Functions of the Salmonella pathogenicity island 2 (SPI-2) type III secretion system effectors

Effectors that affect the vacuole membrane (SifA, PipB2, SseJ, SopD2)

Salmonella-induced filaments (Sifs) are lysosomal glycoprotein-containing membrane tubules induced by Salmonella. They are most easily apparent inside epithelial cells, where they extend centrifugally from SCVs along microtubules (Garcia-del Portillo et al., 1993). The gene responsible for the formation of these structures was discovered using a genetic screen in 1996 and named sifA (Salmonella-induced filament gene A) (Stein et al., 1996). SifA was originally implicated as a possible SPI-2 T3SS effector by virtue of its regulation by SsrA/B, functional link to SPI-2 (Beuzón et al., 2000), and N-terminal amino acid sequence similarity to other SPI-2 effectors (Beuzón et al., 2000; Miao & Miller, 2000). In addition to a complete lack of Sif formation, the sifA mutant has another very interesting phenotype: several hours after uptake by host cells, the vacuolar membrane surrounding these mutant bacteria is gradually lost, revealing that SifA has an essential function in maintaining the integrity of the SCV (Beuzón et al., 2000). As a result, sifA mutant bacteria are released into the cytosol of host cells. In macrophages, cytosolic bacteria are rapidly killed (Beuzón et al., 2002), which explains the strong attenuation of virulence of a sifA mutant in mice. However SPI-2 T3SS null mutants, which also fail to translocate SifA, retain an intact vacuolar membrane, implying that other SPI-2 protein(s) are involved in modifying the SCV membrane (Beuzón et al., 2000). At least three different SPI-2 effectors are now known to be involved in this activity (Ruiz-Albert et al., 2002; Schroeder et al., 2010), and are discussed below.

Following its translocation, SifA is anchored in the SCV membrane and to Sifs through prenylation at a C-terminal hexapeptide motif (Boucrot et al., 2003) by a host cell geranylgeranyltransferase (Reinicke et al., 2005). A major breakthrough in understanding the function of SifA was the discovery that it interacts with a host protein of 1020 aa called SKIP (SifA and kinesin-interacting protein) (Boucrot et al., 2005). The N-terminal region of SifA binds to a C-terminal pleckstrin homology (PH) domain of SKIP (Diacovich et al., 2009; Jackson et al., 2008; Ohlson et al., 2008). The 310 aa N-terminal region of SKIP was shown to interact with the tetratricopeptide repeat (TPR) domain of kinesin light chain (Diacovich et al., 2009; Dumont et al., 2010), a subunit of the kinesin motor that promotes anterograde transport on microtubules. Subsequent work showed that two WD motifs within the N-terminal region of SKIP are involved in binding to kinesin-1 (Rosa-Ferreira & Munro, 2011). The SifA–SKIP interaction seems to promote fission of vesicles from the SCV and anterograde transport of kinesin-1-enriched vesicles (Dumont et al., 2010). This and other work indicates that SKIP is a linker protein that binds and possibly activates kinesin-1. Recent work has shown that SKIP is also an effector of Arl8, a small GTPase present on lysosome membranes (Rosa-Ferreira & Munro, 2011). SKIP therefore provides a link between lysosomal membranes and kinesin-1, and is essential for peripheral distribution of lysosomes in the cell.

SifA also affects two GTPases: it binds GDP-bound RhoA (Ohlson et al., 2008) and competes effectively with the endosomal GTPase Rab9 for binding to SKIP, thereby acting as a G-protein antagonist (Jackson et al., 2008). SifA contains a potential cleavage site for caspase-3 between residues 127 and 141 (Srikanth et al., 2010). This could divide the protein into separate functional domains following its translocation.

Increased recruitment of kinesin to the SCV membrane occurs in the absence of SifA or SKIP (Boucrot et al., 2005). This is mediated by another SPI-2 T3SS effector, PipB2 (Henry et al., 2006). PipB2 was discovered on the basis of sequence similarity to PipB. Both proteins contain several tandem pentapeptide repeats of unknown function in their C-terminal regions (Knodler et al., 2003). PipB is an SPI-2 effector encoded by SPI-5, a pathogenicity island that contains another gene whose product (SopB) is translocated by the invasion-associated SPI-1 T3SS (Knodler et al., 2002; Wood et al., 1998). PipB also localizes to SCV membranes and Sifs (Knodler et al., 2003), but its function is currently unknown. It does not appear to be involved in the development of systemic disease in mice (Knodler et al., 2003), but does contribute to the inflammatory response in the bovine ileum (Wood et al., 1998). On the other hand, PipB2 interacts with kinesin light chain (Henry et al., 2006), which is consistent with the observations that this effector is found not only on the SCV membrane and Sifs, but also on vesicles at the periphery of infected cells (Knodler et al., 2003), and that overexpression of PipB2 results in the redistribution of late endosomal compartments to the cell periphery (Knodler & Steele-Mortimer, 2005). A functional link between PipB2 and SifA was established by the discovery that PipB2 is necessary and sufficient for recruitment of kinesin-1 to vacuoles containing sifA mutant bacteria (Henry et al., 2006). Inhibition of kinesin activity prevented vacuolar membrane rupture around sifA mutant bacteria (Guignot et al., 2004), but surprisingly the lack of PipB2 (and hence a failure to recruit kinesin to the SCV) did not rescue vacuolar membrane loss in a pipB2/sifA double mutant strain (Henry et al., 2006). Hence the mechanism underlying vacuole membrane rupture is still unclear. Possibly some residual kinesin action on the SCV is responsible for the loss of the vacuolar membrane in the double mutant. Nevertheless, it is clear that PipB2 is involved in kinesin-driven SCV membrane dynamics and the radial extension of Sifs along microtubules (Knodler & Steele-Mortimer, 2005), and that it contributes to the systemic phase of growth of S. Typhimurium in the mouse (Knodler et al., 2003).

SseJ is another effector that influences vacuolar membrane dynamics. It contains an N-terminal sequence conserved in several SPI-2 T3SS effectors, and its SPI-2 T3SS-dependent translocation into host cells was demonstrated using a CyaA fusion reporter (Miao & Miller, 2000). SseJ is absent from the genome of S. Typhi (Parkhill et al., 2001), but it contributes to the virulence of S. Typhimurium in mice (Ruiz-Albert et al., 2002). Intracellular vacuoles containing sseJ mutant bacteria appear to be morphologically normal, but when sseJ and sifA mutations were combined, the loss of vacuolar membrane around the double mutant strain was markedly reduced, indicating that loss of vacuolar membrane around sifA mutant bacteria requires the destabilizing action of SseJ (Ohlson et al., 2005; Ruiz-Albert et al., 2002; Schroeder et al., 2010). Moreover, deletion of sseJ leads to an increased number of Sifs, suggesting that SseJ inhibits Sif formation (Birmingham et al., 2005). The influence of SseJ on the vacuolar membrane and Sif formation indicates that SseJ attenuates or opposes the activity of SifA.

SseJ is a glycerophospholipid : cholesterol acyltransferase (GCAT) that esterifies cholesterol in infected cells (Lossi et al., 2008; Nawabi et al., 2008). Interestingly, this activity is strongly potentiated by a eukaryotic protein (Lossi et al., 2008), which was subsequently identified as RhoA (Christen et al., 2009). In S. Typhimurium-infected cells, a large proportion of total cellular cholesterol is recruited to the SCV (Catron et al., 2002). The phospholipase A activity of SseJ (Lossi et al., 2008) is likely to act on SCV membrane phospholipids, with the GCAT transferring acyl chains to cholesterol. Host cells infected with wild-type bacteria contain significantly more lipid droplets than cells infected with an sseJ mutant (Nawabi et al., 2008). Since lipid droplets are enriched in esterified cholesterol, this suggests that the action of SseJ causes the removal of esterified cholesterol from SCVs to lipid droplets. This could then influence the anchoring of prenylated proteins such as SifA and Rab proteins to the SCV membrane (Chen et al., 2008), which in turn would have an effect on the recruitment of microtubule motors to SCVs. Furthermore, the rigidity of membranes is greatly affected by their cholesterol content (Petrache et al., 2005). By depleting the SCV of cholesterol, SseJ would increase its rigidity, which could also help explain the role of SseJ in rupture of the vacuole in a sifA mutant. The effect of SseJ could also be mediated by its deacylation activity on membrane phospholipids, as the curvature of membranes reflects the composition of their phospholipid bilayer (McMahon & Gallop, 2005). Finally, changes at the level of host cell signalling pathways could be initiated by an effect on lipid rafts or by generation of lysophospholipid secondary messengers. The discovery that SifA binds GDP-RhoA, favouring GTP exchange, and that GTP-RhoA interacts with SseJ (Ohlson et al., 2008), provides a potential biochemical link between these two effectors, and could point to the existence of a multi-protein signalling complex comprising SifA, SKIP, RhoA, SseJ and possibly other effectors such as PipB2 and SopD2, which could regulate the stability, expansion and tubulation of the vacuolar membrane.

SopD2 was identified through its conserved N-terminal secretion sequence (Brumell et al., 2003). SopD2 localizes to the SCV membrane, Sifs, late endosomes and lysosomes. Deletion of sopD2 interferes with Sif formation at an intermediate stage and has therefore been suggested to be necessary for fusion of late endosomes, leading to Sif formation (Jiang et al., 2004). A sopD2 mutant is attenuated for virulence in mice (Jiang et al., 2004) and, as seen with an sseJ mutation, deletion of sopD2 reduces the loss of membrane around sifA mutant bacteria, which implies that SopD2 has an important function in vacuolar membrane dynamics (Schroeder et al., 2010). Interestingly, like sseJ, sopD2 is a pseudogene in S. Typhi (Parkhill et al., 2001), which suggests that S. Typhimurium and S. Typhi use different mechanisms to regulate membrane dynamics on their respective vacuoles.

The phenotypes and biochemical activities of SifA, PipB2, SseJ and SopD2 reveal that their apparently opposing actions actually work together to control vacuolar membrane dynamics. This maintains the integrity of the vacuolar membrane and presumably allows its controlled expansion, which must accompany bacterial cell division. It is also possible that by regulating vesicular fusion on the SCV these proteins ensure delivery of nutrients to the SCV lumen, thereby facilitating bacterial replication.

Effectors that affect SCV positioning (SseF, SseG)

In epithelial cells, maturation of SCVs is accompanied by their displacement towards the juxtanuclear region. This centripetal movement occurs along microtubules in the direction of the microtubule-organizing centre (MTOC), where Golgi stacks accumulate (Harrison et al., 2004; Ramsden et al., 2007). SCV movement is dependent on the recruitment of dynein to the SCV via a Rab7–RILP interaction (Guignot et al., 2004; Harrison et al., 2004). Close apposition between SCVs and the Golgi/MTOC seems to be important for bacterial replication (Ramsden et al., 2007; Salcedo & Holden, 2003), possibly by enabling interactions between SCVs and Golgi- or endosomal-associated vesicular traffic (Mota et al., 2009; Salcedo & Holden, 2003). Bacterial cell division leads to the formation of microcolonies that are retained in the Golgi/MTOC region (Ramsden et al., 2007) through the action of two SPI-2 T3SS effectors: SseF and SseG. Strains lacking either of these effectors undergo initial juxtanuclear migration and then scatter throughout the cell (Deiwick et al., 2006; Ramsden et al., 2007; Salcedo & Holden, 2003). Centrifugal scattering of sifA mutants also occurs, and this can be explained by the accumulation of kinesin on their vacuolar membranes prior to rupture (Boucrot et al., 2005).

SseF and SseG are encoded by genes within SPI-2 (Hensel et al., 1998) and were among the first proteins suggested to be effectors of the SPI-2 T3SS (Hensel et al., 1998). The use of epitope-tagged versions of SseF and SseG confirmed their translocation into macrophages via the SPI-2 T3SS (Hansen-Wester et al., 2002). sseF and sseG single mutant strains have similar levels of growth defect in macrophages and virulence attenuation in mice (Hensel et al., 1998; Kuhle & Hensel, 2002), but an sseF/sseG double mutant is not more attenuated than either single mutant (Deiwick et al., 2006). The similar phenotypes of single mutant strains and the lack of an additive effect on virulence attenuation in the double mutant strain support the hypothesis that SseF and SseG contribute to the same function. SseF and SseG are unusual effectors in that they are integral membrane proteins (Kuhle & Hensel, 2002) and interact (Deiwick et al., 2006). They have also been shown to be involved in Sif formation (Kuhle & Hensel, 2002), and to co-localize with the microtubule network, leading to the creation of microtubule bundles (Kuhle et al., 2004).

There are currently two models to explain how SseF and SseG could control the positioning of SCVs inside cells. The first is based on opposing activities of the microtubule motors kinesin and dynein. In this scenario, recruitment of kinesin (driving centrifugal movement) is modulated by SifA and PipB2, while that of dynein (driving centripetal movement) is controlled by SseF and SseG. The balance of these activities is normally in favour of SseF and SseG, which results in juxtanuclear positioning of SCVs. In support of this model, kinesin accumulates around vacuoles enclosing sifA mutant bacteria prior to their rupture (Boucrot et al., 2005), and dynein recruitment to SCVs is reduced in sseF and sseG mutants (Abrahams et al., 2006). Another possibility is a Golgi-tethering model, in which the complex formed by SseF and SseG, anchored in SCV membranes, interacts stably with Golgi-associated molecules to retain SCVs in that region of the cell. This model is based on live imaging observations (Ramsden et al., 2007) and the fact that treatment of cells with Brefeldin A (which disassembles the Golgi complex) mimics the sseF and sseG positioning phenotype (Ramsden et al., 2007). However, to date, no Golgi-associated protein has been shown to interact with SseF or SseG.

The positioning of Salmonella microcolonies at the MTOC/Golgi region could have an important physiological significance in enabling the acquisition of nutrients and membrane. Indeed a SPI-2-dependent redirection of post-Golgi traffic towards the SCV has been observed (Kuhle et al., 2006), and the secretory carrier membrane protein (SCAMP)-3, which mainly localizes to the trans-Golgi network (TGN) in uninfected cells (Castle & Castle, 2005), is redistributed to the SPI-2 effector-induced tubular network emanating from SCVs (Mota et al., 2009). This indicates that Sifs contain both endosomal and secretory pathway-derived molecules, and that SCVs interact with post-Golgi trafficking pathways (Mota et al., 2009).

Effectors involved in ubiquitin modification (SspH1, SspH2, SlrP, SseL)

The post-translational reversible modification of eukaryotic proteins by covalent linkage of ubiquitin to lysine residues or to itself (to form poly-ubiquitin chains) regulates many important cellular processes (Fang & Weissman, 2004; Pickart & Eddins, 2004). Ubiquitination occurs via the sequential action of three enzymes: a ubiquitin-activating enzyme (E1), a conjugating enzyme (E2) and a ligase (E3). While there are only two E1 enzymes and E2 enzymes are relatively few, many E3 ligases exist which provide substrate specificity to the process. Ubiquitination can be reversed by the action of deubiquitinases. Ubiquitin is not found in proteobacteria, but several important pathogens, including S. enterica, have evolved both E3 ligase and deubiquitinating activity (Steele-Mortimer, 2011). These effectors are translocated via the SPI-1 and SPI-2 T3SSs and interfere with ubiquitin signalling in host cells.

Three S. enterica proteins have been discovered with E3 ligase activity: SspH1, SspH2 and SlrP. SspH1 and SspH2 share 69 % identity at the amino acid level (Miao et al., 1999), and both contain leucine-rich repeats (LRRs), which are domains implicated in protein–protein binding (Kobe & Kajava, 2001). SlrP also contains LRRs, and was discovered through a signature-tagged mutagenesis screen for S. Typhimurium genes involved in virulence in mice and calves (Tsolis et al., 1999). SlrP has been linked to the SPI-2 T3SS on the basis of similarity to SspH1 and SspH2, and is required for growth within murine but not bovine hosts (Tsolis et al., 1999). The three proteins have been shown to be effectors of the SPI-2 T3SS through the use of CyaA fusion reporters (Miao et al., 1999; Miao & Miller, 2000). However, SspH1 and SlrP can be translocated by both the SPI-2 and the SPI-1 T3SS. Similar LRRs are also present in some other T3SS effectors, including Shigella IpaH and Yersinia YopM (Tsolis et al., 1999).

An important breakthrough with respect to these proteins was the discovery that IpaH and SspH1 have ubiquitin ligase activity (Rohde et al., 2007). IpaH and SspH1 have no primary amino acid sequence similarity to eukaryotic counterparts, and the crystal structures of IpaH (Singer et al., 2008; Zhu et al., 2008) and SspH2 (Quezada et al., 2009) reveal that they constitute a novel class of E3 ligase, containing a unique fold in the C-terminal region enclosing the catalytic site. Another unusual feature of these ligases is an autoinhibitory mechanism mediated by sequestering of the E3 ligase fold by the LRR domain. In the case of SspH2, the LRR motif also mediates localization of the protein to the apical side of the host cell plasma membrane (Quezada et al., 2009). Taking into account their different localization in the cell (nuclear for SspH1, cytoplasmic for SlrP, at the cell periphery for SspH2) and temporal regulation (SPI-1 or SPI-2 T3SS delivery), these effectors seem likely to affect host cell pathways in different ways. Nonetheless, each appears to have an impact on bacterial virulence. An slrP mutant is attenuated in mouse virulence (Tsolis et al., 1999) and an sspH1, sspH2 double mutant results in less mortality in calves (Miao et al., 1999).

Two lines of evidence indicate that SspH1 interferes with host cell inflammatory responses. Infection of epithelial cells with an sspH1 mutant strain results in increased secretion of the proinflammatory cytokine IL-8, compared with the wild-type strain (Haraga & Miller, 2003). Secondly, transfection-based expression of SspH1 in host cells causes an inhibition of NF-κB pathway-dependent gene expression (Haraga & Miller, 2003). A subsequent yeast two-hybrid screen identified the human serine/threonine protein kinase 1 (PKN1) as an interacting target for SspH1, and this interaction was confirmed by co-immunoprecipitation of endogenous PKN1 in infected HeLa cells (Haraga & Miller, 2006). The interaction is mediated by the LRR domain of SspH1 and is necessary for inhibition of NF-κB activation (Haraga & Miller, 2006). Interestingly, PKN1 has been reported to be involved in the TNFα-induced NF-κB signalling pathway (Gotoh et al., 2004). SspH1 has E3 ubiquitin ligase activity towards PKN1 (Rohde et al., 2007), but the relationship between ubiquitination of PKN1 and its possible effect on NF-κB activation remains to be established.

Unlike SspH1, the E3 ligases SspH2 and SlrP do not appear to influence activation of the NF-κB pathway (Haraga & Miller, 2003). SspH2 was instead found to localize to the host cell actin cytoskeleton and to function in actin polymerization (Miao et al., 2003). In vitro, SspH2 induces polymerization of K48-linked ubiquitin chains, which suggests that it can target proteins for proteasomal destruction (Levin et al., 2010).

SlrP has been reported to interact with the small redox protein thioredoxin (Bernal-Bayard & Ramos-Morales, 2009) and with ERdj3 (Bernal-Bayard et al., 2010), a chaperone of the Hsp40/DnaJ family. These two potential targets were identified by yeast two-hybrid screening and co-immunoprecipitation in host cells overexpressing SlrP. SlrP has been shown to ubiquitinate thioredoxin in vitro (Bernal-Bayard & Ramos-Morales, 2009). Thioredoxin is involved in functions such as cell proliferation, survival and apoptosis. In HeLa cells infected with S. Typhimurium, thioredoxin activity decreased, and this appeared to correlate with increased cell death (Bernal-Bayard & Ramos-Morales, 2009). However slrP mutants were not used to determine whether these phenotypes were dependent on SlrP or its E3 ligase activity. ERdj3 appears to be located in the endoplasmic reticulum (ER) (Bernal-Bayard et al., 2010). It has been suggested that SlrP, despite its largely cytoplasmic distribution, might also localize within the ER, where it could interact with the ERdj3. Although this interaction appears to be independent of the ubiquitination activity of SlrP, SlrP is able to compete for binding with ERdj3 substrates in vitro (Bernal-Bayard et al., 2010). However, neither ERdj3 nor thioredoxin has been shown to interact with, or be modified by, SlrP in host cells infected with bacteria. Therefore, these findings should be interpreted with caution, given the frequency with which yeast two-hybrid screens generate false positives and the fact that these conclusions were based on experiments involving overexpression of proteins following transfection and in vitro assays.

In addition to three ubiquitin E3 ligases, the SPI-2 T3SS translocates a protein with deubiquitinase activity. SseL was discovered as an SPI-2 T3SS effector on the basis of screens for genes regulated by SsrA/B (Coombes et al., 2007; Rytkönen et al., 2007). Purified SseL cleaved ubiquitin substrates in vitro with a clear preference for K63-linked chains over K48 linkages, which suggests that it is not involved in protecting proteins from proteasomal degradation (Rytkönen et al., 2007). Infection of epithelial cells and macrophages with an sseL mutant strain led to the accumulation of ubiquitin-modified proteins, indicating that SseL functions as a deubiquitinase in infected cells (Rytkönen et al., 2007). A strain lacking sseL did not display an intracellular replication defect in RAW 264.7 macrophages, but was defective for a SPI-2-dependent late-stage cytotoxic effect in macrophages and attenuated for virulence in the systemic phase of infection in mice (Rytkönen et al., 2007). How the deubiquitinase activity of SseL leads to cytotoxicity remains to be established. Although SseL did not appear to influence cytokine production or interfere with activation or degradation of IκBα upon infection of macrophages (Rytkönen et al., 2007), subsequent work by another group claimed that SseL counteracts IκBα ubiquitination and suppresses proinflammatory cytokine signalling (Le Negrate et al., 2008). Clearly, further work is required to resolve this controversy. A recent study involving stable isotope labelling of amino acids in cell culture (SILAC) and MS to screen for SPI-2 T3SS effectors and interacting host proteins identified oxysterol-binding protein (OSBP) as a possible interacting partner for SseL (Auweter et al., 2011). This could be related to the finding that the deubiquitinase activity of SseL alters the lipid metabolism of infected cells, since lipid droplets accumulate in the gallbladders of infected mice in the absence of SseL (Arena et al., 2011).

Effectors targeting the host cytoskeleton (SteC, SspH2, SrfH, SpvB)

Several SPI-1 effectors interact with host proteins to cause dramatic rearrangements of the actin cytoskeleton (Patel & Galán, 2005). The resulting membrane ruffling drives bacterial uptake. However, these changes are transient, and within a short period of time following bacterial uptake the cortical actin cytoskeleton resumes its normal morphology. Several hours later, S. Typhimurium induces a different form of actin reorganization. As the bacteria begin to replicate, a dense F-actin meshwork assembles in the vicinity of SCVs in several cell types, including epithelial cells, fibroblasts and macrophages (Méresse et al., 2001). This meshwork depends on the SPI-2 T3SS (Méresse et al., 2001) and SteC was subsequently identified as the effector mediating this phenomenon (Poh et al., 2008), indicating that other SPI-2 T3SS effectors are not required for gross actin reorganization. SteC was first discovered in a CyaA-based transposon screen for translocated effectors (Geddes et al., 2005). It is a kinase with weak similarity to human Raf-1 and other mammalian kinases (Poh et al., 2008). Its kinase activity is necessary for actin meshwork formation, and overexpression of SteC in host cells causes massive alterations to the actin cytoskeleton (Poh et al., 2008). However, an steC mutant strain does not have an obvious effect on virulence (Geddes et al., 2005; Poh et al., 2008) and its potential significance as an effector is currently unclear.

SspH2 and SrfH (also called SseI) (Miao & Miller, 2000; Worley et al., 2000) are two SPI-2 T3SS effectors that also localize to the SCV-associated F-actin meshwork (Miao et al., 2003). A yeast two-hybrid assay identified two actin-binding proteins (α-filamin and profilin-1) as interacting partners for SspH2, where the N-terminal region of SspH2 mediated binding to filamin and its C-terminal region interacted with profilin (Miao et al., 2003). As SrfH shares a virtually identical N-terminal region with SspH2 (Miao & Miller, 2000), its interaction with filamin observed in the same study was not surprising (Miao et al., 2003). However, while the binding between SspH2 and profilin was confirmed by glutathione S-transferase (GST) pull-down experiments, binding to filamin was not detected by this method (Miao et al., 2003). Profilin enhances actin polymerization and filamin dimers act by cross-linking actin filaments in areas of active polymerization (Winder & Ayscough, 2005). SspH2 and SrfH did not appear to contribute to the formation of the actin meshwork, but purified SspH2 inhibited the rate of actin polymerization in vitro (Miao et al., 2003), which suggests that it might counteract the effects of SteC.

SrfH has since been reported to influence host cell migration, although in contradictory ways. Yeast two-hybrid and overexpression assays in mammalian cells revealed an interaction between SrfH and thyroid receptor-interacting protein 6 (TRIP6) (Worley et al., 2006). TRIP6 is an adaptor protein that localizes to focal adhesion sites and along actin stress fibres, serving as a platform for recruitment of molecules involved in actin polymerization and cell motility, amongst other functions. SrfH enhanced the ability of bacteria to induce motility of RAW 264.7 macrophages and facilitated bacterial spreading in the mouse model of systemic infection (Worley et al., 2006). However, although TRIP6 was required for the effect of SrfH on cell motility, paradoxically, TRIP6-depleted cells carrying wild-type or srfH mutant bacteria migrated at levels similar to TRIP6-containing cells infected with wild-type bacteria (Worley et al., 2006). This effect of SrfH on cell migration was not replicated in a more recent study in which SrfH was shown to interfere with the direction of movement of primary macrophages, and to inhibit the overall migration of both macrophages and dendritic cells (McLaughlin et al., 2009). This inhibition was dependent on IQGAP1, a regulator of cell migration, which was shown to interact with SrfH both in a GST pull-down assay and when the effector was translocated from intracellular bacteria, which provides strong evidence that IQGAP1 is a physiological target of SrfH (McLaughlin et al., 2009). However, the same study showed that, in RAW 264.7 macrophages, SrfH does not bind IQGAP1 and that its effect in this cell type is to mediate cell detachment (McLaughlin et al., 2009), implying that the function of this effector is not the same in the two cell types. Moreover, SrfH has been shown to be required for long-term systemic infection in mice (Lawley et al., 2006; McLaughlin et al., 2009), and SrfH-dependent inhibition of dendritic cell migration in vivo correlates with a decrease in CD4+ T cells in the spleen of infected mice (McLaughlin et al., 2009).

SpvB is a third effector that appears to interfere with the actin cytoskeleton. This protein is encoded by the spv operon in the Salmonella virulence plasmid (Krause et al., 1992). While this operon (spvRABCD) has been known for many years to be important for S. Typhimurium virulence in the mouse (Gulig et al., 1993; Gulig & Doyle, 1993; Matsui et al., 2001; Rhen et al., 1989; Roudier et al., 1992), it has become clear only recently that at least two of its proteins are substrates for Salmonella T3SSs. SpvB has been studied in some detail and appears to have at least two functional domains: an N-terminal region that has similarity to a toxin from Photorhabdus luminescens, and a C-terminal region that acts on actin as an ADP-ribosylating toxin, inhibiting its polymerization (Lesnick et al., 2001; Tezcan-Merdol et al., 2001; Tezcan-Merdol et al., 2005). SpvB causes cytotoxicity of macrophages, and although SPI-2 T3SS-dependent translocation of SpvB has not yet been demonstrated, this process has also been found to be dependent on the SPI-2 T3SS (Browne et al., 2002, 2008). Interestingly, a Salmonella spvB mutant caused increased actin meshwork formation in HeLa cells (Miao et al., 2003), suggesting that the target of SpvB might be the F-actin meshwork formed as a result of the action of SteC. Furthermore, another link between SPI-2 and SpvB comes from work studying P-bodies, which are RNA/protein aggregates involved in post-transcriptional regulation. In Salmonella-infected human epithelial cells, the SPI-2 T3SS and SpvB were both found to be required for P-body disassembly (Eulalio et al., 2011). However, infection did not affect P-body integrity in mouse macrophages, showing that its effects are cell type-specific.

Effector involved in immune signalling (SpvC)

SpvC is also encoded by the spv operon (Krause et al., 1992). Its secretion from Salmonella requires either the SPI-1 or SPI-2 T3SS, depending on the conditions in which bacteria are grown (Mazurkiewicz et al., 2008). However, SPI-2 T3SS-dependent translocation of SpvC into macrophages was only detected when its degradation was prevented by use of a proteasome inhibitor, which suggests that it is present in relatively small quantities. SpvC is a member of a family of effectors that have phosphothreonine lyase activity on MAPKs (Li et al., 2007; Mazurkiewicz et al., 2008). These enzymes use a process of irreversible phosphate removal, and are therefore potentially extremely powerful. This might provide an explanation for the apparently low amounts of SpvC in host cells. In the same study, SpvC was shown to be active on MAP kinases ERK and JNK, and overexpression of SpvC downregulated the release of TNF-α and IL-8 from infected macrophages and epithelial cells, respectively (Mazurkiewicz et al., 2008). The anti-inflammatory effect of SpvC was confirmed in a recent study using a streptomycin-treated mouse model of enterocolitis (Haneda et al., 2012). In the early stages of infection, SPI-1 T3SS-translocated SpvC led to reduced levels of cytokine mRNA and inflammation compared with an spvC mutant strain, and this correlated with systemic spread of the bacteria in mice.

It is very likely that additional SPI-2 effectors interfere with immune signalling. For example, it has been shown that infection of antigen-presenting cells by S. Typhimurium leads to poly-ubiquitination of major histocompatibility complex class II (MHC-II) molecules, resulting in removal of mature, peptide-loaded molecules from the cell surface. Although this process requires the SPI-2 T3SS, the specific effector(s) responsible have yet to be identified (Lapaque et al., 2009).

Other effectors of the SPI-2 T3SS

Early studies involving S. Typhimurium and mouse macrophages implicated the SPI-2 T3SS in avoidance of SCV–lysosome fusion (Uchiya et al., 1999), the respiratory burst (Vazquez-Torres et al., 2000) and nitric oxide (Chakravortty et al., 2002). The effectors mediating the latter two processes are currently unknown. The SPI-2-encoded protein SpiC was originally proposed to be responsible for SPI-2-mediated phagolysosome fusion avoidance (Uchiya et al., 1999). Additional evidence pointed to SPI-2 T3SS-dependent translocation of SpiC into the cytosol of macrophages (Shotland et al., 2003). Two host cell proteins, Hook3 (Shotland et al., 2003) and NIPSNAP (NSP4) (Lee et al., 2002), have been proposed to interact with SpiC. However, while Golgi-associated Hook3 has been implicated in vesicular trafficking (Xu et al., 2008), NSP4 is a mitochondrial protein (Verhagen et al., 2007). It is not clear how binding of SpiC to these proteins causes inhibition of phagolysosome fusion. It has also been established that SpiC is part of the three-protein regulatory complex that functions within the bacterial cell to mediate the switch from translocon protein secretion to effector translocation (Yu et al., 2004, 2010). Moreover, attempts by our group and others to show that SpiC is itself a translocated protein have failed (Freeman et al., 2002; Yu et al., 2002), suggesting that SpiC is not an effector of the SPI-2 T3SS. The notion that the SPI-2 T3SS is involved in avoidance of or resistance to macrophage killing mechanisms is also difficult to reconcile with more recent work from our laboratory showing that an SPI-2 T3SS null mutant has an intracellular replication defect but does not undergo an increase in killing by macrophages (Helaine et al., 2010). The link between the SPI-2 T3SS and avoidance of the respiratory burst has also been questioned recently (Aussel et al., 2011).

Several other proteins have been implicated as being SPI-2 T3SS effectors, although their functions and physiological contributions to Salmonella virulence remain to be established. These include SifB, PipB, GogB, SseK1, SseK2, SseK3, GtgE, GtgA, SpvD, SteA, SteB, SteD, SteE and CigR.

Like SspH1 and some other effectors mentioned above, SifB was identified by its conserved, translocation-specific N-terminal sequence (Miao & Miller, 2000). It was named after its similarity to SifA, as they share 30 % identity over the length of the protein, and SifB was shown to be translocated into host cells via the SPI-2 T3SS (Miao & Miller, 2000). SifB localizes to the SCV and Sifs (Freeman et al., 2003), although its potential function is unknown and a strain lacking sifB displayed the same level of virulence as a wild-type strain in mice (Ruiz-Albert et al., 2002).

GogB is encoded by the gifsy-1 bacteriophage and localizes to the cytoplasm of infected cells in an SPI-2 T3SS-dependent manner (Coombes et al., 2005). The protein has a partial LRR domain in its N-terminal region (Coombes et al., 2005), and its C-terminal domain is similar to the T3SS effector EspK of enterohaemorrhagic Escherichia coli (EHEC) (Coombes et al., 2005; Vlisidou et al., 2006). A gogB mutant strain showed no major defect in colonization of the small intestine or spleen of infected mice (Figueroa-Bossi et al., 2001).

SteA and SteB were, like SteC, identified in a screen based on transposon mutagenesis coupled with the CyaA fusion method (Geddes et al., 2005). Both proteins were shown to be translocated by the SPI-1 and SPI-2 T3SSs, and SteA localized to the TGN in transfected epithelial cells (Geddes et al., 2005). In the same study, an steA mutant strain was found to be significantly attenuated in infection of mice spleens compared with a wild-type strain (Geddes et al., 2005). However, no functional studies on SteA or SteB have been reported to date.

SseK1 and SseK2 were identified due to their similarity to the E. coli T3SS effector NleB, and their translocation via the SPI-2 T3SS was shown using the CyaA reporter system (Kujat Choy et al., 2004). Recently, a third effector encoded on the Salmonella phage ST64B, named SseK3, was also shown to be translocated by the SPI-2 T3SS (Brown et al., 2011). The three effectors show remarkable similarity at the amino acid level, but their functions are currently unknown. Interestingly, NleB contributes to inhibition of NF-κB activation in epithelial cells infected with EHEC (Newton et al., 2010). Strains lacking SseK1 and SseK2 or all three SseK family members are strongly attenuated for virulence in mice (Brown et al., 2011).

Another group of effectors was identified recently in a proteomic study of the secretome of an ssaL mutant strain, which is derepressed for effector secretion under conditions of SPI-2 activation (Niemann et al., 2011; Yu et al., 2010). Secreted proteins were analysed by MS, and six previously unidentified effectors of the SPI-2 T3SS were detected, namely CigR, GtgA, GtgE, SpvD, SteD and SteE (Niemann et al., 2011). Translocation of all these proteins by the SPI-2 T3SS was confirmed by the CyaA method, and GtgE, SpvD and SteE were also found to be translocated by the SPI-1 T3SS (Niemann et al., 2011). Moreover, mutant strains lacking gtgE, spvD, steD or steE were significantly attenuated for virulence in mice (Niemann et al., 2011). In other work, GtgE was shown to be a protease that specifically cleaves Rab29 (Spanò et al., 2011). In cells infected with S. Typhi, which does not contain gtgE, Rab29 is recruited to the SCV, and exogenous expression of GtgE in S. Typhi leads to its increased replication in macrophages.

SrfJ is another candidate SPI-2 T3SS effector. It is part of the SsrA/B regulon and has sequence similarity to a human glucosyl ceramidase (Worley et al., 2000). A strain carrying a mutation in srfJ was mildly attenuated for virulence in mice (Ruiz-Albert et al., 2002). However, translocation of this protein by the SPI-2 T3SS is yet to be established.