Type III export: new uses for an old pathway

Authors


Abstract

Gram-negative bacteria use type III secretion (TTS) systems to translocate proteins into the extracellular environment or directly into eukaryotic cells. These complex secretory systems are assembled from over 20 different structural proteins, including 10 that have counterparts in the flagellar export pathway. Secretion substrates are directed to the TTS machinery via mRNA and/or amino acid secretion signals. TTS chaperones bind to select secretion substrates and assist in the export process. Recent progress in the understanding of TTS is reviewed.

Introduction

A number of Gram-negative animal and plant pathogenic bacteria use type III secretion (TTS) systems to target bacterial effector proteins into eukaryotic cells. The secretion process in many of these virulence-associated TTS systems is triggered when a pathogen comes in contact with a eukaryotic cell and, hence, has been called ‘contact-dependent secretion’. Once delivered into the host cell, effector proteins modulate host cell functions to the bacterium's advantage (Cornelis and Van Gijsegem, 2000). In addition to their role in pathogenesis, TTS systems are required for biogenesis of the bacterial flagella (Macnab, 1999) and are essential for the establishment of the symbiotic relationship between rhizobia and various legumes (Viprey et al., 1998). The similarity between the flagellar and virulence-associated TTS systems suggests that the virulence-associated systems have evolved from the more ancient flagellar systems (Macnab, 1999; Nguyen et al., 2000). Recent studies of the different TTS systems have provided new and exciting information regarding the mechanism of protein secretion across both the bacterial and eukaryotic membranes and have yielded the first view of the assembled TTS machinery.

This review will present a brief overview of the TTS pathway, focusing on the mechanism of secretion and the assembly/architecture of the secretion machinery. For information on translocation of proteins across the eukaryotic membrane and effector protein function, readers are referred to other recent reviews (Galán and Collmer, 1999; Cornelis and Van Gijsegem, 2000). As a result of restrictions on space and references, we have been forced to limit the amount of information covered under specific topics and have concentrated our coverage on several of the more thoroughly characterized TTS systems. Unfortunately, we were not able to include many fine studies from numerous laboratories that were relevant to the subject matter of this review: for this we apologize.

Members of the TTS system family

TTS systems were first identified in the human pathogenic yersiniae (Michiels et al., 1990), and have since been identified in a large number of other mammalian pathogens including Salmonella typhimurium, Shigella flexneri, Pseudomonas aeruginosa, enteropathogenic and enterohemorrhagic Escherichia coli, Bordetella pertussis and Chlamydia trachomatis (Hueck, 1998). Salmonella typhimurium and Yersinia enterocolitica each encode two separate virulence-associated TTS systems (Shea et al., 1996; Haller et al., 2000). Some plant pathogens, such as Erwinia amylovora, P. syringae, Ralstonia solanacearum and Xanthomonas campestris, require TTS systems to elicit the hypersensitive response in resistant plants and to be pathogenic on sensitive plants (Hueck, 1998). The flagellar export apparatus represents an additional member of the TTS family that has been hypothesized to be the evolutionary precursor of the virulence-associated TTS systems (Macnab, 1999; Nguyen et al., 2000). The Y. enterocolitica flagellar export system was recently shown to have the ability to export at least one virulence protein to the external environment, indicating that three distinct TTS systems are involved in the export of virulence factors by this organism (Young et al., 1999).

Genes encoding the secreted and structural components that make up each individual TTS system are generally found on a single plasmid or within a distinct pathogenicity island on the bacterial chromosome, suggesting that these genes were inherited as a single unit. Sequence similarities suggest that many of the structural components of the TTS apparatus are conserved among the different bacterial species that employ these systems; however, the proteins delivered by these pathogens are quite diverse (Galán and Collmer, 1999). The similarity among the structural components that make up the different TTS machineries suggests that a common mechanism of substrate recognition and export is shared by these systems.

Recognition of TTS substrates

TTS systems use multiple signals to target proteins for secretion and translocation across eukaryotic cell membranes. This process is best understood in the pathogenic yersiniae. These organisms secrete Yop effector proteins in vitro at 37°C in the absence of calcium, a condition that obviates the normal requirement for host cell contact (Michiels et al., 1990). Experiments carried out with YopE, YopH and YopN of Y. enterocolitica and Y. pseudotuberculosis revealed the existence of two pathways capable of targeting proteins to the TTS apparatus (Sory et al., 1995; Schesser et al., 1996; Woestyn et al., 1996; Cheng et al., 1997). The N-terminal 15- to 17-amino-acid residues of YopE, YopH and YopN were sufficient to direct the export of heterologous reporter proteins. However, no shared amino acid sequence was present in the targeting sequences, and frameshift mutations that completely altered the amino acid sequences within these regions failed to prevent secretion (Anderson and Schneewind, 1997). These results led to the conclusion that the secretion signals may reside in the 5′ ends of the yop mRNAs rather than in the polypeptide sequences, suggesting that Yops can be secreted in a co-translational manner. The 5′ ends of several yop mRNAs are predicted to form stem–loop structures that could conceal translational start codons and thus prevent Yop translation until the transcripts can interact with the secretion machinery. However, no direct evidence that yop mRNAs form stem–loop structures has been presented and no interaction between yop mRNAs and TTS components has been reported.

Several Yops have a second amino acid targeting domain that is dependent upon the binding of a specific TTS chaperone (Wattiau and Cornelis, 1993; Wattiau et al., 1996; Cheng et al., 1997). For YopE this chaperone is termed SycE (YerA in Y. pseudotuberculosis). SycE is a small homodimeric cytosolic protein that specifically binds within amino acid residues 15–50 of YopE and functions to solubilize and stabilize YopE in the bacterial cytosol (Frithz-Lindsten et al., 1995; Schesser et al., 1996; Woestyn et al., 1996; Cheng et al., 1997). Yersinia strains lacking SycE secrete and translocate reduced levels of YopE or YopE-hybrid proteins that carry the chaperone-binding domain. Removal of the SycE-binding domain from YopE alleviates the SycE requirement for secretion; however, these deletants are not translocated by wild-type Yersinia (Woestyn et al., 1996).Boyd et al. (2000) recently demonstrated that YopE deletants that lack the chaperone-binding domain, but carry the N-terminal signal can be translocated by a yop polymutant strain that no longer expresses five other effector Yops. These data suggest that the role of SycE in wild-type Yersinia could be to introduce a hierarchical order or preference in Yop delivery.

Lloyd et al. (2001) have recently challenged the role of mRNA signals in YopE secretion, suggesting that N-terminal amino acid residues, and not mRNA signals, are involved in the export of YopE. They show that frameshift mutations that completely alter the N-terminal 11 residues of YopE fail to prevent YopE secretion in wild-type Yersinia, but drastically reduce the N-terminal signal-dependent secretion of YopE in a sycE mutant. Furthermore, a series of mutations that extensively alter the mRNA sequence of this region without changing the amino acid sequence had no effect on YopE secretion. A comparison of the N-terminal sequences of the various secreted Yops led these researchers to suggest that an N-terminal amphipathic helical region may be involved in Yop export. Interestingly, a synthetic amphipathic sequence consisting of alternating serine and isoleucine residues was shown to function as an N-terminal TTS targeting signal in Yersinia. From these observations, Lloyd et al. (2001) conclude that YopE is targeted for secretion by N-terminal amino acids and the SycE chaperone.

Multiple secretion signals have also been proposed for the anti-sigma factor FlgM, a substrate of the flagellar TTS system that inhibits σ28-dependent expression of late flagellar genes until the completed hook–basal body complex becomes competent for its export (Hughes et al., 1993). Chilcott and Hughes (1998) demonstrated that FlgM expressed from constructs that lack the endogenous 5′ untranslated region can be secreted and that its efficient secretion is dependent upon specific N-terminal amino acids, indicating that amino acid signals can target FlgM for secretion. Additionally, the efficiency of FlgM secretion can apparently be influenced by its 5′ untranslated region. FlgM translated from flagellar class 2 flgAMN transcripts was primarily intracellular; however, FlgM translated from class 3 flgMN transcripts was primarily extracellular, suggesting that FlgM translation and secretion can be coupled (Karlinsey et al., 2000). Interestingly, the flagellar-specific chaperone FlgN (Fraser et al., 1999) is required for efficient translation of FlgM from flagellar class 3 transcripts but not from class 2 transcripts, indicating that TTS chaperones may participate in the coupling of translation and secretion. Further analysis of multiple TTS substrates from different TTS systems will be required to determine the functional significance and relative contribution of mRNA and amino acid secretion signals.

The high level of similarity/identity shared between many of the components from the various TTS systems suggests that these systems recognize similar secretion signals. In fact, it has been shown that secreted proteins from one TTS system can often be secreted from a heterologous TTS system (Rosqvist et al., 1995). For example, IpaB, a Shigella TTS substrate, is secreted via the Yersinia TTS system in the presence, but not in the absence, of its chaperone, IpgC. Likewise, YopE is secreted via the Salmonella pathogenicity island (SPI)-1 TTS system in a SycE-dependent manner. Similarly, P. syringae export substrates AvrB and AvrPto were secreted via the Yersinia TTS system and the Erwinia chrysanthemi TTS system cloned in E. coli (Anderson et al., 1999). These findings suggest that the various TTS systems use a relatively well-conserved delivery apparatus to dispense a wide variety of effector proteins by recognizing common secretion signals.

TTS chaperones

TTS chaperones are found in almost all bacteria that use flagellar and/or virulence-associated TTS systems; however, not all TTS substrates utilize chaperones. These chaperones share common features, including small size (12–18 kDa), an acidic pI and an overall α-helical character (Wattiau et al., 1996; Bennett and Hughes, 2000). Each chaperone binds non-covalently to a cognate substrate protein that is usually encoded just upstream or downstream of that chaperone's structural gene. In the absence of the chaperone, secretion of the cognate protein is prevented or dramatically reduced; however, the exact function of these chaperones in the secretory process remains elusive (Bennett and Hughes, 2000). TTS chaperones in virulence-associated systems are thought to function as homodimers that bind to an N-terminal region of their substrate (Cheng and Schneewind, 1999), whereas in the flagellar system the chaperone-binding site is located at the C-terminus (the last 40 amino acids) (Fraser et al., 1999).

The foregoing generalizations do not hold true for every TTS chaperone. Some chaperones seem to function as heterodimers. One such example is the YscB/SycN complex of Y. pestis, which has been shown to serve as the specific chaperone for YopN, the putative plug or cap of the TTS apparatus (Day and Plano, 1998). Several TTS chaperones also bind more than one secretion substrate. For example, SycH, the specific chaperone for YopH, is also responsible for binding and TTS-targeting of LcrQ/YscM1, a regulator of the Yersinia yop virulon (Cambronne et al., 2000).

Regulation of TTS

Secretion of proteins via the TTS pathway is not a constitutive process; instead, export is triggered by exogenous signals. Initiation of the Yop secretion process in Yersinia is triggered by contact between the bacterium and a eukaryotic cell. Secreted effector Yops are targeted directly into the eukaryotic cell and are not found in substantial amounts in the extracellular milieu, indicating that the delivery process, termed translocation, is polarized (Rosqvist et al., 1994). Salmonella typhimurium and S. flexneri also release effector proteins after host cell contact; however, in these cases, effector proteins are found both in the extracellular environment and within host cells, suggesting that secretion and translocation are not tightly coupled (Ménard et al., 1994; Galán and Collmer, 1999).

In Yersinia, Yop secretion is blocked in the presence of calcium ions prior to contact with a host cell. The block in Yop secretion is dependent upon the secreted YopN protein, TyeA and LcrG (Forsberg et al., 1991; Nilles et al., 1997; Iriarte et al., 1998; Cheng and Schneewind, 2000). Mutational inactivation of yopN, tyeA or lcrG results in uncontrolled secretion prior to host cell contact and in a loss of polarized translocation after host cell contact. TyeA and LcrG are small proteins (< 100 residues) that bind to a C-terminal domain of YopN and LcrV, respectively (Nilles et al., 1997; Iriarte et al., 1998). A small amount of the secreted YopN protein is surface exposed in the presence of calcium and prior to contact with host cells. Surface-exposed YopN is thought to function as a regulatory plug that prevents Yop secretion prior to host cell contact (Forsberg et al., 1991). On the other hand, LcrG has been proposed to block the secretion apparatus from the cytoplasmic face of the inner membrane. Sequestration of LcrG by cytosolic LcrV is believed to prevent its secretion-blocking activity, thus the ratio of LcrV to LcrG is thought to play a critical role in regulating Yop secretion (Nilles et al., 1998).

In S. flexneri, the secreted IpaB and IpaD proteins are required to prevent TTS prior to cell contact (Ménard et al., 1994). A small portion of the total cell-associated IpaB and IpaD are complexed with one another in the Shigella envelope fraction prior to their secretion; however, upon secretion IpaB preferentially interacts with IpaC. The membrane associated IpaB–IpaD complex has been proposed to function as a plug that prevents secretion prior to contact with the surface of an eukaryotic cell. At this time, the mechanism by which the aforementioned Yersinia and Shigella proteins block secretion and the physiological signals that release the block in secretion are unknown.

Expression of TTS components and substrates in the various virulence-associated TTS systems are regulated by a multiplicity of signals and signal-response systems. In general, virulence-associated TTS systems are expressed under conditions that the pathogen encounters during the infectious process. For example, in several plant pathogens, including R. solanacearum and E. amylovora, the expression of TTS components and substrates is controlled by a regulatory cascade that responds to environmental conditions and specific plant products (Wei et al., 2000). In fact, Aldon et al. (2000) have recently shown that transcriptional activation of TTS genes in R. solanacearum is controlled, in part, by an outer membrane protein, termed PrhA, that recognizes a non-diffusible plant cell wall product. Thus, expression of TTS components and substrates is upregulated upon contact between the bacterium and plant cells. In this case, transcriptional induction of TTS genes occurs even in the absence of a functional TTS apparatus, suggesting that transcription is not coupled to the TTS process in this organism.

In Yersinia, on the other hand, increased transcription of TTS genes is directly coupled to the cell contact-dependent opening of the Yop export apparatus through a secreted negative regulatory protein, designated LcrQ (YscM in Y. enterocolitica) (Pettersson et al., 1996). Upon contact with a eukaryotic cell, the TTS process is initiated and LcrQ is secreted, triggering increased transcription of TTS genes. The mechanism by which cytosolic LcrQ downregulates gene transcription has not been established; however, YopD and SycD must be present for transcriptional downregulation to occur (Williams and Straley, 1998). These proteins have not been demonstrated to bind DNA and no DNA-binding motifs are present in their deduced amino acid sequences; therefore, they probably exert their regulatory effects in an indirect manner, possibly through interactions with the yop virulon-specific transcriptional activator LcrF (Yother et al., 1986).

Secretion of flagellar TTS substrates is also regulated both at the level of expression (transcriptional and translational regulation) and directly at the level of secretion (Chilcott and Hughes, 2000). Flagellar TTS substrates required to complete the hook–basal body complex (middle genes) are expressed prior to expression of proteins required to complete the external flagellar filament (late genes). The secreted anti-sigma factor FlgM plays a key role in preventing late gene expression prior to completion of the hook–basal body complex (Hughes et al., 1993). Thus, the flagellar structural substrates are expressed in a temporal manner that ensures that the proper secretion substrates are available at specific developmental stages.

Secretion of the two main categories of flagellar TTS substrates (rod hook-type proteins and filament-type proteins) is also regulated by the ability of the export apparatus to change its substrate specificity (Williams et al., 1996). After completion of the hook–basal body complex, the flagellar TTS apparatus changes its substrate specificity from rod hook-type proteins to filament-type proteins. This process is dependent upon the secreted FliK protein and the TTS component FlhB (a YscU family member) (Minamino et al., 1999). fliK null mutants are non-motile, defective at filament assembly and express abnormally long hook structures (polyhook phenotype). Motile revertants of fliK mutants include extragenic suppressors that map to the flhB locus. The mechanism by which FliK and FlhB sense completion of the hook–basal body complex and change the substrate specificity is unknown.

TTS machinery

A comparison of the individual secretion apparatus components encoded within the various virulence-associated TTS system gene clusters has revealed as many as 11 conserved proteins, and 10 of these also share amino acid similarity/identity with elements of the flagellar export apparatus (Table 1). Members of the YscC, YscD, YscJ, YscL, YscN, YscQ, YscR, YscS, YscT, YscU and YscV families have counterparts in almost every TTS system (Hueck, 1998). Other secretion components, such as the Yersinia YscX and YscY proteins, are unique to only one or two closely related systems (Day and Plano, 2000).

Table 1. Broadly conserved and system-specific type III secretion componentsa.
Yersinia
pestis b
Salmonella
SPI-1b
Pseudomonas
syringae b
Flagellar
biosynthesisb

Cellular locationc
  • a

    . Proteins required to form a functional TTS system in Y. pestis, S. typhimurium (SPI-1), P. syringae pv syringae or tomato and in the flagellar TTS system of S. typhimurium.

  • b

    . Numbers in parentheses indicate the length of the protein in amino acids.

  • c

    . Locations are indicated by IM for inner membrane and OM for outer membrane.

  • d

    . May correspond to YscD/FliG protein family.

  • e

    . May correspond to YscF/PrgI protein family.

YscC (607)InvG (563)HrcC (700) OM secretin
YscD (419) HrpQ (310)FliG (331)IM
YscJ (224)PrgK (252)HrcJ (268)FliF (560)IM/OM
YscL (221)OrgB (227)HrpE (194)FliH (235)cytoplasmic
YscN (439)InvC (432)HrcN (454)FliI (456)cytoplasmic/ATPase
YscQ (307)SpaO (303)HrcQ (338)FliN (137)IM
YscR (217)SpaP (224)HrcR (208)FliP (245)IM
YscS (88)SpaQ (86)HrcS (88)FliQ (89)IM
YscT (261)SpaR (263)HrcT (264)FliR (264)IM
YscU (354)SpaS (356)HrcU (359)FlhB (383)IM
YscV(704)InvA (685)HrcV (697)FlhA (692)IM
YscW (131)InvH (147)  OM lipoprotein
YscK (209)OrgA (200)HrpD (175) cytoplasmic
YscO (154)InvI (147)HrpO (148) secreted
YscP (455)InvJ (336)HrpP (192) secreted
YscF (87)PrgI (80)  secreted; needle subunit
YscE (66)   cytoplasmic
YscG (115)   cytoplasmic
YscI (115)PrgJ (101)HrpB (124) secreted
YscX (122)   secreted
YscY (114)   cytoplasmic
HrpA (108)e secreted; Hrp pilin subunit
HrpF (74) cytoplasmic
HrpG (143) cytoplasmic
HrpT (67) OM
PrgH (392)d  IM
  FliO (125)IM
  FliM (334)IM

Of the 11 conserved TTS component families, only members of the YscL, YscN and YscQ families are predicted to be cytoplasmic or peripheral membrane proteins (Hueck, 1998). YscN family members are cytoplasmic ATPases that are highly conserved among the different TTS systems, contain consensus ATP-binding motifs (Walker boxes A and B) and are related to the catalytic subunits of bacterial FoF1 ATPases (Woestyn et al., 1994). The role of YscN family members in the export process is unknown; however, it has been suggested that they may play a role in protein targeting and/or energization of the export process, a function similar to that performed by SecA in the general secretory pathway. The flagellar YscN family member, FliI, has recently been shown to form a soluble complex with a homodimer of FliH, a YscL family member (Minamino and Macnab, 2000a). The interaction of FliH with FliI inhibited FliI ATPase activity, suggesting that FliH may regulate this activity. Other studies demonstrated that FliI also interacts with the flagellar secretion chaperone FliJ and with a number of secreted proteins including flagellin and hook protein (Silva-Herzog and Dreyfus, 1999; Minamino and Macnab, 2000b). The (FliH)2–FliI soluble complex could directly interact with secretion substrates and their chaperones and facilitate delivery to the secretion machinery. Interaction with the secretion machinery may remove FliH, thus stimulating ATPase activity and protein export.

Yeast two- and three-hybrid analyses have identified protein interactions between the Y. pestis YscN and YscL, YscL and YscQ, and YscQ and YscK proteins, suggesting that the YscN–YscL (FliI–FliH) interaction is conserved between virulence-associated and flagellar TTS systems (Jackson and Plano, 2000). The widely conserved YscQ family of proteins has a variable system-specific N-terminal domain and a highly conserved C-terminal domain that shows similarity to FliN, a component of the flagellar switch complex. It has previously been suggested that YscQ may represent a link between the widely conserved and system-specific components of the various TTS complexes (Hueck, 1998), a hypothesis that is supported by the finding that YscQ interacts with the relatively conserved YscL protein and the more system-specific YscK protein.

The YscD, YscJ, YscR, YscS, YscT, YscU and YscV family members have been shown, or are strongly predicted, to be integral inner membrane proteins (Hueck, 1998). YscJ family members carry sec-dependent signal sequences with lipoprotein attachment motifs (Leu–Xaa–Gly–Cys), suggesting that these proteins are lipoproteins. These lipoproteins are predicted to have a single C-terminal transmembrane domain that is immediately followed by two to three basic residues, which may function as a stop transfer signal (Hueck, 1998). The YscJ counterpart in P. syringae was localized both to the inner and outer membranes (Deng and Huang, 1999), suggesting that these proteins may function as a bridge across the periplasmic space. Membrane topology models of YscR, YscU and YscV predict these proteins to possess large hydrophilic cytoplasmic domains that could interact with more peripherally associated secretion components, namely members of the YscL and YscN families.

YscC family members are outer membrane proteins that multimerize to form ring-shaped structures with an external diameter of about 200 Å and apparent central pores of about 50 Å (Koster et al., 1997; Crago and Koronakis, 1998). These structures, which require an outer membrane lipoprotein (YscW family member) for proper insertion, are hypothesized to form channels for secretion across the bacterial outer membrane (Nouwen et al., 1999). Members of the YscC family represent a subset of the secretin superfamily of outer membrane channel-forming proteins, whose members are responsible for the transport of a diverse subset of large macromolecules across the bacterial outer membrane (Genin and Boucher, 1994). The flagellar export apparatus lacks this ring-like structure and homologues of the proteins required for its formation; instead, the flagellar rod and hook structures are thought to form the conduit for export across the outer membrane (Macnab, 1999). Penetration of the outer membrane by the flagellar rod is facilitated by the L-ring lipoprotein, FlgH, which some consider the flagellar equivalent of the virulence-associated YscC family (Minamino and Macnab, 1999).

Proteins exported via the TTS pathway appear to be exported via flagellar hook-filament, pilin-like, or needle-like structures. Structural components of the bacterial flagellum are exported through the flagellar rod-hook-filament to assemble at its distal end (Macnab, 1999). Similarly, several components of the virulence-associated TTS and translocation machineries are also export substrates. YscO and YscP are members of two poorly conserved families of secretion components that are exported to an extracellular location (Payne and Straley, 1998; Stainier et al., 2000). YscX, which is essential for Yop secretion in the yersiniae, was recently demonstrated to interact with cytosolic YscY and to be exported in a TTS-dependent manner (Day and Plano, 2000). Likewise, InvJ (YscP family), PrgI (YscF family) and PrgJ (YscI family) are exported via the S. typhimurium SPI-1 system (Collazo et al., 1995; Kimbrough and Miller, 2000; Kubori et al., 2000). The exported components of the secretion machinery are thought to be required for the assembly of external TTS structures.

Molecular architecture of the assembled TTS machinery

The bacterial flagellum consists of a basal body, a hook and a filament. The recently visualized needle complex of S. typhimurium bears a striking resemblance to the flagellar hook–basal body structure (Fig. 1); however, two of the three predominant constituents of this complex (PrgH and InvG) have no counterparts in the flagellar system, and the third, PrgK (YscJ family), shows only limited amino acid similarity/identity to FliF (Kimbrough and Miller, 2000; Kubori et al., 2000). The majority of the structural components that are conserved between these two TTS systems have been localized to the innermost part of the flagellar basal body, the MS-ring and the C-ring (Macnab, 1999). Analysis of purified S. typhimurium needle complexes (Kubori et al., 1998) revealed a membrane-bound base structure composed of two pairs of rings that are joined by a central rod. The rings, which are ≈ 40 nm in diameter, are anchored to the bacterial inner and outer membranes. A hollow straight needle structure 80 nm in length extends outward from the base. A supramolecular structure similar in size and appearance to that of the Salmonella needle complex has also been isolated from S. flexneri (Tomano et al., 2000). Visualization of the S. flexneri TTS complex in bacterial membrane ghosts revealed an additional structure, a large cytoplasmic bulb, that was absent from the isolated S. flexneri and S. typhimurium needle complexes (Blocker et al., 1999). This structure may represent the YscN family-containing complex that is associated loosely or that interacts dynamically with the more stable needle complex base.

Figure 1.

Hypothetical model of the S. typhimurium SPI-1 (left) and flagellar (right) TTS systems. Conserved secretion components are labelled with the S. typhimurium designations and according to the Yersinia nomenclature (parenthesis). A flagellar FliI (ATPase) (FliH)2 complex may deliver substrates and energize export via an integral inner membrane secretory complex contained in a patch of membrane within the MS-ring structure (Macnab, 1999). Secreted substrates are delivered into a central channel within the flagellar basal body-hook-filament structure where they eventually assemble at the distal end of this structure. A similar targeting pathway may function to deliver effector proteins to the SPI-1 TTS apparatus (Kubori et al., 1998). Effector proteins exported across the inner membrane may enter a channel within the needle complex base and eventually be targeted to the extracellular environment or into host cells with the help of the SipB, SipC and SipD translocators.

The major constituents of the S. typhimurium needle complex base (InvG, PrgH and PrgK) carry sec-dependent secretion signals, and assembly of this substructure therefore does not require a functional TTS apparatus (Kimbrough and Miller, 2000; Kubori et al., 2000). On the other hand, assembly of the external needle requires type III export, and the major constituent of this structure (PrgI) is a secreted protein (Kimbrough and Miller, 2000; Kubori et al., 2000). The length of the external needle is negatively regulated by the secreted InvJ protein in S. typhimurium, such that an invJ null mutant expresses abnormally long needle segments (Kubori et al., 2000). This function is similar to that performed by the secreted FliK protein, which negatively regulates flagellar hook length (Minamino et al., 1999). Salmonella typhimurium mutants defective in expression of PrgI or PrgJ assemble a needle complex base that lacks the external needle, suggesting that both of these proteins are involved in needle assembly. PrgI exhibits sequence similarity to a family of proteins (YscF family) found only in animal pathogenic bacteria (Kubori et al., 2000). Similarly, S. flexneri mxiH mutants (MxiH is also a YscF family member) produce a defective base structure that lacks a needle (Tamano et al., 2000). Overexpression of MxiH in the mxiH mutant results in production of abnormally long needles that resemble those of the S. typhimuriuminvJ mutant. Surface appendages, Hrp pili, that are indispensable for the TTS process in plant pathogenic bacteria are assembled from an unrelated protein termed HrpA in P. syringae (Roine et al., 1997) and HrpY in R. solanacearum (Van Gijsegem et al., 2000). Mutational inactivation of prgI, mxiH, hrpA or hrpY prevents secretion of additional TTS substrates from each gene's respective TTS system, indicating that secretion of needle and/or pilin-like subunits and assembly of external structures must be completed before other proteins can be secreted. Furthermore, these results imply that some form of substrate specificity switching, similar to that seen in the flagellar system, also occurs in the virulence-associated TTS systems (Kimbrough and Miller, 2000; Kubori et al., 2000; Tamano et al., 2000). Thus, in each TTS system a delivery apparatus is assembled from a combination of conserved and unique secretion components that function together to deliver specific proteins to the extracellular environment and beyond.

Concluding remarks

TTS systems have the remarkable ability to transfer proteins across three biological membranes, a process that requires a complex, dedicated delivery apparatus. The basic secretion machinery required for this process appears to be derived from the more ancient flagellar export pathway. Much progress has been made on determining the signals that target proteins to the secretion machinery; however, there is still much to be discovered concerning the mechanism of secretion, the process of protein translocation across the eukaryotic membrane, and the structure/assembly of the TTS machinery. In addition, little information is currently available on the mechanism of sensing cell contact, the sensors themselves, or the molecular events that trigger secretion.

The development of inverted membrane vesicle systems or eventually the reconstitution of TTS in proteoliposome systems will allow a closer look at the molecular events involved in the secretion process and will allow researchers to examine the energetics of the TTS process. Identification, purification and visualization of additional assembled TTS complexes will permit direct structural and functional comparisons among these macromolecular structures, which will shed light on the structure and function of both the unique and conserved secretion components. Further understanding of the molecular events involved in the TTS process will allow microbiologists and cell biologists to exploit these systems as tools to study cell biology, as vaccine delivery systems, and as targets for disease prevention and therapeutic intervention.

Acknowledgements

We are very grateful to the anonymous referees for helping us to improve this manuscript and to H. Wolf-Watz for providing a forthcoming manuscript. Work in the authors' laboratory was supported by Public Health Service Grant AI39575.

Ancillary