(NAS Colloquium) Virulence and Defense in Host--Pathogen Interactions (National Academy of Sciences Colloquium Series)

About this PDF file: This new digital representation of the original work has been recomposed from XML files created fr...

Author: Proceedings of the National Academy of Sciences

89 downloads 781 Views 5MB Size Report

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

DOWNLOAD PDF

About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please use the print version of this publication as the authoritative version for attribution.

i

COLLOQUIUM ON Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals

NATIONAL ACADEMY OF SCIENCES WASHINGTON, D.C.


ii

NATIONAL ACADEMY OF SCIENCES Colloquium Series In 1991, the National Academy of Sciences (NAS) inaugurated a series of scientific colloquia, several of which are held each year under the auspices of the NAS Council's Committee on Scientific Programs. Each colloquium addresses a scientific topic of broad and topical interest, cutting across two or more traditional disciplines. Typically two days long, colloquia are international in scope and bring together leading scientists in the field. Papers from colloquia are published in the Proceedings of the National Academy of Sciences (PNAS).


CONTENTS

iii

PNAS Proceedings of the National Academy of Sciences of the United States of America

Contents

COLLOQUIUM Papers from the National Academy of Sciences Colloquium on Virulence and Defense in Host—Pathogen Interactions: Common Features Between Plants and Animals

INTRODUCTION Pathogens and host: The dance is the same, the couples are different Noel Keen, Brian Staskawicz, John Mekalanos, Frederick Ausubel, and R. James Cook COLLOQUIUM PAPERS Striking a balance: Modulation of the actin cytoskeleton by Salmonella Jorge E. Galán and Daoguo Zhou Structure and function of pectic enzymes: Virulence factors of plant pathogens Steven R. Herron, Jacques A. E. Benen, Robert D. Scavetta, Jaap Visser, and Frances Jurnak Pseudomonas syringae Hrp type III secretion system and effector proteins Alan Collmer, Jorge L. Badel, Amy O. Charkowski, Wen-Ling Deng, Derrick E. Fouts, Adela R. Ramos, Amos H. Rehm, Deborah M. Anderson, Olaf Schneewind, Karin van Dijk, and James R. Alfano Molecular and cell biology aspects of plague Guy R. Cornelis A framework for interpreting the leucine-rich repeats of the Listeria internalins Michael Marino, Laurence Braun, Pascale Cossart, and Partho Ghosh Acyl-homoserine lactone quorum sensing in Gram-negative bacteria: A signaling mechanism involved in associations with higher organisms Matthew R. Parsek and E. Peter Greenberg Phenotypic variation and intracellular parasitism by Histoplasma capsulatum Silke Kügler, Tricia Schurtz Sebghati, Linda Groppe Eissenberg, and William E. Goldman Exploitation of host cells by enteropathogenic Escherichia coli B. A. Vallance and B. B. Finlay Genetic complexity of pathogen perception by plants: The example of Rcr3, a tomato gene required specifically by Cf-2 Mark S. Dixon, Catherine Golstein, Colwyn M. Thomas, Erik A. van der Biezen, and Jonathan D. G. Jones Plants and animals share functionally common bacterial virulence factors Laurence G. Rahme, Frederick M. Ausubel, Hui Cao, Eliana Drenkard, Boyan C. Goumnerov, Gee W. Lau, Shalina Mahajan-Miklos, Julia Plotnikova, Man-Wah Tan, John Tsongalis, Cynthia L. Walendziewicz, and Ronald G. Tompkins Role of the cystic fibrosis transmembrane conductance regulator in innate immunity to Pseudomonas aeruginosa infections Gerald B. Pier Bad bugs and beleaguered bladders: Interplay between uropathogenic Escherichia coli and innate host defenses Matthew A. Mulvey, Joel D. Schilling, Juan J. Martinez, and Scott J. Hultgren AvrPto-dependent Pto-interacting proteins and AvrPto-interacting proteins in tomato Adam J. Bogdanove and Gregory B. Martin Reactive oxygen and nitrogen intermediates in the relationship between mammalian hosts and microbial pathogens Carl Nathan and Michael U. Shiloh Nitric oxide and salicylic acid signaling in plant defense Daniel F. Klessig, Jörg Durner, Robert Noad, Duroy A. Navarre, David Wendehenne, Dhirendra Kumar, Jun Ma Zhou, Jyoti Shah, Shuqun Zhang, Pradeep Kachroo, Youssef Trifa, Dominique Pontier, Eric Lam, and Herman Silva The role of antimicrobial peptides in animal defenses Robert E. W. Hancock and Monisha G. Scott Suramin inhibits initiation of defense signaling by systemin, chitosan, and a β-glucan elicitor in suspension-cultured Lycopersicon peruvianum cells Johannes Stratmann, Justin Scheer, and Clarence A. Ryan

8752

8754 8762 8770

8778 8784 8789

8794 8799 8807

8815

8822

8829

8836 8841

8849

8856 8862


CONTENTS

iv

NATIONAL ACADEMY OF SCIENCES COLLOQUIUM Virulence and Defense in Host-Pathogen Interactions: Common Features Between Plants and Animals1 DECEMBER 10–11, 1999 Friday, December 10 Virulence Mechanisms in Pathogens—Chair, R. James Cook Welcome and Expectations for the Colloquium, Noel T. Keen Jorge E. Galan, Yale University School of Medicine, “Modulation of the host-cell actin cytoskeleton by the Salmonella type III secretion system” Frances Jurnak, University of California, Irvine, “Structure and function of pectic enzymes—virulence factors of plant pathogens” Daniel A. Portnoy, University of California, Berkeley, “Mechanisms of Listeria monocytogenes pathogenesis” Alan Collmer, Cornell University, “Pseudomonas syringae effector proteins and their type III secretion and translocation” Guy Cornelis, Catholic University, Brussels, Belgium, “Type III secretion and translocation of Yersinia Yops” Ulla Bonas, University of Halle, Germany, “Type III secretion and targeting of bacterial proteins from plant and animal pathogens by Xanthomonas campestris pv. vesicatoria” Virulence Mechanisms in Pathogens—Chair, John Mekalanos Jeff Miller, University of California, Los Angeles, “Signal transduction during the Bordetella infectious cycle” Partho Ghosh, University of California, San Diego, “Structure/function studies with internalin B of Listeria monocytogenes” Peter Greenberg, University of Iowa, “Communication systems and group behavior in Pseudomonas aeruginosa” William E. Goldman, Washington University, St. Louis, Missouri “Phenotypic variation and intracellular survival of Histoplasma capsulatum” Brett Finlay, University of British Columbia, “Exploitation of host cells by enteropathogenic E. coli” David A. Relman, Stanford University, “Global host gene expression responses during infection” Saturday, December 11 Active Defense Mechanisms in Hosts—Chair, Brian Staskawicz Kathryn Anderson, Sloan–Kettering Institute, “Drosophila toll receptor pathways” Jonathan Jones, John Innes Institute, Norwich, England, “Role of toll-like proteins in disease resistance of plants”

1 The colloquium organizers wish to thank the following sponsors for their generous financial support for this meeting: Aventis Research & Technologies Gmbh & Co. DuPont Pharmaceuticals Company University of California Biotechnology Program The Cystic Fibrosis Foundation SmithKline Beecham Pharmaceuticals


CONTENTS

Lory Rahme, Harvard University, “Common themes of pathogenesis among plants, insects and mammals” Gourisankar Ghosh, University of California, San Diego, “The NF- B pathway in vertebrates” Jeff Dangl, University of North Carolina, “Perception of pathogen signals by plants” Gerald Pier, Harvard University, “Innate defense mechanisms on mucosal surfaces: The Pseudomonas aeruginosa—CFTR paradigm” Active Defense Mechanisms in Hosts—Chair, Frederick Ausubel Matthew Mulvey, Washington University School of Medicine, St. Louis “Innate host defenses against uropathogenic E. coli” Gregory B. Martin, Boyce Thompson Institute, Cornell University, “Pathogen recognition and signal transduction mediated by the product of the Pto disease resistance gene” Carl Nathan, Weill Medical College, Cornell University, NY, “Reactive oxygen and nitrogen species in animal defense: Mechanisms of microbial resistance” Dan Klessig, Rutgers University, “NO and salicylic acid signaling in plant defense” Robert Hancock, University of British Columbia, “Antimicrobial peptides in animal defense” Clarence A. Ryan, Washington State University, “Defense signaling and response pathways in plants against pests” Keynote Address—Chair, Brian J. Staskawicz David Baltimore, California Institute of Technology, “Isn't microbiology out-dated?”

v


INTRODUCTION

8752

Introduction

Noel Keen*, Brian Staskawicz†, John Mekalanos‡, Frederick Ausubel§, and R. James Cook¶ * Department of Plant Pathology, University of California, Riverside, CA 92521; † Department of Plant and Microbial Biology, University of California, Berkeley, CA 94720; ‡ Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA 02115; § Department of Molecular Biology, Massachusetts General Hospital, Boston, MA 02114; and ¶ Departments of Plant Pathology and Crop and Soil Sciences, Washington State University, Pullman, WA 99164 History was made December 9–11, 1999, at the Beckman Center in Irvine, CA with the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals.” This was the first colloquium dedicated to the discussion of virulence mechanisms shared by plant and animal pathogens and defense mechanisms shared by plants and animals. It has become clear from the commonality in microbial virulence mechanisms and the occurrence of similar innate resistance systems in animals and plants that all of these mechanisms have an ancient and intertwined history. It also is becoming increasingly evident that susceptibility or resistance to disease involves subtle and highly specific exchanges of molecular signals between pathogens and their hosts and that understanding them can provide new approaches to controlling diseases. The colloquium provided a remarkable closure to a century that began with only a primitive understanding of the microorganisms that cause disease in plants and animals. Indeed, it was only because of breakthroughs of the past decade in understanding the molecular biology of microbial virulence and eukaryote defense that the need to bring the plant and animal fields together became apparent. In a meeting highlight, David Baltimore emphasized in his closing keynote lecture that lessons from the study of microbiology and microbial pathogens continue to greatly influence science. The animal systems discussed during the colloquium ranged from humans to insects and an array of microbial pathogens. Several plant–pathogen systems were considered, including those involving the genetic model plant, Arabidopsis thaliana. The genetic tractability of this plant, particularly its amenability for efficient mutant screens, offers experimental advantages not present in many other eukaryote model systems. A feature of pathogenic microorganisms that attracted considerable attention was the frequent use of conserved type III secretion systems by both plant and animal bacterial pathogens to introduce virulence determinants into host cells. Indeed, some of the delivered effector molecules are also functionally similar in plant and animal cells. For example, some pathogen effector molecules block animal and plant defense reactions or alter host cell structure and function to accommodate pathogen development. Pathogen effector molecules also often have interesting and unique structures, leading to the suspicion that they have resulted from long and intense evolution. There is currently great interest in determining the precise localization and functions of these molecules in animal and plant cells because such information opens the door for therapy. The acquisition and evolution of these molecules by microbial pathogens is also a topic receiving considerable attention. Although pathogen effector molecules generally are considered to increase virulence or otherwise abet development of a pathogen in its host, plants and animals have evolved surveillance systems to co-opt microbial effectors and use them as cues for initiation of defense mechanisms. This long-standing evolutionary cat and mouse game is being rapidly elucidated, particularly in plant–pathogen systems. Pathogens appear to have evolved functionally overlapping and redundant effector virulence molecules to confound host surveillance, and plants have responded by directing surveillance mechanisms to the particular subcellular sites of effector molecule virulence activity. The common features of inducible defense used by plants and animals are, in some ways, even more surprising than the pathogen side. For example, active oxygen species and nitric oxide are shared signaling mechanisms, as are lipid systems involving phospholipase activations. Furthermore, it recently has also become clear that leucine-rich repeat (LRR) proteins (such as Toll in Drosophila and so-called “disease resistance gene” proteins in plants) are conserved in active defense of vertebrates, insects, and plants, and these systems all show features of the well-studied vertebrate NF- B pathway. For animal systems, attention was focused on innate immunity, which is a rapid defense response independent from the well-known acquired immunity involving circulating antibodies. In insects, the corresponding pathogen defense system is called humoral immunity and involves the ultimate production of antimicrobial peptides. For plant systems, the equivalent mechanism of defense is the hypersensitive response, again culminating with the release of pathogen-antagonistic molecules. Structural studies on the conserved LRR proteins discussed at the colloquium promise to greatly influence our view of disease control, including the eventual design of custom LRR proteins to target new pathogen effector ligands. Plant resistance phenotypes expressed as a hypersensitive response typically follow a “gene-for-gene” model first described genetically by H. H. Flor in the late 1940s. According to this model, that probably occurs in most eukaryotes, plant genes coding for resistance surveillance proteins are matched by complementary genes in the pathogen that lead to production of the recognized effector ligands. As noted earlier, these pathogen effectors are generally, if not always, virulence determinants— the plant strategy is to impose a penalty on the pathogen if it mutates to lose the effector ligand. At the Irvine colloquium, there was considerable discussion of pathogen genes that biochemically interact with genes in vertebrates, insects, and plants. In many cases, pathogen effectors (often introduced by a type III secretion system) are thought to interact either at the host cell surface or at specific subcellular locations within host cells. In vertebrates and insects, as well as plants, these initial recognition events trigger

This paper is the introduction to the following papers, which were presented at the National Academy of Sciences colloquium “Virulence and Defense in Host-Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA.


INTRODUCTION

8753

cascades of signal tranduction events, eventually leading to the activation of genes whose products antagonize pathogen development. It was clear at the colloquium that the advent of microarray gene expression technologies will greatly assist in inventorying these genes and those encoding microbial effector functions. Such technical advances and the conceptual realization of evolutionary commonalities have put scientists interested in diseases of plants and animals on a solid footing for future manipulation of the respective defense responses to minimize the threat of diseases.


STRIKING A BALANCE: MODULATION OF THE ACTIN CYTOSKELETON BY SALMONELLA

8754

Colloquium Striking a balance: Modulation of the actin cytoskeleton by Salmonella Jorge E. Galán* and Daoguo Zhou Section of Microbial Pathogenesis, Boyer Center for Molecular Medicine, Yale School of Medicine, New Haven, CT 06536-0812 Salmonella spp. have evolved the ability to enter into cells that are normally nonphagocytic. The internalization process is the result of a remarkable interaction between the bacteria and the host cells. Immediately on contact. Salmonella delivers a number of bacterial effector proteins into the host cell cytosol through the function of a specialized organelle termed the type III secretion system. Initially, two of the delivered proteins, SopE and SopB, stimulate the small GTP-binding proteins Cdc42 and Rac. SopE is an exchange factor for these GTPases, and SopB is an inositol polyphosphate phosphatase. Stimulation of Cdc42 and Rac leads to marked actin cytoskeleton rearrangements, which are further enhanced by SipA, a Salmonella protein also delivered into the host cell by the type III secretion system. SipA lowers the critical concentration of G-actin, stabilizes F-actin at the site of bacterial entry, and increases the bundling activity of the host-cell protein T-plastin (fimbrin). The cellular responses stimulated by Salmonella are short-lived; therefore, immediately after bacterial entry, the cell regains its normal architecture. Remarkably, this process is mediated by SptP, another target of the type III secretion system. SptP exert its function by serving as a GTPase-activating protein for Cdc42 and Rac, turning these G proteins off after their stimulation by the bacterial effectors SopE and SopB. The balanced interaction of Salmonella with host cells constitutes a remarkable example of the sophisticated nature of a pathogen/host relationship shaped by evolution through a longstanding coexistence. When examining the strategies used by microbial pathogens to colonize and multiply within a host, it is useful to make the distinction between pathogens that may only accidentally encounter a particular host and those that have sustained a longstanding association. The distinction is relevant because in the first case, the terms of the host/pathogen interactions have not been shaped by evolutionary forces. Therefore, infections by this type of microorganisms oftentimes lead to serious or lethal disease. In contrast to “accidentally encountered” pathogens, the interaction between “host-adapted” or “quasi-adapted” pathogens and their hosts has been shaped by evolution, resulting in balanced encounters that most often lead to infections that are either subclinical or self-limiting. In some of these cases, the process of coevolution and adaptation has effectively precluded this type of microorganisms from exploring other niches. To be sure, the evolutionary forces may have dictated that infection leads to a certain degree of host pathology that is not acceptable for society, and therefore the outcome of such interaction is referred to as “disease.” However, this outcome does not mean that the interactions between these microorganisms and their hosts are “unbalanced” in evolutionary terms. It may simply mean that a certain degree of host damage is required for these pathogens to replicate and to move on to a new host and/or for the host to expel them. Salmonella enterica is an excellent example of a bacterial pathogen whose interactions with vertebrate animals have been shaped by millions of years of close coexistence. There are different serotypes of Salmonella enterica that cause different diseases. The type of disease caused by these bacteria range from self-limiting gastroenteritis (commonly referred to as “food poisoning”) to more systemic illness such as typhoid fever. The outcome of Salmonella infections depends on both the serotype of the infecting Salmonella as well as the species and/or the immunological status of the infected hosts. Central to the pathogenesis of Salmonella is the function of a specialized protein secretion system, termed type III, that is encoded within a pathogenicity island located at centisome 63 of its chromosome (1). The nucleotide composition of this pathogenicity island suggests that these genes have been acquired by horizontal transfer from another microorganism. This event most likely took place early in the evolution of Salmonella, resulting in a significant niche expansion for these bacteria, perhaps marking the beginning of its close association with vertebrate hosts. This type III protein secretion system directs the secretion and translocation into the host cell of a number of bacterial proteins that have the capacity to modulate a variety of host cellular functions (2, 3). Among these functions is the ability of Salmonella to modulate the host cell actin cytoskeleton to induce its own internalization into nonphagocytic cells, which is essential for its pathogenicity. This process is the result of the coordinated activities of several bacterial effector proteins that alternatively stimulate and down-regulate host cell responses in a remarkable “yin and yang.” In this article, we will discuss the molecular mechanisms that lead to Salmonella internalization into non-phagocytic cells. However, this paper is not intended to be a comprehensive review of the literature on Salmonella–host cell interaction. Instead, we have chosen to focus on key aspects of the entry process that better illustrate the remarkable balance between a pathogen and its host.

THE CENTISOME 63 TYPE III SECRETION SYSTEM: SALMONELLA'S KEY TO ENTER INTO HOST CELLS At the center of the mechanisms used by Salmonella to gain access into nonphagocytic cells is the type III secretion system encoded at centisome 63 of its chromosome (1). Salmonella encodes another type III secretion system at centisome 31 that is only functional at later stages of the pathogenic cycle after the bacteria has entered host cells (4, 5). Type III secretion systems are widely distributed among plant and animal pathogenic bacteria that share the property of engaging host cells in an intimate manner (for detailed reviews see refs.2 and 3). Composed of more than 20 proteins, these systems stand among the most complex protein secretion systems known. Such complexity is caused by their specialized function, which is not only to secrete proteins from the bacterial cytoplasm but

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviation: GAP, GTPase activating protein. * To whom reprint requests should be addressed. E-mail: [email protected].



8755

Fig. 1. Needle complex of S. typhimurium type III secretion system. (A) Electron micrographs of osmotically shocked S. typhimurium showing needle complexes in the bacterial envelope (arrows). (B) Electron micrographs of purified needle complexes. (C) Schematic representation of the needle complex and its components. The location of the different components is hypothetical and more proteins may be present in the complex. [Figure reprintzed with permission from ref. 2 (Copyright 1999, American Association for the Advancement of Science).] also to deliver them to the inside of the eukaryotic host cell. Perhaps a more important factor contributing to their complexity is the temporal and spatial restrictions that govern their activity. Thus, the function of type III secretion systems requires poorly characterized signals that cue the bacteria to secrete and deliver proteins at the appropriate time and in the appropriate environment. A number of components of the type III secretion system assemble into an organelle, appropriately termed the “needle complex,” that spans both the inner and outer membranes of the bacterial envelope (ref.6; Fig. 1). The architecture of the needle complex resembles that of the flagellar hook-basal body (7, 8). It is composed of two pairs of inner and outer rings that presumably anchor the structure to the inner and outer membranes of the bacterial envelope. The rings are connected by a rod-like structure, which together form the base of the needle complex. A needle-like structure of 80 nm in length protrudes outwards from the base of the needle complex. The entire structure is 100 nm in length and 40 nm in diameter at its widest section. The identification of the needle complex has provided important clues in understanding the transit of type III secreted proteins through the bacterial envelope. However, this information has been less useful in elucidating the mechanisms by which this system mediates the translocation of the secreted proteins into the host cell. The latter process depends on the function of a subset of type III secreted proteins that, although dispensable for protein secretion, are essential for the translocation of effector proteins into the host cell. In Salmonella, those proteins are SipB, SipC, and SipD (9, 10, 11 and 12). The mechanisms by which these “protein translocases” mediate the delivery of effector proteins into the eukaryotic host cell are not understood. However, functionally equivalent proteins in other type III secretion systems (e.g., Yersinia spp.) have been proposed to form a pore or channel through which the effector proteins cross the eukaryotic cell membrane (13, 14). Another important component of type III secretion systems is a family of small molecular weight acidic polypeptides that bind a specific subset of cognate cytosolic proteins (15). Although absolutely required for the function of the type III secretion systems, the actual mechanisms by which these proteins exert their chaperone-like function is poorly understood and is the subject of some controversy. At least two functions have been proposed for these proteins: (i) partitioning factors that prevent the premature association of type III secreted proteins within the bacterial cytoplasm, and (ii) secretion pilots that “guide” the cognate secreted protein to the secretion machinery. It is possible that different chaperones may exert different functions. The Salmonella invasion-associated type III secretion system encodes at least three such chaperonelike proteins: SicP (16), SicA (12), and SigD (17, 18). SicP serves as chaperone for the effector protein SptP (16). Consistent with this role, SicP binds SptP, which in its absence completely is degraded within the bacterial cytoplasm. Thus, SicP seems to function as a partitioning factor for SptP, perhaps preventing it from interacting with an as-yet-unidentified protein. SicA, however, appears to play a more complex role (19). One of its functions is to prevent the association of SipB and SipC in the bacterial cytosol that would target these proteins for degradation. Absence of SicA results in the degradation of both SipB and SipC. Interestingly, in the absence of both SicA and SipC, SipB is not only stable but also is secreted at wild-type levels, indicating that SicA is not essential for SipB secretion per se. In addition, absence of SicA results in the lack of expression of several genes that encode type III effector proteins, suggesting another function for this protein (19). It is possible that some of the chaperones associated with type III secretion systems exert a role in the temporal regulation of substrate secretion by the secretion machinery. Indeed, evidence has been presented that indicates that there is a hierarchy in the secretion of type III proteins in Salmonella (20, 21).



8756

Fig. 2. Interaction of S. typhimurium with intestinal epithelial cells. (A) Scanning electron micrograph of S. typhimuriuminfected intestinal epithelial Caco-2 cells (S. Olmsted, C. Ginocchio, C. Wells, and J.E.G., unpublished data). (B and C) Actin cytoskeleton rearrangements in S. typhimurium-infected intestinal epithelial Caco-2 cells. Filamentous actin was stained with rhodamine phalloidin (red) and S. typhimurium with an FITC-conjugated antibody (green). REVVING UP RHO GTPASES: SIGNALING FOR ENTRY AND BEYOND Salmonella entry into host cells strictly depends on the function of the actin cytoskeleton as addition of drugs that interfere with actin dynamics effectively block bacterial internalization (22, 23). Furthermore, immediately after contact with host cells, Salmonella induces rapid and marked actin cytoskeleton rearrangements (Fig. 2). It was recognized early on that these changes resemble the membrane ruffles induced by the stimulation of cells with growth factors or the expression of activated oncogenes (24, 25). An important insight into the understanding of the cellular responses that lead to bacterial internalization came from the finding that Cdc42 and Rac, two members of the Rho subfamily of actinorganizing small GTP-binding proteins, were essential for Salmonella entry into host cells (26). These small molecular weight GTP-binding proteins can cycle between two states, a GDP-bound (inactive) and a GTP-bound (active) conformation that can bind a variety of downstream effector proteins. Therefore, they can act as effective molecular switches

Fig. 3. Actin cytoskeleton reorganization induced by the expression of SopE and SopB in cultured mammalian cells. Cos-1 cells were transfected with plasmids expressing SopE or SopB or with the empty vector and were stained with rhodamine phalloidin to visualize the actin cytoskeleton.



8757

to control signaling events in a temporal and spatial manner. Expression of dominant-negative mutants of these GTPases either completely (Cdc42N17) or partially (Rac-1N17) blocked bacterial entry into cultured cells. Stimulation of these GTPases by Salmonella also leads to the activation of the downstream mitogen-activated protein kinases Jnk and p38 (26, 27). This finding is significant for Salmonella pathogenesis because the type III secretion-dependent activation of the mitogen-activated protein kinase pathways leads to the stimulation of the transcription factors NF-KB and AP-1 (27) and to the production of proinflammatory cytokines (27, 28 and 29). These events are presumed to be essential for the establishment of the inflammatory diarrhea that follows Salmonella infection. How does Salmonella engage the Rho-GTPase signaling pathways? The answer to this question became clear when it was found that one of the proteins delivered by the invasionassociated type III secretion system, termed SopE, was capable of directly activating both Cdc42 and Rac-1 (30). SopE was found to catalyze the exchange of GDP for GTP on several (but not all) members of the Rho GTPase subfamily of small G proteins (Rac-1, Rac-2, Cdc42, RhoG, and RhoA, but not RhoB or RhoC). Consistent with this activity, microinjection or transient expression of SopE in cultured cells leads to marked actin cytoskeleton rearrangements and membrane ruffling that resemble the changes induced by Salmonella infections (Fig. 3). These cytoskeleton rearrangements can be blocked by coexpression of dominant-negative forms of either Cdc42 or Rac-1 (30). These results position SopE as a key stimulator of signaling events leading to bacterial entry. Introduction of a loss-of-function mutation in sopE reduced but did not abolish the ability of Salmonella to enter host cells (31, 32). Because expression of dominant-negative Cdc42 abolished the ability of Salmonella to enter into host cells, these results indicated that, in addition to SopE, there must be another bacterial protein capable of stimulating small GTPase signaling (26). Recently, such a factor was identified as SopB, another protein delivered by Salmonella into host cells by its invasionassociated type III secretion system (D.Z., L.-M. Chen, S. Shears, and J.E.G., unpublished data). SopB is an inositol phosphate polyphosphatase with the potential to generate several products with second messenger activity (17, 34). Consistent with its role in the actin-mediated bacterial internalization process, transient expression of SopB in mammalian cells leads to actin cytoskeleton rearrangements resembling those induced by the bacteria (D.Z., L.-M. Chen, S. Shears, and J.E.G., unpublished data, and Fig. 3). Furthermore, a Salmonella strain defective in both sopE and sopB is completely defective in its ability to stimulate actin cytoskeleton rearrangements and membrane ruffling. Interestingly, expression of a dominant-negative mutant of Cdc42 but not Rac-1 effectively blocked SopB-mediated actin cytoskeleton rearrangements. Consistent with this finding, internalization of a sopE mutant strain of Salmonella (i.e., a strain internalized only by the function of SopB) was blocked by expression of a dominant-negative mutant of Cdc42 but not Rac-1 (D.Z., L.-M. Chen, S. Shears, and J.E.G., unpublished data). These results indicate that unlike SopE, which exerts its effect on both Cdc42 and Rac-1 (see above), SopB appears to restrict its function to Cdc42. At this point it is not known how SopB stimulates Cdc42 activation. It is possible that some of the phosphoinositide products of the SopB activity either may activate an endogenous cellular exchange factor (e.g., Dbl) or may directly stimulate nucleotide exchange.

Fig. 4. Host-cell signal transduction pathways stimulated by Salmonella through the invasion-associated type III secretion system.

Fig. 5. Role of the Salmonella type III secreted protein SipA and the host-cell protein T-plastin (fimbrin) in bacterial-induced actin remodeling. (Left) Panels show electron micrographs of negatively stained actin, actin plus SipA, and actin plus SipA and plastin. Note the increased bundling activity of T-plastin in the presence of SipA. (Right) Panels show the recruitment of Tplastin to the membrane ruffles induced by either Salmonella or the transient expression of the bacterial effector SopE. Cells were transfected with hemagglutinin-epitope tagged T-plastin (fimbrin) and either infected with S. typhimurium or cotransfected with a plasmid expressing SopE. Cells were stained with a monoclonal antibody directed to the epitope tag. For details on these experiments, see ref. 42.



8758

Fig. 6. SptP mediates the recovery of the normal organization of the actin cytoskeleton after S. typhimurium internalization. Ref52 cells were infected with either wild-type or a ∆sptP mutant strain for 3 h, and the actin cytoskeleton was visualized by rhodamine phalloidin staining. Bacteria were visualized by 4,6-diamidino-2-phenylindole (DAPI) staining. Phalloidin and DAPI images were captured individually and merged using Adobe PHOTOSHOP. [Figure reprinted with permission from ref. 21 (Copyright 1999, Macmillan Magazines, Ltd.), http://www.nature.com.] DOWNSTREAM SIGNALING: EFFECTORS OF CDC42 AND RAC-1 FUNCTION Stimulation of Rho GTPases leads to a variety of cellular responses, including actin remodeling, stimulation of nuclear responses, and modulation of cell cycle progression (35, 36). Although the mechanisms by which small GTP-binding proteins modulate all these activities are not well understood, it is assumed that they exert their functions by engaging a variety of downstream effectors that bind the GTP-bound (active) form of these regulators. In many instances, mutation analysis of these GTPases has allowed the identification of specific residues, termed “effector loops,” that are involved in engaging specific downstream signaling pathways. Stimulation of Rho GTPases by Salmonella leads to at least two clearly defined responses: (i) actin cytoskeleton rearrangements resulting in bacterial uptake, and (ii) the stimulation of the mitogenactivated protein kinase pathways leading to nuclear responses. It has been shown that these responses are mediated by different downstream effectors (ref.37; Fig. 4). For example, the stimulation of nuclear responses depends on the activity of p21-activated kinase (PAK) (37). Consistent with this involvement, Salmonella infection or transient SopE expression lead to a robust and rapid activation of PAK (37). Furthermore, expression of a dominant-negative kinase-defective mutant form of PAK effectively blocks Salmonella or SopE-induced JNK activation. However, expression of dominant-negative PAK does not block Salmonella-induced actin cytoskeleton rearrangements or bacterial internalization, indicating that the nuclear and morphological responses induced by the bacteria are mediated by different downstream effectors. The identity of the downstream targets of Cdc42 and Rac-1 that mediate bacterial-induced actin rearrangements are not known. However, the effector loop of Cdc42 that is necessary to stimulate bacterial-induced actin rearrangements has been identified (37). This loop specifically binds effector proteins that contain a conserved 16-aa motif termed CRIB (CDC42/Rac interacting binding) or PBD (p21-binding domain). Therefore, it is expected that the effector protein(s) involved in the stimulation of actin rearrangements must contain such a domain. A very good candidate is the Wiskott Aldrich Syndrome protein (WASP), which has been shown to be involved in Cdc42-mediated modulation of the actin cytoskeleton (38). Another downstream effector potentially involved in actin remodeling is phospholipase A2 (PLA2). Previous studies have shown that Salmonella infection of cultured intestinal cells leads to the activation of PLA2 and to the subsequent production of arachidonic acid metabolites and Ca2+ fluxes, events required for bacterial entry (39). Subsequent studies have demonstrated that Rac stimulation leads to the activation of PLA2 and the production of arachidonic acid metabolites (40). It is therefore likely that the calcium fluxes stimulated by Salmonella are also the consequence of the activation of Cdc42 and Rac by SopE and SopB.

FINE-TUNING THE ACTIN CYTOSKELETON REARRANGEMENTS: THE ROLE OF THE ACTINBINDING PROTEIN SIPA Microinjection or transient expression in host cells of the bacterial-encoded exchange factor SopE leads to actin cytoskeletal rearrangements and membrane ruffling. These morphological changes resemble the actin reorganization induced by Salmonella infections but they differ in one important aspect. Whereas the morphological changes induced by the transient expression of SopE are distributed throughout the cells, those stimulated by Salmonella are localized to the point of bacterial host cell contact. Although localized delivery of effector proteins by the bacteria may account in part for signal localization, evidence indicates that at least one Salmonella type III secreted effector protein is actively engaged in this process. A strain of Salmonella typhimurium that carries a loss-of-function mutation in sipA, a gene that encodes a type III secreted protein, is significantly impaired in its ability to induce localized actin cytoskeleton reorganization, despite the fact that this strain is not impaired in its ability to deliver other effector proteins (41). The sipA mutant induces diffuse actin cytoskeleton rearrangements and consequently is impaired in its ability to enter cells. A search for a cellular target of this bacterial effector protein revealed that SipA binds to filamentous (F) actin (41). The binding of SipA to actin results in a significant reduction in the concentration of monomeric actin required for polymerization (critical concentration) and a marked increase in the stability of F-actin (Fig. 5). In addition, SipA binding to F-actin increases the bundling activity of Tplastin (fimbrin), an actin-binding protein that is actively recruited to membrane ruffles on Salmonella infection or transient expression of SopE (ref.42; Fig. 5). These activities may contribute to the cytoskeletal changes induced by Salmonella in various ways. Lowering the critical concentration of actin may facilitate actin nucleation on bacterial stimulation of Cdc42 and Rac. Furthermore, increasing the bundling activity of plastin as well as stabilizing F-actin may enhance the formation of membrane ruffles, resulting in more pronounced outward extension of these structures and more efficient bacterial uptake.



8759

Fig. 7. Model for S. typhimurium interaction with host cells. (I–V) Contact with host cells activates the invasion-associated type III secretion system resulting in the delivery of effector proteins (e.g., SopE, SopB, SipA, and SptP) (I). Delivery of different effectors may not happen at the same time. Introduction of the exchange factor SopE and the inositol polyphosphatase SopB results in the activation of Cdc42 and Rac-1 (II), the stimulation of downstream signaling pathways, and the recruitment of plastin and other ruffling-associated molecules (III) initiating the actin cytoskeleton reorganization. The bacterial effector protein SipA helps this process by lowering the critical concentration of actin, stimulating the bundling activity of plastin, and stabilizing F-actin (IV). This activity results in a localized and more pronounced outward extension of the ruffling process (IV). Subsequent to the stimulation of Cdc42 and Rac-1 by SopE and SopB, Salmonella delivers another effector protein, SptP, which reverses the activation of these small G proteins by stimulating their intrinsic GTPase activity and therefore facilitating cell recovery (V). An additional type III secreted protein, SipC, recently has been reported to have actin-nucleating activity (43). It is possible that this activity may aid bacterial entry by contributing to the actin nucleation events that lead to membrane ruffling and bacterial uptake. However, SipC by itself is not sufficient to induce actin cytoskeleton rearrangements because an S. typhimurium sopE sopB double-mutant strain is unable to stimulate these cellular responses despite having an intact type III secretion system and wild-type SipC expression (D.Z., L.-M. Chen, S. Shears, and J.E.G., unpublished data). The requirement of SipC for the translocation of all type III secreted proteins (9, 10) has precluded the use of standard genetic analysis to evaluate the contribution of SipC's actin-nucleating activity to bacterial internalization. The isolation of a mutant defective in actin nu



8760

cleation but proficient in protein translocation will be required to assess the contribution and significance of the actin-binding activity of SipC in bacterial entry.

PUTTING ON THE BRAKES: A LESSON ON SELF-RESTRAINT In his pioneering electron microscopic description of the interaction of S. typhimurium with the intestinal epithelium of infected animals, Akio Takeuchi made the observation that the morphological changes of the brush border induced by the bacteria were reversible (44). He observed that shortly after infection, the brush border of intestinal epithelial cells infected by Salmonella regained their normal appearance despite the presence of numerous bacteria within an intracellular compartment. The molecular bases for the reversion of these marked morphological changes remained unknown until recently. The finding that Salmonella induces its own internalization by activating the small GTP binding proteins Cdc42 and Rac suggested the possibility that the endogenous cellular mechanisms controlling the down-regulation of these GTPases may mediate the reversion of the bacterial-induced responses. However, micro-injection into cultured cells of purified SopE in amounts equivalent to those injected by the bacterial type III secretion system led to actin cytoskeleton rearrangements and membrane ruffling that continued on for periods of time that far exceeded those resulting from bacterial infection (30). This result suggested a more active participation of Salmonella in mediating cellular recovery after bacterial infection. The finding that cells infected with a S. typhimurium strain carrying a null mutation in sptP failed to recover the integrity of their actin cytoskeleton after bacterial infection further supported this hypothesis (ref.21; Fig. 6). SptP is secreted by the invasion-associated type III secretion system and contains two modular domains: (i) an amino terminal domain with sequence similarity to the YopE and ExoS toxins of Yersinia spp. and Pseudomonas aeruginosa, respectively, and (ii) a carboxyl terminal domain homologous to Yersinia YopH and eukaryotic tyrosine phosphatases (10, 45). How does SptP exert its function? A mechanism for its biochemical function was suggested by the observation that the actin cytoskeleton rearrangements stimulated by the exchange factor SopE could be effectively blocked by the comicroinjection of SptP, suggesting a Cdc42/Rac-1 antagonistic function for this protein (21). G proteins have an intrinsic GTPase activity that allows them to turn themselves off after activation by switching to the GDP-bound (inactive) conformation (46). However, such intrinsic activity is very low unless in the presence of GTPase activating proteins (GAPs), which can stimulate the intrinsic GTPase activity by several orders of magnitude (47). Consistent with its antagonistic function, SptP was found to be a potent GAP for Cdc42 and Rac-1 but not for other GTPases from the Rho or more distantly related families (21). The GAP activity is encoded in the amino terminal domain of SptP. Consistent with this finding, it now has been reported that the related bacterial toxins ExoS and YopE are also GAPs for Rho GTPases (33). Although different GAPs often do not share strong sequence similarity, they do share a short sequence motif that contains a conserved arginine that is essential for catalysis (47). SptP, ExoS, and YopE exhibit such a motif indicating a similarity between the catalytic mechanism of these bacterial proteins and those of eukaryotic GAP proteins (21). Consistent with this hypothesis, an SptP mutant in which the invariant arginine has been changed to an alanine is totally devoid of GAP activity. This mutant also is unable to reverse the actin cytoskeleton rearrangements that follow Salmonella infection, therefore confirming the role of the SptP GAP activity in this process. SptP not only reverses the actin cytoskeletal changes but also prevents the potential harm to the host cell derived from excessive signaling through Cdc42 and Rac that may lead to apoptosis (21). This function may allow the bacteria to preserve the integrity of its intracellular niche long enough to permit its replication and to allow reprogramming of gene expression, a necessary step to continue with the next phase of its pathogenic life cycle.

LESSONS LEARNED FROM SALMONELLA Salmonella entry into host cells requires the coordinated action of several bacterial effector proteins that, on delivery into the host cell, exert their activity in a temporally coordinated manner (Fig. 7). Thus, activation of Cdc42 and Rac by SopE and SopB is followed by actin cytoskeleton rearrangements that are further modulated by the activity of the actin-binding protein SipA. The bacterial-induced cellular responses subsequently are reversed by another bacterial effector protein, SptP, which opposes the activities of SopB and SopE. In addition, it is likely that the effector molecules themselves may have regulatory domains that control their activity. The existence of other putative effector proteins of unknown function that also are delivered into host cells by the centisome 63 type III secretion system indicates that we are just beginning to understand the complexities of this system. The interaction of Salmonella with host cells is therefore an eloquent example of the sophisticated nature of the mechanisms used by bacterial pathogens that have sustained long-standing associations with their hosts. Such sophistication is the result of evolutionary forces operating over extended periods of time, leading to a rather balanced interaction that allows bacterial replication while preventing excessive harm to the host. Because the study of microbial pathogens commonly focuses in the examination of the events that lead to overt harm to the host, it is frequently overlooked that the interaction of these “adapted pathogens” with their hosts most often does not lead to overt disease. In fact, the Salmonella example indicates that the bacteria can be actively engaged in preventing overt harm. This fact is particularly important when considering the design of novel therapeutic and prevention strategies because inadvertent interference with “downmodulators of virulence” could result in more harm to the host. As we increase our understanding of the molecular mechanisms that govern the interaction of adapted pathogens with their hosts, we will undoubtedly find more examples of “pathogen self-restraint.” We thank members of J.E.G.'s laboratory for careful review of this manuscript. Work in J.E.G.'s laboratory that was discussed in this article was supported by Grants AI30492 and GM52543 from the National Institutes of Health. 1. Galán, J. E. (1999) Curr. Opin. Microbiol. 2, 46–50. 2. Galán, J. E. & Collmer, A. (1999) Science 284, 1322–1328. 3. Hueck, C. J. (1998) Microbiol. Mol. Biol. Rev. 62, 379–433. 4. Ochman, H., Soncini, F. C., Solomon, F. & Groisman, E. A. (1996) Proc. Natl. Acad. Sci. USA 93, 7800–7804. 5. Shea, J. E., Hensel, M., Gleeson, C. & Holden, D. W. (1996) Proc. Natl. Acad. Sci. USA 93, 2593–2597. 6. Kubori, T., Matsushima, Y., Nakamura, D., Uralil, J., Lara-Tejero, M., Sukhan, A., Galán, J. E. & Aizawa, S.-I. (1998) Science 280, 602–605. 7. Aizawa, S. I. (1996) Mol. Microbiol 19, 1–5. 8. Macnab, R. M. (1996) in Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, ed. Neidhardt, F. C. (Am. Soc. Microbiol., Washington DC), 2nd Ed., pp. 123–145. 9. Collazo, C. & Galán, J. E. (1997) Mol. Microbiol. 24, 747–756. 10. Fu, Y. & Galán, J. E. (1998) Mol. Microbiol. 27, 359–368. 11. Kaniga, K., Trollinger, D. & Galán, J. E. (1995) J. Bacterial. 177, 7078–7085. 12. Kaniga, K, Tucker, S. C., Trollinger, D. & Galán, J. E. (1995) J. Bacteriol. 177, 3965–3971. 13. Hakansson, S., Schesser, K., Persson, C, Galyov, E. E., Rosqvist, R., Homble, F. & Wolf-Watz, H. (1996) EMBO J. 15, 5812–5823. 14. Neyt, C. & Cornelis, G. R. (1999) Mol. Microbiol. 33, 971–981. 15. Wattiau, P., Woestyn, S. & Cornelis, G. R. (1996) Mol. Microbiol. 20, 255–262.



8761

16. Fu, Y. & Galán, J. E. (1998) J. Bacteriol. 180, 3393–3399. 17. Norris, F. A., Wilson, M. P., Wallis, T. S., Galyov, E. E. & Majerus, P. W. (1998) Proc. Natl. Acad. Sci. USA 95, 14057–14059. 18. Hong, K. H. & Miller, V. L. (1998) J. Bacteriol. 180, 1793–1802. 19. Tucker, S. & Galán, J. E. (2000) J. Bacteriol., in press. 20. Collazo, C. & Galán, J. E. (1996) Infect. Immun. 64, 3524–3531. 21. Fu, Y. & Galán, J. E. (1999) Nature (London) 401, 293–297. 22. Kihlstrom, E. & Nilsson, L. (1977) Acta Pathol. Microbiol. Scand. 85, 322–328. 23. Buckholm, G. (1984) Acta Pathol. Microbiol. Immun. Scand. 92, 145–149. 24. Galán, J. E., Pace, J. & Hayman, M. J. (1992) Nature (London) 357, 588–589. 25. Francis, C. L., Ryan, T. A., Jones, B. D., Smith, S. J. & Falkow, S. (1993) Nature (London) 364, 639–642. 26. Chen, L. M., Hobbie, S. & Galán, J. E. (1996) Science 274, 2115–2118. 27. Hobbie, S., Chen, L. M., Davis, R. & Galán, J. E. (1997) J. Immunol. 159, 5550–5559. 28. Eckmann, L., Stenson, W. F., Savidge, T. C., Lowe, D. C., Barrett, K. E., Fierer, J., Smith, J. R. & Kagnoff, M. F. (1997) J. Clin. Invest. 100, 296–309. 29. Jung, H. C., Eckmann, L., Yang, S.-K., Panja, A., Fierer, J., Morzycka-Wroblewska, E. & Kagnoff, M. F. (1995) J. Clin. Invest. 95, 55–65. 30. Hardt, W.-D., Chen, L.-M., Schuebel, K. E., Bustelo, X. R. & Galán, J. E. (1998) Cell 93, 815–826. 31. Hardt, W.-D., Urlaub, H. & Galán, J. E. (1998) Proc. Natl. Acad. Sci. USA 95, 2574–2579. 32. Wood, M. W., Rosqvist, R., Mullan, P. B., Edwards, M. H. & Galyov, E. E. (1996) Mol. Microbiol. 22, 327–338. 33. Goehring, U. M., Schmidt, G., Pederson, K. J., Aktories, K. & Barbieri, J. T. (1999) J. Biol. Chem. 274, 36369–36372. 34. Galyov, E. E., Wood, M. W., Rosqvist, R., Mullan, P. B., Watson, P. R., Hedges, S. & Wallis, T. S. (1997) Mol. Microbiol. 25, 1903–1912. 35. Hall, A. (1998) Science 279, 509–514. 36. Van Aelst, L. & D'Souza-Schorey, C. (1997) Genes Dev. 11, 2295–2322. 37. Chen, L. M., Bagrodia, S., Cerione, R. A. & Galán, J. E. (1999) J. Exp. Med. 189, 1479–1488. 38. Symons, M., Derry, J. M. J., Karlak, B., Jiang, S., Lemahieu, V., McCormick, F., Francke, U. & Abo, A. (1996) Cell 84, 723–734. 39. Pace, J, Hayman, M. J. & Galán, J. E. (1993) Cell 72, 505–514. 40. Peppelenbosch, M. P., Qiu, R. G., de Vries-Smits, A. M., Tertoolen, L. G., de Laat, S. W., McCormick, F., Hall, A., Symons, M. H. & Bos, J. L. (1995) Cell 81, 849–856. 41. Zhou, D., Mooseker, M. & Galán, J. E. (1999) Science 283, 2092–2095. 42. Zhou, D., Mooseker, M. S. & Galán, J. E. (1999) Proc. Natl. Acad. Sci. USA 96, 10176–10181. 43. Hayward, R. D. & Koronakis, V. (1999) EMBO J. 18, 4926–4934. 44. Takeuchi, A. (1967) Am. J. Pathol. 50, 109–136. 45. Kaniga, K., Uralil, J., Bliska, J. B. & Galán, J. E. (1996) Mol. Microbiol. 21, 633–641. 46. Bourne, H. R., Sanders, D. A. & McCormick, F. (1990) Nature (London) 348, 125–132. 47. Scheffzek, K., Ahmadian, M. R. & Wittinghofer, A. (1998) Trends Biochem. 23, 7257–7262.


STRUCTURE AND FUNCTION OF PECTIC ENZYMES: VIRULENCE FACTORS OF PLANT PATHOGENS

8762

Colloquium Structure and function of pectic enzymes: Virulence factors of plant pathogens Steven R. Herron*, Jacques A. E. Benen†, Robert D. Scavetta‡, Jaap Visser†, and Frances Jurnak*§ * Department of Physiology and Biophysics, University of California, Irvine, CA 92697; † Department of Molecular Genetics of Industrial Microorganisms, Wageningen Agricultural University, Dreijenlaan 2, 6703 HA Wageningen, The Netherlands; and ‡ Department of Pharmacology, C236, University of Colorado Health Sciences Center, 4200 East Ninth Avenue, Denver, CO 80262 The structure and function of Erwinia chrysanthemi pectate lysase C, a plant virulence factor, is reviewed to illustrate one mechanism of pathogenesis at the molecular level. Current investigative topics are discussed in this paper. Plant cell walls are primarily polysaccharide in composition. A simple but major pathogenic mechanism in plants involves degradation of the cell wall by a battery of polysaccharidases secreted by pathogens. Most of the degradative enzymes are glycoside hydrolases, which degrade the cellulose and pectate matrices by the addition of water to break the glycosidic bonds. The pectate network is also degraded by polysaccharide lyases, which cleave the glycosidic bonds via a β-elimination mechanism. To better understand the latter virulence mechanism, research has been carried out on pectate lyase C, a pectolytic enzyme secreted by the pathogenic bacterium Erwinia chrysanthemi. The story of pectate lyase C illustrates how structural techniques have contributed to a detailed understanding of polysaccharide recognition and the lyase cleavage mechanism. In the process, a novel protein structural fold and a unique catalytic role for an arginine have been discovered. The structural results have also provided the first atomic description of a pectate fragment, which differs considerably from the popular view in conformation as well as the mode of interactions with Ca2+ ions. Finally, the growing structural database of pectolytic enzymes is enabling researchers to elucidate subtle structural differences that are responsible for the specific recognition of a unique oligosaccharide sequence from a heterogeneous mix in the plant cell wall. Such knowledge will ultimately lead to a better understanding of the characteristics that render the host susceptible to attack by a particular pathogen.

DIFFERENCE IN OUTER BARRIERS OF PLANT AND MAMMALIAN CELLS To be successful in attacking a host cell, a pathogen must pass the outer barrier of a cell. In plants, the outer barrier is the cell wall, composed primarily of polysaccharides. In mammals, the outer barrier is a membrane, composed primarily of lipids. Polysaccharides do serve a function at the mammalian membrane, either as cell surface components involved in molecular recognition or as the primary components of the gelatinous intercellular milieu. Given the prevalence of polysaccharides in the plant cell wall, it is not surprising to find that many plant pathogens secrete a battery of saccharidases as a major mode of attack. The analogous mechanism in mammalian systems is the secretion of enzymes that degrade the intercellular matrix to allow the pathogen access to the cell membrane and cellular cytoplasm.

DEGRADATIVE ENZYMES OF PLANT CELL WALL COMPONENTS The plant cell wall is composed of two types of polysaccharide matrices: the pectate network and the cellulose network (1, 2). The pectate network consists of the smooth region composed of homogalacturonans and the hairy region composed of highly branched rhamnogalacturonans. The cellulose network consists of microfibrils composed of 1-4-linked β-D-glucan units as well as xyloglucan and arabinoxylan, the two hemicelluloses that coat the cellulose microfibrils to prevent excessive aggregation. Not only does the basic saccharide unit differ in the two types of network, but the saccharide units in the pectate network also have a uronic acid or an esterified uronate moiety at the C-6 position of the saccharide units. Enzymes, which degrade the pectate network, belong to two classifications: glycoside hydrolases (3) and polysaccharide lyases (4). Glycoside hydrolases incorporate a water molecule via a general acid catalysis during the cleavage of the glycosidic bond between the two saccharide units. In contrast, polysaccharide lyases cleave the glycosidic bond via a β-elimination reaction that removes a proton. The final product contains an unsaturated bond between C-4 and C-5 of the saccharide unit at the nonreducing end. Enzymes, which degrade the cellulose network, all function as glycoside hydrolases. Generally, hydrolases have an acidic pH optima, using aspartic and glutamic acid groups during catalysis, whereas lyases have a basic pH optima, using catalytic amino acids that are still under active investigation. The degradation of plant cell walls by secreted enzymes from pathogens has been a fertile area of research for many years (5). Although it is well known that certain pathogens are effective against specific hosts, the mechanisms of host–pathogen interactions remain elusive. In certain organisms, such as Erwinia chrysanthemi, the genetic organization and regulation of many secretory saccharidases have been elucidated (6, 7). One surprising finding is that many pathogenic organisms secrete multiple isozymes of the same enzyme but the transcription of the genes is often independently regulated. Since 1990, research has been directed at the protein level, to understand the structure and enzymatic mechanism of the degradative cell wall enzymes. Additional efforts are being made to elucidate the molecular basis for protein recognition of the composition and conformation of the individual saccharide units in the substrate. Another major question, yet to be resolved, is whether there is isozymespecific recognition of a unique polysaccharide sequence and if this specificity plays an important role in pathogen–host cell interactions. The work will be illustrated by the research on Erwinia chrysanthemi pectate lyase C (PelC).

THREE-DIMENSIONAL STRUCTURES OF PLANT CELL WALL DEGRADATIVE ENZYMES PelC is one of several isozymes secreted by E. chrysanthemi that cleave the pectate component of the plant cell wall, causing

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviation: PelC, pectate lyase C. § To whom reprint requests should be addressed at: Department of Physiology and Biophysics, Room 350-D Med Sci I, University of California, Irvine, CA 92697-4560. E-mail: [email protected].



8763

Fig.1. Five examples of plant cell wall degradative enzymes that fold into a parallel β helix motif. The predominant secondary structural features of the proteins are illustrated as β strands, and the coils represent α helices. (A) E. chrysanthemi pectate lyase C. (B) E. chrysanthemi pectate lyase E. (C) A. niger pectin lyase B. (D) E. carotovora polygalacturonase. (E) A. aculeatus rhamnogalacturonase A. soft-rot disease in a variety of crops (8). The pelC gene has been cloned into Escherichia coli, with the latter constructs exhibiting maceration activity comparable to E. chrysanthemi (9). By using the recombinant form from E. chrysanthemi, the threedimensional structure of PelC has been determined to fold into a novel topology, not observed before 1993 or predicted during the first 30 years of crystallography (10, 11). PelC folds into a large right-handed coil termed the parallel β helix. There are eight coils in the helix, and each is comprised of three β strands, connected by three turns with unique features. When the coils are stacked, the structure has the appearance of three parallel β sheets, stabilized by an extensive network of interstrand hydrogen bonds. Another notable feature of the PelC fold is the internal organization of the amino acids that form the core of the parallel β helix. All of the amino acids, which are oriented toward the interior, are regularly aligned with amino acids from neighboring coils, giving rise to long ladders of hydrophobic, aromatic, or polar amino acids. In contrast, the exterior amino acids are randomly oriented and comprise loops, of varying length and composition, which protrude from the central core. The initial postulate that the protruding loops form the substrate and active site has proven to be correct, as will be discussed in a later section. From the perspective of an effective plant virulence factor, the most important feature of the parallel β helix fold is the stability that it confers upon an enzyme that must function in the hostile extracellular environment. Most enzymes that degrade the plant cell wall are members of the glycoside hydrolase superfamily. Of the 62 families, organized according to sequence similarities, three-dimensional structures for members of 28 families have been reviewed (12). The enzymes that degrade the cellulose network of the plant cell wall are represented by a diversity of protein folds. The predominant structural fold, represented by endoglucanases and xylanases, is the (β/α)8 barrel. In this topology, eight parallel β strands, arranged in a circular, barrel-like motif, are connected by α helices that cover the exterior of the β barrel. The polysaccharide substrate site is located at the carboxy terminal base of the barrel, with the catalytic residues often found on β strands 4 and 7 (13). The structures of other cellulases fold into antiparallel β motifs as well as into domains comprised entirely of α helices. For example, cellobiohydrolase folds into a distorted antiparallel β barrel, sometimes called a β sandwich (14). Loops of different lengths and conformations fold over one face of the β sandwich and form the substrate-binding tunnel. Another example is endoglucanase CelD from Clostridium thermocellum, a cellulase that is structured into a motif best described as an (α/α)6 barrel (15). In contrast to the cellulase and hemicellulase families, the enzymes that degrade the pectate network all share the same parallel β helix topology initially found in PelC (Fig. 1). These include two additional pectate lyases, E. chrysanthemi PelE (16), Bacillus subtilis Pel (17); two pectin lyases from Aspergillus niger, PLA (18) and PLB (19); one rhamnogalacturonase, Aspergillus aculeatus RGase A (20); and two polygalacturonases, Erwinia carotovora polygalacturonase (21) and A. niger endopolygalacturonase II (22). Although the mechanism of pectic cleavage differs for the hydrolases and the lyases, the substrate binding sites are all found in a similar location within a cleft formed on the exterior of the parallel β helix. Moreover, parallel β helix folds, with analogous substrate-binding clefts, have been found in enzymes that degrade oligosaccharides found on cell surface receptors, such as P22 tailspike endorhamnosidase (23), and in the intercellular matrix, such as chondroitinase B (24).

STRUCTURAL APPROACHES TO THE ELUCIDATION OF THE ENZYMATIC MECHANISM OF PELC Before 1990, publication of a plethora of biochemical data usually preceded the publication of a three-dimensional structure. The two types of information are mutually beneficial. A structure explains the biochemical data and the biochemical data helps to elucidate the active site of an enzyme or the functionally relevant parts of a protein. However, in the last decade, threedimensional structures have been determined at such a rapid pace, as a consequence of dramatically improved x-ray diffraction and NMR techniques, that it is now more common to solve a structure, in the absence of any supporting biochemical data. Thus, new strategies are being developed to extract functional information from static threedimensional images. PelC has been among the first group of protein structures solved, without the benefit of relevant biochemical data to interpret the functional aspects of the three-dimensional structure. When the PelC structure was reported, details of its enzymatic mechanism could be summarized in the following few sentences. PelC is composed of 353 amino acids, with a molecular mass of 37,676 daltons and two disulfide bonds (9). As shown in Fig. 2, PelC catalyzes the endo- and exolytic cleavage of the α-1-4 glycosidic bond of polygalacturonic acid, generating an unsaturated trimer end product (25). Ca2+ is essential for the in vitro reaction and is likely required in the in vivo reaction in the plant cell wall in which the Ca2+ concentration is estimated to be as high as 1.0 mM. PelC lyase activity can be detected from a pH of 6.2 to a pH of 11.2, with an optimal pH at 9.5 (26). No amino acids, which participate in catalysis or substrate binding, had been identified by classical techniques. Consequently, neither the active site nor the substrate-binding pocket could be deduced with certainty from the initial PelC structure. The native PelC structure does provide two clues about the location of the region involved in catalysis (10). First, a Ca2+-like, heavy atom derivative binds in



8764

a site that has coordination compatible with a Ca2+ binding site. This has been the first indication that the essential Ca2+ might bind directly to the protein, rather than to the polygalacturonic acid substrate, as previously believed. The hypothesized Ca2+ binding site was ultimately shown to be analogous to the Ca2+ site captured directly in the B. subtilis Pel structure (17). The second revelation is that all surface charges are localized to an elongated groove surrounding the Ca2+ site. The shape, length, and charge of the groove are complementary to a negatively charged, oligogalacturonic moiety, suggesting the region was the substrate-binding pocket. The next insight into the enzymatic mechanism has occurred with the report of the PelE structure (16). PelE is one of the E. chrysanthemi isozymes but belongs to a different subfamily of pectate lyases. The PelE subfamily is characterized by a single disulfide bond and cleaves the polygalacturonic acid substrate to a dimer. PelE shares 22% sequence identity with PelC. Not surprisingly, PelE folds into the same parallel β helix topology and has a similar Ca2+ binding pocket. However, in contrast to PelC, the charged amino acids on PelE are randomly distributed on the surface and are not localized into an elongated groove. This finding suggests that either the active site region has been deduced incorrectly from the PelC structure or that the optimal in vivo substrate for PelE is not the same as for PelC. To extract additional clues about the location of the catalytic site, the structures of PelC and PelE have been superimposed, and the similarities as well as the differences have been analyzed. In general, key amino acids that are essential to a structural or functional property of an enzyme are invariant and the region around an active site is structurally conserved. The superposition of the PelC and PelE structures has permitted the correction of the evolutionary-based amino acid sequence alignment for the two proteins as well as for the entire superfamily (27). The superfamily includes 14 extracellular pectate lyases, 7 pectin lyases, and 12 plant pollen and style proteins. The corrected sequence alignment has identified 10 invariant amino acids, 5 of which are amino acids that could potentially be involved in catalytic activity. In PelC nomenclature, these five potentially catalytic amino acids include Asp-131, Asp-144, His-145, Thr206, and Arg-218. The five chemically inert amino acids include Gly-6, Gly-12, Gly-13, Trp-142, and Pro-220. With most enzymes, the invariant amino acids cluster around the same general region, that of the active site. However, the situation is different with PelC. The 10 invariant amino acids cluster in two distinctly different regions (Fig. 3). Asp-131, Arg-218, and Pro-220 are grouped around the Ca2+ binding site that has been postulated to be part of the active site. The other seven amino acids cluster on the opposite side of the parallel β helix and are too distant to be part of a catalytic site near the Ca2+ ion. Instead of identifying the active site, the cluster analysis has posed new questions. If PelC has two active sites, which one is responsible for pectolytic activity and what is the function of the second active site? Site-directed mutagenesis has been used to conclusively identify the pectolytic active sites (28). Both invariant and highly conserved amino acids have been targeted for mutations. Moreover, several mutations have been made at each position. In all, 32 PelC mutants have been prepared, purified, and characterized. Those amino acids clustering around the Ca+2 site have been shown to simultaneously affect maceration of plant tissue and pectolytic activity. Moreover, all mutations in two amino acids, Asp-131 and Arg-218, reduce the pectolytic activity to less than 0.5% of the wild type, suggesting that these amino acids have some type of catalytic role. Mutations in the invariant amino acids in the second cluster, including Trp-142, Asp-144, His-145, and Thr-206, have also been prepared, but most of the mutants cannot be purified in high yields. From preliminary experiments, it appears that most of these mutant proteins remain bound to the membrane fraction and are in an unfolded state, as demonstrated by their sensitivity to trypsin. Only very



8765

Fig. 4. Stereoview of the active site of the PelC R218K-(Ca2+)4-pentaGalpA complex. The color code is the same as that used in Fig. 3. small quantities of W142H, D144N, and T206A could be purified and characterized. The latter mutants retain pectolytic activity, demonstrating that the second cluster of invariant amino acids is not involved in the enzymatic cleavage of a polysaccharide. The function of the second cluster remains a mystery.

STRUCTURE OF THE PELC R218K-(CA2+)4-PENTAGALPA COMPLEX With the active site of PelC now identified, attention has turned toward the elucidation of the atomic details of the protein– saccharide interactions in the substrate binding site and the details of the enzymatic mechanism. In an ideal world, a crystallographer prefers to study the structure of a complex between a native protein and a transition state analogue or an inhibitor, to work out atomic details of the active site of the enzyme. Unfortunately, no efficient inhibitors of the pectate lyases are known. Therefore, an alternate approach, that of forming complexes between catalytically impaired mutants of PelC and an oligosaccharide fragment, has been used. X-ray diffraction data for several mutant PelColigosaccharide complexes have been collected. However, only one data set, using the R218K PelC mutant and pentagalacturonic acid (pentaGalpA), has yielded a view of an ordered oligosaccharide composed of four of the five GalpA units (29). The PelC mutantoligosaccharide structure held many new surprises (Fig. 4). First, instead of finding only one Ca2+ ion, as in the native PelC structure, a total of four Ca2+ sites have been found. All Ca2+ ions are bridging acidic groups on the protein to uronic acid moieties on the oligosaccharide. One Ca2+ ion also forms a link between two adjacent uronic acid moieties. The Ca2+-oligosaccharide interactions are substantially different from the popular “eggbox” model in which Ca2+ ions cross-link the uronic moieties of adjacent chains of pectate together in the plant cell wall (1). The second surprising feature is that the conformation of the tetraGalpA fragment is different from the pectate conformation previously predicted (30, 31). Instead of having a 21 or 31 helical conformation, the pectate fragment is a mixture of both 21 and 31 helices. Third, the complex represents an enzymesubstrate interaction and thus provides a detailed view of a Michaelis intermediate in the reaction pathway. The scissile bond has been assigned to the glycosidic bond between the third and fourth GalpA units from the reducing end of the ordered tetrasaccharide by coupling the structural results with enzymatic cleavage patterns of oligogalacturonates of defined lengths. In the substrate cleft, all invariant and conserved amino acids in the pectate lyase family could be assigned either a role in catalysis or a role in specific recognition of the GalpA units. The GalpA units on either side of the scissile bond form the most interactions with the protein. A conserved Arg-223 recognizes the galactose epimer and a Ca2+ network surrounds each uronic acid group. The invariant aspartic acid, Asp-131 in PelC, is instrumental in coordinating strongly to the primary Ca2+, which in turn forms a linked network with other Ca2+ ions and uronic acid moieties. The most surprising finding is that Arg-218, invariant in the pectate lyase superfamily, is the amino acid that initiates proton abstraction from C-5 in the GalpA unit adjacent to the scissile bond. Such a role for an arginine has not been observed previously but is consistent with the high pH optimum of the in vitro pectolytic reaction. That arginine can serve as a proton abstractor is made possible by the local electronic environment surrounding C-5 and the adjacent uronic acid moiety. As predicted by Gerlt and Gassman (32), positive charges, in the form of a Ca2+ ion and a conserved lysine, Lys-190, neutralize the uronic acid group and stabilize an enolic intermediate. Together, these positive charges lower the pKa of the C-5 proton to levels that approximate the pKa of an arginine. Overall, the structure of the R218K-(Ca2+)4-pentaGalpA complex provides many details about the enzymatic reaction as well as the nature of the protein–saccharide interaction. Yet there remain unanswered questions that can best be addressed with structural techniques. One major puzzle is the identification of the group responsible for transferring a proton to the glycosidic bond, simultaneous with bond cleavage. The answers must await structures of new complexes that mimic alternate stages of the reaction.

SIGNIFICANCE OF MULTIPLE ISOZYMES FOR PATHOGENESIS One enigma, yet to be resolved in the pectate lyase field, is the function of multiple, independently regulated pectate lyase isozymes and the role, if any, in pathogenesis. The composite results of the structural studies suggest that the answer may lie in the heterogeneous nature of the substrate. The pectate



8766

component is composed of repeating units of negatively charged galacturonic acid. The uronic acid group on C-6 is frequently esterified with a methyl group. Methylation neutralizes the negative charge of an individual GalpA unit and, consequently, alters the surface charge of the pectate polymer. The percentage and the positional sequence of methylated GalpA units (mGalpA) vary during the life cycle of the plant and from one type of plant to another. Thus, a pathogen that secretes a battery of pectic enzymes, each of which uses a similar catalytic mechanism, but recognizes a different sequence of methylated and nonmethylated oligogalacturonate units, would be expected to have a broader host range. The structural studies provide tentative evidence that the pectate lyase isozymes cleave the same in vitro substrate but differ in the preferred in vivo substrate. In the majority of enzyme families, the most conserved region, in terms of primary and tertiary structure, is the active site and the surrounding region. The pectate lyase family appears to be somewhat different. Comparisons of the pectate lyase structures have revealed that, although the immediate region around the catalytic site is highly conserved, the substrate groove extending on either side is structurally the most diverse (16). The diversity arises from the variation in the sequence and length of the loops protruding from the parallel β barrel. One set of loops, termed the T3 loops, form the binding pockets required for recognition of multiple saccharide units in the substrate. The surface charges in PelC are localized to an elongated groove that complements the shape and length of an oligosaccharide composed of repeating GalpA units (10). In contrast, the analogous region in PelE is highly charged only around the primary Ca2+ binding site, with a predominance of neutral groups lining the remaining portion of the groove (16). Analogous differences are found in three-dimensional models constructed from the sequences of other pectate lyase isozymes. Thus, polygalacturonic acid may serve as a suitable in vitro substrate for all isozymes because the saccharide binding pockets near the active site recognize only galacturonic acid moieties. The structural differences between PelC, PelE, and other isozymes at more distant subsites suggest that the optimal in vivo substrate may differ in length, charge, and possibly type of saccharide unit from one isozyme to another. Differences among isozyme substrates need not be as great as a different type of saccharide unit. Smaller differences, such as the degree of methylation of the pectate component, may be sufficient to explain the physiological significance of multiple pectate lyase isozymes.

MODEL OF THE PLB-MGALPA4 COMPLEX Given the difficulty of obtaining defined lengths and specific methylated sequences, it will be difficult to ascertain the optimal substrate of each pectic isozyme through structural studies of additional complexes. Instead, some information may be gleaned from modeling saccharide units in substrate subsites. The first step is to elucidate the structural determinants that are necessary to distinguish between a methylated and a nonmethylated uronic acid moiety on GalpA units. This goal can be achieved by comparing the substrate subsites in PelC with those in a pectin lyase. Pectin lyases share a similar enzymatic mechanism with the pectate lyases but recognize a different substrate, that of pectin, the fully methylated, neutral form of pectate. Given the sequence and functional similarities, it is not surprising that the two known pectin lyase structures, PLA (18) and PLB (19), are homologous to the pectate lyases. However, no structural complexes containing pectin are currently available. To answer the question about the atomic determinants of methylation, the PLB structure has been superimposed upon the R218K-(Ca2+)4pentaGalpA structure, minimizing the atomic distances between all conserved and invariant amino acids shared by the two structures. The superposition has allowed for a feasible placement of an oligosaccharide substrate in the PLB substrate. The Ca2+ ions from the R218L model have been removed, and methyl groups have been added to the uronate moiety. The contacts between PLB and the modeled pectin fragment have been visually and energetically optimized to avoid unfavorable contacts. The resulting model is shown in Fig. 5 and Fig. 6; a schematic representation of the modeled PLB-pectin contacts is shown in Fig. 7. Three striking features emerge from the model. First, there is no ambiguity in positioning the methylated uronate group. For three of the four mGalpA units, a hydrophobic pocket, large enough to accommodate a methyl group, exists around one oxygen of the uronate moiety and a polar environment surrounds the other oxygen. The hydrophobic pockets surrounding the methyl groups are composed primarily of tryptophan and tyrosine residues, the unusual cluster of which was noted in the original report of the PLB structure. The second notable feature of the model is the conservation of a positive charge between the uronate moieties of the third and four saccharide units. In pectin lyase, the positive charge is conferred by Arg-176, a conserved arginine in the pectin lyase subfamily. In pectate lyases, the positive charge is conferred by a Ca2+ ion, which is essential for catalytic activity within the pectate lyase subfamily. The high degree of conservation of the positive charge at analogous positions in the structures suggests that the positive charge is likely to play a significant role during catalysis, possibly stabilizing a negative intermediate or participating in the transfer of a proton to the glycosidic bond. The third notable feature is confirmation of the roles of the three amino acids that are invariant in the pectate lyase superfamily: Asp-154, Arg-236, and Pro-238 in PLB nomenclature. As in the pectate lyases, Pro-238 is found in the unusual cis-conformation, which helps to orient Arg-236 toward the α proton on C-5 of the saccharide unit adjacent to the scissile bond. Although it is unusual, the involvement of Arg-236 in a catalytic step has been confirmed by site-specific mutational data presented in Table 1. Both R236Q and R236K PLB mutants are catalytically inactive. The invariant aspartic acid, Asp-154, does not have a catalytic role but, rather, as suggested from the structures, appears to stabilize the position of a positive charge near the scissile bond. The lack of a catalytic role for the invariant aspartic acid has also been confirmed by

Fig. 5. Overview of the PLB-(mGalpA)4 model. Pectin lyase B is shown as yellow ribbons. The substrate and the interacting amino acids are represented by rods using the color code described in Fig. 3.



8767

mutational data (Table 1). Both PLB mutants, D154E and D154N, retain significant catalytic activity. The role of Arg-176 remains to be confirmed through mutational experiments.

Fig. 6. Stereoview of the active site of the PLB-(mGalpA)4 model. The view of the active site is the same as that used in Fig. 4 to allow for comparisons. The hydrophobic pockets around three of the four methyl groups esterified to the uronate moieties are visible. The color code is the same as that used in Fig. 6. CONCLUSIONS The story of PelC, although incomplete, illustrates how structural information can contribute to an understanding of the multifaceted steps of pathogenesis. In the process, a novel protein structural fold and a unique catalytic role for an arginine have been discovered. The structural results have also provided the first atomic description of a pectate fragment, finding the conformation as well as the mode of interactions with Ca2+ ions to be considerably different from the popular view. The growing structural databases of pectolytic enzymes should allow researchers to elucidate the structural motifs that recognize specific types of saccharides. The initial attempt to elucidate the structural principles that permit an enzyme to distinguish between pectate and pectin is the first step in understanding how saccharidases recognize a unique oligosaccharide sequence from a heterogeneous mix in the plant cell wall. Such knowledge should ultimately lead to a better understanding of the unique properties of the host that render it more vulnerable to attack by a given pathogen.

Fig. 7. Schematic representation of the pectin lyase B with (GalpA)4 at a distance of ≤3.0 Å. mGalpA (1) is the reducing saccharide and mGalpA (4) is the nonreducing terminus. Hydrogen bonds are represented with dotted lines and hydrophobic interactions, with boldface lines. Oxygen atoms and methyl groups are represented by circles, with the corresponding number, and the carbon atoms are assumed at the intersection of bonds designated in boldface lines. METHODS PLB-(mGalpA)4 Modeling. Amino acids in PLB, corresponding to the invariant amino acids in the pectate lyase superfamily, were superimposed upon their counterparts in the structure of the PelC R218K-(Ca2+)4-pentaGalpA complex. The superposition was improved by LSQMAN (33). The Ca2+ ions were removed from the R218K structure, and methyl groups were added to one of the uronate oxygens. AUTODOCK was used to maximize the favorable interactions and to allow carbohydrate flexibility (34). The pectin substrate was first introduced into the PLB active cleft as two dimers to improve the fit. Then, the two fragments were rejoined and subjected to several rounds of energy minimization. All model figures were prepared with SETOR (35).



Table 1. Specific activities and kinetic parameters of wild-type and mutated PLB Specific activity, units/mg Km, mg/ml Enzyme PLB wild type 698 6.3 PLB D154E 306 7.8 PLB D154N 91 14.8 PLB R236Q 0 – PLB R236K 1.4 –

Vmax, units/mg 1,429 1,143 512 – –

8768

pH optimum 8.5 8.5 9.5 8.5

DNA Manipulations. Cloning experiments for the preparation of PLB mutants were performed in E. coli DH5α (36) by using standard protocols (37). Restriction enzymes were used as described by the supplier (GIBCO/BRL). Nucleotide sequences were determined by using a Cy5 AutoCycle Sequencing Kit (Amersham Pharmacia) with universal and reverse primers or gene specific primers. The reactions were analyzed with an ALFred DNA sequencer (Amersham Pharmacia). Computer analysis was done by using the program GENERUNNER (Hastings Software). The cloning of A. niger N400 pelB PLB was described previously (38). A translational promoter gene fusion was constructed by using the pyruvate kinase promoter (39). The pki-pelB fusion (pPK-PLB) was described previously (38). Site-Directed Mutagenesis. Site-directed mutagenesis of PLB was carried out by using the Altered Sites II kit (Promega) and synthetic oligonucleotides (Isogen, Maarsen, The Netherlands). The procedure was performed as described (40). The pki-pelB promoter gene fusion was excised with pPK-PLB by using restriction endonucleases BamHI and HindIII and was ligated into BamHI- and HindIII-digested pALTER I, resulting in plasmid pIM3550. Plasmid DNA was isolated and sequenced to confirm the desired mutations and to check the gene for undesired mutations. Those plasmids showing the correct sequence and the expected mutation were used to transform A. niger strain NW1888 (cspAI, pyrA6, leu-13, prtF28), a derivative of A. niger N400 (CBS. 120.49). Transformations were carried out as described (41). Culture Conditions and Enzyme Purifications. Mutant PLB producing transformants were selected by growing individual transformants in minimal medium as described (42). Large scale cultivation of transformants producing mutant PLB was performed as outlined (43) in multiple 300 milliliter batches in one liter Erlenmeyer flasks incubated in an orbital shaker (250 revolutions per minute) at 30°C. Wild-type and mutant PLB were purified essentially as described by Kester and Visser (43) with the following modification. After dialysis of the solubilized ammonium sulfate precipitate, the dialysate was loaded onto a Source Q anion exchanger (15.5-milliliter bed volume) (Pharmacia) preequilibrated with 10 mM TrisHCl (pH 7.5). Proteins were eluted with a 0–200 mM sodium chloride linear gradient in 10 mM TrisHCl (pH 7.5). PLB-containing fractions were pooled and exhaustively dialyzed against 20 mM sodium phosphate (pH 6.0). Enzyme solutions were stored at − 20°C. The purity of the mutant enzyme was confirmed by SDS/PAGE and Coomassie brilliant blue staining. The concentration of purified mutant enzymes was determined by measuring the absorbance at a wavelength of 280 nm, using a molar absorption coefficient (ε) of 50,220 M−1 cm−1 for PLB, as calculated from the tryptophan, tyrosine, and cysteine content (44). Enzyme Assay and Determination of Kinetic Parameters. Standard PLB assays were carried out in 50 mM TrisHCl and 0.06 M sodium chloride (pH 8.5), containing 3 mg/ml (wt/vol) lime pectin with 75% methyl esterification (Copenhagen Pectin Factory, Lille Skensved, Denmark) in a total volume of 1.0 ml. The assay buffer was equilibrated at 30°C, and the reaction was started by the addition of 20 µl of enzyme solution. The activity was determined by measuring the increase in absorbance at 235 nm (ε = 5200 M−1cm−1). The kinetic traces were corrected for spontaneous chemical β-elimination. Km and Vmax values were determined from triplicate initial rate measurements in the same way as described for the standard assays, with the exception that the pectin concentrations varied from 0.5 to 7.0 mg/ml. The research was supported by the U.S. Department of Agriculture (Grant 98-35304). The research was conducted in part at the Stanford Synchrotron Radiation laboratory, which is operated by the Office of Basic Energy Science of the U.S. Department of Energy. 1. Carpita, N. C. & Gibeaut, D. M. (1993) Plant J. 3, 1–30. 2. Albersheim, P., Darvill, A., O'Neill, M., Schols, H. A. & Voragen, A. G. J. (1996) in Progress in Biotechnology: Pectins and Pectinases, eds. Visser, J. & Voragen, A. G. J. (Elsevier, Amsterdam), Vol. 14, pp. 47–53. 3. Davies, G. & Henrissat, B. (1995) Structure (London) 3, 853–859. 4. Kiss, J. (1974) Adv. Carbohydr. Chem. Biochem. 29, 229–230. 5. He, S. Y., Lindeberg, M. & Collmer, A. (1993) in Biotechnology in Plant Disease Control ed. Chet, I. (Wiley–Liss, New York), pp. 39–64. 6. Reverchon, S., Nasser, W. & Robert-Baudouy, J. (1991) Mol. Microbiol. 5, 2203–2216. 7. Tamaki, S. J., Gold, S. Robeson, M., Manulis, S. & Keen, N. T. (1988) J. Bacteriol. 170, 3468–3478. 8. Barras, F., Van Gijsegem, F. & Chatterjee, A. K. (1994) Annu. Rev. Phytopathol. 32, 201–234. 9. Keen, N. T., Dahlbeck, D., Staskawicz, B. & Belser, W. (1984) J. Bacteriol. 159, 825–831. 10. Yoder, M. D., Keen, N. T. & Jurnak, F. (1993) Science 260, 1503–1507. 11. Yoder, M. D., Lietzke, S. E. & Jurnak, F. (1993) Structure (London) 1, 241–251. 12. Davies, G. & Henrissat, B. (1995) Structure (London) 3, 853–859. 13. Jenkins, J., Leggio, L. L., Harris, G. & Pickersgill, R. (1995) FEBS Lett. 362, 281–285. 14. Divne, C., Ståhlberg, J., Teeri, T. T. & Jones, T. A. (1998) J. Mol. Biol. 275, 309–325. 15. Juy, M., Amit, A., Alzari, P., Poljak, R. J., Claeyssens, M., Bguin, P. & Aubert, J.-P. (1992) Nature (London) 357, 89–91. 16. Lietzke, S. E., Keen, N. T., Yoder, M. D. & Jurnak, F. (1994) Plant Physiol. 106, 849–862. 17. Pickersgill, R., Jenkins, J., Harris, G., Nasser, W. & Robert-Baudouy, J. (1994) Nat. Struct. Biol. 1, 717–723. 18. Mayans, O., Scott, M., Connerton, I., Gravesen, T., Benen, J., Visser, J., Pickersgill, R. & Jenkins, J. (1997) Structure (London) 5, 677–689. 19. Vitali, J., Schick, B., Kester, H. C. M., Visser, J. & Jurnak, F. (1998) Plant Physiol. 116, 69–80. 20. Petersen, T. N., Kauppinen, S. & Larsen, S. (1997) Structure (London) 5, 533–544. 21. Pickersgill, R., Smith, D., Worboys, K. & Jenkins, J. (1998) J. Biol. Chem. 273, 24660–24664. 22. van Santen, Y., Benen, J. A. E., Schröter, K. H., Kalk, K. H., Armand, S., Visser, J. & Dijkstra, B. W. (1999) J. Biol. Chem. 274, 30474–30480. 23. Steinbacher, S., Seckler, R., Miller, S., Steipe, B., Huber, R. & Reinemer, P. (1994) Science 265, 383–386. 24. Huang, W., Matte, A., Li, Y., Kim, Y. S., Linhardt, R. J., Su, H. & Cygler, M. (1999) J. Mol. Biol. 294, 1257–1269. 25. Preston, J. F., Rice, J. D., Ingram, L. O. & Keen, N. T. (1992) J. Bacteriol. 174, 2039–2042. 26. Tardy, F., Nasser, W., Robert-Baudouy, J. & Hugouviewx-Cotte-Pattat, N. (1997) J. Bacteriol. 179, 2503–2511. 27. Henrissat, B., Heffron, S. E., Yoder, M. D., Lietzke, S. E. & Jurnak, F. (1995) Plant Physiol. 107, 963–976. 28. Kita, N., Boyd, C. M., Garrett, M. R., Jurnak, F. & Keen, N. T. (1996) J. Biol Chem. 271, 26529–26535. 29. Scavetta, R. D., Herron, S. R., Hotchkiss, A. T., Kita, N., Keen, N. T., Benen, J. A., E., Kester, H. C. M., Visser, J. & Jurnak, F. (1999) Plant Cell 11, 1081– 1092.



8769

30. Morris, E. R., Powell, D. A., Gidley, M. J. & Rees, D. A. (1982) J. Mol. Biol. 155, 507–533. 31. Walkinshaw, M. D. & Arnott, S. (1981) J. Mol. Biol. 153, 1055–1085. 32. Gerlt, J. A. & Gassman, P. G. (1992) J. Am. Chem. Soc. 114, 5928–5934. 33. Kleyegt, G. J. & Jones, T. A. (1995) Structure (London) 3, 535–540. 34. Morris, G. M., Goodsell, D. S., Huey, R. & Olson, A. J. (1996) J. Comput. Aided Mol. Des. 10, 293–304. 35. Evans, S. V. (1993) J. Mol. Graphics 11, 134–138. 36. Woodcock, D. M., Crowther, P. J., Doherty, J., Jefferson, S., DeCruz, E., Noyer-Weidner, M., Smith, S. S., Michael, M. Z. & Graham, M. W. (1989) Nucleic Acids Res. 17, 3469–3478. 37. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989) in Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Lab. Press, Plainview, NY), pp. 1–125. 38. Kusters-van Someren, M., Flipphi, M., de Graaff, L. H., van den Broeck, H., Kester, H. C. M., Hinnen, A. & Visser, J. (1992) Mol. Gen. Genet. 234, 113– 120. 39. de Graaff, L. H., van den Broeck, H. & Visser, J. (1992) Curr. Genet. 22, 21–27. 40. Armand, S., Wagemaker, M. J., Sánchez-Torres, P., Kester, H. C. M., van Santen, Y., Dijkstra, B. W., Visser, J. & Benen, J. A. E. (2000) J. Biol. Chem. 275, 691–696. 41. Kusters-van Someren, M. A., Harmsen, J. A. M., Kester, H. C. M. & Visser, J. (1991) Curr. Genet. 20, 293–299. 42. Parenicová, L., Benen, J. A. E., Kester, H. C. M. & Visser, J. (1998) Eur. J. Biochem. 251, 72–80. 43. Kester, H. C. M. & Visser, J. (1994) FEMS Microbiol. 120, 63–68. 44. Edelhoch, H. (1967) Biochemistry 6, 1948–1954.


PSEUDOMONAS SYRINGAE HRP TYPE III SECRETION SYSTEM AND EFFECTOR PROTEINS

8770

Colloquium Pseudomonas syringae Hrp type III secretion system and effector proteins Alan Collmer*†, Jorge L Badel*, Amy O. Charkowski‡, Wen-Ling Deng*, Derrick E. Fouts*, Adela R. Ramos*, Amos H. Rehm*, Deborah M. Anderson§, Olaf Schneewind§, Karin van Dijk¶, and James R. Alfano¶ * Department of Plant Pathology, Cornell University, Ithaca, NY 14853-4203; ‡ United States Department of Agriculture, Agricultural Research Service, Western Regional Research Center, 800 Buchanan Street, Albany, CA 94710; § Department of Microbiology and Immunology, University of California School of Medicine, Los Angeles, CA 90095; and ¶ Department of Biological Sciences, University of Nevada, Las Vegas, NV 89154-4004 Pseudomonas syringae is a member of an important group of Gram-negative bacterial pathogens of plants and animals that depend on a type III secretion system to inject virulence effector proteins into host cells. In P. syringae, hrp/hrc genes encode the Hrp (type III secretion) system, and avirulence (avr) and Hrpdependent outer protein (hop) genes encode effector proteins. The hrp/hrc genes of P. syringae pv syringae 61, P. syringae pv syringae B728a, and P. syringae pv tomato DC3000 are flanked by an exchangeable effector locus and a conserved effector locus in a tripartite mosaic Hrp pathogenicity island (Pai) that is linked to a tRNALeu gene found also in Pseudomonas aeruginosa but without linkage to Hrp system genes. Cosmid pHIR11 carries a portion of the strain 61 Hrp pathogenicity island that is sufficient to direct Escherichia coli and Pseudomonas fluorescens to inject HopPsyA into tobacco cells, thereby eliciting a hypersensitive response normally triggered only by plant pathogens. Large deletions in strain DC3000 revealed that the conserved effector locus is essential for pathogenicity but the exchangeable effector locus has only a minor role in growth in tomato. P. syringae secretes HopPsyA and AvrPto in culture in a Hrp-dependent manner at pH and temperature conditions associated with pathogenesis. AvrPto is also secreted by Yersinia enterocolitica. The secretion of AvrPto depends on the first 15 codons, which are also sufficient to direct the secretion of an Npt reporter from Y. enterocolitica, indicating that a universal targeting signal is recognized by the type III secretion systems of both plant and animal pathogens. T ype III protein secretion systems underlie the pathogenicity of many Gram-negative bacteria, including important animal pathogens in the genera Yersinia, Salmonella, Shigella, and Escherichia, and plant pathogens in the genera Pseudomonas, Erwinia, Xanthomonas, and Ralstonia (1, 2). The plant pathogens cause diverse diseases in hosts that range from apple trees to the model weed Arabidopsis thaliana, and they all share an ability to colonize the intercellular spaces of plant tissues and to cause death (sometimes delayed) in plant cells. A fundamental difference between this group of plant pathogens and most of the animal pathogens is that the former do not enter living host cells, but rather interact with the host cytoplasm from outside of an approximately 200-nm-thick plant cell wall. The ability to deliver effector proteins across this barrier via the type III secretion system is likely to be unique to plant pathogens, and it is key to their pathogenicity. Pseudomonas syringae is a widespread and representative plant pathogen. It is host specific and elicits leaf spots and other foliar necroses in host plants and the hypersensitive response (HR) in nonhosts (3). In host plants, disease symptoms typically develop after several days of bacterial growth in leaf intercellular spaces. In nonhosts, the defense-associated programmed cell death that characterizes the HR occurs within 24 h in plant cells that are in contact with the bacterium (4). Underlying both types of P. syringae interactions with plants are hrp (HR and pathogenicity) and hrc (HR and conserved) genes that encode the type III secretion system and avirulence (avr) and Hrp-dependent outer protein (hop) genes that encode effector proteins injected into plant cells by the system (three-letter suffixes often indicate the strain of origin for the effector) (5). Avr proteins are so named because in some potential hosts they betray the parasite to the R (resistance) gene surveillance system of plants, thereby triggering the HR (6). P. syringae is divided into more than 40 pathovars based on pathogenic specificity for various plant species, and some pathovars are further divided into races on the basis of host range among differential cultivars of the host species. Although the basis for host range at the pathovar– plant species level has not been established, host range at the race-plant cultivar level is determined by combinations of Avr-R genes interacting in a gene-for-gene manner (6). That is, if the interactants contain corresponding Avr and R genes, then HR-associated defenses will be triggered. The R gene-encoded surveillance system, which appears to be arrayed primarily against the antihost proteins of parasites, is a key determinant of defense against highly adapted “stealth” pathogens like P. syringae. Similar gene-for-gene pathosystems involving multiple races and cultivars occur with many pathogenic fungi, nematodes, and viruses. The HR is similarly triggered in “incompatible” interactions with many of these parasites, and it is noteworthy that the HR is typically triggered in plants only by potential pathogens, not by encounters with the far more numerous nonpathogenic microbes in the environment. Our research has focused on three strains of P. syringae: (i) P. syringae pv syringae (Psy) 61 is a weak pathogen of bean and is the source of the hrp/hrc gene cluster cloned on cosmid pHIR11 that contains all of the genes necessary for nonpathogenic bacteria like Pseudomonas fluorescens and Escherichia coli to elicit the HR in tobacco (7). (ii) Psy B728a is closely related to strain 61 but is a highly virulent model for studying epiphytic fitness and pathogenicity (brown spot of bean) in the field (8, 9). (iii) P. syringae pv. tomato (Pto) DC3000 is a well-studied pathogen of tomato and Arabidopsis (bacterial speck) that is taxonomically quite divergent from pathovar syringae (10), and it produces AvrPto, one of the best-studied Avr proteins (11). Thus, we can compare two closely related strains and one highly

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviations: Psy, Pseudomonas syringae pv syringae; Pto, P.s. tomato; HR, hypersensitive response; hrp, HR and pathogenicity; hrc, HR and conserved; Pai, pathogenicity island; EEL, exchangeable effector locus; CEL, conserved effector locus; Hop, Hrp-dependent outer protein; Avr, avirulence. †

To whom reprint requests should be addressed. E-mail: [email protected].



8771

Fig. 1. The cluster of Psy 61 genes carried on pHIR11 that enables nonphytopathogenic bacteria to elicit the HR in tobacco. hopPsyA (checkered) encodes an effector protein that apparently is delivered into plant cells. Other genes encode regulatory factors (shaded), Hrc components associated with export across the inner membrane (diagonal hatching) or outer membrane (cross hatching), extracellular Hrp proteins (stippled), or proteins with unknown function (open boxes). Squares on arrows denote the presence of HrpL-activated promoters (55). divergent strain in our investigation of the evolution and function of Hrp systems. In the last decade, research on the evolution and function of type III secretion systems in Salmonella and Yersinia spp. has yielded two revolutionary insights. First, genes associated with pathogenicity, such as those encoding type III secretion systems, are often clustered in horizontally acquired pathogenicity islands (Pais) that may enable the evolution of virulence in “quantum leaps” (12, 13). Second, type III secretion systems have the remarkable ability to inject bacterial proteins into the cytoplasm of eukaryote host cells (14, 15). In this article, we will describe our progress in understanding how the P. syringae Hrp system expressed from pHIR11 enables a nonpathogen like E. coli to make a quantum leap in its ability to interact with plants by eliciting the HR, how hrp/hrc genes are arranged in Hrp Pais that also encode a variety of putative effectors, and how universal targeting signals and genetically dissectable secretion mechanisms underlie effector protein traffic through the pathway.

HOPPSYA, PHIR11, AND THE MINIMUM GENETIC UNIT FOR BACTERIAL ELICITATION OF THE HYPERSENSITIVE RESPONSE Cosmid pHIR11 was seminal in establishing the minimum genetic requirements and relative role of the Hrp system and effectors in HR elicitation. pHIR11 was cloned from Psy 61 on the basis of its ability to complement several hrp::Tn5 mutations in that strain (7). It also enables P. fluorescens, P. putida, and E. coli (and probably many other Gram-negative bacteria) to elicit the HR in tobacco. However, pHIR11 does not enable nonpathogens to multiply or cause disease in any plants tested. For example, P. fluorescens (pHIR11) does not cause any symptoms in tobacco leaves unless inoculated at a very high level (≥5 × 106 cell/ml), such that enough individual plant cells undergo the HR to produce a confluent collapse. The DNA sequence of pHIR11 reveals a 25-kb cluster of hrp/hrc genes linked to an apparent operon encoding hopPsyA (hrmA) and ORF1 (16, 17, 18, 19, 20, 21 and 22) (Fig. 1). The hrp/hrc clusters of Psy B728a and Pto DC3000 are arranged similarly (further discussed below), but HopPsyA is unique to Psy 61 (18, 23). Three proteins, the HrpZ and HrpW harpins and HrpA pilin, are secreted by the P. syringae Hrp pathway in culture more abundantly than other Hrp-dependent proteins (24, 25, 26 and 27). Harpins are glycine-rich cysteinelacking proteins that possess heat-stable HR elicitor activity when infiltrated at relatively high concentration into the intercellular leaf spaces of many plants (5, 28). However, in P. syringae their HR-elicitation activity does not correlate with bacterial host range, and these proteins appear to have an ancillary role in plant interactions (21). HrpA forms a Hrp-specific pilus that is 6–8 nm in diameter and is essential for all Hrp phenotypes (26). Through a series of observations, HopPsyA was identified as the HR-triggering effector that is injected into plant cells by the pHIR11 Hrp system, and it was simultaneously shown to have salient characteristics of known Avr proteins: (i) Mutations in hopPsyA abolish the ability of pHIR11 to direct HR elicitation without affecting HrpZ production or secretion, indicating that the essential role of HopPsyA is not as a component of the Hrp secretion system (29). (ii) HopPsyA travels the Hrp pathway, as demonstrated by its secretion in culture (discussed below) (30). (iii) HopPsyA has no apparent effect when delivered exog

Fig. 2. Summary of evidence that HopPsyA functions like an avirulence protein that interacts inside plant cells with the product of an R gene present in N. tabacum but not N. benthamiana. The upper squares, labeled “pHIR11,” indicate the responses in leaves of N. tabacum (N.t.) and N. benthamiana (N.b.) to P. fluorescens 55 carrying pHIR11 (+) or a hopPsyA::TnphoA derivative (−) after infiltration at a concentration of 5 × 107 cells/ml. “HR” indicates rapid confluent collapse of infiltrated tissue; “Null” indicates no visible response. The next two photographs, labeled “Agrobacterium,” show the effect in N.t. and N.b. of A. tumefaciens GV3101-mediated transient expression of hopPsyA via glucocorticoidinducible expression vector pTA7002 (85). Plants receiving pTA7002 (−) or pTA7002::hopPsyA (+) were sprayed with the glucocorticoid dexamethasone 48 h after infiltration and then photographed 24 h later. Note that the confluent tissue collapse indicative of the HR is observed only when hopPsyA is expressed in the N.t. leaf. The lower two photographs, labeled “P. s. tabaci,” show the effect in N.t. leaves of P. syringae pv tabaci 11528 carrying empty vector pDSK519 (−) or pCPP2349 (hopPsyA+) (+) at 1 and 5 days after inoculation (23). The level of inoculum was 5 × 108 in the lowest sector on each side of each leaf and 5 × 106 cell/ml in the next sectors up. Note that the HR developed by the end of day 1 in the sector infiltrated with 5 × 108 cells/ml of P. syringae pv tabaci (pCPP2349), whereas disease symptoms caused by P. syringae pv tabaci (pDSK519) developed later, with a lower level of inoculum, and were uniquely marked with the bright yellow chlorosis characteristic of wildfire. [Reproduced with permission from ref. 23 (Copyright 1997, Mol. Plant–Microbe Interact.).]



8772

enously to tobacco cells but is lethal if expressed inside them via either biolistic- or Agrobacterium tumefaciens-mediated transient expression (23) (Fig. 2). (iv) Unlike tobacco (Nicotiana tabacum L. cv. Xanthi), Nicotiana benthamiana does not respond with the HR to either P. fluorescens (pHIR11) or hopPsyA transiently expressed inside plant cells after delivery by A. tumefaciens (Fig. 2). (v) When transformed into P. syringae pv tabaci, hopPsyA causes the tobacco wildfire pathogen to become avirulent in tobacco, as would be expected if tobacco possessed an R gene directing recognition of HopPsyA (23) (Fig. 2). (vi) A hopPsyA mutation in Psy 61 does not abolish virulence in bean or HR elicitation in tobacco, which is typical of known avr genes apparently because of their redundant contribution to parasitic fitness in hosts and HR elicitation in nonhost species (31). (vii) P. fluorescens and E. coli strains carrying pHIR11 fail to elicit the HR in soybean, Arabidopsis, and tomato, suggesting that those plants lack an R gene corresponding to HopPsyA; however, bacteria carrying pHIR11 do elicit an HR in those plants if transformed with avr genes that they recognize (32, 33) (Fig. 3). (viii) For AvrB and AvrPto, two of the Avr proteins demonstrated to work with pHIR11 in triggering an R gene-dependent HR, there are multiple lines of evidence that recognition by the R gene system occurs inside plant cells (33, 34, 35 and 36) (Fig. 3). Thus, pHIR11 directs heterologous HR elicitation in tobacco because it happens to encode a complete Hrp system plus an effector recognized by tobacco. The minimum genetic requirements for being a bacterial parasite of plants are unknown. Parasitism apparently requires the delivery of multiple effector proteins that suppress general antibacterial defenses and/or promote nutrient release from plant cells. The number of effector proteins secreted by a given strain and the virulence targets of those proteins are unknown. More than 30 avr/hop genes from various P. syringae and Xanthomonas strains have been cloned and sequenced (37). Members of the AvrRxv/AvrBsT family are unique in being similar to animal pathogen effectors—the Yersinia YopJ/Salmonella Avr A family (38). The ability of YopJ to inhibit MAP kinase kinases suggests a potential role of the AvrRxv proteins in suppressing plant defenses (39). The Xanthomonas AvrBs2 is unique in having a sequence that predicts an enzymatic activity, and the similarity of AvrBs2 to A. tumefaciens agrocinopine synthase suggests a role in bacterial nutrition (40). Finally, the P. syringae AvrD protein family is unique in directing the synthesis of syringolide elicitors of an Rpg4-specific HR in soybean (41). Other Avr proteins offer little clue to their function as effectors, although it is noteworthy that many make a quantitative contribution to virulence (37), and they can be deleterious when overexpressed in plant cells lacking cognate R genes (33, 42) (Fig. 3).

FUNCTIONS OF HRP SYSTEM COMPONENTS P. syringae hrp genes were initially characterized on the basis of plant reaction phenotypes: typical mutants no longer elicited the HR in nonhosts or were pathogenic (or parasitic) in hosts (43, 44). Subsequently, levels of hrp expression and the secretion of HrpZ and Avr/Hop effectors provided phenotypes for dissecting the functions of Hrp system components. Table 1 summarizes the phenotypes of representative Hrp system mutants and indicates the following genetically distinguishable functions (Fig. 1): (i) positive and negative regulation of the Hrp regulon (hrp/hrc genes and known avr/hop genes); (ii) export of harpins and effectors across the inner membrane via a translocator apparently evolved from the flagellar biogenesis system; (iii) export of harpins and effectors across the outer membrane through a channel formed by secretin multimers; (iv) translocation of effectors across the plant cell wall and plasma membrane into the host cytoplasm by an unknown system.

Fig. 3. Summary of evidence that AvrB elicits an Rpg1-dependent HR whether delivered by P. syringae pv glycinea, P. fluorescens (pHIR11), or biolistic transformation. Top indicates the responses of soybean cultivars Acme and Harosoy to P. syringae pv glycinea and P. fluorescens (pHIR11) strains with or without avrB (33, 86). “P” indicates pathogenicity (bacterial blight symptoms); “HR” indicates rapid confluent collapse with an inoculum level of 5 × 107 cells/ml and no disease development at any inoculum level; “Null” indicates no visible response. Bottom shows the effects on β-glucuronidase (GUS) activity of transient coexpression of avrB in leaf cells of Acme and Harosoy. The leaves were biolistically cobombarded with tungsten particles coated with the indicated plasmids, incubated for 24 h, and then histochemically stained for GUS activity (23), which is an indicator of the viability of the transformed cells (87). Note that the histochemically stained spots are much smaller in the Acme leaves expressing avrB and completely absent in Harosoy leaves expressing avrB. Recent observations highlight the complexity of Hrp system functions. Regulation involves not only the positive regulators HrpR, HrpS, and HrpL (45), but also HrpA and HrpV: multiple hrp genes are activated by the HrpA pilin and repressed by HrpV. Psy 61 and Pto DC3000 strains with nonpolar deletions of hrpA no longer express hrp genes in culture, and they have a Hrp− phenotype in plant bioassays (46). In contrast, constitutive expression of hrpV in Psy 61 represses the production of multiple Hrp components in culture, although it does not abolish HR elicitation, which suggests significant differences between Hrp regulation in culture and in plant a (47). The repressive effects on hrp gene expression of deleting hrpA or overexpressing hrpV can be overcome by constitutive expression of hrpL or hrpRS, which suggests that HrpA and HrpV act upstream of the HrpRS-HrpL activation cascade (although effects on hrpRS expression have not yet been tested) (46, 47). The ability of constitutively produced HrpRS to restore the expression of the Hrp regulon in a Pto DC3000 hrpA mutant enabled testing of the role of HrpA in the secretion of the HrpW harpin and AvrPto (46). Surprisingly, HrpA is required for both of these proteins to be secreted



8773

in culture, which complicates genetic dissection of the role of HrpA in the translocation of effector proteins into plant cells. Furthermore, no mutations in P. syringae have identified factors specifically involved in the translocation step, as would be indicated by a block in effector protein translocation into plant cells (detectable by an R gene-dependent HR) without a block in secretion in culture. Mutants of this class have been extensively explored in Yersinia (yopB and yopD) (48). Finally, it is noteworthy that Psy B728a hrpJ::ΩSpR and hrcC::nptII (nonpolar) mutants are strongly reduced in their ability to colonize bean leaves grown in the field from surfaceinoculated seeds (9). The ability of Psy to achieve threshold population levels as an epiphyte on the surface of bean leaves has been shown to be important in pathogenesis in the field (8). Interestingly, B728a hrpJ::ΩSpR and hrcC::nptII mutants achieve high population levels on occasional leaves, and at a similarly low frequency they cause brown spot symptoms (9). This suggests that the Psy B728a Hrp system has a larger role in growth in planta than virulence per se, which is consistent with the finding of gacS (lemA) mutants of Psy B728a that do not produce disease lesions even though they grow to wild-type levels in bean and produce the HR in nonhost tobacco (49).

THE TRIPARTITE MOSAIC STRUCTURE OF THE P. SYRINGAE HRP PAI To further characterize the Hrp system genes and any candidate effector genes linked to them, we have investigated the sequence of the hrp/ hrc gene clusters and flanking regions of Psy B728a and Pto DC3000 (50). The hrp/hrc cluster resides at the center of a Hrp Pai with three distinct loci that make different contributions to pathogenicity. The hrp/hrc genes of the divergent strains 61 and DC3000 are similar in arrangement, although the hrpA genes are notably different (28% amino acid identity). In contrast, the hrpA genes of strains 61 and B728a are 100% identical. However, B728a is distinguished by a 3.6-kb insert containing homologs of bacteriophage λ genes Ea59 and Ea31 (50). The entire hrp/hrc cluster (hrpK-hrpR) was deleted from Pto DC3000 by a marker-exchange strategy using PCR-amplified DNA from the regions bordering hrpK and hrpR (Fig. 4). As expected, the mutant failed to grow significantly or to cause bacterial speck disease in tomato (Lycopersicon esculentum Mill. cv. Moneymaker) and Arabidopsis (Col-0), and it did not elicit the HR in tobacco (D.E.F. and A.C., unpublished data). Three nucleotides downstream of the hrpK stop codon, the DNA sequence of Psy 61, Psy B728a, and Pto DC3000 is completely divergent (50). This divergent region, the exchangeable effector locus (EEL) further described below, has a significantly lower G + C content than the rest of the Hrp Pai and the P. syringae genome. The EELs have variable lengths of 2.5 kb (Psy 61), 7.3 kb (Psy B728a), and 5.9 kb (Pto DC3000), and they are bounded by hrpK and tRNALeu -queA-tgt sequences. The latter are also found in Pseudomonas aeruginosa but without linkage to any Hrp Pai genes. On the other side of the hrp/hrc cluster, beyond hrpR, resides a conserved effector locus (CEL) of 17-kb (further discussed below). Comparison of the CEL regions sequenced in the divergent strains B728a and DC3000 revealed that the first seven ORFs are arranged identically and have an average DNA sequence identity of 78% and a G + C content that is similar to that of the hrp/hrc region and the rest

Fig. 4. The Hrp Pai of Pto DC3000 and the phenotype of large deletions affecting each of the three major regions. The shaded boxes denote genes considered to be outside of the Hrp Pai. Squares on arrows denote the presence of Hrp boxes. hrpK is presented on the same line as the EEL because ORF1 is in the same apparent operon. Open boxes denote genes in the EEL and CEL. The structure of the hrpJ and hrpU operons is based on partial sequence data and colinearity with this region in Psy 61. Dashed lines indicate the regions deleted, and the Inset (Left) shows the effect of each deletion on the ability of DC3000 to elicit the HR in tobacco or cause disease symptoms in tomato.



8774

of the P. syringae genome. Overall, the Hrp Pai of P. syringae has key properties of Pais possessed by animal pathogens (13), including: (i) the presence of many virulence-associated genes (several with relatively low G + C content) in a large (50-kb) chromosomal region, (ii) linkage to the 3 end of a tRNA gene, (iii) absence from the corresponding locus in a closely related species, and (iv) instability and possession of many sequences related to mobile genetic elements (specifically in the EEL, discussed below).

AN EEL MAKES A SMALL CONTRIBUTION TO PARASITIC FITNESS The Psy B728a EEL possesses three ORFs predicting products similar to known Avr proteins: P. syringae pv phaseolicola AvrPphC and P. syringae pv glycinea AvrC (ORF1); P. syringae pv phaseolicola AvrPphE (ORF2); and Xanthomonas AvrBsT and AvrRxv (ORF5) (50). avrPphE illustrates the instability of the EEL region in being absent from the EELs of Psy 61 and Pto DC3000 and present in P. syringae pv phaseolicola 1302A but in a different location, immediately downstream of hrpK (hrpY) in that strain (51). Although Psy 61 and B728a are in the same pathovar, the strain 61 EEL is completely different and carries only hopPsyA and ORF1, which are present in only a few Psy strains (18, 23). The ORFs in the Pto DC3000 EEL predict no products with similarity to known Avr proteins; however, the ORF1 protein is secreted in a hrp-dependent manner by E. coli (pCPP2156), which expresses an Erwinia chrysanthemi Hrp system and secretes P. syringae Avr proteins (52) (J.R.A. and K.v.D., unpublished data). Several ORFs in these EELs are preceded by Hrp boxes indicative of HrpL-activated promoters (53, 54 and 55). The EELs of these three strains also contain sequences homologous to various mobile genetic elements (50). The Psy B728a EEL carries sequences similar to those in a P. syringae pv phaseolicola plasmid that harbors several avr genes (56) and to sequences homologous to insertion elements that are typically found on plasmids, which suggests plasmid integration via an insertion sequence element in this region (57). Psy B728a ORF3 and ORF4 show similarity to sequences implicated in the horizontal acquisition of the LEE Pai by pathogenic E. coli strains, and the Pto DC3000 EEL carries a TnpA fragment similar to Pseudomonas stutzeri TnpA1 (50). These ORFs are not preceded by Hrp boxes and are unlikely to encode effector proteins. Cosmid pCPP2346, which carries the B728a hrp/hrc region and flanking sequences (4 kb on the left and 13 kb on the right), enabled P. fluorescens to secrete the B728a HrpZ harpin in culture and to elicit the HR in tobacco leaves. However, confluent necrosis developed more slowly than with P. fluorescens (pHIR11). These observations suggested that the product of at least one of the effector genes in the B728a EEL was recognized by an R gene in tobacco. In agreement with this hypothesis, a derivative of plasmid pCPP46 carrying the B728a EEL renders P. syringae pv tabaci avirulent in tobacco (W.-L.D. and A.C., unpublished data). The contribution of the various EELs to the parasitic fitness of P. syringae strains was assayed with appropriate mutants (50). A hopPsyA::TnphoA Psy 61 mutant had previously been shown to be only partially impaired in Hrp phenotypes (31). Deletions of the entire EEL regions of Pto DC3000 (50) and Psy B728a (W.-L.D, unpublished work) were constructed by marker exchange with appropriate border regions subcloned on either side of an ΩSpR cassette. The growth in host plants of mutant and wild-type strains was compared after inoculation by syringe infiltration. The Pto DC3000 ∆EEL mutant was slightly reduced in the final population it achieved in tomato (cv. Moneymaker) (50), but no significant reduction was observed with the Psy B728a ∆EEL mutant in bean (Phaseolus vulgaris L. cv. Eagle) (W.-L.D. and A.C., unpublished data). The mutants also retained the ability to elicit the HR on various nonhosts. Thus, additional effectors encoded elsewhere in the genome apparently contribute to parasitic fitness in hosts and betray the parasite to R gene surveillance in nonhosts.

THE CEL IS IMPORTANT FOR PATHOGENICITY The region to the right of hrpR in DC3000 had been known for several years to contain the avrE locus, which is comprised of two transcriptional units and, when heterologously expressed, causes P. syringae pv glycinea to become avirulent on all soybean cultivars tested (58). Also known to be in this region is the hrpW gene encoding a second harpin, which is distinguished by its C-terminal domain with homology to class III pectate lyases and its ability to bind to calcium pectate (27). A previous sequence analysis of the 5 sequences for the first four transcriptional units beyond hrpR (58) was extended to include the first 14 ORFs to the right of hrpR in Pto DC3000 and a partial sequence of the corresponding region in Psy B728a (50). Unlike the EEL, this region contains no sequences similar to known mobile genetic elements, and it appears conserved between Psy and Pto because the first seven ORFs are arranged identically in these divergent strains and have an average DNA sequence identity of 78%. In Fig. 4, the outer border of the CEL is given tentatively as around ORF10. ORF8 is preceded by a Hrp box and is therefore a candidate effector. In contrast, the gene beyond ORF10 shows homology to a family of bacterial GstA proteins (50). Because glutathione S-transferase activity is common in nonpathogenic fluorescent pseudomonads (59), this gene is not likely to be an effector or part of the CEL. The ORF5 protein is secreted in a hrp-dependent manner by E. coli (pCPP2156), but mutation with an ΩSpr cassette has little effect on either HR elicitation in tobacco or pathogenicity in tomato (A.O.C., J.L.B, and A.C., unpublished data). Notably, six operons in this region are preceded by a Hrp box, which is characteristic of known avr genes in P. syringae (53, 54 and 55, 58) (Fig. 4). To assess the collective contribution of the CEL ORFs that were both partially characterized and likely to encode effectors, we constructed a mutation in Pto DC3000 that replaced avrE through ORF5 with an ΩSpr cassette (50). The ∆CEL mutant still elicited the HR in tobacco, but tissue collapse was delayed 5 h. The mutant no longer elicited disease symptoms in tomato when infiltrated at a concentration of 104 cfu/ml, and growth in plant a was strongly reduced (50). Pathogenicity was restored to the ∆CEL mutant by a plasmid carrying ORF2 through ORF10, and the mutant was able to secrete AvrPto in culture. All of these observations suggest that the ∆CEL mutation does not interfere with Hrp secretion functions and that the loss of pathogenicity can be attributed to the loss of multiple effectors. Finally, although avrE and several other candidate effector genes are located in the Hrp Pai of Pto DC3000, additional effector genes, such as avrPto, are located somewhere else (60), and the complete inventory of effector genes in this strain remains unknown. Because most of the known P. syringae avr genes are associated with mobile genetic elements (61), the avr composition of various P. syringae strains may vary considerably, presumably as a result of opposing selection pressures to promote parasitism while evading host R gene surveillance.

HRPK AND CEL ORF1 The hrpL and hrpR genes bracket a cluster of operons that contain both hrp and hrc genes and appear sufficient to encode a complete Hrp type III secretion system. hrpK and the CEL ORF1 reside in the two borders between this core hrp/hrc cluster and known effector genes, and the functions of these two genes are unknown. The HrpK proteins of Psy and Pto are 79% identical, which makes them more conserved than several of the



8775

proteins encoded by the core hrp/hrc cluster (50). hrpK mutants have a variable Hrp phenotype (51, 62), and a Psy B728a hrpK mutant still elicits the HR in tobacco and secretes HrpZ in culture (Fig. 5). These observations suggest that HrpK is a conserved effector rather than a component of the Hrp system. It is noteworthy that candidate effector genes appear to reside downstream of hrpK in the same operon in Psy B728a and Pto DC3000 (50). In contrast to HrpK, the CEL ORF1 is more likely to be an ancillary component of the Hrp system than an effector, because it is most similar to E. coli murein lytic transglycosylase MltD and shares a lysozyme-like domain with the product of ipgF (63), which is a Shigella flexneri gene linked to type III secretion system genes (64). Although mutations in these genes in Pto DC3000 and S. flexneri have no obvious phenotype (58, 64), other peptidoglycan hydrolases may mask the phenotype (65). The region to the right of hrpR in pHIR11 has not been sequenced and may harbor ORF1. However, TnPhoA mutations in this region have no apparent phenotype (31).

Fig. 5. Secretion of the HrpZ harpin by Psy B728a hrpK mutant CUCPB5092. Bacteria were grown under Hrp-inducing conditions and fractionated into cell-bound (C) and supernatant (S) fractions as previously described (30). Proteins were resolved by SDS/PAGE and analyzed by immunoblotting with anti-HrpZ antibodies. EFFECTOR PROTEIN SECRETION AND A UNIVERSAL TYPE III TARGETING SIGNAL The ability of pHIR11 to deliver the products of avr genes from other P. syringae pathovars suggested that the Hrp system recognizes a universal targeting signal in these proteins. Indeed, the cluster of hrp/hrc genes from the soft-rot pathogen E. chrysanthemi, cloned in cosmid pCPP2156, enables E. coli to secrete AvrB and AvrPto (52), and Erwinia amylovora and P. syringae can interchangeably deliver their respective DspE and AvrE proteins to plants, as indicated by appropriate plant reactions (66). Moreover, Xanthomonas campestris pv vesicatoria can secrete AvrB (67), Yersinia enterocolitica can secrete both AvrB and AvrPto (68), and X. campestris pv vesicatoria and E. coli (pCPP2156) can secrete YopE, a Yersinia effector (67, 68). These observations extend the original discovery of heterologous delivery of effectors by the type III secretion systems of Yersinia, Salmonella, and Shigella, and they strongly suggest the existence of universal targeting signals in proteins traveling all type III secretion pathways (69). Yersinia secretes multiple Yops (Yersinia outer proteins) via the type III secretion pathway, and Yops carry an mRNA targeting signal in their first 15 codons (70, 71). Fusion of the first 15 codons of YopE to an Npt reporter is sufficient for type III secretion of the hybrid to the bacterial milieu, and mutations that shift the reading frame of these codons do not abolish secretion, which indicates that the targeting information resides in the mRNA rather than the peptide (71). Several Yops, including YopE, have a second targeting domain, which depends on a customized chaperone and is required for translocation into host cells (72). However, YopQ has only an mRNA targeting signal (73), and it is also secreted by E. coli (pCPP2156). Several observations support the hypothesis that the first 15 codons of avrPto similarly carry an mRNA targeting signal: (i) deletion of the first 10 codons abolishes the secretion of AvrPto by E. coli (pCPP2156) and Y. enterocolitica; (ii) fusion of the first 15 codons of AvrPto to an Npt reporter is sufficient for type III secretion of the hybrid to the Y. enterocolitica milieu; and (iii) mutations that shift the reading frame of the AvrPto codons (+1, +2, and −1) do not abolish secretion of the Npt reporter (68). Thus, the mRNA signal recognized by type III secretion systems appears to be shared by the effectors of both plant and animal pathogens. The efficiency with which different P. syringae Avr proteins are secreted in culture by different type III systems varies considerably (as indicated by the proportion of total effector protein released to the medium), and secretion by native P. syringae Hrp systems has been reported only recently (30, 74). AvrB and AvrPto secretion illustrates this variability. Both proteins are secreted much more strongly by Y. enterocolitica than by E. coli (pCPP2156) (52, 68), and AvrPto is the only one of these two that is secreted by P. syringae (30). Heterologously expressed AvrRpt2 is similarly secreted by E. coli (pCPP2156) and Pto DC3000 (74), but much less efficiently than AvrPto is secreted by P. syringae (30). These differences in secretion behavior in culture bear no apparent relationship to the biological activity of the effectors. For example, although AvrB secretion from P. syringae has yet to be observed in culture, AvrB is almost certainly delivered into plant cells (33, 34). Thus, it remains unclear whether the differing secretion behaviors of AvrB and AvrPto reflect some form of effector sorting by the pathway or whether it is peculiar to secretion in culture. In Yersinia, the type III pathway can be activated by growth at 37°C in low-calcium medium, and these conditions obviate the normal requirement for host cell contact (1). In P. syringae, the hrp/hrc genes are induced in minimal media that do not support rapid growth (75), and two environmental factors relevant to pathogenesis have been found to be critical for the secretion of HopPsyA and AvrPto by Psy 61 and Pto DC3000, respectively (30). That is, both proteins are secreted at pH 6.0 and 20°C but not at pH 7.0 or 30°C. These conditions correspond to the low pH of plant intercellular fluids and the cool temperatures that favor disease development with these bacteria (76). However, the secretion capacity of the type III systems of plant pathogens in culture, even under optimal conditions, seems much less than that of Yersinia, regardless of the effector protein. One explanation is that the Hrp systems are not fully activated until contact with plant cells, and the appropriate mutants or signals needed to unlock that capacity have not been found. Alternatively, the secretion capacity of the Hrp system may be reduced by adaptations for delivery through the plant cell wall matrix. In support of the first hypothesis, the Hrp regulon in Ralstonia solanacearum is induced maximally in culture by cocultivation with plant cells (77, 78). We have similarly observed that a PhrpA-uidA fusion is induced 20-fold when suspension-cultured tobacco cells are added to Hrp-inducing minimal medium (W.-L.D. and A.C., unpublished data), which suggests that contact-dependent induction may be widespread with the Hrp systems of plant pathogenic bacteria. The differing abilities of type III systems to translocate effector-reporter hybrid proteins provides support for the second hypothesis. Early evidence for the translocation of Yop proteins into host cells was obtained with a YopE-CyA hybrid that produced adenylate cyclase activity in a calmodulin-dependent manner (15). This reporter system is a powerful tool for investigating translocated proteins and their targeting signals. Unfortunately, translocation into host cells of effector-reporter hybrids has not been described for any plant pathogens, and fusion of the C terminus of AvrRpt2 with Myc6, Gfp, or CyA blocks avrRpt2-Rps2-dependent HR elicitation and diminishes the virulence of Pto DC3000 in Arabidopsis plants lacking the cognate Rps2 gene (74). It seems that the fusion of large polypeptides to the C terminus of Avr proteins disrupts Hrp functions. Thus, it appears that effector proteins can be targeted to the type III pathway by a universal mRNA targeting signal and



8776

secreted across inner and outer membranes by machinery that is common to all type III systems. However, translocation into host cells is likely to be unique because of adaptations to the fundamentally different surfaces of plant and animal cells.

CONCLUSIONS Our investigation of the basis for P. syringae phytopathogenicity has focused on the mechanisms underlying elicitation of the HR, a signature of plant encounters with incompatible phytopathogens, and it has revealed the modular nature of the process and its underlying genetics. Thus, the requirements for HR elicitation can be reduced to two components: a functional Hrp type III secretion system and an injected effector protein that is recognized by the R-gene surveillance system of the test plant. Hrp protein secretion in culture can be further dissected genetically, revealing two operons directing export across the inner membrane and another directing export across the outer membrane. The effector proteins also appear modular in their possession of a universal type III targeting signal in the 5 ends of their cognate mRNAs. This modularity has several experimental consequences: a cloned P. syringae Hrp system is sufficient to direct heterologous secretion and delivery of effector proteins by nonphytopathogenic bacteria; effector proteins from P. syringae can be heterologously delivered into plants by Erwinia Hrp systems or secreted in culture by the Yersinia type III system; and the need for any Hrp system for HR elicitation can be circumvented entirely by delivery of effector protein genes into plant cells by biolistics or Agrobacterium-mediated transformation. The modular nature of the Hrp/effector system is also seen in the tripartite mosaic architecture of the P. syringae Hrp Pai, which features both exchangeable and conserved effector loci. The EEL represents a region in flux because of its high frequency of recombination, and this probably allows fine tuning of pathogenicity. On the basis of its similar G + C content to the hrp/hrc cluster and the rest of the P. syringae chromosome, the CEL was probably acquired at the same time as the core hrp/hrc cluster, and it encodes effectors that contribute more significantly to pathogenicity than the EEL. The modular nature of the Hrp/effector system suggests that it functions universally in a broad range of potential plant hosts and with a frequently changing pool of effectors. Effector gene instability may be driven by the evolution of R gene surveillance systems and changes in effector targets in plants. The next challenge is to identify all of the effector proteins produced by model strains of P. syringae, to understand how these proteins promote parasitism, and to understand how the type III system of phytopathogens has been adapted to deliver these proteins across plant cell walls. We thank Nam-Hai Chua (The Rockefeller University, New York, NY) for providing pTA7002. This work was supported by National Science Foundation grant MCB 97–35303-4488 (A.C.), National Research Initiative Competitive Grants Program United States Department of Agriculture grants 98–35303-6464 (J.R.A.) and 98–35303-6662 (D.E.F.), and United States Public Health Service Grant AI 42797 from the National Institutes of Health—National Institute of Allergy and Infectious Diseases Branch (O.S.). 1. Hueck, C. J. (1998) Microbiol. Mol Biol. Rev. 62, 379–433. 2. Galán, J. E. & Collmer, A. (1999) Science 284, 1322–1328. 3. Hirano, S. S. & Upper, C. D. (1990) Annu. Rev. Phytopathol. 28, 155–177. 4. Richael, C. & Gilchrist, D. (1999) Physiol. Mol. Plant Pathol. 55, 5–12. 5. Alfano, J. R. & Collmer, A. (1997) J. Bacteriol. 179, 5655–5662. 6. Keen, N. T. (1990) Annu. Rev. Genet. 24, 447–463. 7. Huang, H.-C, Schuurink, R., Denny, T. P., Atkinson, M. M., Baker, C. J., Yucel, I., Hutcheson, S. W. & Collmer, A. (1988) J. Bacteriol. 170, 4748–4756. 8. Hirano, S. S, Rouse, D. I., Clayton, M. K. & Upper, C. D. (1995) Plant Dis. 79, 1085–1093. 9. Hirano, S., Charkowski, A. O., Collmer, A., Willis, D. K. & Upper, C. D. (1999) Proc. Natl. Acad. Sci. USA 96, 9851–9856. 10. Manceau, C. & Horvais, A. (1997) Appl. Environ. Microbiol. 63, 498–505. 11. Martin, G. B. (1999) Curr. Opin. Plant Biol. 2, 273–279. 12. Groisman, E. A. & Ochman, H. (1996) Cell 87, 791–794. 13. Hacker, J, Blum-Oehler, G., Muhldorfer, I. & Tschape, H. (1997) Mol. Microbiol. 23, 1089–1097. 14. Rosqvist, R., Magnusson, K. E. & Wolf-Watz, H. (1994) EMBO J. 13, 964–972. 15. Sory, M.-P. & Cornelis, G. R. (1994) Mol. Microbiol. 14, 583–594. 16. Huang, H.-C., He, S. Y., Bauer, D. W. & Collmer, A. (1992) J. Bacteriol. 174, 6878–6885. 17. Huang, H.-C., Xiao, Y., Lin, R.-H., Lu, Y., Hutcheson, S. W. & Collmer, A. (1993) Mol. Plant-Microbe Interact. 6, 515–520. 18. Heu, S. & Hutcheson, S. W. (1993) Mol. Plant–Microbe Interact. 6, 553–564. 19. Lidell, M. C. & Hutcheson, S. W. (1994) Mol. Plant–Microbe Interact. 7, 488–497. 20. Huang, H.-C., Lin, R.-W, Chang, C.-J., Collmer, A. & Deng, W.-L. (1995) Mol. Plant-Microbe Interact. 8, 733–746. 21. Preston, G., Huang, H.-C, He, S. Y. & Collmer, A. (1995) Mol. Plant-Microbe Interact. 8, 717–732. 22. Deng, W.-L., Preston, G., Collmer, A., Chang, C.-J. & Huang, H.-C. (1998) J. Bacteriol. 180, 4523–4531. 23. Alfano, J. R., Kim, H.-S, Delaney, T. P. & Collmer, A. (1997) Mol. Plant– Microbe Interact. 10, 580–588. 24. He, S. Y., Huang, H.-C. & Collmer, A. (1993) Cell 73, 1255–1266. 25. Yuan, J. & He, S. Y. (1996) J. Bacteriol. 178, 6399–6402. 26. Roine, E, Wei, W., Yuan, J., Nurmiaho-Lassila, E.-L., Kalkkinen, N., Romantschuk, M. & He, S. Y. (1997) Proc. Natl. Acad. Sci. USA 94, 3459–3464. 27. Charkowski, A. O., Alfano, J. R., Preston, G., Yuan, J., He, S. Y. & Collmer, A. (1998) J. Bacteriol. 180, 5211–5217. 28. Wei, Z.-M., Laby, R. J., Zumoff, C. H., Bauer, D. W., He, S. Y., Collmer, A. & Beer, S. V. (1992) Science 257, 85–88. 29. Alfano, J. R., Bauer, D. W., Milos, T. M. & Collmer, A. (1996) Mol. Microbiol. 19, 715–728. 30. van Dijk, K., Fouts, D. E., Rehm, A. H., Hill, A. R., Collmer, A. & Alfano, J. R. (1999) J. Bacteriol. 181, 4790–4797. 31. Huang, H.-C., Hutcheson, S. W. & Collmer, A. (1991) Mol. Plant–Microbe Interact. 4, 469–476. 32. Pirhonen, M. U., Lidell, M. C., Rowley, D. L., Lee, S. W., Jin, S., Liang, Y., Silverstone, S., Keen, N. T. & Hutcheson, S. W. (1996) Mol. Plant-Microbe Interact. 9, 252–260. 33. Gopalan, S., Bauer, D. W., Alfano, J. R., Loniello, A. O., He, S. Y. & Collmer, A. (1996) Plant Cell 8, 1095–1105. 34. Leister, R. T., Ausubel, F. M. & Katagiri, F. (1996) Proc. Natl. Acad. Sci. USA 93, 15497–15502. 35. Tang, X., Frederick, R. D., Zhou, J., Halterman, D. A., Jia, Y. & Martin, G. B. (1996) Science 274, 2060–2062. 36. Scofield, S. R., Tobias, C. M., Rathjen, J. P., Chang, J. H., Lavelle, D. T., Michelmore, R. W. & Staskawicz, B. J. (1996) Science 274, 2063–2065. 37. Leach, J. E. & White, F. F. (1996) Annu. Rev. Phytopathol. 34, 153–179. 38. Ciesiolka, L. D., Hwin, T., Gearlds, J. D., Minsavage, G. V., Saenz, R., Bravo, M., Handley, V., Conover, S. M., Zhang, H., Caporgno, J., et al. (1999) Mol. Plant–Microbe Interact. 12, 35–44. 39. Orth, K., Palmer, L. E., Bao, Z. Q., Stewart, S., Rudolph, A. E., Bliska, J. B. & Dixon, J. E. (1999) Science 285, 1920–1923. 40. Swords, K. M. M., Dahlbeck, D., Kearney, B., Roy, M. & Staskawicz, B. J. (1996) J. Bacteriol. 178, 4661–4669. 41. Yucel, I., Midland, S. L., Sims, J. J. & Keen, N. T. (1994) Mol Plant–Microbe Interact. 7, 148–150. 42. McNellis, T. W., Mudgett, M. B., Li, K., Aoyama, T., Horvath, D., Chua, N.-H. & Staskawicz, B. J. (1998) Plant J. 14, 247–257. 43. Lindgren, P. B, Peet, R. C. & Panopoulos, N. J. (1986) J. Bacteriol. 168, 512–522. 44. Lindgren, P. B., Panopoulos, N. J., Staskawicz, B. J. & Dahlbeck, D. (1988) Mol Gen. Genet. 211, 499–506. 45. Hutcheson, S. W. (1997) in Plant–Microbe Interactions, Vol. 3, eds. Stacey, G. & Keen, N. T. (Chapman & Hall, New York), pp. 145–179. 46. Wei, W., Plovanich-Jones, A., Deng, W.-L., Collmer, A., Huang, H.-C. & He, S. Y. (2000) Proc. Natl Acad. Sci. USA 97, 2247–2252. 47. Preston, G., Deng, W.-L., Huang, H.-C. & Collmer, A. (1998) J. Bacteriol. 180, 4532–4537. 48. Cornelis, G. R. & Van Gijsegem, F. (2000) Annu. Rev. Microbiol, in press. 49. Willis, D. K., Hrabak, E. M, Rich, J. J., Barta, T. M., Lindow, S. E. & Panopoulos, N. J. (1990) Mol. Plant–Microbe Interact. 3, 149–156. 50. Alfano, J. R., Charkowski, A. O., Deng, W.-L., Badel, J. L., Petnicki-Ocwieja, T., van Dijk, K. & Collmer, A. (2000) Proc. Natl. Acad. Sci. USA 97, 4856– 4861. 51. Mansfield, J., Jenner, C., Hockenhull, R., Bennett, M. A. & Stewart, R. (1994) Mol. Plant–Microbe Interact. 7, 726–739. 52. Ham, J. H., Bauer, D. W., Fouts, D. E. & Collmer, A. (1998) Proc. Natl. Acad. Sci. USA 95, 10206–10211. 53. Shen, H. & Keen, N. T. (1993) J. Bacteriol. 175, 5916–5924.



8777

54. Innes, R. W., Bent, A. F., Kunkel, B. N., Bisgrove, S. R. & Staskawicz, B. J. (1993) J. Bacteriol. 175, 4859–4869. 55. Xiao, Y. & Hutcheson, S. (1994) J. Bacteriol. 176, 3089–3091, and correction (1994) 176, 6158. 56. Jackson, R. W., Athanassopoulos, E., Tsiamis, G., Mansfield, J. W., Sesma, A., Arnold, D. L., Gibbon, M. J., Murillo, J., Taylor, J. D. & Vivian, A. (1999) Proc. Natl. Acad. Sci. USA 96, 10875–10880. 57. Szabo, L. J. & Mills, D. (1984) J. Bacteriol. 157, 821–827. 58. Lorang, J. M. & Keen, N. T. (1995) Mol. Plant–Microbe Interact. 8, 49–57. 59. Zablotowicz, R. M., Hoagland, R. E., Locke, M. A. & Hickey, W. J. (1995) Appl. Environ. Microbiol. 61, 1054–1060. 60. Salmeron, J. M. & Staskawicz, B. J. (1993) Mol. Gen. Genet. 239, 6–10. 61. Kim, J. F., Charkowski, A. O., Alfano, J. R., Collmer, A. & Beer, S. V. (1998) Mol. Plant–Microbe Interact. 11, 1247–1252. 62. Bozso, Z., Ott, P. G., Kecskes, M. L. & Klement, Z. (1999) Physiol. Mol. Plant Pathol. 55, 215–223. 63. Mushegian, A. R., Fullner, K. J., Koonin, E. V. & Nester, E. W. (1996) Proc. Natl. Acad. Sci. USA 93, 7321–7326. 64. Allaoui, A., Menard, R., Sansonetti, P. J. & Parsot, P. (1993) Infect. Immun. 61, 1707–1714. 65. Dijkstra, A. J. & Keck, W. (1996) J. Bacteriol. 178, 5555–5562. 66. Bogdanove, A. J., Kim, J. F., Wei, Z., Kolchinsky, P., Charkowski, A. O., Conlin, A. K., Collmer, A. & Beer, S. V. (1998) Proc. Natl. Acad. Sci. USA 95, 1325–1330. 67. Rossier, O., Wengelnik, K., Hahn, K. & Bonas, U. (1999) Proc. Natl. Acad. Sci. USA 96, 9368–9373. 68. Anderson, D. M., Fouts, D. E., Collmer, A. & Schneewind, O. (1999) Proc. Natl. Acad. Sci. USA 96, 12839–12843. 69. Rosqvist, R., Hakansson, S., Forsberg, A. & Wolf-Watz, H. (1995) EMBO J. 14, 4187–4195. 70. Anderson, D. M. & Schneewind, O. (1999) Curr. Opin. Microbiol. 2, 18–24. 71. Anderson, D. M. & Schneewind, O. (1997) Science 278, 1140–1143. 72. Wattiau, P., Woestyn, S. & Cornelis, G. R. (1996) Mol. Microbiol. 20, 255–262. 73. Anderson, D. M. & Schneewind, O. (1999) Mol. Microbiol. 31, 1139–1148. 74. Mudgett, M. B. & Staskawicz, B. J. (1999) Mol. Microbiol. 32, 927–941. 75. Huynh, T. V., Dahlbeck, D. & Staskawicz, B. J. (1989) Science 245, 1374–1377. 76. Rudolph, K., Burr, T. J., Mansfield, J. W., Stead, D., Vivian, A. & von Kietzell, J. (1997) Pseudomonas syringae Pathovars and Related Pathogens (Kluwer, Dordrecht, The Netherlands), p. 663. 77. Marenda, M., Brito, B., Callard, D., Genin, S., Barberis, P., Boucher, C. & Arlat, M. (1998) Mol. Microbiol. 27, 437–453. 78. Brito, B., Marenda, M., Barberis, P., Boucher, C. & Genin, S. (1999) Mol. Microbiol. 31, 237–251. 79. Xiao, Y., Lu, Y., Heu, S. & Hutcheson, S. W. (1992) J. Bacteriol. 174, 1734–1741. 80. Xiao, Y., Heu, S., Yi, J., Lu, Y. & Hutcheson, S. W. (1994) J. Bacteriol. 176, 1025–1036. 81. Charkowski, A O., Huang, H.-C. & Collmer, A (1997) J. Bacteriol. 179, 3866–3874. 82. Deng, W.-L. & Huang, H.-C. (1998) J. Bacteriol. 181, 2298–2301. 83. Grimm, C. & Panopoulos, N. J. (1989) J. Bacteriol. 171, 5031–5038. 84. Rahme, L. G., Mindrinos, M. N. & Panopoulos, N. J. (1991) J. Bacteriol. 173, 575–586, erratum (1992) 174, 3840. 85. Aoyama, T. & Chua, N.-H. (1997) Plant J. 11, 605–612. 86. Staskawicz, B., Dahlbeck, D., Keen, N. & Napoli, C. (1987) J. Bacteriol. 169, 5789–5794. 87. Mindrinos, M., Katagiri, F., Yu, G.-L. & Ausubel, F. M. (1994) Cell 78, 1089–1099.


MOLECULAR AND CELL BIOLOGY ASPECTS OF PLAGUE

8778

Colloquium Molecular and cell biology aspects of plague Guy R. Cornelis* Microbial Pathogenesis Unit, Christian de Duve Institute of Cellular Pathology, and Faculté de Médecine, Université Catholique de Louvain, Avenue Hippocrate, 74, UCL 74.49, B-1200 Brussels, Belgium A 70-kb virulence plasmid (sometimes called pYV) enables Yersinia spp. to survive and multiply in the lymphoid tissues of their host. It encodes the Yop virulon, a system consisting of secreted proteins called Yops and their dedicated type III secretion apparatus called Ysc. The Ysc apparatus forms a channel composed of 29 proteins. Of these, 10 have counterparts in almost every type III system. Secretion of some Yops requires the assistance, in the bacterial cytosol, of small individual chaperones called the Syc proteins. These chaperones act as bodyguards or secretion pilots for their partner Yop. Yop proteins fall into two categories. Some are intracellular effectors, whereas the others are “translocators” needed to deliver the effectors across the eukaryotic plasma membrane, into eukaryotic cells. The translocators (YopB, YopD, LcrV) form a pore of 16–23 Å in the eukaryotic cell plasma membrane. The effector Yops are YopE, YopH, YpkA/YopO, YopP/YopJ, YopM, and YopT. YopH is a powerful phosphotyrosine phosphatase playing an antiphagocytic role by dephosphorylating several focal adhesion proteins. YopE and YopT contribute to antiphagocytic effects by inactivating GTPases controlling cytoskeleton dynamics. YopP/YopJ plays an anti-inflammatory role by preventing the activation of the transcription factor NF-KB. It also induces rapid apoptosis of macrophages. Less is known about the role of the phosphoserine kinase YopO/YpkA and YopM. bacterial pathogenesis | Yersinia | Yops | translocation | type III secretion Y ersinia pestis has in the past caused social devastation on a scale unmatched by any other infectious agent. Although it is presently not a major public health problem, there are still at least 2,000 cases of plague reported annually, and plague has recently been recognized as a reemerging disease by the World Health Organization. The pathogenicity of Yersinia results from its impressive ability to overcome the defenses of the mammalian host and to overwhelm it with massive growth. Multiplication of Y. pestis is largely extracellular (1). In infected mice, significant levels of interferon γ (IFN-γ) and tumor necrosis factor α (TNF-α) arise only just before death. In contrast, prompt and marked synthesis of these cytokines is observed upon infection with avirulent strains (2). All these observations suggest that the pathogenicity arsenal of Y. pestis protects the bacterium from phagocytosis and slows down the onset of the inflammatory response. The closely related food-borne pathogens Yersinia pseudotuberculosis and Yersinia enterocolitica cross the intestinal barrier and multiply in the abdominal lymphoid tissues. Although they cause infections that are generally self-limited, they share with Y. pestis the Yop virulon, the core of the Yersinia pathogenicity arsenal. This Yop virulon allows extracellular Yersinia docked at the surface of a host cell to inject specialized proteins, called Yops, across the plasma membrane. The injected Yops disturb the dynamics of the cytoskeleton and block the production of pro-inflammatory cytokines, thereby favoring the survival of the invading Yersinia. The Yop virulon is thus a complex weapon for close combat with cells of the immune system (for an exhaustive review see ref.3). It is the archetype of the so-called “type III secretion” virulence mechanisms now identified in more than a dozen major animal or plant pathogens (for review see ref.4).

A DEVICE TO INJECT BACTERIAL PROTEINS ACROSS EUKARYOTIC CELL MEMBRANES The Yersinia Ysc Secretion Apparatus. “Yop secretion” was discovered around 1990 by trying to understand the mysterious phenomenon of Ca2+ dependency: when incubated at 37°C in the absence of Ca2+ ions, Yersinia bacteria do not grow but, instead release large amounts of proteins called Yops into the culture supernatant (5). Although it is generally referred to as Yop “secretion,” it is not a physiological secretion but rather a massive leakage resulting from the artificial opening of an otherwise tightly controlled delivery apparatus. Despite the fact that it is presumably artifactual, this observation turned out to be of paramount importance because it allowed the genetic analysis that led to the identification of 29 ysc (Yop secretion) genes involved in the process of Yop release. Among the 29 Ysc proteins, 10 (YscC, -J, -N, -O, -Q, -R, -S, -T, -U, and -V) appear to have counterparts in almost every type III secretion apparatus. YscC belongs to the family of secretins, a group of outer membrane proteins involved in the transport of various macromolecules and filamentous phages across the outer membrane. Similar to other secretins, it forms a ring-shaped structure with an external diameter of about 200 Å and an apparent central pore of about 50 Å (6). At least one disulfide bond is essential for its assembly (7), and its proper insertion in the outer membrane requires the presence of an ancillary lipoprotein called YscW (6). Four proteins (YscD, -R, -U, and -V, formerly called LcrD) have been shown, and two other proteins (YscS and -T) have been predicted to span the inner membrane. The secretion process absolutely requires YscN, a 47.8-kDa protein with ATP-binding motifs (Walker boxes A and B) resembling the β catalytic subunit of F0F1 proton translocase and related ATPases (8). YscJ is a lipoprotein that has not been localized yet, but its counterpart in Pseudomonas syringae has been shown to span the inner and outer membranes (9). Little is known about the YscL, YscQ, and Ysc proteins, which are less conserved. Finally, the two proteins YscO and YscP are themselves released upon Ca2+ chelation, suggesting that they belong to the external part of the apparatus (10, 11, 69). Fig. 1 summarizes current knowledge of the Yop virulon. Assembly of the bacterial flagellum also involves a type III secretion system. This system has no secretin but it has counterparts to the nine other conserved Ysc proteins (YscJ, -N, -O, -Q, -R, -S, -T, -U, and -V). All these proteins belong to the most internal part of the basal body—i.e., the MS ring, the C ring, and the ATPase (reviewed in ref.12), which is in good agreement with the localization proposed for the homologous Ysc proteins. Thus, the similarity between the Ysc apparatus and the flagellum export apparatus resides in their most inner part. While the Salmonella and Shigella “injectisomes” can be visualized by electron microscopy (13, 14), such visualization is not yet the case for the Yersinia Ysc apparatus. Little is known about the actual mechanism of export, but it is generally assumed that the Ysc apparatus serves as a hollow conduit through which the exported proteins travel to cross the two membranes and the peptidoglycan barrier, in one step. Whether proteins travel folded or unfolded has not yet been demonstrated but, given the size of channel, it is likely that they travel at least partially unfolded.

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviation: TNF-α, tumor necrosis factor α. * E-mail: [email protected].



8779

Fig. 1. The Yop virulon. When Yersinia are placed at 37°C in a rich environment, the Ysc secretion channel is installed. Proteins YscD, -R, -S, -T, -U, and -V are localized in the inner membrane (IM), whereas YscC and YscP are exposed at the bacterial surface. Lipoprotein YscW stabilizes YscC. YscN belongs to the family of ATPases. A stock of Yop proteins is synthesized, and some of them are capped with their specific Syc chaperone. As long as there is no contact with a eukaryotic cell, a stop-valve, possibly made of YopN, TyeA, and LcrG, blocks the Ysc secretion channel. On contact with a eukaryotic target cell, the bacterium attaches tightly by interaction between its YadA and Inv adhesins and β-integrins, and the secretion channel opens. The Yops are then transported through the Ysc channel, and the Yop effectors are translocated across the plasma membrane, guided by the translocators YopB, YopD, and LcrV. Translocation of Effectors Across Animal Cell Membranes. Purified secreted Yops have no cytotoxic effect on cultured cells, although live extracellular Yersinia have such an activity. Cytotoxicity nevertheless depends on the capacity of the bacterium to secrete YopE and YopD, and YopE alone is cytotoxic when microinjected into the cells (15). This observation led to the hypothesis that YopE is a cytotoxin that needs to be injected into the eukaryotic cell's cytosol by a mechanism involving YopD to exert its effect (15). This hypothesis was demonstrated by confocal laser scanning microscopy (16) and by the adenylate cyclase reporter enzyme strategy, an approach that is now widely used in “type III secretion” (17): infection of eukaryotic cells with a recombinant Y. enterocolitica producing hybrid proteins consisting of the N terminus of various Yops (other than YopB and YopD) fused to the catalytic domain of a calmodulin-dependant adenylate cyclase (Yop-Cya proteins) leads to an accumulation of cyclic AMP (cAMP) in the cells. Since there is no calmodulin in the bacterial cell and culture medium, this accumulation of cAMP signifies the internalization of Yop-Cya into the cytosol of eukaryotic cells (17). The phenomenon is strictly dependant on the presence of YopD and YopB. Thus, extracellular Yersinia inject Yops into the cytosol of eukaryotic cells by a mechanism that involves at least YopD and YopB (18, 19). Yops are thus a collection of intracellular “effectors” (YopE, YopH, YopM, YpkA/YopO, YopP/YopJ, and YopT) and “translocators” (including YopB and YopD) which are required for the translocation of the effectors across the plasma membrane of eukaryotic cells (20). This model of intracellular delivery of Yop effectors by extracellular adhering bacteria is now largely supported by a number of other results, including immunological observations. During a mouse infection by wild-type Y. enterocolitica, the epitope formed by amino acid residues 249–257 of the YopH effector protein is presented by MHC class I molecules, as cytosolic proteins are, and not by MHC class II molecules, as antigens are that are processed in phagocytic vacuoles (21). A Pore Formed by Translocators. The translocators YopB and YopD have hydrophobic domains, suggesting that they could act as transmembrane proteins (16, 17 and 18, 22). In agreement with this possibility, Yersinia has a contact-dependent lytic activity on sheep erythrocytes, depending on YopB and YopD (19, 23), which suggests that the translocation apparatus involves some kind of a pore in the target cell membrane by which the Yop effectors pass through to reach the cytosol. This YopB- and YopD-dependent lytic activity is higher when the effector yop genes are deleted, suggesting that the pore is normally filled with effectors (19, 23). The idea of a translocation pore is further supported by the observation that the membrane of macrophage-like cells infected with an effector polymutant Y. enterocolitica becomes permeable to small dyes (23). If the macrophages are preloaded before the infection with a low-molecular weight fluorescent marker, they release the fluorescent marker but not cytosolic proteins, indicating that there is no membrane lysis but rather insertion of a small pore (diameter 16–23 Å) into the macrophage plasma membrane (23). The hypothesis of a channel is reinforced by the observation that artificial liposomes that have been incubated with Yersinia contain channels detectable by electrophysiology (24). All these events are dependent on the presence of the translocators YopB and YopD. These two hydrophobic Yops seem thus to be central for the translocation of the effectors and for the formation of a channel in lipid membranes. They presumably play different roles in pore formation. Indeed, YopB alone can disturb artificial membranes, whereas YopD cannot. Moreover, YopD has been shown to end up in the cytosol of eukaryotic cells (25). YopB and YopD are encoded by a large operon that also encodes LcrV, LcrG, and the chaperone SycD. LcrV is a secreted Yop that has a different name for historical reasons. The fact that LcrG and LcrV are encoded together with translocators suggests that they could also be involved in translocation. Not surprisingly, LcrV interacts with YopB and YopD (26), is surfaceexposed before target cell contact (27), and is also required for translocation (26). In contrast with YopB, YopD, and LcrV, LcrG is not a released protein, but its exact localization in the bacterium remains elusive. It is required for efficient translocation of Yersinia Yop effector proteins into the eukaryotic cells but it is not required for pore formation. It binds to heparan sulfate proteoglycans (28), but the significance of this binding is not clear yet. The Cytosolic Chaperones. Type III secretion often involves a new type of small cytosolic chaperone (29, 30 and 31) (Fig. 1). In Yersinia, these chaperones are called “Syc” (for specific Yop chaperone) (31). Generally, they are encoded by a gene that is located close to the gene encoding the Yop protein they serve, and this is a useful indication to recognize them. These chaperones may not form a single homogeneous group but rather could belong to two



8780

different subfamilies, one devoted to effectors and one devoted to translocators. SycE, the chaperone of YopE, is the archetype of the first family (31). The other representatives of this family in Yersinia are SycH (30), SycT (32), and SycN (33, 34). They are small (14–15 kDa), acidic (pI 4.4–5.2) proteins with a putative C-terminal amphiphilic α-helix. They specifically bind to their cognate Yop and, in their absence, secretion of this Yop is severely reduced, if not abolished. Until now, research has focused mainly on SycE and SycH, but their exact roles remain elusive. Three hypotheses have been proposed, based on different types of observations. SycE and SycH have been shown to bind to their partner Yop (YopE and YopH) at a unique site spanning roughly residues 20–70 (35). Surprisingly, when this site is removed, the cognate Yop is still secreted and the chaperone becomes dispensable for secretion (36). This observation indicates that the binding site itself creates the need for the chaperone and suggests that the chaperone acts as a “bodyguard” protecting this site from premature associations that would lead to degradation. In agreement with this first hypothesis, SycE has been shown to protect YopE from intrabacterial degradation: the half-life of YopE is longer in wild-type bacteria than in sycE mutant bacteria (37, 38). The partners in these hypothetical premature associations could be the translocators (36), but such interactions could not be demonstrated. Moreover, the hypothesis of premature associations with translocators is not sufficient to explain the need for SycE. Indeed, YopE can be secreted by the plant pathogen Xanthomonas campestris (see below) and, although X. campestris does not synthesize proteins resembling the Yersinia translocators, SycE is still necessary to ensure intrabacterial stability of YopE in X. campestris (39). According to a second hypothesis, discussed below, SycE could act as a secretion pilot leading the YopE protein to the secretion locus. Finally, a recent observation suggests a third hypothesis. Both SycE and SycH are required for efficient translocation of their partner Yops into eukaryotic cells (35). However, when YopE is delivered by a Yersinia polymutant strain that synthesizes an intact secretion and translocation apparatus but no other effector than YopE, it appears that YopE is delivered even in the absence of its chaperone and chaperone-binding site (70). Thus, the SycE chaperone appears to be needed only when YopE competes with other Yops for delivery. This observation suggests that the Syc chaperones could be involved in some kind of hierarchy for delivery. This third hypothesis about the role of the Syc chaperones fits quite well with the observation that only a subset of the effectors seems to have a chaperone, but it still needs to be strengthened. Little is known about the role of SycT and SycN. However, there is an unexpected complexity for SycN in the sense that it requires YscB working as a cochaperone (34, 40). SycD is the archetype of the second group of “type III chaperones”. In its absence, translocators YopB and YopD are not secreted and they are less detectable inside the bacterial cell (30, 41). SycD appears to be different from SycE and SycH in the sense that it binds to several domains on YopB, reminiscent of SecB, a molecular chaperone in Escherichia coli that is dedicated to the export of proteins and has multiple binding sites on its targets (41). IpgC, the related chaperone from Shigella flexneri, can prevent the intrabacterial association between translocators IpaB and IpaC (29). The similarity between IpgC and SycD suggested that SycD could play a similar role and would thus prevent the intrabacterial association of YopB and YopD. However, this turned out not to be the case (41). Because YopB and YopD have also the capacity to bind to LcrV, one could speculate that SycD prevents the premature association, not between YopB and YopD but rather between YopB, YopD, and LcrV, but this possibility has not been shown yet. Recognition of the Transported Proteins. Effectors delivered by type III secretion systems have no classical cleaved N-terminal signal sequence (5). Instead, it was demonstrated in 1990 that Yops are recognized by their N terminus and that no sequence is cleaved off during Yop secretion (5). The minimal region shown to be sufficient for secretion was gradually reduced to 17 residues for YopH (35), to 15 residues for YopE (35), and to 15 residues for YopN (42). A systematic mutagenesis of the secretion signal by Anderson and Schneewind (42, 43) led to doubts about whether this signal was encoded in the protein. No point mutation could be identified that specifically abolished secretion of YopE, YopN, and YopQ. Moreover, some frameshift mutations that completely altered the peptide sequences of the YopE and YopN signals also failed to prevent secretion. Anderson and Schneewind (42, 43) concluded from these observations that the signal leading to the secretion of these Yops could be in the 5 end of the messenger RNA rather than in the peptide sequence. Secretion would thus be cotranslational, and translation of yop mRNA might be inhibited either by its own RNA structure or as a result of its binding to other regulatory elements. If this is correct, one would expect that no Yop could be detected inside bacteria. However, while this is reported to be true for YopQ (43), it is certainly not true for other Yops, such as YopE. To determine whether this N-terminal (or 5-terminal) signal is absolutely required for YopE secretion, Cheng et al. (37) deleted codons 2–15 and they observed that 10% of the hybrid proteins deprived of the N-terminal secretion signal were still secreted. They inferred that there is a second secretion signal and they showed that this second, and weaker, secretion signal corresponds to the SycE-binding site. Not surprisingly, this secretion signal is functional only in the presence of the SycE chaperone (37). Whether this signal plays a role in vivo remains to be elucidated. Control of the Injection. Yersinia secrete their Yops in vitro under conditions of Ca2+ deprivation. What is the triggering signal in vivo? Most probably contact with a eukaryotic cell. Several reports have shown that Yop delivery by Yersinia is a “directional” phenomenon in the sense that most of the load is delivered inside the eukaryotic cell and that there is little leakage (22). According to the assays used, there is some discrepancy on the degree of “directionality” (18), but there is no doubt that the bulk of the released Yops load ends up inside the eukaryotic cell, indicating that contact must be the signal. Pettersson et al. (44) provided a nice visual demonstration of the phenomenon. By expressing luciferase under the control of a yop promoter, they showed that active transcription of yop genes is limited to bacteria that are in close contact with eukaryotic cells. However, although contact is clearly the triggering event, it is not clear yet whether a specific receptor is involved. Pore formation in artificial membranes (24) tends to suggest that there is none.

EFFECTOR YOPS AND HOST RESPONSE The Array of Yop Effectors. Six effector Yops have been characterized: YopE, YopH, YopM, YopJ/YopP, YopO/YpkA, and YopT (Fig. 2). Only two of them have a known enzymatic activity: YopH is a powerful phosphotyrosine phosphatase resembling eukaryotic phosphatases. The catalytic activity is exerted by the C-terminal domain (200 residues), which contains a phosphate-binding loop including a critical cysteine residue (Cys-403) (45). YpkA/YopO is a serine-threonine kinase (46) which shows some similarity with the COT (Cancer Osaka Thyroid) oncogene product, a cytosolic serine/threonine protein kinase expressed in hematopoietic cells and implicated in signal transduction by growth factors. YpkA catalyzes auto



8781

phosphorylation of a serine residue in vitro. Infection of HeLa cells with a multiple yop mutant overproducing YpkA leads to a morphological alteration of the cells, different from those mediated by YopE and YopH. The cells round up but do not detach from the extracellular matrix. Inside the HeLa cells the YpkA protein is targeted to the inner surface of the plasma membrane (47). No target protein corresponding to YpkA/ YopO has been identified yet.

Fig. 2. Inhibition of phagocytosis by YopE, YopH, and YopT. (A) Phagocytosis of an invading bacterium by a macrophage. The process involves phosphorylation of focal adhesion proteins (p130cas, Fak, Fyn, paxillin) and actin polymerization controlled by GTPases such as RhoA and Rac. Phagocytosis is followed by killing of the bacterium. (B) Resistance to phagocytosis by Yersinia. On contact, Yersinia injects Yop effectors. YopH dephosphorylates proteins from the focal adhesion (PTPase, phosphotyrosine phosphatase); YopE inactivates Rac and cdc42 by stimulating their GTPase activity (GAP, GTPaseactivating protein); YopT deactivates RhoA. YopM is a strongly acidic protein containing leucine-rich repeats (LRRs) whose action and target remain unknown. It belongs to a growing family of type III effectors that has several representatives in Shigella (ipaH multigene family) and Salmonella (48). YopM has been shown to traffic to the cell's nucleus by means of a vesicle-associated pathway (49), but its action in the nucleus remains unknown. The Cytoskeleton Is a Target of YopE, YopH, and YopT. Three effectors, of six identified so far, exert a negative role on cytoskeleton dynamics and, by doing so, contribute to the strong resistance of Yersinia to phagocytosis by macrophages (ref.15; N. Grosdent and G.R.C., unpublished observations). Studies using HeLa cells have shown that YopH dephosphorylates p130cas, paxillin, and the focal adhesion kinase (FAK) (50, 51 and 52), leading to disruption of the focal adhesion and a reduced invasinmediated engulfment by HeLa cells (a phenomenon called “invasion”). YopH is specifically targeted to the focal complexes; residues 223–226, which are known to be surface-exposed, are involved in this process. Deletion of these targeting residues affects the anti-invasion effect. These observations also apply to phagocytosis by the J774 macrophage–monocyte cell line, at least in the absence of opsonization (53). In the latter cells, a catalytically inactive YopH coprecipitates not only with p130cas but also with FYB (54).

Fig. 3. Effects of YopP/YopJ. Bacterial lipopolysaccharide (LPS), bound to the LPS-binding protein (LBP), interacts with its receptor CD14 and coreceptor from the Toll-like family, which leads to phosphorylation cascades resulting in the activation of mitogen-activated protein kinases (MAPKs) and of the kinase of the inhibitor of NF-κB (IκB). Phosphorylation of IκB is followed by its degradation, and NF-κB migrates to the nucleus and activates transcription of proinflammatory cytokines, including TNF-α. Translocated YopP/YopJ prevents the activation of the two phosphorylation cascades, and thus blocks the release of TNF-α. YopP/YopJ also induces macrophage apoptosis. See text for details and references. YopE has been known for a long time to disrupt actin filaments (15, 55), but its exact target has not yet been identified. However, YopE is homologous to the N-terminal domain of SptP from Salmonella, another type III effector, and it has been shown recently that this N-terminal domain of SptP acts as a GTPase-activating protein (GAP) for Rac-1 and Cdc42 (56). It is thus likely that YopE exerts its negative effect on the dynamics of the cytoskeleton by exerting the same GAP activity. Finally, YopT exerts a dramatic depolymerizing effect on actin (32) by modifying RhoA, a GTPase that regulates the formation of stress fibers (57). The exact nature of the modification is not yet known. YopP/YopJ Down-Regulates the Inflammatory Response. As shown schematically in Fig. 3, YopP/YopJ is a key player in the downregulation of the inflammatory response that is observed during Yersinia infection. In vitro, YopP/YopJ has been shown to counteract the normal proinflammatory response of various cell types. Its injection reduces the release of TNF-α by macrophages (58) and of IL-8 by epithelial (59, 60) and endothelial (G.R.C. and S. Tötemeyer, unpublished results) cells. It also reduces the presentation of adhesion molecules ICAM-1 and Eselectin at the surface of endothelial cells (G.R.C. and S. Tötemeyer, unpublished results) and hence presumably reduces the recruitment of neutrophils to the sites of infection. All these events result from the inhibition of the activation of NF- B, a transcription factor known to be central in the onset of inflammation (59, 61). The inhibition of NF- B activation was recently shown to result from YopP/YopJ-mediated inhibition of IKKβ, a kinase that phosphorylates I B, the inhibitor of NF- B (62). By preventing



8782

phosphorylation of I B, YopP/YopJ prevents its degradation and the translocation of NF- B to the nucleus. The inhibition of NF- B activation is accompanied by a lack of activation of the mitogen-activated protein (MAP) kinases (MAPKs) c-Jun-N-terminal kinase (JNK), p38, and extracellular signalregulated kinase (ERK) 1 and 2 (58, 63, 64) that is observed upon infection of macrophages by a Yersinia producing YopP/ YopJ. Lack of activation of these MAPKs results from the inhibition of the upstream MAPK kinases (MAPKKs) by binding of YopP/YopJ (62). Last but not least, YopP/YopJ also induces apoptosis in macrophages (65, 66). This apoptosis is accompanied by cleavage of the cytosolic protein BID, the release of cytochrome c, and the cleavage of caspase-3 and -7 (C. Geuijen, W. Declerq, A. Boland, P. Vandenabeele, and G.R.C., unpublished results). The release of cytochrome c and the cleavage of BID can both be inhibited by caspase inhibitors, suggesting that YopP/YopJ interferes with a signaling pathway upstream of the mitochondria (C. Geuijen, W. Declerq, A. Boland, P. Vandenabeele, and G.R.C., unpublished results). The reduction in the release of TNF-α is not simply the consequence of apoptosis, because it occurs even when apoptosis is prevented by inhibiting the activity of caspases (61). On the contrary, apoptosis may result from the loss of the anti-apoptotic factor NF- B (61); however, this hypothesis still awaits demonstration. It is thus not yet clear whether YopP/YopJ causes apoptosis by activating a death mechanism or by inhibiting an NF- B-dependent survival mechanism. Interestingly, YopP/YopJ share a high level of similarity with AvrXv and AvrBsT from X. campestris and a protein from the nitrogen-fixing Rhizobium. Inhibition of Antigen-Specific T and B Lymphocytes Responses. While they colonize and multiply in Peyer's patches or lymph nodes, Yersinia must also encounter lymphocytes. Artificial in vitro systems demonstrated that B and T lymphocytes are indeed targets for Yersinia injections (ref.67; A. P. Boyd and G.R.C., unpublished results). Yao et al. (68) observed that T and B cells transiently exposed to Yersinia were impaired in their ability to be activated by means of their antigen receptors. T cells are inhibited in their ability to produce cytokines, and B cells are unable to up-regulate surface expression of the costimulatory molecule B7.2, in response to antigenic stimulation. This block of activation results from the inhibition of early phosphorylation events (68). Through the analysis of various mutants, YopH appeared to be the main effector involved in these events. Thus YopH not only contributes to the evasion of the innate immune response but it could also incapacitate the host adaptive immune response. I thank S. Bleves, G. Denecker, and C. Josenhans for a critical reading. Our work on Yersinia is supported by the Belgian Fonds National de la Recherche Scientifique Médicale (Convention 3.4595.97), the Direction Générale de la Recherche Scientifique-Communauté Française de Belgique (Action de Recherche Concertée 99/04–236), and the Interuniversity Poles of Attraction Program–Belgian State, Prime Minister's Office, Federal Office for Scientific, Technical and Cultural Affairs (PAI 4/03). 1. Simonet, M., Richard, S. & Berche P. (1990) Infect. Immun. 58, 841–845. 2. Nakajima, R. & Brubakker, R. R. (1993) Infect. Immun. 61, 23–31. 3. Cornelis, G. R., Boland, A., Boyd, A. P., Geuijen, C, Iriarte, M., Neyt, C. Sory, M.-P. & Stainier, I. (1998) Microbiol. Mol. Biol. Rev. 62, 1315–1352. 4. Hueck, C. J. (1998) Microbiol. Mol. Biol. Rev. 62, 379–433. 5. Michiels, T., Wattiau, P., Brasseur, R., Ruysschaert, J. M. & Cornelis, G. R. (1990) Infect. Immun. 58, 2840–2849. 6. Koster, M., Bitter, W., de Cock, H., Allaoui, A., Cornelis, G. R. & Tommassen, J. (1997) Mol. Microbiol. 26, 789–798. 7. Jackson, M. W. & Plano, G. V. (1999) J. Bacteriol. 181, 5126–5130. 8. Woestyn, S., Allaoui, A., Wattiau, P. & Cornelis G. R. (1994) J. Bacteriol. 176, 1561–1569. 9. Deng, W. L. & Huang, H. C. (1999) J. Bacteriol. 181, 2298–2301. 10. Payne, P. L. & Straley, S. C. (1998) J. Bacteriol. 180, 3882–3890. 11. Payne, P. L. & Straley, S. C. (1998) J. Bacteriol. 181, 2852–2862. 12. Minamino, T. & Macnab, R. M. (1999) J. Bacteriol. 181, 1388–1394. 13. Kubori, T., Matsushima, Y., Nakamura, D., Uralil, J., Lara-Tejero, M., Sukhan, A. & Galan, J. E. (1998) Science 280, 602–605. 14. Blocker, A., Gounon, P., Larquet, E., Niebuhr, K., Cabiaux, V., Parsot, C. & Sansonetti, P. (1999) J. Cell. Biol. 147, 683–693. 15. Rosqvist, R., Forsberg, A. & Wolf-Watz, H. (1991) Infect. Immun. 59, 4562–4569. 16. Rosqvist, R., Magnusson, K. E. & Wolf-Watz, H. (1994) EMBO J. 13, 964–972. 17. Sory, M.-P. & Cornelis, G. R. (1994) Mol. Microbiol. 14, 583–594. 18. Boland, A., Sory, M.-P., Iriarte, M., Kerbourch, C., Wattiau, P. & Cornelis, G. R. (1996) EMBO J. 15, 5191–5201. 19. Hakansson, S., Schesser, K., Persson, C., Galyov, E. E., Rosqvist, R., Homble, F. & Wolf-Watz, H. (1996) EMBO J. 15, 5812–5823. 20. Cornelis, G. R. & Wolf-Watz, H. (1997) Mol. Microbiol. 23, 861–867. 21. Starnbach, M. N. & Bevan, M. J. (1994) J. Immunol. 153, 1603–1612. 22. Persson, C., Nordfelth, R., Holmström, A., Hakansson, S., Rosqvist, R. & Wolf-Watz, H. (1995) Mol. Microbiol. 18, 135–150. 23. Neyt, C. & Cornelis, G. R. (1999) Mol. Microbiol. 33, 971–981. 24. Tardy, F., Homblé, F., Neyt, C., Wattiez, R., Cornelis, G. R., Ruysschaert, J.-M. & Cabiaux, V. (1999) EMBO J. 18, 6793–6799. 25. Francis, M. S. & Wolf-Watz, H. (1998) Mol. Microbiol. 29, 799–813. 26. Sarker, M. R., Neyt, C., Stainier, I. & Cornelis, G. R. (1998) J. Bacteriol. 180, 1207–1214. 27. Pettersson, J., Holmström, A., Hill, J., Leary, S., Frithz-Lindsten, E., von Euler-Matell, A., Carlsson E., Titball, R., Forsberg, A. & Wolf-Watz, H. (1999) Mol. Microbiol. 32, 961–976. 28. Boyd, A. P., Sory, M.-P., Iriarte, M. & Cornelis, G. R. (1998) Mol. Microbiol. 27, 425–436. 29. Ménard, R., Sansonetti, P. J., Parsot, C. & Vasselon, T. (1994) Cell 79, 515–525. 30. Wattiau, P., Bernier, B., Deslee, P., Michiels, T. & Cornelis, G. R. (1994) Proc. Natl. Acad. Sci. USA 91, 10493–10497. 31. Wattiau, P. & Cornelis, G. R. (1993) Mol. Microbiol. 8, 123–131. 32. Iriarte, M. & Cornelis, G. R. (1998) Mol. Microbiol. 29, 915–929. 33. Iriarte, M. & Cornelis, G. R. (1999) J. Bacteriol. 181, 675–680. 34. Day, J. B. & Plano, G. V. (1998) Mol. Microbiol. 30, 777–788. 35. Sory, M.-P., Boland, A., Lambermont, I. & Cornelis, G. R. (1995) Proc. Natl. Acad. Sci. USA 92, 11998–12002. 36. Woestyn, S., Sory, M.-P., Boland, A., Lequenne, O. & Cornelis, G. R. (1996) Mol. Microbiol. 20, 1261–1271. 37. Cheng, L. W., Anderson, D. M. & Schneewind, O. (1997) Mol. Microbiol. 24, 757–765. 38. Frithz-Lindsten, E., Rosqvist, R., Johansson, L. & Forsberg, A. (1995) Mol. Microbiol. 16, 635–647. 39. Rossier, O., Wengelnik, K., Hahn, K. & Bonas, U. (1999) Proc. Natl. Acad. Sci. USA 96, 9368–9373. 40. Jackson, M. W., Day, J. B. & Plano, G. V. (1998) J. Bacteriol. 180, 4912–4921. 41. Neyt, C. & Cornelis, G. R. (1999) Mol. Microbiol. 31, 143–156. 42. Anderson, D. M. & Schneewind, O. (1997) Science 278, 1140–1143. 43. Anderson, D. M. & Schneewind, O. (1999) Mol. Microbiol. 31, 1139–1148. 44. Pettersson, J., Nordfelth, R., Dubinina, E., Bergman, T., Gustafsson, M., Magnusson, M. & Wolf-Watz, H. (1996) Science 273, 1231–1233. 45. Guan, K. & Dixon, J. E. (1990) Science 249, 553–556. 46. Galyov, E. E., Hakansson, S., Forsberg, A. & Wolf-Watz, H. (1993) Nature (London) 361, 730–732. 47. Hakansson, S., Galyov, E. E., Rosqvist, R. & Wolf-Watz, H. (1996) Mol. Microbiol. 20, 593–603. 48. Miao, E. A., Scherer, C. A., Tsolis, R. M., Kingsley, R. A., Adams, G., Baumler, A. J. & Miller, S. I. (1999) Mol. Microbiol. 34, 850–864. 49. Skrzypek, E., Cowan, C. & Straley, S. C. (1998) Mol. Microbiol. 30, 1051–1065. 50. Black, D. S. & Bliska, J. B. (1997) EMBO J. 16, 2730–2744. 51. Persson, C, Carballeira, N., Wolf-Watz, H. & Fällman, M. (1997) EMBO J. 16, 2307–2318. 52. Black, D. S., Montagna, L. G., Zitsmann, S. & Bliska, J. B. (1998) Mol. Microbiol. 29, 1263–1274. 53. Persson, C., Nordfelth, R., Andersson, K., Forsberg, A., Wolf-Watz, H. & Fällman, M. (1999) Mol. Microbiol. 33, 828–838. 54. Hamid, N., Gustavsson, A., Andersson, K., McGee, K., Persson, C., Rudd, C. E. & Fallman, M. (1999) Microb. Pathog. 27, 231–242. 55. Rosqvist, R., Forsberg, A., Rimpilainen, M., Bergman, T. & Wolf-Watz, H. (1990) Mol. Microbiol. 4, 657–667. 56. Fu, Y. X. & Galan, J. E., (1999) Nature (London) 401, 293–297. 57. Zumbihl, R., Aepfelbacher, M., Andor, A., Jacobi, C. A., Ruckdeschel, H., Rouot, B. & Heeseman, J. (1999) J. Biol Chem. 274, 29289–29293. 58. Boland, A. & Cornelis, G. R. (1998) Infect. Immun. 66, 1878–1884.



8783

59. Schesser, K., Spiik, A.-K., Dukuzumuremyi, J.-M., Neurath, M. F., Pettersson, S. & Wolf-Watz, H. (1998) Mol. Microbiol. 28, 1067–1079. 60. Schulte, R., Wattiau, P., Hartland, E. L., Robins-Browne, R. M. & Cornelis, G. R. (1996) Infect Immun. 64, 2106–2113. 61. Ruckdeschel, K., Harb, S., Roggenkamp, A., Hornef, M., Zumbihl, R., Kohler, S., Heesemann, J. & Rouot, B. (1998) J. Exp. Med. 187, 1069–1079. 62. Orth, K., Palmer, L. E., Qin Bao, Z., Stewart, S., Rudolph, A. E, Bliska, J. B. & Dixon, J. E. (1999) Science 285, 1920–1923. 63. Ruckdeschel, K., Machold, J., Roggenkamp, A., Schubert, S., Pierre, J., Zumbihl, R., Liautard, J.-P., Heesemann, J. & Rouot, B. (1997) J. Biol. Chem. 272, 15920–15927. 64. Palmer, L. E., Pancetti, A. R., Greenberg, S. & Bliska, J. B. (1999) Infect. Immun. 67, 708–716. 65. Mills, S. D., Boland, A., Sory, M.-P., Van der Smissen, P., Kerbourch, C., Finlay, B. B. & Cornelis, G. R. (1997) Proc. Natl. Acad. Sci. USA 94, 12638– 12643. 66. Monack, D. M., Mecsas, J., Ghori, N. & Falkow, S. (1997) Proc. Natl. Acad. Sci. USA 94, 10385–10390. 67. Chaux, P., Luiten, R., Demotte, N., Vantomme, V., Stroobant, V., Traversari, C., Russo, V., Schultz, E., Cornelis, G. R., Boon, T. & van der Bruggen, P. (1999) J. Immunol. 163, 2928–2936. 68. Yao, T., Mecsas, J., Healy, J. I., Falkow, S. & Chien, Y.-H. (1999) J. Exp. Med. 190, 1343–1350. 69. Stainier, I., Bleves, S., Josenhans, C., Karmani, L., Kerbourch, C., Lambermont, I., Tötemeyer, S., Boyd, A. & Cornelis, G. R. (2000) Mol. Microbiol., in press. 70. Boyd, A. P., Lambermont, I. & Cornelis, G. R. (2000) J. Bacteriol., in press.


A FRAMEWORK FOR INTERPRETING THE LEUCINE-RICH REPEATS OF THE LISTERIA INTERNALINS

8784

Colloquium A framework for interpreting the leucine-rich repeats of the Listeria internalins Michael Marino*, Laurence Braun†, Pascale Cossart†, and Partho Ghosh*‡ * Department of Chemistry and Biochemistry, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0314; and † Institut Pasteur, Unité des Interactions Bactéries-Cellules, 28 rue du Dr. Roux, 75015 Paris, France The surface protein InlB of the bacterial pathogen Listeria monocytogenes is required for inducing phagocytosis in various nonphagocytic mammalian cell types in vitro. InlB causes tyrosine phosphorylation of host cell adaptor proteins, activation of phosphoinositide 3-kinase, and rearrangements of the actin cytoskeleton. These events lead to phagocytic uptake of the bacterium by the host cell. InlB belongs to the internalin family of Listeria proteins, which also includes InlA, another surface protein involved in host cell invasion. The internalins are the largest class of bacterial proteins containing leucine-rich repeats (LRR), a motif associated with protein– protein interactions. The LRR motif is found in a functionally diverse array of proteins, including those involved in the plant immune system and in the mammalian innate immune response. Structural and functional interpretations of the sequences of internalin family members are presented in light of the recently determined x-ray crystal structure of the InlB LRR domain. A number of intracellular pathogens have the ability to invade mammalian cells that are normally nonphagocytic by inducing phagocytic behavior (1). The cellular machinery required for phagocytosis appears to be present in many types of non-phagocytic cells but in a constitutively unassembled or inactive state. Certain pathogens possess means by which to induce assembly of the phagocytic machinery necessary to bring about their own uptake. Among these pathogens is the facultative intracellular bacterium Listeria monocytogenes, a cause of meningitis and abortion in humans (2). L. monocytogenes induces its own phagocytosis in a large number of nonphagocytic cell types in vitro through the actions of the bacterial surface proteins InlA and InlB (also called internalin and internalin B, respectively) (3). The related proteins InlA and InlB act on different sets of cell types. Although InlA is required for inducing phagocytosis in the enterocytelike epithelial cell line Caco-2 (4), InlB is required for inducing phagocytosis in some hepatocyte, endothelial, epithelial, or fibroblast-like cell lines (e.g., Vero, HeLa, HEp-2, Chinese hamster ovary) (5, 6, 7, 8, 9 and 10). The different cell type specificities likely reflect the fact that InlA and InlB have different receptor specificities. The receptor for InlA is E-cadherin (11), and, whereas the receptor for InlB has not been identified, it is clear that it is not E-cadherin. Nevertheless, InlB has been shown to have potent effects on host cell signaling events. InlB promotes activation of host cell phosphoinositide (PI) 3-kinase and tyrosine phosphorylation of certain adaptor proteins (Gab1, Shc, and Cbl) implicated in the membrane localization of phosphoinositide 3-kinase (PI 3-kinase) (12, 13). PI 3-kinase is known to have effects on actin polymerization (14), a process that is required for L. monocytogenes invasion of host cells (12). Reorganization of the actin cytoskeleton presumably leads to changes in the membrane that cause phagocytosis of the bacterium.

INTERNALIN FAMILY InlA and InlB belong to the internalin family of Listeria proteins, which includes at least seven additional members in L. monocytogenes (InlC, InlC2, InlD, InlE, InlF, InlG, and InlH) (4, 15, 16, 17 and 18). Internalin proteins are also found in Listeria ivanovii, a pathogen of ruminants (19, 20), and also in nonpathogenic species of the genus Listeria. Except for InlC, which is a soluble and secreted protein, the L. monocytogenes internalins have sequence motifs that are consistent with bacterial cell surface localization. In contrast to InlB, which has been shown to be important to proliferation in hepatocytes in vivo (21, 22), the role of InlA in virulence has not been clearly demonstrated (23, 24). The biochemical functions and exact virulence properties of the other L. monocytogenes internalins remain to be determined. Proteins of the internalin family share several sequence motifs, the most distinctive of these being leucine-rich repeats (LRR). The LRR motif is associated with protein–protein interactions and is found in both bacterial and eukaryotic proteins (25, 26 and 27). The majority of bacterial LRR proteins are found in Listeria and belong to the internalin family, but LRR proteins have also been identified in Yersinia spp. (28), Shigella flexneri (29), Salmonella typhimurium (30), and Burkholderia (Pseudomonas) solanacearum (31). Among eukaryotes, the LRR motif is found in proteins encoded by the disease resistance (R) genes of the plant immune system (32, 33 and 34) and by the toll and toll-like genes of Drosophila and mammals, which are involved in innate immune responses (35, 36). The LRR motif is not limited to virulence- or immunityrelated proteins but is instead found in a large group of proteins with highly diverse functions and cellular localizations (25). The LRR regions of InlA and InlB play critical roles in interactions with mammalian cells. The LRR region of InlA is necessary to promote invasion of permissive cells in vitro (37). Likewise, the LRR region of InlB is necessary for the activation of phosphoinositide 3-kinase, rearrangement of the actin cytoskeleton, and invasion of permissive cells in vitro (38). The recently determined x-ray crystal structure of the InlB LRR domain (39) provides a framework for the interpretation of sequences of internalin family members.

INTERNALIN LRR The internalin LRR regions are highly regular, with almost all repeats being 22 residues in length (Fig. 1). The LRRs are tandemly repeated from 6 times in InlG to 16 times in InlA. The 7.5 LRRs present in InlB give it its elongated and curved shape (Fig. 2). The repeats are similar to those observed in the LRRs of porcine and human ribonuclease inhibitors (RI) (40, 41 and 42), the U2LRR fragment of U2 small nuclear ribonucleoprotein (43), and the Ran GTP-ase activating protein rnalp (44). As in these structures, each repeat alternates between a short β-strand and an opposing antiparallel helical segment; the β-strands and helices are connected to each other by coils. The main chain wraps around in a right-handed sense, with the β-strands forming a parallel β-sheet and the helices stacking on each other, giving rise to the elongated shape of the LRR.

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviations: LRR, leucine-rich repeat; RI, ribonuclease inhibitor. ‡




8785

Fig. 1. Sequence alignments of L. monocytogenes internalin LRR regions. Stars above the sequences denote conserved internalin LRR residues, and bars denote the position of β-strands and extent of 310-helices observed in the InlB structure (39). Residues predicted to form the concave or convex face of these structures are enclosed in boxes. Hydrophobic, negatively charged, and positively charged residues predicted to be surface exposed are highlighted in yellow, red, and cyan, respectively. Sequences for InlA, InlB, InlC, InlC2, InlD, InlE, InlF, InlG, and InlH are shown. The repeats are 22 residues in length except for repeat 4 of InlC and repeat 6 of InlA, which are 21 residues, and repeat 5 of InlF, which is 23 residues. These small deletions and insertions are predicted to occur at loop regions. Although the β-strands are a highly conserved structural feature of LRR proteins, the helical segments are more divergent. The β-strands are almost always composed of three residues that are in precise register. On the other hand, the helical segments are variable in length, register, and type. In fact, one repeat of U2LRR contains a second β-strand instead of a helix. In InlB, the helical segments are short (three to five residues) and are exclusively composed of 310-helices. U2LRR possesses 23-and 25-residue repeats that also form 310-helices. This contrasts with RI and rnalp, which have long (10–14 residues) α-helices and also have longer repeat lengths (28 residues or greater). Apparently the more tightly wound 310-helix accommodates formation of short repeats more favorably than does the α-helix. The curvature of the LRR region arises from the β-strands of adjacent repeats being packed more closely together than opposing helices. The β-strands are within interrepeat hydrogen bonding distance, whereas the helices form intrarepeat hydrogen bonds. The β-strands form the concave face and the helices the



8786

convex face of the molecule (Fig. 2). This curvature would appear to limit the possible number of tandem LRRs, because a closed circle would be formed. However, the InlB β-strands are twisted and give the entire structure a right-handed superhelical twist, thereby placing no obvious limits on the possible number of tandem repeats (39). The β-strands of RI, U2LRR, and rnalp are not as extensively twisted and indeed place a limit on the number of possible tandem repeats. It is not clear at present what gives rise to this difference.

Fig. 2. Structure of the InlB LRR region (residues 77–242). The right-handed coil of the LRR alternates between β-strands and 310-helices. The β-strands form the concave face of the molecule and have a superhelical twist not observed in other LRR proteins. This figure and Fig. 4 were generated with MOLSCRIPT (48) and rendered with RASTER3D (49). The coil regions connecting β-strands and 310-helices constitute the great majority (14–17 residues) of the 22-residue InlB LRR. Even though these regions lack conventional secondary structure, they form highly regular structures (rms deviation of 0.32 Å in Cα positions). This regularity may in part be caused by the presence of water molecules that act as extensions of secondary structure. These waters are intercalated between repeats and form bridging hydrogen bonds between main-chain atoms of adjacent repeats. Structural determination of InlB to high resolution (1.86 Å) allowed visualization of three distinct spines of water molecules that run the length of the repeat region (39). Some of these hydrogen bond-mediating waters are also observed in the 2.0-Å resolution structure of human RI (42). A regular pattern of hydrogen bondmediating waters is expected to be present in other LRR proteins and to be observed in high-resolution structures of other LRR proteins.

LRR FLANKING SEQUENCES The LRR regions are flanked by sequences that are conserved in the internalin family and that most likely play a role in providing stability. At the N terminus of the LRR region is a segment of 40 residues termed the “N-terminal cap” in InlB, because it provides a hydrophilic cap on the hydrophobic core of the first LRR (39). The hydrophobic core would be exposed to solvent at either end of the LRR were it not for other portions of the protein. The N-terminal cap forms numerous polar and hydrophobic interactions with the first LRR. The most prominent of these is a hydrophobic interaction with Tyr at position 20 of the first LRR, a highly conserved residue in the internalin family (Fig. 1). In addition to its structural role, the N-terminal cap has been suggested to be involved in function based on the observation of two highly solvent-exposed calcium ions that bind to this region (39). The calciums are not required for formation of protein structure and may instead act as metal ion bridges between InlB and mammalian cell surface receptors or binding proteins. The N-terminal cap and LRR regions of InlB are sufficient to elicit mammalian cell effects and to induce phagocytosis (38). At the C terminus of the LRR region is a conserved sequence of 100 residues termed the interrepeat (IR) region (17). The IR region is dispensable for the mammalian cell signaling activities of InlB but is required in InlA for inducing phagoctyosis (37). The IR region has not yet been visualized, but crystals of intact InlB have been obtained that should elucidate the role of this as well as other regions of internalins (M.M. and P.G., unpublished data). At least a portion of the IR region is likely to provide a hydrophilic cap on the last LRR.

LRR PATTERN: STRUCTURAL RESIDUES A total of 10 positions of the 22-residue internalin LRR are highly conserved and serve structural rather than functional roles. Seven positions (2, 5, 7, 12, 15, 18, and 21) of the repeat contain mainly Leus or Iles that point inward and compose the hydrophobic core of the protein (Fig. 3). Other hydrophobic residues are accommodated at these positions, including Val, Phe, Met, and Ala (Fig. 1). Position 10 also faces inward and usually contains Asns or Glns, which form the so-called “Asn ladder” (45). The Asn hydrogen bonds to main-chain atoms in its own repeat and to those in the preceding repeat. This explains why position 10 in the first repeat is not constrained to be an Asn or Gln. The seven hydrophobic positions along with the Asn-ladder position define the internalin LRR sequence motif (Fig. 4). This motif corresponds almost identically to the 22-residue repeat

Fig. 3. Structure of a single InlB LRR. Positions 2, 5, 7, 10, 12, 14, 15, 17, 18, and 21 are conserved for structural reasons. The remaining positions are solvent exposed and variable. The register of the InlB LRR is that described for porcine RI (40).



8787

motif identified for proteins of the right-handed β-helix family (Fig. 4) (46, 47). This correspondence led to a suggestion that the internalins would form a β-helix structure rather than the observed LRR structure (47). The β-helix structure is found in pectate lyases and forms a distinctive L-shaped repeat containing three β-strands. The greatest difference between the two repeat motifs occurs at position 19, which faces outward and is variable in the internalin LRR but is occupied by an inward-facing conserved hydrophobic residue in the β-helix motif. The close correspondence of these two motifs suggests a means by which to probe the basis for their formation.

Fig. 4. Consensus motifs of the internalin LRR and the β-helix repeat. The two differ primarily at position 19. Black bars above the sequences represent β-strands, and the gray bar represents the 310-helix. Black arrows in the cartoons represent β-strands and the gray zigzag the 310-helix. The register for the β-helix repeat has been adjusted for purposes of comparison. Two structural positions of the InlB LRR are unique to the internalin family. Position 14 is outward facing and is usually occupied by an Asp that hydrogen bonds to main-chain atoms at the beginning of 310-helices (Fig. 3). Other residues capable of hydrogen bonding, including Ser, Thr, Asn, and Gln, are also found at this position (Fig. 1). Position 17 is generally occupied by a small amino acid, usually Gly, Pro, Ala, or Ser. It is located on the 310-helix, and a larger residue at this position would sterically clash with the preceding 310-helix. These two positions add to the eight inward-facing ones to yield the 10 positions conserved for structural reasons.

LRR PATTERN: FUNCTIONAL RESIDUES More than half the residues of the internalin LRR face outward, are variable, and can serve to define protein–protein interaction surfaces specific to each internalin. As with other LRR proteins, the interaction surfaces of the internalins are likely to be formed by the β-strand regions that constitute the concave faces of these molecules. The concave face has been observed to form the major binding surface in complexes of RI and U2LRR with their target proteins (41, 43) and is inferred to form the binding surface for rna1p (44). The sides of RI adjacent to the concave face are also involved in binding. In InlB, the concave face contains separate patches of hydrophobic and negatively charged residues that could constitute protein–protein interaction surfaces (39). Sequence characteristics of the internalins also point to the concave face as being important to protein–protein interactions. Three-quarters of the hydrophobic residues predicted to be surface exposed in the L. monocytogenes internalins are located on the concave face (positions 3, 4, 6, 8, and 9) (Fig. 1). Likewise, greater than three-quarters of the negatively charged residues predicted to be surface exposed are also located on the concave face (Fig. 1). In general, the concave faces of the internalins possess the hydrophobic and negative charge characteristics observed for InlB. The sides adjacent to the concave face (positions 11 and 13 on one side, and positions 21 and 1 on the other) are highly enriched in positively charged residues, as also observed in InlB. In contrast to the distinctive pattern of residues found on the concave face and sides, the convex face (positions 16, 19, and 20) of the internalins appears to be relatively featureless (Fig. 1). The exception to this is InlC, for which the convex face is predicted to contain a small hydrophobic patch. Although the high sequence conservation of the internalin LRR allows strong predictions about protein–protein interaction sites based on the InlB structure, it does not lead as easily to predictions about target specificity. The challenge for the future will be in identifying binding partners for individual internalins and in determining how the LRR structural motif is used to generate a variety of specific functional interactions. M.M. was supported by National Institutes of Health Training Grant GM07240-24. This work was in part supported by a W. M. Keck Distinguished Young Scholar award and an American Heart Association grant (9930135N) to P.G. 1. Finlay, B. B. & Cossart, P. (1997) Science 276, 718–725. 2. Lorber, B. (1997) Clin. Infect. Dis. 24, 1–9. 3. Cossart, P. & Lecuit, M. (1998) EMBO J. 17, 3797–3806. 4. Gaillard, J. L., Berche, P., Frehel, C., Gouin, E. & Cossart, P. (1991) Cell 65, 1127–1141. 5. Dramsi, S., Biswas, I., Maguin, E., Braun, L., Mastroeni, P. & Cossart, P. (1995) Mol. Microbiol. 16, 251–261. 6. Lingnau, A., Domann, E., Hudel, M., Bock, M., Nichterlein, T., Wehland, J. & Chakraborty, T. (1995) Infect. Immun. 63, 3896–3903. 7. Braun, L., Ohayon, H. & Cossart, P. (1998) Mol. Microbiol. 27, 1077–1087. 8. Greiffenberg, L., Goebel, W., Kim, K. S., Weiglein, I., Bubert, A., Engelbrecht, F., Stins, M. & Kuhn, M. (1998) Infect. Immun. 66, 5260–5267. 9. Müller, S., Hain, T., Pashalidis, P., Lingnau, A., Domann, E., Chakraborty, T. & Wehland, J. (1998) Infect. Immun. 66, 3128–3133. 10. Parida, S. K., Domann, E., Rohde, M., Muller, S., Darji, A., Hain, T., Wehland, J. & Chakrabory, T. (1998) Mol. Microbiol. 28, 81–93. 11. Mengaud, J., Ohayon, H., Gounon, P., Mege, R. M. & Cossart, P. (1996) Cell 84, 923–932. 12. Ireton, K., Payrastre, B., Chap, H., Ogawa, W., Sakaue, H., Kasuga, M. & Cossart, P. (1996) Science 274, 780–782. 13. Ireton, K., Payrastre, B. & Cossart, P. (1999) J. Biol. Chem. 274, 17025– 17032. 14. Carpenter, C. L. & Cantley, L. C. (1996) Curr. Opin. Cell Biol. 8, 153–158. 15. Engelbrecht, F., Chun, S. K., Ochs, C., Hess, J., Lottspeich, F., Goebel, W. & Sokolovic, Z. (1996) Mol. Microbiol. 21, 823–837. 16. Domann, E., Zechel, S., Lingnau, A., Hain, T., Darji, A., Nichterlein, T., Wehland, J. & Chakraborty, T. (1997) Infect. Immun. 65, 101–109. 17. Dramsi, S., Dehoux, P., Lebrun, M., Goossens, P. L. & Cossart, P. (1997) Infect. Immun. 65, 1615–1625. 18. Raffelsbauer, D., Bubert, A., Engelbrecht, F., Scheinpflug, J., Simm, A., Hess, J., Kaufmann, S. H. & Goebel, W. (1998) Mol. Gen. Genet. 260, 144–158. 19. Engelbrecht, F., Domainguez-Bernal, G., Hess, J., Dickneite, C., Greiffenberg, L., Lampidis, R., Raffelsbauer, D., Daniels, J. J., Kreft, J., Kaufmann, S. H., et al. (1998) Mol. Microbiol. 30, 405–417. 20. Engelbrecht, F., Dickneite, C., Lampidis, R., Götz, M., DasGupta, U. & Goebel, W. (1998) Mol. Gen. Genet. 257, 186–197. 21. Gaillard, J. L., Jaubert, F. & Berche, P. (1996) J. Exp. Med. 183, 359–369. 22. Gregory, S. H., Sagnimeni, A. J. & Wing, E. J. (1997) Infect. Immun. 65, 5137–5141. 23. Jonquières, R., Bierne, H., Mengaud, J. & Cossart, P. (1998) Infect. Immun. 66, 3420–3422. 24. Lecuit, M., Dramsi, S., Gottardi, C., Fedor-Chaiken, M., Gumbiner, B. & Cossart, P. (1999) EMBO J. 18, 3956–3963. 25. Kobe, B. & Deisenhofer, J. (1994) Trends Biochem. Sci. 19, 415–421. 26. Buchanan, S. G. & Gay, N. J. (1996) Prog. Biophys. Mol. Biol. 65, 1–44. 27. Kajava, A. V. (1998) J. Mol. Biol. 277, 519–527. 28. Leung, K. Y. & Straley, S. C. (1989) J. Bacteriol. 171, 4623–4632. 29. Hartman, A. B., Venkatesan, M., Oaks, E. V. & Buysse, J. M. (1990) J. Bacteriol. 172, 1905–1915. 30. Miao, E. A., Scherer, C. A., Tsolis, R. M., Kingsley, R. A., Adams, L. G., Bäumler, A. J. & Miller, S. I. (1999) Mol. Microbiol. 34, 850–864. 31. Van Gijsegem, F., Gough, C., Zischek, C., Niqueux, E., Arlat, M., Genin, S., Barberis, P., German, S., Castello, P. & Boucher, C. (1995) Mol. Microbiol. 15, 1095–1114.


A FRAMEWORK FOR INTERPRETING THE LEUCINE-RICH REPEATS OF THE LISTERIA INTERNALINS 32. Dixon, M. S., Jones, D. A., Keddie, J. S., Thomas, C. M., Harrison, K. & Jones, J. D. (1996) Cell 84, 451–459. 33. Boyes, D. C., Nam, J. & Dangl, J. L. (1998) Proc. Natl Acad. Sci. USA 95, 15849–15854. 34. Gassmann, W., Hinsch, M. E. & Staskawicz, B. J. (1999) Plant J. 20, 265–277. 35. Belvin, M. P. & Anderson, K. V. (1996) Annu. Rev. Cell. Dev. Biol. 12, 393–416. 36. Hoffmann, J. A., Kafatos, F. C., Janeway, C. A. & Ezekowitz, R. A. (1999) Science 284, 1313–1318. 37. Lecuit, M., Ohayon, H., Braun, L., Mengaud, J. & Cossart, P. (1997) Infect. Immun. 65, 5309–5319. 38. Braun, L., Nato, F., Payrastre, B., Mazié, J. C. & Cossart, P. (1999) Mol. Microbiol. 34, 10–23. 39. Marino, M., Braun, L., Cossart, P. & Ghosh, P. (1999) Mol. Cell 4, 1063–1072. 40. Kobe, B. & Deisenhofer, J. (1993) Nature (London) 366, 751–756. 41. Kobe, B. & Deisenhofer, J. (1995) Nature (London) 374, 183–186. 42. Papageorgiou, A. C., Shapiro, R. & Acharya, K. R. (1997) EMBO J. 16, 5162–5177. 43. Price, S. R., Evans, P. R. & Nagai, K. (1998) Nature (London) 394, 645–650. 44. Hillig, R. C., Renault, L., Vetter, I. R., Drell, T., IV, Wittinghofer, A. & Becker, J. (1999) Mol. Cell 3, 781–791. 45. Kobe, B. & Deisenhofer, J. (1995) Curr. Opin. Struct. Biol. 5, 409–416. 46. Yoder, M. D., Keen, N. T. & Jurnak, F. (1993) Science 260, 1503–1507. 47. Heffron, S., Moe, G. R., Sieber, V., Mengaud, J., Cossart, P., Vitali, J. & Jurnak, F. (1998) J. Struct. Biol. 122, 223–235. 48. Kraulis, P. (1991) J. Appl. Crystallogr. 24, 946–950. 49. Merrit, E. A. & Bacon, D. J. (1997) Methods Enzymol. 277, 505–524.

8788


ACYL-HOMOSERINE LACTONE QUORUM SENSING IN GRAM-NEGATIVE BACTERIA: A SIGNALING MECHANISM INVOLVED IN ASSOCIATIONS WITH HIGHER ORGANISMS

8789

Colloquium Acyl-homoserine lactone quorum sensing in Gram-negative bacteria: A signaling mechanism involved in associations with higher organisms Matthew R. Parsek* and E. Peter Greenberg†‡ * Department of Civil Engineering, Northwestern University, Evanston, IL 60208; and † Department of Microbiology, University of Iowa, Iowa City, IA 52242 Recent advances in studies of bacterial gene expression have brought the realization that cell-to-cell communication and community behavior are critical for successful interactions with higher organisms. Species-specific cell-to-cell communication is involved in successful pathogenic or symbiotic interactions of a variety of bacteria with plant and animal hosts. One type of cell-cell signaling is acylhomoserine lactone quorum sensing in Gram-negative bacteria. This type of quorum sensing represents a dedicated communication system that enables a given species to sense when it has reached a critical population density in a host, and to respond by activating expression of genes necessary for continued success in the host. Acyl-homoserine lactone signaling in the opportunistic animal and plant pathogen Pseudomonas aeruginosa is a model for the relationships among quorum sensing, pathogenesis, and community behavior. In the P. aeruginosa model, quorum sensing is required for normal biofilm maturation and for virulence. There are multiple quorumsensing circuits that control the expression of dozens of specific genes that represent potential virulence loci. bacterial signaling | biofilms | pathogenesis | Pseudomonas aeruginosa | virulence mechanisms I t was once held that most bacteria function only as individuals designed to compete with one another and to multiply rapidly under appropriate conditions. After all, dense cultures of bacteria can be grown from a single cell. This concept has given way to the view that, like other creatures, bacteria can communicate with each other and form communities that represent more than the sum of the individuals (for example, see refs.1, 2 and 3). It is now clear that many bacterial species use chemicals to signal each other and to coordinate their activities. Several different types of signals have now been described. Many Gram-positive bacteria use small peptides in signaling one another (4, 5), Gram-negative bacteria appear to use small molecule signals of various sorts (1, 6, 7, 8 and 9). Perhaps the best-studied signaling system is the acyl-homoserine lactone (acyl-HSL) system used by a large number of Gramnegative bacterial species. This type of bacterial cell-to-cell communication was discovered in the context of microbial ecology, but it is now evident that acyl-HSL signaling is important in plant and animal (including human) diseases. The signaling pathway is an enticing target for the development of antipathogenic therapies. As will be discussed below, acyl-HSL signaling represents a dedicated communication system that is used by bacteria to control specific genes in response to population density. Acyl-HSLs are small molecule signals with no other known function. These chemical signals are produced by specific enzymes, and they are detected by specific receptors. Because acyl-HSL signaling provides a mechanism by which a bacterial species can monitor its own population density, this type of signaling and other signaling systems that achieve the same purpose have been termed quorumsensing systems (10).

OVERVIEW OF ACYL-HSL QUORUM SENSING In most described cases, acyl-HSL signals are generated by the activity of a single enzyme that uses as substrates S-adenosylmethionine and an intermediate of fatty acid biosynthesis, acyl-acyl carrier protein (11, 12, 13, 14 and 15). The enzyme is generally a member of the LuxI family of acyl-HSL synthases. Different LuxI homologs generate different acyl-HSLs. Thus Pseudomonas aeruginosa RhlI primarily catalyzes the synthesis of N-butyryl-HSL (C4-HSL), and P. aeruginosa LasI directs the synthesis of N-(3-oxododecanoyl)-HSL (3OC12-HSL). The acyl sidechain length and the substitutions on the side chain provide signal specificity. Acyl side chains of these signals can be fully saturated, they can have hydroxyls or carbonyls on the third carbon, and they can have lengths of 4 to 16 carbons (ref.7; A. Schaefer and E.P.G., unpublished data). Short-chain signals such as C4-HSL diffuse freely through the cell membrane (16, 17), and 3OC12-HSL partitions into cells, presumably in the membrane. The 3OC12-HSL signal can diffuse into the surrounding environment but export is enhanced by the mexAB-oprM, and perhaps other, efflux pumps (17, 18). Regardless, the cellular concentration of an acyl-HSL is defined by the environmental concentration, and environmental concentrations can rise only when there is a sufficient population of the signal-producing bacterium. The specific receptors for acyl-HSL signals are members of the LuxR family of transcriptional regulators. LuxR family members have been proposed to consist of two domains, a C-terminal DNA-binding domain, and an N-terminal acyl-HSL-binding domain (for a review see ref.19). A simple model depicting an acyl-HSL quorum-sensing circuit is shown in Fig. 1. Quite often the two regulatory genes (the R and I genes) are linked, but not always. The orientation of the two genes with respect to each other is variable. Acyl-HSL quorum sensing is commonly found in Gramnegative bacteria that interact with plant and animal hosts. Quorum sensing was first discovered to control the luminescence of Vibrio fischeri, a bacterium that forms a mutualistic light organ symbiosis with certain marine animals (20, 21). Here quorum sensing is critical to the symbiosis. Acyl-HSL signaling is critical for virulence of the plant pathogen Erwinia carotovora (22) and for virulence of P. aeruginosa in mouse models of lung (23) and burn infections (24), in invertebrates (25, 26 and 27), and in plants (28). Thus acyl-HSL quorum sensing appears as a common theme in the interaction of several different bacterial species with eukaryotic hosts. We will describe the elements of quorum sensing, and discuss some of the factors controlled by quorum sensing in P. aeruginosa. In this paper P. aeruginosa will serve as a model for the role of bacterial communication in community behaviors important in pathogenesis.

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviations: acyl-HSL, acyl-homoserine lactone; C4-HSL, N-butyryl-HSL; 3OC12-HSL, N-(3-oxododecanoyl)-HSL. ‡




8790

Fig. 1. Generalized scheme for an acyl-HSL quorum-sensing circuit in a bacterial cell. The orange square indicates an acyl-HSL synthase-Luxl homolog. The diamonds are a LuxR homolog. Yellow diamonds on the bacterial chromosome are the LuxR homolog activated by the acyl-HSL signal. The arrows on the chromosome are qsc genes. The acyl-HSL (AHSL) signal can ). The diffuse in and out of cells. The compound in the box is an acyl-HSL (R1, H, OH, or O; R2, (CH2)2-14, or substrates for the acyl-HSL synthase are an acylated acyl carrier protein (Acyl-ACP) and S-adenosylmethionine (SAM). QUORUM SENSING IN P. AERUGINOSA P. aeruginosa can be isolated from soil and water. It is also an opportunistic pathogen of humans, other animals, and plants. One of the reasons P. aeruginosa is a successful opportunistic pathogen is that it produces a battery of secreted virulence factors. These virulence factors include exoproteases, siderophores, exotoxins, and lipases. Many of these virulence factors are regulated by quorum sensing (for reviews see refs.1, 29, and 30). Of what advantage to P. aeruginosa is quorum sensing control of virulence factors? First, it is economical to produce extracellular factors only after a critical population has been achieved. A mass of cells is required to produce sufficient quantities of these factors to influence the surrounding environment. Furthermore, in the host, timing of the deployment of virulence factors may be critical. The pathogen can amass without displaying its virulence factors, and then the pathogen can mount a surprise attack in which the arsenal of virulence factors is deployed in a coordinated and overwhelming fashion. Genetic studies have revealed two quorum-sensing systems in P. aeruginosa. Both of these systems, LasR-I and RhlR-I, have linked R and I genes. They are the quorum-sensing systems (31, 32, 33, 34, 35 and 36). In addition, the recently completed P. aeruginosa genome sequencing project has revealed a third LuxR homolog that is adjacent to a cluster of quorum-sensing-controlled (qsc) genes (37). However, a third LuxI homolog is not evident from the sequence, and the function of the third LuxR homolog is as yet unknown. LasR is a transcriptional regulator that responds primarily to the LasI-generated signal, 3OC12-HSL, and RhlR is a transcriptional regulator that responds best to the RhlI-generated signal, C4-HSL. The current model for quorum sensing in P. aeruginosa is as follows: at low population densities LasI produces a basal level of 3OC12-HSL. As density increases, 3OC12-HSL builds to a critical concentration, at which point it interacts with LasR. This LasR-3OC12-HSL complex then activates transcription of a number of genes. The list of target genes includes lasB, toxA, rhlR, and lasI (29, 32, 36, 38, 39). A curious fact is that different target genes are activated at different 3OC12-HSL concentrations (39). Activation of lasI by LasR creates a positive autoregulatory loop. The activation of rhlR by LasR results in a quorum-sensing regulatory cascade, in which activation of the rhl system requires an active las system. RhlR responds best to the RhlI-generated C4-HSL. RhlR then activates expression of genes required for production of a variety of secondary metabolites such as hydrogen cyanide and pyocyanin (for a review see ref.29). A DNA sequence with dyad symmetry called a lux-box-like sequence can easily be identified in the promoter regions of many quorum-sensing-controlled (qsc) genes (10, 37, 40, 41). By analogy to other acyl-HSL quorum-sensing systems we deduce that the lux-box-like sequences function as binding sites for LasR and RhlR. It is not yet clear how RhlR and LasR discriminate between their respective binding sites. In fact, many genes show partial activation with either LasR or RhlR and the appropriate acyl-HSL (for example see refs.30 and 37). One explanation for the partial activation or incomplete specificity is that binding site discrimination is less than perfect and either LasR or RhlR can bind with varying efficiency to any lux-box-like element. However, lux-box-like sequences are not apparent in the promoter regions of all qsc genes. This observation suggests that LasR or RhlR may also bind to identified sequences, or that some qsc genes are controlled by LasR or RhlR indirectly. As discussed above, many genes have been reported to come under the control of quorum sensing in P. aeruginosa. For some genes such as lasB there is a considerable amount of evidence in support of this conclusion (36, 38). For other genes, the data are limited, and in many cases the degree of transcriptional control reported is low. A recent study used a random mutagenesis approach to identify 39 genes that were highly regulated (minimum 5-fold induction, maximum 740-fold induction) by quorum sensing (37). The genes were divided into four different classes, two of which respond to 3OC12-HSL, and two of which required both C4-HSL and 3OC12-HSL for maximal induction. The qsc genes map throughout the P. aeruginosa chromosome (Fig. 2), confirming the view that quorum sensing in this bacterium represents a global regulatory system (29). The 39 genes revealed by the random mutagenesis study represent only a subset of the qsc genes in P. aeruginosa. It was estimated that as many as 4% of the roughly 6,000 P. aeruginosa genes are controlled by quorum sensing (37). One report indicates that transcription of rpoS, a gene encoding an RNA polymerase σ subunit involved in expression of stationary-phase factors, is activated by RhlR and C4-HSL (42). This finding raises the possibility that many genes may be controlled indirectly rather than directly by quorum sensing. It is also an enticing hypothesis because it lends itself to the idea that one specific cue that enables a cell to anticipate stationary phase is crowding. Unfortunately, quorum-sensing control of rpoS transcription is an example for which there is limited evidence. It is also an example for which there are low levels of induction (at best 3-fold). In fact, recent investigations suggest that quorum sensing may have no significant influence on rpoS transcription in P. aeruginosa (43).



8791

Fig. 2. Map of qsc genes or operons on the P. aeruginosa chromosome. Arrowheads indicate the direction of transcription. The different colors indicate different regulatory classes. Black, genes that respond primarily to 3OC12-HSL and that can respond to added signal in early logarithmic phase. Red, genes that respond primarily to 3OC12-HSL, but only after cultures have entered stationary phase (late genes). Blue, genes that respond best to both 3OC12-HSL and C4-HSL, and that can respond to added signal in early logarithmic phase. Green, late genes that require both signals for full expression. The lasR, lasl, and rhlR genes are shown in gold. The locations of lux-box-like elements are shown as black dots between the two DNA strands. These elements were identified in putative promoter regions of some but not all of the qsc genes. The numbers are distance in megabases (Mb) on the approximately 6.3-Mb chromosome (numbering starts at oriC). This figure is from ref.37. There are other control elements that affect the quorumsensing regulatory circuit. The gacA gene product is a transcriptional activator that among other things induces C4-HSL production (44). The rsaL gene is downstream of the lasR gene and it is involved in negatively regulating lasR (45). Vfr is a global regulator that affects a mild activation of lasR (46). The environmental signals that these regulators respond to are unknown. Furthermore, the physiological significance of the observed levels of regulation by these factors remains to be determined. Further complicating the quorum-sensing regulatory scheme is the recent discovery that a specific quinolone produced by P. aeruginosa can serve as an extracellular signal to activate lasB (8). The mechanisms that underlie quinolone signaling remain unknown.

Fig. 3. Diagram of the P. aeruginosa biofilm-maturation pathway. Unattached cells that approach a surface may attach. Attachment involves specific functions. Attached cells will proliferate on a surface and use specific functions to actively move into microcolonies. The high-density microcolonies differentiate into mature biofilms by a 3OC12-HSL-dependent mechanism. REGULATION OF VIRULENCE BY QUORUM SENSING IN P. AERUGINOSA Mutations in elements of the quorum-sensing machinery in P. aeruginosa do not markedly influence the growth of this bacterium in the laboratory. For example, the growth rate of the LasI−, RhlI− double mutant PAO-MW1 in Luria–Bertani broth at 30°C or 37°C is similar to that of the wild-type PAO1 under normal laboratory culture conditions. Yet quorum-sensing mutant strains show severe virulence defects in various mouse models, in invertebrate models, and in a plant model system. We will briefly describe the results of experiments in which colonization of the lungs in a neonatal mouse model by a LasR mutant was compared with colonization by the wild-type parental strain (23). When neonatal BALB/cByJ mice were inoculated intranasally with wild-type P. aeruginosa, the bacteria colonized the lung, causing an acute pneumonia, bacteremia, and death. In contrast, a LasR mutant strain could colonize the lung, but it did not achieve high densities and it did not cause pneumonia, bacteremia, or death. The mutant did not have a growth defect under laboratory conditions, it could invade the lung and survive but it could not cause disease.

BIOFILMS AND QUORUM SENSING Bacteria often tend to attach to surfaces and form communities enmeshed in a self-produced polymeric matrix. These communities are called a biofilm (2, 47). P. aeruginosa is often found in naturally occurring biofilms. Under the appropriate laboratory conditions, P. aeruginosa forms characteristic biofilms that can be several hundred micrometers thick (Fig. 3). Development of a mature biofilm proceeds through a programmed series of events (for a recent review see ref.2). After attachment, cells multiply to form a layer on a solid surface. Individuals in the layer then exhibit a surface motility called twitching. Twitching depends on type IV pili. As a result of twitching motility, small groups of P. aeruginosa called microcolonies form. Microcolonies then differentiate to form a mature biofilm. Microcolonies in a mature biofilm have towerand mushroom-shaped architectures. The cells in these structures are encased in an extracellular polysaccharide matrix. Water channels that allow the flow of nutrients into and waste products out of the biofilm innervate these structures. There is a significant physiological heterogeneity within biofilms. In P. aeruginosa biofilms there is a steep oxygen gradient. Oxygen is present at measurable concentrations mainly at the periphery of the biofilm. Oxygen microelectrode studies have also shown that water channels serve to bring oxygen to deeper areas of the biofilm. Similar gradients may be expected for pH and nutrients. These gradients dictate physiological variability among individual cells in the biofilm, with slower-growing cells present deeper within the biofilm and more actively growing cells at the periphery. This heterogeneity in physiological activity makes studying biofilms



8792

with traditional molecular microbiological techniques difficult. Bacteria in these mature biofilms are phenotypically resistant to microbicidal agents, including antibiotics. Thus biofilms cause many different types of chronic or persistent bacterial infections (for a recent review of biofilm infections and biofilm physiology see ref.2). Recent studies have linked quorum sensing and biofilm maturation (48). This is a particularly gratifying finding because quorum sensing functions to control gene expression in groups of bacteria, and biofilms are just that, organized groups of bacteria. A mutation in lasI has a dramatic affect on biofilm maturation. LasI mutants are incapable of 3OC12-HSL synthesis, and the development of LasI mutant biofilms is arrested after microcolony formation but before maturation of the microcolonies into thick structured assemblages. Thus LasI mutant biofilms appear flat and undifferentiated. The normal biofilms architecture can be restored to the mutant by addition of the LasI-generated quorum-sensing signal 3OC12-HSL. A RhlI mutant exhibits normal biofilm development and architecture. The 3OC12-HSL-responsive qsc genes involved in biofilm maturation remain unknown. Of interest, the LasI mutant biofilms were susceptible to treatment with the detergent SDS, whereas wildtype biofilms were resistant. LasR− mutants have biofilm phenotypes similar to that of LasI− mutants (Fig. 4). These observations suggests that antimicrobial therapies targeting the quorum-sensing mechanism of P. aeruginosa may result in the formation of abnormal biofilms that are more amenable to treatment. Another study showed that acyl-HSLs could be detected in clinical biofilm isolates and on catheters colonized by P. aeruginosa in mice (49). Some critical questions remain to be answered regarding the quorum-sensing mechanism in biofilms: What constitutes a quorum in a biofilm? Does a LasR–LasI, RhlR–RhlI regulatory cascade exist within a biofilm? Do acyl-HSL synthesis patterns change within a biofilm? Biofilm and quorum-sensing research has led us to appreciate the fact that P. aeruginosa can behave as a community. Because evidence supports the hypothesis that both biofilms and quorum sensing play integral roles in P. aeruginosa pathogenesis, more studies of the community activities of this bacterium are needed.

FUTURE CHALLENGES It is now generally appreciated that bacteria possess specific communication systems and that they are capable of organizing into functional communities. We have described one type of signaling system, an acyl-HSL system, in some detail in this article. Other signaling systems are known, but it seems apparent that many more remain to be discovered. Aside from the discovery of different types of chemical communication systems for intraspecies signaling, a challenge for the future is to begin to address the possibility that there is significant interspecies communication. The idea of interspecies communication is supported by a limited body of information. For example, we know that many different bacterial species make a signal to which Vibrio harveyi responds; however, we do not yet know the nature of the signal (for a recent review see ref.9). Because quorum sensing is required for virulence of P. aeruginosa and other bacteria, the quorum-sensing system is a target for development of new types of therapeutics, antipathogenic agents, agents that do not kill bacterial pathogens but that do interfere with their ability to cause infections. A challenge that faces us is to identify inhibitors of quorum sensing and test their effectiveness in the treatment of infections, particularly persistent biofilm infections. Another challenge is to better understand the network of genes regulated by quorum sensing, and to identify qsc genes involved in normal biofilm maturation and infection by organisms such as P. aeruginosa.

Fig. 4. Scanning confocal microscope images of a mature P. aeruginosa wild-type biofilm (Upper) and a quorum-sensing mutant biofilm (Lower). In this case the quorum-sensing mutant was a lasR, rhlR double mutant. The perspective is from above the biofilm on a glass surface. The glass surface is red, and the green is from the green fluorescent protein encoded by the gfp gene in the recombinant P. aeruginosa. The wild-type biofilm consists of thick microcolonies. The immature mutant biofilm appears thinner, and more of the glass surface is exposed. With the lasR, rhlR mutant shown here (but not with lasI, rhlI mutants) zones of clearing around microcolony towers are often observed. Other experiments have shown that these zones are filled with extracellular polysaccharide (M.R.P., unpublished data). The biofilms are in flow-through reaction vessels similar to those described in ref. 48. The colors were applied to the image by computer enhancement with Adobe PHOTOSHOP 5.0. The black marker bar is 100 µm in length. We thank M. Hentzer, A. Heydorn, M. Givskov, and S. Mølin for help with the Biofilm experiments. Work by the authors was supported by grants from the National Institutes of Health (GM59026), the National Science Foundation (MCB 9808308), and the Cystic Fibrosis Foundation. 1. Fuqua, C. & Greenberg, E. P. (1998) Curr. Opin. Microbiol. 1, 183–189. 2. Costerton, J. W., Stewart, P. & Greenberg, E. (1999) Science 284, 1318–1322. 3. Salmond, G. P. C., Bycroft, B. W., Stewart, G. S. A. B. & Williams, P. (1995) Mol. Microbiol. 16, 615–624. 4. Dunny, G. M. & Leonard, B. A. (1997) Annu. Rev. Microbiol. 51, 527–564. 5. Novick, R. P. & Muir, T. W. (1999) Curr. Opin. Microbiol. 2, 40–45. 6. Fuqua, W. C. & Winans, S. C. (1994) J. Bacteriol. 176, 2796–2806. 7. Fuqua, W. C., Winans, S. C. & Greenberg, E. P. (1996) Annu. Rev. Microbiol. 50, 727–751. 8. Pesci, E. C., Milbank, J. B., Pearson, J. P., Kende, A. S., Greenberg, E. P. & Iglewski, B. H. (1999) Proc. Natl. Acad. Sci. USA 96, 11229–11234. 9. Bassler, B. L. (1999) Curr. Opin. Microbiol. 2, 582–587. 10. Fuqua, W. C, Winans, S. C. & Greenberg, E. P. (1994) J. Bacteriol. 176, 269–275. 11. Schaefer, A. L., Val, D. L., Hanzelka, B. L., Cronan, J. E., Jr., & Greenberg, E. P. (1996) Proc. Natl. Acad. Sci. USA 93, 9505–9509.



8793

12. Parsek, M. R., Val, D. L., Hanzelka, B. L., Cronan, J. E., Jr., & Greenberg, E. P. (1999) Proc. Natl. Acad. Sci. USA 96, 4360–4365. 13. Moré, M. I., Finger, D., Stryker, J. L., Fuqua, C., Eberhard, A. & Winans, S. C. (1996) Science 272, 1655–1658. 14. Hanzelka, B., Parsek, M. R., Val, D. L., Dunlap, P. V., Cronan, J. E., Jr., & Greenberg, E. P. (1999) J. Bacteriol. 181, 5766–5770. 15. Hoang, T. T., Ma, Y., Stern, R. J., McNeil, M. R. & Schweizer, H. P. (1999) Gene 237, 361–371. 16. Kaplan, H. B. & Greenberg, E. P. (1985) J. Bacteriol. 163, 1210–1214. 17. Pearson, J. P., Van Delden, C. & Iglewski, B. H. (1999) J. Bacteriol. 181, 1203–1210. 18. Evans, K., Passador, L., Srikumar, R., Tsang, E., Nezezou, J. & Poole, K. (1998) J. Bacteriol. 180, 5443–5447. 19. Stevens, A. M. & Greenberg, E. P. (1998) in Cell-Cell Signaling in Bacteria, eds. Dunny, G. & Winans, S. C. (Am. Soc. Microbiol., Washington, DC), pp. 231–242. 20. Nealson, K. H. & Hastings, J. W. (1979) Microbiol. Rev. 43, 469–518. 21. Ruby, E. G. (1996) Annu. Rev. Microbiol. 50, 591–624. 22. Pirhonnen, M., Flego, D., Heikiheimo, R. & Palva, E. T. (1993) EMBO J. 12, 2467–2476. 23. Tang, H. B., DiMango, E., Bryan, R., Gambello, M., Iglewski, B. H., Goldberg, J. B. & Prince, A. (1996) Infect. Immun. 64, 37–43. 24. Rumbaugh, K. P., Griswold, J. A. & Hamood, A. N. (1999) J. Burn Care Rehabil. 20, 42–49. 25. Tan, M. W., Rahme, L. G., Sternberg, J. A., Tompkins, R. G. & Ausubel, F. M. (1999) Proc. Natl. Acad. Sci. USA 96, 2408–2413. 26. Tan, M. W., Mahajan-Miklos, S. & Ausubel, F. M. (1999) Proc. Natl. Acad. Sci. USA 96, 715–720. 27. Mahajan-Miklos, S., Tan, M. W., Rahme, L. G. & Ausubel, F. M. (1999) Cell 96, 47–56. 28. Rahme, L. G., Stevens, E. J., Wolfort, S. F., Shao, J., Tompkins, R. G. & Ausubel, F. M. (1995) Science 268, 1899–1902. 29. Pesci, E. C. & Iglewski, B. H. (1997) Trends Microbiol. 5, 132–135. 30. Van Delden, C. & Iglewski, B. H. (1998) Emerg. Infect. Dis. 4, 551–560. 31. Brint, J. M. & Ohman, D. E. (1995) J. Bacteriol. 177, 7155–7163. 32. Gambello, M. J., Kaye, S. & Iglewski, B. H. (1993) Infect. Immun. 61, 1180–1184. 33. Latifi, A., Winson, K. M., Foglino, M., Bycroft, B. W., Stewart, G. S. A. B., Lazdunski, A. & Williams, P. (1995) Mol. Microbiol. 17, 333–344. 34. Ochsner, U. A., Koch, A. K., Fiechter, A. & Reiser, J. (1994) J. Bacteriol. 176, 2044–2054. 35. Ochsner, U. A. & Reiser, J. (1995) Proc. Natl. Acad. Sci. USA 92, 6424–6428. 36. Passador, L., Cook, J. M., Gambello, M. J., Rust, L. & Iglewski, B. H. (1993) Science 260, 1127–1130. 37. Whiteley, M., Lee, K. M. & Greenberg, E. P. (1999) Proc. Natl. Acad. Sci. USA 96, 13904–13909. 38. Gambello, M. J. & Iglewski, B. H. (1991) J. Bacteriol 173, 3000–3009. 39. Seed, P. C., Passador, L. & Iglewski, B. H. (1995) J. Bacteriol. 177, 654–659. 40. Pearson, J. P., Pesci, E. C. & Iglewski, B. H. (1997) J. Bacteriol. 179, 5756–5767. 41. Pesci, E. C., Pearson, J. P., Seed, P. C. & Iglewski, B. H. (1997) J. Bacteriol. 179, 3127–3132. 42. Latifi, A., Foglino, M., Tanaka, K., Williams, P. & Lazdunski, A. (1996) Mol. Microbiol 21, 1137–1146. 43. Whiteley, M., Parsek, M. & Greenberg, E. P. (2000) J. Bacteriol. 182, in press. 44. Reimmann, C., Beyeler, M., Latifi, A., Winteler, H., Foglino, M., Lazdunski, A. & Hass, D. (1997) Mol. Microbiol. 24, 309–319. 45. de Kievit, T., Seed, P. C., Nezezon, J., Passador, L. & Iglewski, B. H. (1999) J. Bacteriol. 181, 2175–2184. 46. Albus, A. M., Pesci, E. C., Runyen-Janecky, L. J., West, S. E. & Iglewski, B. H. (1997) J. Bacteriol. 179, 3928–3935. 47. Costerton, J. W., Lewandowski, Z., Caldwell, D. E., Korber, D. R. & Lappin-Scott, H. M. (1995) Annu. Rev. Microbiol. 49, 711–745. 48. Davies, D. G., Parsek, M. R., Pearson, J. P., Iglewski, B. H., Costerton, J. W. & Greenberg, E. P. (1998) Science 280, 295–298. 49. Stickler, D. J., Morris, N. S., McLean, R. J. & Fuqua, C. (1998) Appl. Environ. Microbiol. 64, 3486–3490.


PHENOTYPIC VARIATION AND INTRACELLULAR PARASITISM BY HISTOPLASMA CAPSULATUM

8794

Colloquium Phenotypic variation and intracellular parasitism by Histoplasma capsulatum Silke Kügler*, Tricia Schurtz Sebghati*, Linda Groppe Eissenberg*, and William E. Goldman† Department of Molecular Microbiology, Washington University School of Medicine, St. Louis, MO 63110 The success of Histoplasma capsulatum as an intracellular pathogen depends completely on successful conversion of the saprophytic mycelial (mold) form of this fungus to a parasitic yeast form. It is therefore not surprising that yeast phase-specific genes and gene products are proving to be important for survival and proliferation of H. capsulatum within macrophages. In this study, we have focused on the role and regulation of two yeast-specific characteristics: α-(1,3)-glucan, a cell wall polysaccharide modulated by cell-density (quorum) sensing, and a secreted calcium-binding protein (CBP) that is essential for pathogenicity. H istoplasma capsulatum is the best studied of the dimorphic fungal pathogens, all of which undergo a reversible morphological variation that is tightly linked to their lifestyle and pathogenesis. The normal home for these fungi is the soil, where they exist in a mycelial (mold) form. Humans and other mammals inhale aerosolized conidia and hyphal fragments, prompting a dramatic conversion of the mycelial form to budding yeasts. The phenotypic variation between mycelial and yeast forms corresponds to a complete switch in lifestyle, from a saprophytic soildwelling fungus to a parasitic form that is highly adapted for growth at higher temperature (37°C) and for avoidance of host defense mechanisms. In fact, H. capsulatum thrives within the normally harsh environment of the phagolysosomes of macrophages (1), the primary host cell type. In this regard, this yeast is unique from other intracellular pathogens in its ability to prevent the acidification of the phagolysosome (2), presumably limiting the antifungal effectiveness of this compartment. Little else is known about how H. capsulatum survives and proliferates inside macrophages, but clues to these strategies have come from studying characteristics that are specific to the yeast form. Because the yeast cell is functionally dedicated to intracellular parasitism, it is likely that identification and characterization of genes that are exclusively expressed in the yeast phase will lead to a better understanding of H. capsulatum virulence. Our laboratory has focused on two yeast phase-specific characteristics: an unusual cell wall polysaccharide that is intimately linked to strain-specific pathogenicity and a secreted calcium-binding protein (CBP) that has become, to our knowledge, the first formally defined virulence determinant of H. capsulatum.

MODULATION OF -(1,3)-GLUCAN IN THE CELL WALL Many virulent strains of H. capsulatum possess α-(1,3)-glucan in the cell wall of the yeast form, although it is absent in the mycelial form. This polysaccharide is likewise present in the yeast phase of two other pathogenic dimorphic fungi, Paracoccidiodes brasiliensis (3) and Blastomyces dermatitidis (4). In each of these species, spontaneous variants that have lost their α-(1,3)-glucan have also lost virulence (3, 5, 6). Whether this polysaccharide directly influences virulence is unknown; its absence may simply alter cell wall architecture, permeability, or secretion. Regardless, its presence correlates with strain-specific virulence in a mouse model of infection and influences the dynamics of intracellular survival and proliferation (5, 7). Strains of H. capsulatum regulate α-(1,3)-glucan production such that it seems to be constitutively synthesized when the yeasts are proliferating inside macrophages. This regulation is in sharp contrast to the modulation of this phenotype during broth culture growth. When yeasts from a dense culture were washed and then diluted to a low density in standard growth medium, only an estimated 30% of the yeasts had α(1,3)-glucan detectable in their cell walls 24 h later (Fig. 1). The percentage remains low until the culture begins to enter stationary phase, when α-(1,3)-glucan is once again detectable in the wall of virtually every yeast cell. This phenotypic variation is a consistent feature of broth culture growth, with one important exception. When yeasts from a dense culture were diluted into medium that contained filtrate from a stationary phase culture, most yeasts remained positive for α-(1,3)-glucan 24 h later (data not shown). They tested positive even when the filtrate comprised as little as 4% of the new medium, negating the possibility that nutrient deprivation was responsible for blocking this modulation. These results suggest that H. capsulatum yeast cells release a factor that, when present in sufficient concentration, promotes α-(1,3)-glucan incorporation into the cell wall. Growth-dependent modulation of α-(1,3)-glucan production resembles the “quorum-sensing” phenomenon seen in many bacteria (8, 9 and 10). In those systems, microbes release an “autoinducer” at a constant rate, such that its concentration in the external environment is directly proportional to the cell density. Eventually, a sensor molecule detects a critical concentration of the autoinducer and signals a transcriptionally regulated phenotypic variation. Among bacteria, the autoinducer is usually an acyl-homoserine lactone (10) or a peptide (9), small mediators that are diffusible across membranes. Our initial characterizations suggest that the cell density autoinducer released by H. capsulatum is a larger molecule (molecular weight 6,000, based on dialysis experiments) and therefore unlikely to diffuse across membranes. It is tempting to speculate about a regulatory role for this system during intracellular parasitism; a high intraphagosomal concentration of the autoinducer could signal α-(1,3)-glucan synthesis to remain active, similar to the regulation in vitro when broth cultures become dense. Histoplasma may therefore be the first characterized example of an organism that uses a quorum-sensing mechanism to detect that it is within a host cell compartment.

CALCIUM AND INTRACELLULAR PARASITISM Another Histoplasma yeast phase-specific phenotype is the production of CBP, a secreted calcium-binding protein. Secretion of CBP is correlated not only with yeast/mycelial morphology but also with dependence on calcium for growth; H. capsulatum in the yeast phase is capable of growing in the presence of high levels of EGTA, whereas mycelial growth in limiting calcium is inhibited (11). This dependence may reflect a general problem faced by intracellular pathogens, because both Salmonella and Toxoplasma have specific responses to low calcium conditions that are duplicated during host cell parasitism (12, 13). Still, defining the role of CBP in Histoplasma virulence will depend on more than correlative and circumstantial evidence, and our efforts have focused on developing a formal molecular genetic proof.

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. * S.K., T.S.S., and L.G.E. contributed equally to this work. † To whom reprint requests should be addressed. E-mail: [email protected].



8795

Fig. 1. Growth-dependent modulation of α-(1,3)-glucan in H. capsulatum yeast cell walls. H. capsulatum G186AR yeasts were washed and used as an inoculum (Left) for a culture initiated at low density (2 × 106 cells per ml). At 24 h (Center) and 96 h (Right) after inoculation, an aliquot of yeasts was removed and monitored for α-(1,3)-glucan by comparing immunofluorescence (Upper) and differential interference contrast (Lower) images (magnification ×1,700). Murine monoclonal IgM antibody MOPC104E was used to detect α-(1,3)-glucan, and FITC-tagged goat anti-murine IgM was used as a fluorescent secondary antibody. In many fungi, the ability to evaluate protein function by gene disruption strategies is complicated by “illegitimate” recombination events. Simple allelic replacement is therefore difficult to detect among the high background of ectopic insertions, and marker-based selection strategies become complicated by duplications, rearrangements, and deletions that can accompany the desired recombination event. For molecular genetics in H. capsulatum, we have developed a range of strategies that are based on transformation with linear telomeric plasmids (14, 15). This genetic system has been adapted most recently as an allelic replacement tool, allowing us to disrupt the CBP1 locus in H. capsulatum and test its role in calcium acquisition and virulence. To disrupt CBP1 in a virulent strain of H. capsulatum (G186AR), a linear telomeric plasmid was designed with several unique features to enrich for homologous recombination events. Most importantly, inverted telomeric repeats at each end of the linear plasmid help maintain the plasmid extrachromosomally and nearly eliminate ectopic integration events. For selection and disruption, the linear plasmid contains a URA5 gene, and the CBP1 gene has an internal fragment replaced by a hygromycin resistance marker (hph). In addition, over 5 kilobases of flanking DNA (upstream and downstream of the CBP1 coding sequence) was included to increase the frequency of the desired double crossover event. This construct was transformed into a uracil auxotroph of H. capsulatum G186AR, and transformants were selected initially as uracil prototrophs. Cultures were then grown in the presence of 5-fluoroorotic acid (selecting against URA5 on the plasmid vector) and hygromycin (selecting for retention of the disrupted CBP1 gene). After this two-step selection strategy, cbp1-null mutants were isolated at high frequency, and their genotypes were confirmed by PCR, Southern analysis, and protein gels (data not shown). Two cbp1-null isolates were first tested for their ability to grow in the presence of EGTA. The growth of these knockout strains was inhibited in medium containing EGTA at a concentration as low as 150 µM (Fig. 2). (Normal growth medium used for Histoplasma contains 300 µM calcium.) In contrast, wild-type H. capsulatum strains were able to grow in medium containing greater than 1 mM EGTA. CBP1 expression was restored in one of the disrupted strains by transformation with another telomeric plasmid carrying wild-type CBP1. Complementing the knockout with CBP1 restores growth in calciumlimited medium (Fig. 2). To determine the virulence of the CBP knockout strain, we evaluated its interaction with a macrophage-like cell line (P388D1 cells). The ability of yeast strains to kill P388D1 cells has been correlated previously with virulence as measured in a mouse model of histoplasmosis (7). When tested in this in vitro virulence model, the cbp1-null strain could be seen inside of the macrophages but was unable to destroy the macrophage monolayer. When the knockout strain was complemented with CBP1 in trans, the macrophages were killed as efficiently as they were by the wild-type virulent strain (Fig. 3). These results describe the first successful disruption of a virulence determinant in H. capsulatum. The actual function of CBP in vivo still needs to be elucidated. CBP may act in a siderophore-like manner, binding calcium ions and providing



8796

yeasts with calcium needed for growth. Alternatively, CBP may bind calcium to modulate the phagolysosomal environment such that yeast proliferation is possible.

Fig. 2. CBP and calcium dependency of H. capsulatum growth. Cultures of H. capsulatum were grown for 4 days in normal medium or in medium with EGTA added to chelate calcium. Cumulative growth was measured (by absorbance at 600 nm) for the cbp1-null strain (cbp1::hph), the trans-complemented strain cbp1::hph(CBP1), and the wild-type parental strain G186AR. Each bar represents the mean for triplicate samples, with error bars indicating standard deviations from the mean. Growth inhibition by EGTA could be reversed by adding equimolar calcium chloride to the cultures (data not shown). REGULATION OF CBP CBP1 is regulated at the level of transcription, and we have shown previously by Northern blot analysis that mRNA is present in the yeast phase but not detectable in the mycelial phase (16). Further studies with reverse transcription–PCR demonstrated that CBP1 continues to be expressed while H. capsulatum yeasts grow inside P388D1 cells. This expression is consistent with experiments with mice inoculated with H. capsulatum; splenocytes harvested later from these mice respond to purified CBP in proliferation assays, implying that CBP is secreted during infection of mammalian hosts (17). All of these regulatory studies have used either mycelial or yeast cultures of H. capsulatum, but there has been no means to examine carefully the developmental programming of CBP1 expression during the transition between these two forms. For this purpose, we have developed a reporter system with a synthetic gene encoding the green fluorescent protein (gfp). A promoter fusion was constructed, consisting of 1.3 kilobases of the 5 untranslated region of CBP1, including the start codon ATG, fused to the gfp gene (0.7 kilobases). By using the telomeric plasmid system, constructs containing different modified gfp sequences were tested for fluorescence in H. capsulatum. One human codon-optimized version of gfp with the amino acid exchanges F64L, S65T, and H231L (18) yielded brightly fluorescent yeast cells. As expected, mycelial cells showed only very weak or no fluorescence, consistent with the CBP phenotype and with Northern blot analysis of CBP1 expression.

Fig. 3. Role of CBP in parasitism of macrophages. P388D1 macrophage-like cells were inoculated with strains of H. capsulatum at a multiplicity of one yeast per five macrophages and then cocultured for 5 days as described (7). (A) cbp1-null strain cbp1::hph; (B) trans-complemented strain cbp1::hph(CBP1); and (C) wild-type parental strain G186AR (magnification × 800). To follow the regulation of CBP1 during the transition from yeast cells to mycelia, a similar plasmid construct was randomly integrated into the chromosome to stabilize gfp copy number from cell to cell. Expression of gfp was monitored by fluorescence microscopy after placing a yeast culture at 25°C to trigger transition to the mycelial form (Fig. 4). During the first 24 h, the yeast cells and their initial hyphal extensions remained brightly fluorescent. By the next day, fluorescence of the yeast cells was reduced, and the longer hyphae had almost no detectable fluorescence. At 72 h after the temperature shift, hardly any fluorescence was visible, although the residual fluorescence was localized only to yeast cells. These results demonstrate that CBP1 is not simply temperature regulated, because fluorescence from preexisting CBP-GFP would have been obviously reduced at 24 h (particularly in the hyphal extensions). Instead, it seems that CBP1 is developmentally



Fig. 4. Regulation of CBP1 during transition from yeast to mycelial forms. H. capsulatum strain G186ASura5 carrying the CBP1 promoter-gfp fusion was grown as yeast at 37°C and then transferred to 25°C at time 0 h. (A) Differential interference contrast microscopy, (B) Fluorescence microscopy (magnification ×1,200).

8797



8798

down-regulated as the morphological conversion to the mycelial form progresses.

HISTOPLASMA AS A MODEL SYSTEM Among the fungi, only a few primary pathogens cause systemic mycoses, and among those, H. capsulatum has received the most research attention in terms of biology, biochemistry, and molecular genetics. Historically, the mycelia–yeast transition and the ability of yeasts to destroy macrophages have been the areas of primary focus. With the development of molecular genetic tools such as freely replicating plasmids, reporter genes, and gene disruption strategies, it is now possible to probe this organism's fascinating biology with genetic precision and functional proof. It is also likely that many of the lessons learned from H. capsulatum will be applicable to the other dimorphic fungal pathogens, most of which are closely related and cause similar clinical syndromes. The existence of a morphological form dedicated to parasitism provides a highly visible target in the hunt for genes and regulatory mechanisms that are most likely to be involved in fungal pathogenesis. This work was supported by Public Health Service Grants AI25584 (to W.E.G.), AI07172 (to Washington University), and HL07317 (to Washington University). W.E.G. is a recipient of the Burroughs–Wellcome Fund Scholar Award in Molecular Pathogenic Mycology. 1. Eissenberg, L. G., Schlesinger, P. H. & Goldman, W. E. (1988) J. Leukocyte Biol. 43, 483–491. 2. Eissenberg, L. G., Goldman, W. E. & Schlesinger, P. H. (1993) J. Exp. Med. 177, 1605–1611. 3. San-Blas, G., San-Blas, F. & Serrano, L. E. (1977) Infect. Immun. 15, 343–346. 4. Kanetsuna, F. & Carbonell, L. M. (1971) J. Bacteriol. 106, 946–948. 5. Klimpel, K. R. & Goldman, W. E. (1987) Infect. Immun. 55, 528–533. 6. Hogan, L. & Klein, B. (1994) Infect. Immun. 62, 3543–3546. 7. Eissenberg, L. G., West, J. L., Woods, J. P. & Goldman, W. E. (1991) Infect. Immun. 59, 1639–1646. 8. Greenberg, E. P. (1997) Am. Soc. Microbiol. News 63, 371–377. 9. Dunny, G. M. & Leonard, B. A. B. (1997) Annu. Rev. Microbiol. 51, 527–564. 10. Fuqua, C, Winans, S. C. & Greenberg, E. P. (1996) Annu. Rev. Microbiol. 50, 727–751. 11. Batanghari, J. W. & Goldman, W. E. (1997) Infect. Immun. 65, 5257–5261. 12. García Véscovi, E., Soncini, F. C. & Groisman, E. A. (1996) Cell 84, 165–174. 13. Sibley, L. D., Weidner, E. & Krahenbuhl, J. L. (1985) Nature (London) 315, 416–419. 14. Woods, J. P. & Goldman, W. E. (1992) Mol. Microbiol. 6, 3603–3610. 15. Woods, J. P. & Goldman, W. E. (1993) J. Bacteriol. 175, 636–641. 16. Patel, J. B., Batanghari, J. W. & Goldman, W. E. (1998) J. Bacteriol. 180, 1786–1792. 17. Batanghari, J. W., Deepe, G. S., Jr., Di Cera, E. & Goldman, W. E. (1998) Mol. Microbiol. 27, 531–539. 18. Haas, J., Park, E.-C. & Seed, B. (1996) Curr. Biol. 6, 315–324.


EXPLOITATION OF HOST CELLS BY ENTEROPATHOGENIC ESCHERICHIA COLI

8799

Colloquium Exploitation of host cells by enteropathogenic Escherichia coli B. A. Vallance and B. B. Finlay* Biotechnology Laboratory, University of British Columbia, Vancouver, BC, Canada V6T 1Z3 Microbial pathogens have evolved many ingenious ways to infect their hosts and cause disease, including the subversion and exploitation of target host cells. One such subversive microbe is enteropathogenic Escherichia coli (EPEC). A major cause of infantile diarrhea in developing countries, EPEC poses a significant health threat to children worldwide. Central to EPEC-mediated disease is its colonization of the intestinal epithelium. After initial adherence, EPEC causes the localized effacement of microvilli and intimately attaches to the host cell surface, forming characteristic attaching and effacing (A/E) lesions. Considered the prototype for a family of A/E lesion-causing bacteria, recent in vitro studies of EPEC have revolutionized our understanding of how these pathogens infect their hosts and cause disease. Intimate attachment requires the type III-mediated secretion of bacterial proteins, several of which are translocated directly into the infected cell, including the bacteria's own receptor (Tir). Binding to this membrane-bound, pathogen-derived protein permits EPEC to intimately attach to mammalian cells. The translocated EPEC proteins also activate signaling pathways within the underlying cell, causing the reorganization of the host actin cytoskeleton and the formation of pedestal-like structures beneath the adherent bacteria. This review explores what is known about EPEC's subversion of mammalian cell functions and how this knowledge has provided novel insights into bacterial pathogenesis and microbe-host interactions. Future studies of A/E pathogens in animal models should provide further insights into how EPEC exploits not only epithelial cells but other host cells, including those of the immune system, to cause diarrheal disease. T he study of bacterial pathogenesis has undergone a dramatic resurgence in interest, in part because of the reemergence of old diseases such as tuberculosis, the emergence of new bacterial diseases, and the development of antibiotic resistance in many bacterial pathogens. Much of this recent work has focused on defining the molecular and cellular mechanisms underlying how microbes cause disease and has led to a new appreciation of bacterial pathogenesis. An emerging theme in this field is the ability of many bacteria to exploit host cell signal transduction pathways and cytoskeletal/membrane components to allow colonization and invasion of their hosts. In most cases, the bacteria that take this approach are intracellular pathogens such as Shigella, Listeria, and Salmonella (1). These microbes obtain entry into the host by triggering their own uptake by both phagocytic and nonphagocytic host cells. In contrast, enteropathogenic Escherichia coli (EPEC) is an extracellular pathogen that causes disease by binding to the surface of host cells and directly injecting virulence factors into the underlying cell through its type III secretion system (2). These translocated bacterial proteins then interact with host cell components and alter signaling pathways, resulting in disease. EPEC is a serious and widespread cause of infantile diarrhea, particularly in developing countries (3). During infection, EPEC induces a characteristic “attaching and effacing” (A/E) histopathology on gut enterocytes. A/E lesions are characterized by the localized effacement of microvilli and marked cytoskeletal changes, including the accumulation of polymerized actin, directly beneath the adherent bacteria (4, 5). The reorganization of actin forms a pedestal-like structure upon which the bacterium resides (4, 5). A/E lesion formation thus firmly anchors the bacterium to the host cell, and this intimate attachment is thought to be essential for EPEC pathogenicity. In recent years, our laboratory and others have made significant progress determining the mechanisms by which EPEC attaches to mammalian cells in culture and in defining the role of EPEC's virulence factors in the regulation of host cytoskeletal rearrangements and gene expression during infection. These secreted bacterial proteins are thought to corrupt host cell systems, redirecting the cell's own structural components to support the attachment of EPEC. They also induce changes in host cell signaling pathways that likely act not only in pedestal formation, but also in mediating the diarrheal response to EPEC infection. Perhaps the most significant finding of recent years was the discovery that EPEC does not bind to a host receptor during the process of intimate attachment, but instead inserts its own receptor [translocated intimin receptor (Tir)] into the membrane of the target host cell (6). Since this novel finding, other A/E lesion-causing bacterial pathogens have also been shown to produce Tir homologues (7), suggesting that the process of intimate attachment is conserved amongst many enteric pathogens. These include the hemolytic uremic syndrome causing enterohemorrhagic E. coli (EHEC; O157:H7), as well as the rabbit enteropathogenic E. coli (REPEC) and Citrobacter rodentium. Studies investigating EPEC's virulence factors are the most advanced, with EPEC considered the prototype for the family of A/E inducing pathogens. In this paper, we will review the recent progress made in elucidating the mechanisms of EPEC virulence and its exploitation of host cells. Further, we will discuss the role of these factors in the intact host and how future studies may aid our understanding of the contributions that bacterial, as well as host factors, make to EPEC mediated disease.

CLINICAL SYMPTOMS AND PATHOLOGY As a human pathogen, outlining the symptoms and pathology associated with EPEC infection provides context to the recent advances made defining the molecular basis for EPEC mediated disease. One of several categories of diarrheagenic E. coli, EPEC

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviations: EPEC, enteropathogenic Escherichia coli; REPEC, rabbit enteropathogenic Escherichia coli; EHEC, enterohemorrhagic Escherichia coli; A/E, attaching and effacing; Tir, translocated intimin receptor; LEE, locus of enterocyte effacement; Esp, EPEC-secreted protein; BFP, bundle forming pilus; WAS, Wiskott-Aldrich syndrome; PMN, polymorphonuclear cell. * To whom reprint requests should be addressed at: Biotechnology Laboratory, Room 237, Wesbrook Building, 6174 University Boulevard, University of British Columbia, Vancouver, BC, Canada V6T 1Z3. E-mail: [email protected].



8800

is a well established cause of human diarrhea, particularly in young children. Although outbreaks were still frequent in developed countries until the 1940s and 1950s (8), the incidence of EPEC infection in the United States and United Kingdom has since declined. However, EPEC is still responsible for occasional outbreaks in daycare centers and pediatric wards (9). EPEC has remained an important cause of infant mortality in developing countries, with recent outbreaks reporting a mortality rate of 30% (10). Thus, EPEC infection is estimated to cause the deaths of several hundred thousand children per year (2). The hallmark of EPEC infection is the A/E histopathology often observed in small bowel biopsy specimens from infected patients, and seen after the infection of epithelial cells in tissue culture (2, 3). Infection generally causes acute diarrhea, but severe cases can lead to a protracted disease (3). Aside from profuse watery diarrhea, both vomiting and the development of fever are common symptoms of EPEC infection (3). Based on the morbidity and mortality associated with this microbe, EPEC strains remain a significant health threat to children worldwide.

THE LOCUS OF ENTEROCYTE EFFACEMENT Unlike the nonpathogenic strains of E. coli found within the human intestine, EPEC and other pathogenic E. coli strains contain pathogenicity islands within their genome. All of the genes necessary for the formation of A/E lesions and pedestals are contained within a 35kbp pathogenicity island termed the locus of enterocyte effacement (LEE) (2, 5). The G + C content of the LEE is 38.4%, significantly lower than the 50% composition of the nonpathogenic E. coli K-12 chromosome. This discrepancy suggests that the LEE was originally acquired from a foreign source and was subsequently inserted into EPEC's chromosome. The insertion site for the LEE region in the E. coli K-12 genome is at the site encoding the tRNA for selenocysteine. Interestingly, this location appears to be a hot spot for the insertion of several virulence factor genes, including a large but different pathogenicity island found in uropathogenic E. coli (11). The complete LEE region has been sequenced and contains a type III secretion system (2), as well as other genes necessary for pedestal formation. These include several genes coding for type IIIsecreted proteins, termed Esps (EPEC-secreted proteins), including EspA, EspB, EspD, and EspF, as well as an adhesin, intimin, and its translocated receptor, Tir. Mutation of any of these bacterial factors, with the exception of EspF (12), prevents A/E lesion formation in epithelial cell culture models (3). As with other type III secretion systems, cytosolic chaperone proteins have been shown to be required for the translocation of secreted effector proteins. Two chaperones have been identified in the LEE, CesD for EspB and EspD (13), and CesT that chaperones Tir (14). DNA sequences with a high degree of homology to the EPEC LEE have been found in the other A/E lesion-causing bacteria, including EHEC, as well as the mouse pathogen C. rodentium, suggesting a common pathway underlying A/E lesion formation (2). This pathway is also self contained because the introduction of the cloned LEE of EPEC into a previously nonpathogenic E. coli strain conferred the ability to form A/E lesions (15).

LOCALIZED EPEC ADHERENCE TO EPITHELIAL CELLS Interactions between EPEC and host cells entail several distinct steps and have classically been viewed as a three-stage process. The first stage in EPEC pathogenesis involves the initial adherence of the bacterium to the host's intestinal epithelium. In this stage, EPEC form dense microcolonies on the surface of tissue culture cells in a pattern known as localized adherence (3). The bacterium is thought to initially attach to the host cell through a plasmid-encoded bundle forming pilus (BFP). Although mutants lacking this plasmid still attach to host cells, they do not form microcolonies and produce fewer A/E lesions than wild-type EPEC (5, 16). Even so, BFP remains an important virulence factor because BFP mutant strains show severe impairment in their ability to cause diarrhea in human volunteers (17). This loss of virulence probably indicates that both initial adherence to host cells as well as microcolony formation are critically involved in the ability of EPEC to successfully infect its host. Curiously, the mediators of initial attachment appear to vary among the family of A/E pathogens. The related pathogen EHEC lacks BFP and, unlike EPEC, infects the human colon rather the ileum (3). Therefore, whether bacterial colonization occurs preferentially in the small or the large bowel may be influenced by the expression of BFP and other adhesins as well as by environmental factors regulating the expression of other virulence factors (18).

EPEC-SECRETED PROTEINS The second stage of EPEC pathogenesis involves the production of bacterial proteins including EspA, EspB, and EspD. These proteins are translocated from the bacterial cytoplasm to the external environment by a type III secretion system (Fig. 1), encoded by the esc and sep genes, also found within the LEE pathogenicity island. The type III secretion machinery is thought to generate a pore permitting this translocation to occur (2, 5). Type III secretion systems also play an important role during infection by other Gram-negative pathogenic bacteria such as Yersinia and Salmonella, enabling virulence factors to be translocated directly from the bacterial cytoplasm to the host-cell membrane or cytoplasm. Although the majority of the Esps produced by EPEC are necessary for A/E lesion formation, their precise role in EPEC pathogenesis is not well defined. EspA makes filamentous appendages surrounding the bacterium that are transiently present on the bacterial surface (19). These filaments interact with the host cell, possibly forming a translocation tube reaching into the host cell. In support of this theory, EspB is translocated into the host cytosol and membrane by a process dependent on EspA (19). EspD is known to be inserted into the host cell membrane (20). Although the exact functions of EspB and EspD are unknown, their sequence homology to the

Fig. 1. Translocation of EPEC-secreted proteins (Esps) occurs through a type III secretion system that forms a pore through EPEC's membranes. Once translocated outside the bacteria, EspA forms a filamentous translocation tube whereas EspB and EspD are inserted into the host cell membrane, putatively forming a pore structure, allowing the passage of other effector proteins, such as Tir into the host cell membrane.



8801

YopD and YopB system in Yersinia, and their ability to lyse red blood cells (21), suggests they function as components of the translocation apparatus machinery, forming a pore structure in the host membrane. Thus, the primary function of EspB and EspD may be to deliver other virulence proteins to the host cell, rather than acting as translocated effectors themselves. However, transfection of EspB into HeLa cells also leads to changes in cellular morphology and the reorganization of stress fibers at late time points (22), suggesting that EspB may also act as a cytoskeletal toxin. EspF is also a LEE-encoded EPEC secreted protein, although its role in EPEC pathogenesis is undefined because mutants lacking EspF still form A/E lesions (12).

INTIMATE ADHERENCE AND THE ROLE OF TIR The third stage of EPEC infection is characterized by enterocyte effacement, pedestal formation, and intimate bacterial attachment to the host cell. Intimate attachment requires the outer bacterial membrane protein, intimin. A 94-kDa outer membrane protein encoded by the eae gene, intimin mediates intimate attachment to the host cell by binding to a 90-kDa protein in the host membrane (23). This receptor, now called Tir, was originally thought to be a host protein but was recently shown to be a bacterial protein that is translocated into the host cell membrane (6). The bacterial form of Tir migrates as 78 kDa when analyzed by SDS/PAGE, but, after translocation into the host membrane, Tir undergoes phosphorylation on tyrosine (6) and probably serine and threonine residues (24), and these modifications account for the apparent shift to 90 kDa. EPEC's use of Tir represents the first example of a pathogen injecting its own receptor into mammalian cells and has revised our concepts of bacterial pathogenesis. By ignoring the usual dependence on the expression of a host-derived receptor, EPEC can, at least in tissue culture, infect cells of most species and tissue origins. Based on its novel role in EPEC pathogenesis, Tir has undergone intense study since its identification. Much like EspB and EspD, the transfer of Tir to the host cell requires the type III secretion system and the Esps critical for the formation of A/E lesions. Tir is required for pedestal formation because its deletion prevents EPEC from forming A/E lesions in tissue culture (6). Tir has two predicted transmembrane domains with a hairpin model proposed for Tir conformation in the host membrane (2). As predicted, the N- and Cterminal regions of Tir are located within the cell whereas the intervening region between the transmembrane domains forms an extracellular loop. Several groups have recently shown that intimin binds to this extracellular loop termed the intimin binding domain, via its C-terminal region (24, 25 and 26). Tir-intimin binding has been shown to be essential for pedestal formation and actin condensation (6). Further, Liu et al. recently used latex beads coated with the C-terminal region of EHEC intimin to trigger A/E lesion formation on HEp-2 cells (27). Pedestals only formed when cells were preinfected with an EPEC intimin mutant, which translocates Tir into the host cell. This confirms the essential role of Tirintimin interactions in pedestal formation. Studies examining A/E lesion formation in tissue culture have shown that only a portion of surface intimin is required to interact with translocated Tir. After A/E lesion formation, the expression of surface intimin not bound to Tir is down-regulated (28). Intimin has also been shown to bind to host cells in vitro through its C-terminal region (int280) in a Tir-independent manner (26, 29), suggesting more than one receptor for intimin on epithelial cells. Such binding requires the cysteine 937 residue (26, 29). β1 integrins have been proposed to be the host cell receptor mediating such binding. Although integrins are not present on the apical surface of enterocytes, they are expressed by other cell types, including on the apical surface of M cells located on the luminal surface of Peyer's patches (30). Interestingly, recent studies have shown that, during C. rodentium infection of mice, intimin from C. rodentium or EPEC can induce colonic epithelial hyperplasia concurrent with a strong T helper cell 1 immune response (31). These responses did not depend on bacterial viability because exposure to formalin fixed bacteria still elicited these events (32) whereas an intimin substitution mutant for cysteine 937 failed to do so. This suggests that intimin binding to a host cell receptor precipitates the host response. In fact, C. rodentium and EPEC may directly interact with mucosal immune cells because many mucosal T lymphocytes express β1 integrins. Despite these findings, the question of a host-derived receptor for intimin remains controversial, with reports from other laboratories finding that intimin does not bind β1 integrins and that Tir is a necessary requirement for EPEC binding to host cells (23, 33).

STRUCTURE OF THE A/E LESIONS The host cell undergoes a number of alterations during infection by EPEC, but the most striking change is the formation of actin pedestals. In fact, the resulting localized actin accumulation is so distinct that it forms the basis of an in vitro diagnostic test for EPEC (34). The process of pedestal formation begins following the adherence of EPEC to epithelial cells in vitro or in vivo, in concert with host cell microvilli effacement. Within 3 h of tissue culture cell infection by EPEC, pedestals begin to form directly beneath the bacteria (2, 5). The epithelial membrane beneath the adherent organisms is raised to form pedestal-like structures that can extend up to 10 µm away from the cell to form a pseudopodlike structure (35) (Fig. 2). Beyond providing strong attachment of EPEC to the cell surface, presumably to prevent being dislodged during the ensuing diarrheal response by the host, the role of pedestals is unclear. It may, however, represent a strategy by EPEC to remain extracellular, in keeping with the ability of EPEC to block its own phagocytosis by macrophages (36). Immunofluorescence studies have shown that, in addition to membrane bound Tir, pedestals contain predominantly filamentous (F)-actin (37) (Fig. 3). α-actinin, talin, ezrin, and villin also accumulate along the length of the pedestal, as these cytoskeletal components act in the crosslinking of actin microfilaments. Nonmuscle myosin II and tropomyosin are also found, but at the base of the pedestal (37). Based on its location at the pedestal tip, as well as its predicted structure, Tir is the most likely

Fig. 2. Transmission electron micrograph of A/E lesions in rabbit intestinal epithelial tissue (Peyer's patch) caused by REPEC O103. Bacteria labeled with “B”; pedestal labeled with “P.” (Photograph is courtesy of Ursula Heczko, Biotechnology Laboratory, University of British Columbia.) (×20,000.)



8802

bacterial candidate to link EPEC to the host cytoskeleton and direct actin accumulation and pedestal formation. To accomplish this, Tir may exploit regulators of actin dynamics to initiate actin polymerization. Members of the Rho family of small GTP-binding proteins were considered potential candidates for this role, but investigations have shown that Rac, Rho, and Cdc42-dependent pathways are not involved in pedestal formation (38). Recent studies have instead implicated members of the Wiskott-Aldrich syndrome (WAS) family of proteins (WASP and NWASP) as well as an actin nucleating factor, the heptameric Arp2/3 complex. These factors are recruited to the pedestal tip (39) (Fig. 4), and mutation of the GTPase binding domain of WASP prevented the recruitment of the Arp2/3 complex, and pedestal formation. Although host cell cytoskeletal rearrangements are responsible for pedestal formation, they are probably also responsible for microvilli effacement. Although this remains to be proven, the disappearance of microvilli may result from a bacterial-triggered depolymerization of microvilli actin, which is then used to form pedestals. However, some proteins found in pedestals are not derived from the microvilli, suggesting a more complex process than a simple rebuilding of microvilli beneath the adherent bacteria. It is currently unclear whether all of the cytoskeletal proteins identified within pedestals are essential for their formation.

Fig. 3. Pedestal formation on epithelial cells induced by enteropathogenic Escherichia coli. Fluorescence microscopy of the EPEC pedestal, triple-labeled for actin (green), EPEC (blue), and Tir (red) (courtesy of Danika Goosney, Biotechnology Laboratory, University of British Columbia), (×1,000.) SIGNAL TRANSDUCTION Another critical form of cellular exploitation used by EPEC involves the subversion of host signaling pathways to aid in infection. As a result, several signal transduction pathways are stimulated within epithelial cells after EPEC infection. One such pathway results in tyrosine phosphorylation of substrates that co-localize with the accumulated actin beneath adherent bacteria. The major phosphorylation substrate detected in EPEC-infected cells is Tir (6). This phosphorylation event requires both the type III secretion system, as well as the Esps, because their mutation prevents EPEC from localizing tyrosine phosphorylated proteins such as Tir beneath adherent EPEC (2, 6). This suggests that EPEC triggers these responses through bacterial effector molecules. Because phosphorylation occurs after Tir enters the host cell, tyrosine kinase activity must be recruited to the vicinity of Tir within the cell. Although the identity of the involved tyrosine kinase is unknown, it is probably of host origin, presumably recruited or activated by bacterial effector proteins, or by Tir mimicking an endogenous substrate. Alternatively, a bacterial effector with tyrosine kinase activity could be translocated into the cell along with Tir. There is speculation that Tir is also serine/ threonine-phosphorylated (24). The tyrosine phosphorylation event during EPEC infection is critical for actin nucleation because, in its absence, A/E lesions do not form (2, 6). Surprisingly, tyrosine phosphorylation of Tir does not occur with the related pathogen EHEC O157:H7 (7), although this pathogen readily forms pedestals. This divergence in signaling requirements suggests that subtle but important differences exist between these two pathogens in their modes of pedestal formation, although this awaits further examination.

Fig. 4. The structure of the EPEC pedestal. EPEC intimately attaches to the host cell through intimin-Tir binding. N-WASP and the Arp 2/3 complex are recruited to the pedestal tip, nucleating actin. F-actin, α-actinin, talin, ezrin, and villin are found along the length of the pedestal whereas nonmuscle myosin II and tropomyosin are found at the pedestal base. EPEC also induces other signaling cascades within the host cell, including inositol phosphate fluxes, activation of protein kinase C, phospholipase-Cγ, and NF- B (2, 5). One consequence of inositol trisphosphate fluxes is the release of Ca2+ from intracellular stores. Several groups have investigated the effect of EPEC infection on changes in intracellular calcium, but the results have been inconclusive. Crane and Oh measured protein kinase C activity and found an enhancement in membrane-associated protein kinase C. However, increasing protein kinase C activity required intimate adherence of the bacterium because intimin mutants did not affect activity (40). EPEC also induces the activation of NFB in a T84 epithelial cell culture model, in association with increased interleukin-8 production and the recruitment of polymorphonuclear cells (PMNs) (41, 42). Although the importance of many of these signaling events in infection is demonstrated by their requirement for pedestal formation, much further characterization is required. In fact, it remains unclear whether these responses are specifically triggered by bacterial effector proteins or are a nonspecific consequence of EPEC infection. In either case, changes in cell signaling likely play a major role in the symptomatology of EPEC-mediated disease, and particularly in mediating the resulting diarrhea.



8803

Fig. 5. Putative mechanisms underlying EPEC induced diarrhea include increased epithelial permeability and alterations in Cl− and HCO3− ion secretion. Contributing structural changes include loss of absorptive surfaces, reduced tight junction integrity, and tissue damage. MECHANISMS OF EPEC-MEDIATED DIARRHEA The primary symptom of EPEC infection is diarrhea. Unfortunately, despite the advances made in our understanding of EPEC pathogenesis at the genetic and cellular level, how EPEC triggers diarrhea is still uncertain. In fact, it remains unclear if the resulting diarrhea is triggered by a specific manipulation of the host by EPEC virulence factors, or if EPEC is the recipient of a stereotyped host response to bacterial adhesion. Whatever the cause, diarrhea may prove of benefit to A/E lesion-causing bacteria and particularly those that inhabit the large bowel. As they colonize the intestine, A/E lesion-causing pathogens not only must interact with the cells of their eukaryotic hosts, but they are often in competition with normal intestinal flora and even other pathogens. Mechanisms like diarrhea that disturb the normal host-prokaryote equilibrium presumably provide EPEC with an advantage over competing flora. By intimately binding to the host's enterocytes, EPEC can remain attached to the host's intestinal surface whereas other less adherent microbes are flushed away. The diarrhea seen during EPEC infection could be caused by the dramatic loss of absorptive microvilli in the A/E lesion (3). Alternatively, at least one study has shown that EPEC infection reduces the tight junction integrity of epithelial cell monolayers, based on altered distribution of tight junction proteins such as zona occludens (ZO)-1 (43). This observation remains controversial because other studies showed no change in such proteins during infection (44). However, in volunteer studies, the incubation period between EPEC ingestion and the onset of diarrhea is less than 4 h, suggesting that a more active secretory response may be involved (45). Studies examining other diarrheal pathogens have identified changes in chloride ion secretion as one of the most common mechanisms leading to secretory diarrhea (46). Interestingly, reports have shown that EPEC can actively alter ion transport, causing a rapid but transient increase in short circuit current (Isc) in intestinal epithelial cell monolayers mounted in Ussing chambers, with chloride ion secretion implicated in this effect (47, 48). Mutation of espB but not the gene encoding intimin abrogated these ionic changes (47, 48). This agrees with studies of EPEC infection in adult human volunteers. Although EPEC intimin mutants were less virulent, diarrhea still developed in 4 of 11 volunteers who ingested the intimin mutant (45), indicating the involvement of other virulence factors in EPEC induced diarrhea. It should be noted that not all studies have supported a role for chloride ion secretion. Hecht and Koutsouris recently implicated changes in bicarbonate (HCO3−) ions rather than chloride ions in the EPEC mediated changes in intestinal ion transport (49). Finally, other host factors beyond those present in epithelial cell cultures may also contribute to diarrhea. There is substantial recruitment of neutrophils and other PMNs to the site of in vivo infection (50). The inflammatory response may be attributable to bacterial triggered signals from infected cells because EPEC activates both NF- B and interleukin-8 expression in tissue culture cells (41, 42). These signals were associated with transmigration of PMNs through epithelial cell monolayers. Increased paracellular permeability and stimulation of chloride secretion could be a consequence of this EPEC-induced PMN infiltration (Fig. 5). In vivo, an inflammatory response should take longer than 3 h to develop, suggesting it is not the mechanism that initiates EPEC-mediated diarrhea, although inflammation may contribute to the duration and severity of the diarrheal response. Care must be taken interpreting these responses, however, because many enteric pathogens, including Salmonella (51), activate NF- B and IL-8 expression simply by adhering to the surface of host cells. This response could therefore be a generalized mechanism used by pathogens to initiate diarrhea, or could reflect a stereotyped innate host response to bacterial adhesion. Whatever the cause, the PMNs and other inflammatory cells recruited to the infected intestine also cause considerable tissue damage through the release of toxic inflammatory mediators (46). Although the resultant pathology probably contributes to changes in epithelial function, many of the cytokines expressed during A/E lesion causing bacterial infections (31), such as γ-interferon and TNF-α, have also been shown to directly alter epithelial cell function in tissue culture (52).

PATHOGENESIS IN ANIMAL MODELS Despite the progress made defining the molecular basis for EPEC-host cell interactions, we still know very little about EPEC-mediated disease. Infections affect more than a single cell type in isolation and need to be considered in the context of the complexity of their hosts. Because EPEC is primarily a pediatric pathogen, ethical considerations have generally precluded tak



8804

ing biopsy samples during infection. As a result, only a few studies of EPEC infection have been carried out using adult human volunteers (45). These findings have been recently reviewed by Nataro and Kaper (3). These studies used very large bacterial doses and, although important, represent an artificial situation. There is now an urgent need to examine the new information regarding EPEC pathogenesis in the context of animal models. Unfortunately, EPEC is a human specific pathogen and does not infect most laboratory animal species. In this case, the family of enteric A/E lesion-causing pathogens can be used to draw conclusions about EPEC, by using microbes that naturally infect other animal species. E. coli capable of forming A/E lesions have been isolated from rabbits (REPEC), as well as calves, sheep, pigs, dogs, and mice. The rabbit pathogen REPEC produces a disease similar to EPEC, infecting the small bowel of weanling rabbits and causing both diarrhea and weight loss (53). Despite these advantages, there are limitations to the model, as genetic and immunological resources are not plentiful for the rabbit. In this respect, C. rodentium (formerly known as Citrobacter freundii biotype 4280) offers an advantage. C. rodentium produces A/E lesions in mice and, like EHEC in humans, colonizes the large rather than the small bowel. Unlike REPEC, this pathogen induces a strong host Th1 immune response as well as epithelial cell hyperplasia rather than diarrhea (31). Although the symptomatology may differ, the basic mechanisms of A/E lesion formation and host response likely remain the same. As a result, the wide array of reagents available for the mouse, including antibodies, recombinant cytokines, and gene knockout strains, makes this model highly suited to study the host response to infection. Although animal models offer clear opportunities to improve our understanding of how A/E lesion-causing pathogens cause disease, in vivo studies have lagged behind in vitro studies of EPEC. We are only beginning to identify the bacterial factors necessary for disease, but, so far, results from animal models have been encouraging, as they have validated those findings made in tissue culture. Using the REPEC model, Abe et al. demonstrated the crucial nature of the bacterial proteins EspA and EspB in both A/E lesion formation and diarrheal disease (53). REPEC strains lacking either gene lost the ability to form A/E lesions, although they were still able to locally adhere to the small bowel. The essential role of EspB in the formation of A/E lesions has also been shown in the C. rodentium model in mice (54). Interestingly, the recent studies on C. rodentium infection by Higgins et al. identified an important component of the bacterial triggered host response not dependent on A/E lesion formation (32). The discovery that intimin expression was required to induce both the epithelial hyperplasia and T helper cell 1 immune activation confirms the need to assess potential virulence factors of EPEC in animal models.

EPEC INTERACTIONS WITH THE INTACT HOST As described above, the focus of research on EPEC pathogenesis has been on the microbe. Through the mutation of genes encoding potential bacterial virulence factors, their role in the infectious process has been surmised by the ability to form A/E lesions in tissue culture, or their ability to cause diarrhea in vivo. This has proven successful because most of EPEC's LEE-encoded genes have been assessed in vitro for their importance in A/E lesion formation (2, 3, 5), and several have been tested in rabbit models for their importance in the diarrheal response (53). Unfortunately, these studies only demonstrate the necessity of a bacterial protein for pedestal formation or diarrhea but do not identify its actual role(s) in the infectious process. As a result, only a few of EPEC's arsenal of effector proteins have been well characterized, and, likely, even these factors have as yet unrecognized functions, as exemplified by the discovery of intimin's role in generating the immune response during C. rodentium infection (32). Characterizing the functions of EPEC's secreted and translocated effectors is the next step in the field of EPEC pathogenesis. This will require continued intensive cellular and biochemical analysis of the changes elicited within infected host cells. To fully appreciate EPEC's capacity to exploit mammalian cells, these studies also need to be examined within the complexity of an intact host. There are a number of areas that need to be assessed, including how EPEC interacts with and manipulates the array of cell types present within the intestine. The mammalian intestinal epithelium is a highly specialized tissue that maintains complex and selective secretory and absorptive functions while interfacing with the external luminal environment. Particularly in the colon, the epithelium exists within a diverse microfloral ecology, which contributes to and regulates the physiology of the lower gut. As a result, intestinal epithelial cells have evolved selective physical, chemical, and immunological barriers that permit this mutually beneficial co-existence (51). Not surprisingly, studies using epithelial cell cultures can only model the intestinal epithelium in a limited fashion. The epithelial lining of the intestine is also a dynamic system, with epithelial cells undergoing rapid turnover every 3–4 days (55). Thus, the epithelial layer of the gut contains cells at varying stages of differentiation, ranging from immature crypt cells to mature enterocytes. As well, more specialized forms of epithelial cells are found interspersed among the columnar epithelium. These include the mucus-secreting goblet cells and the antigen-sampling M cells that overlie the gut-associated lymphoid tissues. Although EPEC is probably capable of interacting with these cells to form pedestals (56), whether EPEC subverts their function by other means has yet to be examined. Besides interactions with the host's intestinal epithelium, A/E lesion-causing bacteria must also encounter and presumably circumvent innate host defenses. Unfortunately, little is known about innate immunity against EPEC infection, but histological examination of infected tissues has identified both neutrophils and macrophages responding to both EPEC and C. rodentium infection (31, 50). This recruitment of inflammatory cells may be in response to signals sent by the infected epithelium because epithelial cells in culture can produce the neutrophil chemoattractant IL-8 in response to infection by EPEC (41, 42). An active role for intestinal epithelial cells in host defense is not new; many studies have shown that epithelial cells can secrete a number of pro-inflammatory and antimicrobial agents in response to bacterial infection or after immune stimulation (46, 52). Other aspects of innate immunity against EPEC, such as the actions of antimicrobial peptides, have yet to be studied. However, both human colostrum and milk have been shown to strongly inhibit the adhesion of EPEC to HEp-2 cells in vitro with the inhibitory activity found in both the sIgA and oligosaccharide fractions (57). These results suggest that breast-feeding may protect infants from EPEC infection, although a better understanding of the mechanisms involved is required. Just as the host has developed protective measures, EPEC has evolved measures to counteract and subvert the host's immune response. EPEC can block its own uptake by professional phagocytes like macrophages (36), presumably to inhibit antigen presentation by macrophages. This requires a functioning type III secretion system and the expression of EspA, EspB, and EspD, but not Tir. These requirements suggest that EPEC disrupts the phagocytic process by directly contacting macrophages, rather than through a soluble mediator. EPEC can also inhibit the host's immune response, with Malstrom and James reporting that EPEC lysates inhibited IL-2, IL-4,



8805

and IFN-γ production by both mucosal and splenic lymphocytes (58). This effect did not depend on the expression of either EspA or EspB. However, it is unclear how relevant these interactions may be when taken in the context of the massive immune activation seen in the related C. rodentium model (32).

CONCLUSIONS Research in recent years has made remarkable progress in our understanding of the molecular mechanisms underlying EPEC pathogenesis. Molecular biology, genetics, and cell biology have provided many new insights into how EPEC and related pathogens interact with and exploit host cells during infection. The identification of a type III secretion system within EPEC's genome, as well as the discovery of Tir, has improved our understanding of how EPEC subverts the host cytoskeleton to permit bacterial attachment. This research has identified features common to other enteric pathogens as well as strategies apparently unique to A/E lesion-causing bacteria, such as the translocation of the bacterial receptor Tir into the host membrane. Although studying the molecular and cellular aspects of bacterial virulence factors has informed us about the mechanisms of bacterial disease, these factors can also function as tools to study various aspects of mammalian cell functions. As such, they have received much interest in the field of microbial pathogenesis, as well as from cell biologists interested in the mechanisms underlying actin dynamics and cytoskeletal rear-rangements (39). These studies also have relevance to unrelated pathogens, such as Helicobacter pylori, which produces A/E-like lesions (59). However, H. pylori-induced pedestals lack the intense actin accumulation observed with the A/E intestinal pathogens, and no homologues of the A/E genes have been found in H. pylori. There are also similarities between the immunopathology associated with A/E lesion-inducing bacteria, and that seen in inflammatory bowel diseases (31). Although no infectious causal agent has yet been identified in inflammatory bowel diseases, the potential contribution of maladaptive microbial-host interactions to the chronicity of these diseases has long been of great interest to gastroenterologists (60). Despite this progress, more studies are needed characterizing EPEC's effector molecules as well as their role in causing diarrhea and disease within the intact host. Other gene products that contribute to colonization of the host must be identified, and their role in the infectious process examined in vivo. Although it is true that infections may differ between animals and humans, the similarities will probably prove greater than the differences. Studies integrating host genetics, physiology, and the immune system, all of which are critical determinants to the outcome of infection, should provide a better understanding of EPEC and other A/E pathogens, and together these developments may lead to new therapeutic strategies. With our present knowledge of the factors that mediate bacterial adhesion to the host cell, and the demonstration that preventing bacterial adherence prevents most aspects of the disease, we have already identified potential targets for vaccination. Although the pathogenic effects of A/E lesion formation still need to be separated from other bacterial actions during infection, a successful approach may involve vaccination against the factors involved in bacterial adhesion such as the components and effectors of the type III secretion machinery. Alternatively, with the recent interest in microbe-microbe interactions, and in the use of probiotics, identifying bacterial species that can outcompete EPEC for attachment to host cells may be an attractive option. By continuing to characterize EPEC's effector molecules, their specific effects, and the host's response to infection, we should look forward to new advances in the prevention and treatment of A/E pathogen mediated diarrhea. We thank Jean Celli, Annick Gauthier, and Carrie Rosenberger for helpful discussions and critical reading of the manuscript and Danika Goosney, Ursula Heczko, and Fern Ness for figures. Work in our laboratory is supported by operating grants from the Medical Research Council of Canada and a Howard Hughes International Scholar award (to B.B.F.). B.B.F. is a Medical Research Council Scientist. 1. Steele-Mortimer, O., Knodler L. A. & Finlay, B. B. (2000) Traffic 1, 107–118. 2. DeVinney, R., Gauthier, A., Abe, A. & Finlay, B. B. (1999) Cell Mol. Life Sci. 55, 961–976. 3. Nataro, J. P. & Kaper, J. B. (1998) Clin. Microbiol. Rev. 11, 142–201. 4. Kaper, J. B. (1998) Trends Microbiol. 6, 169–172. 5. Frankel, G., Philipps, A. D., Rosenshine, I., Dougan, G., Kaper, J. B. & Knutton, S. (1998) Mol. Microbiol. 30, 911–921. 6. Kenny, B., DeVinney, R. D., Stein, M., Reinscheid, D. J., Frey, E. A. & Finlay, B. B. (1997) Cell 91, 511–520. 7. DeVinney, R., Stein, M., Reinscheid, D., Abe, A., Ruschkowski, S. & Finlay, B. B. (1999) Infect. Immun. 67, 2389–2398. 8. Robins-Browne, R. M. (1987) Rev. Infect. Dis. 9, 28–53. 9. Bower, J. R., Congeni, B. L., Cleary, T. G., Stone, R. T., Wanger, A., Murray, B. E., Mathewson, J. J. & Pickering, L. K. (1989) J. Infect. Dis. 160, 243–247. 10. Senerwa, D., Olsvik O., Mutanda L. N., Lindqvist, K. J., Gathuma, J. M., Fossum, K. & Wachsmuth, K. (1989) J. Clin. Microbiol. 27, 1307–1311. 11. Blum, G., Ott, M., Lischewski, A., Ritter, A., Imrich, H., Tschape, H. & Hacker, J. (1994) Infect. Immun. 62, 606–614. 12. McNamara, B. P. & Donnenberg, M. S. (1998) FEMS Microbiol. Lett. 166, 71–78. 13. Wainwright, L. A. & Kaper, J. B. (1998) Mol. Microbiol. 27, 1247–1260. 14. Abe, A., de Grado, M., Pfuentzer, R. A., Sanchez-SanMartin, C., DeVinney, R., Puente, J. L., Strynadka, C. J. & Finlay, B. B. (1999) Mol. Microbiol. 33, 1162–1175. 15. McDaniel, T. K. & Kaper J. B. (1997) Mol. Microbiol. 23, 399–407. 16. Donnenberg, M. S., Kaper, J. B. & Finlay, B. B. (1997) Trends Microbiol. 5, 109–114. 17. Levine, M. M., Nataro, J. P., Karch, H., Baldini, M. M., Kaper, J. B., Black, R. E., Clements, M. L. & O'Brien, A. D. (1985) J. Infect. Dis. 152, 550–559. 18. Kenny, B., Abe, A., Stein, M. & Finlay, B. B. (1997) Infect. Immun. 65, 2606–2612. 19. Knutton, S., Rosenshine, I., Pallen, M. J., Nisan, I., Neves, B. C., Bain, C., Wolff, C., Dougan, G. & Frankel, G. (1998) EMBO J. 17, 2166–2176. 20. Kresse, A. U., Rohde, M. & Guzman, C. A. (1999) Infect. Immun. 67, 4834–4842. 21. Warawa, J., Finlay, B. B. & Kenny, B. (1999) Infect. Immun. 67, 5538–5540. 22. Taylor, K. A., Luther, P. W. & Donnenberg, M. S. (1999) Infect. Immun. 67, 120–125. 23. Rosenshine, I., Ruschkowski, S., Stein, M., Reinscheid, D. J., Mills, S. D. & Finlay, B. B. (1996) EMBO J. 15, 2613–2624. 24. Kenny, B. (1999) Mol. Microbiol. 31, 1229–1241. 25. de Grado, M., Abe. A., Gauthier, A., Steele-Mortimer, O., DeVinney, R. & Finlay B. B. (1999) Cell. Microbiol. 1, 7–17. 26. Hartland, E. L., Batchelor, M., Delahay, R. M., Hale, C., Matthews, S., Dougan, G., Knutton, S., Connerton, I. & Frankel, G. (1999) Mol. Microbiol. 32, 151–158. 27. Liu, H., Magoun, L., Luperchio, S., Schauer, D. B. & Leong, J. M. (1999) Mol. Microbiol. 34, 67–81. 28. Knutton, S., Adu-Bobie, J., Bain, C., Philipps, A. D., Dougan, G. & Frankel, G. (1997) Infect. Immun. 65, 1644–1652. 29. Frankel, G., Philips, A. D., Novakova, M., Batchelor, M., Hicks, S. & Dougan, G. (1998) Mol. Microbiol. 29, 559–570. 30. Clark, M. A., Hirst, B. H. & Jepson, M. A. (1998) Infect. Immun. 66, 1237–1243. 31. Higgins, L. M., Frankel, G., Douce, G., Dougan, G. & MacDonald, T. T. (1999) Infect. Immun. 67, 3031–3039. 32. Higgins, L. M., Frankel, G., Connerton, I., Goncalves, N. S., Dougan, G. & MacDonald, T. T. (1999) Science 285, 588–591. 33. Liu, H., Magoun, L. & Leong, J. M. (1999) Infect. Immun. 67, 2045–2049. 34. Knutton 359 Knutton, S., Baldwin, T, Williams, P. H. & McNeish, A. S. (1989) Infect. Immun. 57, 1290–1298. 35. Moon, H. W., Whipp, S. C., Argenzio, R. A., Levine, M. M. & Gianella R. A. (1983) Infect. Immun. 41, 1340–1351. 36. Goosney, D. L., Celli, J., Kenny, B. & Finlay, B. B. (1999) Infect. Immun. 67, 490–495. 37. Sanger, J. M., Change, R., Ashton, F., Kaper, J. B. & Sanger J. W. (1996) Cell Motil. Cytoskeleton 34, 279–287. 38. Ben-Ami, G., Ozeri, V., Hanski, E., Hofmann, F., Aktories, K., Hahn, K. M., Bokoch, G. M. & Rosenshine, I. (1998) Infect. Immun. 66, 1755–1758. 39. Kalman, D., Weiner, O. D., Goosney, D. L., Sedat, J. W., Finlay, B. B., Abo, A. & Bishop J. M. (1999) Nat. Cell Biol. 1, 389–391.



8806

40. Crane, J. K. & Oh, J. S. (1997) Infect. Immun. 65, 3277–3285. 41. Savkovic, S. D., Koutsouris, A. & Hecht, G. (1996) Infect. Immun. 64, 4480–4487. 42. Savkovic, S. D., Koutsouris, A. & Hecht, G. (1997) Am. J. Physiol. 273, C1160–C1167. 43. Philpott, D. J., McKay, D. M., Sherman, P. M. & Perdue, M. H. (1996) Am. J. Physiol. 270, G634–645. 44. Canil, C., Rosenshine, I., Ruschkowski, S., Donnenberg, M. S., Kaper, J. B. & Finlay, B. B. (1993) Infect. Immun. 61, 2755–2762. 45. Donnenberg, M. S., Tacket, C. O., James, S. P., Losonsky, G., Nataro, J. P., Wasserman, S. S., Kaper, J. B. & Levine M. M. (1993) J. Clin. Invest. 92, 1412– 1417. 46. Perdue, M. H. & McKay, D. M. (1994) Am. J. Physiol. 267, G151–G156. 47. Collington, G. K., Booth, I. W., Donnenberg, M. S., Kaper, J. B. & Knutton, S. (1998) Infect. Immun. 66, 6049–6053. 48. Collington, G. K., Booth, I. W. & Knutton, S. (1998) Gut 42, 200–207. 49. Hecht, G. & Koutsouris, A. (1999) Am. J. Physiol. 276, G781–G788. 50. Ulshen, M. H. & Rollo, J. L. (1980) N. Engl. J. Med. 302, 99–101. 51. Gewirtz, A. T., Rao, A. S., Simon, P. O., Merlin, D., Carnes, D., Madara, J. L. & Neish, A. S. (2000) J. Clin. Invest. 105, 79–92. 52. McKay, D. M. & Baird, A. W. (1999) Gut 44, 283–289. 53. Abe, A., Heczko, U., Hegele, R. G. & Finlay, B. B. (1998) J. Exp. Med. 188, 1907–1916. 54. Newman, J. V., Zabel, B. A., Jha, S. S. & Schauer, D. B. (1999) Infect. Immun. 67, 6019–6025. 55. Karam, S. M. (1999) Front. Biosci. 15, D286–D298. 56. Van Moll, L. K. & Cantey, J. R. (1997) Infect. Immun. 65, 3788–3793. 57. Delneri, M. T., Carbonare, S. B., Silva, M. L., Palmeira, P. & Carneiro-Sampaio, M. M. (1997) Eur. J. Pediatr. 156, 493–498. 58. Malstrom, C. & James, S. (1998) Infect. Immun. 66, 3120–3127. 59. Hessey, S. J., Spencer, J., Wyatt, J. I., Sabala, G., Rathbone, B. J., Axon, A. T. & Dixon, M. F. (1990) Gut 31, 134–138. 60. Sartor, R. B. (1997) Ailment Pharmacol. Ther. 11, 17–22.


GENETIC COMPLEXITY OF PATHOGEN PERCEPTION BY PLANTS: THE EXAMPLE OF RCR3, A TOMATO GENE REQUIRED SPECIFICALLY BY CF-2

8807

Colloquium Genetic complexity of pathogen perception by plants: The example of Rcr3, a tomato gene required specifically by Cf-2 Mark S. Dixon*, Catherine Golstein, Colwyn M. Thomas, Erik A. van der Biezen†, and Jonathan D. G. Jones‡ The Sainsbury Laboratory, John Innes Centre, Norwich Research Park, Colney Lane, Norwich NR4 7UH, United Kingdom Genetic analysis of plant–pathogen interactions has demonstrated that resistance to infection is often determined by the interaction of dominant plant resistance (R) genes and dominant pathogen-encoded avirulence (Avr) genes. It was postulated that R genes encode receptors for Avr determinants. A large number of R genes and their cognate Avr genes have now been analyzed at the molecular level. R gene loci are extremely polymorphic, particularly in sequences encoding amino acids of the leucine-rich repeat motif. A major challenge is to determine how Avr perception by R proteins triggers the plant defense response. Mutational analysis has identified several genes required for the function of specific R proteins. Here we report the identification of Rcr3, a tomato gene required specifically for Cf-2mediated resistance. We propose that Avr products interact with host proteins to promote disease, and that R proteins “guard” these host components and initiate Avr-dependent plant defense responses. M any plant pathogens are highly adapted biotrophic parasites that require living hosts to complete their life cycle. Flor, in his pioneering work in the 1940s, studied the interaction between flax and flax rust. He showed that recessive virulence genes enable mutant strains to overcome specific disease resistance (R) genes, and that avirulence (Avr) genes are dominant (1). This “gene-for-gene” interaction was subsequently demonstrated for many other plant–pathogen interactions. Plant R proteins are postulated to provide a surveillance system that can detect Avr determinants from diverse viral, prokaryotic, and eukaryotic pathogens. Here we review recent studies that reveal the genetic complexity of pathogen perception by plants, and we report the identification and preliminary characterization of Rcr3, a gene specifically required for tomato Cf-2 function.

VIRULENCE AND AVIRULENCE It would be surprising if pathogens were to carry Avr genes that had no function other than to enable recognition by plants that carry the matching R genes. During infection, pathogens make an array of virulence factors. If plant R genes can evolve to recognize one of these molecules to trigger a defense response, any such virulence factor could become an avirulence determinant. Avr genes of the bacterial plant pathogens Xanthomonas campestris and Pseudomonas syringae encode hydrophilic proteins that are delivered inside the plant cell by a specialized type III secretion mechanism (2). In the absence of recognition by corresponding R genes, some of these Avr genes have been shown to confer enhanced virulence (3, 4). Viral pathogens also present potential ligands intracellularly, and for some R genes, the corresponding viral Avr protein has been defined as the replicase domain (5) or the coat protein (6). We study resistance to two biotrophic pathogens, the tomato leaf mold fungus Cladosporium fulvum and the oomycete Peronospora parasitica, which causes downy mildew on cruciferous plants, including Arabidopsis. Studies on C. fulvum, which colonizes the intercellular spaces of infected leaves, led to the cloning of Avr4 and Avr9 that confer avirulence on tomato plants carrying the corresponding Cf-4 and Cf-9 R genes (see below). Avr4 and Avr9 encode small secreted peptides that trigger Cf gene-dependent resistance (7, 8). No evidence to support a role for Avr4 or Avr9 as virulence determinants has been demonstrated. In contrast, ECP2 is a secreted peptide that has a virulence function in all C. fulvum strains analyzed (9) and is recognized as an avirulence factor in certain tomato lines (10). P. parasitica, like many other parasites, forms hyphae that produce feeding structures inside the plant cell wall, termed haustoria, that provide intimate association with the plant plasma membrane (11). Avr effectors may be delivered through haustoria inside the plant cell. This is plausible because their cognate R proteins are predicted to be intracellular (see next section).

R PROTEINS Tomato Cf-2, Cf-4, Cf-5, and Cf-9 genes confer recognition of different C. fulvum Avr genes. They encode an extracellular leucine-rich repeat (eLRR) domain, a transmembrane domain, and a short cytoplasmic domain with no sequence similarity to known signaling domains (Fig. 1A) (12). Cf-9 has recently been shown to be a plasma membrane-localized glycoprotein (13). Xa21 is a rice gene that confers resistance to bacterial blight caused by X. campestris pv. oryzae and also encodes eLRRs and a transmembrane domain, but in addition has a cytoplasmic Ser/ Thr protein kinase domain (Fig. 1B) (14). The presence of eLRRs in these R proteins is consistent with a presumed role in the recognition of extracellular ligands. Interestingly, Pto, a tomato gene that confers resistance to P. syringae pv. tomato strains expressing the AvrPto gene, also encodes a Ser/Thr protein kinase (Fig. 1C) (15). Pto lacks a signal peptide (but carries a putative myristoylation site), and its interaction with AvrPto in yeast suggests a cytoplasmic recognition capacity (16, 17).

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviations: AFLP, amplified restriction fragment polymorphism; (e)LRR, (extracellular) leucine-rich repeat; NB, nucleotidebinding; ARC, Apaf-1, R proteins, and CED4 homology; R, resistance; TIR, Toll and interleukin-1 receptors homology; TLR, Toll-like receptor; LZ, leucine zipper; CARD, caspase recruitment domain; GUS, β-glucuronidase. * Present address: School of Biological Sciences, University of Southampton, Southampton SO16 5YA, U.K. † Present address: Aventis CropScience, B-9000 Gent, Belgium. ‡ To whom reprint requests should be addressed. E-mail: [email protected].



8808

Fig. 1. Tomato Cf proteins and structurally related proteins. (A) Cf proteins contain different numbers of N-terminal eLRRs, a transmembrane (TM) domain, and a short cytoplasmic domain (CD). (B) Rice Xa-21 and tomato Pto bacterial R proteins. (C) eLRR-TM domain-containing proteins with different cytoplasmic domains (CD). See text for descriptions. Both the Cf class and Xa21 resemble the Drosophila Toll receptor and its human homologs, the Toll-like receptors (TLRs) (18). TLR2 and TLR4 (Fig. 1C) function in innate immunity and activate microbial defense pathways to trigger inflammatory responses after recognition of conserved molecular structures of distinct microbial pathogen classes (19). TLRs have therefore been designated “pattern recognition receptors” (20) and serve an analogous role to plant R proteins in detecting pathogen molecules and activating defense pathways. Interestingly, TLRmediated inflammatory responses resemble plant defense responses, including production of antimicrobial active oxygen species and nitric oxide and ultimately cell death (21, 22). The significance of these homologies for plant R protein mechanisms has yet to be established, but it is tempting to suggest that, whereas vertebrates use the TLRs to recognize conserved structures of different pathogen classes, plants have evolved these to detect strain-specific pathogen Avr products. Arabidopsis RPP1 and RPP5 genes confer resistance to P. parasitica and belong to the largest class of R genes that encode nucleotidebinding leucine-rich repeat (NB-LRR) proteins (Fig. 2A). Genome sequencing has shown that approximately 200 NB-LRR-encoding genes are present in Arabidopsis (23). NBLRR proteins can be divided into two subclasses on the basis of their N-terminal domain. The leucine zipper (LZ)-NB-LRR is a broad class of NB-LRR proteins (with perhaps two subclasses) that contains an N-terminal putative heptad LZ or coiled-coil domain (23). Members of this class have been identified for resistance to bacteria, viruses, fungi, oomycetes, and even nematodes and aphids. Examples include Arabidopsis RPS2, RPM1, and RPS5 for resistance to P. syringae (24), RPP8 for resistance to P. parasitica (25), and maize Rp1 for resistance to Puccinia sorghi (26). Members of the TIR-NB-LRR class carry N-terminal homology with the cytoplasmic domain of the Toll and interkeukin-1 receptors (Fig. 2B) involved in vertebrate innate immunity (27, 28). Besides RPP1 and RPP5, this class also includes tobacco N for resistance to tobacco mosaic virus, flax L6 for resistance to flax rust, and Arabidopsis RPS4 for resistance to P. syringae (24, 29). After activation, the TIR domain of human TLR2 and TLR4 binds, through homophilic interactions, to the TIR domain of the MyD88 adaptor protein (Fig. 2B) (30). MyD88 also carries a death domain (DD) that then binds to the DD of the Ser/Thr protein kinase IRAK (the human ortholog of Drosophila Pelle, and interestingly, also homologous to tomato Pto; Fig. 1B), leading to the translocation of the transcription factor NF- B and induction of the inflammatory response (27). These homologies may suggest a role for the TIR domain of the plant R proteins in homophilic TIR–TIR interactions with TIR domaincontaining plant adaptor proteins, but these have yet to be reported in plants.

Fig. 2. Arabidopsis RPP1, RPP5, and structurally related proteins. (A) RPP1 and RPP5 have an N-terminal TIR domain, a nucleotide-binding Apaf-1, R proteins and CED4 homology (NB-ARC) domain, and a LRR domain. The RPP1 family differ by their N-terminal domains, which are absent in RPP5: RPP1A has a putative signal anchor (SA) domain, RPP1B,C have hydrophobic domains (HD). (B) Human proteins that function in the NF-κB pathway. eLRR, extracellular LRR domain; TM, transmembrane domain. (C) Human proteins with CARDs that function in NF-κB and/or apoptosis pathways. See text for descriptions. An intriguing homology has been noted between plant NB-LRR proteins and the apoptotic adaptor CED4 from Caenorhabditis elegans (31). This region of homology also extends to its human ortholog Apaf-1 (Fig. 2C) (32) and has been designated the NB-ARC domain (Fig. 2C) (33) or Ap-ATPase domain (34). The NB-ARC domain is an ancient motif, because it is also present in proteins of Gram-positive bacteria (34). Another protein that resembles plant NB-LRR proteins is human caspase recruitment domain (CARD)4/Nod1 (35, 36), which contains LRRs and a NB site (the homology to the rest of the NB-ARC domain is less convincing). Instead of TIR or LZ domains at their N termini, Apaf-1 and CARD4/Nod1 carry CARDs (Fig. 2C). Induction of apoptosis by both Apaf-1 and CARD4/Nod1 occurs through homophilic CARD–CARD interactions with caspase-9, a CARD-containing cysteine protease (36, 37). CARD4/Nod1, in addition, plays a role in activation of the NF- B pathway through homophilic CARD–CARD interactions with the Ser/Thr kinase RICK (Fig. 2C) (35, 36). Caspases and CARDs have not yet been reported in plants; however, the use of peptide inhibitors specific to caspases in animal cells suggests plants may possess caspase activity (38). Although the significance of these homologies is not clear, they provide interesting paradigms for NB-LRR protein mechanisms. For example, the WD40 repeat domain of Apaf-1 negatively regulates its NBARC domain, which is alleviated after binding to cytochrome c, then allowing ATP-dependent oligomerization and caspase-9 activation (37). By analogy, pathogen Avr signals perceived by the LRR domain of plant NB-LRR proteins might alleviate negative regulation of their NB-ARC domain such that ATP hydrolysis and oligomerization can be initiated, possibly leading to homo- or heterodimerization (through



8809

homophilic interactions) of the N-terminal TIR or LZ domains.

R GENE EVOLUTION Plants can activate both localized and systemic defense mechanisms in response to pathogen infection. This activation is often a consequence of R protein perception of a pathogenencoded Avr determinant in cells targeted by the pathogen (24). This contrasts with the situation in mammals where circulating defender cells migrate to the sites of pathogen ingress. Furthermore, in mammals, the elaboration of recognition capacity and the consequent resistance to an array of potential pathogens occurs somatically through the adaptive immune system and germinally in genes encoding the MHC (39). In plants, R gene sequence evolution takes place exclusively through the germline. Population geneticists have proposed that in natural populations, where plants and their pathogens coevolve, disease is kept in check by “balancing polymorphisms” at R gene loci (40). In this model, no single R gene allele is present at high frequency because, as a result of selection pressure, it may be overcome by a novel pathogen variant (40). For example, R gene monocultures in crops eventually succumb to pathogen infection, and breeding durable disease resistance remains a major challenge. Overdominance, or heterozygote advantage, has been proposed to explain the polymorphism at the human MHC locus. However, overdominance cannot explain polymorphism at R gene loci in inbreeding species. We and others have proposed that frequency-dependent selection may account for the maintenance of sequence polymorphism at the Arabidopsis RPP5, RPM1, and other R gene loci (41, 42). Recent analyses have revealed the molecular organization of a number of R gene loci and provided insight to the molecular mechanisms that generate sequence diversity. We have analyzed two Cf loci in tomato and two RPP loci in Arabidopsis. The Cf-0, Cf-9, and Cf-4 haplotypes of tomato contain distinct complements of tandemly duplicated Cf-9 homologs (Hcr9s; refs.43 and 44). Several Hcr9s have been shown to confer resistance to C. fulvum through recognition of distinct Avr determinants (43, 45). A similar comparison has revealed the polymorphic nature of Cf-2 homologs (Hcr2s) at the tomato Cf-2/Cf-5 locus (46). The sequence polymorphism at Cf loci is because of interspecific variation (12). However, in Arabidopsis the RPP1 and RPP5 loci (41, 47), which are also comprised of linked multigene families, show pronounced intraspecific polymorphism when different landraces are compared by DNA gel blot analysis (Fig. 3). DNA sequence analysis of tomato Hcr9s and Hcr2s (43, 46), the Arabidopsis RPP1, RPP5, and RPP8 families (25, 41, 47, 48) and NB-LRR homologs at the lettuce Dm3 locus (49) have recently been reported. Like RPP5 and RPP1 homologs, Hcr2s contain variable numbers of LRRs, suggesting this is an important mechanism to generate novel recognition specificities (12, 46). Hcr9s, however, contain similar numbers of LRRs, and most sequence variation occurs in nucleotides that encode the putative solvent-exposed amino acids of a conserved β-strand/β-turn structural motif of LRR proteins. These amino acids are thought to be critical for the recognition specificity of ligand binding in LRR proteins (50). Sequence comparisons of NB-LRR multigene families such as RPP1, RPP5, and Dm3 have also shown high ratios of nonsynonymous to synonymous substitutions in sequences encoding the corresponding amino acids of their LRRs. These sequences within the LRRs have clearly undergone diversifying selection, whereas sequences encoding putative signaling domains have undergone purifying selection (23, 41, 47). As in studies on the evolution of the mammalian MHC locus (39), the relative contribution of interallelic recombination, gene conversion, unequal crossing over, and the accumulation of point mutations in generating R gene diversity is controversial. Classical genetic experiments suggested that interallelic sequence exchange at the flax L locus (51) or chromosomal mispairing and unequal crossing over between homologs at the maize Rp1 locus (52) could generate R gene variation. This was subsequently verified by molecular analyses of the L and Rp1 loci (26, 51). At some R gene loci, a “birth-and-death” model of R gene evolution has been proposed (53), based on a model for evolution of the mammalian MHC locus (54). In this model, expansion or contraction of members of a multigene family is caused by unequal crossing over, and evolution of alleles proceeds by sequence exchange between gene orthologs, random mutation and selection.

Fig. 3. Polymorphic R gene families revealed by DNA gel blot analysis. Comparisons of tomato lines with introgressed chromosome segments containing different Cf genes show interspecific polymorphisms (44). Comparisons of different Arabidopsis landraces show intraspecific polymorphisms at the RPP1 (47) and RPP5 loci (48). A comparison of the Cf-0, Cf-4 and Cf-9 haplotypes in tomato showed that the copy number of Hcr9s at this locus can vary. In the Cf-4 and Cf-9 haplotypes (each containing five Hcr9s), evidence for extensive intergenic sequence exchange between Hcr9 paralogs was shown that generated recombinant Hcr9s. This sequence exchange could have contributed to the creation of novel recognition specificities. Also, the order of putative gene orthologs differed in the two haplotypes (43) and is inconsistent with the proposal that negligible sequence exchange occurs between paralogs. Hcr9s have also accumulated multiple point mutations that generate sequence diversity in the solvent-exposed residues of their LRRs (see above), and this appears to be another important source of sequence variation. The comparison of Col-0 and La-er haplotypes of the Arabidopsis RPP5 locus also revealed extensive sequence exchange between RPP5 homologs (41). With the exception of homologs at the ends of the cluster, orthologous genes were difficult to discern. Furthermore, many “mutated” homologs were identified. It is possible that these sequences provide a reservoir for intergenic sequence exchange and the generation of sequence diversity. It is likely that the rate and molecular mechanism for generating R gene novelty will depend on a number of factors. Further insight into these mechanisms will come from comparative analysis of natural R gene variants such as the numerous recombinant Rp1 alleles in maize (26, 52). These analyses should reveal the range of molecular mechanisms that generate R gene diversity in different plants, at different R gene loci within plants,



8810

and the molecular basis for R protein-mediated pathogen perception.

Fig. 4. Infection phenotypes of rcr3 mutants. Seedlings were immersed in a solution containing 3.5 × 105 ml−1 spores of the appropriate C. fulvum race. Inoculated plants were kept for 3 to 6 days in a phytochamber at saturated humidity and subsequently at 70% relative humidity, 24°C, during the 16-hr day-light, 18°C overnight. (A) Comparison of infection phenotype of the rcr3 mutants and the Cf0-sensitive and Cf2-resistant control lines. Plants were inoculated with C. fulvum race 5 15 days after sowing, and leaves were photographed 17 days after inoculation. Fungal colonization is visible on the abaxial surface of the leaves as white mycelium. (B) Comparison of fungal biomass accumulation in rcr3–1 (X), rcr3–2 (), rcr3–3 (), Cf0 (), and Cf2 () lines by using a fluorometric GUS assay (56) over a time course after inoculation. Twelve-day-old plants were inoculated with C. fulvum race 4 GUS (57). No GUS activity (4-methyl umbelliferone mg−1 protein min−1) was detected on Cf2 controls in contrast to Cf0 controls. Three cotyledons were harvested and analyzed separately for each sample at each time point (the mean value for each set of triplicate samples is shown and error bars indicate the standard deviation). RESULTS AND DISCUSSION Mutational Analysis Identifies Rcr3, a Gene Required for Cf-2 Function. Mutational analysis is a powerful tool to dissect signaling pathways and has been used to identify genes required for R gene function. Mutational analysis of genes required for Cf-9 function has revealed two loci required for full C. fulvum resistance, Rcr1 and Rcr2 (55). Here, we report the results of mutagenesis of the tomato Cf2 line. Twenty-five M2 seeds from 3,200 diepoxybutane- or ethyl methanesulfonate-treated Cf2 M1 plants were screened for their response to C. fulvum race 4 βglucuronidase (GUS) infection, and four M2 families that segregated for resistance or sensitivity to infection were identified (Table 1). PCR analysis of disease-sensitive plants confirmed these plants contained Cf-2. To determine the genetic basis of disease sensitivity in these mutants, all four were crossed to the Cf0 near-isogenic line that lacks known resistance genes to C. fulvum. In crosses to Cf0, mutants at Cf-2 should give susceptible F1 progeny as neither parent would have contained a functional gene for resistance to C. fulvum infection. Crosses between the four mutant lines and Cf0 gave resistant F1 progeny, confirming the mutations were not in Cf-2. Progeny from intercrosses between all four mutants were sensitive to C. fulvum race 4 GUS infection, demonstrating that these mutations are allelic. This gene has been designated Rcr3 (required for C. fulvum resistance 3; Table 1).

Fig. 5. Comparison of fungal development and plant cell death response during the C. fulvum infection of rcr3 mutants and control lines. At least three cotyledons and two leaves were sampled every 3 days from each genotype during the experiment described in Fig. 4A and were stained with lactophenol-trypan blue (58). Photomicrographs of cotyledons were taken 7 days after inoculation at the same magnification. Scale bar = 100 µm. (A) Fungal growth in the Cf0 line: C. fulvum hyphae (h) tend to localize close to plant veins (v) where they start swelling, (B) Characteristic early resistance response in the Cf2 line: patches of dead mesophyll cells (p) develop at the vicinity of the plant veins, whereas the fungus is undetectable. (C) Infection phenotype of rcr3–3, and rcr3–2 (not shown): once the fungus has penetrated the leaf apoplast, the fungus develops as in Cf0. However, these fungal foci appear to occur at a lower frequency on rcr3–3 and rcr3–2 than on Cf0. (D) rcr3–1 intermediate phenotype: although rcr3–1 allows sparse fungal growth in some areas; the fungus remains undetectable on the major part of the cotyledon. Discrete dead cells (d) were scattered around the upper mesophyll of the mutants but not in wild type, irrespective of C. fulvum infection (noninoculated material not shown). This cell death phenotype was poorly reproducible. Calcium oxalate crystals (c) in some mesophyll cells are common in tomato leaves. These phenotypes were observed in three independent infections with C. fulvum race 5 and in an infection with C. fulvum race 4 GUS, where over 300 infection sites were analyzed in cotyledon and leaf samples. Strong and Weak Mutant Alleles of Rcr3. Inoculation with C. fulvum race 5 revealed that rcr3–2 (the line carrying the rcr3–2 mutation) and rcr3–3 allowed as much fungal growth as Cf0, whereas rcr3–1 appeared significantly less susceptible (Fig. 4). From 2 weeks after inoculation, C. fulvum growth on rcr3–1 was intermediate to Cf0 and Cf2 (Fig. 4B). Fungal development was barely detectable on rcr3–1 at 1 week after inoculation, whereas hyphae had already invaded the mesophyll layers and were proliferating around the vascular tissue of rcr3–2, rcr3–3, and Cf0 cotyledons (Fig. 5). Microscopic inspection of 5-bromo-4-chloro-3-indolyl-β-D-glucuronic acidstained tissue confirmed that the fungus was able to sporulate on all three mutants (not shown). The rcr3–1, rcr3–2, and rcr3–3 mutants were each back crossed



8811

to Cf 2 and selfed to produce F2 seed. Approximately one-quarter of the F2 progeny from crosses between Cf2 and rcr3–2 or rcr3–3 appeared to be fully susceptible to C. fulvum race 4 GUS infection. One-quarter of the F2 progeny from the Cf2 × rcr3–1 cross showed an intermediate level of disease sensitivity as observed in the original M2 seedlings, suggesting this phenotype is caused by allelic variation at rcr3 and not by another unlinked mutation (data not shown). Therefore, rcr3–1 is a weak, and rcr3–2 and rcr3–3 are strong, suppressors of Cf-2 function. Rcr3 Maps to Chromosome 2. To generate a mapping population where rcr3 and molecular markers were segregating, rcr3–1 was crossed to Lycopersicon pennellii (59). F1 progeny from this cross were resistant to C. fulvum race 4 GUS infection, demonstrating that the L. pennellii Rcr3 allele can complement the rcr3 mutation. A single F1 plant was back crossed to rcr3–1 to give back cross (BC1) individuals that segregated 1:1 for resistance or susceptibility to C. fulvum race 4 GUS infection. DNA gel blot analysis was used to screen resistant BC1 plants for Cf-2 homozygotes, and one was used as male parent in a back cross to rcr3–1. In the resulting BC2 population, all plants were homozygous for Cf-2 and, as expected, segregated 1:1 for resistance or susceptibility to C. fulvum race 4 GUS infection. The chromosomal location of Rcr3 was determined by amplified restriction fragment polymorphism (AFLP) analysis (60) of bulked segregant pools. Fifty-eight BC2 individuals were inoculated with C. fulvum race 4 GUS and screened for resistance or susceptibility to infection. DNA was extracted from pools of 24 resistant and 34 susceptible plants and subjected to AFLP analysis. Several AFLP markers linked to the L. pennellii Rcr3 allele were identified (Fig. 6A). AFLP analysis of L. esculentum × L. pennellii introgression lines (61) showed these markers were located on tomato chromosome 2 (Fig. 6B). Rcr3 Is Required Specifically for Cf-2 Function. To determine whether Rcr3 is required for the function of other Cf genes, rcr3 mutants were crossed to Cf5 and Cf9 near isogenic lines. Rcr3 will segregate independently of Cf-5 and Cf-9, because they map to different chromosomes (12). In a cross between Cf9 and rcr3 lines, F2 progeny should segregate 3 resistant to 1 susceptible when inoculated with C. fulvum race 2,4 if Rcr3 is not required for Cf-9 function (C. fulvum race 2,4 lacks Avr2 and Avr4 and is virulent on tomato lines carrying either Cf-2 or Cf-4, so in a cross between Cf9 and rcr3, Cf-2 confers no effective resistance to C. fulvum race 2,4). If Rcr3 is required for Cf-9 function, the progeny should segregate 9 resistant to 7 susceptible when inoculated with C. fulvum race 2,4. Progeny from this cross segregated 3 resistant to 1 susceptible, demonstrating Rcr3 is not required for Cf-9-mediated resistance (Table 2). Further confirmation of this result was obtained from analysis of backcross progeny. As predicted, the progeny segregated 1:1 for resistance and susceptibility to C. fulvum race 2,4 infection (Table 2), confirming that Rcr3 is not required for Cf-9 function. DNA gel blot analysis confirmed that none of the susceptible progeny contained Cf-9 (data not shown). Table 1. C. fulvum-sensitive Cf2 mutants Mutant allele rcr3-1* rcr3-2* rcr3-3† rcr3-4†

Mutagen DEB‡ EMS§ EMS EMS

Frequency¶ 5/23 2/22 5/23 4/23

* Identified with fluorometric GUS assay. † Identified by visual inspection. ‡ Diepoxybutane. § Ethyl methanesulfonate. ¶ Number of susceptible individuals in total M family. 2

Fig. 6. Rcr3 maps to chromosome 2. AFLP analysis was carried out as previously described (59). (A) Identification of AFLP markers linked to Rcr3 (arrows). Susceptible (S) and resistant (R) pools of the BC2 mapping populations were subjected to AFLP analysis. Pairs of lanes (1 to 7) represent analysis with different AFLP primer combinations. (B) Mapping of a linked AFLP marker by using L. pennellii introgression lines (1–2 to 4–4) with primer combination 4. Susceptible and resistant pools of the BC2 mapping populations were analyzed as controls. The arrows indicate the AFLP fragment linked to L. pennellii Rcr3 present in introgression lines 2–3 and 2–4 that localizes that marker to chromosome 2. A similar analysis was used to determine whether Rcr3 is required for Cf-5-mediated resistance. Cf-5 and Cf-2 are allelic, and their products are more than 90% identical (46). F2 progeny from a cross between rcr3–2 and Cf5 were screened by inoculation with C. fulvum race 2,4 and segregated at a ratio of 3 resistant to 1 susceptible (Table 2), demonstrating Rcr3 is not required for Cf-5-mediated resistance. PCR analysis confirmed that all C. fulvum race 2,4 susceptible plants lacked Cf-5 (data not shown). To confirm the authenticity of the cross, the same F2 plants were reinoculated with C. fulvum race 5 (a race that lacks Avr5 and is virulent on lines carrying only Cf-5, so only Cf-2 will confer resistance to infection in this test). The segregation ratio of resistant to sensitive progeny was close to a 9:7 ratio predicted on the basis of independent segregation of Cf-2 and Rcr3 (Table 2). Cf Protein Function. One approach to understanding Cf protein function is to analyze the biochemical and physiological re Table 2. Effect of rcr3 mutations on Cf-5 and Cf-9 functions Segregation Cf line Cross R:S Cf9* F2‡ 41:17 Cf9* BC1‡ 16:10 Cf5† F2¶ 91:32 Cf5† F2§ 61:52 S, susceptible; R, resistant. * Crosses with rcr3-1. † Crosses with rcr3-2. ‡ Screened with C. fulvum race 2,4. § Screened with C. fulvum race 5. ¶ Significance level at P = 0.05 for a χ2 with one degree of freedom is 3.84.

Goodness of fit (χ2)¶ 1:1 3:1 – 0.57 1.38 18.5 – 0.068 – 25

9:7 4.9 – 15.7 0.24



8812

sponses that are activated on Avr-dependent elicitation. R proteins appear to activate similar cellular responses that are effective against a broad range of plant pathogens. For example, in transgenic tobacco, Cf-9 can confer resistance to Potato Virus X expressing Avr9 (62). We have defined five main Avr9-dependent responses that ensue on elicitation in tobacco Cf-9 plants and cell cultures. These include stimulation of ion fluxes (63), production of active oxygen species (64), activation of two mitogen-activated protein kinases (65), and a calciumdependent protein kinase (66). We have also defined a large set of rapidly induced genes (67) and are currently trying to identify plasma membrane proteins that interact with epitope-tagged Cf-9 in tobacco (13). Another approach is to dissect the signaling pathway of Cf-mediated resistance by using mutational analysis. Previously we identified two genes (Rcr1 and Rcr2) that are required for full Cf-9 function (55). In the present study, mutations in Rcr3, a gene required for Cf-2 function, were identified. No Cf-2 mutants were recovered because the Cf2 line carries two functional copies of the gene (68). Four mutations in Rcr3 were identified, and no mutations were identified in other genes, indicating that the genetic screen for suppressors of Cf-2 function may be saturated. The lack of additional mutants in the Cf-2-mediated signaling pathway may reflect functional redundancy or lethality. Genetic analysis demonstrated that Rcr3 is not required for Cf-9 function, therefore Rcr3 is not a component of a conserved Cf signal transduction pathway. We suggested previously that Cf-2 and Cf-5 may interact with a common partner to activate an Avr-dependent plant defense response (12, 46). Because Rcr3 is not required for Cf-5 function, Rcr3 cannot be a Cf-2/Cf-5 common interacting partner. In Arabidopsis, several genes appear to be required for the function of either of the two classes of NB-LRR genes. EDS1 is required for the function of the TIR-NB-LRR genes RPP5, RPP1, and RPS4, whereas NDR1 and Pbs2 are required for the function of the LZ-NB-LRR genes RPS2, RPS5 and RPM1 (29, 69, 70). These studies appeared to identify at least two parallel signaling pathways for R genes from the two different NB-LRR classes. Pbs3 is required for the function of RPS2, RPS5 and RPM1 (all LZ-NB-LRR class) and also RPS4 (TIR-NB-LRR class), thus establishing a possible convergence point for these pathways (70). Another gene, Pbs1, has been identified that is specifically required for RPS5 (70). Rcr3 and Pbs1 are therefore unique because they appear to suppress the function of single R genes. These analyses have revealed additional genetic complexity in the pathogen perception mechanism of plants. There are several possible roles for Rcr3 in Cf-2/Avr2-mediated resistance to C. fulvum. Rcr3 could be specifically required for the expression, stability, or modification of Cf-2. Rcr3 may also play a role in the perception of Avr2 or in its proteolytic processing/modification to generate a mature ligand that can be recognized by Cf-2. It is known that maturation of Avr9 and Avr4 requires the activity of both fungal and plant proteases (71, 72). Proteolytic processing of ligand molecules is common in other systems, e.g., processing of the Spätzle ligand of the Drosophila Toll receptor (73). However, if Rcr3 functions solely as an extracellular protease or modifying enzyme, infiltrating cotyledons of rcr3 mutants with intercellular fluids isolated from a compatible tomato–C. fulvum race 5 interaction should restore the characteristic chlorotic response observed in Cf2 plants (74). No complementation of the mutant phenotype was observed, suggesting Rcr3 does not function as an extracellular modifying enzyme (data not shown). It is possible that Cf-2 and Cf-5 utilize independent signaling pathways to activate the plant defense response, and that Rcr3 is a component of the Cf-2-dependent signaling pathway, downstream of Avr2 perception. An analogous situation may occur in barley, where at least two distinct classes of Mla alleles can be identified on the basis of their requirement for the Rar1 gene (75). Alternatively, Rcr3 may interact directly with Cf-2 to function as a signaling partner. A paradigm for Cf protein function was suggested from studies on three genes (CLV1, CLV2, and CLV3) that condition the Arabidopsis “clavata” mutant phenotype (76, 77 and 78). CLV1 is an Xa21-like extracellular receptor kinase, and CLV2 is an extracellular membrane anchored Cf-like protein (Fig. 1C). CLV1 and CLV2 are proposed to form a heterodimer that may be activated on binding of a peptide ligand (CLV3) resulting in phosphorylation of the kinase domain of CLV1 and association with downstream factors such as a kinase-associated protein phosphatase, KAPP, and the small GTPase, Rop (77, 79). Likewise, the truncated rice Xa21D protein for resistance to X. campestris pv. oryzae may signal through a full-length Xa21 homolog (80). A similar model has been postulated for the sporophytic selfincompatibility mechanism in Brassica species to explain the requirement of SLG, a secreted S-locus glycoprotein, and SRK, an S-locus receptor kinase that contains an extracellular SLG domain. We with others have investigated whether Cf-9 and Avr9 directly interact with each other. These experiments have failed to reveal a direct interaction between Avr9 and Cf-9, suggesting that other factors are required for Avr9 perception (J.D.G.J., P. J. G. M. De Wit & T. Nürnberger, unpublished work). These factors and other proteins required for Cf-function are conserved in solanaceous plants because Cf-9 confers responsiveness to Avr9 in tobacco and potato (81) and Cf-2 to Avr2 in tobacco (M.S.D., unpublished work). Furthermore, studies using labeled Avr9 have identified a high-affinity binding site on plasma membranes of tomato and tobacco that is present irrespective of their Cf-9 genotype (82). Similarly in soybean, a 34-kDa protein has been identified that binds syringolide elicitors, avirulence factors recognized by Rpg4, irrespective of the Rpg4 genotype (83). By analogy, Avr2 may not interact directly or exclusively with Cf-2 but may interact with Rcr3. Another example where accessory proteins are required for full function of recognition proteins is provided by the mammalian CD14 surface protein. CD14 contributes to the perception of microbial products and acts in concert with specific TLRs that discriminate between pathogens and initiate transmembrane signaling (84). Pathogen-encoded (a)virulence factors could interact with host proteins to modify their functions to access nutrients or to suppress defense mechanisms. For example, the human pathogen Yersinia uses a type III secretion system, in common with many plant pathogenic bacteria (2), to deliver YopJ into host cells to suppress host defense mechanisms (85). R proteins may have evolved to specifically recognize the physical association of pathogen-encoded virulence factors with their plant cellular targets to subsequently activate defense mechanisms, i.e., R proteins may “guard” host proteins and initiate an Avr-dependent defense response (86, 87). The physical interaction of Pto with AvrPto provides the only published evidence for an Avr product binding to its corresponding R protein (16, 17). We have proposed that Prf, a NB-LRR protein required for Pto function (88), might best be thought of as an R protein that recognizes the AvrPto/Pto pathogenicity complex to activate a resistance response (86). A similar model has been proposed to account for yeast two-hybrid interactions between RPP5 and an Arabidopsis RelA/SpoT homolog (87). Accordingly, Rcr3 may be the pathogenicity target of AVR2, and Cf-2 could specifically “guard” Rcr3. The tendency for a lower disease sensitivity observed 10 days after C. fulvum infection in rcr3–2 and rcr3–3 mutants compared with the Cf0 susceptible control (Fig. 4B) could be caused by absence of the Rcr3 pathogenicity target. The characterization of a Cf0/rcr3 mutant will reveal whether Rcr3 is necessary for full pathogenicity of C. fulvum. The Guard hypothesis predicts that for each R protein there is both a corresponding pathogen Avr product and a host target. Such a model would explain the dual recognition capacity of some NB-LRR proteins such as RPM1 and Mi-1 if they “guard” the same host component targeted by unrelated Avr products. Evolutionary mechanisms sustaining R gene diversity are essential for the plant to be able to detect distinct pathogen (a)virulence products that target R protein-“guarded” host components. All members of the Jones lab are thanked for useful discussion. We thank Sara Perkins, Margaret Shailer, and Justine Campling for their excellent horticultural service. This work was supported in part by the United Kingdom Biotechnology and Biological Sciences Research Council (M.S.D.) and by a Gatsby Ph.D. studentship (C.G.). The Sainsbury Laboratory is supported by the Gatsby Charitable Foundation.



8813

1. Flor, H. H. (1971) Annu. Rev. Phytopathol. 9, 275–296. 2. Galan, E. J. & Collmer, A. (1999) Science 284, 1322–1328. 3. Kearney, B. & Staskawicz, B. J. (1990) Nature (London) 346, 385–386. 4. Ritter, C. & Dangl, J. L. (1995) Mol. Plant-Microbe Interact. 8, 444–453. 5. Erickson, F. L., Holzberg, S., Calderon-Urrea, A., Handley, V., Axtell, M., Corr, C. & Baker, B. (1999) Plant J. 18, 67–75. 6. Bendahmane, A., Kohm, B. A., Dedi, C. & Baulcombe, D. C. (1995) Plant J. 8, 933–941. 7. Van den Ackerveken, G. F. J. M., Van Kan, J. A. L. & De Wit, P. J. G. M. (1992) Plant J. 2, 359–366. 8. Joosten, M. H. A. J., Cozijnsen, T. J. & De Wit, P. J. G. M. (1994) Nature (London) 367, 384–386. 9. Lauge, R., Joosten, M. H. A. J., Van den Ackerveken, G. F. J. M., Van den Broek, H. W. J. & De Wit, P. J. G. M. (1997) Mol. Plant-Microbe Interact. 10, 725–734. 10. Lauge, R., Joosten, M. H. A. J., Haanstra, J. P. W., Goodwin, P. H., Lindhout, P. & De Wit, P. J. G. M. (1998) Proc. Natl Acad. Sci. USA 95, 9014–9018. 11. Koch, E. S. & Slusarenko, A. (1990) Plant Cell 2, 437–445. 12. Thomas, C. M., Dixon, M. S., Parniske, M., Golstein, C. & Jones, J. D. G. (1998) Philos. Trans. R. Soc. London 353, 1413–1424. 13. Piedras, P., Rivas, S., Droge, S., Hillmer, S. & Jones, J. D. G. (2000) Plant J., 21, 529–536. 14. Song, W.-Y., Wang, G.-L., Chen, L.-L., Kirn, H.-S., Pi, L.-Y., Holsten, T., Gardner, J., Wang, B., Zhai, W.-X., Zhu, L.-H., et al. (1995) Science 270, 1804– 1806. 15. Bogdanove, A. J. & Martin, G. B. (2000) Proc. Natl Acad. Sci. USA 97, 8836–8840. 16. Scofield, S. R., Tobias, C. M., Rathjen, J. P., Chang, J. H., Lavelle, D. T., Michelmore, R. W. & Staskawicz, B. J. (1996) Science 274, 2063–2065. 17. Tang, X., Frederick, R. D., Zhou, J., Halterman, D. A., Jia, Y. & Martin, G. B. (1996) Science 274, 2060–2063. 18. Rock, F. L., Hardiman, G., Timans, J. C., Kastelein, R. A. & Bazan, J. F. (1998) Proc. Natl. Acad. Sci. USA 95, 588–593. 19. Janeway, C. A. & Medzhitov, R. (1999) Curr. Biol. 9, R879–R882. 20. Medzhitov, R. & Janeway, C. A. (1997) Curr. Opin. Immunol. 9, 4–9. 21. Brightbill, H. D., Libraty, D. H., Krutzik, S. R., Yang, R. B., Belisle, J. T., Bleharski, J. R., Maitland, M., Norgard, M. V., Plevy, S. E., Smale, S. T., et al. (1999) Science 285, 732–736. 22. Aliprantis, A. O., Yang, R. B., Mark, M. R., Suggett, S., Devaux, B., Radolf, J. D., Klimpel, G. R., Godowski, P. & Zychlinsky, A. (1999) Science 285, 736– 739. 23. Meyers, B. C., Dickerman, A. W., Michelmore, R. W., Sivaramakrishnan, S., Sobral, B. W. & Young, N. D. (1999) Plant J. 20, 317–332. 24. Hammond-Kosack, K. E. & Jones, J. D. G. (1997) Annu. Rev. Plant Physiol. Plant Mol. Biol. 48, 575–607. 25. McDowell, J. M., Dhandaydham, M., Long, T. A., Aarts, M. G. M., Goff, S., Holub, E. B. & Dangl, J. L. (1998) Plant Cell 10, 1861–1874. 26. Collins, N., Drake, J., Ayliffe, M., Sun, Q, Ellis, J., Hulbert, S. & Pryor, T. (1999) Plant Cell 11, 1365–1376. 27. O'Neill, L. A. J. & Greene, C. (1998) J. Leukocyte Biol. 63, 650–657. 28. Hoffmann, J. A., Kafatos, F. C., Janeway, C. A. & Ezekowitz, R. A. B. (1999) Science 284, 1313–1318. 29. Gassmann, W., Hinsch, M. E. & Staskawicz, B. J. (1999) Plant J. 20, 265-277. 30. Medzhitov, R., Preston-Hurlburt, P., Kopp, E., Stadlen, A., Chen, C. Q., Ghosh, S. & Janeway, C. A. (1998) Mol. Cell 2, 253–258. 31. Chinnaiyan, A. M., Chaudhary, D., O'Rourke, K., Koonin, E. V. & Dixit, V. M. (1997) Nature (London) 388, 728–729. 32. Zou, H., Henzel, W. J., Liu, X. S., Lutschg, A. & Wang, X. D. (1997) Cell 90, 405–413. 33. Van der Biezen, E. A. & Jones, J. D. G. (1998) Curr. Biol. 8, 226–227. 34. Aravind, L., Dixit, V. M. & Koonin, E. V. (1999) Trends Biochem. Sci. 24, 47–53. 35. Bertin, J., Nir, W. J., Fischer, C. M., Tayber, O. V, Errada, P. R., Grant, J. R., Keilty, J. J., Gosselin, M. L., Robison, K. E., Wong, G. H. W., et al. (1999) J. Biol. Chem. 274, 12955–12958. 36. Inohara, N., Koseki, T., Del Peso, L., Hu, Y. M., Yee, C., Chen, S., Carrio, R., Merino, J., Liu, D., Ni, J. & Nunez, G. (1999) J. Biol. Chem. 274, 14560– 14567. 37. Hu, Y. M., Benedict, M. A., Ding, L. Y. & Nunez, G. (1999) EMBO J. 18, 3586–3595. 38. Del Pozo, O. & Lam, E. (1998) Curr. Biol. 8, 1129–1132. 39. Hughes, A. L. & Yeager, M. (1998) Annu. Rev. Genet. 32, 415–435. 40. Leonard, K. J. (1997) in The Gene-for-Gene Relationship in Plant-Parasite Interactions, eds. Crute, I. R., Holub, E. B. & Burdon, J. J. (CAB, Wallingford, U. K.), pp. 211–230. 41. Noel, L., Moores, T. L., Van der Biezen, E. A., Parniske, M., Daniels, M. J., Parker, J. E. & Jones, J. D. G. (1999) Plant Cell 11, 2099–2012. 42. Stahl, E. A., Dwyer, G., Mauricio, R., Kreitman, M. & Bergelson, J. (1999) Nature (London) 400, 667–671. 43. Parniske, M., Hammond-Kosack, K. E., Golstein, C, Thomas, C. M., Jones, D. A., Harrison, K., Wulff, B. B. H. & Jones, J. D. G. (1997) Cell 91, 821–832. 44. Thomas, C. M., Jones, D. A., Parniske, M., Harrison, K., Balint-Kurti, P. J., Hatzixanthis, K. & Jones, J. D. G. (1997) Plant Cell 9, 2209–2224. 45. Takken, F. L. W., Thomas, C. M., Joosten, M., Golstein, C., Westerink, N., Hille, J., Nijkamp, H. J. J., De Wit, P. J. G. M. & Jones, J. D. G. (1999) Plant J. 20, 279–288. 46. Dixon, M. S., Hatzixanthis, K., Jones, D. A., Harrison, K. & Jones, J. D. G. (1998) Plant Cell 10, 1915–1925. 47. Botella, M. A., Parker, J. E., Frost, L. N., BittnerEddy, P. D., Beynon, J. L., Daniels, M. J., Holub, E. B. & Jones, J. D. G. (1998) Plant Cell 10, 1847–1860. 48. Parker, J. E., Coleman, M. J., Szabo, V., Frost, L. N., Schmidt, R., Van der Biezen, E. A., Moores, T., Dean, C., Daniels, M. J. & Jones, J. D. G. (1997) Plant Cell 9, 879–894. 49. Meyers, B. C., Shen, K. A., Rohani, P., Gaut, B. S. & Michelmore, R. W. (1998) Plant Cell 10, 1833–1846. 50. Kobe, B. & Deisenhofer, J. (1994) Trends Biochem. Sci. 19, 415–421. 51. Ellis, J., Lawrence, G., Ayliffe, M., Anderson, P., Collins, N., Finnegan, J., Frost, D., Luck, J. & Pryor, T. (1997) Annu. Rev. Phytopathol. 35, 271–291. 52. Hulbert, S. H. (1997) Annu. Rev. Phytopathol. 35, 293–310. 53. Michelmore, R. W. & Meyers, B. C. (1998) Genome Res. 8, 1113–1130. 54. Nei, M., Gu, X. & Sitnikova, T. (1997) Proc. Natl. Acad. Sci. USA 94, 7799–7806. 55. Hammond-Kosack, K. E., Jones, D. A. & Jones, J. D. G. (1994) Plant Cell 6, 361–374. 56. Jefferson, R. A., Kavanagh, T. A. & Bevan, M. W. (1987) EMBO J. 6, 3901–3907. 57. Oliver, R. P., Farman, M. L., Jones, J. D. G. & Hammond-Kosack, K. E. (1993) Mol. Plant-Microbe Interact. 6, 521–525. 58. Keogh, R. C., Deverall, B. J. & Mcleod, S. (1980) Trans. Brit. Mycol. Soc. 74, 329–333. 59. Thomas, C. M., Vos, P., Zabeau, M., Jones, D. A., Norcott, K. A., Chadwick, B. P. & Jones, J. D. G. (1995) Plant J. 8, 785–794. 60. Vos, P., Hogers, R., Bleeker, M., Reijans, M., Van de Lee, T., Hornes, M., Frijters, A., Pot, J., Peleman, J., Kuiper, M. & Zabeau, M. (1995) Nucleic Acids Res. 23, 4407–4414. 61. Eshed, Y. & Zamir, D. (1994) Euphytica 79, 175–179. 62. Thomas, C. M., Tang, S., Hammond-Kosack, K. E. & Jones, J. D. G. (2000) Mol. Plant-Microbe Interact. 13, 465–469. 63. Blatt, M. R., Grabov, A., Brearley, J., Hammond-Kosack, K. E. & Jones, J. D. G. (1999) Plant J. 19, 453–462. 64. Piedras, P., Hammond-Kosack, K. E., Harrison, K. & Jones, J. D. G. (1998) Mol. Plant-Microbe Interact. 11, 1155–1166. 65. Romeis, T., Piedras, P., Zhang, S. Q., Klessig, D. F., Hirt, H. & Jones, J. D. G. (1999) Plant Cell 11, 273–287. 66. Romeis, T., Piedras, P. & Jones, J. D. G. (2000) Plant Cell 12, 803–815. 67. Durrant, W., Rowland, O., Piedras, P. & Jones, J. D. G. (2000) Plant Cell 12, 963–977. 68. Dixon, M. S., Jones, D. A., Keddie, J. S., Thomas, C. M., Harrison, K. & Jones, J. D. G. (1996) Cell 84, 451–459. 69. Aarts, N., Metz, M., Holub, E., Staskawicz, B. J., Daniels, M. J. & Parker, J. E. (1998) Proc. Natl. Acad. Sci. USA 95, 10306–10311. 70. Warren, R. F., Merritt, P. M., Holub, E. & Innes, R. W. (1999) Genetics 152, 401–412. 71. Van den Ackerveken, G. F. J. M., Vossen, J. P. M. J. & De Wit, P. J. G. M. (1993) Plant Physiol. 103, 91–96. 72. Joosten, M. H. A. J., Vogelsang, R., Cozijnsen, T. J., Verberne, M. C. & De Wit, P. J. G. M. (1997) Plant Cell 9, 367–379.



8814

73. Van Eeden, F. & St Johnston, D. (1999) Curr. Opin. Genet. Dev. 9, 396–404. 74. Hammond-Kosack, K. E. & Jones, J. D. G. (1994) Mol. Plant-Microbe Interact. 7, 58–70. 75. Jorgensen, J. H. (1996) Genome 39, 492–498. 76. Clark, S. E., Williams, R. W. & Meyerowitz, E. M. (1997) Cell 89, 575–585. 77. Jeong, S., Trotochaud, A. E. & Clark, S. E. (1999) Plant Cell 11, 1925–1933. 78. Fletcher, L. C., Brand, U., Running, M. P., Simon, R. & Meyerowitz, E. M. (1999) Science 283, 1911–1914. 79. Trotochaud, A. E., Hao, T., Wu, G., Yang, Z. B. & Clark, S. E. (1999) Plant Cell 11, 393–405. 80. Wang, G. L., Ruan, D. L., Song, W. Y., Sideris, S., Chen, L. L., Pi, L. Y., Zhang, S. P., Zhang, Z., Fauquet, C, Gaut, B. S., et al. (1998) Plant Cell 10, 765– 779. 81. Hammond-Kosack, K. E., Tang, S. J., Harrison, K. & Jones, J. D. G. (1998) Plant Cell 10, 1251–1266. 82. Kooman-Gersmann, M., Honee, G., Bonnema, G. & De Wit, P. J. G. M. (1996) Plant Cell 8, 929–938. 83. Ji, C, Boyd, C., Slaymaker, D., Okinaka, Y., Takeuchi, Y., Midland, S. L., Sims, J. J., Herman, E. & Keen, N. (1998) Proc. Natl. Acad. Sci. USA 95, 3306– 3311. 84. Ulevitch, R. J. (1999) Nature (London) 401, 755–756. 85. Orth, K., Palmer, L. E., Bao, Z. Q., Stewart, S., Rudolph, A. E., Bliska, J. B. & Dixon, J. E. (1999) Science 285, 1920–1923. 86. Van der Biezen, E. A. & Jones, J. D. G. (1998) Trends Biochem. Sci. 23, 454–456. 87. Van der Biezen, E. A., Sun, J., Coleman, M. J., Bibb, M. J. & Jones, J. D. G. (2000) Proc. Natl. Acad. Sci. USA 97, 3747–3752. 88. Salmeron, J. M., Oldroyd, G. E. D., Rommens, C. M. T., Scofield, S. R., Kim, H.-S., Lavelle, D. T., Dahlbeck, D. & Staskawicz, B. J. (1996) Cell 86, 123– 133.


PLANTS AND ANIMALS SHARE FUNCTIONALLY COMMON BACTERIAL VIRULENCE FACTORS

8815

Colloquium Plants and animals share functionally common bacterial virulence factors Laurence G. Rahme*†‡, Frederick M. Ausubel§, Hui Cao†‡, Eliana Drenkard§, Boyan C. Goumnerov†‡, Gee W. Lau†‡, Shalina Mahajan-Miklos§¶, Julia Plotnikova§, Man-Wah Tan§ ||, John Tsongalis‡, Cynthia L Walendziewicz‡, and Ronald G. Tompkins†‡ † Department of Surgery, Harvard Medical School, ‡ Department of Surgery, Massachusetts General Hospital and Shriner's Burn Institute, Boston, MA 02114; § Department of Genetics, Harvard Medical School, and Department of Molecular Biology, Massachusetts General Hospital, Boston, MA 02114; and || Society of Fellows, Harvard University, 78 Mount Auburn Street, Cambridge, MA 02138 By exploiting the ability of Pseudomonas aeruginosa to infect a variety of vertebrate and nonvertebrate hosts, we have developed model systems that use plants and nematodes as adjuncts to mammalian models to help elucidate the molecular basis of P. aeruginosa pathogenesis. Our studies reveal a remarkable degree of conservation in the virulence mechanisms used by P. aeruginosa to infect hosts of divergent evolutionary origins. B acterial pathogens infect a wide variety of evolutionary distinct hosts, including both lower and higher eukaryotes. In all of these cases, the pathogen must have the ability to recognize, become associated with, exploit the nutrient reserves of, and combat the defense responses of its specific host. To accomplish these tasks, pathogens use an extensive arsenal of virulence-related factors. Many pathogens cause disease in a single or limited number of host species as a consequence of a long coevolutionary history. However, the interactions between host and pathogen that limit host range and determine host resistance or susceptibility are poorly understood. Although many bacterial virulence components are thought to be host-specific, numerous studies have demonstrated the existence of what appear to be universal virulence mechanisms used by diverse bacterial species (1). Similarly, recent work has revealed common features underlying host defense responses against pathogens in plants, insects, and mammals (2). Thus, some of the underlying virulence mechanisms of pathogens, as well as the host defenses against them, are likely to have ancient evolutionary origins preserved across phylogeny. There remains a great deal to learn about the molecular nature of antagonistic encounters between pathogenic bacteria and their hosts. Despite our limited knowledge, significant advances have been made in developing techniques that facilitate our understanding of virulence mechanisms and the critical role of the host during pathogenesis (3). Our laboratories have developed a technique, which we refer to as multihost pathogenesis, to study pathogens that cause disease in both vertebrate and nonvertebrate hosts (4, 5, 6, 7 and 8). Epidemiological studies carried out in the 1970s suggested that clinical isolates of Pseudomonas aeruginosa might be capable of causing disease in plants (9, 10). Based on this premise, we developed a model pathogenesis system by using a human clinical isolate of P. aeruginosa, strain UCBPP-PA14 (PA14) that elicits disease in plants, nematodes, insects, and mice (4, 5, 6 and 7, 11). P. aeruginosa is the most common causative organism of sepsis in burned patients and the leading cause of pulmonary infections and mortality in cystic fibrosis patients (12). In addition, this important human opportunistic pathogen infects injured, immunodeficient, or otherwise compromised individuals (13). The pathophysiology of infections caused by P. aeruginosa is complex as shown by the clinical diversity of diseases associated with this organism and the multiplicity of virulence factors it produces. Although a natural soil inhabitant, P. aeruginosa is versatile in its metabolic potential, which allows it to survive in a variety of natural and hospital environments. It appears that the combination of environmental persistence, versatility in virulence mechanisms, and multiple virulence factors allows P. aeruginosa to be effective both as an opportunistic human pathogen and as a plant pathogen. None of the currently available animal models fully mimics all aspects of any single human disease caused by P. aeruginosa. Moreover, it is not practical to use any current vertebrate animal model to systematically screen for pathogenicity-related P. aeruginosa mutants. This latter problem has become acutely apparent as the Pseudomonas Genome Project nears completion, and the relevance of thousands of P. aeruginosa genes to human pathogenesis needs to be determined. In contrast to vertebrate animal models, host-pathogen interaction models that use nonvertebrate hosts permit genomewide screening for P. aeruginosa virulence factors required for pathogenesis. If there is substantial overlap in the virulence factors required for pathogenesis in invertebrate and mammalian hosts, large-scale screening in invertebrate hosts can become an efficient method for identifying previously unknown virulence-related genes relevant to human pathogenesis. Using P. aeruginosa strain PA14, we provided direct evidence that P. aeruginosa uses a shared subset of virulence factors to elicit disease in both plants (Arabidopsis thaliana) and animals (mice) (4, 11). More recently, we extended the Arabidopsis-P. aeruginosa model to additional nonvertebrate hosts, including the nematode Caenorhabditis elegans (7, 14, 15) and the insects Galleria melonella (8) and Drosophila melanogaster (S.M.-M., G.W.L., L. Perkins, F.M.A., and L.G.R., unpublished data). This paper summarizes the use of a plant pathogenesis model to identify previously unknown virulence factors and highlights the remarkable conservation in the virulence mechanisms used by P. aeruginosa to infect evolutionary divergent hosts.

P. AERUGINOSA PATHOGENESIS Despite quite detailed knowledge about some of the extracellular proteins and several surface-associated components synthesized by P. aeruginosa, our understanding of P. aeruginosa

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host-Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. * To whom reprint requests should be addressed. E-mail: [email protected]. ¶ Present address: Micrbobia, Inc., One Kendall Square, Building 1400W, Suite 1418, Cambridge, MA 02139.



8816

infections is only rudimentary. Much remains to be learned about individual virulence factors essential for pathogenicity and the mechanism by which these factors work together during pathogenesis. P. aeruginosa virulence factors facilitate tissue invasion and systemic spread and include pili and flagella, endotoxins, exotoxins, vascular permeability factors, and a variety of excreted enzymes (16). P. aeruginosa proteases degrade a variety of host proteins and have a direct destructive effect on skin tissue (17, 18). Elastase (LasB), a potent metalloprotease with broad substrate specificity, degrades host proteins, such as elastin, collagen, transferrin, Ig, and some of the components of complement (19). The LasB elastase acts in concert with both LasA protease and alkaline protease to cause efficient elastolysis, which is required for the tissue damage associated with P. aeruginosa pathogenesis (20). The P. aeruginosa alginate capsule plays an essential role in chronic pulmonary infection in cystic fibrosis patients (21, 22). Finally, the P. aeruginosa lipopolysaccharide also has been shown to be an important virulence factor (23). The synthesis of several P. aeruginosa-secreted virulence factors is transcriptionally regulated by environmental stimuli (24, 25 and 26). For example, phospholipase C is regulated by the availability of phosphate (27, 28), whereas exotoxin A and elastase are regulated by the concentration of iron in the growth medium (29, 30, 31 and 32). Moreover, the production of a large number of exoproducts, including elastase, alkaline protease, the LasA protease, hemolysin, pyocyanin, and rhamnolipids is cell densitydependent and is regulated by the so-called quorumsensing cascade (33). The P. aeruginosa quorum-sensing cascade is mediated by low molecular mass homoserine lactones that are synthesized by the products of the lasI and rhlI genes. The concentration of these homoserine lactones is monitored by the products of the lasR and rhlR genes, which serve as global transcriptional activators of a variety of exoproducts relevant for P. aeruginosa pathogenesis. In addition to homoserine lactones, it recently has been shown that 2-heptyl-3-hydroxy-4-quinolone also serves as an intracellular signal molecule involved in the activation of pathogeneicity-related factors (34). This molecule belongs to the 4-quinolone chemical family, best known for the antibiotic activity of many of its members.

Fig. 1. Macroscopic and microscopic symptoms elicited by P. aeruginosa strain UCBPP-PA14 infiltrated into Arabidopsis leaves. (A) The upper part of an Arabidopsis leaf (ecotype LL-O) was infiltrated with bacterial suspensions at a titer of 103 cfu/ cm2 and photographed 2 days postinfection. A chlorotic zone (white arrow) surrounds the soft-rot symptoms. (B and C) Arabidopsis leaves infiltrated with bacterial suspension and stained with trypan blue 2 days postinfection. Whole leaves were examined by using a Zeiss Axioskop and photographed. Bacterial movement was observed along leaf veins (arrows indicate vessel parenchyma filled with bacteria). Bacteria are absent in xylem vessels (arrowheads). In addition to regulating the production of a variety of specific virulence-related factors, the quorum-sensing cascade as well as other cell-tocell signaling mechanisms are involved in the differentiation of P. aeruginosa biofilms (35, 36), matrix-enclosed bacterial populations that attach to each other and to biotic, or abiotic surfaces (37). P. aeruginosa is protected from adverse environmental conditions and from antibacterial agents while growing in this matrix-enclosed structure that has been implicated as a contributing factor in the persistence and severity of P. aeruginosa infections. It is thought that when nutrient conditions become limiting, bacteria are shed from the biofilm and enter the planktonic (free-living) phase (37). In this manner, cells are able to colonize new habitats and form new biofilms. Another quorum-sensing regulated virulence factor is pyocyanin, a blue-green-colored phenazine (33, 38). This secondary metabolite has antimicrobial activity against several species of bacteria, fungi, and protozoa, a quality attributed to its redoxactive potential (39). Although little is known about the nature of the enzymes that catalyze the formation of pyocyanin in P. aeruginosa, the conversion of chorismate to anthranilate is thought to be a key step in the pathway that is most likely catalyzed by the anthranilate synthetase encoded by the phnA and phnB genes (40). Even though pyocyanin-induced free radical production appears to be responsible for much of its antimicrobial activity (41), the role of pyocyanin in P. aeruginosa-associated tissue injury is less clear.

DEVELOPMENT OF THE ARABIDOPSIS-P. AERUGINOSA MODEL As described briefly above, we developed the Arabidopsis-P. aeruginosa pathogenesis model based on the P. aeruginosa strain PA14. This was accomplished as follows. A collection of 75 P. aeruginosa strains, of which 30 were human isolates, was screened for its ability to elicit disease on leaves of at least four different Arabidopsis ecotypes (4). Several strains elicited disease characterized by varying degrees of soft-rot symptoms in all four ecotypes (wild varieties) tested. Importantly, two strains, PA14 (a human isolate) and UCBPP-PA29 (a plant isolate), caused severe soft-rot symptom in some, but not all of the ecotypes tested. The finding that these two P. aeruginosa strains exhibit ecotype specificity is typical of plant pathogen interac



8817

tions and reflects the selection of plant accessions (ecotypes) that are resistant to particular strains (races) of a pathogen. These observations support the notion that P. aeruginosa is a natural plant pathogen in the wild.

Fig. 2. P. aeruginosa entering the mesophylic compartment of an Arabidopsis leaf through a stomatal opening. Detached Arabidopsis leaves were soaked in a bacterial suspension at a titer of 103 cfu/ml. At 24 h postinfection, the leaves were fixed in osmium tetroxide, followed by dehydration in ethanol. The leaves then were coated in Technics Sputter Coater Hummer II and subsequently examined and photographed by using an Amray (Bedford, MA) 1000 scanning electron microscope. Bacteria are shown to be primarily concentrated on the surface of guard cells, above the stoma. Fig. 1A illustrates the symptomatology observed on a susceptible 6-week-old Arabidopsis plant when the upper part of a leaf is inoculated with P. aeruginosa strain PA14. The severe symptoms elicited by PA14 include a water-soaked reaction zone and chlorosis (yellowing). Soft-rot symptoms begin as small water-soaked lesions, which enlarge rapidly in diameter. Complete maceration and collapse of the infiltrated leaf occurs 4–5 days after infection. As shown in Fig. 1B, trypan blue staining of infected leaves reveals the presence of P. aeruginosa in the intercellular spaces of the leaf as well as bacterial movement along the veins and bacterial proliferation in vessel parenchyma and companion cells. No bacteria are detected in xylem. Fast propagation of P. aeruginosa in the vessel parenchyma is the prelude to systemic infection and maceration of the entire plant. As illustrated in Fig. 2, detached leaves of Arabidopsis exposed to PA14 culture suspension show that P. aeruginosa cells attach to the leaf epidermis and congregate at the stomal openings where they gain entry to the leaf interior. Similar to other phytopathogenic bacteria, P. aeruginosa strains that elicit disease symptoms are also able to proliferate in Arabidopsis leaves. Table 1 lists the maximal levels of growth reached by several P. aeruginosa strains and PA14 mutants at the fourth day postinfection. The degree of proliferation was correlated with the severity of disease symptoms. In each case, reduced symptoms were associated with reduced bacterial counts in leaves (Table 1; ref.4). Table 1. Animal mortality and proliferation studies of P. aeruginosa strains and isogenic UCBPP-PA14 mutants in Arabidopsis and in a mouse burned model P. aeruginosa strain cfu/cm2 leaf area* Mean titer ± SD in biopsies adjacent to burn † % Mouse mortality‡ 7 7 7 7 UCBPP-PA14 2.6 × 10 ± 2.0 × 10 6.0 × 10 ± 2.1 × 10 77 UCBPP-PA29 2.7 × 107 ± 1.3 × 107 8.2 × 107 ± 2.0 × 107 5 6.0 × 107 ± 1.2 × 107 20 PAK 6.0 × 104 ± 3.0 × 104 4.0 × 107 ± 1.8 × 107 25 PAO1 8.0 × 105 ± 2.7 × 105 UCBPP-PA14 plcS 1.1 × 105 ± 0.5 × 105 6.7 × 104 ± 7.5 × 104 40 1.4 × 108 ± 1.7 × 108 40 UCBPP-PA14 toxA 1.5 × 106 ± 1.0 × 106 UCBPP-PA14 gacA 6.0 × 103 ± 2.1 × 103 NT 0 For both quantitation and mortality studies, mice were injected with 5 × 104 cfu/ml. NT, not determined. * Means of four samples ± SD of maximum bacterial counts obtained at 5 days postinfection with 103 cells. No viable bacterial cells were retrieved from the underlying rectus abdominus muscle immediately after bacterial injection or in animals that received a sham injury in other studies. ‡ The number of animals that died of sepsis was monitored each day for 10 days. †

In our efforts to identify additional plant species susceptible to P. aeruginosa infection, we found that soft-rot symptoms also were observed when PA14 was inoculated into the midrib of lettuce leaf stem (Fig. 3B). In addition to PA14, lettuce is susceptible to the well-characterized P. aeruginosa strains PAK and PA01. In later stages of infection, all three P. aeruginosa strains invade the entire midrib of lettuce leaves, resulting in complete maceration and collapse of the tissue (Fig. 3B).

PATHOGENICITY OF PA14 IN A MOUSE BURN MODEL We primarily have used a mouse full-thickness skin burn model to assess PA14 pathogenicity in a mammalian host (42). Several factors facilitate the colonization and establishment of the bacteria in this model. First, the inoculum is delivered in a scald eschar induced on the abdominal surface of an anesthetized mouse. This method provides the bacteria with necrotic tissue as nutrients and with protection from the active components of the immune system, because the circulation in this area is impeded. Second, it is well known that burn trauma itself induces a generalized decrease in the efficiency of the immune system, which aids in the development of systemic infections and sepsis. The first signs of developing infection and sepsis can be observed clinically and histopathologically 12–24 h postinoculation. Mortality studies were conducted to assess the presence of systemic infection (Table 1). Tissue biopsies from the site of inoculation were obtained and compared to assess local invasiveness. As presented in Table 1, strains PA14 and PA29 as well the P. aeruginosa human isolates PA01 and PAK all proliferated and invaded the muscle underlying the eschar, but strain PA14 resulted in a higher mortality than the other strains. Animals usually succumbed to the infection 36–48 h postinoculation. We conducted a series of morphological studies to further investigate the dynamics of PA14 infections in the burned mouse model. Specimens from control animals and from burned and inoculated animals were processed by standard histologic techniques. Inoculated mice were killed at varying time intervals and histopathologic examination of parenchymal organs was undertaken. The findings revealed that a local inflammatory response developed during the first 24 h at the site of inoculation and that systemic infection, including multiple organ involvement, occurred between 24 and 36 h. Histologic sections removed from livers and kidneys 36 h postinoculation revealed massive perivascular bacterial invasion of the organs and associated tissue



8818

necrosis (Fig. 4). These studies demonstrate the ability of PA14 to overcome the immune defense mechanisms in the host and establish a generalized infection.

Fig. 3. (A) Strategy used to screen for P. aeruginosa virulence-associated genes. (B) Midribs of lettuce leaves were inoculated with 10 µl of a bacterial suspension at a titer of 103 cfu/ml and photographed 4 days postinfection. Bottom leaf was inoculated with the wild-type strain PA14 (black arrowhead) and top leaf was inoculated with the PA14-toxA mutant. Brown spots (white arrowheads) on the midrib of the top leaf reveal inoculation sites. SHARED VIRULENCE FACTORS IN PLANT AND ANIMAL PATHOGENESIS The relevance of the Arabidopsis-P. aeruginosa model to mammalian pathogenesis initially was demonstrated by testing whether two P. aeruginosa virulence factors, exotoxin A and phosopholipase C, thought to influence tissue invasion and systemic spread in mammals (4), also play a role in Arabidopsis pathogenesis. Exotoxin A (toxA) is a potent mammalian protein synthesis inhibitor (43). Phospholipase C (plcS) preferentially degrades phospholipids found in eukaryotic cell membranes (44), which leads to a systemic inflammatory response (45). Conversely, a known virulence determinant important for plant pathogenesis in Pseudomonas syringae, gacA, was tested for its role in mouse pathogenesis. GacA encodes a transcriptional activator of genes encoding extracellular products involved in pathogenicity (46). Isogenic PA14 plcS, toxA, and gacA mutants were constructed by marker exchange, and the resulting mutants were tested for pathogenicity in the Arabidopsis leaf infiltration assay and in the burned mouse model. Decreased pathogenicity was observed in both hosts with all three of the isogenic PA14 mutants. Unlike the wild-type PA14, none of the mutants caused maceration and collapse of Arabidopsis leaves, and the growth of the mutants was significantly reduced (4). Similarly, all three mutants exhibited significantly lower mortality than the wildtype strain in the mouse model. Consistent with its known role in pathogenesis, the plcS mutant exhibited reduced ability to proliferate and invade the tissue adjacent to the burn area (Table 1). Subsequent to these experiments, it was shown that GacA functions upstream of lasI-lasR and rhlI-rhlR quorumsensing regulatory systems (47) to regulate virulence gene expression (33) and type 4 pilus-mediated twitching motility in P. aeruginosa (48).

Fig. 4. Histologic section of mouse liver and kidney taken 36 h postinoculation with P. aeruginosa strain PA14. (A) Massive perivascular bacterial infiltration and associated necrosis of the surrounding hepatocytes (arrow) are visible in the liver section. Brown and Hopps (B&H) Gram stain at ×40 magnification, (B) Section of kidney from the same animal showing large bacterial infiltrates within the tubules of the renal medulla (arrow). B&H Gram stain at ×40 magnification. USE OF NONVERTEBRATE HOSTS TO IDENTIFY VIRULENCE-ASSOCIATED GENES INVOLVED IN MAMMALIAN PATHOGENESIS The observation that P. aeruginosa uses common virulence genes to infect both animals and plants led us to theorize that previously unknown virulence determinants required for P. aeruginosa pathogenesis in animals could be identified by screening randomly mutagenized PA14 clones for ones that exhibited



8819

decreased virulence in plants. This experimental approach overcomes a major limitation imposed by an animal model, that is, the large number of animals required to undertake a systematic genetic screen for the identification of virulence-related genes. Table 2. P. aeruginosa mutants causing decreased virulence in plants Symptoms elicited in Arabidopsis * % Mouse mortality † Strain PA14 Severe 100 Group I mutants isolated from plant screens 33C7 None 0 1D7 Weak 50 25A12 plcS toxA 33A9

Weak Moderate Moderate Moderate

87 40 40 0

25F1 Moderate 34H4 Moderate pho34B12 Moderate pho15 Moderate 16G12 Moderate Group II mutants isolated from C. elegans screens 36A4 None 50E12 Weak

20 33 56 62 100

48D9

Weak

50

41C1 23A2 41A5 6A6, 3E8 12A1 8C12

Weak Weak Weak Moderate Moderate Moderate

81 85 100 18 50 63

0 0

Gene identity and comments Wild type Unknown‡ gacA, quorum-sensing regulator, two-component system lemAgacA Unknown plcS, degrades phospholipids of eukaryotic membranes toxA, inhibits eukaryotic protein synthesis Unknown, absent in PA01, reduced motility, increased surface attachment Homologue of Chlorobium tepidum orfT Unknown, contains a bipartite nuclear localization signal Unknown, phenazine regulator, contains a HTH motif dsbA, periplasmic S-S forming enzyme Unknown Homologue of P. syringae hrpM Homologue of A vinelandii PstP protein, required for accumulation of poly-β-hydroxybutyrate Homologue of lemA, sensor of the two-component system lemA-gacA Homologue of putative E. coli integral membrane protein AefA mexA, non-ATPase membrane efflux pump Unknown phzB, phenazine biosynthesis lasR, quorum-sensing regulator Unknown

* Symptoms observed 4–5 days postinfection. None, no symptoms; weak, localized water soaking and chlorosis of tissue circumscribing the inoculation site; moderate, moderate water-soaking and chlorosis with most of the tissue softened around the inoculation site at 2–3 days postinfection. † Mice were injected with 5 × 105 cells. ‡ Unknown, BLASTX analysis yielded no encoded proteins with significant homology. P. aeruginosa strain PA14 was mutagenized with transposon TnphoA, and 2,500 prototrophic mutants were screened for impaired virulence in a lettuce stem assay. In these highthroughput screens, we substituted Arabidopsis with lettuce because several mutants could be tested on one lettuce midrib by “tooth-picking” single colonies directly from plates that contain the mutant clones (Fig. 3; ref.5). Following the methodology depicted in Fig. 3, the virulence-attenuated mutants identified subsequently were tested for their ability to elicit disease symptoms and proliferate in Arabidopsis leaves. As summarized in Table 2, in addition to toxA and plcS a total of 11 virulencerelated genes were identified among the 2,500 prototrophs screened that elicited null, weak, or moderate rotting symptoms on lettuce and Arabidopsis plants (Table 2, group I). Importantly, the reduced ability of the mutants to cause disease symptoms correlated with reduced bacterial counts in leaves (5). In addition to demonstrating its ability to cause disease in plants and mice, we have shown that PA14 kills C. elegans when it is presented to the nematodes as a food source (15). In low-salt medium, PA14 kills worms over a period of 2–3 days (slow killing) by an infection-like process that correlates with the accumulation of bacteria in the worm intestine (14). Alternatively, PA14 grown in a high-salt, rich medium kills worms within 4–24 h (fast killing) by a toxin-mediated process (7). Fast and slow killing appear to be mechanistically distinct because most P. aeruginosa mutants that are attenuated in slow killing are not attenuated in fast killing and vice versa. Altogether, 13 P. aeruginosa genes have been identified that when mutated exhibit reduced virulence in the C. elegans slow or fast killing assays. Of these 13 genes, nine also are required for maximum pathogenicity in the Arabidopsis leaf infiltration assay and are listed inTable 2 (group II genes; gacA was identified in both the plant and nematode screens). Eight of the 20 P. aeruginosa genes listed in Table 2 that were identified by screening in plants and nematodes do not correspond to previously identified proteins from other species (5, 6 and 7). Remarkably, at least 15 of these 20 genes are required for pathogenesis in the burned mouse model.

WHAT TYPES OF VIRULENCE MECHANISMS ARE CONSERVED THROUGH EVOLUTION? The “multihost” pathogenesis screens performed in our laboratories identified a variety of virulence-associated genes that encode proteins involved in transcriptional control, posttranscriptional control, efflux systems, biosynthetic enzymes involved in phenazine production, toxins, and proteins of unknown function. Table 2 highlights the striking conservation in the virulence mechanisms used by P. aeruginosa to infect plants, nematodes, and mammals. It is remarkable that at least 15 of the P. aeruginosa mutants isolated in either a plant or a nematode screen were found to be required for full virulence in the burned mouse model. One virulence-related factor found to play a critical role in pathogenesis in plants, nematodes, and mice is the periplasmic disulfide bondforming enzyme encoded by the dsbA gene (49), whose function is likely to affect several periplasmic virulencerelated proteins. An important role for dsbA in pathogenesis has been described in several human pathogens, including Shigella flexneri (50) and Vibrio cholera (51), and in the bacterial phytopathogen Erwinia chrysanthemi (52). Three virulence-related genes identified in our screen, hrpM, gacS, and gacA, previously had been shown by other researchers to be important in phytopathogenesis. However, our studies found that these genes also play an important role in mammalian



8820

pathogenesis. The three genes also are required for effective killing of C. elegans (6, 7). The hrpM gene of P. syringae pv. syringae was identified as a homologue of the Escherichia coli mdoH gene, which is part of an operon involved in the synthesis of membrane-derived oligosaccharides (MDO) (53). Although MDO have been found to have different functions in a variety of Gram-negative bacteria, the role of MDO in bacterial pathogenesis is not well understood. Mutation in the hrpM locus of the plant pathogen P. siryngae pv. syringae abolishes both the development of disease symptoms on host plants as well as the hypersensitivity response in nonhost plants (54). The GacS protein is a sensor kinase of a two-component bacterial family of regulators (55). GacA, the cognate response regulator of GacS, initially was identified as a global regulator of secondary metabolites in P. fluorescens (56, 57). It has been shown that mutations in both the gacA and gacS genes in P. siryngae lead to decreased lesion formation in beans and no production of the toxin syringomycin (46, 55, 58). The requirement of the lasR, gacS, and gacA gene products for pathogenicity in plants, nematodes, and mice provides evidence that quorum sensing and regulated export of proteins are general features of pathogenesis for all three hosts. The lasR, gacS, and gacA genes are present in numerous plant and animal bacterial pathogens as well as in saprophytes. It is likely that LasR, GacS, and GacA initially served as master regulators, enabling ancestral Gram-negative organisms to adapt to their environment. Subsequently, these proteins evolved to regulate a variety of genes that allowed prokaryotes to invade and establish their presence in eukaryotic hosts. Another class of conserved virulence factors important in plant and nematode pathogenesis is the genes involved multidrug efflux pumps. The mexA gene corresponding to mutant 23A2 encodes a component of a multidrug efflux pump; it recently has been shown to be involved in active efflux of P. aeruginosa autoinducers that are not freely diffusable (59). However, mutant 23A2 was only marginally compromised in the mouse burn model. Further studies using a lower inoculum need to be conducted to ascertain its role as a virulence factor in mammalian pathogenesis. An additional protein relevant to plant pathogenesis, but not to mammalian pathogenesis, is a homologue of the putative E. coli integral membrane protein AefA (Proposite: PS01246). The function of this protein is yet unknown. Other known proteins identified in our screens, but not previously shown to be involved in pathogenesis, include a PtsP homologue of A. vinelandii required for poly-β-hydroxybutyrate accumulation. The ptsP gene is predicted to encode enzyme INtr, a presumptive transcriptional regulator of RpoN-dependent operons (60). One of the novel virulence factors identified in the plant screen (Pho34B12) affects hemolytic and elastolytic activity as well as pyocyanin production (refs.5 and 7; H.C. and L.G.R., unpublished work), all of which are under quorum-sensing regulation. The protein encoded by pho34B12 contains a helixturn–helix DNA binding motif similar to that found in the LysR family of transcriptional regulators (ref.5; H.C. and L.G.R., unpublished work). This class of proteins includes regulators involved in both mammalian and plant pathogenesis (1). Our multihost pathogenesis study results indicate that phenazines are an important class of virulence-related effector molecules. Despite intensive in vitro analyses of phenazines, the physiological significance of their role in P. aeruginosa pathogenesis in mammals has been controversial (61). Before our studies, there had been no demonstration of their role in vivo. Mutants 3E8 and 6A6 correspond to the previously identified phzB gene in P. fluorescens and phzY gene in P. aureofaciens. Both phzB and phzY are present in operons known to regulate production of phenazine-1-carboxylate (62). Additional effector molecules that play a role in multihost pathogenesis include the ToxA protein and the previously unidentified virulence gene encoding the 34H4 protein. Current work indicates that 34H4 contains a bi-partite nuclear localization signal required for translocation of the protein into the nucleus of mammalian cells (G.W.L. and L.G.R., unpublished work). A common phenotype of at least two of the mutants, 33A9 and 3E8, is reduced motility and altered surface attachment ability (Table 2). Mutant 3E8 exhibits reduced attachment to abiotic surfaces, such as polyvinilchloride plastic surfaces (ref.63; S.M.-M. and F.M.A., unpublished work) whereas 33A9 exhibited increased surface attachment ability (E.D., G. O'Toole, F.M.A., and L.G.R., unpublished work). Bacterial adhesion is an essential step in biofilm formation and consequently, in bacterial virulence. No significant similarity to any known genes was found in the gene corresponding to mutant 33A9. Interestingly, DNA sequence analysis of the region containing the 33A9 gene revealed that this gene is not present in the recently sequenced genome of the P. aeruginosa strain PAO1.

CONCLUSIONS In summary, we can draw the following conclusions from our studies of P. aeruginosa multihost pathogenesis. First, the variety of virulenceassociated genes described in our experiments indicates that the multihost strategy has few limitations with regard to the categories of virulencerelated functions that can be identified. Second, the fact that most of the genes identified were not previously known to be involved in pathogenesisrelated functions demonstrates that the multihost strategy is particularly efficient in identifying novel P. aeruginosa virulence factors. Third, and perhaps most importantly, even though many pathogens cause disease in a single or limited number of host species these studies provide strong evidence that there exists several universal bacterial virulence mechanisms highly conserved across phylogeny. Although some of the virulence-associated genes identified have homologues in other pathogenic bacteria, the exact role of these genes in pathogenesis remains unclear in most cases. The use of genetically tractable host systems, such as plants, nematodes, and insects, will generate important information about host responses and lead to a better understanding of the fundamental molecular mechanisms that underlie bacterial pathogenesis. 1. Finlay, B. B. & Falkow, S. (1997) Microbiol. Mol. Biol. Rev. 61, 136–169. 2. Kopp, E. B. & Medzhitov, R. (1999) Curr. Opin. Immunol. 11, 13–18. 3. Strauss, E. J. & Falkow, S. (1997) Science 276, 707–711. 4. Rahme, L. G., Stevens, E. J., Wolfort, S. F., Shao, J., Tompkins, R. G. & Ausubel, F. M. (1995) Science 268, 1899–1902. 5. Rahme, L. G., Tan, M.-W., Le, L., Wong, S. M., Tompkins, R. G., Calderwood, S. B. & Ausubel, F. M. (1997) Proc. Natl. Acad. Sci. USA 94, 13245–13250. 6. Tan, M.-W., Rahme, L. G., Sternberg, J. A., Tompkins, R. G. & Ausubel, F. M. (1999) Proc. Natl. Acad. Sci. USA 96, 2408–2413. 7. Mahajan-Miklos, S., Tan, M.-W., Rahme, L. G. & Ausubel, F. M. (1999) Cell 96, 47–56. 8. Jander, G., Rahme, L. G. & Ausubel, F. M. (2000) J. Bacteriol. 182, 3843–3845. 9. Kominos, S. D., Copeland, C. E., Grosiak, B. & Postic, B. (1972) Appl. Microbiol. 24, 567–570. 10. Cho, J. J., Schroth, M. N., Kominos, S. D. & Green, S. K. (1975) Phytopathology 65, 425–431. 11. Finlay, B. B. (1999) Cell 96, 315–318. 12. Doring, D. (1993) in Pseudomonas aeruginosa as an Opportunistic Pathogen, ed. Campa, M. (Plenum, New York), pp. 245–273. 13. Wood, R. E. (1976) Hosp. Prac. 11, 91–100. 14. Tan, M.-W., Mahajan-Miklos, S. & Ausubel, F. M. (1999) Proc. Natl. Acad. Sci. USA 96, 715–720. 15. Tan, M.-W. & Ausubel, F. M. (2000) Curr. Opin. Microbiol. 3, 29–34. 16. Fink, R. B., Jr., ed. (1993) in Pseudomonas aeruginosa the Opportunist: Pathogenesis and Disease (CRC, Boca Raton, FL), pp. 1–5.



8821

17. Kawaharajo, K., Homma, J. Y., Aoyama, Y., Okada, K. & Morihara, K. (1975) Jpn. J. Exp. Med. 45, 79–88. 18. Holder, I. A. & Neely, A. N. (1991) Antib. Chemother. 44, 99–105. 19. Galloway, D. R. (1991) Mol. Microbiol. 5, 2315–2321. 20. Peters, J. E. & Galloway, D. R. (1990) J. Bacteriol. 172, 2236–2240. 21. Ohman, D. E., Cryz, S. J. & Iglewski, B. H. (1980) J. Bacteriol. 142, 836–842. 22. Zielinski, N. A., DeVault, J. D., Roychoudhury, S., May, T. B., Kimbara, K., Kato, J., Shinabarger, D., Kitano, K., Berry, A., Misra, T. & Chakrabarty, A. M. (1990) in Pseudomonas: Biotransformations, Pathogenesis, and Evolving Biotechnology, eds. Silver, S., Chakrabarty, A. M., Iglewski, B. & Kaplan, S. (Am. Soc. Microbiol., Washington, DC), pp. 15–27. 23. Pitt, T. L. (1989) Antibiot. Chemother. 42, 1–7. 24. Nicas, T. I. & Iglewski, B. H. (1986) J. Clin. Microbiol. 23, 967–969. 25. Jones, S., Yu, B., Bainton, N. J., Birdsall, M., Bycroft, B. W., Chhabra, S. R., Cox, A. J. R., Golby, P., Reeves, P. J., Stephens, S., et al. (1993) EMBO J. 12, 2477–2482. 26. Passador, L., Cook, J. M., Gambello, M. J., Rust, L. & Iglewski, B. H. (1993) Science 260, 1127–1130. 27. Pritchard, A. E. & Vasil, M. L. (1986) J. Bacteriol. 176, 291–298. 28. Shortridge, V. D, Lazdunski, A. & Vasil, M. L. (1992) Mol. Microbiol. 6, 863–871. 29. Grant, C. C. R. & Vasil, M. L. (1986) J. Bacteriol. 168, 1112–1119. 30. Lory, S. (1986) J. Bacteriol. 168, 1451–1456. 31. Frank, D. W., West, S. E. H. & Iglewski, B. H. (1990) in Molecular Basis of Bacterial Pathogenesis, eds. Iglewski, B. H. & Clark, V. L. (Academic, San Diego), Vol. 11, pp. 427–455. 32. Iglewski, B. H., Rust, L. & Bever, R. A. (1990) in Pseudomonas: Biotransformations, Pathogenesis, and Evolving Biotechnology, eds. Silver, S., Chakrabarty, A. M., Iglewski, B. & Kaplan, S. (Am. Soc. Microbiol., Washington, DC), pp. 36–43. 33. Pesci, E. C. & Iglewski, B. H. (1997) Trends Microbiol. 4, 132–134. 34. Pesci, E. C., Milbank, J. B. J., Pearson, J. P., McKnight, S., Kende, A. S., Greenberg, E. P. & Iglewski, B. H. (1999) Proc. Natl. Acad. Sci. USA 96, 11229– 11234. 35. McLean, R. J. C., Whiteley, M., Stickler, D. J. & Fuqua, W. C. (1997) FEMS Microbiol. Lett. 154, 259–263. 36. Davies, D. G., Parsek, M. R., Pearson, J. P., Iglewski, B. H., Costerton, J. W. & Greenberg, E. P. (1998) Science 280, 295–298. 37. Costerton, J. W., Lewandowski, Z., Caldwell, D. E., Korber, D. R. & LappinScott, H. M. (1995) Annu. Rev. Microbiol. 49, 711–745. 38. Fuqua, W. C., Winans, S. C. & Greenberg, E. P. (1994) J. Bacteriol. 176, 269–275. 39. Sorensen, R. U., Klinger, J. D., Cash, H. A., Chase, P. A. & Dearborn, D. G. (1983) Infect. Immun. 41, 321–330. 40. Essar, D. W, Eberly, L., Hadero, A. & Crawford, I. P. (1990) J. Bacteriol. 172, 884–900. 41. Hassan, H. M. & Fridovich, I. (1980) J. Bacteriol. 141, 1556–1163. 42. Stevens, E. J., Ryan, C. M., Friedberg, J. S., Barnhill, R. L., Yarmush, M. L. & Tompkins, R. G. (1994) J. Burn Care Rehabil. 15, 232–235. 43. Iglewski, B. H. & Kabat, D. (1975) Proc. Natl. Acad. Sci. USA 72, 2284–2288. 44. Berka, R. M. & Vasil, M. L. (1982) J. Bacteriol. 152, 239–245. 45. Besterman, J. M., Duronio, V. & Cuatrecasas, P. (1986) Proc. Natl. Acad. Sci. USA 83, 6785–6789. 46. Rich, J. J., Kinscherf, T. G., Kitten, T. & Willis, D. K. (1994) J. Bacteriol. 176, 7468–7475. 47. Reimmann, C., Beyeler, M., Latifi, A., Winteler, H., Foglino, M., Lazdunski, A. & Haas, D. (1997) Mol. Microbiol. 24, 309–319. 48. Glessner, A., Smith, R. S., Iglewski, B. H. & Robinson, J. B. (1999) J. Bacteriol. 181, 1623–1629. 49. Bardwell, J. C. A., McGovern, K. & Beckwith, J. (1991) Cell 67, 581–589. 50. Watarai, M., Tobe, T., Yoshikawa, M. & Sasakawa, C. (1995) Proc. Natl. Acad. Sci. USA 92, 4927–4931. 51. Peek, J. A. & Taylor, R. K. (1992) Proc. Natl. Acad. Sci. USA 89, 6210–6214. 52. Shevchik, V. E., Bortoli-German, I., Robert-Baudouy, J., Robinet, S., Barras, F. & Condemine, G. (1995) Mol. Microbiol. 16, 745–753. 53. Loubens, I., Debarbieux, L., Bohin, A., Lacroix, J.-M. & Bohin, J.-P. (1993) Mol. Microbiol. 10, 329–340. 54. Anderson, D. M. & Mills, D. (1985) Phytopathology 75, 104–108. 55. Hrabak, E. M. & Willis, D. K. (1992) J. Bacteriol. 174, 3011–3020. 56. Laville, J., Voisard, C., Keel, C., Maurhofer, M., Défago, G. & Haas, D. (1992) Proc. Natl. Acad. Sci. USA 89, 1562–1566. 57. Gaffney, T. D., Lam, S. T., Ligon, J., Gates, K., Frazelle, A., DiMaio, J., Hill, S., Goodwin, S., Torkewitz, N., Allshouse, A. M., et al. (1994) Mol. PlantMicrobe Interact. 7, 455–463. 58. Hrabak, E. M. & Willis, D. K. (1993) Mol. Plant-Microbe Interact. 6, 368–375. 59. Pearson, J. P., Van Delden, C. & Iglewski, B. H. (1999) J. Bacteriol 181, 1203–1210. 60. Reizer, J., Reizer, A., Merrick, M. J., Plunkett, G., Rose, D. J. & Saier, M. H. J. (1996) Gene 181, 103–108. 61. Sorensen, R. U. & Joseph, F. J. (1993) in Pseudomonas aeruginosa as an Opportunistic Pathogen, eds. Campa, M., Bendenelli, M. & Friedman, H. (Plenum, New York), pp. 43–57. 62. Mavrodi, D. V., Ksenzenko, V. N., Bonsall, R. F., Cook, R. J., Boronin, A. M. & Thomashow, L. S. (1998) J. Bacteriol. 180, 2541–2548. 63. O'Toole, G. A. & Kolter, R. (1998) Mol. Microbiol. 28, 449–461.


ROLE OF THE CYSTIC FIBROSIS TRANSMEMBRANE CONDUCTANCE REGULATOR IN INNATE IMMUNITY TO PSEUDOMONAS AERUGINOSA INFECTIONS

8822

Colloquium Role of the cystic fibrosis transmembrane conductance regulator in innate immunity to Pseudomonas aeruginosa infections Gerald B. Pier* Channing Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, MA 02115-5899 Chronic Pseudomonas aeruginosa infection occurs in 75–90% of patients with cystic fibrosis (CF). It is the foremost factor in pulmonary function decline and early mortality. A connection has been made between mutant or missing CF transmembrane conductance regulator (CFTR) in lung epithelial cell membranes and a failure in innate immunity leading to initiation of P. aeruginosa infection. Epithelial cells use CFTR as a receptor for internalization of P. aeruginosa via endocytosis and subsequent removal of bacteria from the airway. In the absence of functional CFTR, this interaction does not occur, allowing for increased bacterial loads in the lungs. Binding occurs between the outer core of the bacterial lipopolysaccharide and amino acids 108–117 in the first predicted extracellular domain of CFTR. In experimentally infected mice, inhibiting CFTR-mediated endocytosis of P. aeruginosa by inclusion in the bacterial inoculum of either free bacterial lipopolysaccharide or CFTR peptide 108–117 resulted in increased bacterial counts in the lungs. CFTR is also a receptor on gastrointestinal epithelial cells for Salmonella enterica serovar Typhi, the etiologic agent of typhoid fever. There was a significant decrease in translocation of this organism to the gastrointestinal submucosa in transgenic mice that are heterozygous carriers of a mutant ∆F508 CFTR allele, suggesting heterozygous CFTR carriers may have increased resistance to typhoid fever. The identification of CFTR as a receptor for bacterial pathogens could underlie the biology of CF lung disease and be the basis for the heterozygote advantage for carriers of mutant alleles of CFTR. N either novices nor experts dispute the complexity of the immune system. A multitude of physiologic, cellular, and molecular factors work together to identify foreign antigens (particularly of the harmful microbial variety), respond to them, and eliminate them. Responses start almost immediately after contact, and during the earliest phases of infection, mammals rely on the innate immune system to rid themselves of harmful microbes. Fortunately, this system works efficiently most of the time; however, susceptibility to serious microbial infection increases when breakdowns occur in this system. Compromising key innate immune factors such as the physical barrier of skin (e.g., as a result of severe wounds or trauma), disruption of mucosal barriers (e.g., by injury, chemotherapy, or radiation), or acquired or induced suppression of phagocytic cell function all lead to markedly enhanced rates of microbial infection, particularly with bacterial pathogens. Among the more puzzling failures of the innate immune system are the ones associated with defects in the gene encoding the cystic fibrosis (CF) transmembrane conductance regulator (CFTR), leading to chronic lung infection with an unusual variant of the ubiquitous Gram-negative pathogen Pseudomonas aeruginosa. Over 80% of patients with CF acquire a P. aeruginosa infection that results in progressive loss of lung function and early death (Fig. 1). The initially acquired strain is typically an environmental isolate (1) that expresses a smooth lipopolysaccharide (LPS) containing O side chains and little or no extracellular mucoid exopolysaccharide (alginate). After an indeterminate—but probably short— time, the organism phenotypically changes to an LPS-rough (i.e., lacking O side chains), mucoid exopolysaccharide-hyperexpressing (i.e., mucoid) variant. These changes are correlated with an acceleration in the decline in lung function in patients with CF (2, 3). The challenge in connecting innate immunity at mucosal surfaces with the clinical aspects of CF is to explain how defects in a chloride ion channel lead to such a high level of infection with one predominant microbial pathogen.

CF CF arises from mutations in the gene encoding CFTR (4, 5), a plasma membrane protein involved in chloride ion secretion (6) and regulation of other ion channels (7). The consequences of these mutations at the protein level are varied, given that hundreds of different mutations in CFTR (898 as of April 2000) have been reported. The ∆F508 CFTR allele, which results from an in-frame 3-bp deletion in the CFTR gene and the subsequent loss of a phenylalanine at position 508, accounts for two-thirds of all mutant CFTR alleles. About one-half of patients with CF are homozygous for this genotype. In general, nonsense or stop mutations in CFTR result in severe disease because of a lack of plasma-membrane CFTR. The ∆F508 CFTR allele gives rise to a misfolded protein that does not make it out of the endoplasmic reticulum and is degraded in cytoplasmic proteosomes after multiubiquitination. Some mutations associated with severe disease nonetheless give rise to a mutant protein in the membrane, but the mutant protein cannot properly mediate chloride ion conductance and perhaps other critical functions of CFTR. Before modern medical management of this disease, the major clinical manifestations of CF occurred in the gastrointestinal (GI) tract, and intestinal blockade and malnutrition were prominent factors in death before 1 year of age. As might be expected, there is high-level expression of CFTR in the GI epithelium, principally in the crypts. For the past 30–40 years, the GI symptoms have been managed adequately, and the major clinical problem for patients with CF has been progressive loss of pulmonary function over many years caused by chronic bacterial infection with mucoid P. aeruginosa (4, 5, 8). Thus, the molecular and cellular connections between lung infection and defects in CFTR have been of great interest as the primary determinant of the overall clinical status of patients with CF.

This paper was presented at the National Academy of Sciences colloquium “Virulence and Defense in Host–Pathogen Interactions: Common Features Between Plants and Animals,” held December 9–11, 1999, at the Arnold and Mabel Beckman Center in Irvine, CA. Abbreviations: CF, cystic fibrosis; CFTR, cystic fibrosis transmembrane conductance regulator; ASL, airway surface liquid; LPS, lipopolysaccharide; GI, gastrointestinal. * E-mail: [email protected].



8823

Fig. 1. Inverse relationship between isolation of mucoid P. aeruginosa (but not Staphylococcus aureus or Haemophilus influenzae) and decline in percentage of predicted FEV1, as compiled from the CF Foundation Patient Registry database for 1998. MICROBIAL ASPECTS OF LUNG INFECTION IN CF Despite a complex sputum bacteriology, the progressive decline in pulmonary function that is the hallmark of CF is mostly attributable to a single pathogen, mucoid P. aeruginosa (8). Patients with CF become colonized and sometimes infected with a variety of potential pathogens; S. aureus and nontypable H. influenzae, for example, are common bacterial isolates with pathogenic potential from cultures of CF respiratory tract secretions (Fig. 1). However, there are no compelling data that indicate that either S. aureus or H. influenzae contributes to lung function decline in CF except on the rare occasions when they cause acute pneumonia, empyema, or a similar infection. Indeed, it remains to be determined whether antistaphylococcal therapy in patients with CF confers clinical benefit or harm (9). One unpublished but completed clinical trial of daily antistaphylococcal therapy found no clinical benefit from potent suppression of S. aureus carriage in CF respiratory secretions (H. Stutman, unpublished work). Often after prolonged mucoid P. aeruginosa infection, patients with CF become superinfected with organisms such as Burkholderia cepacia, Aspergillus spp., and atypical mycobacteria (8, 10). In rare instances, patients become infected with virulent microbial pathogens in the absence of P. aeruginosa. Any microbial pathogen can potentially cause serious infection in patients with CF, but only mucoid P. aeruginosa appreciably contributes to the characteristic chronic and progressive decline in pulmonary function (Fig. 1). Studies clearly show that patients with CF harboring only nonmucoid P. aeruginosa and S. aureus maintain 80% of their predicted lung function (2, 3) and that the presence of S. aureus in the absence of mucoid P. aeruginosa actually predicts long-term survival for patients with CF (11).

Fig. 2. Molecular consequences of CFTR mutations. [Reproduced with permission from ref. 17 (Copyright 1995, Lap Chee Tsui)]. The reason that patients with CF initially acquire and fail to eliminate environmental strains of P. aeruginosa is enigmatic. Patients with CF have normal immune function, and the relationship between defects in chloride ion conductance of mutant CFTR and hypersusceptibility to P. aeruginosa infection is not fully elucidated. Recent work supports the idea that, starting at an early age, many patients with CF harbor microbial pathogens in their lungs that can be detected only by invasive techniques such as bronchoalveolar lavage; thus, it may be difficult to know exactly when infection is initiated (12, 13 and 14). Infecting strains of P. aeruginosa do seem to be fairly stable, and most patients harbor a single major clone of P. aeruginosa for many years (15, 16). Little is known about how this otherwise virulent and highly immunogenic pathogen establishes a low-level infection that fails to elicit an appropriate host response leading to prompt bacterial elimination. It is not unreasonable to consider that patients with CF may, in fact, respond to and eliminate many strains of P. aeruginosa until some combination of factors interferes with elimination of one particular strain and chronic infection is initiated. It is not known when the switch to the LPS-rough, mucoid phenotype occurs; however, this change may take place quite early after the initial infection, inasmuch as most patients harbor the nonmucoid phenotype for short periods and the mucoid phenotype for years (1).



8824

Fig. 3. Invasion of transformed airway epithelial cell lines by three strains of P. aeruginosa (two clinical isolates from patients with CF, 324 and 149, and a laboratory strain, PAO1). Cells were grown for 72 h at the temperature indicated in each figure (“Cells”). Bacteria were allowed to invade the epithelial cells for 3 or 4 h at the temperature indicated by “Invasion” on the figures. Bars indicate the means of the determinations, and error bars indicate the standard deviation of the mean. (A) Cells grown at 37°C, a temperature that inhibits membrane expression of ∆F508 CFTR (44); the invasion assay was carried out for 4 h at 37°C. Ingestion of P. aeruginosa by cells expressing wild-type CFTR was significantly higher than in cells lacking wildtype CFTR (P < 0.01, ANOVA). (B) Cells grown at 26°C and invasion assessed at 26°C. (C) Cells grown at 26°C and invasion assessed at 37°C. In both assays where cells were grown at 26°C to promote surface expression of ∆F508 CFTR (44), there were no significant differences (P 0.2, ANOVA) in bacterial invasion among the cell lines for any P. aeruginosa strain tested. CFTR AND INITIATION OF P. AERUGINOSA INFECTION The consequences of mutations in the CFTR gene to the production of CFTR protein vary widely because of the multitude of mutations in this gene (Fig. 2, originally published in ref.17). Most—but not all (18)—of the mutations that lead to severe pulmonary disease result in a lack of CFTR protein in the plasma membrane. Because the principal function of CFTR identified to date has been as a chloride ion channel, many of the recent theories connecting loss of CFTR protein in the apical membrane of airway cells with chronic bacterial infections have invoked disruption of airway fluid composition subsequent to decreased transport of chloride ions as a primary mechanism leading to lung infections. For example, Smith et al. (19) proposed that defective chloride ion transport might lead to an increased NaCl content of airway surface liquid (ASL) and that elevated levels of NaCl might interfere with a putative antimicrobial factor present in the airways of both healthy persons and patients with CF. However, the principal antibacterial factors in ASL isolated by this group turned out to be lysozyme and lactoferrin (20). The antimicrobial activity of individual factors was inhibited at high ionic strengths, but this effect, in turn, was overcome with increased concentrations of lysozyme and lactoferrin. Goldman et al. (21) identified a salt-sensitive human β-defensin in the lung that could be compromised in CF if the ASL was hypertonic. However, an elevated salt content in the ASL of patients with CF was not seen in studies by Knowles and colleagues (22), nor was this elevated salt content seen in tracheal xenografts from CF fetuses grown in the flanks of immunodeficient mice (23, 24). Boucher and colleagues (25) proposed that the CF airway epithelia had abnormal levels of fluid adsorption and were actually dehydrated and less able to clear microbes because of loss of mucus transport. Whether nonfunctional antimicrobial constituents of the ASL or increased viscosity may contribute to hypersusceptibility of patients with CF to infection has not yet been tested in animal models or clinical trials. Although disruptions to the normal composition and/or physiologic function of ASL could clearly be a component of the pathology of lung infection in CF, the lack of specificity for P. aeruginosa of these components of innate immunity leaves a gap in our understanding of why most patients with CF are chronically infected by this one bacterial pathogen. It has also been suggested that increased adherence of P. aeruginosa to airway epithelial cells is a critical component for the initiation of infection in patients with CF (26, 27). The putative receptors are cell-surface residues expressing fucose (28) and asialo GM1 (29); these residues reportedly are more numerous on CF cells (26, 28, 30). Increased asialo GM1 levels have been reported to result from defective organelle acidification that prevents proper addition of neuraminic acid to asialo GM1 residues (31). However, other studies have not confirmed defects in either increased adherence of P. aeruginosa to CF cells (32) or defective cellular physiology (33). In addition, the differences in adherence of P. aeruginosa to CF and non-CF cells is modest; P. aeruginosa is only 10–15% more adherent to ∆F508 (FTR homozygous cells) than to normal cells (34). Even if this increased bacterial binding is an accurate reflection of the biology of the CF cells, it does not seem likely that this small difference contributes to the 80% infection rate in CF versus essentially no mucoid P. aeruginosa infection in age-matched, healthy patients. Moreover, increased P. aeruginosa adherence to cells has been observed only when using primary cultures of nasal polyp cells from patients with CF homozygous for the ∆F508 CFTR allele when adherence is compared with that seen with primary respiratory epithelial cell cultures from healthy patients without CF and heterozygous carriers of mutant CFTR alleles (30, 35). Cultured nasal polyp cells from patients with CF who are homozygous for other mutant CFTR alleles did not bind P.



8825

aeruginosa any differently than cells from individuals without CF, but the clinical status of homozygous ∆F508 CFTR patients and that of the patients with CF with other genotypes was indistinguishable (34). A human tracheal epithelial cell line engineered to overproduce the regulatory domain of CFTR also had modestly increased adherence of a single laboratory strain of P. aeruginosa (36), but this cell line is the only non∆F508 CFTR cell line to have comparable increased binding of P. aeruginosa. It seems difficult to reconcile the finding that increased adherence of P. aeruginosa to epithelial cells early in the course of disease contributes significantly to the onset of infection with the observation that this increased adherence is measured only in cells from patients homozygous for the ∆F508 CFTR allele and a laboratory cell line not representative of any known human CFTR mutation. No in vivo studies in humans or transgenic CF mice have compared the adherence of P. aeruginosa to fucosylated or asialo GM1 receptors on epithelial cells with a CF phenotype to adherence on wild-type cells; thus, the role of binding of P. aeruginosa to these receptors as a component of CF lung disease is predicated only on limited in vitro observations.

LUNG EPITHELIAL CELLULAR INTERNALIZATION OF P. AERUGINOSA REQUIRES CFTR BINDING TO BACTERIAL LPS CORE OLIGOSACCHARIDE Internalization by epithelial cells may be a mechanism for clearing bacteria from the lung via cellular desquamation of internalized organisms. Other cellular reactions subsequent to ingestion may also modulate the host response to P. aeruginosa on the lung epithelium. Along these lines, Hultgren and colleagues (37) showed that piliated Escherichia coli binds to uroplakins on bladder epithelial cells to initiate infection, and shedding of these cells with bound and internalized organisms promotes their removal. Similar findings for bladder infections were reported by Aronson and colleagues (38, 39 and 40). We have proposed that epithelial cell ingestion of P. aeruginosa may result in cellular desquamation or shedding, with removal of the ingested bacteria from the epithelium, and that ingestion of P. aeruginosa is highly compromised in cells expressing the ∆F508 allele of CFTR compared with that in cells expressing wild-type CFTR (refs.41 and 42; Fig. 3A). The dependence of efficient P. aeruginosa internalization on intact CFTR expression was documented in a number of studies (41, 42 and 43). The ∆F508 CFTR allele encodes a CFTR protein with a temperature-sensitive defect (44). We found that ingestion of P. aeruginosa by ∆F508 CFTR cells was enhanced by conditions that increased cell-surface expression of the ∆F508 CFTR protein (growth at 26°C; Fig. 3B). When ∆F508 CFTR protein was expressed in the cell plasma membrane, uptake of P. aeruginosa by these airway cells was comparable to uptake by cells expressing wild-type CFTR in the plasma membrane. We then identified the outer-core oligosaccharide of the P. aeruginosa LPS as the bacterial ligand needed for efficient uptake (41). Further studies indicated that CFTR itself is the epithelial cell receptor that binds to P. aeruginosa to promote bacterial internalization (42). Because CFTR is thought to have only a modest number of exposed extracellular loops projecting from the plasma membrane, we investigated whether any of these loops could function as a receptor for P. aeruginosa. We found that amino acids 103–117— and probably only amino acids 108–117, predicted to reside in the first extracellular loop of CFTR—serve as the receptor component for binding and internalization of P. aeruginosa (42). Synthetic CFTR peptide 103–117 and purified P. aeruginosa LPS with a complete-core oligosaccharide bind to each other in a specific fashion (42). Recent studies have shown that, within a collection of cultured epithelial cells, there is nonhomogeneous expression of CFTR, and expression levels correlate with uptake of P. aeruginosa (45). Using a Madin–Darby canine kidney cell line transfected with a

Fig. 4. Effect of adding inhibitors of the CFTR–P. aeruginosa interaction on the course of infection in neonatal mouse lungs. (A) Addition of LPS core oligosaccharides to the P. aeruginosa challenge inoculum reduces cellular uptake and increases bacterial loads in the lung. Closed circles, no inhibitor; open squares, LPS core oligosaccharide, 10 µg/ml; closed squares, LPS incomplete core oligosaccharide control, 10 µg/ml. Each symbol indicates the median number of bacterial colony-forming units for 8–10 lungs obtained from each group, and the bars indicate the upper and lower quartiles. Differences among groups were analyzed by nonparametric statistics [P < 0.0001; Kruskal–Wallis nonparametric ANOVA; P < 0.001; Dunn procedure for individual pairwise differences between the groups at 1 and 24 h; also at 1 h, the group receiving the incomplete core oligosaccharide had a reduced level (P = 0.05; Dunn procedure) of intracellular bacteria compared with the group receiving nothing along with the inoculum]. At 48 h, the group treated initially with complete-core inhibitor had significantly more bacteria in the lungs (P = 0.003; Kruskal–Wallis; P < 0.05; Dunn procedure for all pairwise comparisons). (B) Effect of addition of synthetic peptides to the bacterial inoculum on P. aeruginosa infection in neonatal mice. (Upper) Amount of P. aeruginosa internalized by lung cells 24 h after infection. (Lower) Total amount of P. aeruginosa found in lungs 24 h after infection. Box plots indicatefrom bottom to top–the 10th, 25th, 50th (median), 75th, and 90th percentiles. Circles above or below the 90th or 10th percentile indicate individual points outside this range. There were 12–14 total lung samples used in each group. For both groups of comparisons (A and B), the overall differences were significant at P < 0.001 (Kruskal–Wallis nonparametric ANOVA test), and the difference between the group receiving the first-domain peptide and the other three groups was significant at P < 0.001 (Dunn procedure for pairwise comparisons).



8826

green fluorescent protein-tagged CFTR (46), we found that cells were distributed into a population that expressed large amounts of CFTR (

(NAS Colloquium) Virulence and Defense in Host--Pathogen Interactions (National Academy of Sciences Colloquium Series)

(Sackler NAS Colloquium) Neural Signaling (National Academy of Sciences Colloquium Series)

(Sackler NAS Colloquium) Nanoscience

(Sackler NAS Colloquium) Nanoscience

(NAS Colloquium) Plants and Population

(NAS Colloquium) Computational Biomolecular Science

(NAS Colloquium) Geology, Mineralogy, and Human Welfare

National Academy of Sciences Colloquium, Carbon Dioxide and Climate Change; Proceedings of the National Academy of Sciences

(NAS Colloquium) Proteolytic Processing and Physiological Regulation

(NAS Colloquium) Science, Technology and the Economy

(NAS Colloquium) The Neurobiology of Pain

(NAS Colloquium) Science, Technology and the Economy

(NAS Colloquium) Molecular Kinesis in Cellular Function and Plasticity

(NAS Colloquium) Genetic Engineering of Viruses and Viral Vectors

Molecular Kinesis in Cellular Function and Plasticity (NAS Colloquium)

(NAS Colloquium) Auditory Neuroscience: Development, Transduction, and Integration

Logic Colloquium '73

Colloquium on Methods of Optimization

Logic Colloquium 1984: Proceedings

Logic Colloquium '69

Dynamical Systems (Colloquium Publications)

Dynamical Systems (Colloquium Publications)

Algebraic Topology (Colloquium Pbns. Series, Vol 27)

Algebraic Topology (Colloquium Pbns. Series, Vol 27)

Logic: Colloquium Proceedings, 1977

Logic Colloquium 1982: Proceedings

Logic 1985: Colloquium Proceedings

Logic colloquium '88

Logic Colloquium '88: Proceedings

Logic colloquium 2007

Logic Colloquium 1976: Proceedings

(NAS Colloquium) Virulence and Defense in Host--Pathogen Interactions (National Academy of Sciences Colloquium Series)

(Sackler NAS Colloquium) Neural Signaling (National Academy of Sciences Colloquium Series)

(Sackler NAS Colloquium) Nanoscience

(Sackler NAS Colloquium) Nanoscience

(NAS Colloquium) Plants and Population

(NAS Colloquium) Computational Biomolecular Science

(NAS Colloquium) Geology, Mineralogy, and Human Welfare

National Academy of Sciences Colloquium, Carbon Dioxide and Climate Change; Proceedings of the National Academy of Sciences

(NAS Colloquium) Proteolytic Processing and Physiological Regulation

(NAS Colloquium) Science, Technology and the Economy

(NAS Colloquium) The Neurobiology of Pain

(NAS Colloquium) Science, Technology and the Economy

(NAS Colloquium) Molecular Kinesis in Cellular Function and Plasticity

(NAS Colloquium) Genetic Engineering of Viruses and Viral Vectors

Molecular Kinesis in Cellular Function and Plasticity (NAS Colloquium)

(NAS Colloquium) Auditory Neuroscience: Development, Transduction, and Integration

Logic Colloquium '73

Colloquium on Methods of Optimization

Logic Colloquium 1984: Proceedings

Logic Colloquium '69

Dynamical Systems (Colloquium Publications)

Dynamical Systems (Colloquium Publications)

Algebraic Topology (Colloquium Pbns. Series, Vol 27)

Algebraic Topology (Colloquium Pbns. Series, Vol 27)

Logic: Colloquium Proceedings, 1977

Logic Colloquium 1982: Proceedings

Logic 1985: Colloquium Proceedings

Logic colloquium '88

Logic Colloquium '88: Proceedings

Logic colloquium 2007

Logic Colloquium 1976: Proceedings

Recommend Documents