences of positively charged amino acids, lysine or arginines). The induction of the immune response in the PLL system as well as the stimulation of the response of sensitized cells in vim or in vitro requires the identical immunogenic antigen with eight or more lysyl residues (57-60). (Blocking of all the €-amino groups of DNP-PLL renders the molecule nonantigenic in responder animals.) Furthermore, in this system as in other systems controlled by single dominant genes, the immune response observed is heterogcneous both as to immunoglobulin class and as to specificity and affinityof the antibodies produced. This is particularly clear in the case of anti-DNP-PLL antibodies ( 3 5 ) . c. The immunogenicity requirements for the response to DNP-PLL ( a minimum of eight lysyl residues) and the specificity of cellular immunity to this antigen arc similar (59-61) but differ from the specificity of the anti-DNP-PLL antibodies produced ( 5 8 ) , suggesting either that the inimune receptors on the antigen-sensitive cells and specific serum
56
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
antibodies later produced do not have identical specificities or, more probably, that the DNP-PLL antigen undergoes interaction at two levels of specificity, one of which is under the control of the PLL gene. d. In favor of this last interpretation is the observation that immune responses to haptens in hapten-protein systems require previous or simultaneous responses to the carrier molecule and that tolerance to carrier molecules abolishes the responses to haptens they bear (24, 43, 64, 66). According to this view, immune responses to PLL, GL, polyarginine, or to hapten conjugates of these molecules would require an identical interaction of the antigen with a molecule of which the synthesis is controlled by the PLL gene and which is specific for the sequences of charged amino acids that characterize these antigens, before specific antibodies can be synthesized against other determinants they carry. It is conceivable, in addition, that such an interaction can determine which determinants are recognized. An analysis of where such an interaction takes place await the precise determination of the type of cells capable of transferring responder status.
MULTICHAINAMINO ACID C. THE RESPONSEOF MICE TO BRANCHED, COPOLYMERS The ability of inbred strains of mice to make antibodies to branched, multichain, synthetic polypeptide antigens bearing a restricted range of antigenic determinants is a quantitative genetic trait (67). This trait appears to be controlled to a large extent by an autosomal dominant gene, which has been designated immune response-1 ( Ir-1) . A schematic structural diagram of the type of antigen used in these studies is shown in Fig. 2. These antigens are synthesized by polymerizing
j4olylysine
--l-rT-
Po Iy - D, L- a I a n in e
- Poly (tyrosine,
glutornic ocid)
FIG. 2. A schcniatic tliagrarii of the structural pattern of (T,G)-A--L 509. (From H. 0. McDevitt and M. Sela, J. Exptl. hled. 122, 517531, 1965.)
GENETIC CONTROL OF SPECIFIC IM M U N E RESPONSES
57
side chains of poly-D,L-alanine on the E-amino groups of a long backbone of PLL. This results in a branched, multichain polypeptide, polyalanyl-polylysine (A-L), which obeys thc general rulcl for linear amino acid polymers that copolymers of one or two amino acids arc not immunogenic in mice (38). The addition of short, random sequences of tyrosinc and glutamic acid to the amino termini of the poly-o,L-alanine side chains converts A--L into (T,C)-A-L, a good antigen in some strains of mice. It should be noted that the sequences of tyrosine and glutamic acid at the tips of the side chains are random and not identical on every side chain. Substitution of histidine or phenylalanine for tyrosine in the sidechain termini produces (H,G)-A--L or (Phe,G)-A--L, closely related antigens in which the nature of the antigenic determinant is varied on a similar structuraI background. When CBA and C57 mice are immunized with the maximal immunizing dose of (T,G)-A--L. CBA's respond poorly while C57's respond well. Exactly the opposite result is found with (H,G)-A--L: CBA mice respond well, whereas C57 mice respond poorly. The F, hybrid between CBA and C57 responds well to both antigens, and reciprocal backcross progeny segregate as a 1:1 mixture of the F, and the respective homozygous parent animals (67, 68). Figure 3 illustrates the genetic segregation in antibody response to (T,G)-A--L. The horizontal axis plots the antibody response of individual mice in tcrnis of the percent of antigen bound in an antigcn-binding assay, The vertical axis represents the number of animals falling into a given percentile of percent antigcnbound values. Since a wide variety of control experiments have excluded dose-response differences (67) or different responses due to adjuvant, age, or sex, these results indicate that the ability of mice to respond to ( T,G)-A--L is a genetically controlled, quantitative, dominant trait. Since substitution of histidine for tyrosine in the antigenic determinant results in a reversal of the high- and low-responding strains, as noted above, this genetic trait, Ir-1, may be concerned with the recognition of the antigenic determinants. Preliminary analysis of the Ir-1 gene showed that there was no significant quantitative difference in distribution among the immunoglobulin classes of the antibodies produced to (T,G)-A--L in high- and lowresponding strains (69) and that in a segregating backcross population [ (CBA x C57) F, CBA], no linkage was found between Ir-1 and the Tg region coding for the mouse immunoglobulin allotypes. This indicates that the Ir-I gene is not associated with the known structural genes coding for the F, fragments of mouse immunoglobulin heavy chains (69). Further analysis of the mechanism of action of the Ir-1 locus has been
x
58
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
81
8l
0
c 57
CBA
CBA x C57 FI
25
81
50
75
I00
CBA x C 5 7 FI x C 5 7
4
0
25
50
75
100
% ANTIGEN BOUND
FIG. 3. Immune response of mice given 10 pg. (T,G)-A--L 509 in complete Freund's adjuvant, and boosted with 10 p g . of the same antigen in saline. (From H. 0. McDevitt and M. Sela, J. Erptl. Med. 122, 517-53'1, 1965.)
carried out, as indicated in Section 111, by the use of cell transfer studies, linkage studies, and studies of the specificity of the antibodies produced.
1. Cell Transfer Studies Since preliminary analysis (69) failed to show any association of the Ir-1 gene with the level of immunoglobulin class or with immunoglobulin allotype, cell transfer studies were undertaken to find out whether the ability to respond well to a particular synthetic polypeptide antigen was a trait closely related to the process of antibody formation and, therefore, transferable with immunocompetent cells. McDevitt and Tyan ( 7 0 ) found that it is possible to transfer the ability to respond well to (T,G)-A--L from (C57BL/6 x C3H) F, responder animals into lethally irradiated, nonresponder C3H parental recipients by the transfer of 100 to 150 million adult unfractionated
GENETIC CONTROL OF SPECIFIC IMMUNE RESPONSES
59
TABLE XI ADOPTIVE TRZTWER O F i PRIMARY RESPONSETO (T,G)-A--L TO LETHALLY IRRADIATED C3H HOSTSW I T I I IC3H X C37BL/6) PI SPLEENCELLP~
Recipiei it3 (No. of ariimdsj
Primary stimiiliisc (lay, after radiatioii
Thymectomized C3H 9 (7) Normal C3H 0 (7) Normal C3H 61 (9) Thymectomized C3H 9 (5) n’orxnal (C3H X C57BL/6) FI 0 ( 5 ) Sormal (CSII X C37BL/6) 171 0 (4)
0 0 21 21 21 0
Summary:
Titer 10 days after iecoiidary stimiiluid 52, 19, 10, 58, 0, 0, 65 ( 1/50) 13, 60, 41, 51, 50, 8, 54 (1/50) 39, 8, 49, 45, 0, 63, 52, 50, 73 (1/50) 46, 17, 54, 72, 26 ~ 5 0 ) 76, 59, 58, 50, 47 (1/50) 69, 76, 71, 6.5 (1/.50)
F,+ C3H
=
FI C31I + C3II
= =
171 --t
38% (19/2X) 63% (9/9) 8%
From IT. 0. McDevitt and M. L. Tyaii, J. Esptl. dled. 128, 1-11 (1068). &Spleen cell dose in all these transfers wab 100-150 X lo6 (C3H X C57BL/6) F, spleen cells per iecipient. (T = tyrosine; G = glr1tltmlc acid; A = polyalanyl; L = polylysine.) c First stimulus was 10 pg. (T,G)-A-L 309 iri complete Freurd’s adjuvant; second stimulus was 10 pg. (T,G)-.4-L-309 i i i aqueous solution given 3 weeks later. d Titer is average percent antlgeil I,ourid. Figures in parerilheses indicate dilution at which antisera were titered.
spleen cells. These results are summarized in Table XI. This transfer was carried out using normal, nonimmunized spleen cells and was successful in 19 of 28 attempts whether the recipients were immunized immediately after transfer or 3 weeks after transfer and whether or not the recipient animals were thymectomized. These results strongly support, but do not prove, the conclusion that this genetic control (Ir-1) is directly related to the process of antibody formation. The possibility still rcmains that the successful transfer of any type of cell from a responder into a nonresponder strain (e.g., liver cells and skin cells) would result in the transfer of an cnzyme system required for proper metabolism of the antigen. Conclusive support for the hypothesis that this genetic control is an integral part of the process of antibody formation was obtained from another type of cell transfer experiment. In these experiments, thymectomized, lethally irradiated (C57BL/6 x DBA/2) F, mice [who normally respond well to (T,G)A-L] were given fetal liver cells and a thymus implant from CBA (nonresponder) fetuses and then immunized with (T,G)-A--L at 60 and 100
60
HUGH 0. MC DEVITT AND BARUJ BENACERRAF
days after irradiation and cell transfer. All the chimeras failed to respond to (T,G)-A--L, but 12 of 13 subsequently responded well to (H,G)-A--L, to which the CBA strain normally responds well (71 ). This experiment indicates that the genetic control of the ability to respond to (T,G)-A--L and (H,G)-A--L is exerted through a mechanism directly related to the immune response, since thymectomized, irradiated mice which were responders to (T,G)-A--L and nonresponders to (H,G)-A--Lcould be converted into the opposite phenotype by the transfer of fetal liver and thymus from a strain of the opposite phenotype ( 71 ). Preliminary evidence (M. L. Tyan and H. 0. McDevitt, unpublished data) indicates that it is possible to transfer the ability to respond well to (T,G)-A--L by the transfer of partially purified peripheral blood lymphocytes, although efforts to accomplish this transfer with thoracic duct lymphocytes ( 71 ) have been unsuccessful. Further experiments will be required before the site of action of the Ir-1 gene can be definitely localized to a particular cell type. In summary, the evidence indicates that the Ir-1 gene acts in a cell type directly involved in the process of antibody formation, now tentatively identified as a peripheral blood lymphocyte. 2. The Linkage of Ir-1 with the Major Histocompatil~ility ( H - 2 ) Locus This linkage was first suspected when it was found that C3H.SW mice responded well to (T,G)-A--L,although they are genetically identical with C3H/DiSn mice ( a known low-responding strain), except for the H-2 locus-C3H.SW being H-2” and C3H mice being H-2k. Similar anomalous results were found with B1O.BR mice ( H-2k) which are congenic with C57BL/10 ScSn mice (H-2”). The B1O.BR mice respond poorly to (T,G)-A--L, although C57BL/10 mice respond well to this antigen, These anomalous results suggested that the ability to respond well or poorly to (T,G)-A--L had been bred into these strains by the breeding process which put a different H-2 allele on the same genetic background. This suggestion implies that ability to respond (Ir-1) is linked to the H-2 locus. Linkage was established in backcross tests in which it was shown that the ability to respond well to (T,G)-A--L is linked to the H-2h allele, whereas the ability to respond well to (H,G)-A--Lis linked to the H-2“ allele ( 7 0 ) .Although three recombinant animals were tentatively identified in the initial backcross test, only one has been fertile and this animal was not, in fact, a recombinant. Further testing of backcross populations is continuing, but at present there is no conclusive evidence that Ir-1 and H-2 are not identical. Extensive testing has shown that there is a regular correlation be-
TABLE XI1 THERELATIONSHIP BETWEEN H-2 TYPEA N D IMMUNE RESPONSETO
THE
A-L SERIESOF ANTIGENS" ~~
(T,G)-A-L
Strain A/J A.BY c57 D1.LP C3H.SW BALB/c DBA/2 CBAb C3H/HeJ B1O.BR AKR DBA/lC SJL A.SW WB/Re SWR
Antigenbound
(H,G)-A-L
(%)
Range
No. of animals
k
10 78 69 59 79 28 34 12 17 6
-
5-15 62-87 53-82 40-95 52-91 0-55 11-53 0-27 9-26 2-14
-
9 9 10 10 10 8 10 10 10 10
-
9
6
S
5
4-12 3-7
S
0 0 0
8 10 6 10 10
H-2 Type
a b b b
b d d
k k k
W
?
Antigenbound
(Phe,G)-A-L
KO.of
Antigenbound
( %)
Range
animals
(%I
Range
No. of animals
77 0
61-83
FIG. 6. Formation of mi~ltipleholes (clusters ) on sheep erythrocytes by rabbit IgG anti-Forssinan antibody and guinea pig C'. The experimental design was similar to that in Fig. 5, except that pnrified IgG anti-Forssman antibody was employed, and the final guinea pig C' coccentration was 1/67, Several multiples of the 50%heinolytic amount of antibody were used. This figure illustrates how, at low ratios of anti-
c',
some nienibrane fragments have inany more holes than would be expected body to if one lesion corresponded to one hole. (Hatched Mocks-plates with no holes; black blocks-plates with the stated ntmilier of holes per square micron. )
of lesions per whole eiythrocyte membrane (26 p ? ) would be about 0.7, averagcd ovcr all thc cells (h4aycr, 1961b). It is striking, espccially at the lowcr multipIicitics of antibody, th'it some membrane fragments were seen which contained far more than the expectcd number of holcs. The distribution shown in this figure does not, however, give a true picture of thc average number of holcs per wholc erythrocyte, since the memb r m c fragmcnts with no or few holes werc generally larger than those
94
TOHN 13. HUMPHREY AND ROBERT R . DOURhlASHKIN
with many holes. Nevertheless, the skew distribution of holes indicates that some occurred in clusters, and that clusters were more frequent as the amount of antibody decreased while the concentration of C’ was held constant. It was not clear whether cluster formation with IgG antibody was duc to the W R Y in which C’ was fixed or to local movement of antibody on the cell surfacc. (Although the antibody used was avid, evidence was obtained that transfer from cell to cell could occur to a small extent when erythrocytes to which this IgG antibody was attached, and which wcre then washed, wcre incubated with untreated erythrocytes). In other experiments reported by Rosse et at. (1966), erythrocytes from patients with PNH and from nonnaI human donors were lysed with human TgM (anti-I) antibody and large amounts of human C’. Again, electron-microscopic examination of the membrane fragments revealed a mean number of holes per cell very much higher than that expected from the extent of hemolysis. Since there is no reason to believe that the onehit theory does not apply to human cells and C’, these observations are not compatibIe with the hypothesis that a single lesion is always represented by a single hole. They might, however, be explained in one of two ways: either the terminal stage of human C’ activation at a single site generates a sufficient amount of some surface-active agent to cause extensive local micelle formation in the membrane (see Section VIII) or there is multiple binding of C’3 at a single site, as has been indicated b y Mullcr-Eberhard et al. (1966), and a variable number of the bound C’3 molecules activate the succeeding C’ components. We have attempted to tcst the second posibility by comparing the lysis of human erythrocytes by huinan C’ and human anti-I antibody [prepared as described by Rosse et al. (1966), but from a less hemolytically active serum] under conditions in which one or the other reagent was limiting. It was nece5sary to increase the sensitivity of the erythrocytes to lysis b y human C’ treatment with 2-aminoethylisothiouronium bromide (AET), which renders normal cells as sensitive to C’ as PNH cells (Sirchia and Dacie, 1967). Thc experimental results in Table I1 indicate that when C’ was limiting thcre were few clusters (not exceeding those due to the action 011 these cells of human C’ alone, to which they are abnormally sensitive) and that the proportion of membranes with single holes increased with the degree of hemolysis. However, when antibody was limiting and c’ in excess, hemolysis was accompanied by the presencc of more meinliranes with clusters than with single holes. Frank, Dourmashkin, and Humphrey ( unpublished observations ) havc also txamincd the lesions produced by human C’ on sheep erythrocytes sensitized with rabbit anti-Forssman Igh4 antibody. When antibody WRS limiting and C’ prcsent in cxccss the membmnes contained clustcrs
95
THE LESIONS I N CELL MEMBRANES CAUSED BY COMPLEMENT
IiYhIi
OF
T l R L E I1 dET TREATED HVMAYERYTHROCYTE5 B Y A \ D B Y H ~ U AAYTI-I L ANTI BODY'^
IIUMAN
c’
S O .of membiaiie\ o r fragment\ with Dilutioii of human anti-I 1/ 1/9600 1/4X00 1/2400 l/l200 1/600 1/300 1/1>0 I/ m 1/9600 1/4XOO 1/2400 1/1200 1/600 1/300 l/l>O a
Dilutioiiof
Hemolysih
SO
Siiigle
hiimnn C’
( %)
holes
holes
Clusters
1/YO 1/ : 3 0 1/:10 1/30
6 ;i 20 34 50 64 7;.
77
4
-
-
2 -
1/30 1/30 1/30 1/30 1/60 1/60 1/60 1/60
1/60 1/60 1 /60 1/60
8:) X,.
4.5 3 X 10 3 19 28 .i 44 .54 63
-
-
-
-
-
-
-
-
-
48 59
9
2
17
2
Hiimaii eryt~hrorytesmade seirsitive to C’ by treatment, wit.h ANT were incubated
at, 37°C. with varying amoiuits of purified hiinian IgM antibody at. different, concerit.ratioiis of hiininn C’. After iiicxibat,ioii any intact, cells were lysed osmotically, and the
washed memhraiies and membrane fragments were examined for the presence of holes either singly or iu chisters.
iii
the electroii microscope
of holes, whereas when the degree of lysis was limited by the C’ concentration, clusters were not observed and the nuniber of holes was approximately that expected from the number of lcsions predicted from the onehit theory, Thus the formation of multiple holes in clusters appears ( a ) with human C’ in cxcess, and lysis is limited by the amount of human or rabbit IgM antibody, and ( b ) with guinea pig C’, when this is in excess and lysis is limited by the aiiiount of rabbit IgG, but not IgM antibody. Although these findings might be taken to support the idea that clusters of holes are formed by multiple activation of C’3 at a single site, they certainly do not prove its correctncss. VI.
The Number of Antibody Molecules Required to Produce a Lesion
Humphrey and Dourmashkin ( 1965) used purified rabbit IgG and IgM antibodies against the Forssman antigen to correlate directly the
96
JOHN H. HUMPHREY AND ROBERT R. DOURMASHKIN
number of antibody molecules attached to sheep erythrocytes with the number of holes per cell membrane detected by electron microscopy after lysis in the presence of guinea pig C’. The antibodies were labeled with lZ5I,and known numbers of molecules were added per cell, more than sufficient to lyse all the erythrocytes in 2 hours at 365°C. in the presence of a 1:50 dilution of C’, preabsorbed with sheep erythrocytes. Many membrane fragments were examined; the number of holes seen md the membrane areas were measured; and the number of holes per whole membrane (26 p 2 ) calculated. Using their purest IgM antibody, the ratio of the number of antibody molecuIes to the number of holes was found to be 1.15 to 3.4 (mean 2.1), whereas with their best IgG antibody the ratio varied from 1: 308 to 1:4100. Using the same antibody preparations, they also measured the number of antibody molecules per erythrocyte required for 50%lysis, and calculated on the basis of Mayer’s one-hit theory how many molecules were on average necded to cause 1 lesion per cell. The numbers for IgM antibody were 2.4 and for IsG about 4000. The maximum number of rabbit IgM Forssinan antibody molecules which can attach to a sheep erythrocyte was estimated to be about 100,000, and of IgG molecules about 600,000 (Humphrey and Dourmashkin, 1965). In the case of rabbit IgM antibody and guinea pig C’, it had already been shown that the number of holes corresponded to the number of theoretical hits. They argued that in the case of IgM antibody it was very unlikely that two molecules would arrive by chance at 2 neighboring sites out of 100,000, and that a single molecule would suffice to c a s e a lesion. The fact that about 2 molecules were required in their experiments could be attributed to working in the absence of a sufficient excess of the C’ components. The validity of this explanation was supported by subsequent experiments (see Section IV) in which it was found that, in the presence of higher concentrations of C’ ( 1:12.5), at various multiplicities of antibody molecules to erythrocytes the number of holes per cell was linearly related to the number of antibody molecules per cell. Borsos and Rapp (19651) introduced the technique of C’la fixation and transfer, by which the number of C’1 molecules activated in a system can be measured by transfer of the C’la to EAC’4 cells, which are then lysed by addition of a large excess of the remaining C’ components. They found a linear relationship between the amount of IgM antibody and the amount of C’1 activated over a wide range of antibody concentrations and aIso concluded that 1 molecule of antibody was sufficient to sensitize a shcep erythrocyte to the action of guinea pig C’. Thus this conclusion
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
97
seems to be soundly based, although in those systems where a lesion is represented by a cluster rather than by a single hole (see Section V ) , a one-to-one relationship between the number of holes and of IgM antibody molecules is not to bc expected. Similar measurements have not been reported for other lytic systems, but Rowley and Turner (1966) have examined how many molecules of rabbit IgM antibody against the 035 lipopolysaccharide antigen of Salnioitella oddaide were required on average to opsonize one bacterium, so that it would be ingested and killed by macrophages in the peritoiieal cavity of a living mouse. They found that a minimum of 8 molecules were needed and suggest that C’ activation is a part of the opsonic mechanism, as proposed by Nelson ( 1962). In Humphrey and Dourmashkin’s ( 1965) attempts to correlate directly the number of holes and the number of IgG rabbit anti-Forssman antibody molecules per erythrocyte, their results showed a much wider variation than in the case of IgM. This may have been partly due to the tendency discussed in Section IV for IgG antibody to give rise to clusters of holes, so that a few fragments of membrane with clusters may greatly exaggerate the number of holes estimated per cell. It may also have bccn due to the tendency reported by Coltrn et nl. (1967) for IgG antibodies to fix C’la to a marked but variable extent less well at elcvated temperatures (30°C. ) than at lower temperatures (6°C. ). No account was taken of this in Humphrey and Dourmashkin’s experiments. However, they suggested that their findings could be explained if 2 adjacent IgC, molecules on the erythrocyte surface were required to activate C’ and cause a hole. They calculated that if a sheep erythrocyte has 600,000 combining sites for IgG antibody, about 800 moIecules would be required to attach at random so that there would be an even chance of 2 molecules being adjacent ( assuming that each site has two effective neighboring sites) and concluded that their findings were coinpatible with this cstimated figurc. Humphrey ( 1967) showed that the amount of rabbit IgG anti-Forssman antibody required to lyse sheep erythrocytes in the presence of a given amount of guinca pig C’ was approximately 860 times greater than the amount needed to produce a similar degree of lysis when the crythrocytes, with IgG antibody attached, were also trcated with an excess of guinea pig y - antirabbit IgG. On thc assumption that a solitary molccule of rabbit IgG antibody on the erythrocyte surfacc would not activate C’ but that subsequcnt attachment of 3 or 4 molecules of thc anti-IgG would do so effectively, this finding supports the calculation that roughly 800 nioleculcs of IgG antibody per shcep erythrocyte are reqtlired to activate. C’ at one site,
98
JOHN H. HUMPHREY AND ROBERT R . DOURMASHKIN
Clearer, though still indirect evidence that 2 or more 7s antibody molecules in close proximity on the cell are required to fix 1 molecule of C’la was provided by Borsos and Rapp (1965b) who found that the number of C’la molecules fixed was proportional to a power of the antibody concentration slightly over 2 (2.1-2.6). That this conclusion is also valid for IgG antibody in immune complexes is suggested by the observation of Cohcn (1968) that IgG rabbit antiovalbumin, of which the antibody-combining sites were intact but of which the C’-fixing ability had been destroyed by amidation and benzylation, inhibited C’ fixation by intact antibody in thc presence of ovalbumin to an cxtent predictablc from calculations based on the assumption that 2 intact, adjacent, antibody molecules arc necessary for c’fixation. Most of the experiments discussed in this section were done with rabbit IgM and IgG antibodies and guinea pig C’. Gcneralization of the conclusions to other systems may not be justified; indeed, even among rabbit IgM antibodies, some antibodies that fail to activate C’ have been described ( Hoyer et al., 1968).
V11.
The Stage of C’ Action at
Which Holes
Are Formed
If the holes are, indeed, the lesions that lead to the loss of osmotic control by cell membranes, it should follow that they are formed at the final stage of C’ activation, i.e., by the action of C’9 (MidIer-Ebcrhard, 1968). Jn the case of C’ holes made in isolated bacterial lipopolysaccharides, where no question of osmotic regulation arises, Gewurz et a!. (1968) showed that all the components of guinea pig C’, from C’1 to C’9, were consumed. Interestingly, they found that lipopolysaccharide incubated with fresh normal serum (containing, presumably, natural antibody) consumed relatively little C’l, 4, and 2 but relatively much more of components C’3 to C’9, compared with aggregates composed of rabbit anti-RSA and BSA made at equivalence. In this respect lipopolysaccharide behaved like zymosan. Gewurz et al. found that consumption of thc tcrminal C’ components depended, nevertheless, upon the prior activation of C’l, 4, and 2. Although the extensive activation of the C’3 components onward remains to be explained, it is consistent with the present authors’ observation th‘it remarkably large numbers of holcs are produced by C‘ on lipopolysacchiiride substrates in the presence of quite small amounts of natural antibody. Examination of erythrocyte membranes in thc electron microscope at the stagc EAC’142 by Rorsos et a?. (1962) did not reveal any abnornialities, Mullrr-Ebcrhard ( 1965) stcites that electron microscopy rcvealed
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
99
the characteristic ultrastructural lesioiis following human C’9 action, and quotes a note by Hadding et 01. (1966) reporting that 110 such lesions could be detected with certainty in membranes of cells containing C’8 sites. Increased fragility of erythrocytes at the C’8 stage has been reported by Linscott and Nishioka (1963), by Stolfi (1967), and by Gotze et al. (1968), and it would be interesting to know whether any changes in the membrane are present at this stage. Frank et nl. (1964) showed that 0.09 hl EDTA inhibits the transforination of a cell in the E* state (i.e., which has been acted upon by all the C’ components and has one or more irrtprable lesions) to a ghost which has leakcd its hemoglobin. They also showed (1965) that the conversion of E D to a ghost involved three steps. The first, which was temperature-dependent, took place in the presence of 0.09 M EDTA; the second, which was not temperature-dependcnt occurred only when the EDTA concentration was lowered to 0.01 M ; and the third, which was also indcpendent of temperature, was blocked by the presence of 25% BSA in the external medium. This last is presumably the osmotic swelling which occurs after the holes are formed. The first two steps could be the chemical or enzymatic action of C’9 on the cell membrane, followed by some rearrangement of the constituents to form the holes. It would be of considcrable interest to verify this by examining membranes fixed at each stage, but, unfortunately, in the present author’s experience, the number of lesions per cell which may be expected to be produced by thc methods described is so small that definitc conclusions could not be drawn. Information about the nature either of C’9 or of the holes should cast light on each other. It is suggested by Miiller-Ebcrhard that C’S is bound firmly to the cell surfacc and that C’9 reacts with ceII-bound C’S. In the case of sheep crythrocytes which had reacted with human C’S, Hadding and Muller-Eberhard (1967) found that lysis resembling that causcd by C’9 could be brought about by 1,lO-phenanthrolinc and they postulated that chelation of ferrous ions was involved in the final step. Giitze et al. (1968) using EAC’1-8 prepared with guinea pig C’ were unablc to confirm this suggestion. These workers prepared highly purified C’9 from pig serum and found that its activity was abolished by 0.1 111mercaptoethanol but not by reagents rcactiiig with -SH or with serine hydroxyl groups. It was also inhibited rcyersibly by 10F A1 Cu2+or 10;”M Zn’+, but irreversibly by either lo-” A1 Fez+or 10.’ h l Fe:“ ions. Apart from thc suggestion that heavy metal ions may be involvcd, the modc of action of C’9 is at present unclcar.
100
JOHN H. HUMPHREY AND ROBERT R. DOURMASHKZN
FIG.7. Electron micrograph of Escheiichia cola lipopolysaccharide ( LPS ) applied directly to grid, then treated with human complement ( 1:20) containing
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
VIII.
The Nature of
101
C’ Holes
A. QUALITATIVE STUDES Evidence has been accumulated that C’ holes represent a rearrangement of the predominantly lipid outer layer of the cell membrane, in which small micelles of lipid constituents form circular or spherical structures, the centers of which become filled with stain and can be viewed in the electron microscope. It was found soon after the holes were first observed that their appearance was not affected-except that it was actually sharpened-by treatment of membranes bearing holes with trypsin or with buffer at pH 2.5 to detach antibody and/or C’ components. We have since treated holes in Esclierichia coli lipopolysaccharide on electron microscope grids with pronase, and, again, the only effect is to sharpen but not to alter their appearance (Fig. 7, authors unpublished work). It was also found that typical holes could be pro-
FIG.8a. Formalin-fixed sheep erythrocyte nieml)ranes, bearing very many complement holes (guinea pig), were fixed for 10 minutes in OsOl and negatively stained. Magnification: X400,OOO. ( From Humphrey ct al., 1968.) natural antibody. The grids were subsequently floated on 0.1%pronase for 1 hour. Pronase did not change the appearance of the holes. Coinpleiiient holes are seen both on LPS particles and also on LPS spread oiit on grid; their bizes are different (see Table I ) . Magnification: X225,OOO.
102
JOHN H. HUMPHREY AND ROBERT R. DOURMASHKIN
duced by the action of anti-Forssman antibody and C’ on erythrocyte membranes prepared by osmotic lysis and then hardened by treatment with 2% formaldehyde in thc cold (Humphrey, 1963; Humphrey et a]., 1968). Such treatment, which does not destroy thc Forssman antigen, made it possible to obtain stabilized erythrocyte membranes bearing holes which could be subjected to extraction procedures with organic solvents and yet leave the membrane matrix sufficiently intact for subsequent electron-microscopic examin a t’ion. Extraction of an aqueous suspension of the formalin-fixed membranes with ether did not affect the holes, but alcohol-ether removed them to a large but variable extent, and chloroform-methanol (2: 1 ) removed them completely ( Fig. 8 ) . These observations were interpreted to mean that the holes are present in a predominantly lipid layer on the membrane, some or all of the constituents of which are removed by solvent extraction. After formalin-fixed, sheep erythrocyte membranes bearing C’ holes
FIG.81). A preparation similar to that in Fig. 8a after extraction with chloroform-methanol for 1 hour. After this treatment, coniplemeiit holes could not be found. Magnification: X400,OOO. (From Humphrey et al., 1968.)
THE LESIONS IN CELL MEMBRANES CAUSED BY COhfPLEMENT
103
had been extracted with alcohol-cther in the cold and shown to have few or no remaining holes, thcy wcie treated afresh with anti-Forssnian antibody and guinca pig c’ and rc-examined. Holes were again visible, though less numerous than before thc extraction procedure ( Humphrey et al., 1968). Since about 5x of the Forssman antigen on the membranes had survived alcohol-ether extraction and was available for combination with antibody ( unpublished observation ), this finding implies that a fresh substrate in which C’ hole$ could be formed was furnished during the vcond treatment with antiserum and C’. Erythrocyte membranes contain a complex mixture of lipids, which differs mxkedly from species to species (e.g., human and shcep), yet the C’ holes in each arc similar. It seemed probable, thercfore, that further analysis of the nature of the holes would be easier if a simpler substrate werc used. Humphrey et at. (1968) chose the native Iipopolysaccharide complex secreted by Escherichia coli 12408 when grown in a lysinedeficient medium, since this material was a typical rough strain 0 antigen the composition of which had been characterized by Knox et nl. (1967). It contained 60%of Iipopolysaccli~uide,26% of extractable phospholipid (largely phosphatidylethanolamine ), and 11%of peptide. The presence of peptide permitted the material to be tracc labeled with radioactive iodine, and so enabled the amount adsorbed onto carbon-coated electron microscope grids or glass surfaces or onto bentonite particles to be nieasured easily. As already mentioned in Section 111, C’ holes were readily produced by the action of fresh normal human or guinca pig serum on this material, even without the addition of extra antibody from specifically immunized rabbits. Such holes could be completely removed by extraction with chloroform-methanol, but only if acetic acid was also added. The requirement for acetic acid was interpreted as due to the need to overcome tight bonding of the lipids by the layer of activated carbon on the grids; the labeled lipopolysaccharide itself was not removed by the solvent. These findings confirmed that the holey were formed in some lipid substrate on the surface of the grid, supplied by the lipopolysaccharide complex and,/or by the scruni used as the source of C’. In order to tcst the second possibility, freshly prepared carbon-coated grids were floated on solutions of highly purified simple antigens, namely crystallized human serum albumin (HSA) or Type 3 pneumococcus polysaccharide, and washed. The grids with adsorbed antigens were then incubated with various dilutions of specific rabbit antiserum and guinea pig C’. Over a fairly narrow range of antibody dilutions the grids showed scanty but typical C’ holes, controls without antibody or with heated C’ showed none (Humphrey ct al., 1968). The original HSA, even though
104
JOHN H. HUMPHREY AND ROBERT R. DOURMASHKIN
crystallized, was found by thin-layer chromatography to contain about 1%of a mixture of lipids resembling those in normal serum; however, substantially more serum lipids than would be accounted for by the adsorbed HSA became adsorbed onto the grids during the contact with antiserum and/or C’. These findings were interpreted as showing that adsorbed serum lipids are able to provide a substrate for the formation of holes by activated C’, though not necessarily to imply that they are the usual substrate on cell membranes. There is evidence that at least some plasma phospholipids exchange quite rapidly with those on the surface of erythrocytes (Lovelock et aZ., 1960) and that plasma unesterified cholesterol exchanges with that on a variety of biological membranes (Hagerman and Gould, 1951; Graham and Green, 1967). The possibility arises, therefore, that the difference in the size of holes made by human C’ and guinea pig C’, noted in Section 11, is due not to differences in the end product of C’ action but because the membrane lipids are affected by lipids in the serum used as the source of C’. We attempted to test this by examining holes made on Escherichia coli lipopolysaccharide complex by small amounts of fresh guinea pig serum in the presence of a 20-fold excess of heated human serum (1hour at 56°C.) and vice versa. In each case the size of the holes was characteristic of the active C’ and was unaffected by the presence of lipids derived from the heated serum ( authors’ unpublished observations).
B. CHEMICAL STUDIES Humphrey et al. (1968) reported preliminary studies in which Escherichia coli-lipopolysaccharide ( LPS ) complex was firmly adsorbed onto carbon-coated glass cover slips, which were treated with fresh or heated human serum or buffer solution. Controls included cover slips without LPS treated in parallel. Six hundred square centimeters of cover slip surface would adsorb about 1 mg. of LPS and so provide sufficient material for analysis of the lipids by thin-layer chromatography. Similarly treated electron microscope grids were floated on the solutions at each stage, so as to check that many holes were made on the LPS by fresh serum but none in the other preparations. The lipids extracted with chloroform-methanol-acetic acid were examined in several solvent systems. A wide variety of lipids from heated or unheated serum was found to be adsorbed onto the cover slips, whether or not they had LPS on them. The phosphatidylethanolamine of the LPS was not evidently affected by the formation of holes, and no lysophosphatides were present in any of the samples; there appeared, however, to be some small changes in the distribution of cholesterol and its esters and of free fatty acids.
THE LESIONS
IN CELL
MEMBRANES CAUSED BY COMPLEMENT
105
The use of carbon-coated surfaces had two disadvantages. The first was that the amount of serum lipids adsorbed was so large as to seem likely to ohscure any possi1)le minor cffclcts due to activation of C’; thc second was that such surfaces contained, besides carbon, variable amounts of a mixture of lipidlike materials which were shown to be derived from the carbon arc rods uscd for coating the cover slips. (These survived heating the rods to high tcmperaturcs and are prcsumed. though withoat proof, to be higher paraffins from the bonding material used in making the rods.) In further unpublished experiments, the LPS was, therefore, adsorbed onto bentonite, which could firmly bind the complex up to 40%of its dry weight. The sm‘dler bentonite particles were suitable for electron-microscopic examination, and it was found that abundant holes were produced by C’ on the adsorbed LPS. When experiments similar to those described in the preceding paragraph wcre performcd with LPS on bentonite, the bentonite was found to adsorb serum lipids, though to a lcsser extent than the carbon-coated cover slips. Again, thinlayer chromatography of the extracted lipids failed to show any clear differences betwcen those from LPS trcated with fresh serum with many holes or with heated serum, with none (Fig. 9). Subsequent re-cxtraction of the bentonite with solvents containing 1.0 N HCI or with Na dodecylsulfate and urea gave no indication that any lipid remained after the first extraction. The position at present is that, despite the evidence that C’ holes are formed in a surface lipid layer, no chemical evidcnce of changes in the lipids has been detected. It is possible to add two further pieces of negative evidence. The first relates to the possibility that some altcmtion in cholesterol is involved in the formation of holes. This was tested by treating erythrocyte membranes with saponin, which solvatcs in cholesterol layers to produce the hexagonal micellcs described in Section 11, or with digitonin, which combines with cholesterol to form pseudocrystalline structures on erythrocyte membranes (Dourmashkin and Rosse, 1966). Neither saponin nor digitonin affected the appearancc of preformed C’ holes, nor did pretreatment of the membranes prevent the subsequent formation of holes by antibody and C’ ( Dounnashkin, unpublished observations). Although these findings do not cxclude thc poqsibility that cholesterol is an important constituent of the lipid layer in which hoIes are formed, they make it improbable that any specific effect on cholesterol is involved. The second possibility, advocated by Munder et al. (1965) but disputed by Keller ( 1965), is that c’ acts on cell membranes by the generation of lysolecithin. Humphrey et al. (1968) looked for evidencc that lysophosphatides were present in erythrocyte mcmbranes as a result of
FIG. 9. Effect of C' on surfacc lipids. Thin-laycr chromatogram on silica gel G.
THE LESIONS IN CELL MEMBRANES
CAUSED BY COMPLEMENT
107
the action of C’. Although whcn added to crythrocyte suspensions, even in sublytic amounts, lysolecithin or lysophosphatidylethanolarnine could readily be recovered and detected in solvent extracts of the cell membranes, no increase in lysophosphatides in the membranes was found to accompany extensive production of holes by C’. In fact, whcn the sera used a s the sources of antibody and of C’ containcd no detectable lysophosphatides initially, none were detected in the C’-treated cell memhranes; furthermore, no 1ysophosphatidc.s were found in extracts of LPS even when extensive hole formation had occurred. An additional, though Icss compelling argument against the participation of lysophosphatides is that the menibrimes of erythrocytcs lysed by adding lysolecithin to the suspending ineclium did not show any typical C’ lesions. Smith and Becker (1968) examined the changes in total lipids that occurred when 2 x 10’” sheep erythrocytes scnsitized with rabbit Forssman antibody were incubated with guinea pig C’. They detected an increase in titratable acid, in parallel with a dccrease in serinc or choline phosphatides, which obeyed first-order kinetics. Antibody and activc C‘ were both required for thcsc. changes to occur. No such changes wcrc found when cells which had reacted with C’ as far as C’3 or even C’7 wcre incubatcd alone, but thcy did occur on addition of partially purified C’8 and C’9 or of EDTA guinea pig serum. The authors suggest that the lipid alterations were due to the action of one of the terminal components of C’ on the cellular intermediates. Because thc lipids examined were cxtracted from the mixture of erythrocytes and scruni (or scrim fractions), it is difficult to relate them to specific changes in the cell membranes. Calculations made from the authors’ data suggest that 50%lysiis of 2 X of the lipids extracted I,y chloroforiii-iiic.tlianol 2 : 1 ( v / v ) from two systems which h;id I,een treated with antibody and heated C‘ ( n o holes) or active C’ ( m a n y holes). Thc systems were as follows: ( I ) forlnalin-fixed sheep erythrocyte mernbranes ra1,bit anti-Forssnian Ig51 antibody f heated guinea pig C’; ( 3 ) as ( I ), h i t using fresh guinea pig C’; ( 4 ) Escherichin coli LPS ac1sorl)ecl onto Iientonite particles heated human seruin; ( 5 ) as ( 4 ) , but using fresh human ser~iiii (eqiial ainorints of reagents were employed far coinparison in each system); ( 3 ) reference prcparation of rat hrain polar lipids and standard neiitial lipids. I.ipids were identified as n, cholesterol rsters; 17, triglycerides: c, fatty acids; d, cholesterol; c, polar lipids remaining at the origin on neutr;il lipid plates; f , nrutral lipids at the first solvent front ( C :51 :W ) ; g , cerebrosides; h, phosphatidylcthanolaniines; i , sulfatides; i, phosphatidylcholines; k, sphingomyelins; I , phosphatitlylseriiie; m, Iysophospliatidylcholine; n, gangliosides. [Neutral lipids \\‘ere separatcd in petioleiim ct1ier:etlier :glacial acetic acid ( 7 3 :25 : 2) ; polar lipids were sepalatrd i i i chloroforin : methanol :\vater ( 14 : 16:1 ) to 15 ciii a i d then in it-propanol : 12.5%acpeotls h’H,OH (4:1 ) to 8 cm.] T l ~ scliroinato~rupl~y wits rarricatl ont by Sheila N. Payiw.
+ +
108
JOHN H. HUM P HR E Y AND ROBERT R . DOURMASHKIN
1O1O cells was accompanied by the release of about 0.025 p M of titratalile acid, i.e., about 1.5 x 10” molecules of acid were releascd for each cell lysed. Since only half the cells were lysed and their experimental conditions were siich that single holes would be expected to accompany each lesion, this would imply a change in the improbably large number of 10” molecules of phosphatide per hole if the substrate were phosphntide in the membrane. Furthermore, when heated guinea pig serum was used, increases in titratable acid were sometimes observed up to half those found with fresh serum, although no hemolysis occurred. For numcrous reasons, especially those advanced in Section VII1,A above, we propose that C’ holes are due to a rearrangement occurring to form micellar structures in a predominantly lipid surface layer. The fact that holes are similar in appearance and size when generated by C’ from a given species on a wide variety of membranes, and even on a pseudomembrane formed by thc adsorption of serum lipids onto a carbon-coated surface, argues in favor of the micelles being formed by the local action of a similar or identical detergentlike agent in each case. Nevertheless, the consistent difference observed between the size of the holes formed by human C’, on the one hand, and by guinea pig C’ (and, from very limited data, rabbit and calf C’), on the other, suggests that the end products of C’ from different sources may differ. The hypothetical agent could be generated from adsorbed C’ components (either C’9 or earlier) or could arise by alteration or removal by C’8 or 9 of substances preexisting in thc membrane. Unfortunately there are at present few clues as to what this hypothetical agent might be. IX.
Artificial Membrane Models
Model lipid membranes have been used increasingly for electronmicroscopic study of the action of surface-active agents, such as lysophosphatides ( Bangham and Home, 1964), saponin (Lucy and Glauert. 1964), or filipin (Kinsky et al., 1967). They have recently been applied to the study of antigcn-antibody reactions in the presence of C’ by Barfort et d. (1968) and by Haxby et d . (1968). Barfort et a2. prepared himolecular membranes from sphingomyelin and a-tocopherol and nieitsiired the decrease of electrical resistance in 0.1 A{ NaCl when solutions of various antigens were on one side and of specific antisera, on the other. Very marked changes of resistance occurred when active C’ was present in the antiserum, but not in its absence. Haxby et al. (1968) prepared . hposonics,” a term describing an aqueous colloidal suspension of sphcrules of lipid mixtures, a certain proportion of which may be completely closrd, thereby tr‘ipping n quantity of aqueous solution. These “
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
109
authors used the lipids obtained by extraction of sheep or becf crythrocytes with chloroform-nieth;~nol, they also prepared liposomes with sphingomyelin ( o r ~~1losphaticlylcholinc), cholesterol, and dicetyl phosphate, into the membranes of which was also incorporated a methanolsoluble extract of the erythrocytes (presumed to contain Forssman antigen). The liposomes wcre formed in an aqueous medium containing glucose, which became trapped within them, so that subsequent damage to the liposomc membranes could be nionitorcd by the release of glucose, assayed by a sensitive spectrophotomctric method. Haxby et al. found that on incubation with anti-Forssman antibody and C’, glucose was released from the liposomes. The requirement for antibody and active C’ was absolute. In the case of liposomes prepaied from sphingomyelin, cholesterol, and dicetyl phosphate, but without Forssinan antigen, no release occurred even when antibody and C’ were present. These authors mention that clectron-microscopic examination of lipo~omeswhich released glucose rapidly, revealed “pits” typic‘il of C’ action. Such experiments would seem to prove that C’ holes are formed in a lipid substratc. They do not, however, enable conclusions to be drawn concerning the detailed nature of the substrate in which the holes are formed, became of the inevitable adsorption of lipids or lipoproteins from the C’ and/ or the antiserum. It would be of considerable interest to test the effect of purified antibody and C’ frce from lipid on a simple meinbranc system or even on isolated bacterial lipopolysaccharides. Unfortunately all the usual methods for removing lipids from serum or plasma cause inactivation of hemolytic C’. Dalmasso and Muller-Eberhard (1966) succeeded in removing nearly all thc lipids from human serum, apart from those associated with serum albumin, by prolonged high-speed centrifugation after adjusting the density of serum to 1.21 with NaBr. This procedure inactivated C’3, 4, and 5, hut hemolytic activity could be restored by adding back the missing coinponmts in a highly purified form containing no detcctable lipids. Dalmasso and Mdller-Eberhard found that if albumin was also removed b y ammonium sulfate fractionation from the lipid-depleted plasma, rendering it practically lipid-frec, hcmolytic activity could still be restored with purified C’3, 4, and 5. Thus, lipid-free C’ can evidently lyse sensiti7cd sheep erythrocytes. However, their membranes already contain a complcx mixture of lipids, and it does not follow that lesions would equally be foimed on a simpler substratum. Wc attemptcd ( Humphrey. Dourinashkin, and Payne, unpublished observations) to test this using LPS on electron microscope grids. The denser fraction of human srrum, subjected to prolonged centrifugation at a density of 1.21, retilined its macroglobulin
110
JOHN H. HUMPHREY AND ROBERT R. DOURMASHKIN
antibodies but lost its hemolytic activity (tested on AET-treated human group A erythrocytes) and its capacity to make holes in LPS. Both hemolytic activity and the capacity to make holes in LPS were restored by adding purified human C’3, 4, and 5 kindly provided by Dr. MullerEberhard. However, our reconstituted human C’ still contained some lipids, about 10% of those in the original serum. When the albumin fraction was also removed, and the serum was lipid-free, it still possessed its macroglobulin antibodies, and about one-tenth of the original hemolytic activity was restored by adding C’3, 4, and 5, but it no longer formed holes when applied to LPS under similar conditions. So far as they go, these observations imply that when the substrate is LPS some essential ingredient for the formation of holes is supplied by the serum lipids and, also, are consistent with the findings discussed in the previous section. There is considerable uncertainty at present about the structural organization of the numerous components of which cell mernbrancs are coniposed and about the way in which their selective permeability properties are maintained (e.g., see Chapman, 1968). The phenomcna discussed in this chapter have not generally been considered in relation to thc more gencrnI probIem of cell membrane structure. However, they may be relevant in two respects: first in emphasizing the readiness with which plasma lipids or lipoproteins can be adsorhed onto a variety of surfaces, so as to provide a layer that resembles a cell membrane at least in its susceptibility to the action of C’; second, in showing that a single hole or bubble about 100A. in diameter, occurring in the outer predominantly lipid Iayer, can cause an irreparable loss of the selective permcability properties of the cell. X.
Biological Significance of the Terminal C’ Lesion
Despite the fact that rabbits without C’6, mice without C’5, and men with little or no C’2 can survive in good health (see Miiller-Eberhards review, 1968), it is difficult to suppose that the complex mechanism involving the terminal C’ components and culminating in cell lysis would have evolved and siirvived in so wide a variety of vertebrate species if it did not have significant biological value. In this chapter, two points relcvant to this have been brought forward. The first is that a single IgM antibody molecule against ccll surface component9 can activate C’ so as to damage the cell lethally. It is probable that this occurs especially, and perhaps only, when the antigenic groups on the cell surface form a closely spaced, rcyeating pattern, allowing multipoint attachment of the pluri-
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
111
valent IgM antibody ( Frank and Humphrey, 1968)-a condition satisfied by antigcms such as thc Forssman antigen on erythrocytes and the 0 antigens of grain-negative bacteria. A similar requirement for closely spaced similar antigenic groups is also likely in the case of IgG antibodies, since otherwise it would not be possible) for 2 molecules to b c ~ o m eattached adjacently, as appears to be necessary for C’ activation by IgG to take place. In tlie case of gram-negative bacteria, the action of C’ is enhanced by the subsequent action of lysnzyme, which not only erodcs the underlying cell wall but renders the inncr protoplasinic inembranc accessible, in turn, to the action of antihody and C’ (Glynn, 1969). The whole mcchanism seems well designed to make the best uscb of weakly avid “natural” IgM antibody, and of the IgM antibodies produced early in a primary response (especially against the surface antigens on particulate materials). The second point is that the lesion produced by activation of C’ at a single site may consist not of one hole but of a cluster. It is conceivable that this may represent a useful amplification of the action of C’; for example, if a cell or a microbe were able to withstand the damage caused by one hole but not that caused by several. However, we have no evidence about this, and thr circumstances in which clusters are produced have not yet been adequately defined. Two important riders need to be attached to any generalizations about the damage to cells by C’. The first is that the lytic action of C’ appear5 only to be effective whcn C’ activation occurs actually at or very close to the susceptible surface of the membrane. For example, attachment of bacterial lipopolysaccliaride 0 antigens to sliecp erythrocytes or to PNH human erythrocytes (Rosse et ul., 1966) rendcrs the cells extremely sensitivc>to the lytic action of C’ and specific antibody against the 0 antigen. However, we have ohserved in unpublished experiments, that when Type 3 pneumococcal capsular polysaccharide ( S3) is attached to erythrocytes by the method of Askonas et al. (1960), lieniolysis with C’ occurs only in the presence of vcry much higher concentrations of anti-S3 than are needed for liemagglutination or for fixation of C’ by the erythrocytes coated with S:3. Electron microscopy of tlie menibrancs of cells treated with anti-S3, with or without C’, revealed that the polysaccharide coating became aggregated by the antibody into iinev(~imasses which were no longcr closely applied to the membranes. Consequently, although C’ was fixed extensivcly, little or none of it was attached to the membrane. Rowley and Turner ( 1968) report cxperinients in which living Salnionella rrtlelnide organisms n ~ r econjugated with RSA, mouse 7-globulin or aggregated human IgG. Thc organisms wcrc thcn treatcd with guinea pig
112
JOHN H. HUMPHREY AND ROBERT R . DOURMASHKIN
C’ and rabbit antisera against the attached proteins or against Salmonella adelaide itself, and the amount of antibody needed to kill them was estimated. Coated organisms retained their sensitivity to the antibacterial antibody but were only killed by lo5 times more anti-BSA or lo6 times more antimouse y-globulin and were not killed by any quantity of antibody against aggregated IgG. Thcse findings were taken to imply that the effectiveness of C’ was inversely related to the distance of the site of activation from the surface. Similar conclusions could perhaps be drawn from the observation that human iso- and autoantibodies are more effective at causing C’-dependent lysis of human erythrocytes pretreated with trypsin, papain, or ficin than of normal ceIIs (Mollison, 1967), although in this case increased binding of antibody after enzyme treatment may be a sufficient explanation. The conclusion that holes are formed only when C’ activation occurs very close to the susceptible membrane implies that one or more of the activated C’ components remains lytically effective in free solution only for a very short duration-as appears to be the case for C’3 ( Muller-Eberhard, 1968). The second rider is that antibodies in some classes of immunoglobulin fail to activate C’ and that not all antibodies in those classes that do activate C’, namely, IgM and IgG, are equally efficient. It has already been mentioned that Hoyer et al. (1968) were able to distinguish a hemolytic and nonhemolytic form of rabbit anti-Forssman IgM antibody. In some species, notably guinea pigs, one kind of antibody ( y , ) can inhibit C’ fixation by another ( y 2 ) by competition for the same antigenic sites (Benacerraf et al., 1963). There is evidence that this may occur in other species also (Humphrey, 1968), and it would be interesting to know whether human antibodies of classes such as IgG4 and IgA which do not fix C’, could inhibit C’ fixation by antibodies of the other IgG subclasses. Among human isoantibodies, it is well recognized that some, such as IgM immune anti-A, are much more frequently capable of causing heinolysis than others, such as IgM natural anti-A, and that anti-Rh antibodies of any class are scarcely ever hemolytic (Mollison, 1967). Furthermore, even among the IgM anti-I antibodies found in cases of macroglobulinemia with cold agglutinins, there is surprisingly little correlation between the hemolytic and agglutinating capacities of sera from different subjects (Dacie, 1962). Some of the differences may be due to the way in which the antigenic groups are disposed at the cell surface, and some, for example, in macroglobulinemia with cold agglutinins, to selective removal of the more avid antibodies by the patient’s own erythrocytes in uiuo. However, it is evident that there is still much
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
11.3
to learn before the significance of C’-dependent lytic reactions is fuIIy understood. ACKNOWLEDGMENT W e wish to express our thanks to Mrs. Sheila N. Payne for collaboration in some of the experiments discussed above and for reading the manuscript.
REFERENCES Ackers, G. K. (1964). Biochemistry 3, 723. .4mano, T., Inai, S., Seki, T., Kashiba, S., Fujikawa, K., and Nishimura, S. (1954). Mcd. J. Osaka Univ. 4, 401. Andrews, P. (1966). Brit. Med. Bull. 22, 109. “The Separation of Biological Materials” (R. A. Kekwick, ed.). British Council, London. Askonas, B. A,, Farthing, C. P., and Humphrey, J. H. (1960). lmmzrnology 3, 336. Baker, R. F. (1964). J. Ultractrzrct. Res. 11, 494. Bangham, A. D., and Horne, R. W. (1962). Nature 196, 952. Bangham, A. D., and Horne, R. W. (1964). J. Mol. Biol. 8, GO. Barfort, P., Arquilla, E. R., and Vogelhut, P. 0. (1968). Science 160, 1119. Becker, E. L., and Austen, K. F. (1966). J . Exptl. Med. 124, 379. Benacerraf, B., Ovary, Z., Block, K. J., and Franklin, E . C . (19F3). J. Exptl. Med. 117, 937. Berry, D. M., and Almeida, J. D. (1968). J. Gen. Virol. 3, 97. Bladen, 13. A., Evans, R. T., and Mergenhagen, S. E. (1966). J . Bactcriol. 91: 2.377. Bladen, H. A., Gewurz, H., and Mergenhagen, S. E. (1967). J . Exptl. Med. 125, 767. Borsos, T., and Rapp, H. J. ( 1965a). J. lmmunol. 95, S59. Borsos, T., and Rapp, H. J. (1965b). Science 150, 505. Borsos, T., Dourmashkin, R. R., and Humphrey, J. H. (1964). Nature 202, 251. Brenner, J., and Horne, R. W . (1959). Biochim. Biophys. Acta 34, 103. Chapman, D. ( ed. ) ( 1968 ). “Biological Memliranes-Physical Fact and Function.” Academic Press, New York. Cohen, S. (1968). J. lmmtmol. 100, 407. Colten, H. R., Borsos, T., and Rapp, H. J. (1967). I n “Protides of the Biological Fluids” ( H . Peeters, ecl.), Vol. 15, p. 471. Elsevier, Amsterdam. Dacie, J. V. (1962). “The Haeinolytic Anaemias: Congenital and Acquired,” 2nd ed. Part 2, p. 466. Churchill, London. Dalinasso, A. P., and Miiller-Eberhard, H. J. (1966). J. Irnrnmol. 97, 680. Dourmashkin, R. R., and Rosse, W. F. (1966). Am. J. Med. 41, 699. Dourmashkin, R. R., Dongherty, R. M., and Harris, R. J. C. (1962). Nrztirre 194, 11 16. Frank, M. M., and Humphrey, J. H. (1968). J . E x p t l . M c d . 127, 967. Frank, M. M., Rapp, H. J., and Borsos, T. (1964). J. Z t t ~ m t m d 93, . 409. Frank, M. M., Rapp, H. J., and Borsos, T. (1965). J . Zrnrntmol. 94, 295. Gewurz, H., Shin, H. S., and Mergenhagen, S. E. (1968). J. E x p t l . Med. 128, 10.19. Glauert, A. M., and Lucy, J. A. ( 1969). J . Microscopy 89, 1 . Glauert, A. M., Dingle, J. T., arid Lucy, J. A. (1962). Nature 196, 95.3. Glynn, A. A. (1969). Irnm~rtdogy16, 463. Glynn, A. A., and Milne, C. M. (1967). Imtnrrtiohg!! 12, 639.
114
JOHN H. HUMPHREY AND ROBERT R. DOURMASHKIN
Gotze, O., Haupt, I., and Fischer, H. (1968). Nature 217, 1165. Graham, J. M., and Green, C. (1967). Biochem. J. 103, 16C. Green, H., Fleischer, R. A., Barrow, P., and Goldberg, B. (1959a). J. Exptl. Med. 109, 511. Green, H., Barrow, P., and Goldberg, B. (1959b) J. Exptl. Med. 110, 699. Hadding, U., and Miiller-Eberhard, H. J. (19’67). Science 157, 442. Hadding, U., Miiller-Eberhard, H. J,, and Dalmasso, A. P. (1966). Federation Proc. 25, 485. Hagerman, J. S., and Gould, R. G. (1951). Proc. Soc. Exptl. Biol. Med. 78, 329. Haxby, J. A., Kinsky, C. B,, and Kinsky, S. C. (1968). Proc. NatZ. Acad. Sci. U . S. 61, 300. Hoyer, L. W., Borsos, T., Rapp, H. J., and Vannier, U’. E. (1968). J. Exptl. Med. 127, 589. Humphrey, J. H. (1963). 3rd Intern. Symp. ImmunopathoZ. p. 369. Schwabe, Basel. Humphrey, J. H. (1967). Nature 216, 1295. Humphrey, J. H. (1968). In “Biochemistry of the Acute Allergic Reactions” (K. F. Austen ancl E. L. Becker, eds.), p. 249. Blackwell, Oxford. Humphrey, J. H., and Dourmashkin, R. R. (1965). Ciha Found. Symp. Complenierit p. 175. Churchill, London. Humphrey, J. H., Dourmashkin, R. R., and Payne, S. N. (1968). 5th Intern. Symp. Immunopathol. p. 209. Schwabe, Basel. Keller, R. (1965). Intern. Arch. Allergy Appl. Irnmuno!. 28, 201. Kemp, C. L., and Howatson, A. F. (1966). Virology 30, 147. Kinsky, S. C., Luse, S . A., Zopf, D., van Deenen, L. L. M., and Haxby, J. (1967). Biochim. Bi0phy.y. Acta 135, 844. Knox, K. W., Cullen, J., and Work, E. (1967). Biochem. J . 1803, 192. Linscott, W. D., and Nishioka, K. ( 1 9 6 3 ) . J. Ex&. Med. 118, 795. Lovelock, J. E., James, A. T., and Rowe, C. E. (1960). Biochcm. J. 74, 137. Lucy, J. A,, and Glauert, A. M. (1964). J. Mol. Biol. 8, 727. Mayer, M. M. (1961a). In “Immunochemical Approaches to Problems in Microbiology” ( M . Heidelberger and 0. Plescia, eds.), p. 268. Rutgers Univ. Press, New Brunswick, New Jersey. Mayer, M. M. (196lb). In “Experimental Immunochemistry” ( E . A. Kahat and M. M. Mayer, eds.), 2nd ed., p. 180. Thomas, Springfield, Illinois. Mollison, P. L. (1967). “Blood Transfusion in Clinical Medicine,” 4th ed., p. 250. Blackwell, Oxford. Morgan, T. E., and Huher, G. L. (1967). J. Cell B i d . 32, 757. Muller-Eberhard, H. J. (1968). Adoall. Immunol. 8, 1. Miiller-Eberhard, H. J., Dalmasso, A. P., and Calcott, M. A. (1966). J. Exptl. Med. 123, 33. Mnnder, P. G., Ferber, E., and Fischer, H. (1965). Z . Natirrforsch. 20b, 1049. Muschel, L. H., Carey, W. F., and Baron, L. S. (1959). J. Immrmol. 82, 38. Nelson, R. A . ( 1962). 2nd Intern. Symp. Immunopathol. p. 245. Schwabe, Basel. Padgett, F., nncl Levine, A. S. (1965). Virology 27, 633. Rosse, W. F., Dourmashkin, R. R., and Humphrey, J. H. (1966). J. Erptl. Med. 123, 969. Rowley, D., and Turner, K. J. (1966). Nature 210, 496. Rowlcy, D., and Turner, K. J. (1968). Notcrre 217, 657.
THE LESIONS IN CELL MEMBRANES CAUSED BY COMPLEMENT
115
Sears, D. A., Weed, R. J., Swisher, S. N., and Trabold, N. (1964). j . Clin. lnoest. 43, 975. Sessa, C . , and Weissman, C. ( 1968). J. Biol. Chem. 243,4364. Simpson, R. W., and Hauser, R. E. (1966). Virology 30, 684. Sirchia, G,, and Dacie, J. V. (1967). Nature 215, 747. Smith, J. K., and Becker, E. L. (1968). j . Immvnd. 100, 459. Stolfi, R. (1967). Federation Proc. 26, 362. Taylor, A,, Knox, K. W., and Work, E. ( 1966). Biochem. J. 99, 53. Wardlaw, A. C. (1962). J. Exptl. Med. 115, 1231.
This Page Intentionally Left Blank
Cytotoxic Effects of Lymphoid Cells in Vitro PETER PERLMANN AND GORAN HOLM Deportment o f Immunology. The W e n n e r - G r e n Institute. University o f Stockholm. Stockholm. Sweden
I . Introcluction . . . . . . . . . . . . . I1 Methods . . . . . . . . . . . . . . A . Target Cells . . . . . . . . . . . . B. Effector Cells . . . . . . . . . . . . C . Survey of Methods . . . . . . . . . . . D General Considerations . . . . . . . . . . E Concluding Remarks . . . . . . . . . . I11. Different in Vitro Models . . . . . . . . . . A . Cytotoxic Effects on Antigenic Target Cells of Lymphoid Cells from . . . . . . . . . . Sensitized Donors . 3. Induction of Cytotoxicity of Lymphoid Cells from Norma1 Donors . . . . . . by Antibodies to Target Cell Antigens C. Cytotoxic Effects of Lymphoid Cells Triggered by Target CellBound Complement . . . . . . . . . . D . Nonspecific Cytotoxicity of Lymphoid Cells Activated by Phyto. . . . . . . hemagglutinin or Other Stimulants E Target Cell Destruction by Lymphoid Cells from Nornial Donors . . . . . . . . after “in Vitro Sensitization” . . . . . . IV Some in Viuo Implications of the in Vitro Models 4 . Delayed Hypersensitivity . . . . . . . . . B. Autoimmunity . . . . . . . . . . . C. Graft-versus-Host Reactions . . . . . . . . . D. Allograft Rejection . . . . . . . . . . E . Tumor Defense . . . . . . . . . . . V Summary . . . . . . . . . . . . . . References . . . . . . . . . . . . .
.
. .
.
.
.
I.
117 119 119 120 121 125 126 127 127
144 153 156 168 172 172 175 177 179 181 183 185
Introduction
Lymphoid cells havc effector functions in certain tissue-damaging immune reactions . Cell-mediated tissue injury is believed to be of prime importance in delayed hypersensitivity. in some of the experimental and human autoimmunities. in the various allograft phenomena. and in somc forms of tumor rejection . The basis for this belief are. in brief. the typical histopathological picture of the tissue Iesions seen in these states. the various graft-versus-host phenomena. and the fact that tissue-destructive immune reactions can often be provoked in nonsensitized recipients by 117
118
PETER PERLMANN AND GORAN HOLM
transfer of lymphoid cells from sensitizcd donors rather than by immune serum. During recent ycars, a number of morc direct in vitro models have been designed in order to throw light on the mechanism of these tissuedamaging reactions. It is now well established that sensitized lymphocytes, upon contact with antigen, will inhibit thc in vitrn migration of macrophages and other leukocytes (Rich and Lewis, 1932; David et al., 1964). This inhibition is brought about by soluble mediators released from the reacting lymphocytes (Bloom and Bennett, 1966; David, 1966; Svejcar et al., 1968) and is believed to be of importance for the initiation of the delayed hypersensitivity reaction. In a different in vitro model, it has been shown that lymphoid cells from sensitized donors will destroy tissue culture cells carrying the antigen to which the cell donor is sensitized. This type of cytolytic reactions has been encountered in a great variety of immune situations, comprising all those mentioned in the first paragraph. The reactions are, therefore, believed to be in vitro correlates to corresponding effector activities of lymphoid cells in vivo. The cell that initiates the in vitro cytotoxic reaction has been assumed to be the sensitized lymphocyte, equipped with its own recognition sites for antigen on the cells which are destroyed. Although this may be true in many situations, it now seems clear that “normal” lymphoid cells can become cytotoxic to other cells by a variety of pathways. There is little doubt that this “nonspecific” cytotoxicity in vitro also is of importance in wivo. The study of the various pathways by which lymphoid cells can become cytotoxic has been helpful for our beginning understanding of their effector role in cell-destructive reactions in general. This review will deal with the in vitro cytotoxicity of lymphoid cells. After a brief discussion of some methodological issues, various in vilro models will be described. In a final section, the possible bearing of these models on cell-mediated tissue destruction in vivo will be discussed. Parts of this subject have previously been treated in this series (Dutton, 1967; Wilson and Billingham, 1968). In this article, the cells which are destroyed upon addition of lymphoid cells are called target cells or targets. The cells which achieve this are called effector celk. The designation lymphoid cells rather than lymphocytes has been chosen in order to put some emphasis on the fact that the effector cell populations used in many experiments are relatively crude mixtures, containing not only lymphocytes, heterogeneous by themselves, but monocytes or macrophages as well. Frequently, polymorphonuclear leukocytes are also present. The significance of lymphocytes and other
CYTOTOXIC EFFECTS OF LYMPHOID CELLS ill
VitrO
119
cell types for the cytotoxic rcactions t\dI lie discussed in the different sections. The tcrm setrsitizcrl cell or .sen.sitixtl hp1]hOC~/tedcsignate thc antigeIi-reactivc lymphocytc~swith mtiliody-like reccytors believcd to bc produccd by thc cells which carry them. In other words, it applies to those lymphocytes that are thought to bc) instrumental in delaycd hypersensitivity and similar reactions. In contrast, the term cellmediated reaction is used in a hroad sense. It applies here to those reactions in which lymphoid cells function as the effectors, regardless of whether or not sensitized lymphocytes are involved. It draws a border line between cellmediated reactions and those which are brought about hy humoral antibodies and complement but without participation of lymphoid cells. Obviously, this definition does not exclude participation of either humoral antibody or complement acting in conjunction with lymphoid cells. II.
Methods
Cell-mediated cytotoxicity in vitro is the manifestation of complicated cellular interactions. It is tested by adding lymphoid cells in excess and under tissue culture conditions to target cells of various kinds. After n certain time which varies depending on the particular system under study, destruction of the target cells takes place and can bc recorded. The many different methods which have been applied for the assay of cytotoxicity do not necessarily always nicasure the same reactions. Thcy are also subjected to different kinds of artifacts which must be taken into account when the results obtained with different methods are to 1~ compared.
A. TARGET CELLS All kinds of tissue culture cells may serve as target cells. When the stt1dic.s involve lymphoid cells from donors sensitized to target cell antigens which are stable in vitro, the cultures can be propagated for a long time. This is truc for the major transplantation antigens (Rosenau and Moon, 1964) and for some of thc tumor-spccific antigcns. I n contrast, some “tissue-specific” antigcns disappear within 1 week of culture (for references, see Dumonde, 1966). Studics of cytotoxic reactions in autoimmune diseases, therefore, usually requirc cultures of freshly explantcd tissues. Established cell lines of normal tissucs or tumors are frequently uscd (Brunner et a?., 1966; Holm and Perlmann, 1967a). Such cell lines have the advantage of being adapted to tissue culture conditions. They arc easy to culture as monolayers which in most instances can bc>converted
120
PETER PERLMANN AND GORAN HOLM
to suspension cultures. Cells grown in nionodisperse suspension are more suitable target cells than cells attached to glass. They are not influenced by enzymes or other factors which may detach the cells which grow in monolayer (Lundgren et al., 1968a). They are probably available for contact with the effector cells from all sides, and this may increase the sensitivity of the method. Furthermore, established cell lines consist of one cell type, in contrast to the mixture of two or more cell types usually encountered in primary cultures. Chicken erythrocytes have also been employed as target cells. When these erythrocytes are passively coated with antigen, they are excellent targets for lymphoid cells from donors sensitized to that antigen or, after reaction with humoral antibody, for lymphoid cells from normal donors ( Perlmann and Holm, 1968). It is a notable finding that macrophages, cultured on glass, may serve as target cells for lymphoid cells from donors sensitized to thcir transplantation antigens ( Brondz, 1964). In other situations, macrophages from sensitized animals are cytotoxic to antigenic target cells (Granger and Weiser, 1964). Antibody-forming target cells have been used by Friedman (1964). The reduction of plaque-forming cells in the Jerne assay after contact with lymphoid cells from sensitized donors was taken as a measure of cell destruction. Different types of target cells may vary in susceptibility to cell-mediated lysis (Holm, 1967a; Brunner et al., 1969b). However, even target cells of one type may sometimes vary in susceptibility when different batches are compared. Comparison of the susceptibility of target cells of different types may, therefore, be difficult to quantitate. B. EFFECTOR CELLS Lymphoid cells are obtained from the lymphoid organs or from the circulation. When prepared from lymph nodes or spleen, the common techniques for preparation of cell suspensions are applied ( for references, see Ling, 1968). Comparison of the cytotoxic activity of cells from different origin must be done with caution. With spleen and lymph node cells it has sometimes been found advantageous to place small pieces of tissue on a grid (Trowell, 1959) from which lymphocytes are allowed to fall on a target cell monolayer underneath (Vainio et al., 1964; Holm, 1966). However, this method is difficult to quantitate. It is essential to start the experiments with cell suspensions in which !N-100% of the lymphoid cells are viable. Cells from different animals may vary highly in their viability in vitro (for references, see Ling, 1968). Mouse cells in particular will die much faster during the course of an
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitl.0
121
experiment than human or guinea pig cells. Since the cytotoxicity of lymphoid cells is a fmiction of viable cells, differenccs in viability will affect the results. This has to be takcn into account when cytotoxicity of lymphoid cells from cliff erent specics are comparcd. In most experiments, the lymphoid cells consist of mixtures of lymphocytes of various sizes, monocytes, and macrophagcs. Polymorphonuclear leukocytcs and red cells are also present in many cases. Some of these cells (monocytes, macrophages, poIymorphonucIear leukocytes) may be cytotoxic by themselves (Granger and Weiser, 19M; Lo Buglio et al., 1967) or may die and release cytotoxic factors (Lundgren et al., 1968a). Moreover, these cell types may interact with each other in the course of the cytotoxic rcaction. To ascribe effector functions to a special cell type, it is essential to use homogeneous suspensions with as little admixture of other cell types as possible. Thoracic duct lymph is, therefore, preferred as the source of lymphocytic effector cells. In other cascs, the analysis requires the application of purification procedures. Even then it may be difficult to exclude the participation of small numbcrs of nonlymphocytic cells in the reaction, particularly when the number of lymphoid cells added to target cells is high.
C. SURVEYOF METHODS 1 . Destruction of Manolnyers
In these methods, disruption of target cell monolayers is evaluated after incubation with lymphoid cells. In a common modification, part of the monolayer is selected undcr the microscope and photographed before, during, and after incubation with the effcctor cells (Govaerts, 1960; Koprowski and Fernandes, 1962; Berg and KZllkn, 1963). Thesc methods have considerable drawbacks: the reactions are slow (48 hours or more) and evaluation of cell damage is subjective. Due to variations in target and effector cell densities, the important effector cell/ target cell ratios cannot be controlled from one part of the monolayer to anothcr. In the phque technique by Granger and Weiser (1964, 1966), later applied by Moller and Moller ( 1965), drops of suspensions containing known numbers of lymphoid cells are added to the target monolayer in well-defined areas. A cytotoxic reaction is recorded as local reduction of target cell density ( = plaque formation). The intensity of thr plaques is This allows a semiquantitative evaluusually scored from to ation of cell damage. With the plaque techniquc, the ratio of effector cells to target cells can be controlled to some extent. Small numbers of effcctor cells are
+ ++++.
122
PETER PERLMANN AND GORAN HOLM
needed for each plaque and several drops of effector cells can be placed on each monolayer. The method is rapid and the results are easy to read. However, long incubation times are often required for development of the plaques (Moller and Moller, 1965; Moller et al., 1966; Lundgren, 1969) . In all monolayer techniques the detachment of cells from the substrate is taken as reflecting the cytotoxic activity of the effector cells. Detached cells are usually seen to be damaged or dying (Rosenau and Moon, 1961; Biberfeld et al., 1968; Ax et al., 1968). However, various nonspecific factors, such as enzymes released from dying polymorphonuclear leukocytes or from other cells may also be responsible for detachment of undamaged target cells (Lundgren et al., 1968a). Non-specifically affected cells will usually not survive very long either, and their detachment will introduce errors in the evaluation of the results. 2. Cell Counting
Target cells from stock cultures are counted and added to flatbottomed tubes or petri dishes. When the cells have become attached to the glass, the effector cells are added. After incubation, the supernatant is usually discarded. The remaining target cells are removed from the glass mechanically or by trypsinization. Dead cells are stained with vital dye and the number of unstained cells are counted (Taylor and Culling, 1963; Brondz, 1964; Moller, 1965h). Alternatively, target cell nuclei are counted after digestion of the cytoplasm with citric acid (Rosenau and Moon, 1961; Wilson, 1965a). Since the culture medium is usually discarded prior to counting, only target cells attached to the gIass will be enumerated. When this is the case, these methods do not distinguish between cell death and cell dctachment ( see preceding paragraph). Moreover, the number of surviving ceIls is influenced by the number of cell divisions during incubation. Lymphoid cells may have growth-promoting feeder effects on the target cells (Taylor and Culling, 1963; Wilson, 1965a; Roseuau, 1968) and this may tend to give erronous results. Target cell multiplication can be suppressed by X-irradiation before incubation with the effector cells (Wilson, 1965a). Counting methods are relativcly laborious. Their methodical error is rather big and several tubes (usually 3-6) must be run in parallel. They can only be used when target cells and effector cells can be distinguished microscopically by differences in size or shape. However, in spite of these disadvantages, the methods have been used by several authors and form the basis of quantitative studies (Wilson, 1965a; Brunner et nl., 1966; Berke et al., 1969a). Several of the previous disadvantages are avoided
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
123
in some recent micromodifications ( Ulrich and Kieler, 1969; Takasugi and Klein, 1969).
3. Inhibition of Colony Formation These methods can be regarded as variants of the cell counting techniques. Hellstrom and Sjogren (1965) assayed the plating efficiency of tumor target cells after incubation with antibodies and complement. In a later modification (Hellstrom, 1967), thc target cells are allowed to form a noiicoiifluent monolayer overnight before lymphoid cells are added. The number of target cell colonies is counted after 3-4 days of culture. Brunner et al. (1966) described a slightly different technique where the target crlls after contact with effector cells were suspended arid cultured in a semisolid medium. The colonies were counted 1 week later. The colony inhibition method is very sensitive. When the effector cclls are added to nonconfluent nionolayers, the effective lymphocyte/ target ccll ratio may be very low, even when the number of lymphocytes added for each target cell is as high as 5000 (Hellstrom, 1967). Rapidly growing target cclls, adapted to tissue culture conditions, are more suitable than primary cultures. The method has mostly been used for studies of cytotoxic reactions directed against tumor cells. The colony inhihition techniques measure the survival of single cells which are capable of attaching to glass and of forming colonies by niultiplication. The method can, therefore, only be applied to certain types of monolayers. As in other monolayer methods, errors may arise because of en7yinatic or otherwise “nonspecific” detachment of cells from the glass.
4 . Isotope Release from Labeled Target Cells Release of radioactive markers from target cells is frequently used for quantitative evaluation of cell damage. In general, after incubation of isotopically lnbcled target cells with cytotoxic effector cells, the total radioactivity of the sample and that of thc cell-free supernatant (or of the insoluble residue) are determined. Rclcased (or retained) isotope, ~is~ially expressed as perceiitage of total radioactivity, reprcseiits a cuinulative measure of cell darnage. Target cell growth and multiplication will influence the results only marginally. Isotope can be introduced into target cells by making use of their metnbo’fism, or chemically. a. Deox!/ribonticleic Acid (DNA ) . \Vhcn DNA-synthcsizing target cclls are incubated with ”or I ’C-thymidine, thc label is incorporated into DNA. Dnmaged cells clo not release DNA unless completely disin-
124
PETER PERLMANN AND GORAN HOLM
tegrated ( Green et al., 1959a; Klein and Perlmann, 1963). Since damaged cells arc susceptible to proteolytic enzymes (Hirata, 1963), isotope rcIease from affected cells is measured after treatment of the target cells with trypsin ( Klein and Perlmann, 1963). Isotope released from damaged cells during the experiment is reutilized b y DNA-synthesizing cells. Horn'ever, reutilization can be blocked by addition of cold thymidine (Holm and Perlmann, 1965). The advantage of the method is the stability of the marker. Dcoxyribonucleic acid-labeled cells can be used for long-term incubations with the effector cclls (Holm et nl., 1964; Vainio et al., 1964; Holm and Perlmann, 1965; Holm, 1966). However, the method is rather insensitive. Thus, with cells, doubly labeled with thymidine-'% and chromate-'j'Cr and exposed to lymphocytes, significant damage was detected with 'ICr within 1 to 3 hours, but with l'C-thyniidine damage was first detected after 18 hours of incubation (Holm and Perlmann, 1967a). 13. Proteins. These can be labeled with radioactive amino acids; protein-bound label is readily released from damaged cells. The spontaneous release of label may vary for different cell types and reutilization may be difficult to control (Bickis et al., 1959; Perlmann and Broberger, 1963). c. 3'P-Pliosphate. This label is incorporated into P-containing constituents of both cytoplasm and cell nucleus. It can be used for determination of cell-niediatcd cytotoxicity in short-term experiments ( Ellem, 1958; Perlmann and Broberger, 1963). Spontaneous release and reutilization may be difficult to prevent in long-term incubat'ions. d. "Cr-Chromate. This has been used as a routine label in studies of the survival of red cells in hemolytic diseases. Sanderson (1964) and Wigzell ( 1965) adapted this label for quantitation of antibody-induced lysis of nucleated cells in vitro. Later, "'Cr-chromate was also found to be an excellent label for detcrmination of cell-mediated lysis of tissue culture cells growing in suspension (Holm and Perlmann, 1967a) or in monolayer ( Holm, 1967c) and of chicken erythrocytes (Perlniann et al., 1968). The method has since been adapted for the same purpose by other authors (Brunner et al., 1968a; MacLennan and Loewi, 1968a; Berke et al., 1969a). Chromium-51 is noncovalently bound to proteins and other cell constitucnts. The chromate is reduced during binding and isotope is not reutilized (Bunting et al., 1963; Holm and Perlmann, 1967a). From 80 to 95%of the radioactivity is released from dead cells (Wigzcll, 1965; Holm and Perlmann, 1967a). Labeling is rather stable; thus, only 1 5 4 0 % is spontancously releascd from tissue culture cells in suspension during 24 hours of incubation at 37°C. Most of this comes probably from spon-
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitio
125
taneously damaged or dying cells. Under optimal conditions, spontaneous W r relrase from chicken erythrocytes usually does not exceed 5%in 1 to 2, days (Perlmann et al., 1968). Because of a minimum of handling, the error of this method is very small. Therefore, release of “Cr represciits a very sensitive index of cell damage, which sometimes: can bc measured within less than 3 hours (Holm and Perlmann, 1967a: Brunner et al., 1969b). 5. Inhibition of Isotope Incorporation Changes of the nietabolism of tissue culture cells upon contact with lymphoid cells can be assayed as inhibition of incorporation of labeled precursors. Incorporation into protein of radioactive amino acids, added to the incubation mixture, has recently been used for this purpose (Granger and Williams, 1968; Granger and Kolb, 1968). This method has also bcen used to study the effects of 5oluble cytotoxic factors (Williams and Granger, 1969). When measured in this way, cytotoxicity reflects primarily inhibition of growth. The method suffers from the drawback that both the effector cclls and the target cells may synthesize protein. Incorporation of isotope into the cffcctor cells can be avoided by the use of precursors which are not utilized by lymphoid cells. Incorporation of radioactive thymidinc into DNA in a cell mixture during the first 24 hours of incubation will takc place primarily in the target cells (Ming et al., 1967). However, even in this c x e , the results may be difficult to evaluate.
D. GENERAL CONSIDERATIONS I . Nature of Cell Injurg The principal methods discussed above measure in part different types of target cell injuries. This raises the question to what extent results obtainccl with different methods are comparable. It will be noted, that detachment of target cells from the substrate is a component of all methods in which usc is made of target cells growing in monolayer. Dctached cells are usually dead or will rapidly die. AS slready statcd, detachment will often lead to loss of target cells for causes which may be unrelated to the activity under study. In the colony inhibition assay and in ccll counting methods, cxctxpt when nondividing target cells are used, growth inhibition is one manifestation of injury. When supravital staining is used, changes in membrane perme,ibility arc recorded. This is also the casc with all iiicthods in which iso t op rclcase i s mcwxircd. Growth inhibition nnd changes of mcmilirane
126
PETER PERLMANN AND GORAN HOLM
permeability, provoked by the effector cells, may be related and result in cytolysis (Briinner et ul., 196th; Berke et al., 1969a). Present data speak in favor of the opinion that membrane permeability increases carly during lysis (Green et al., 1959a,b). Methods that predominantly measure such alterations seem to be the most sensitive ones. In conclusion, there may be doubts about the interpretation of cytotosicity measured by methods based mainly on detachment of cells from glass. Growth inhibition and increased membrane permeability may reflect related phenomena, provided the experiments are performed under otherwise comparable conditions. 2. Choice of Controls
The choice of controls is crucial. Usually, target cell damage in the presencc of effector cells is compared with that in the samples which contain target cells without additions. Alternatively, target cells are incubated with lymphoid cells which are not expected to be cytotoxic. This second type of control would seem to be preferable since it provides culture conditions similar to those prevailing in the test samples. However, the lymphoid cells may have feeder effects on the monolayer, and these controls then grow and multiply faster than the controls free of lymphoid cells (Taylor and Culling, 1963; Wilson, 19651; Rosenau, 1968). In primary cultures and with low plating efficiency, only 10%of the cells may sometimes be found in the controls at the end of an experiment (Moller, 1965a). The results of such experiments must be evaluated with caution since there is no real base line to which the cytotoxic effects of thc effector cells can be referrcd. The release of radioactivity from cells labeled with isotope which is not reutilizcd is independent of most of the factors discussed above. Thc base line for estimation of the cytotoxic effects of lymphoid cells is usually provided by samples in which lymphoid cells are absent.
E. CONCLUDING REMARKS The interaction of lymphoid cells and target cells represents a dynamic situation which continuously changes during incubation. Some of the variables that govern the reactions have been described above. Others, such as differences in target cell susceptibility, state of sensitization of the donor of the lymphoid cells, death of effector cells, or transformation or selection of lymphoid cells during incubation will he discussed in the following sections. Thc discussion in this chapter can be summarized with the prescntntion of some criteria for the ideal method to determine cell-mediatetl
CYTOTOXIC EFFECTS OF L Y m w o I D CELLS
in Vitro
127
cytotoxicity: ( 1 ) high sensitivity to enablc detection of cell damage within the shortest possible time of incubation; ( 2 ) low error allowing measurement of small differences; ( 3 ) quantitative and kinetic measurements of cell damage; ( 4 ) low spontaneous cell damage in the controls (“background”); ( 5 ) independence from detachment of target cells from substrate; and ( 6 ) inhibition or promotion of target cell growth b y culture conditions do not affect the results. None of the methods available at present fulfills all these criteria. Monolayer methods and methods measuring the incorporation of radioactive precursors are far from thc ideal. Cell counting techniques and the colony inhibition method are better. Most critclria are fulfilled by the method that measures release of ”Cr. It can be applied in most experimental situations and with all types of target cells. Ill.
Different in Vifro Models
A. CYTOTOXIC EFFECTS ON ANTIGENICTAHGET CELLS CELLS FROhf SENSITIZED DOKORS
OF
LYhrPHOID
The first experimental approach along the lines discussed above was that by Govaerts (1960). This author studied the cytotoxic effect of lymphoid cells from the thoracic duct of dogs, sensitized by a kidney allograft, on donor kidney cells in tissue culture. Recipient lymphoid cells became aggregated to the cultured cells which were destroyed within 24 to 48 hours. Lymphoid cells from normal dogs or dogs who had received a third party graft were not cytotoxic. However, in contrast to what since has been found by many, recipient serum strongly enhanced the weak cytotoxicity exhibited by recipient lymphoid cells in normal serum. Moreover, addition of complcmcnt further potentiated the cellmediated cytotovic effect. A large number of similar inwstigations has since been performed in many laboratories. In Table I, sollie of this work has been compiled. The different papers have bcen ordered according to the immune systems studied. No attempt has been made to present a complete list. Rather, reference has been made to papers considered to be representative for work done in this field. 1. Aggregation of L~jinplioirlCel1.Y to Target Cells.
Whcwever microscopic observation h a s been included in the work, aggregation of lymphoid cellc to target cells has been sccn. This is most c y d y observed when the target c ~ l l sare tissue culture cell5 in monolayer. In these cacc’s, lymphoid cc.11~from normal donors, or from donors sen-
128
PETER PERLMANN AND GORAN HOLM
TABLE I I n Vilro CYTOTOXICITY OF LYMPHOID CELLSIN DIFFERENTIMMUNE SYSTEMS Authors
Target cells
Effector cells
Method
1. Transplantation Antigens
Govaerts (1960)
Dog, kidney cells Peritoneal or thoracic Monolayer dact cells, graft recipients Spleen cells, BALB/c Monolayer, cell Rosenau and Moon Mouse, L-cells (1961, 1964) (CJH), other immiinized with C3I-I counting cell lines or C57B1 tissue Wilson (1963) Mouse or rat Lymph node cells Cell coiinting kidney cells (mouse), thoracic duct cells (rat), sensitization by allogeneic skin grafts Wilson (1965n) Rat, tnmor cell Lymph node cells, 6ho- Cell counting lines racic duct, cells, rats sensitized by allogeneic skin grafts or immunization Vairiio et ol. (1964) Mouse, fibroLymph node and spleen Isotope release blasts, various cells, allogeneic immu(W-thymidine) H-2 genotypes iiization Monse sarcomas, Lymph node cells, mice Cell countiiig Broridz (1964, 1968) macrophages, immunized with alloH-2 genotypes geneic tumor cells Granger and Mouse, A/Jax Peritoneal cells, C57B1 Monolayer (plaqiie Weiser (1964) fibroblasts mice immunized with technique) Sarcoma I (A/Jax) Miiller (1965h) Mouse sarcomas, Spleen and lymph node Cell coiinting cells, mice immuiiized various H-2 genot,ypes with allogeneic cells Brunner el (11. Mouse, mastoSpleen cells, C57B1, Cloiiing assay, cell (1!)66) cyt.oma immiinized with cotinti Iig (DBA/2 urigiri) DBA/2 cells Monse, mashSpleen cells, C57B1, Isohpe release ( 5 0 ) Bruniier et c t l . (196%) cytomn immunized with (DBA/2 origin) I)BA/2 cells Hashimoto and Rat,, sarcoma Peritoneal exudate, frac- Cell couiit,ing Sudo (1968) (Yoshida strain) tionated into different cell t,ypes, Donryu mt,s immunized wiUi t,iimor Lurrdgren (I!KN) H\im:iii skin Blood lymphocytes, skin Munolayer (Plaqiie fil)rol)lnsts transplanted patients technique)
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
.bit
hors
Target cells ~
Effector cells
in Vitro
129
Method
~~~
2. Expe~iinentul$utoimnziine Diseases
Koprowski and Dog, brain tissue E:sperimeitt.nl allergic Sloiiolayer Fernandes (1962) etirephalit,is. Lymph node cell, Lewis ra(s imnirinizecl with giiinea pig spinal cord Rose 6.1 ul. (1969) Rabbit, thyroid .iiitoimmriiie thyroitlit,is. hloiiolaycr Lymph node cells, cells rnk)l)its imniuiiized with thyroglohulin, thyroid esiracts Experimental allergic hloiiolayer Berg and KallBii Eat, glia cells encephalitis. Blood (1963) motlonuclear cells, t.horacic duct cells, rabbits or rats immiinized with porcine spinal rord Aiit,oimmune thyroidit.is. Monolayer Biorklund (1964, R,zt,, thyroid Thoracic duct, and explants 1968) lymph node c,ells, rats immiiiiized with rat t.iiyroid Munolayer Rat, myelinated Ksperiment al allergic fibers of trineriritis. Lymph iiotle geminal ganglia cells, rats imniuiiized with rat or rabhit peripheral nerve Rat, kidney cells I1:xperinieiital “a~itoirn- Isotope release Holm (1966) (14C-thymidine) mrine” nephrosis. B h J d , mononuclear cells, lymph node, spleen cells, rals immruiized with rat, kitliiey l~oc:dmuscle lesions, Monolayer Rat., skeletal Kakulas (1966) muscle nitisclr fiber necrosis. Lymph node cells, rats immiinizetl with rat skeletal muscle I
-
~
3. I‘arious Antzgens Taylor atid Crilling RIorise (L-stlain), Splecn cells, B.\LB/c Cell countiiig ( 1963) guinea pig mice or guinea pigs imniritiized with L-cells fihroltlast s
Contitiired
130
PETER PERLMANN AND GORAN HOLM
TABLE I (Conlinucd Authors
Target cells
Effector cells
Method
Holtzer and Winlrler Human amnion Blood mononuclear cells, Monolayer (1968) (cell line), lymph node o r spleen guinea pig kidcells, guinea. pigs senriey cells sitizetl with human amnion cells MacLennan and Chang cells (hu- Lymph node cells, rats Isotope release ( W r ) immunized with Chang Loewi (1968b) man cell line) cells Perlmann et nl. Chicken erythro- Blood moiioriiiclear cells, Isotope release ("'Cr) spleen cells, guinea (1968, 19691)) cytes coated pigs immunized wit,h with purified protein derivaMyeobucteriu tttberculosis tive (PPD)
4. H umun Diseases Perlmanii and Human colon cells Ulcerat,ive colit,is. Blood Isotope release Broberger (1963) nionoiiuclear cells (W-amino acid, 32P)
Hedberg and KalIBn (1964)
Human fibroblasts R.heuniatoid arthritis, systemic I L I ~ U Serythematosiis, psoriatic arthropat,liy. hlononuclear cells from syiiovial fluid Berg and KallBn Rat, glia cells Miihiple sclerosis. Blood mononuclear cells (1964) Rraunst,einer el uI. Human amtiion R heiimat.oid arthrit,is. (1964) cells Lymph node cells Trayanova et al. Human fibrohlasls Systemic lupus eryt,he(1966) and kidney cells matosus. Blood mononuclear cells Watson el ul. (1966) Human colon cells Irlcerative colitis. Blood mononurlear cells Snkernick el a / . Humail fibrol)lasts Iiheiimat,oitl arthritis. (1968) 131ood monorruclear cells
Monolayer
Monolayer Monolayer hlonolayer
Monolayer h'lonolayer
5 . Anamal Tumors (Tumor-Spccijc fi'euctzons)
Itosenau and Morton (1966)
Mouse, methylSpleen cells, syngeneic cholanthrenemice immunized with induced sarcoma tumor i n C3I-I or C.57Bl
Cell counting
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitl.0
131
TABLE I (C‘ontinicetl)
Authors Alexander et 01. (1966h) Phillips (1967)
IIellstroni et al. (1968%)
Hellstroni et nl. (1969b) Heppner and Pierce (1969)
Target cells
bIouse letiltemiai Spleeii cells, dlogeneic cells of D13.412 mice ininirinizetl with origiii tiinlor Mouse, Sarcoma I 1,ymph node cells, C57Bl (A/J:ix origin) mice iinmiiriized with tiimor hIoilSe, iIlet.hyl1,yrnpfi node cells, cholaii t.tiretre o r t unior-t)earing alloplastic tlisc-ingeneic, syngeiieic, 01’ ducetl samiinas auit,ologviis mice in C3H or BhLB/c Mouse, Noloiiey Lymph node cells, t umoi-bearing synsarconia in BALB/c or C3H geneic mice Mouse, spoiitaLymph node cells, autolnenw mammary ogoris tiimor virusluniors carrying BAL13/c mice, foster-fed on C3H; or virus-free BX LR /c 6. Himuin Tiimors
Hellstrnrri rt ( 1 96x1))
rtl.
Hellstriiiri et a / . (196%)
Bubenik et al. (1969,)
Effector cells
Seiirolil:tst,oiiiat
(7‘li?tL(JI’-SpC’CL$C
Method Cell counting
Cell c o i i n h g
Coloiiy inhibition
Colony inhibil ioii
Colony inhibition
Keuctions)
Xritdogoiis or allogeiieic Colony irihiljitioii bli~odinononuclear cells Xut,ologr)iisor allogeiieic Coloiiy inhibition M)od mononuclear cells
Wilm’s tiimor, vnrioiis aclenocxrcinoin:ts :tiid sarcomits Bladder cnrciAutologoiis or allogeneic Cell coiinting iioma blood moiioiiiic1e:i.r (microniodificacells tion)
sitized to unrelated target cells, can be washed off. In contrast, lymphoid cells from donors sensitized to target cell antigens cannot easily be removed by washing (Rosenau and Moon, 1961; Koprowski and Fernandcs, 1962; Taylor and Culling, 1963; Wilson, 1963; Brondz, 1964; Vainio et al., 1964). An increased stickiness of lymphoid cells from donors injected with Freund’s complete adjuvant has been described (Koprowski and Fernandes, 1962; Holtzer and Winkler, 1968) but is not a rcgular finding. Aggregation usually precedes morphological changes and death of the target cells ( Rosenau and Moon, 1961; Koprowski and Fernandes, 1962).
132
PETER P E R L M A N N A N D GORAN HOLM
2. Target Cells
a. Nature and Properties Affecting tlie Cytotuxic Reaction. By definition, a common feature for the target cells discussed under this heading is their possession of antigens to which the donors of the lymphoid cells are sensitized. In all c a m studied thus far, these antigens arc supposed to be surface components. Thcy can be naturally occurring constituents of the target cell surfacc; antigens, foreign to the cclls and artificially attached, may also be effective (Perlmann and Holm, 1968; Perlmann et al., 1969b). This is of some importance since it simulates conditions which may occur in an organism in connection with delnycd hypersensitivity. Lymphoid cells from donors immunized to serum proteins in tlie incubation medium in which target cells were growing were not cytotoxic: to these cells (Rosenau and Moon, 1964; Taylor and Culling, 1965). (However, in Section III,D,I, it will be shown that antigens unrelated to target cell antigens under difEerent experimental conditions can induce cell-mediated cytotoxicity. ) A sufficiently high dcnsity of antigenic determinants on the cell surface is known to be essential €or the cytolytic reaction induced by humoral antibody and complement (Moller and Moller, 1962; Winn, 1962). Although there is some evidence that the density of antigenic receptors influences the course of cell-mediated cytotoxicity, other properties of the target cells and the experimental conditions are probably also of importance (Brunner et al., 1969b). b. Allogeneic Inhibition. From experiments to be discussed latcr (Moller, 1965a,b; Moller and Moller, 1965) (see Section III,D,2), it was suggested some years ago that in vitru destruction of antigenic target cells by lymphoid cells from sensitized donors constitutes a case of allogeneic inhibition. Thc latter is defined as a nonimmunological phenomenon in which cells may be killed through an unknown (possibly suicidal) mechanism when coming into close contact with cells of different genctic (i.e., allogcneic) origin (Hellstriim and Moller, 1965; K. E. Hellstrom and I. Hellstrom, 1967). Thus, although close contact between lymphoid cells from immune donors and antigenic target cells may be established by immunologically specific reccptors on the cell types, the actual killing would be due to confrontation of the target cells with foreign histocompatibility antigens on the lymphoid cells ( Moller, 1965a,b; Moller and Moller, 1965). However, since the time when this concept was advanced, several papers have been published in which lymphoid cells and target cells were from genetically compatible donors. Rosenau and Morton (1966), I. Hellstrom and K. E. Hellstrom (1967), Hellstrom
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
133
et al. ( 1968a), and Brunner et al. (1969b) have shown that experimentally induced tumors in micc arc destroyed in tumor-specific reactions by syngeneic lymphoid cells from scnsitized animals. Tumor-specific reactions in autologous combinations have recently also becn dcscribed for human tumors (Hellstrom et al., l W b , c ; Bubenik et al., 1969a). Other examples are organ-specific destruction of kidney or thyroid cells by syngeneic lymphoid cells from rats immunized with kidney or thyroid, respectively (Holm, 1967c; Biorklund, 1968), and of human colon cells by autologous lymphoid cells from patients with ulcerative colitis (Watson et al., 1966). Such findings do not rule out the possibility that differences in histocompatibility or, perhaps, other cell-type-specific differences in surface architecture may influence the course of lymphoid cell-target cell interactions. Most importantly, however, it is obvious &at lymphoid cells may become cytotoxic to other cclls within an organism in a strictly autologous situation. 3. Immunological Specificity and the Significance of Humoral Antibodies
a. Specificity. Different types of controls have been used to prove the immunological specificity of the cytotoxic reaction. Thus, lymphoid cells from tumor-bearing animals, or from animals immunized with X-irradiated tumors, have been shown to destroy tumor cells in a specific reaction. Tumor cells with noncross-reacting tumor antigens were not affected (Rosenau and Morton, 1966; I. Hellstrom and K. E. Hellstrorn, 1967; Hellstrom et al., 1968a). Similar results have recently been obtained with several human tumors. In nitro destruction of certain tumors has been obtained both with the patients’ own lymphoid cells and with those from other patients carrying a tumor of the same type. In contrast, these lymphoid cells did not affect cells from healthy organs of the type from which the tumors originated. Neither did they destroy cells from unrelated tumors (Hellstrbm et al., 1968b,c). Among other things, these findings provide convincing evidence for the occurrence of tumor-specific antigens and 3 corresponding immune response for a variety of human tumors, Unexpectedly, in view of what was heretofore known of animal tumors, these antigens seem to be both tumor and organ specific. For a full discussion of these problems thc recent review by Hellstrijm and Hellstrom (1969) should be consulted. In work with other immune systems (Table I ) , the specificity of the reactions has usually been controlled in a similar manner. A more detailed analysis of immunological specificity has thus far only been possible in those cases in which the target cell$ carried serologically and gcnetically well-defined transplantation antigens. Rosenau and Moon ( 1964) found
134
PETER PERLMANN AND GORAN HOLM
that lymphoid cells from BALB/c mice (genotype H-2d), immunized with spleen cells from C3H mice (H-2k), killed cells from established cell lines of C3H origin but did not affect fibroblasts originating from thc strains DBA-2 (H-2d) or C57B1 ( H-2b). The results were reversed when the lymphoid cells wcre from donor mice immunized with C57B1 spleen. Target cells of rat or human origin were not affected. Lundgrcii (1969) has reported similar results for the human HL-A system. A detailed analysis, within the H-2 locus of the mouse, was recently reported by Brondz ( 1968) who used methylcholanthrene-induced sarcoma cells as immunogen, lymph node cells as effector cells, and peritoneal macruphages as target cells. Serum from the sensitized mice was taken on the same day as the lymph node cells. Several strain combinations within the H-2 system were tested. In all cases, the resulting cell-mediated cytotoxic reaction was specific for the H-2 antigens of the target cells. Moreover, in order to be cytotoxic, the lymph node cells had to be taken from donors sensitized to all or almost all of the foreign H-2 antigens on the target cells, When the donors were sensitized to only half of the antigcns or less, thc targets were not killed, in spite of the fact that hcmagglutinating and cytotoxic antibodies were shown to be present in the serum of the lymph node donors. Rrunner et al. (1969b) have made similar fiiidings. No cytotoxic reaction could be produced by combining lymph node cells from two donors, each sensitized against a different fraction of the H-2 antigens on the target cells (Brondz, 1968). These findings establish a high degree of specificity of the cell-mediated reaction, with exacting requirements exceeding those of the serological reaction. Thcy may possibly indicate certain requirements of carrier specificity, resembling those operating in delayed hypersensitivity (for references, see Turk, 1967). The author assumed that, in order to trigger a cytotoxic reaction, the activc lymphoid cell must carry multivalent receptors, each receptor with specificity for all foreign H-2 antigens on the targct cells. Other more appealing explanations may he found, however. b. Hutnorcil Antibodies. It is often stated that the cytotoxic action of lymphoid cells from sensitized donors 011 antigenic target cells is an in wiiro expression of delayed hypersensitivity or a similar state rather than of a humoral antibody response. It is implied that cytotoxicity reflects an activity of sensitized cells rather than one of humoral antibody. This argument is based on several lines of evidence. Thus, the optimal time for harvesting active cclls from sensitized donors may coincide with the development of allograft rejection rather than with humoral antibody production (Brondz, 1964; Brunner et al., 196%). When humoral antibodies are present they may not he cytotoxic (Broberger and Perlmann,
CYTOTOXIC EFFECTS OF LYMPHOD CELLS
in Vitro
135
1963; IVilson, 196%; Holm, 1966; Hellstriim et al., 196Sa,b). However, in many cases no correlations liavc, bcen cst,il)lishcd and cytotoxic antibodies have sonietimes hcmi seen in high titcrs (Rosenau, 1963; Berg and Kallch, 1964, Hellstrbm et al., 196Sb,c, 1969b). In an alloimmunc>system (H-2) of the mouse, Biunner et al. (196Sa,b) found an cxccllent correlation between formation of 19 S (but not 7 S ) antibodies and thc appearance of cytotoxic cells. In the guinea pig, Perlmann et al. (1969b) found lymphoid cells cytotoxic to purified protein derivative (PPD) coated target cells cvcn when delayed hypersensitivity in the donors could not he detected by means of skin testing. Attempts to elute nntibody-like activity from lymphoid cells of sensitized donors ( Taylor and Culling, 1963; Wilson, 196%; Brondz and Bartova, 1966) or to confer activity on lymphoid cells fioin normal donors by exposure to antiserum in vitro have mostly been unsuccessful (Rosenau, 1963; Perlmann and Broberger, 1963; Brondz, 1964; Wilson, 1965a; Moller, 1965b); see, however, Section III,B. For cytotoxicity, humoral antibodies require participation of complement. In experiments with cell-mcdiated cytotoxicity, heat inactivated serum is usually included in the incubation mixtures and addition of complemciit has no effect ( Brondz, 1964; Wilson, l965a, Moller, 1965b; Rosenau, 196s). Exceptions to this rule have becn noted ( Govaerts, 1960; Perlmann and Broberger, 1963). There are also cases in which cytotoxic effects of lymphoid cells in uitro have been demonstrated in complete absence of serum (Rosenau and Moon, 1961; Rosenau, 1968). Other arguments for cell-mediated cytotoxicity being the expression of a cellular activity, independent of humoral antibody, are derived from observation of the inhibitory effects of the lattcr. Thus, treatment of mouse target cells with heat-inactivatcd alloantibodies against their H-2 antigens before or during cxposiuc to lymphoid cells from immunized allogeneic mice strongly reduced the cell-mediated cytotoxic effects ( Moller, 1965b; Brunner et al., 1967, 196Sa). The immunological specificity of this inhibition was clearly established. Moreover, only when a major part of the relevant antigens was blocked by antibody, the cellmediated rcaction w'is abolished (Mauel et al., 1969). Both 7 S and 19 S alloantibodies have been found to be protective ( Bruiiner et a]., 1968a). Similar although less clear-cut inhibitions have occasionally been observed in other imniiuie systems (Holm, 1967c; MacLeiirian and Loewi, 1968b). Most recently, an immunologically specific inhibition of cellmediated cytotoxicity in vitro by humoral antibodies has been described for several human and animal tumors ( Hellstriim et al., 1969a). Thesc in vitro results are indicative of an enhancement-like inhibition
136
PETER PERLMANN AND GORAN HOLM
(Kaliss, 1958) by antibodies on the efferent ( =effector) side of the immune response (Brunncr et al., 1968b). An afferent or central inhibition (Billingham et al., 1956) by humoral antibody on the inductive phase of the immune response may also be reflected by a reduction of cellmediated cytotoxicity in uitro. Brunner et al. (1968b, 1969a) found that humoral alloantibody ( mouse), passively administered in connection with immunization, reduced the in vitro cytotoxic effects of lymphoid cells harvested from the treated animals within 1 to 3 weeks after grafting. Both 7 S and 19 S antibodies were suppressive. However, formation of humoral antibodies ( 1 9 s ) in the treated animals was also strongly reduced. Since course and manifestation of the immune response reflect a balance between cellular and humoral factors, it is not really contradictory that inhibition by antibody of in vitro cytotoxicity does not represent a general finding. Thus, Brondz (1965) was not able to achieve inhibition of target cell killing by lymph node cells with alloantibodies within the H-2 system of the mouse. In some of the tumor-specific systems, recently described by Hellstrom et al. (1969a), humoral antibodies did not inhibit in vitro cytotoxicity of lymphoid cells on the tumor cells (colony inhibition). These “inactive” sera were from donors in which the tumors had regressed and which often contained antibodies which were cytotoxic to the cells in the presence of complement (Hellstrom et al., 1969b). In summary, cell-mediated cytotoxicity in vitro may reflect a state of delayed hypersensitivity or a related condition of a sensitized donor. However, from what has been said it is also clear that too far-reaching generalizations in this direction are unwarranted. That humoral antibodies in certain cases can induce in vitro cytotoxicity of normal lymphoid cells will be more fully discussed in Section II1,B. It would be surprising if such reactions would not also take place in some of the cases where the lymphoid cells are obtained from sensitized donors in which antibody formation takes place concomitantly. In this sense, the in vitro reactions are not less complex than are the in vivo manifestation of delayed hypersensitivity, tumor or allograft rejection, or tissue damage in autoimmunity.
4. Kinetics Target cell destruction by lymphoid cells generally proceeds slower than lysis induced by humoral antibody and complement. This does not by itself constitute a valid argument against participation of antibody and Complement in cell-mediated cytotoxicity. The rate of target cell destruction will depend on a variety of factors such as nature of target cells, immune state of the donor of the effector cells, the number of these
CYTOTOXIC EFFECTS
OF LYMPHOID CELLS
in Vitro
137
cells, and the assay system. The course of the reaction will also be dependent on the experimental conditions that affect the viability of the lymphoid cells. A certain excess of lymphoid cells over target cells seems always to be necessary to obtain measurable target cell destruction. The most complete kinetic analysis has first been reported by Wilson (1965a) who used suspensions of irradiatcd tumor cells (cell lines) from inbred rats. The lymphoid cells were from lymph nodes or thoracic duct of allogeneic rats, sensitized against the histocompatibility antigens of the target cells by means of skin allografts. The lymphoid cells werc added in excess and target cell destruction was evaluated by counting of nuclei (Section II,D,2). Detectable destruction of target cells was first noted after about 20 hours of incubation but was virtually complete at 50 hours. Target cell survival was inversely related to the number of Iymphoid cells added. The exponential relationship found was similar to a “single-hit” inactivation phenomenon, suggesting that one single “sensitized lymphocyte would suffice to affect adversely one target cell. It could also be extrapolated that 142%of the lymphoid cells added in these experiments were immunologically active ( i.e., destructive for target cells), This course of target cell destruction is typical for most of the reports in which cell-mediated cytotoxicity was studied under similar conditions. The influence of the assay system on the reaction rate is brought out by some experiments of Brunner et d. (1966), using mouse mastocytoma cells of the DBA/2 strain as targets and, as effectors, spleen cells of C57B1 mice, sensitized against DBA/Z. The assay system was either microscopic counting of surviving target cells, or measurement of their ability to form colonies in semisolid medium after contact with the effector cells (Section II,D,3). Like Wilson, the authors found an inverse and exponential relationship of target cell survival to number of spleen cells added. However, whereas it took from 24 to 4s hours to demonstrate a reaction by microscopic counting, this was demonstrable, after 3 hours and complete after 12 hours in the cloning assay. In later experiments (Brunner et nl., 196Sa,b) these authors compared the cloning assay with 51Cr-releasefrom labeled mastocytoma cclls. Interestingly, in these different assays, target cell destruction had very similar kinetics. In the same immune system with spleen cells taken at peak response after one immunizing injection, Brunner et al. (196913) have recently been able to achieve almost complete target cell destruction within 1 to 2 hours, when the effector cells were added to the mastocytoma cells at a ratio of 100:1. At lower ratios, a certain time lag was observed and the rate of target cell destruction decreased. However, even at lymphocyte/
138
PETER PERLMANN AND GORAN HOLM
target cell ratios as low as 1:1, up to 20%of the targets were lysed within 6 hours. The data suggested that one “sensitized effector cell killed more than one target cell. It was also shown that the rate of destruction was different for target cells of different types. Thus, at lymphocyte/ target cell ratios of 100: 1, embryonic fibroblasts in suspension or allogeneic spleen cells were destroyed at much lower rate than mastocytoma cells, in spite of the fact that the same H-2 antigens were involved.
5. Effector Celh
a. Origin and Activity of Lymphoid Cells. Spleen cells, lymph node cells, blood lymphocytes, or thoracic duct cells may vary somewhat in cytotoxicity when their effects are compared on the basis of equal numbers of lymphocytes. Moreover, activity levels in cell populations of different origin will also vary at different times after immunization (Wilson, 1965a; Moller, 1965b, Bruiiner et al., 1969b). Thymus cells from immunized donors are only slightly active or inactive ( Brunner et al., 1969b ) . It is assumed that the activity of a cell sample reflects the concentration therein of active cells rather than varying degrees of cytotoxic activity on the level of the single cells. In experimental animals, all authors have found peaks of activity at about 1 to 2 weeks after antigenic challenge (Brondz, 1964; Wilson, 1965; Brunner et al., 196913). Mode and intensity of immunization are decisive for the cytotoxic activity of the effector cells, with cells from heavily immunized donors exhibiting the strongest effects (Brunner et al., 196913; Perlmann et al., 1969b). The nature of the immunogen is also important, with living antigenic cells being more efficient than X-irradiated cells or extracts (Hellstrom et al., 1968a; Brunner et al., 1969b). This may be the reason for the finding of relatively active lymphoid cells in human tumor patients and in some autoimmune diseases. It may be noted that the highly efficient cytotoxicity described by Brunner et al. (1969b) was achieved by immunizing C57B1 mice with living mastocytoma cells of DBA/B origin. When DBA/2 spleen cells were the immunogen, the cytotoxic activity of the lymphoid cells, tested against several DBA/S targets, was much less. b. Lymphocytic Cells. The facts reported in the preceding paragraphs and the common microscopic picture of clustering of lymphocytes to target cells have led to the general assumption that lymphocytes also are cytotoxic effector cells in uitro. From the data published thus far it cannot be concluded which cells within the functionally heterogeneous populations of lymphocytes (Gowans and McGregor, 1965; Miller and Osoba, 1967) participate in in vitro cytotoxicity. From indirect evidence already referred to (Section III,A,4,b) the assumption may be made that
CYTOTOXIC EFFECTS OF L Y m w o m CELLS
in Vitro
139
the lymphocytes that trigger delayed hypersensitivity and similar conditions in vivo, or some of their descendants, may function as cytotoxic cffector cells in vitro. However, such indirect evidence does not help to resolve the question of the relative role in in vitro cytotoxicity of antigenreactive and thymus-derived lymphocytes as compared to that of bone marrow-derived precursors of antibody-producing cells ( Mitchell and Miller, 1968). It is not known, for instance, whether or not antibodyproducing plasma cells or their prccursors exhibit cytotoxic activities in vitro. Even if not directly cytotoxic by themselves, they may contribute indirectly by releasing small amounts of antibody capable of inducing cytotoxic activity of normal lymphocytes by one of the pathways described in the following paragraphs and in Section II1,B. c. The Contribution of Monocytes, Alacrophages, and Polymorphonuclear Leukocytes. At the present stage of our knowledge, the conclusions that lymphocytes are acting as effector cclls in in vitro cytotoxicity requires many qualifications. Thus, although some cell populations, such as those from thoracic duct, consist of 98 to 99% lymphocytes, most authors have made experiments with ccll populations in which lymphocyte concentrations may have varicd from 75 to 95%. Since the effector cells are added to the target cells in excess, sometimes at very high ratios (1000: 1 or more), a small fraction of active nonlymphocytic ceIls may very significantly contribute to target cell damage. There is evidence that macrophages from sensitized donors also are cytotoxic to target cells in vitro by phagocytic and “nonphagocytic” mechanisms ( Bennett et al., 1963; Old et nl., 1963; Granger and Weiser, 1964, 1966). It is likely that the same applies to monocytes of the pcriphcral blood. Whether or not polymorphonuclear leukocytes from sensitized donors may exhibit specific cytotoxicity in vitro is unknown. When antibody-producing cells are present, and this is the rule rather than the exception, cytotoxic activities of macrophages, monocytes, and polymorphonuclear leukocytes will most likely be enhanced. Interactions between lymphocytes and other cells cannot be disregarded in this context. Cellular interactions of this type have recently attracted growing attention in several relevant situations, such as in connection with the induction of humoral immunity by macrophages ( Fishman and Adler, 1963; Ford et al., 1966; Mitchison, 1968; Unanue and Askonas, 1968). I t has been shown that the reaction of a small number of sensitized lymphocytes with antigen will lead to profound changes in activity of macrophage populations by cmission of a soluble factor (David, 1966; Bloom and Bennett, 1966; Svcjcnr et al., 1968). Conversely, lc~ the reaction of a few niacrophagcs with antigrw sCm1s to be c a p ~ ~ b to
140
PETER PERLMANN AND GORAN HOLM
induce morphological transformation and DNA synthesis in lymphocytes (Cline and Swett, 1968). It is obvious that the relative role of these and similar interactions for the cytotoxicity of lymphoid cells on target cells in vitro requires further studies. 6. Cytotoxic Mechanisms a. Contact. It has already been described that a characteristic feature of cell-mediated cytotoxicity is the antigen-specific aggregation of lymphoid cells to target cells. Instrumental in bringing about contact are antibody-like receptors on some of the lymphoid cells. When contact between lymphoid cells and target cells is prevented, damage of the latter does not take place (Rosenau, 1963; Wilson, 1965a). When two target cell populations, carrying different antigens, are simultaneously exposed to lymphoid cells from a donor immunized against only one of the antigens, only the target cells carrying the relevant antigen are destroyed (Brunner, personal communication). Although these findings indicate that contact is necessary for the cytotoxic reaction they do not reveal its mechanism. Microscopic observations and time-lapse cinematography with target cells in monolayer shows that some lymphoid cells, after random movement, will become temporarily attached to the target cells which finally seem to undergo osmotic lysis (Rosenau, 1963). These studies do not prove whether or not lysis is actually produced by the attached cclls. Treatment of mouse lymphoid cells with alloantibodies against H-2 antigens that distinguished them from the target cells in a transplant a t'ion immune system did not inhibit their cytotoxic effect ( Mauel et al.. 1969). Thus, when the lymphoid cells are from sensitized donors, allogeneic inhibition i.e., contact of the targets with foreign alloantigens on the lymphoid cells, does not lead to target cell death (Section III,A,S,b). On the other hand, addition of heat-inactivated antilymphocyte serum, made in a diffcrent species, efficiently prevented the cytotoxicity of both mouse and guinea pig spleen cells (Mauel et al., 1969; Lundgren, 1969). It remains to be established whether this inhibitory effect was due to prevention of efficient contact bctween the cell types or to physiological inactivation of the lymphoid cells. A possible immunoglobulin nature of receptors on the lymphoid cells is suggested by the rcsults of Winkles and Arnason ( 1966). These authors inhibited the demyelinating activity in vitro of lymph node cells from rats immunized with nerve tissue b y treatment with rabbit anti-rat IgA. b. Phyyiologicnl State of Lymphoid Cells. Many authors have shown that the lymphoid cells have to be alive in order to be cytotoxic. No
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
141
specific cytotoxic effects have thus far been obtained with dead cells or with extracts or eluates of cells from sensitized donors (Brondz, 1964; Wilson, 1965a; Brondz and Bartova, 1966). Although the lymphoid cells may die during the experiment when the conditions are unfavorable for their survival, availahle evidence suggests that they are not killed in the course of thc cytotoxic reaction (Wilson, 1963). In Fig. 1 it is shown that the cytotoxic potential of a suspension of lymphoid cells from sensitized mice remained unchanged when the cells were added to a fresh batch of target cells after having killed the first one. Figure 1 demonstrates a good similarity of the kinetics of target cell destruction on thc two occasions. This speaks against death of cffector cells as being a major event, at least in short-term experiments. It possibly favors the
I
2
3
4
5
6
HOURS
FIG. 1. Cytotosic effect of lymphoid cells (spleen) from sensitized mice (C57BI anti-DBA/?) on srispensions of DBA/B mastocytoma cells, labeled with ”Cr. Target 1-100: specific isotope release (ordinate) during 2 hours of incubation of targets with an excess of 100 spleen cells per target cell. During this period most target cells were lysed. In parallel tubes, unlal,eled target cells were incubated with lymphoid cells under identical conditions. After 2 hours of incubation, the suspensions were ccntrifuged and the siiiiie niunber of fresh hut “‘Cr-labeled cells was now added. Isotope release from these cells is shown h y ciiii’e cksignated “Target 2-100.” Control-100: cytotoxic effect of a second sample of lymphoid cells from the same batch, first incubated for 2 hours without targets. The cells were then centrifuged and “Cr-labeled target cells were ndded as clcwxibed. Target 1-30 (etc. ) : parallel experiments with cells from the same batches but perforrned a t spleen cell/target cell ration of 30: 1. Tiiiie of experiiiient shown on abscissa. For other cxperimental details, see Brunner rt al. ( l968a). (Froni Brunncr ct nl., 1AliOa. )
142
PETER PERLMANN AND G ~ R A NHOLM
opposite supposition that reaction with antigen may enhance the cytotoxic potential of the effector population. If true, this could be due to either activation of previously "sensitized cells or recruitment. Some authors have succeeded in making normal lymphoid cells cytotoxic by exposing them to ribonucleic acid ( R N A ) or ribosomes from sensitized cells (Gerughty et al., 1966; Wilson and Wecker, 1966; Bondevik and Mannick, 196s). These preliminary results suggest one possible mechnnism for recruitment. Since the fraction of cytotoxically activc cells in a sensitized population is relatively large (Wilson, 1965a; Biunner et al., 1966), recruitment of originally inactive cells may play a role (MacLennan and Harding, 1969). However, more experiments are needed to establish this important point. Lymphoid cells from sensitized donors will aggregate to target cells at low temperatures, but undcr these conditions 110 destruction of the latter ensues (Wilson, 1967a). This reflects the fact that metabolic processes are necessary for the cytotoxic reaction. Experiments with metnbolic inhibitors support this notion. Antimycin A, suppressing electron transport in the respiratory chain (Chance and TVilliams, 1956) also impaired cytotodcity when applied to the lymphoid cclls (pretreatment) at concentrations that inhibit respiration without being toxic during the experimental period (Perlmann et al., 1969b). Inhibitors of RNA and protein synthesis, such as Imuran (Wilson, 1965b), actinomycin D, and cycloheximide (Brunncr et al., 1968a), inhibit cytotoxicity when applied at concentrations that blocked these metabolic activities without being toxic for the cells in the incubation mixture. However, inhibition with cycloheximide was partial and reversible. The residual activity ( approximately 50%) found after cycloheximide treatment was Explained as reflecting the presence in or on the cells of a pool of protein needed for cytotoxicity. When the lymphoid cells werc treated with low concentrations of trypsin, no inhibition was obtained in long-term experinients ( Brondz and Bartova, 1966). However, trypsin treatment abolished the cytotoxicity of lymphoid cells in short-term experiments ( Mauel et al., 1969). Inhibition by trypsin was only temporary and was coniplctely reversible. When trypsin treatment was followed by cycloheximide, cytotoxicity was strongly and irreversibly supprcssed. This suggested that protein, removed by trypsin, could not bc replaced at necessary concentration when protein synthesis was blocked. It may be that thc protein needed for the lytic reaction represents the antibody-like receptors, rcquired for the initiation of thc cytotoxic rcaction. Cytotoxicity of lymphoid cells has also Iwen suppressed by cthylcnc-
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitl.0
143
diaminetetracetic acid (EDTA) (Mauel et al., 1969). Inhibition was reversible and activity could be fully restored either by washing or by addition of Ca2+and Mg". It is not known whether this treatment affects contact or the lytic process. In long-term experiments (48 hours), X-ray at high doses ( 15,000 r ) seemed to be inhibitory, probably because it affected the viability of the lymphoid cells under the experimental conditions applied ( Rosenau and Moon, 1966). In short-term experiments, D N A synthesis and cell division are not required for the manifestation of cytotoxicity in uitro ( Maud et al., 1969). Contradictory results have been obtained by hydrocortisoiic treatment, probably because of differences in experimental conditions ( Rosenau and Moon, 1962; Mauel et al., 1969). The mode of action of this drug in this system is open to speculation.
7. Conclusions Cytotoxicity mediated by lymphoid cells from scnsitized donors may be viewed as a two-step phenomenon. The immunological specificity of target cell destruction resides in the first step in which contact between effector cells and targc t cells is established. A prerequisite for specificity is the presence of antibody-likc receptors on some cells in the effector cell population. Available evidence points to lymphocytes as cells possessing such receptors. It is assumed that these receptors also are manufactured by the lymphocytes on which ihey occur but this remains to be proven and is not necessarily a general rule. In any case, small amounts of released antibody may induce aggrcgation of cells, lymphocytes, and others, not originally cquipped with specific receptors synthesized by them (sec Sections III,B,C, and D ) . The results described above strongly support the notion that target cell lysis is brought about in a second step, following the contact established in the first step. There is no evidence that the antibody-like receptors of the first step also participate in the lytic reaction which probably lacks specificity in an immunological sense. It requires the participation of viable and metabolically active effector cells. It is possible that cytotoxic factors which nonspccifically inhibit target cell growth are released from the effector cells upon contact with target cell antigen. However, most of the evidence available to date suggests that cell-to-cell contact or at least close proximity between the ccll types is a principal requirement for the lytic step. This docs not exclude that target cell destruction is brought about by local relcnse of toxic substanccs, liberated from effector cells aggregated to the target cc~lls.Such mediators may be hydrolytic enzymes, components of thc coniplement system (Section III,C), or other factors. It is also possible that early changes in phospholipid metab-
144
PETER PERLMANN AND GORAN HOLM
olism lead to activation of detergent-like agents on thc surface of the effector cells (Fischcr et al., 1968). Entirely different mechanisms may be at play, however. Although phagocytosis does not seem to be a major event, target cell destruction may involve surface processes rclated to pinocytosis or phagocytosis. It remains to be cstablished whether the sensitized lymphocytes that initiate target cell destruction by reacting with antigen in the first step, also are executing the lytic reaction. Some of the microscopic and cinematographic studies wouId seem to suggest this. However, the importance of interactions between lymphocytes or between lymphocytes and other cell types in the effector population has already been pointcd out and cannot be overemphasized. Such interactions may lead to recruitment of originally nonsensitized lymphocytes or of other cell types and thereby actually trigger or amplify the lytic reaction. The question of participation of cells other than lymphocytes in the cytotoxic reaction bccomes particularly pertinent when relatively crude cell mixtures arc used as effector cells. When niacrophages, polymorphonuclear leukocytes, and nntibody-secreting cells are present, it can be safely assumed that target cell destruction may occur along a number of different pathways. Which of these will predominate may depend on the immune state of the cell donor, thc type of target cells used, and the assay system. OF CYTOTOXICITY OF LYMPHOID CELLSFROMNORMAL B. INDUCTION BY ANTIBODIESTO TARGET CELLANTIGENS DONORS
The cytotoxic reactions produced by lymphoid cells from donors sensitized to target cell antigens are immunologically specific and, therefore, require the presence of antibody-like receptors on some cells in the effector population. In spite of some negative results referred to in Section III,A,3,b, recent studies have provided evidence that humoral antibodies may induce cell-mediated cytotoxic reactions in &fro. In brief, antigenic target cells, treated with certain heat-inactivated antisera will be damaged when exposed to a modcrate excess of lymphoid cells from unsensitized donors. The antisera arc effective at dilutions too high to give rise to conventional complement-mediated lysis in the absence of lymphoid cells. 1. Description of Models a. General. Chromium-51-labeled chicken erythrocytes, coated with various protein antigcns, such as PPD or guinea pig thyroglobulin, have been used as target cells. Thc antigens were attached to the red cells by means of tannic acid treatment. When PPD was used (Perlmann and
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
145
Holm, 1968; Perlmann et d., 1969b), active sera were obtained from guinea pigs, immunized with Bxillus Calmette-Guerin ( BCG ) or killed tuberclc bacilli. The c ffector cells were spleen cells or blood lymphocytes from norinal guinea pigs, and cytotoxicity was assayed as isotope release. In the controls, the antisera n7erc replaced by heat-inactivated sera from normal guinea pigs. In othcr controls the lymphoid cells were left out or repIaced by kidney cells or red bIood cells. These as well as guinea pig thymus cells wcrc inactive. There was no species barrier; guinea pig lymphoid cells could be replaced by lymphoid cells from human peripheral blood or even from chicken, autologous to the erythrocytes used as targets. In contrast, blood lymphoid cells from patients with chronic lymphatic leukemia or Burkitt’s lymphoma cells wcre inactive. In thiF model, only sera from hyperimmunized animals wcre active. There was no correlation between their hemagglutinating titers towaid PPD-coated erythrocytcs and their activity in cell-mediated cytotoxicity. The antithyroglobulin sera were also from guinea pigs, immunized with guinea pis thyroglobulin in complete Freund’s adjuvant ( Wasserman and Packal6n. 1965). However, in these cases, one single injection was sufficicnt to produce highly active sera as well as thyroiditis within 3 to 4 weeks after challenge ( Wasserman et nl., 1969). In later work use was made of natural target cell antigens. The cells most widely used in the authors’ laboratory are Chang liver ceIIs (human ccll strain), grown in suspension cultures ( Holm and Perlmann, 1969a,b) and chicken erythrocytcs ( Pcrlinann and Perlmann, 1969). In both systems, antisera are from hyperimmunized rabbits and lymphoid cells from human periphcral blood or thoracic duct. Column-purified blood lymphocytcs are cytotoxic while thymiis cells and leukemic lymphocytes are inactive or only weakly active. Chang cell9 have also been used in a similar system by MacLennan and co-workers ( MacLennm and Loewi, 196%; MacLennan and Harding, 1969; MacLennan et al., 19G9a). These authors used lymphoid cells of blood, lymph node, and spleen origin from humans, rats or rabbits. Antisera were either from rabbits or from rats. Antibody-induced cytotoxicity was seen in some but not all combinations. High concentrations of normal serum or antiseiwn wcw inhibitory. Taylor and Culling (1968) have used mouse fibroblast treated with guinea pig antimouse serum and guinea pig spleen cells. Lundgren et nl. (1968b) used sheep fibroblast monolayers, cuposed to rabbit antishccp erythrocyte serum, and purified human blood lymphocytcs. Cytotoxicity was assayed by the plaque twhnique. In umnbcr of 1ium:un sera, RlncLcnnm et nl. (196911) recently found ‘1 factor \vhich rcndcrs Chang cclls susceptible to lysis by normal human lymphoid cells. The factor had the properties of an antibody (7s)
146
PETER PERLMANN AND GORAN HOLM
specific for Chang cells. Thus, in this system, all components of the reaction were of human origin although the Chang cell antigen involved may have been of viral origin. Bubenik et al. (1969b) have recently obtained similar results in work with inbred ducks. The antibodies were alloantibodies fioni the Campbell strain and were directed against the transplantation antigens of the Peking strain. Purified blood lymphocytes from normal Campbcll ducks lysed Peking targets in the presence of such antibodics. Both eiythrocytes and embryonic fibroblasts were destroyed. Lymphocytes from Campbell ducks tolerant to Peking transplantation antigens were as efficient as normal lymphocytes. b. Kinetics and Morphology. For positive cytotoxic reactions, lymphoid cells are added in numbers exceeding those of the target cells. Optimal ratios vary according to variations in antiserum, lymphoid cells, and target cells, but 25:l-5O:l seems to be optimal in several systems ( Holm and Perlmann, 196913; Perlmann and Perlmann, 1969). Under optimal conditions, isotope release from Chang cells or chicken erythrocytes can be observed within 1 or 2 hours and usually reaches completion ( 100%lysis) within 1 day. Purification of lymphocytes (human blood) affects the time course but not total. lysis. Microscopic observation of mixtures containing normal lymphoid cells, antibodies, and target cells renders results similar to those described for mixtures of lymphoid cells from sensitized donors and antigenic target cells. Thus, when target cells were tissue culture cells in monolayer, lymphocytes were seen to aggregate on their surface. With crude mixtures of blood leukocytes, aggregation was pronounced within a few hours. With purified lymphocytes it was delayed (Holm and Perlmann, 1969b). When both target cells and effector cells were of human origin, the antiserum was first exhaustively absorbed with lymphocytes from the donor providing the effector cells in the cytotoxicity experiment. This treatment neither abolished aggregation nor the cell-mediated cytotoxic effect. With chicken erythrocytes and purified human lymphocytes, erythrophagocytosis by what appears to be lymphocytes has been observed ( < 1%of the Iymphocytcs, with ncver more than one erythrocyte/lymphocyte) (Perlmann and Perlmann, 1969). With Chang livcr cells as target cells, phagocytosis has not been observed. 2. Cytotoxic Mechanisnis
a. Distinction from Complement-Znduced Lysis. Lysis brought about by humoral antibody and complcment is usually completed within less than 1 hour, whereas antibody induced cell-mediated lysis requires several hours even under optimal conditions. However, the validity of
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in V i t T O
147
0 1
0
I
I
I
i:io6
1:105
1:104
I
i:io3
I
i:io2
SERUM D I L U T I O N
FIG. 2. Cytotoxic effect of lymphoid cells from human blood on suspensions of antibody-treated Chang cells. The target cells were labeled with “Cr and treated for 30 minutes with heat-inactivated rabbit anti-Chang cell serum, diluted as indicated on the abscissa. After washing, 5 X 10’ Chang cells were incubated either with 10% fresh guinea pig sernni in the absence of lymphoid cells (squares), or with 125 X 10‘ lymphoid cells arid 10%heat-inactivated guinea pig serum (open circles). Filled circle: lymphoid cells added together with phytohemagglutinin. Ordinate: isotope release in samples, corrected for spontaneous release from the Chang cells (=26.4%) after 20 hours of inculmtion. (From Holm and Pcrlmann, 196Ha.)
such evidence for a distinction between lytic mechanisms may be debated. More important arguments can be based on quantitative considerations. Thus, extremely high dilutions of antiserum, not lytic with complement in the absence of lymphoid cells, are fully effective in the presence of the latter but in the absence of complement. An example is shown in Fig. 2. In this experiment, the heat-inactivated antiserum was from a rabbit, hyperimmunized with Chang cells (Holm and Perlmann, 1969a,b). Figure 2 illustrates the extreme cytotoxic potency of the serum in the presence of lymphoid cells from human blood. As a rule, strong cell-mediated cytotoxic effects with the lymphoid cells in moderate excess
148
PETER PERLMANN AND GORAN HOLM
can still be seen at serum dilutions from lo-" to lo-'. Similar results have been reported by MacLennan et al. (1969a). The same is found with chicken erythrocytcs pretreated with rabbit antiserum and exposed to highly purified human lymphocytes. In contrast, the complement-dependent titers of these antisera at optimal concentrations of either guinea pig or rabbit complement and measured after 3 or 20 hours of incubation never exceeded 1:500 ( Holm and Perlmann, 1969a,b; Perlmann and Perlmann, 1969; MacLennan et nl., 196%). There is other indirect evidence against common lytic mechanisms. Thus, some antisera known to be highly lytic in the presence of complement alone seem to bc quite inactive when tested with lymphoid cells. Rabbit antisheep hemolysin ( anti-Forssman ) at high dilutions lyses chicken erythrocytes with complement but only at very high concentrations with lymphoid cells (human peripheral blood) (Perlmann et al., 1969a; Perlmann and Perlmann, 1969). Since the antibodies in these sera are about 90%of the 19 S class, it may be suspected that complcmentfixing 19 S antibodies are not involved in the cell-mediated cytotoxic reaction. This emphasizes the similarity of cell-mediated cytotoxicity to the IgG-induced red cell sphering and fragmentation by mononuclear cells, recently clescribed by Lo Buglio et al. (1967). In the anti-Chang cell system, MacLennan et al. (1969a) werc able to separate rat antibodies by fractionation through Sephadex. Thc complement fixing lytic antibodies were predominately of macroglobulin nature but had little effect on Chang cells in the presence of lymphoid cells. Antibodies inducing cell-mediated cytotoxicity were predominately of 7S-type. The results support the notion that different cytotoxic mechanisms are operative in conventional complement dependent lysis and cell-mediated cytotoxicity. More work with immunoglobulins belonging to well-defined subclasses is, however, required to establish this. The data do not exclude a nonconventional participation of the complement system in cell-mediated cytotoxicity ( Section II1,C). b. Mode of Antiserum Action. Heat-inactivated antisera at high dilutions can be incorporated in the incubation mixture. Full effect is also obtained when the target cells are pretreated with the antiserum and are washed before incubation. With guinea pig spleen cells and guinea pig anti-PPD serum, it has also been possible to obtain cytotoxicity by first exposing the lymphoid cells to antiserum and adding them to the target cells after washing (Perlmann and Holm, 1968; Perlmann et al., 1969b); in other systems this did not work (Holm and Perlmann, 1969b). The positive results were obtained with spleen cells from guinea pigs, and it is possible that the cytotoxic cells in these reactions were macro-
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in V&O
149
phages, since thcw is good evidence that macrophages (and monocytes ) havc rcceptor sitcs for cytophilic antihoclics ( Beiken and Hcnaccrraf, 1966; Davey and Asherson, 1967). Cytophilic iunti1)otlics with affinities for lymphocytes are not known with certainty. Lo Buglio ct al. (1967) have provided some evidence that “certain lymphocytes” as wc~llas nionocytes specifically bind to the Fc portion of IgG attached to red cells. This binding leads to trapping, deformation, and damage of the red cells. It can be inhibited by IgG in the medium. When lyinphocytes arc the effector cells in the cytotoxic reaction, it is most likely that formation of antigenantibody complexes on the target cell surface provides the sites for induction of cytotoxicity. Available evidence, although limited, suggests that there is no correlation between cytotoxic or hemagglutinating titers of an antiserum and its cytotoxic potency in the cell-mediated test. It is possible, that the antibodies active in this system belong to a special immunoglobulin subclass. This problem has not bcen studied. c. The Zmportance of Contact. In Section III,A,f3, it has bcen shown that contact between lymphoid cells and antigenic target cells is an important feature of the cytotoxic reaction produced by lymphoid cells from sensitized donors. The microscopic appearance of antibody-induced reactions suggests that cell-to-cell contacts are also of importance in this system. When chicken erythrocytes, coated with eithcr PPD or thyroglobulin were mixed in equal proportion and were exposed to lymphoid cells in the presence of antibodies against one of the antigens, only the cells on which the antigen-antibody complexes werc formed became destroyed (Perlmann and Holm, 1969). A similar experiment is shown in Fig. 3. Chicken erythrocytes treated with antibody at high dilution were mixed with untreated chicken erythrocytes and exposed to purified human lymphocytes. When the treated erythrocytes were labeled with 51Cr,isotope release took place rapidly. When thc untreated erythrocytes were labeled, a slowly proceeding lysis first became apparent after several hours of incubation. It could bc assumed that this slow lysis was either due to release of a cytotoxic factor from the lymphocytes, or due to transfcr of antibodies from unlabeled to labeled cells. This latter explanation is strongly supported by thc second diagram in Fig. 3. In this experiment, the target cells were a 1: 1 mixture of chicken and duck erythrocytes. The rabbit anti-chicken erythrocyte antibodies did not cross react with the duck cells and no lysis was seen when the isotopic label was in the duck cells. The results were reversed when rabbit anti-duck antibodies were applied ( Per1m:inn and Perlmann, 1969). Very similar rcsults have been obtained with Chang cells (Holm and Perlmann, 1969b; MacLen-
150
PETER PERLMANN AND GORAN HOLM 0
(Cis
50
t
c )Ly
20
cq, t C )
? D
5
'0
10
20
30
40
50
Hours
Ham
FIG. 3. Cytotoxic effect of column purified lymphocytes from human blood on mixtures of antibody treated and untreated erythrocytes from chicken or ducks. Before incubation with lymphocytes, part of the chicken erythrocytes was treated with heat inactivated anti-chicken erythrocyte serum (rabbit), diluted 1:10'. The other part included in the target cell mixture (chicken or duck erythrocytes, respectively) was treated with heat inactivated normal rabbit serum. 2.5 X 10" lymphocytesjlo? erythrocytes were added after washing. Ordinates: percent isotope release. Abrcissus: hours of incubation. Filled circ2es: target cell mixtures consisting of one part antiserum treated "Cr-labeled chicken erythrocytes and, in upper diagram, one part unlabeled chicken erythrocytes, or, in lower diagram, unlabeled duck erythrocytes. Open circles: the same but with the "Cr-label in the erythrocytes which had not been treated with antiserum. Filled and open triangles: aliquots of the same target cell mixtures incubated without lymphocytes. ( From Perlmann and Perlmann, 1969. )
nan and Harding, 1969). Experiments of this type provide good evidence that antibody-induced cytotoxicity requires contact between effector cells and targets. Local participation of cytotoxic factors in the area of contact is, however, not excluded. d. Inhibition of Cytotoxicity. The lymphoid cells have to be alive in order to be cytotoxic. Inhibition of cytotoxicity by pretreatment of the effector cells with antimycin A indicates that the antibody-induced reaction also is dependent on energy-requiring effector processes ( Perlmann et al., 1969b). The antibody-induced cytotoxicity of normal lymphoid cells is as efficiently inhibited by xenogeneic antilymphocytic serum ( ALS) as that of lymphoid cells from sensitized donors (Section III,A,G,b). An example
CYTOTOXIC EFFECI'S OF LYMPHOID CELLS
in
VitTO
151
is shown in Fig. 4 in which lymphoid cells from human peripheral blood were pretreated with heat-inactivated rabbit ALS and the target cells ( Chang cells) with hcat-inactivated rabbit anti-Chang cell serum ( Holm and Perlmann, 1 9 6 9 ~ ) The . dilution of ALS needed for inhibition is proportional to the dilution of anti-Chang cell serum used. It is of interest that even those ALS conccntrations ( >1:25) that stimulated a major proportion of the lymphocytes in the effector populations to morphological transformation and D N A synthesis (James, 1967), completely inhibited the antibody-induced cytotoxicity (Holm and Perlmann, 1969~). This suggests but does not prove that ALS inhibits the reaction by affecting lymphocyte-target cell contact rather than by a metabolic block. However, it is not known whether the cells that are stimulated by ALS also are involved in the cytotoxic reaction. e. Effector Cekls. CeIl-mediated cytotoxicity, induced by antibody, is brought about by thoracic duct cells and by highly purified lymphocyte
.
NRS 1:5
ALS
13125
1:625
1:125
125
1:5
SERUM D I L U T I O N
FIG.4. Inhibition of lymphocyte-mediated cytotoxicity by antilymphocyte serum ( ALS). Chromium-51-labeled Chang cells were treated with rabbit anti-Chang cell serum diluted 1:lo' ( circles) or 1:10' (triangles ). After washing the target cells were incubated with column-purified human blood lymphocytes, pretreated with normal heat-inactivated rabbit serum or with heat-inactivated rabbit ALS, diluted as indicated on the abscissa. Ordinate: isotope release corrected for spontaneous release from Chang cells (27.4%)after 20 hours of incubation. For further explanations, see legend to Fig. 2. (From Holm and Perlmann, 1969c.)
152
PETER PERLMANN AND GORAN HOLM
preparations (e.g., Fig. 3 ) . Presence of monocytes, macrophages, and polymorphonuclear leukocytes will often speed up the cytotoxic reaction (Holm and Perlmann, 196913; Perlmann and Perlmann, 1969). These cell types are phagocytic for antibody-coated cells. They may become cytotoxic by themselves, although different targets will then vary in susceptibility. In addition to exhibiting direct cytotoxic effects, small numbers of nonlymphocytic cells may participate in in vitro cytotoxicity by interaction with lymphocytes as discussed in Scction III,A,5,c. 3. Conclusions
The antibody-induced cytotoxicity of lymphoid cells from nornial donors resembles that of lymphoid cells from sensitized donors. Certain antibodies are capable of inducing this reaction at very high dilutions and with highly purified lymphocytes as effector cells. Antibodies possessing this capacity may occur in xeno-, allo-, and autoimmune situations. Cytotoxicity does not require differences in histocompatibility between lymphoid cells and target cells. The reaction is most easily induced by antibody bound to antigen on the surface of the target cells which are destroyed. Since antigen-antibody complexes are known to stimulate lymphocytes to blast transformation and cell division ( BlochShtacher et al., 1968; Moller, 1969) and since malignant lymphocytes or thymus cells which are difficult to stimulate (for references, see Ling, 1968; Oppenheim, 1968) arc not cytotoxic, it can be inferred that blast transformation and cytotoxicity of lymphocytes are induced by the same basic reaction. This does not imply that blast transformation or cell division are necessary for cytotoxicity. The lytic reaction requires direct contact bctween effector cells and the antigenic target cells. Available data speak against the rclease of cytotoxic factors into the medium as being the cause of target cell destruction. The lytic process exerted by the effector cells is energy-dependent and may involve surface reactions such as increased motility. When the effector cells are purified lymphocytes, cytotoxicity is not due to phagocytosis but may be based on similar surface activities. It is most likely similar to thc recently described action of monocytes and “lymphocytic monocytes” on human red blood cells coated with IgG in the absence of complement (Lo Buglio et al., 1967). Although this lytic reaction is different from the conventional antibody-complement-induced lysis, the participation of complement cannot be ruled out ( Section II1,C). When present in the effector populations, white cells other than lymphocytes will contribute to target cell destruction. Lysis may then follow a course different from that outlined above.
CYTOTOSIC EFFECTS OF LYMPHOID CELLS
in Vitro
153
C. CYTOTOXIC EFFECTSOF LYMPHOIDCELLSTRIGGERED BY TARGET CELL-BOUND COMPLEMENT
1 . General Activation of the complcment system is necessary for cytolysis produced by humoral antibodies. The coinplement-fixing antibody initiates n relatively well-defined sequence of reactions which takc place on the surface of the target cell and which ultimatcly lcad to lysis. Once complemcnt activation has been triggered, antibody is no longer needed and lysis will be produced without its participation. Decisive for the lytic step is the activation of the terminal components C’8 and C’9 of the complement system (for references, see Miiller-Eberhard, 1968). The complement system comprises a series of interacting proteins. It is known that this system, in addition to its cytolytic role, has other biological functions. I t is well established that partial activation of complement, comprising only its first four components, leads to aggregation phenomena known as immune adherence (R. A. Nelson, 1953; D. S. Nelson, 1963). Instrumental in this reaction is cell surface-bound C’3 (Nishioka and Linscott. 1963). Cells with affinity (“receptor sites”) for activated C’3 will aggregate to those cells on which C’3 activation has taken place. Both macrophages and a certain fraction of lymphocytes have been shown to possess such receptor sites (Lay and Nussenzweig, 1968). Immune adherence of monocytes to C3-carrying erythrocytes is an efficient mechanism for erythrophagocytosis, promoted by xenogeneic 19s antibodies which have no opsoiiizing propertics in the absence of cornplemcnt ( Huber et a]., 1968). Other biologically important functions of the complcnient system are generation of anaphylatoxins, leucotactic factors, and histamine release from mast cells (for references, see MullerEberhard, 1968). The cell-mediated cytotoxic reactions described in Sections II1,A and B are always performed in medium containing heat-inactivated serum or, in some cases, no serum at all. Although there are a few exceptions (Section 111,A,3,11),addition of complement to a mixture of antigenic targct cells and lymphoid cclls from sensitized donors does not promote cytotoxicity. This has been takrn to indicate that ccll-mediated cytotoxicity has a different mechanism from that produced by humoral antibody complement. This argument can probably bc acccpted as valid, although the reservation must be madc th‘it small amounts of complement carried or produced by cells, or present in the heat-inactivated serum, inay contributr to cytotovicity wh(w smill amounts of complcment-fixing anti-
+
154
PETER PERLMANN AND GORAN HOLM
bodies are also present or produced. Howevcr, in view of the complexib of the complement system, the important question is whether C’ participates in cell-mediated lysis by pathways different from those operativc in the conventional lytic reaction. The results described below suggest that this may be the case.
2. Cell-Mediated Lysis of Chicken Red Blood Cells Carrying Partially Activated Coinplenient a. Target Cells Carrying C’3. In these experiments, Wr-labeled chicken red blood cells were used as target cells (Perlmann et al., 1969a) The cells were treated with small amounts of heat-inactivated rabbit antisheep hemolysin, reacting with Forssman antigen common for sheep and chicken eiythrocytes. At this concentration the antibodies, consisting primarily of 19 S immunoglobulin, were not capable to produce cellmediated lysis by themselves ( Section III,B,2,a). The antibody-carrying cells were then treated sequentially with isolated and purified components of human complement and were finally exposed to an excess of lymphoid cells from hunian peripheral blood ( 2 5 lymphocytes per target cell). Target cells treated with antibody alone or with the first three complement components (C’l, C’2, C’4) ( Muller-Eberhard et al., 1966; Polley and Muller-Eberhard, 1968; Cooper and Muller-Eberhard, 1968) were not lysed within 24 hours after addition of the lymphoid cells. The isotope release from such erythrocytes was low and indistinguishable from that seen in the controls to which no effector cells were added. However, whcm C’3 was also bound to the target cells, the lymphoid cells promoted an extensive lysis. This cytotoxic reaction required 10-20 hours to reach completion and was only seen when the effector cells were viable. It is based on energy-dependent effector processes, since pretreatment of the lymphoid cells with antimycin A, which blocks respiration, also inhibited the cytotoxic reaction. About 75%of the effcctor cells used in these experiments were lymphocytes. Fractionation into two fractions-one containing 80% monocytes and approximately 20% lymphocytes (Bennett and Cohn, 1966) and, a second, consisting of glass-bead purified lymphocytes ( Rabinowitz. 1964)-indicated that the monocyte-enriched fraction was highly cytotoxic. In contrast, the purified lymphocytes wcre not cytotoxic ( MiillcrEbcrhard et al., 1969). This suggested that the effector cells in these cxperiments were monocytes. However, since about 20% of the effector cells were lymphocytes, R lynq~hocyte-monocyte interaction in the cytotoxic reaction cannot be cntirely excluded. Cytotoxicity seemed to he due to a nonphagocytic process ( Miiller-Eberhard et al., 1969).
CYI’OTOSIC EFFECTS OF LYMPHOID CELLS
in V i t T O
155
b. Turget Cells Carrying C’7. Isotope-labeled chicken erythrocytes treated with antibody and the first four complement components, were brought in the C’7 stage by addition of isolated C 5 (Nilsson arid MullerEberhard, 1965) and functionally purified C’6 and C’7 (Nilsson and Miiller-Eberhard, 1967). In these experiments the fetal calf serum ( 5 % ) in the medium was fractionated chromatographically in order to remove small amounts of C’8 and C’9 activities, normally present even after heat inactivation. The C’7-carrying erythrocytes were completely lysed within a few hours when exposed to the lymphoid cells (Perlmann et al., 1969a; Miiller-Eberhard et al., 1969). No lysis occurred in the controls even when the C’7-carrying cells were artifically aggregated by means of phytohemagglutinin ( P H A ) . Lysis was not due to small amounts of C’8 bound to human erythrocytes in the effector population. It required viable effector cells and was also inhibitcd by antimycin A. In contrast, inhibition of protein synthesis in the effcctor cells by means of puromycin had no blocking effect. In this system both glass-bead-purified
/
g LL
0
TES 5
10 HOURS AT 3 7 O
15
Frc. 5. Damage of complement-target cell coinplexes by human blood lymphocytes, piirified on glass-bead columns. Target cells consisted of “Cr-labeled chicken erythrocytes, first treated with diluted, rabbit, 19 S antibody to boiled sheep erytlirocyte stroiiiata and, subsequently, with purified Iirtinan complement components. Inc~ilmtion inistiires contained 8 X 10’‘ lymphocytes and 1 X 105 target cells. Ordiii;rte: percent isotope release. Abscissa: time of incubation. EAC‘1-4, EAC’1-3, EAC’1-7: target cell~omplement complexes Imild up by sequential addition of components up to C’4, C’3, ant1 C’7, respectively. Black symbols: incubation mixture containing Iymphocytcs. Open symbols: inculxition mixtiires containing 1 X 10’ unlabeled and untreated chicken erythrocytes instead of lyinphocytes. ( From MullerEberhard et al., 1969. )
156
PETER PERLMANN AND GORAN HOLM
lymphocytes and the monocyte-enriched fractions were cytotoxic (Fig. 5 ) . The velocity of this cytotoxic reaction resembles that produced by addition of C’8 to red cells in the C’7 stage. It is, however, considerably slower than lysis produced by adding both C’8 and C’9. The results can be interpreted to mean that some cells in the effector population carry (or produce) C’8, instrumental as the lytic factor in this cell-mediated cytotoxic reaction. Some preliminary experiments suggest that a thermolabile substance, functionally similar to C’8, can be extracted from human blood lymphocytes and lyse red cells in the C’7 stage (Miiller-Eberhard et al., 1969). 3. Conclusions
These results are limited to a single type of target cells, and presently it is not known if other cell types would behave in a similar way. However, the data indicate that complement in some situations is of importance for induction of cell-mediated cytotoxicity. The two reactions described above probably represent two entirely indcpendent modes of complement action. The cell-mediated destruction of C’3-carrying target cells is related to phagocytosis, following a step of immune adherence. This partial activation of the complement system may be an important mechanism for cell-mediated cytotoxicity in uiuo. The lymphocyte- and monocyte-mediated destruction of target cells carrying C’7 is believed to reflect a different phenomenon. If it is correct that C’8 is the lytic factor in this model, cytotoxicity will be triggered by target cell-bound C’7 (for references, see Miiller-Eberhard, 1968). It may be asked whether C’8, provided by the effector cells, could also act as a cytolytic factor in the cytotoxicity models described in other sections of this chapter. This would either require a buildup of C’8-susceptible sites on the taget cells by an unknown pathway or a direct activation of effector cell-bound C’8. In both cases, the conventional sequence of reactions initiated by antibody would have to bc by-passed. In preliminary experiments, it has been seen that both the antibody and the PHAinduced cytotoxicity of human lymphoid cells (Section III,B,D) may be inhibited by rabbit antiserum to human C’8 and C’5 but not by antiserum to C’lq, C’2, C’4, and C’3 (data to be published). If confirmed by further experiments, activation of C’8 could actually be assumed to be of some general importance for cell-mediated cytotoxicity.
D. NONSPECIFICCYTOTOXICITY OF LYMPHOID CELLSACTIVATED BY PHYTOHEMAGGLUTININ OR OTHERSTIMULANTS It was found some years ago (Holm et al., 1964) that nonspecific aggrcJgation by PHA of human or aninial lymphoid cells to target cells in
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
157
tissue culture led to destruction of the latter. Later studies showed that cytotoxic effects of lymphoid cells from normal donors can be brought about by a variety of nonagglutinating agents, known to stimulate lymphocytes to blast transformation and DNA synthesis. Therefore, “activation” of some cells in the effector population can be considered as the most important basis for cytotoxicity in this model. Lymphoid cells activated by these stimulating agents are destructive to many kinds of target cells including cells originating from the donor.
1 . Diferent Stiniulants a. Reactions Induced by PHA. When lymphoid cells are added to tissue culture cells in monolayer in the presence of PHA, they become aggregated and cannot be removed by washing. The microscopic picture is indistinguishable from that seen after addition of lymphoid cells from sensitized donors to antigenic target cells (Holm et al., 1964). Aggregation is later followed by destruction of the monolayers. Similar reactions have been observed with tissue culture cells from rats and mice (Holm and Perlmann, 1965; Moller and Moller, 1965; Rosenau, 1968; Malchow et al., 1969), tumor cells from mice (Moller, 1965a,b), human fibroblasts (Moller et al., 1966; Lundgren and Moller, 1969) and human tumor cells (Chu et al., 1967), established human cell lines in suspension ( Holm and Perlmann, 1967a,b; Mac Lennan and Loewi, 1968a,b), and with chicken red blood cells (Perlmann et aZ., 1968). Human red blood cells, target lymphocytes, and some human lymphoma cells seem to be relatively resistant to PHA-induced cytotoxicity (Holm, 1967a; Perlmann et al., 1968). The cytotoxic activity of lymphoid cells can also be induced by pretreatment with PHA before adding them to the target cells. The PHA which is added in these experiments is a mixture, consisting of proteins with both hemagglutinating, leukoagglutinating and mitogenic properties (for references, see Naspitz and Richter, 1968). Absorption of the hemagglutinating substances does not abolish mitogenicity (Nordman et al., 1964; Robbins, 1964) and, likewise, does not remove the principle inducing cytotoxicity. The latter is destroyed by heating to 100°C. for 5 minutes (Perlmann et al., 1968). Phytohemagglutinin induces blast transformation and DNA synthesis in a large fraction of the lymphocytes from the peripheral blood of humans and animals. Under optimal conditions, DNA synthesis in human lymphocytes reaches a first peak within 2 to 3 days of culture. The response to PHA is dose dependent and decreases sharply at too high concentrations of the drug (for references, see Ling, 1968). When cytotoxicity of human or chicken lymphoid cells was studied as a function of PHA concentration, there was a good correlation between the dose de-
158
PETEX PERLMANN AND GORAN HOLM
pendency of the cytotoxic reaction and that of DNA synthesis (Holm and Perlmann, 1967b; Perlmann et al., 1968). However, there was no correlation in time, since cytotoxicity was close to its optimum at a time when DNA synthesis had hardly started. This indicates that cytotoxicity and DNA synthcsis are independent expressions of lymphocyte stimulation, Lymphoid cells from some animals such as mice do not respond as well to PHA as do those of humans, guinea pigs, and other animals (Ling, 1968). When special precautions are taken, blast transformation and DNA synthesis may be achieved but the number of responding cells is lower than that found in humans. This is probably one major reason for some of the conflicting results which have been reported for PHAinduced cytotoxicity of lymphoid cells in mice (see below, Section III,D,2). b. Other “Nonspecific” Mitogens. Stimulation of lymphocytes by PHA is assumed to be mediated by immunologically nonspecific cell receptors (for references, see Ling, 1968).Another lymphocyte-stimulating agent, believed to act via similar pathways as PHA, is present in culture filtrate from staphylococci (Ling et al., 19%). This agent induces cytotoxicity of lymphoid cells from human blood to tissue culture cells in monolayer and in suspension ( Holm and Perlmann, 196%).The staphylococcal filtrate induces blast transformation and DNA synthesis in a large fraction of the human circulating lymphocytes but is nonagglutinating; its action is slightly slower than that of PHA. Pretreatment of human lymphocytes with this agent for several hours or longer (2 days) produced strongly cytotoxic cells ( Holm and Perlmann, 1967b). Cytotoxicity was well correlated to the degree of blast transformation induced by the staphylococcal filtrate. It was also noted that the staphylococcal filtrate and PHA had additive effects on cytotoxicity. c. Antigens Unrelated to Target Cell Antigens. Lymphocyte stimulation in vitro to blast transformation and DNA synthesis is induced by antigens to which the donor of the lymphocytes is sensitized (for refer-. ences, see Ling, 1968). By definition, this stimulation requires presence of antibody-like recognition factors on some cells in the lymphocyte population. Antigen-induced stimulation of human peripheral blood lymphocytes proceeds at a slower rate than that induced by PHA and compriscs a smaller fraction of the cells (Ling and Holt, 1967). After pretreatment with PPD, human lymphoid cells from blood of tuberculin-positive donors became cytotoxic to allogeneic tissue culture cells. Cytotoxicity was correlated to the degree of stimulation achieved by the antigen in parallel incubations (Holm and Perlmann, 1967b).
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
159
Since there was no reason to believe that an immunological relationship existed between PPD and the target cells, it was concluded that cytotoxicity was iiii1iiui7ologically nonspecific. Similar results by Ruddle and Waksmxn (1967, 1968a,b,c) confirm and extend these findings. These authors used lymph node cells from rats sensitized to various antigens such as PPD, bovine ./-globulin, or ovalbumin. Treatment of the lymph node cells with antigen induced a cytotoxic activity when they were added to monolayers of rat fibroblasts. Destruction of the latter, recorded by the plaque technique or by cell counting, was first seen after 48 hours and was maximal at 72 hours. The destructive activity was specific for the sensitizing antigen and both allogeneic and syngeneic monolayers were destroyed. Short preincubation of the lymph node cells with antigen was sufficient to produce cytotoxicity. Similar results have also been reported by Lundgren et al. (1968b) who exposed to Salmonella vaccine, purified human lymphocytes from donors sensitized to Salmonella. Lundgren and Moller (1969) also obtained cytotoxicity with human lymphocytes and Streptolysin 0, an antigen against which most humans are sensitized (Ling, 1968). These authors measured plaque formation on autologous or allogeneic fibroblast monolayers. d. Mixed Leukocyte Culture. Mixed culture of lymphoid cells from humans (Bain et al., 1964) or experimental animals (for references, see Dutton, 1967; Wilson and Billingham, 1968) will lead to blast transformation and DNA synthesis. The degree of this stimulation correlates well with differences in the major histocompatibility antigens of the cell donors (Dutton, 1965; Wilson, 1967b; Wilson et al., 1967; Albertini and Bach, 1968). Preincubation of allogeneic lymphoid cells from human peripheral blood in mixed culture for up to 1 week and subsequent transfer of the mixtures to unrelated human tissue culture cells led to lysis of the latter. Lysis was correlated to the degree of blast transformation ( Holm and Perlmann, 1967b). 2. Allogeneic Inhibition
From experiments with target cells and lymphoid cells from normal mice, some authors concluded that the PHA-induced cytotoxicity was an in vitro manifestation of allogeneic inhibition, assumed to be a nonimmunological surveillance mechanism in uiuo (K. E. Hellstrom and I. Hellstrom, 1967). As already stated (Section III,A,2,b), allogeneic inhibition was originally defined as the passive destruction of cells exposed to foreign histocompatibility antigens. Allogeneic inhibition probably
160
PETER PERLMANN AND GORAN HOLM
represents a special case of the phenomenon known as contact inhibition of growth (Eagle, 1965; Eagle and Levine, 1967). This means that it is probably not limited to differences in histocompatibility antigens, although such differences may give rise to more pronounced effects than other smaller differences in cell surface structure or cell surface physiology. In brief, the following evidence has mostly been given in support of this hypothesis; 1. Mouse tumor cells or fibroblasts in tissue culture were destroyed when exposed to lymphoid cells or lymphoma cells from nonimmune donors of allogeneic or F, hybrid origin. Exposure to syngeneic cells seemed to have no effect (Moller, 1965a; Hellstrom et al., 1965, 1967). In order to obtain destruction, the cells were treated with PHA or rabbit antimouse serum. This was done to bring the cell types into close contact. Since F, hybrid lymphoid cells or some of the tumor cells were expected not to react “immunologically” against parental strain antigens, allogeneic inhibition was implicated. 2. Similar cell destructive effects were obtained by exposing mice tumor cells in tissue culture to X-irradiated lymphoid cells from F, hybrids or to extracts of allogeneic or F, hybrid origin (Hellstrom et aZ., 1964, 1967; Moller and Moller, 1965). 3. Cytotoxicity was reduced when the experiments were made in the presence of alloantibodies to the H-2 antigens on the lymphoid cells ( Moller, 1967). Experiments with other material have since shown that allogeneic inhibition and PHA-induced cytotoxicity are separate phenomena. Thus, in the presence of PHA, chicken red blood cells were destroyed by autologous and allogeneic lymphoid cells. No quantitative differenccs in target cell lysis were seen over a wide range of experimental conditions Perlmann et al., 1968). Rapid destruction of human fibroblast monolayers was also obtained with PHA-treated and purified human blood lymphocytes from both autologous and allogeneic donors (Lundgren and Moller, 1969). However, in some cases, in which the effector cells were relatively crude mixtures of human blood cells, allogeneic cells seemed to be more efficient than autologous cells (Mdler et al., 1966; Chu et al., 1967). Malchow et al. (1969) have recently made similar observations with cells from rats and mice. The destructive action of antigen-activated lymph node cells from rats affected both syngeneic and allogeneic monolayers ( Ruddle and Waksman, 1967,1968a). There may be a number of explanations for the differences in results obtained with different materials. The most obvious reasons for the ap-
CYTOTOXIC EFFECTS OF LYMPHOLD CELLS
in VitlO
161
pearance of allogeneic inhibition effects in experiments with mice were> ( 1 ) rcduced viability of both target cells and lymphoid cells during incubation, ( 2 ) marginal response of the latter to PHA ( Albller, 1965a,b), and ( 3 ) choice of syngeiieic lymphoid cells as the base line for the determination of the allogeneic or F, effects (Hcllstrom et al., 1967). Target cell destruction due to allogencic inhibition was weak and most likely due to reactions different from those obtained with lymphoid cells which are easy to stimulate with PIIA. Recent quantitative experiments by Malchow et al. (1969) may serve to illustrate this point. These authors found that addition of syngeneic lymphoid cells to rat or mice targets gave clear-cut cytotoxic effects after PHA treatment. Effector cells from F, hybrids had slightly stronger effects in mice but not in rats. The cytotoxicity of allogeiieic lymphoid cells was slightly stronger than that of syngeneic cells in both cases. Thcse results suggest that differences in cell surface architecture may affect the interaction between lymphoid cells and targets in a quantitative rather than in a qualitative manner.
3. Effector Cells a. Lymphocytic Cells. In the model involving PHA, cytotoxicity is obtained with those populations of lymphoid cells that are easily stimulated to blast transformation and DNA synthesis. Cells which are d a c u l t to stimulate (for references, sec Ling, 1968) were only weakly cytotoxic. This was seen with both human and rat thymus cells (Holm and Perlmann, 1965; Holm, 1967d; Perlmann et al., 1968). Blood lymphocytes from patients with chronic lymphatic leukemia were not cytotoxic under conditions where normal blood lymphocytes had strong effects. Lymphoid cells from blood of patients with Hodgkin’s disease had a reduced cytotoxicity, paralleling a reduced reactivity in skin tests with PPD (Holm et al., 1967). Burkitt lymphoma cells, which have the morphological appearance of transformed lymphoid cells, were not cytotoxic (Perlmann et at., 1968). It is not definitely established which of the functionally heterogeneous lymphocytes from the circulation or the lymphoid organs respond by transformation and proliferation when exposed to PHA or antigen. There is some evidence that the lymphocytes responding to PHA are thymus dependent (Meuwissen et al., 1968; Greaves et d., 1968). Recently, Tursi et a!. (1969) collected lymphoid cells from mice treated with ALS in vivo. These lymphocytes were not cytotoxic when added to target cells in tissue culture in the presence of PHA. This parallels the loss of graft-versus-host reactivity of ALS-treated mice as seen in similar experiments (Boak et al., 1967; Naysmith and James 1968).
162
PETER PERLMANN AND GORAN HOLM
Thus, available evidence suggests that the PHA- and thymus-dependent lymphocytes also trigger in vitro cytotoxicity. This does not imply that they also are the effectors of the lytic reaction. Stimulation by antigen of lymphocytes from sensitized donors often reflects a state of delayed hypersensitivity and is thus believed to be typical for those cells that participate in its induction (Mills, 1966; Oppenheim et al., 1967; Oppenheim, 1968). Antigen-induced cytotoxicity to fibroblasts of lymph node cells from rats sensitized to various proteins has been shown to be correlated to delayed hypersensitivity rather than to Arthus reactivity (Ruddle and Waksman, 1968b). Moreover, in experiments with hapten-conjugated proteins, skin reactivity as well as cytotoxicity in vitro could only be elicited with the hapten bound to the carrier protein used for sensitization. Although these results suggest a relationship between delayed hypersensitivity in vivo and antigen-induced cytotoxicity in vitro, it is not clear whether this is a general rule. Antigeninduced transformation has in some cases been seen to be correlated to antibody formation rather than to delayed hypersensitivity ( Oppenheim, 1968). It is not known whether or not antibody-forming cells or their precursors also are cytotoxic in vitro when exposed to antigen or other stimulants. b. Monocytes, Macrophages, Polymorphonuclear Leukocytes. In the controls of the cytotoxicity experiments, red blood cells or tissue culture cells were sometimes included as “effector cells.” The results have always been negative (MiilIer and Moller, 1965; Holm, 1967a; Perlmann et at., 1968). In PHA-induced cytotoxicity, thoracic duct cells and purified lymphocytes from the peripheral blood were cytotoxic for tissue culture cells ( Holm et al., 1964; Lundgren et al., 1968a; Holm and Perlmann, 196913). Admixture of polymorphonuclear leukocytes, monocytes, or macrophages will affect the course of the interactions in different ways. Human Chang liver cells in suspension seemed to be relatively resistant to polymorphonuclear Ieukocytes (Holm and Perlmann, 1967a, 196%). When human skin fibroblasts were exposed to human peripheral blood lymphocytes contaminated with a few percent of polymorphonuclear leukocytes and monocytes, monolayer destruction took place rapidly regardless of whether or not PHA was present (Lundgren et al., 1968b). Release of toxic substances (hydrolytic enzymes) from adversely affected polymorphonuclear leukocytes was assumed to contribute to monolayer destruction (see Section 11,CJ). Chicken red blood cells, exposed to effector cells consisting of 80-9m polymorphonuclear leukocytes and 1020% mononuclear cells (monocytes and lymphocytes), were rapidly lysed by these cells in the presence but not in the absence of PHA. Kinetically,
CYTOTOXIC EFFECTS OF LYMPIIOID CELLS
in Vitro
163
this lysis was quitc different from that obtained with highly purified lymphocytes and crythrophagocytosis was pronounced ( Perlmann and Perlmann, 1969). How PHA produced these effects is unknown. Although PHA does not induce DNA synthesis or mitosis in polymorphonuclear leukocytes or monocytes, it has been shown to activate metabolic processes in their nuclci (Killander and Rigler, 1965; Darzynkievicz et d., 1969). Whcn agents other than PHA are used for activation, the importance of contaminating cells has not as vet been established. Addition of antigen to the cells from a sensitized donor can be expected to release various cytotoxic activities of both polymorphonuclear leukocytes and macrophages. The question of interaction between lymphocytes and other white cells must not be overlooked. Recent evidence suggests that stimulation of lymphocytes to transformation and mitosis in many situations may be enhanced by other leukocytes (Gordon, 1968: Oppenheim et al., 1968; Hersh and Harris, 1968). Interaction of allogeneic lymphocytes in mixed culture in citro in heat-inactivated serum has bcen shown to lead to the production of leukotactic ( Ramseier, 1967) and macrophage inactivating factors (David, 1966; Bloom and Bennett, 1966; Svejcar et al., 1968). Such interactions may also be of significance for the cytotoxic reactions discussed in this section.
4 . Kinetics The rate of target ccll destruction by activated lymphoid cells is roughly similar to that described for the previous models. It varies with the mode of assay and the naturc of the target cells. With tissue culture targets in monolayers, visible effects may appear within a day, but optimal effects are mostly not scen until after several days (Lundgren and Moller, 1969). An excess of lymphoid cells over target cells is required. Quantitative studies have so far only been performed with PHA-stimulated lymphoid cells froin human peripheral blood and "Cr-labeled Chang liver cells in suspension. Under optimal conditions, significant isotope release could be observed after a few hours and lysis was usually completed within 24 hours. Howevcr, lymphoid cells from different donors may show considerable individual variations ( Holm and Perlmann, 1967a). Lysis at a fixed initial concentration of Chang cells increased signioidally with the log of lymphoid cell concentration-50% lysis at 24 hours required lymphocyte/target cell ratios from 5 : l to 25:l.The results seemed to suggest that a considerable proportion of the effector cells was active and capable of killing target cells within the experimrntal
I64
PETER PERLMANN AND GORAN HOLM
pcriods. However, these data do not exclude that target cell damage is brought about by release of toxic agents or lymphocyte-activating factors from a small number of originally active cells.
5. Cytotoxic Mechanisms a. Contact. When contact between lymphoid cells and target cells was prevented by Millipore membranes, target cell damage did not take place (Holm et al., 1964). On the other hand, mixed aggregation of lymphoid cells and target cells is not by itself detrimental to the latter. No cytotoxic effects were seen when aggregation was brought about by polylysine (Rosenau, 1963; Holm and Perlmann, 1965) which docs not stimulate lymphocytes. Light- and electron-microscopic observations ( Biberfeld P L al., 1968'I and time-lapse cinematography ( A x et al., 1968) of the PHA-induced interaction of lymphoid cells with tissuc culture cells indicated that the aggregated lymphocytes were highly mobile. Adhesion of lymphocytes to target cells by their uropod (McFarland and Heilman, 1965; MacFarland et al., 1966) was relatively rare. Emperipolesis (Pulvertaft, 1960) was not seen. Detachment of target cells from monolayer and/or lysis was observed before lymphocyte transforniation took place. z?. Physiological State of Efector Cells and lnhibitor Studies. Lymphoid cells from human (Holm, 1967a) or chicken blood (Perlmann et al., 1968) do not die in the course of the PHA-induced cytotoxic reaction. As in the previous models, lymphoid cells must be viable in order to produce cytotoxic effects. Cells killed by heat or extracts of normal lymphoid cclls were not cytotoxic (Holm et al., 1964; Holm ant1 Perlmann, 1965; Holm, 1967b; Lundgren and Moller, 1969) . X-irradiation and treatment with hydrocortisone (Moller et al., 1966) or treatment with inhibitors suppressing nucleic acid or protein synthesis did not affect PHA-induced cytotoxicity ( Holm, 196717; Lundgren and Miiller, 1969). In contrast, treatment of the lymphoid cells with antimetabolites that block glycolysis or respiration did inhibit cytotoxicity ( Holm, 19671~; Holm and Perlmann, 1968). It can be concluded that the PHA-induced cytotoxicity is a function of encrgy-requiring processes. Since nucleic acid synthesis or protein synthesis is not necessary, cytotoxicity is obviously not bound to blast transformation or mitosis. There is actually no direct proof that the transforming cells also arc' the executors of the cytotoxic reaction. Stimulation of lymphocytcs by PHA is known to lead to early changes in phoqpholipid mctabolism (Fisher and Mueller, 196s; Kay, 1968) and to early nuclear changes (Killander and Riglcr, 1965; Pogo et al., 1966; Darzynkiewicz et al., 1969; Riiigeitz et al., 1969).
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
165
Particularly the first-mentioned changes are indicative of alterations in cell surface activity and should be of importance for cytotoxicity. It could be expected that activated lymphoid celIs, because of enhanced metabolic requirements, become cytotoxic for target cells by competing with them for nutrients in the medium. Howel-er, when purified human blood lymphocytes were treated with the antibody-like plant protein Concanavalin A (Sumner and O’Kane, 1948; Goldstein and Iyer, 1966) they were strongly stimulated to blast transformation and DNA synthesis but were not cytotoxic for chicken rcd blood cells. Moreover, Concanavalin A potentiated PHA-induced stimulation but completely blocked cytotoxicity. It did not inhibit the PHA induced mixed aggregation of lymphocytes and target cells. The inhibition was completely reversible when Concanavalin A was removed from the lymphocytes by treatment with n-methyl-D-niannoside after one hour but not after 20 lirs of interaction. This is furthcr evidence for the notion that blast transformation and cytotoxicity are separate phenomcnn ( Perlmann ct nl., 1970). Similar results were obtained when human lymphocytes were treated with hc,it-inactivatcd antiserum (rabbit) to human immunoglobulin ( Fig. 6). Under the evperimeiital conditions applied, the antibodies completely suppressed PHA-induced cytotoxicity but did not block PHAinduced transformation and DNA synthesis. Antiserum to human serum albumin had no inhibitory effects. On the contrary, this antiserum induced a slight cytotoxicity of the lymphocytes even in the absence of PHA. The results point to the interesting possibility that lymphocyteassociated y-globulin also may havc rcccptor functions: in PHA-induced cytotoxicity. However, other cxplanations cannot be excluded ( see following paragraph). Rabbit ALS stimulate human lymphocytes to blast transformation and DNA synthesis to almost the same extent as PHA (for references, see James, 1967). Human lymphocytes tieated with heat-inactivated ALS absorbed with Chang cells were not cytotoxic for jlCr-labeled Chang cells ( Holm and Perlmmn, 1969a,c), even \vhen thc antiserum was applied at stimuhting concentrations. Antilymphocyte s c ~ aalso inhibited PHAinduced cytotoxicity. I t is not known whcJther this inhibition of cytotoxicity in vitro is related to the immunosuppressive action of ALS in vivo. Light-microscopic and electron-microscopic observations suggested that the impairment of cytotoxicity was at least partly due to changes in pattern of lymphocyte target cell aggregation and inhibition of peripolesis ( Biberfeld et nl., 1969). Similar inhibitory effects:, recently noted by Lundgrm ( 1,undgren c’t a/., 1968a; Lundgren, 1969) , werc also interpreted in this way.
166
PETER PERLMANN AND GORAN HOLM
i
I6
FIG.6. Inhibition of phytoheniagglutinin ( PHA )-induced cytotoxicity of human blood lymphocytes by rabbit antiserum to human immunoglobulin G (ICG). Ordinate: percent isotope release after 15 hours of incubation of “Cr-labeled chicken erythrocytes with lymphocytes. Abscissa: lymphocyte/erythrocyte ratio. Each tube contained 1 X lo5 erythrocytes. Squares: lymphocytes pretreated with 25% heatinactivated normal rabbit serum for 30 minutes and then washed. Circles: lymphocytes pretreated with 25%rabbit antihuman-ICG. Solid lines: incubation in presence of PHA. Dashed line: incubation without PHA, or without lymphocytes, respectively. (From Perlmann and Holm, 1969.)
c. Cytotoxic Factors. The evidence presented in the preceding paragraphs suggests that cytotoxicity of activated lymphoid cells requires contact between effector cells and targets. In the initial work by Holm et al. ( 1964), no cytotoxic factors could be recovered from the incubation medium of PHA-treated lymphoid cells. However, Ruddle and Waksman ( 1968c) recently presented preliminary evidence for the occurrence of factors inhibiting target cell growth after incubation of rat fibroblasts with PPD-treated lymph node cells from rats sensitized to tubercle bacilli. Similar findings have been reported by Granger and Kolb (1968) and Granger et al. (1969) in various immune and nonimmune cytotoxicity models in mice, guinea pigs, and man. Granger and Williams (1968), Williams and Granger (1968,1969), and Kolb and Granger (1969) found that culture media, obtained from lymphocytes stimulated by PHA, antigen, or mixed culture inhibited L-cell fibroblasts and other target cells. These factors were called nonspecific “cytotoxins” and were assumcd to be effectors of cell-mediated cytotoxicity in general. Their generation would depend on preceding lymphocyte activation. This adds to the already impressive list of factors, such as interferon (Grecn et aZ., 1969),
CYTOTOXIC EFFECTS OF LYMPHOID CELLS
in Vitro
167
lcukotactic factor ( Raniseier, 1967; Ward et d., lS69), specific and nonspecific blastogenic factors, permc&ility factors, and low- and high molecular transfer factors, nll assumrd to he given off from activated lymphoid cells [for other references, see Lawrence ( 1969) 1. The relation of the “cytotoxins” to these factors is prcscntly the matter of some confusion. These findings do not necessarily contradict the notion that the lytic reaction requires close contact between eff cctor cclls and target cells, since cytotoxic factors may primarily exhibit local effects. Much of the evidence referrcd to in this and previous sections points to the fact that remote bystander cells in a target cell culture are not destroyed during incubation under proper conditions. When target cell destruction requires long incubations, thc situation probably becomes more complex. However, even if soluble cytotoxic mediators were involved, the factors referred to above iiiay be of minor significance. In these reports, cytotoxicity was primarily measured as monolayer destruction or a s growth inhibition, both known to be susceptible to changes of thc tissue? culture medium. In PHA-activated lymphocytes, lysosome formation is nn early event (Hirschhoni et al., 1967). When the lymphoid cc.11~die, as in the experiments of Granger and Kolb ( 1968), or when supernatants from blast cell cultures are used, toxic lysosomal enzymcs may become enriched in the medium. Such factors are also released to the mcdiuiii when lymphocyte suspensions are contaminated with other leukocytes (Lundgren et a)., 196%). This has becn the casc in almost all reports referrcd to above. Lysosomal enzymes may cither act on the targct cells directly, or indirectly b y changing the niediuin. Cytolysis may- or inay not follow secondarily to growth inhibition. Possible cytotoxic inediators produced by living and mctabolically active lymphocytes may be of a different nature>.They may be related to the complement system ( Section III,C) or may be ccll-bound surfaceactive agents generated by changes in phospholipid metabolism of the activated cells (Fisher and M~leller, 1968; Kay, 1968; Fischer et d., 1968). Much more work is needed to permit distinction between tissue culture artifacts and possible cytotoxic mediators of biological inqiortaiicc,.
6. Conclusions The PHA-induced cytotoxicity of lymphoid cells from normal donors serves as a model for thc similu cytotoxicity exhibitcd by cells activated by othcxr, morcl physiologic,il stiiniilant\. Common for 2111 models i u the rcquiremcnt of living and mctalmlic~tllyactive effector cc4s, susceptible
168
PETER P E R L h l A N N AND G 6 R A N HOLM
to stimulation. Cytotoxicity is an energy-rcquiring process. It is set in motion by activation of lymphocytcns by antigen or other stimulants. H 0 ~ 7 ever, although the smie lymphocyte-stimulating reactions are probably involvcd, blast transformation and DNA synthesis per se are not required for cytotoxicity. There is no relationship in immunological specificity between the stimulating agents and antigens on the target cells. In this sense, cytotoxicity is nonspecific. Both autologous, syngeneic. and d o geneic target cells are affected. Contact between effector cells and target cells is not needed for activation of the lymphoid cells. However, contact seems to be necessary for target cell destruction at least during carly phases of interaction. Surfacch bound effector sites on the lymphoid cells may be needed for target crll destruction but the nature of such hypothetical sites is unknown. The strongly enhanced surface motility and peripolesis of activated lyniphocytes is considered to be of importance for cytotoxicity. The concept of contactual target cell destruction does not exclude local release of cytotoxic mediators. Mechanisms for target cell destruction, discussed in previous sections ( I I I , A , B , and C ) are also applicable to this model. This is also true for the cytotoxicity of polymorphonuclear leukocytes or niacrophages and for the possible importance of interactions between these cell types and lymphocytes.
E.
CELLDESTRUCTION BY LYMPHOID CELLS FROM NORMAL DONORS AFTER “in Vitro SENSITIZATION” TARGET
1. Cytotoxicity of “Normal” Lyinphoid Cell-s
When lymphoid cells are from donors sensitized to target cell antigens, the cytotoxic reaction is initiated by immunologically spccific reactions. In all the other models, receptor units or lymphocyte-activatiiig agents are introduced experimentally in order to produce cytotoxic effects. Lymphoid cclls froin normal donors are usually not cytotoxic without addition\. On the contrary, growth-promoting fecder effects of normal lymphoid cells ha\^ been o h \ e i i d ( Section II,D,2). Howcver, exceptions to this rule have also been notcd. Thus, Stuart (1962) found that spleen cells from norin,il mice destroycd human tissue culture cells in monolayer within 48 hours. Spleen cells from immunized mice produced thr same effects at a higher rat?. Hcat-killed spleen cells or extracts were inactive. Similar obscrv10,000 to 40,000 mol. wt. The protocol in this study consisted of placing DNase-treated extracts of frozen and thawed leukocytes (0.3 to 0.9 ml., 255-605 loGcells), obtained from donors with the appropriate delayed sensitivity, in a Visking cellulose sac and dialyzing in a (1:1) ratio against distilled water, overnight in the cold room. The dialyzate was then filtered and injected into the shoulder of the tuberculinnegative or coccidioidin-negative recipient and the respective skin tests made in the forearm. It was demonstrated that tuberculin sensitivity was transferred to 11 recipients of dialyzable transfer factor, of which 7 developed marked ( 4f ) reactions and 4 developed moderate (2+ ) reactions to tuberculin. Treatment with 50 pg. RNase had no effect on thc capacity to transfer tuberculin sensitivity and no inhibitors to RNase activity were demonstrable in the dialyzate. The findings obtained with tuberculin were confirmed when coccidioidin was uscd as a test material. Of 11 recipients, 7 developed marked ( 4 + ) reactions and 4 devcloped moderate ( 2 + ) reactions to coccidioidin following transfer with dialyzates prepared from coccidioidin-sensitive donors. It was found that transfer factor could be lyophilized without impairing its activity. This facilitated dialysis with the result that greatcr amounts of transfer factor appeared in the dialyzate as determined by the increased intensity of coccidioidin sensitivity transferred.
x
230
H. S. LAWRENCE
DIALYZATE -ABSORPTION SPECTRA AFTER SEPHADEX G-25
c 9 cc W BC EXTRACT
+
DNASE CIALYLATE
-
A SEPHADEX COLUMN
0 8 8 __
TUBE NO
RECIPIENT 1
1
SKIN REACTICNS-l WEEK PCST-TRANSFER
*I[
-
3+
[3
FIG. 3. Transfer of coccidioidin sensitivity with active fraction (peak 11, Ollt
Tes
18-17 days 3000 Tes
48 hor1rs so 16
7 2 hours A-0 8-10
48
About 60,000/1nm3
2.5,000/mm3 rarely reaches 30,00O/mm3
Can reach 100,000/n1n13
10-1,5,000
9 days
15,000 \-es
s0
I1Olll.h
x
274
IVOR N. BROWN
to the lifetime of the host, an infection of inan with Plasinoclium falciparurn or Plasmoclium vivax is of short duration; P. falcipciruni does not usually persist for more than a year, nor P. vivax for more than 2 years. In contrast, latent Plasmodircm nialarine infection may persist for over 20 years. A first clinical attack of inalaria is often followed by others which in the abscnce of reinfection are termed relapses (see Section V ) . Aftcr P. fulciparum infection, relapses normally occur soon after the primary parasitemia attack. This Plasmodium is thought to have no persistent exoerythrocytic stage and relapses appear to originate from latent blood infection. On thc other hand, after P. vivux and P. malnriae infection, relapses may also occur long after a primary attack. These plasmodia arc thought to have a persistent exoerythrocytic stage which can initiate crythrocytic attacks in addition to those attacks derived from latent blood forms.
B.
SYhlI’TOMS AND PATHOLOGICAL
EFFECTS
Thc characteristic symptoms and pathology of plasmodia1 infection is caused by the blood stage of the Plasmodium. Paroxysms of fever, accompanied by chills and sweating, arc associated with erythrocytic schizogony. The timing of these paroxysms varies according to the species of Plasmodium. For cxample, Plasmodizini vivax undergoes schizogony eveiy 48 hours and paroxysms occiir every third day, whereas Plasmoclium mnlariae segments every 72 hours and paroxysms occiir every fourth day (see Table 11). The pathological eff ccts of infection are numerous, but the principal effects can bc summarized as follows: 1. The production of a moderate to severe anemia often in excess of that due to parasite destruction. This effect may be due to the invasion of erythropoietic stem cells (as occurs in some avian plasmodial infections) or possibly to an immune reaction to normal erythrocytes (see Section X I ) . 2. The disturbancc of tissue function by toxic products released from the parxite. 3. The disturbance of blood circulation because of blockage of fine capillaries by parasiti7cd cells and parasitized cell debris (such as rcd cell stroma and malarial pigment ) lcading to necrosis and hcinorrhage. The fixed tissucs most noticeably affected b y plasmodia1 infection are the splcen, liver, and bone marrow. All become full of infected crythrocytes, ccll debris, and pigmcnt; the splccn and liver often become enI m g d and their tissucs hyperplastic. In the brain, thc acctimiilation of
IMMUNOLOGICAL ASPECTS OF hfALARIA INFECTION
275
large numbers of parasitjzed cells m i i y lead to local necrosis and cerebral malaria (Edington, 1967). This complication is most often seen in young children with Plc~.smodiziinfalcipumm intcxction. The kidney is also particularly affected by malaria (sec Section X I ) . Blackwater fever is nssociated with acute P. fnlcipurum infection and is characterized by a sudden clinical attack followed by hemoglobinuria. Plnsniocliim m a h i e infection which is, in gcneral, low grade and of long duration, is thought to be responsibk for the nephrotic syndrome obscrvcd in children in parts of Africa and the Far East. Ill.
I n n a t e a n d Nonspecific Immunity to M a l a r i a
Innate resistance of mosquitoes will not be discussed here but has been reviewed by Garnham (1964), Huff (1965), and Zuckerman (1968). Vertebrates show varying degrees of innatc immunity ranging from a complete resistance, throuzh a phasic resistance (in which, for example, apparently normal development of exoerythrocytie stages occurs but no erythrocytes are infected) (World Health Organization, 1968), to an incomplete resistance (in which development of all stages of infection occurs but parasite multiplication is restricted). The mechanisms of innate immunity are ill-defined, but among the factors that can affect observed infection are the species of host, its genetic constitution, its age, and its environment. Immunity may also be nonspecifically acquired by infection of the host with another organism. These factors arc briefly discussed below in order that thcy may be borne in mind during subsequent discussion of specific acquired immunity. They are more fully considered in the articles cited in Section 1,B.
A. HOSTSPECIFICITY In general plasmodia show a marked host specificity. The more closely related the prospective host is to the natura1 host the more likely infection is to occur. Unfortunately for laboratory workers, the human plasmodia are probably thc most restricted of all the mammalian subgenera in thcir ability to infect hosts other than man. Human plasmodia] infections can bc produced in some monkeys and in higher apes but development is often rcstricted to the exoerythrocytic stage with no or only subpatent erythrocytic development. Such infections are usually enhanced by splenectomy (Bray, 1958; Garnham et al., 1963; Gould et al., 1966). The owl monkey, Aottrs iriuirgatus, will support infections of Plasmoclium falcipurum (Geiman and Meagher, 1967), and recent studies (Voller et al., 1969) have indicated that these monkeys show a similar pattern of infection to man. Owl monkeys are difficult to keep in captivity but if
976
IVOR N. BROWN
this problcm can be overcome they could provide, for the first time, a model of human plasmodia1 infection for clinical and immunological study. Simian plasmodia, on the other hand, seem more readily to infect man. Plasmoclium knowlesi of macaque monkeys has been used for malaria therapy in neurosyphilitics. Recent reports of natural transmis5ion of monkey malaria to man (Chin et al., 1965; Deane et al., 1966) have revived interest in the zoonotic potential of the siinian parasitrs ( Garnhain, 1967; Bray, 1968). B. GENETICCONSTITUTION Within a susceptible population there are differences in incidence and severity of infection determined by differences in genetic constitution among the individuals of that population. The variation in susceptibility to Plasmodium berghei shown by inbred mouse strains and their hybrids illustrates this ( Greenberg et nl., 1954; Greenberg and Kendrick, 1957). In human malaria there is evidence that some genetic resistance is conferred by both the sickle-cell gene and the glucose-6-phosphate dehydrogenase ( G-6-PD ) -deficiency gene. The sickle-cell gene, which is responsible for the production of an abnormal hemoglobin, is common in several human populations and in particular, in those of Central Africa. The effect of the trait is most noticeable in young children (1-4 years) who lack significant acquired immunity to malarial infection. The proportion of infants showing parasites, as well as the parasite counts, is lower in trait carriers (Allison, 1957, 1961; Edington and WatsonWilliams, 1965; Gilles et nl., 1967). The reason for this conferred resistance is uncertain, but indirect evidence that other abnormal hemoglobin types also protect against malaria suggests that the mechanical effect of sickling is not responsible. Malaria parasites metabolize hemoglobin, and abnormal hemoglobin may prevent normal development of the parasite. The selective advantage conferred by this gene in malarious areas could possibly account for its survival in their human populations “Abnormal” hemoglobin types may, in addition, determine the effect5 observed by Greenberg and his colleagues and, probably, contribute to the observed host specificity of the malaria parasite. The geographic distribution of the G-6-PD-deficiency trait is remarkably similar to that of Plasmodium fakiparum, but evidence is conflicting as to whether possession of the trait protects against malaria. The studies of Allison and Clyde ( 1961) and Gilles et al. (1967) demonstrated that enzyme-deficient children suffer less from infection than do “normal” children. Most of the conflicting evidence comes from studies of older children (where differences were somewhat obscured by acquired im-
IMMUNOLOGICAL ASPECTS OF hIAL4nI.4 INFECTION
277
tnunity ) or from studies on adtilt white voliintccrs (whcrc. only low parasitemias were used). Little is known of other gtmcltic factors involved in natural inimunity to malaria. Amcrican Negrocs show a marked rcxsistancc to infection with Plasmodiiini vivux ( Boyd and Stratn~aii-Thonias,1933), whereas American Cuucasians do not. A similar resistance can bc observed in many West African Negro populations.
C. AGEOF HOST Young persons and young animals tcnd to be more susceptible than adults to a first infection; after apparent cure they are more likely to rekipse. For exmiple, a Plusnzorliuni berghei infection in suckling and weanling albino rats was almost 100%fatal, whereas in adult rats, weighing nearly 200 gin., only 33%of the animals died (Zuckerinan and Yoeli, 1954). The peak parasitcmia was niuch higher (90%compared with 6%) and occurrcd later (15 days coinpared with 8 clays) in the young rats compartd with thc old rats. On the other hand, the adult rats who died of infection died earlier than the young rats (10.3 days compared with 17.0 days).
D. EXVIHONMENTAL CONDITIONS Variation in environmental conditions can affect susceptibility to malaria. Dietary deficiencies, e.g., p-aminobenzoic acid ( Hawking, 1954; Kretschmar, 1965; Jerusalem and Kretschmar, 1967) and ascorbic acid ( McKec and G c h a n , 1946) limit the in vioo multiplication of malaria parasites, and strcss can acccntuate infection ( Kretschmar, 1964).
E. SIhlULTANEOUS INFECTION Evidence is conflicting as to what effect simultaneous infection with another organism has on malaria infection. Typhoid fever has little influencc on malaria ( Giglioli, 1933) altliough typhoid symptoms inay bc morc: severe in malarious subjccts ( Nazario, 1929). Tuberculosis and malaria inny 11v mutually antagonistic ( Yoeli, 1966; Voller and Rossan, 1969d). Also, ;I concurrent Eperythrozootx coccoides infection protects mice against Plastnoclizim berghei and Plasmodium chabaticli infection (Peters, 1965; Voller and Ridwcll, 1968). The reason for this protection is unknown, but it has been suggested that there may be competition for an unknown substrate or that there may be a nonspecific heightening of phagocytosis. On the other hand, if mice are made anemic by infecting with Haemo1)nrtonelln nitiris and then infected with P. berghei the malaria infcction is potentiated (Hsu and Geiman, 1952). This effcct is
278
IVOR N. BROWN
probably duc to known prefcwncc of this specics for infecting thc immature erythrocyte of thc mouse. This p r c h e n c e can also be used to block intection. In normnl mice made polycythcmic by hypertransfusion, ‘1 P. b e r g h i intcvtioii is partidly blocked cven though a large population of mature red cells is availablc for inf-cction. Idcction is coinpletcly blockcd in polycythemic mice by irradiating them 3 days before infection to remove rr~sicliialcrythropoiesis ( Ladda and Lalli, 1966) . Chloroquineresistant P. berghei parasites show a particularly markcd preference for rcticulocytes, and irradiation of normal mice before infection with resistant par‘isites produccs a great diminution in parasitcmia; in addition a high proportion of mice do not become infectcd ( D . C. Warhurst, personal communication) . IV.
A.
POPULATION
immunity Acquired through Infection
STUDIES
How malaria appears in a community depends largely on the inanncr of its transmission. Where transmission is low, the position of the parasite is prccarious. Under these conditions the community does not develop a high level of resistance and, consequently, may suffer infections of epidemic proportions, for example, due to a sudden increase in numbcrs of an effective vector. The parasite most often associated with such conditions of transmission is Plnsmocliuni wivax because of its capacity for late relapses, although Plasmodiiim f a k i p a r u m may appear also. In contrast, in some areas of the world, notably Africa where P. fnlcipariiin predominates, malaria transmission can be continuous and at a high level ( stable nialaria ) although subject to seasonal fluctuation. Studies of hunian populations living in such areas have revealed a remarkably consistent pattern of infection. The incidence (see Fig. 2 ) and density of parasitemia is maximal in young children and declines progressively in the oldcr age groups (Christophcrs, 1924; Taliaferro, 1949; Rrucc-Chwatt, 1963a). During the first few years of life, infections are sevcw and, without treatment, may cause the death of infants. Young children who survive this critical period still suffer heavy parasiteniiar but seem better able to tolerate the infection than infants. This type of immunity is called “clinical” or “antitoxic” immunity. Through older childhood and adolescence, parasite densities decrease until in adult lifc only low levels of parasiteinia are encountered. During this period, when parasite densities are falling, a true antiplasmodial imniunity is thought to be ncquired. MacDonald considers young children as being largely rcsponsihle for continued malarial transmission but knowledgc of the
279
IMMUNOLOGICAL ASPECTS OF MALARIA INFECTION
1 2 5
10
16
30
40
50
60
Age in years
FIG.2. The incidence by age of malaria parasitemia in Keneba villagers (1950). Keneba is situated in Gambia, West Africa, an area endemic for malaria. Plasmodium fulciparum is the dominant parasite. (From McCregor, 1964.)
natural history of malaria is still incomplete ( MacDonald, 1957; Miller, 1958; McCregor, 1965; Rruce-Chwatt, 1965). B.
EXPERIMENTALLY INDUCED INFECTIONS The use of the experimentally induced infection has the advantage that the source and type (infected blood or sporozoites) of infective material and the timing of infection or reinfection can be controlled. Most studies in this respect concern laboratory animals and birds, but man has also been experimentally infected, either as a volunteer or as a paretic undergoing malaria therapy. The effectiveness of an acquired antiplasmodia1 immunity depends largely on the degree of previous infection [this is illustrated by the expcriments of Maier and Coggeshall (1944) described in Section 1x1. Such immunity niay be antitoxic rather than antiparasitic and show as a suppression of symptoms normally associated with erythrocytic attack. The nature of this immunity is not known. Antiparasitic immunity is most evident during the erythrocytic stage of infection and may show as a lengthened prepatent period after parasite inoculation, a shortening of the primary erythrocytic attack, reduced parasitemia, or a complete clearance of parasites from the host tissues. These expressions of antiparasitic immunity niay act independently of one another. Antiplnsmodial immunity is restricted in its specificity and may be STUDIES OF
280
IVOR N. BROWN
effective only against a strain or strains of a given species of Plasmodium (James, 1931; Sinton et al., 1939; Boyd and Kitchen, 1945). This species specificity has been demonstrated in human, simian, and avian malarias ( Gingrich, 1932; Mulligan and Sinton, 1933; Taliaferro and Taliaferro, 1934; Boyd et al., 1936; Manwell, 1938; Taliaferro, 1949; Jeffery, 1966; Voller and Rossan, 1969b,d) and in the rodent malarias (F. E. G. Cox, 1966; Cox and Voller, 1966). After the natural elimination of infection, resistance to further homologous infection lessens although complete susceptibility may never rcturn in some host-parasite combinations. That a true immunity in malaria infection may in reality be an immunity to superinfection has led to much controversy. Sergent et al. (1924) introduced the term “premunition” to describe a nonsterile or coinfectious immunity and compared premunition with sterile or residual immunity. Although useful descriptively, the validity of such a distinction on immunological grounds is questionable. The increased efficacy of immunity to superinfection may be explicable on the basis of an adjuvanted immune response and the waning of such immunity after elimination of the parasite.
C . THEPLASMODIAL LIFECYCLE AND ACQUIRED IMMUNITY Apart from the strain and species specificity of acquired antiplasmodial immunity, there is evidence that those stages of the life cycle that are immunogenic stimulate an immune response specific for themselves. This further specificity is discussed below with reference to the sporozoite, the exoerythrocytic stage, and the blood stage.
1. Sporozoites Whether immunity to sporozoites contributes to immunity acquired in endemic areas is not known. Under natural conditions of transmission they do not appear to provoke an effective immune response. Sporozoites that do not develop within liver parenchyma cells may represent an insufficient antigenic stimulus or a degree of immunity may be developed which is ineffective because injected sporozoites are not in tissue fluids for a long enough period. It could be similarly argued that their brief extracellular life (an hour or less) may not allow for an anamiiestic response. Garnham (1966b) found no evidence of an immunity to Plasmodium cynomolgi bastianelli sporozoites (judging from the number and appearance of liver stages after sporozoite inoculation ) in rhesus monkeys sensitized either by sporozoite-induced infection alone or by sporozoiteinduced infection followed by the injection of large numbers of formalinkilled sporozoites in complete Freund’s adjuvant. However, if sufficient
IMMUNOLOGICAL ASPECTS OF MALARIA INFECTION
281
killed sporozoites of thc avian Plasmoclium, P. gallinaceurn, are injected into fowls, an apparent immunity to sporozoitcs can be induced (Mulligan et al., 1941; Russell et al., 1942). Fowls vaccinated with sporozoites inactivated by ultraviolet light or by grinding, drying, and subsequent reconstitution, gave varying serrim agglutinin titers against the honiologous sporozoites. Those fowls having titers less than 1: 16000 ( this would includr normal birds ) were susceptible to mosquito-borne infection (mortality 51.4%), whereas most fowls having titers greater than 1:32000 were totally resistant or developed mild infections only (mortality 7.7%).Fowls resistant to sporozoite challenge were, however, compIctely susceptible to intravenous challenge with blood containing P. gallinacerim trophozoites. Recent findings with a similar experimental system ( Richards, 1966) have confirmed these results. Sporozoites werc inactivated by the above methods and also by formalin treatment and by freeze-thawing. Young chicks immunized by any of these preparations developed good serum sporozoite agglutinin titers and were partially resistant to sporozoite challenge but were completely susceptible to challenge with erythrocytic parasites. The birds were protected against death rather than infection, for rcsistant birds were found to be carrying latent infections. Also Nussenzweig et al. (1967, 1969) found that mice sensitized by repeated injection of irradiated sporozoites of Plasmodium berghei did not show blood infections if infected with P. berghei sporozoites, whereas mice sensitized with noninfected mosquito salivary gland tissue developed and died of blood infections, In addition, mice similarly sensitized with P . berghei sporozoites did not develop blood infection after the inoculation of Plasmodium vinckei sporozoites, but nevertheless were completely susceptible to a blood-induced infection with either P. berghei or P. vinckei. Possibly the sporozoites of these two species of rodent malaria show more antigenic similarity than do their blood forms. Mice immune to the blood forms of P. berghei are susceptible to infection with the blood forms of P. vinckei. 2. Exoerythrocytic Stages
Not only sporozoites but also exoerythrocytic stages are apparently unaffected by spccific acquired immunity. Immune chickens and pigeons, when inoculated with sporozoites of Plasmodium gallinaceum and Plasmodiuna relicttim, respectively, showed pre-erythrocytic tissuc stages but no significant parasitemia ( Huff and Coulston, 1946). Similarly, if homologous sporozoites were inoculated into a person immune to
282
IVOR N. BROWN
Plasmodium uivax the development of pre-erythrocytic schizonts in the liver occurred but in the absence of patency ( Shortt and Garnham, 1948). Also, rats recovered from (and resistant to) trophozoite (or sporozoite)induced Plasmodium berghei infeetion developed pre-erythrocytic schizonts in liver parenchymal cells on reinoculation with P. berghei sporozoites (Yoeli, 1966). Liver biopsies taken 48 hours after inoculation of sporozoites showed mature and apparently normal exoerythrocytic schizonts, yet erythrocytes did not become infected. This immunity to the blood stage need not be acquired but may be innate. For example, birds vary in their susceptibility to P. gallinaceum (Huff and Coulston, 1946). Chickens are susceptible to both tissue and blood stages, whereas canaries show a complete resistance to the parasite. In intermediate range geese, ducks, and guinea fowl all show apparcntly normal pre-erythrocytic development but the resulting parasitemia is transient in the goose, only subpatent in the duck, and probably nonexistent in the guinea fowl. Similarly the tissue forms of some mammalian parasites may develop in abnormal hosts which show innate resistance to the erythrocytic stage. The nature of this resistance is not known but can often be lowered by splenectomy, when the hosts suffer transient to moderate parasitemias. The red cells of such animals are, therefore, capable of supporting parasites. I n macaque monkeys infected with Plasmodium cynomolgi sporozoites there is a considerable variation in the number of sporozoites that develop into pre-erythrocytic schizonts, but the number of liver stages produced by a given sporozoite inoculum cannot be correlated with any existing state of immunity (Garnham and Bray, 1956). Also the morphology of the liver stages in “immune” animals is identical to that of liver stages in “normal” animals (see also Yoeli, 1966). It has been observed that liver stages become surrounded by phagocytic cells, but whether this signifies a specific immune response is unknown. Within the liver, development of the parasite is largely intracellular and this may preclude an immune response. However, in the brief period when the exoerythrocytic schizont matures and merozoites are released there may possibly be an antigenic stimulus. 3. Blood Stage
The blood stage of the Plasmodium is the most susceptible to thc effects of immunity. Thus, hosts possessing innate immunity often show normal development of exoerythrocytic stages but no or Iittle erythrocytic infection. Similarly, human subjects previously infected wit11 Plasmodium falciparum, Plasmodium uivax, or Plasmodium ovule become
IMMUNOLOGICAL ASPECTS O F MALARIA INFECTION
283
infected after the injection of homologous sporozoites but the remitting parasitemia is transient only (Boyd et al., 1936, Boyd and Kitchen, 1936; Sinton, 1939a,b, 1940) . Studies of antimalarial antibodies, the cellular reactions to infection, and the effect of existing immunity on bloodinduccd mammalian and avian infections have largely confirmed that the blood stage is susceptible to immune attack. Much of the work rcferred to below pertains to immunity and erythrocytic infection. The asexual erythrocytic Plasmodium seems most susceptible just before and at schizogony (Coggeshall, 1943; Taliaferro and Bloom, 1945; Cohen and McGregor, 1963; K. N. Brown and I. N. Brown, 1965) but, in addition, merozoites may be affected by immunity. The engulfment of free P . fulciparuin merozoites by polymorphonuclear leukocytes has been observed in a warm stage preparation of infected human blood drawn from a relapsing patient ( Trubowitz and Masek, 1968), and presumably similar and more specific mechanisms operate in vivo. In his extensive studies of immunity to plasmodia1 blood forms, Taliaferro ( 1948, 1949, 1967) differentiatcd parasiticidal mechanisms from reproduction-inhibiting factors, making the assumption that specific acquired immunity as it developed was superimposed upon existing innate immunity. Antibody levels were not correlated with the observed effects. The effects of acquired immunity were most noticeable during the period of parasite decline, i.e., just after peak acute parasitemia and crisis. These effects were a decrease in the number of merozoites produced per schizont (which may have been an apparent effect due to more e&cient removal of mature schizonts from the blood) and an increase in the number of parasites that died. In the host-parasite combinations studied when the acute parasitemia subsided and only low parasite levels were detectable, the parasite reproductive rate regained its precrisis level, but the number of merozoites that survived to reinfect new erythrocytes remained at low level only. In some simian infections, well marked degeneration is apparent in erythrocytic parasites in semi-immune hosts (these degencrate forms occur mostly just after peak parasitemia and are called “crisis” forms ) . Gametocyte levels fall subsequent to depression of asexual parasite levels, but it is not known whether gametocytes are directly affected by immunity-there is evidence suggesting that they may be. Plasmodium cynoinolgi ganietocytemia persisted at a high level through the crisis of asexual forms in rhesus monkeys, but infectivity for mosquitoes was nevertheless, markedly reduced (Hawking et al., 1966). Similarly, in experiments involving thv transmission of rodent malaria parasites through mosquitoes, a coininon observation ( Wery, 1968) is that transmission is
284
NOR N. BROWN
more easily obtained if mosquitoes are fed on rodents at an early stage of acute infection. One interpretation of these observations (but not the only one) is that immunity affects gametocytogenesis or gametocytes directly. That Plasmodium falciparum gametocyte levels rise after depression of asexual parasitemia levels with passively administered immune 7-globulin (Cohen et al., 1961) would not be inconsistent with this interpretation for the infectivity of gametocytes was not tested in these experiments. Also P . falciparum may be an unusual parasite because its gametocytes are thought to take 8-9 days to develop. On the other hand, high levels of malarial antibody in rodent serum has no apparent effect on the development of rodent malaria infection in mosquitoes up to the oocyst stage ( Killick-Kendrick, personal communication). In addition, adults living in endemic areas may show low gametocyte levels, but these gametocytes are highly infective for mosquitoes. Possibly gametocytes are antigenically labile and can avoid the host’s immune response (Hawking et al., 1966; Section V ) or a developing immune response may stimulate gametocytogenesis. V.
Relapses and Antigenic Variation
Relapses by definition occur in the absence of reinfection and show as a reappearance of patent blood infection (often accompanied by clinical symptoms) after the primary parasitemic attack has subsided. Of the human plasmodia, Plasmodium vivax and Plasmodium mulariae show the greatest tendency to relapse, Plasmodium falciparum the least. There are three main theories of the origin of relapses. The first theory supposes the persistence of small numbers of erythrocytic parasites that escape the action of drugs or immunity. The second theory presumes the existence of persistent exoerythrocytic stages insusceptible to the effects of immunity developed to the blood phase. The third theory implicates a latent stage of the primary exoerythrocytic schizont (or, perhaps, of the sporozoite but there is no evidence for this). On theoretical grounds, relapses are, therefore, to be classified (World Health organization, 1963) as ( a ) recrudescences which derive from persistent blood stages or ( b ) true relapses (or recurrences) which derive from some form of persistent tissue stage. In practice neither of these two mechanisms alone can account for the origin of relapses (see Bray, 1957a, 1963; Garnham, 1966b for discussion), and it is generally accepted that both may occur after sporozoite-induced infections except where no persistent liver stage has been recorded ( e.g., P . falciparum). After blood-induced infection only recrudescences can occur as there is no infection of the liver.
IMMUNOLOGICAL ASPECTS OF MALARIA INFECTION
285
Relapses are one of the puzzling features of malaria infections. Their occurrence indicates the survival of the parasite for often extraordinary lengths of time within the body of the vertebrate host. The reason for relapses has provided a subject for much discussion. Explanations of the chronicity of malaria infections include that the host becomes immunologically tolerant to some malaria antigens, that the parasite is poorly immunogenic, or that the parasite, by virtue of its intracellular habitat, is largely insusceptible to the action of immunity. One of the most common cited reasons for relapses is a waning of protective immunity and this receives some support from observations of the depressed antibody levels just before relapses (Coggeshall, 1943) and the enhancing effect of spIenectomy on malarial infections ( Corradctti, 1963). A neglected factor in many discussions of relapses has been the possibility of antigenic lability of the malaria parasite. It is known that immunologically distinct strains, judging from challenge experiments, may exist within a malaria parasite species, and as already indicated the various stages of the life cycle vary in their susceptibility to the effects of immunity. Microorganisms can, however, show considerable diversity of antigenic structure and some species can change rapidly from one antigenic state to another ( Bcale and Wilkinson, 1961) . An infection may be prolonged if a new variant appears when the original antigenic type is removed by the host’s immune response. Among parasitic protozoa, this type of antigenic change has long been recognized in African trypanosomiasis ( K . N. Brown, 1963). The occurrence of antibody-resistant generations in malarial infections was postulated many years ago (see, for example, Schilling, 1934), and H. W. Cox (1959) put forward indirect evidence for antigenic differences between initial and relapse parasitc populations. Mice harboring chronic infections of Plusmodium berghei were more susceptible to challenge with relapse than with parent strain parasites. Repeated antigenic changes of an order that might account for recrudescing simian and human malaria have only recently been demonstrated in the simian Phasmodium P . knozulesi ( K . N. Brown and I. N. Brown, 1965; I. N. Brown et al., 1968a). A schizont-infected cell agglutination test was used (see Section VII1,A). In a rhcsus monkey suffering a recrudescing P. knowksi infection, the surface antigenic structure of erythrocytes infected with mature asexual parasites isolated from a particular relapse population, differed from that of erythrocytes infected with parasites of other relapses isolated from the same recrudcscing infection. Each popuIntion of parasites (“variant”) stimulatcd specific agglutinating antibodies in the host monkey (Table 111). A low level of nonvariant-specific ag-
TABLE I11 TITERS OF
SCHIZOST-ISFECTED CELL A%GGLUTISINS 1N SERUM S,4MI'LES FRO3f A R H E S U S hfONKEY
SUFFE RI SG A RECRIJDESCISG
Parasit.e stabilates
A
Days afterinitialinfection Parasitesin Moodsmear
B
A
1
1 0
7
-
+
22 -
PfaSmod$izr9?lknowksi
T 73 85 92 94 97
C
D
E
T
T
T
34
49
52 59
65
f
-
+
+ + -+ -+ - +
t-
132 167
196
202
209
218
279
295
328
351
-
-
-
-
_
_
Splen. -
-
+ +
i-
Blood inoculation test
INFECTIOS"
+
1
1
1
1
1
1
1
1
1
1
1
1
Antigen A
1250 10 250 781250 >781250 781250 156250 250 31250 31250 6250 1250 1250 1250 1250 31250 NT
156350 156250 250 250
NT
NT
The monkey was initially infected with parasites of stabilate A, radically cured of the resulting acute infection and then reinfected with A. Parasites were seen in blood smears taken up to day 167 but blood was infective for nmimmUne monkeys until at least day 209. Splenectomy on day 328 did not result in a recrudescence of infection. Stabilates of parasitized cells were isolated either from the host monkey 03 f C) or from recipients of host monkey blood (D E). Schizont-infected cells (antigen) for agglutination tests were collected from nonimnone monkeys inoculated with parasites of the relevant stabiktes. NT = not tested.
+
3
z E0
ifr.osensitization, 168-172 destruction, transfer factor and, 244-
245 Transfer factor, characterization and niechanism of action, confirmation of transfer in man, 22&229 introduction of leukocyte extracts, 217-226 conclusions, 259-261 correlates of cellular immunity and, 234-245 definition and principles, donor selection, 199-200 local transfer, 201-202 protocol, 201
recipient selection, 200-201 systemic transfer, 201 dialyzable, nature and propertic,s of, 229-234 possible applications of, 2,56257 liistorical, 196199, immunological surveillance and tumor immunity, 258-259 mechanisms of cellular iinmune cleficiency diseases and, 248-252 inechanism of action i n vioo and in uitro, 245-248 reconstitution of cellular immune tleficiency and, 252-257 Tuhercnlin sensitivity, transfer of, 212 Tumor defense, lymphoid cell cytotoxicity and, 181-183 Tumor imninnity, transfer factor and, 258-259
V Vaccines, malarial immunity and, 324329 Viiccinia, generalized, transfer factor and, 253-
254 Virus particles, complement and, 88
This Page Intentionally Left Blank