INTRODUCTION

Top Previous Next References

"Things derive their being and nature by mutual dependence and are nothing in themselves."

Nagarjuna, 2nd century Buddhist philosopher

"Such is the power, sometimes called malignant, sometimes benign, that Anastasia, the treacherous city, possesses; if for 8 hours a day you work as a cutter of agate, onyx, chrysoprase, your labour which gives form to desire takes from desire its form, and you believe you are enjoying Anastasia wholly when you are only its slave."

Italo Calvino, Invisible Cities

The elucidation of the molecules and mechanisms underlying bacterial pathogenesis in humans, animals, and plants is a major focus of microbiological research which yields practical applications ranging from refined diagnostics to new antibiotics and improved vaccines. In addition to the pursuit of these practical purposes, recent research on bacterial pathogenesis has allowed insight into the complex beauty of highly adapted interactions between pathogens and their hosts at the cellular and molecular levels. The relative ease of genetic and biochemical analyses of bacteria and widely used tissue culture models of bacterial infection have dramatically increased our knowledge about the molecular components involved in pathogen-host cell interactions, leading to the emergence of cellular microbiology as a new area of microbiological investigation. Importantly, the availability of suitable animal and plant models has allowed the assessment of the contribution of bacterial pathogenicity factors analyzed in vitro to the ultimate outcome of disease.

Genetic analyses of bacterial virulence factors has shown that pathogens are distinguished from their nonpathogenic relatives by the presence of specific pathogenicity genes, often organized in so-called pathogenicity islands, clusters of genes which apparently have been acquired during evolution via horizontal genetic transfer. Thus, distantly related pathogens have turned out to harbor closely related virulence genes. This point has become particularly apparent for a set of approximately 20 genes which together encode a pathogenicity mechanism termed type III secretion. Type III secretion enables gram-negative bacteria to secrete and inject pathogenicity proteins into the cytosol of eukaryotic host cells. Fascinatingly, while the type III secretion apparatus is conserved in pathogens as distantly related as Yersinia and Erwinia, the secreted proteins differ entirely, illustrating how one bacterial pathogenicity mechanism can give rise to a multitude of diseases that range from bubonic plague in humans to fire blight in fruit trees.

Secretion of bacterial pathogenicity proteins by the type III pathway and their injection into the cytosol of animal or plant cells initiates a sophisticated "biochemical cross-talk" (defined by J. E. Galán) between pathogen and host. The injected proteins often resemble eukaryotic factors with signal transduction functions and are capable of interfering with eukaryotic signalling pathways. Redirection of cellular signal transduction may result in disarmament of host immune responses or in cytoskeletal reorganization, establishing subcellular niches for bacterial colonization and facilitating a highly adapted pathogenic strategy of "stealth and interdiction" (defined by J. B. Bliska) of host defense communication lines.

This review comprehensively describes the type III protein secretion systems known to date, which are present in the animal pathogens Yersinia spp., Shigella flexneri, Salmonella typhimurium, enteropathogenic Escherichia coli (EPEC), Pseudomonas aeruginosa, and Chlamydia spp. and in the plant pathogens Pseudomonas syringae, Erwinia spp., Ralstonia (formerly Pseudomonas) solanacearum, Xanthomonas campestris, and Rhizobium spp. For a shorter and more concise overview of type III secretion, refer to the recent review by C. A. Lee (264).

Interaction of bacterial pathogens with host cells is particularly characterized by factors that are located on the bacterial surface or are secreted into the extracellular space. Although the secreted bacterial proteins are numerous and diverse and exhibit a wide variety of functions that include proteolysis, haemolysis, cytotoxicity, and protein phosphorylation and dephosphorylation, only a few pathways exist by which these proteins are transported from the bacterial cytoplasm to the extracellular space. Thus, four pathways of protein secretion (types I to IV) have been described in gram-negative bacteria (114, 122, 391, 447). A fifth system for macromolecular secretion is involved in conjugal transfer of plasmids, T-DNA transfer by Agrobacterim tumefaciens, and secretion of Bordetella pertussis toxin. The last system, which could be named type V secretion, is the least well characterized and comprises only three members so far (481).

In this context, the term "secretion" is used to describe the active transport of proteins from the cytoplasm across the inner and outer membranes into the bacterial supernatant or onto the surface of the bacterial cell. Secretion is distinguished from export, which refers to the transport of proteins from the cytoplasm into the periplasmic space (362, 391).

Type II and type IV sec-dependent secretion pathways. Type II and IV protein secretion pathways involve a separate step of transport across the inner membrane prior to transport across the cell envelope. While these pathways differ in the way in which the proteins are transported across the outer membrane, export to the periplasm occurs via the sec system in both cases. (Secretion of pertussis toxin by B. pertussis, which may be categorized as type V secretion [see above], also involves the sec pathway.) A signature of sec-dependent protein export is the presence of a short (about 30 amino acids [aa]), mainly hydrophobic amino-terminal signal sequence in the exported protein. The signal sequence aids protein export and is cleaved off by a periplasmic signal peptidase when the exported protein reaches the periplasm. In E. coli, the sec pathway comprises a number of inner membrane proteins (SecD to SecF, SecY), a cytoplasmic membrane-associated ATPase (SecA) that provides the energy for export, a chaperone (SecB) that binds to presecretory target proteins, and the periplasmic signal peptidase (see Fig. 1). A number of accessory proteins are also required for normal function (323, 362).

Type III sec-independent pathway. Like the type I secretion pathway, type III secretion is independent of the sec system. (Assembly of the type III secretion apparatus, however, probably requires the sec pathway, since several components of the type III secretion apparatus carry sec-characteristic amino-terminal signal sequences.) The type III secretion apparatus is composed of approximately 20 proteins, most of which are located in the inner membrane, and type III secretion requires a cytoplasmic, probably membrane-associated ATPase. Interestingly, most of the inner membrane proteins are homologous to components of the flagellar biosynthesis apparatus of both gram-negative and gram-positive bacteria, while an outer membrane protein of the type III secretion apparatus is homologous to PulD, the outer membrane secretin of the type II secretion pathway. Although type III secretion does not include distinct periplasmic intermediates of the secreted proteins, transport through the inner membrane is genetically separable from secretion through the outer membrane, since a mutant of the outer membrane PulD homolog of P. syringae was shown to accumulate considerable amounts of a secreted protein in the periplasm (72). As in type I and type II secretion, the genes encoding the type III secretion apparatus are clustered.

View larger version (33K): [in this window] [in a new window]   FIG. 1.   Schematic overview of the type I, II, and III secretion systems as exemplified by alpha-hemolysin secretion by E. coli (type I), pullulanase secretion by Klebsiella oxytoca (type II), and Yop secretion by Yersinia (type III). OM, outer membrane; PP, periplasm; IM, inner membrane; CP, cytoplasm. ATP hydrolysis by HlyB, SecA, and YscN is indicated. The localization of the secretion signals is shown in the secreted proteins (shaded). N, amino terminus; C, carboxy terminus. For type III secretion, the secretion signal may reside in the 5'-region of the mRNA encoding the secreted protein. Type II and type III secretion involve cytoplasmic chaperones (SecB and Syc, respectively) which bind to presecretory proteins. In type II secretion, the amino-terminal signal sequence is cleaved off by a periplasmic peptidase (LspA) after export of the protein via the sec pathway. Type II and type III secretion share a homologous multimeric outer membrane component (PulD, YscC), while the accessory proteins PulS and VirG, which facilitate outer membrane insertion of PulD and YscC, respectively, differ in the two systems. See the text for further details.

A variety of diverse gram-negative pathogens use type III secretion as a conserved and at the same time highly adapted virulence mechanism (Fig. 2). Although these pathogens use additional virulence factors (not discussed here), type III secretion is an essential basic virulence determinant. While the mechanism of protein secretion is conserved, the secreted proteins themselves are highly divergent, and the variety of diseases caused by these pathogens in different hosts is reflected by the multitude of type III secreted proteins. Many of the secreted proteins interact directly with host cell components to alter host cell signal transduction, and most of the secreted proteins act inside the eukaryotic cytosol into which they are translocated by the type III secretion mechanism. This section gives a short overview of the pathogenesis of bacteria which require type III secretion systems for virulence. The emphasis is put on the pathogenicity properties which are directly attributable to or correlated with type III secretion at the cellular and molecular levels.

View larger version (94K): [in this window] [in a new window]   FIG. 2.   Selected phenotypic effects of type III secretion pathogenicity mechanisms on host cells and host tissue. (Top left) Panel B shows how Y. pseudotuberculosis injects (translocates) YopH (immunostained, light) into the cytosol of HeLa cells. Translocation of YopH into macrophages results in inhibition of phagocytosis. Panel A shows a type III secretion mutant: YopH is detected only in association with the bacteria. Reprinted with permission from reference 345. (Top right) Panel B shows the cytotoxic effect of Y. pseudotuberculosis on cultured HeLa cells. Translocation of YopE leads to a collapse of the cytoskeleton. Panel A shows uninfected HeLa cells. Reprinted with permission from reference 383. (Middle left) Invasion of a polarized HEp-2 epithelial cell via the induction of membrane ruffling by S. typhimurium. This figure has previously appeared on the cover of Mol. Microbiol. 1995, vol. 18 no. 3. Reprinted with permission. (Middle right) Pseudopod (pedestal) formation induced by EPEC on HeLa epithelial cells. Reprinted with permission from reference 378. (Bottom left) Induction of apoptosis in macrophages infected with S. flexneri. Panel A shows an uninfected macrophage. Panel B shows an apoptotic macrophage infected with wild-type S. flexneri. Panel C shows a macrophage infected with a S. flexneri type III secretion mutant. Reprinted with permission from reference 509. (Bottom right) Induction of localized tissue necrosis (HR) in a tobacco leaf at sites of infiltration with Erwinia spp. (area 1), buffer alone (area 4), type III secretion mutants (areas 2 and 5), and complemented mutants (areas 3 and 6). Reprinted with permission from reference 34.

Three Yersinia species are pathogenic for humans and rodents. Y. pestis, the causative agent of bubonic plague, enters the host through flea bites or by inhalation, and invades and multiplies in regional lymph nodes corresponding to the infection point. Subsequent dissemination via the lymphatic system and bacteremia with necrotic and hemorrhagic lesions in many organs lead to death of the human or rodent host within 2 to 3 days after infection. Similar pathological symptoms are caused in rodents by Y. enterocolitica and Y. pseudotuberculosis. In humans, Y. enterocolitica causes a broad range of gastrointestinal syndromes, while Y. pseudotuberculosis, which is the least pathogenic of the three species for humans, may (rarely) cause a self-limiting gastroenteritis (86). Both Y. enterocolitica and Y. pseudotuberculosis usually enter their hosts via the oral route of infection and cross the intestinal barrier through specialized epithelial cells called M cells (179). M cells are associated with the follicle-associated epithelium overlying lymphoid follicles throughout the intestine and are especially concentrated in aggregates of lymphoid follicles called Peyer's patches (411). Like Y. pestis, Y. enterocolitica and Y. pseudotuberculosis exhibit a marked tropism for lymphatic tissue and persist and multiply in Peyer's patches, leading to high bacterial titers in these organs 12 to 24 h after infection. Subsequently, the bacteria colonize liver and spleen, and bacterial multiplication in these organs leads to death of the animal 3 to 4 days after infection (206, 429).

All three pathogenic Yersinia spp. exhibit a characteristic ability to resist the host primary immune defense, most notably by inhibiting their own uptake by professional phagocytes (66). Consequently, these pathogens are found mainly in extracellular locations during infection (179, 180, 412). The antiphagocytic effect is mediated by the Yersinia type III secretion system and specifically requires a protein with tyrosine phosphatase activity (169) called YopH (20, 112, 379). After contact of the bacteria with a macrophage, YopH is injected into the cytosol of the target cell (Fig. 2), where it catalyzes a rapid and specific dephosphorylation of several macrophage proteins, whose transient tyrosine phosphorylation appears to be required for normal phagocytosis (20). While earlier studies had observed YopH-dependent overall dephosphorylation of phosphotyrosine proteins in macrophages (50, 51), a more recent analysis identified the cytoskeleton-associated protein paxillin, which is involved in Fc receptor-mediated phagocytosis (165), as a target for YopH in macrophages. In addition to the antiphagocytic effect, Y. pseudotuberculosis resists killing by macrophages by an efficient inhibition of the Fc receptor-mediated oxidative burst, a phenotype which also requires the phosphatase activity of YopH (49). In HeLa cells, YopH was demonstrated to specifically dephosphorylate focal adhesion kinase and the focal adhesion-associated protein p120^cas, and dephosphorylation of these proteins appears to inhibit bacterial uptake by HeLa cells via inhibition of peripheral focal complex formation (48, 344).

In addition to inhibition of phagocytosis, Yersinia spp. are cytotoxic for cultured epithelial cells (Fig. 2) (355), and cytotoxicity also may contribute to the inactivation of macrophages (381). Cytotoxicity is specifically mediated by YopE, another protein secreted by the type III secretion pathway, which, like YopH, is injected into the eukaryotic cytosol (380, 381). YopE does not affect the integrity of the target cell membrane but, rather, causes a collapse of the cytoskeleton via disruption of actin microfilaments (380). The mechanism of actin disruption is probably indirect, since isolated Yops do not affect actin microfilament polymerization or stability in vitro (380).

Mutants with mutations in yopE and yopH of Y. pseudotuberculosis do not reach the same titers in mouse Peyer's patches as the wild type does, and they are cleared from these organs 4 days after oral infection (206). Therefore, both mutants are avirulent when administered orally (59, 130, 131). However, while a yopH mutant is highly attenuated even after intraperitoneal or intravenous (i.v.) infection (59, 381), a yopE mutant exhibits only a partial virulence defect when injected i.v. (381). Thus, YopE appears to function primarily at an early step in Yersinia pathogenesis whereas YopH is required throughout the infectious process.

In addition to YopE and YopH, Yersinia species secrete and translocate other virulence determinants into the cytosol of target cells by the type III secretion mechanism. A protein with serine/threonine kinase activity (YpkA) is required for later stages of infection (149). A ypkA mutant colonizes Peyer's patches similarly to the wild-type strain but is unable to colonize the spleen (150). Similarly, YopM, a translocated protein with thrombin binding activity (56, 268, 368), is required for later stages of mouse infection.

Taken together, the type III secretion mechanism enables Yersinia spp. to inject a number of essential virulence determinants into the cytosol of host target cells. The injected proteins appear to interfere with host cell signal transduction pathways and other cellular processes, allowing Yersinia spp. to obstruct the primary immune response and to establish a systemic infection.

Interestingly, suppression of several proinflammatory cytokines in various cell types appears to be another Yersinia virulence mechanism which has been associated with type III secretion and Yop translocation (44, 325, 407). Suppression of cytokines may account for the fact that infections with pathogenic Yersinia spp. proceed without eliciting a strong inflammatory immune response (324, 325, 412). Recently, the inhibition of an inflammatory host response was tentatively attributed to a low and regulated level of translocation of Yop proteins into target cells (205). A mutant with a mutation in a gene encoding the type III secreted protein YopK (206) was found to translocate significantly larger amounts of Yop proteins. Unexpectedly, the mutant was nevertheless impaired in later stages of mouse infection (206). The virulence defect of this mutant was interpreted to mean that this strain may elicit a stronger and eventually assertive host immune response due to an elevated level of Yop translocation (205).

Shigella spp. cause diarrheal disease with a wide range of clinical symptoms, the most severe form being bacillary dysentery, a bloody diarrhea originating from the colon. Shigellosis is endemic in developing countries, but outbreaks also occur in industrialized nations, especially under conditions of poor hygiene. Children under 5 years of age are the most susceptible victims, with over half a million deaths occurring annually worldwide (275, 292, 508).

Bacterial invasion of the colonic mucosa (258, 435) constitutes an essential step in the pathogenesis of shigellosis. The following model of the cellular basis of shigellosis has been derived from the integration of a variety of in vivo and in vitro observations (147). Similarly to other enteropathogens, shigellae gain initial access to the intestinal epithelium by invading M cells (342, 393, 460). After transcytosis through M cells, the bacteria encounter resident macrophages of the lymphoid tissue underlying the follicle-associated epithelium and resist killing by macrophages by the induction of cell apotosis (Fig. 2) (509, 510). Apoptosis of macrophages in turn leads to the release of significant quantities of interleukin-1 (506), which results in recruitment of polymorphonuclear leukocytes (PMNs) and massive infiltration of the infected tissue with these cells. S. flexneri is unable to invade a polarized epithelial cell monolayer from the apical pole, but transmigration of PMNs across the epithelium destabilizes the integrity of the intestinal barrier and allows S. flexneri to reach the basolateral epithelial cell pole, which is readily invaded by the bacteria (341, 342). Further bacterial invasion and lateral spreading of the bacteria within the epithelium result in the tissue destruction typically associated with Shigella infection (338). The ability of PMNs to kill S. flexneri suggests that these cells may not only contribute initially to the severe tissue damage characteristic of shigellosis but also ultimately participate in clearance and resolution of infection (284).

In in vitro tissue culture infection experiments, S. flexneri provokes its own uptake into nonpolarized epithelial cells by the induction of cytoskeletal rearrangements, resulting in the formation of localized membrane protrusions at the site of bacterial contact with the cell (Fig. 2). These so-called membrane ruffles coalesce around the entering bacterium, leading to its phagocytosis in a membrane-bound vacuole (2). In vivo invasiveness, as well as the induction of membrane ruffles on cultured epithelial cells, depends on the Shigella type III secretion system and has been extensively characterized at the biochemical level.

Four proteins which are secreted via the type III pathway, IpaA, IpaB, IpaC, and IpaD, are specifically involved in bacterial invasiveness (200, 304, 442), and two of them, IpaB and IpaC, are sufficient to induce membrane ruffling on epithelial cells when immobilized on the surface of latex beads (301). A potential receptor for IpaB and IpaC on target cells is the cell adhesion molecule ₅₁ integrin (433). IpaB, IpaC, and IpaD specifically bind to ₅₁ integrin and colocalize with the integrin in Chinese hamster ovary (CHO) cell infection (462). Overexpression of ₅₁ integrin in CHO cells led to enhanced S. flexneri invasiveness, and titration of Ipa proteins by extracellular addition of ₅₁ integrin decreased invasion in a concentration-dependent manner (462). Ligand-induced stimulation and clustering of integrins is known to cause protein tyrosine phosphorylation of a number of cytoskeletal proteins (78, 220), and protein tyrosine phosphorylation is required for efficient uptake of S. flexneri into CHO cells (462). Therefore, binding of the Ipa proteins to ₅₁ integrin may be the primary signal which leads to the induction of membrane ruffles and ultimately results in bacterial internalization.

Induction of membrane ruffling by S. flexneri involves the localized accumulation of a variety of cytoskeletal and signal transduction molecules at the site of bacterial attachment. Shigella attachment induces the accumulation of actin and several actin-binding proteins (myosin, plastin [2], and cortactin [93]). Furthermore, the focal adhesion plaque-associated proteins paxillin, -actinin (300), vinculin (442), talin, and ₅₁ integrin (462) accumulate at the site of bacterial entry. Recently, it was shown that overexpression of vinculin, a protein which is supposed to link the cytoskeleton to the cell membrane, results in increased uptake of S. flexneri by ASML cells and that this effect depends on IpaA (442). Wild-type bacteria were seen to closely associate with a coat of cytoskeletal proteins (F-actin, vinculin, and -actinin) during the entry process, and although recruitment of these proteins to the entry structure is independent of IpaA, tight association of the proteins with the bacteria was abolished in an ipaA mutant. Accordingly, the ipaA mutant is only partially deficient in invasion, while IpaB to IpaD mutants, which do not induce actin polymerization, are noninvasive. IpaA specifically binds to vinculin and thus appears to modulate bacterial invasion by "optimizing" the impact of invasive S. flexneri on host cell cytoskeletal rearrangements (442).

Entry of S. flexneri also leads to differential recruitment of the protein tyrosine kinase pp60^c-src (93) and the GTP-binding protein rho (3) to the entry structure. The focal adhesion kinase pp125^FAK, paxillin (462), and cortactin (93) have been identified as targets for protein tyrosine phosphorylation (93, 300). rho is a central signal transducer in Shigella invasiveness, since inhibition of rho activity results in complete loss of Shigella-induced cytoskeletal rearrangements and of Shigella invasiveness (3, 301, 463). Consistent with the involvement of rho in Shigella invasion, rho-dependent activation of protein kinase C was observed upon Shigella invasion, and inhibition of protein kinase C with various inhibitors greatly reduced Shigella invasiveness (463).

Shortly after internalization, the bacteria lyse the phagocytic membrane and gain access to the cytoplasm (398). In macrophages, S. flexneri induces apoptosis after escape from the phagosome (437, 509). Like invasion of epithelial cells, the ability to lyse the endocytic vacuole and to induce apoptosis in macrophages depends on the S. flexneri type III secretion system. While vacuolar lysis appears to require all three type III secreted Ipa proteins (200, 304), induction of apoptosis relies specifically on IpaB (507), which triggers the apoptotic pathway by direct binding to interleukin-1-converting enzyme (ICE) or a homologous protease (75). The relevance of ICE activation for S. flexneri-induced apoptosis was elegantly shown by reversible inhibition of ICE, which resulted in reversible inhibition of S. flexneri cytotoxicity for macrophages (75). In addition to the induction of apoptosis, binding of IpaB to ICE leads to increased production of mature interleukin-1 by activated ICE (75), which, in vivo, results in attraction of PMN to the site of bacterial infection (see above).

Salmonella spp. infect a variety of vertebrate hosts and cause a broad spectrum of diseases (including gastroenteritis, bacteremia, and enteric fever) which originate from enteric infection. S. typhi causes systemic typhoid fever in humans and constitutes a major health problem in underdeveloped regions of the world, with an estimated 16 to 17 million cases and 0.6 million deaths annually (335), while S. typhimurium and S. enteritidis are increasing causes of food poisoning in industrialized countries (309). Several Salmonella spp. are specifically adapted to a particular host (including S. typhi [humans], S. pullorum [poultry], S. dublin [cattle], and S. arizonae [reptiles]), while others (S. typhimurium, S. enteritidis, and S. choleraesius) exhibit a broader host spectrum but may cause different diseases in different hosts. For example, S. typhimurium causes systemic and lethal infection in susceptible mice at very low doses of infection, while in humans even high doses of this organism only cause a self-limiting gastroenteritis (309).

In systemic infections in mice, Salmonella penetrates the mucosa of the small intestine by preferentially adhering to and invading M cells of the Peyer's patches (79). The bacteria eventually destroy the invaded M cells and the adjacent epithelium (92, 228, 249), thus gaining access to the underlying lymphoid tissue. M-cell invasion is an active event, since the bacteria induce ruffles of the apical cell membrane, which eventually engulf the entering organism (434). Salmonella-induced M-cell membrane ruffles are similar to the localized membrane ruffling observed on cultured epithelial cells after contact with the bacteria (Fig. 2), and, as with epithelial cell invasion by Shigella (see above), the Salmonella-induced ruffles involve rearrangements of the actin cytoskeleton (124, 134). Despite the morphological similarity of membrane ruffles induced by Shigella and Salmonella spp., invasion of epithelial cells by S. typhimurium appears to differ from that by S. flexneri. For example, while Shigella flexneri invasion is inhibited by various protein kinase C inhibitors, S. typhimurium invasion is not (463). Therefore, Salmonella invasion appears to be independent of the rho-controlled signal transduction pathway. In contrast, S. typhimurium requires CDC42 for invasion (73). Furthermore, S. typhimurium does not induce phosphorylation of the focal adhesion kinase p125^FAK or of paxillin (463). A further difference, which may be significant to in vivo invasiveness, is the fact that in contrast to Shigella flexneri, S. typhimurium efficiently invades polarized epithelial cells via the apical cell pole (156).

Induction of membrane ruffles and invasion of M cells and cultured epithelial cells require the type III secretion system encoded by SPI-1 (228, 340), and in parallel to the similar invasive capacities of S. typhimurium and Shigella flexneri, the S. typhimurium type III secreted invasion proteins (called Sip or Ssp) are homologous to the Ipa invasins from Shigella flexneri (216, 233, 234). However, S. typhimurium SPI-1 mutants are only slightly attenuated for mouse virulence after oral infection and show no defect in virulence when injected by the intraperitoneal route (143, 227, 340). Thus, it seems possible that invasion-negative mutants are still taken upat least to some degreeby the naturally phagocytic M cells and that this uptake is sufficient to cause systemic disease. The role of the SPI-1-encoded type III secretion systems in mouse typhoid is therefore restricted to the early stages of infection. Interestingly, tissue tropism of S. typhimurium toward M cells appears to be mediated by species-specific fimbriae (lpf) (36). Like SPI-1 mutants, mutants with mutations in lpf show only a slight virulence defect when administered by the oral infection route, which may be due to SPI-1-mediated invasion of the intestinal epithelium at sites other than M cells. However, combination of lpf and SPI-1 mutations significantly increased the 50% lethal dose for mice (37). Thus, SPI-1 and lpf appear to function synergistically in invasion of the intestinal barrier.

In the lymphoid tissue underneath invaded M cells, salmonellae are phagocytosed by residential macrophages. In contrast to Shigella flexneri, salmonellae do not escape from the phagocytic vacuole. Instead, they survive and replicate inside the vacuole and probably use macrophages as vehicles to disseminate via the host lymphoid system. The bacteria accumulate and massively replicate in the liver, spleen, and bone marrow, organs which are rich in phagocytic cells, leading to organ failure, bacterial sepsis, and death 4 to 6 days after infection. The ability to survive and replicate in professional phagocytes is therefore thought to be an essential virulence determinant of salmonellae that cause systemic infection (16, 119, 273, 311, 331, 406, 453).

Besides its requirement for epithelial cell invasion, the SPI-1-encoded type III secretion system recently has been shown to affect the interaction of Salmonella spp. with cultured murine macrophages (74). However, SPI-1 does not affect the ability to persist and replicate in macrophages but, rather, mediates a cytotoxic effect. Similar to induction of apoptosis by Shigella flexneri, SPI-1 encoded effector proteins induce apoptosis in cultured macrophages (74, 315). However, since SPI-1 mutants are only slightly attenuated for mouse virulence when administered orally (see above), the cytotoxic effect appears to play only a minor role in early stages of infection.

It is interesting that both shigellae and salmonellae use homologous secreted proteins (Ipa and Sip/Ssp, respectively) for similar phenotypes: invasion of epithelial cells and induction of apoptosis in macrophages. The SipB protein from S. typhi, which is highly similar to the respective S. typhimurium protein, even transcomplemented a Shigella flexneri ipaB mutant for invasiveness (197). However, transcomplementation was only partial, indicating that IpaB and SipB are not fully interchangeable. Surprisingly, although Sip/SspB does not allow salmonellae to escape from the phagocytic vacuole, SipB from S. typhi restored the ability of a Shigella flexneri ipaB mutant to lyse the vacuole. Furthermore, Shigella flexneri appears to require vacuolar lysis to be able to induce apoptosis via IpaB, while Salmonella spp., although similarly inducing apoptosis, do not exhibit the same requirement (74). These differences may be due to the fact that Sip/Ssp proteins are translocated across the cell membrane into the eukaryotic cytosol via the type III secretion mechanism (83), while membrane translocation of Ipa proteins has not been reported.

The SPI-1-encoded type III secretion system appears to play a major role in nontyphoidal Salmonella infections of the intestinal epithelium. A characteristic feature of Salmonella-induced enteritis is an intense intestinal secretory and inflammatory response including the induction of PMN transmigration through the intestinal epithelium (295). Transepithelial signalling requires the SPI-1-encoded type III secretion system and has been particularly associated with the type III secreted protein SopB of S. dublin (151).

The second Salmonella type III secretion system, encoded by SPI-2, has been identified by virtue of its large impact on S. typhimurium mouse virulence (195, 410). However, mutants of SPI-2 have no (196) or only a slight (333) effect on bacterial survival and replication in cultured macrophages and are invasive for epithelial cells like the wild-type strain (196). Therefore, although the SPI-2-encoded type III secretion system in Salmonella spp. most significantly affects virulence, its cellular and molecular impact on pathogenesis remains elusive.

EPEC strains form one of several categories of diarrheagenic E. coli strains, which are distinguished from other E. coli strains by their ability to inflict characteristic lesions in small intestine enterocytes, with gross cytoskeletal damage and loss of brush border microvilli (423). In addition, EPEC strains are characterized by their clustered pattern of adherence to epithelial cells, which results in the formation of microcolonies on the cell surface (403), and by their inability to produce Shiga toxin. EPEC strains are a leading cause of diarrhea in infants in the developing world (101, 270).

After the initial adherence to epithelia, EPEC strains attach intimately to the epithelial cell surface, leading to the effacement of microvilli beneath the bacteria. The resulting characteristic histopathologic finding is known as attaching and effacing (A/E) lesions (316). At the zone of contact between bacteria and the epithelial cell surface, cup-like pseudopod structures appear, which form progressively elongating pedestals carrying individual bacteria on their tops (Fig. 2) (378). The pedestal surface conforms to the curvature of the bacterium and maintains a distance of less than 10 nm across much of the bacterial surface (316). Pedestals consist of a densely packed microfilamentous structure (297) and contain cytoskeletal proteins including actin (248), -actinin (123), and myosin light chain (285), in addition to talin and ezrin (123), proteins that link actin filaments to the membrane (65, 367).

Intimate attachment, effacing of microvilli, and formation of pedestals require a bacterial adhesin (called intimin) and EPEC type III secretion. Although intimin is not secreted by the type III secretion pathway, the encoding gene (eaeA) is located within the gene cluster that encodes the EPEC type III secretion system (100, 224, 225), and intimin functions synergistically with type III secretion in pedestal and A/E lesion formation. Intimin specifically binds to a 90-kDa protein (originally named Hp90) which is located in the eukaryotic membrane in direct proximity to the attached bacteria. It was recently discovered that this protein is of bacterial origin and is secreted by the type III secretion mechanism (239). The 549-aa protein, which was named Tir (translocated intimin receptor), is inserted into the eukaryotic membrane. Two other type III secreted proteins, EspA and EspB, which also may be translocated into the eukaryotic cytosol (104, 240, 242), are required for membrane insertion of Tir (239). Intimin was shown to directly bind Tir, demonstrating that EPEC strains transfer their own receptor for intimate attachment into eukaryotic cells.

Concomitant with pedestal formation, adherent EPEC strains induce tyrosine phosphorylation of several proteins in the eukaryotic cell, including Hp90/Tir (376, 378) and phospholipase C-1 (241). While Tir phosphorylation has been thought to facilitate intimin binding (378), activation of phospholipase C-1 induces inositol triphosphate and Ca²⁺ fluxes (30, 133), which might be involved in the cytoskeletal rearrangements involved in pedestal formation. EPEC-induced tyrosine phosphorylation and host cell signalling also depend on the type III secretion of EspA, EspB, and EspC (132, 223, 240, 242, 259, 376). (It should be noted, however, that the signal transduction events leading to A/E lesion formation may be more complicated than implied by this attractive model, since actin accumulation beneath adherent EPEC cells was observed even in the absence of detectable tyrosine phosphorylation due to inhibition of protein tyrosine kinases with genistein [364]. Furthermore, a mutant impaired in Esp secretion in vitro and in the induction of tyrosine phosphorylation of Hp90 was still capable of causing A/E lesions like the wild type [364].)

P. aeruginosa is an opportunistic pathogen capable of infecting immunocompromised individuals, including patients with extensive burns or wounds, cystic fibrosis, and leukemia (53). Recently, secretion of exoenzyme S (ExoS) and the related ExoT (80), two of several virulence determinants of P. aeruginosa (328), has been found to occur via a type III secretion pathway (135, 491). ExoS and ExoT are proteins with ADP-ribosyltransferase activity (221), whose preferred eukaryotic target proteins include members of the H-Ras and K-Ras family of GTP-binding proteins and vimentin, an intermediate filament protein (80). The enzymatic activity of ExoS is absolutely dependent on a eukaryotic cofactor protein, FAS (81), which is a member of a large family of regulatory proteins that play key roles in cell growth and differentiation (142). Although the function of ADP-ribosylation in P. aeruginosa pathogenesis is unclear, the production of ExoS is associated with epithelial cell damage and dissemination of P. aeruginosa within infected hosts (23, 231, 327). It is interesting that ExoS is partially similar to YopE from Yersinia spp. (see "The cytotoxin YopE" below) and that the type III secretion systems of the two pathogens are highly similar (see "Type III secretion system genes in Yersinia species" and "P. aeruginosa type III secretion system genes" below) (135). Indeed, ExoS can be secreted and translocated into mammalian cells via the Yersinia type III secretion pathway and induces host cell cytoskeletal damage similar to that induced by YopE (140). Several other proteins which are secreted by P. aeruginosa and which are highly similar to secreted Yersinia proteins that function in the regulation of secretion and translocation have been identified.

Type III secretion genes have been recently identified in Chlamydia psittaci and other chlamydiae (211). Although these genes have not been shown to be functional, type III secretion could play a major role in host cell interactions of these highly host-adapted bacteria (171). Chlamydiae are a worldwide major cause of sexually transmitted diseases and infectious blindness, and they cause respiratory tract infections and are associated with atherosclerosis; their basic pathogenicity mechanisms remain elusive (171). These obligate intracellular bacteria exhibit a biphasic developmental cycle with a dormant extracellular form and a replicating intracellular form (319). Intracellular bacteria reside within a large vacuole, which, soon after infection, redistributes from the cellular to the nuclear periphery by an F-actin-dependent mechanism (281). The chlamydial inclusion expands by acquiring host membrane components and includes chlamydial proteins. Among the latter, IncA is translocated to the outer surface of the Chlamydia-containing vacuole and may be phosphorylated by a host enzyme (371, 372). IncA does not contain a sec signal sequence, and it has been speculated that this protein may be translocated via the type III secretion pathway (211). Whether type III secretion is involved in tyrosine phosphorylation of host cell proteins observed early after infection (45) remains to be elucidated.

One of the exciting surprises that came with the discovery of type III secretion systems was the finding that a variety of gram-negative phytopathogenic bacteria use type III secretion for pathogenesis, just like the unrelated animal pathogens do. Type III secretion systems are conserved in the four major genera of plant-pathogenic bacteria, Erwinia, Pseudomonas, Ralstonia, and Xanthomonas (60, 85), and are required for the common ability of these pathogens to cause disease in susceptible host plants. The plant diseases caused by these pathogens range from fire blight of rosaceous plants and soft rot caused by various Erwinia spp., bacterial spot disease of pepper and tomato (Xanthomonas campestris), and bacterial speck (P. syringae), to bacterial wilt of solanaceous plants (Ralstonia solanacearum). Interestingly, genes encoding a putative type III secretion system have also been found in Rhizobium spp. (139, 189), where they are involved in cultivar-specific nodulation of leguminous plants (299).

Type III secretion is also essential for the induction of a defense reaction called the hypersensitive response (HR) in resistant plants that are not normally hosts for the particular pathogen. The HR is characterized by localized tissue necrosis and the production of phenolics and antimicrobial agents at the site of bacterial contact (245, 277), which prevent further spread of the infecting bacteria in the plant. The HR is an active response to bacterial infection and requires plant de novo gene expression and protein synthesis, a calcium flux across membranes, and ATPase activity (188). Although the HR is microscopically small under natural conditions, it is macroscopically visible and readily assessed in the laboratory when plant tissue is infiltrated with large numbers of phytopathogenic bacteria (Fig. 2). The bacterial genes required for pathogenicity in susceptible plants and for the elicitation of an HR in resistant plants have operationally been defined as hrp (for hypersensitive response and pathogenicity) (276), and this name is still maintained as a global label of the plant-pathogenic type III secretion systems. In some plant pathogens, proteinaceous elicitors of plant disease and/or HR have been identified and were shown to be secreted via the type III secretion pathway (26, 34, 188, 472). For example, the structurally similar type III secreted proteins from E. amylovora and P. syringae elicit an HR in nonhost plants when infiltrated into plant tissue in a purified form, and mutants with mutations in the encoding genes are avirulent on host plants (188, 472).

Whether bacterial infection leads to plant disease or to an HR resistance phenotype is determined largely by the presence of a matching pair of a dominant resistance gene (R) in the host and a so-called avirulence gene (avr) in the pathogen (424, 455). Disruption of an avr gene usually allows the pathogen to provoke disease in previously resistant plants, while introduction of an R gene into a susceptible plant confers resistance to the pathogen carrying the matching avr locus. The requirement for a matching R-avr gene pair for HR induction has led to the postulation of the gene-for-gene hypothesis (127). In the gene-for-gene concept, the plant R gene is thought to encode a receptor for the pathogen Avr protein, and interaction between these two proteins may trigger a signalling cascade which ultimately results in elicitation of an HR (29, 260, 424).

Interestingly, like the elicitors of plant disease, the action of several bacterial avirulence proteins is phenotypically dependent on functional type III secretion (162, 247, 348, 363). However, secretion of an avirulence protein has not been directly observed. Nevertheless, several avr-encoded proteins have been shown to elicit an HR in an R-gene-dependent manner when expressed inside host cells (162, 267, 446). It is thus likely that the Avr proteins are injected into the target cell via the type III secretion mechanism (29) like the virulence factors injected by Yersinia spp. and other animal pathogens.

R genes from several plant species and numerous bacterial avirulence genes have been cloned (29, 455). As a common structural motif, most R-gene-encoded proteins carry leucine-rich repeats which could mediate interactions with other proteins and therefore could function in signal recognition and transduction (336). In the case of the tomato R gene Pto, the encoded protein is a serine/threonine kinase. Pto confers resistance to bacterial speck disease caused by P. syringae pv. tomato expressing the corresponding AvrPto protein (287). (On the basis of host range, plant-pathogenic bacterial species are subdivided into pathovars [pv.]. Thus, P. syringae pv. tomato causes disease on susceptible tomato plants, while P. syringae pv. phaseolicola causes disease in beans.) Direct interaction between AvrPto and Pto has recently been demonstrated, providing a physical explanation for the gene-for-gene hypothesis and for the specificity of plant-pathogen interactions (408, 436). Pto interacts with and phosphorylates another serine/threonine kinase (Pti1), which also functions in HR induction (504). Thus, after AvrPto is translocated by the type III secretion mechanism into the plant cell cytosol, binding of AvrPto to Pto might stimulate Pto kinase activity to trigger a phosphorylation cascade involved in activating the HR.

This section describes structural and functional aspects of the virulence proteins which are secretedand in many cases injected (translocated) into the eukaryotic cytosolvia the type III secretion pathway. These proteins greatly vary in size, structure, and function and account for the species-specific pathogenicity phenotypes associated with type III secretion. Interestingly, several of the secreted proteins are similar to eukaryotic proteins, implying that the pathogens may have acquired the respective genes from their eukaryotic hosts during evolution. A number of secreted proteins do not exhibit direct antihost functions but, rather, are accessory in that they function in secretion and translocation of the actual virulence factors. (These accessory proteins are discussed in detail under "Translocation of proteins into the eukaryotic cytosol" and "Regulation of type III secretion by contact with eukaryotic cells" below). In several cases, the genes encoding secreted proteins are located outside the gene clusters which encode the type III secretion apparatus, but usually the former genes are transcriptionally coregulated with the latter ones. Several secreted proteins have homologs in different type III secretion systems. Thus, while the Salmonella invasion proteins are homologous to S. flexneri invasins, another secreted Salmonella protein contains two domains which each are similar to a different Yersinia antihost factor, respectively. Some secreted factors are even shared by Salmonella, Yersinia, and the plant pathogen X. campestris pv. vesicatoria. The secreted proteins are summarized in Table 1.

The designation of the secreted virulence factors of Yersinia spp. as Yersinia outer proteins (Yops) originates from the fact that these proteins were first detected in outer membrane preparations of Y. enterocolitica (355) and Y. pseudotuberculosis (428). However, it was later observed that the Yops are secreted into the supernatant of Y. enterocolitica and Y. pseudotuberculosis (190, 191) and that their appearance in the supernatant does not result from cell lysis or from membrane vesiculation (191). Yops exhibit low solubility due to their generally hydrophobic character and tend to form amazingly abundant macromolecular filamentous aggregates in the supernatant of an induced culture (308), a phenomenon which has also been observed for proteins secreted by S. flexneri (337) and S. typhimurium (216, 485). Therefore, the localization of Yops in the outer membrane is probably an artifact of copurification of insoluble Yop aggregates with outer membrane fractions (308).

Interestingly and in contrast to Y. enterocolitica and Y. pseudotuberculosis, the Yops expressed by Y. pestis (392, 428, 484) are rapidly degraded by the surface protease Pla (392, 417), which is encoded by a Y. pestis-specific 9.5-kb plasmid (416). This plasmid and specifically the Pla protease are required for the pathogenesis of plague (418). However, although most Y. pestis Yops are subjected to degradation by Pla, undegraded YopM and YopN are secreted into the bacterial growth medium (368, 427).

The Yopslike the Yersinia type III secretion systemare encoded on a 70-kb virulence plasmid present in all three pathogenic Yersinia spp. (see "Type III secretion system genes in Yersinia species" below). The yop genes are often organized in monocistronic operons, and their expression is induced after a temperature shift from 25 to 37°C, but yop expression at 37°C remains repressed when the bacterial culture medium contains millimolar concentrations of Ca²⁺ (355). The Yops encoded by the plasmids from Y. pestis and Y. pseudotuberculosis are almost identical in sizes and isoelectric points, whereas several of the Y. enterocolitica Yops exhibit some size differences (57, 90). Nevertheless, the Yops of all three Yersinia spp. are immunologically related (57, 58, 191, 294, 415).

A total of 13 Yops and other proteins secreted by the Yersinia type III secretion system have been characterized to date. They can be roughly grouped into (i) proteins with direct antihost functions, the majority of which are translocated into eukaryotic cells (YopE, YopH, YopM, YpkA); (ii) translocatory proteins involved in the translocation process (YopB, YopD, YopK, and perhaps YopR); and (iii) regulatory proteins, which mediate the cell contact-dependent induction of yop gene expression and Yop secretion (YopN, LcrG, LcrV, and LcrQ).

The cytotoxin YopE. Pathogenic Yersinia spp. exhibit a strong cytotoxic effect when adhering to the surface of eukaryotic cells, but cytotoxicity is not observed prior to adherence or when bacterial supernatant containing secreted Yops is added to epithelial cells (380, 381). Since cytotoxicity depends on the presence of YopE, Rosqvist et al. analyzed the localization of this protein before and after cell contact. They found that before cell contact, no YopE was secreted into the supernatant (due to a high concentration of Ca²⁺ in the tissue culture medium, which inhibits Yop expression and secretion in vitro [see "Negative control by Ca²⁺ via feedback regulation" below]), while after bacterial cell contact, YopE was expressed and all detectable YopE was present in the cytosol of the target cell (383). Similarly, it was demonstrated by the use of a YopE hybrid protein fused to the calmodulin-dependent adenylate cyclase CyaA from B. pertussis that YopE is translocated by adherent Y. enterocolitica into the cytosol of cultured epithelial cells (421) as well as into cultured murine macrophages (420). In this elegant approach, the fusion protein, which contained the amino-terminal 130 aa of YopE fused to the adenylate cyclase domain of CyaA (YopE₁₃₀-Cya), catalyzes the formation of cyclic AMP from ATP in a calmodulin-dependent manner (420). Since calmodulin is present in the cytosol of eukaryotic cells but absent from bacterial cells and from the tissue culture supernatant, the observed dramatic accumulation of cyclic AMP in this assay system is indicative of transport of YopE-CyaA out of the bacterial cell and translocation of the fusion protein into the eukaryotic cytosol (421). Interestingly, translocalized YopE protein is not evenly distributed inside the target cell but is enriched in the perinuclear region (345, 383).

View larger version (20K): [in this window] [in a new window]   FIG. 3.   Homologies between the protein tyrosine phosphatase SptP of S. typhimurium, the cytotoxin YopE and the tyrosine phosphatase YopH from Yersinia spp., and the P. aeruginosa ExoS ADP-ribosyltransferase. Identically shaded boxes indicate regions of sequence similarities. The catalytic cysteine residue in the carboxy-terminal parts of SptP and YopH are shown. The predicted catalytic domain of ExoS is located in the carboxy-terminal quarter of the protein (stippled) (253).

The protein tyrosine phosphatase YopH. A computer similarity search by Guan and Dixon (169) identified the 468-aa YopH (59, 305) to be homologous in its carboxy-terminal half to eukaryotic protein tyrosine phosphatases; accordingly, YopH was shown to be a specific (169) and by far the most active tyrosine phosphatase known (501). The catalytic domain is especially conserved, and exchange of the catalytic residue Cys₄₀₃ to Ala (C403A) completely abolishes phosphatase activity (169). YopH confers upon Y. pseudotuberculosis the ability to resist phagocytosis by cultured macrophages (20, 112, 379) in vitro, and the antiphagocytic effect is dependent on its protein tyrosine phosphatase (PTPase) activity (20). Furthermore, the YopH_C403A mutant is avirulent in mice (51), demonstrating the importance of YopH PTPase activity for Yersinia pathogenesis. Like yopE, yopH is transcribed as a monocistronic operon and is regulated at the transcriptional level by temperature and Ca²⁺ (59, 305). The yopH genes are highly conserved between Y. enterocolitica and Y. pseudotuberculosis. However, the sequence similarity stops abruptly 240 bp upstream and 175 bp downstream of the gene, showing that the homology between the virulence plasmids of the two species is confined to blocks of conserved DNA sequences (305). In addition to its homology to eukaryotic PTPases, YopH has a bacterial homolog in the carboxy-terminal half of the S. typhimurium type III secreted SptP, which also exhibits strong PTPase activity (235) (see below) (Fig. 3).

The protein kinase YpkA. Since Yersinia species secrete a protein phosphatase, Galyov et al. reasoned that there may also be a protein kinase in the supernatant of a Y. pseudotuberculosis culture. Indeed, the authors detected kinase activity and cloned the gene by sequence analysis of a small and (as of then) uncharacterized region of the Y. pseudotuberculosis virulence plasmid (149). YpkA is an autophosphorylating protein kinase with homology to eukaryotic protein kinases (149). The amino-terminal putative catalytic domain of YpkA contains structural motifs, called subdomains I to XI, which are common to all eukaryotic protein kinases, and the consensus sequence Asp-Ile-Lys-Pro-Gly-Asn in subdomain VI indicates that YpkA exhibits specificity for phosphorylation of serine and/or threonine (178). The 732-aa YpkA is transcribed as an operon together with the downstream YopJ protein. The operon is transcriptionally regulated analogously to the other Yops, but the level of expression is lower than that observed for YopH and YopE (150). Because YpkA is only weakly expressed and because the strong cytotoxic effects of YopE and YopH mask the effects of YpkA on HeLa cells, a yopE yopH mutant which overexpressed YpkA was used to demonstrate YpkA translocation (174). Interestingly, translocated YpkA was found to be associated with the inner surface of the plasma membrane of the target cell (174).

YopJ/P is required for apoptosis induction and is homologous to Salmonella and Xanthomonas proteins. The 264-aa YopJ is transcribed as an operon together with YpkA in Y. pseudotuberculosis (150). (YopJ is called YopP in Y. enterocolitica [313].) Although the protein has characteristics of other Yops (150), YopJ/P is not required for virulence in mice (150, 431). Recently, YopJ/P was shown to be responsible for the induction of apoptosis in cultured murine macrophages (313). Interestingly, YopJ/P shows similarity to the plant avirulence factor AvrRxv from X. campestris (479) and to the avrA-encoded protein from S. typhimurium (183).

The thrombin binding factor YopM. In addition to YopH and YpkA, YopM is the third secreted Yersinia virulence factor that is similar to eukaryotic proteins. The protein is homologous to the thrombin binding domain of the  chain of human platelet surface glycoprotein Ib (GPIb) and also to a portion of von Willebrand factor (269). GPIb is involved in cross-linking platelets by binding thrombin and von Willebrand factor, causing platelets to aggregate and initiate blood clotting at sites of blood vessel injuries. In addition, thrombin activates platelets, causing them to release a variety of inflammatory mediators (474). Purified YopM was shown to bind thrombin and to inhibit platelet aggregation in vitro (368). Although Y. pestis and Y. enterocolitica yopM mutants were strongly attenuated after i.v. infection of mice (269, 321), the significance of thrombin binding for Y. pestis pathogenesis is not clear. It is conceivable that YopM might compete with platelets for thrombin binding in vivo and that the resulting prevention of blood clot formation could contribute to the dissemination of the bacteria throughout the body. In addition, YopM could inhibit platelet activation in vivo and thereby might mute the local inflammatory response to the bacteria (368). More recently, it was demonstrated that YopM is also translocated into the cytosol of macrophages (56). To