Several effectors contain a conserved N-terminal translocation motif
DNA sequence analysis of the chromosomal region surrounding the PhoP-activated gene pagJ led to the identification of a gene encoding a 700-amino-acid protein that was designated SspH1 (Miao et al., 1999). Probes derived from sspH1 detected a second hybridizing sequence that contained a gene, named sspH2, which encodes a protein of similar size and 69% identity to SspH1. The SspH1 and SspH2 proteins contain leucine-rich repeats. SspH1 was shown to be a target of the SPI-1 and SPI-2 TTSSs, whereas SspH2 translocation was restricted to the SPI-2 TTSS. Both proteins are similar to the Shigella flexneri and Yersinia TTSS effectors IpaH and YopM. A third protein that contains leucine-rich repeats and is similar to SspH1 and SspH2 is SlrP, which was identified as contributing to virulence in mice by an STM screen for genes involved in host adaptation to mice and calves (Tsolis et al., 1999). In contrast to SlrP, SspH1 and SspH2 are not involved in murine systemic infection but do contribute to the virulence of S. typhimurium in calves (Miao et al., 1999).
In a later report by Miao and Miller (2000), a further four ORFs which encode likely effector proteins were identified on the basis of the similarity of their amino termini to those of SspH1, SspH2 and SlrP. These were SifA (Stein et al., 1996) and the previously undescribed proteins SseI, SseJ and SifB. The sequence of SseI gives no clue to its possible function. SseJ is 30% identical to the glycerophospholipid:cholesterol acyltransferase of Aeromonas and includes three residues that make up the catalytic triad of the enzyme, within three conserved regions that exist in all members of this family of acyl transferases. SifB shares 30% overall amino acid identity with SifA, but SifA has no significant similarity to other database entries.
Use of transcriptional reporter fusions showed that expression of sspH2, sseI, sseJ, sifA and sifB is dependent on the SPI-2 TTSS regulator SsrA-SsrB. In contrast, sspH1 and slrP expression was independent of both SsrA-SsrB and the SPI-1 regulator HilA, suggesting that these two genes might be expressed constitutively (Miao and Miller, 2000). To determine if the amino-terminal conserved region of these proteins is required for their translocation, fusions were made to the Bordetella pertussis cyaA gene. Protein fusions to catalytic domains of adenylate cyclase toxin catalyse the conversion of ATP to cAMP only in the presence of calmodulin in the cytoplasm of eukaryotic cells. Hence, the cyaA gene acts as a very useful and sensitive translocation reporter (Sory and Cornelis, 1994). Increased cAMP levels inside host cells, caused by the translocation of SspH1, SspH2, SlrP, SseI and SseJ fusions across the phagosome membrane, was shown to be dependent upon the SPI-2 TTSS. SspH1 and SlrP fusions were also translocated across the plasma membrane by the SPI-1 TTSS. An alignment of the putative translocation signals of all seven proteins revealed a conserved region (WEKI(I/M)XXFF) present in all proteins at amino acid residues 34–41 of the alignment. An SspH1-CyaA fusion carrying a deletion in this motif eliminated its translocation by both SPI-1 and SPI-2 TTSSs (Miao and Miller, 2000). SPI-2-dependent translocation of SifB to the SCV membrane and Sifs has been demonstrated by immunofluorescence microscopy of an epitope-tagged protein (Freeman et al., 2003). Inactivation of sseI or sifB did not result in a detectable virulence defect in systemic infection of mice (Ruiz-Albert et al., 2002), although the high inoculum and relatively short duration of infection may have masked subtle defects. The lack of a virulence defect for sseI and sifB mutants has been confirmed in separate studies (Ho et al., 2002; Freeman et al., 2003). Both sifA and sseJ contribute to systemic infection and their functions are discussed further below.
More recently, another SPI-2 effector (SopD2) has been identified that is 42% identical in amino acid sequence to SopD, a SPI-1 effector that mediates inflammation and fluid secretion in bovine ileal mucosa (Jones et al., 1998). An S. typhimurium sopD2 mutant has a virulence defect in mice, and the protein was shown to be translocated via the SPI-2 TTSS in epithelial cells to the SCV membrane, Sifs and late endocytic compartments (Brumell et al., 2003). Interestingly, the N-terminal 150 amino acids of SopD2 fused to GFP was sufficient to target the protein to late endocytic compartments and to induce their swelling. This was an important finding because it shows that the N-terminal region of SopD2 is likely to contribute to effector function as well being required for secretion/translocation (Brumell et al., 2003).
sspH1 , sspH2 and sseI are all located within lysogenic bacteriophages ( Miao and Miller, 2000 ; Figueroa-Bossi et al., 2001 ; Hansen-Wester et al., 2002b ). This is also true for at least one SPI-1 effector ( Mirold et al., 2001 ), and may represent a common mechanism for the dissemination of Salmonella effectors. sspH2 is encoded within an uncharacterized bacteriophage whereas sseI and sspH1 are within the Gifsy-2 and Gifsy-3 prophages respectively ( Figueroa-Bossi et al., 2001 ). These prophages are competent for lytic growth and lysogeny of naive Salmonella ( Figueroa-Bossi and Bossi, 1999 ). This suggests that sseI and sspH1 are actively disseminating throughout the salmonellae. In accordance with this, sspH1 and sseI are present in a minority of Salmonella serotypes, whereas sspH2 is more widely distributed ( Miao et al., 1999 ; Tsolis et al., 1999 ; Hansen-Wester et al., 2002b ). Indeed, the seven non-SPI-2 genes encoding translocated effectors have a variable G + C content, ranging from 38% for sifB to 55% for sspH2 , implying that they have been acquired by horizontal transmission at different times or from different sources. sopD2 and sseJ are pseudogenes in Salmonella typhi ( Parkhill et al., 2001 ), suggesting that their functions might be host-specific.
The extensive reorganization of host cell lgps into Sifs and SCV membranes in infected epithelial cells was first observed by García-del Portillo et al. (1993). Subsequently, a genetic screen carried out by the same laboratory resulted in the identification of a gene, designated sifA, that is required for Sif formation (Stein et al., 1996). The location of sifA in the potBCD operon, its low G + C content, and the presence of direct repeats flanking the gene all suggest that sifA was acquired as a result of a horizontal transfer. sifA was shown to be necessary for Salmonella virulence, although curiously the mutant grew better than the wild-type strain in epithelial cells (Stein et al., 1996). The discovery that SifA is an important SPI-2 effector has emerged from work of several laboratories. Miao and Miller (2000) found that SifA is regulated by the SPI-2 SsrA-SsrB two component system and that it contains the conserved N-terminal amino acid sequence discussed above. Further evidence for a functional link between SifA and SPI-2 came from: (i) the finding that SPI-2 null mutants fail to make Sifs (Beuzón et al., 2000; Guy et al., 2000; Brumell et al., 2001a); (ii) a genetic test indicating that the two loci interact functionally in vivo (Beuzón et al., 2000); (iii) the finding that sifA expression occurs following bacterial entry into host cells in an ssrA-dependent manner (Beuzón et al., 2000) and (iv) that sifA mutants have a replication defect in macrophages (Beuzón et al., 2000; Brumell et al., 2001a). The replication defect occurs because the sifA mutant loses its vacuolar membrane (Beuzón et al., 2000). In macrophages, this exposes bacteria to a cytosolic killing activity (Beuzón et al., 2002) which might be ubiquicidin, a small cationic protein derived from the ribosomal protein S30 (Hiemstra et al., 1999). Interestingly, although the vacuolar membrane enclosing sifA mutants is also lost in epithelial cells, the killing activity is not present, providing an explanation for the observation that mutant bacteria grow at a greater rate than wild-type bacteria (Beuzón et al., 2002; Brumell et al., 2002). The sifA mutant phenotype can be complemented if SifA is expressed ectopically in host cells infected with mutant bacteria (Beuzón et al., 2000). SPI-2 TTSS-dependent translocation of a haemagglutinin (HA)-tagged SifA to the SCV membrane and Sifs was subsequently shown by fractionation experiments and immunofluorescence microscopy (Brumell et al., 2002). SifA function is dependent on the six C-terminal amino acids of the protein (CLCCFL). This peptide is required to anchor SifA to membranes and is a probable site for lipid modification (Boucrot et al., 2003). Expression of SifA in HeLa cells results in vacuolation of LAMP1-positive compartments and formation of structures resembling Sifs (Brumell et al., 2001b), suggesting that SifA is largely sufficient for Sif formation.
What is the function of SifA? This effector seems to be required both for recruitment of LAMP1-containing vesicles to the bacterial microcolony (Ruiz-Albert et al., 2002), and fusion of these with the SCV membrane, or its stabilisation (Beuzón et al., 2000). These activities could ensure a progressive increase of SCV membrane surface area to accommodate replicating bacterial cells. However, it should also be borne in mind that despite the requirement for a net gain of vacuolar membrane during bacterial replication, bacterial replication per se does not play a significant role in loss of vacuolar membrane of sifA mutant bacteria (Ruiz-Albert et al., 2002). At the biochemical level, SifA might stimulate or inhibit a protein that controls vesicle targeting and membrane fusion in the cell. Clearly, identification of the cellular target(s) of SifA is the essential next step towards an understanding of its function.
Mutation of sseJ results in a modest attenuation of virulence during systemic infection of mice, which is correlated with a small replication defect in primary macrophages (Ruiz-Albert et al., 2002; Freeman et al., 2003). Like SifA, expression of SseJ in Hela cells produces a dramatic reorganization of LAMP1-positive vesicles. However, in the case of SseJ, these take the form of globular membranous compartments (GMCs) which are composed of material that is characteristic of the SCV membrane and Sifs, enclosed by SseJ. SseJ also localizes to the vacuolar membrane and Sifs in infected, transfected cells (Ruiz-Albert et al., 2002), and when expressed from bacteria in infected cells (Kuhle and Hensel, 2002; Freeman et al., 2003). When SseJ and SifA are co-expressed in HeLa cells, GMC-like structures are produced, with SifA enclosed by SseJ. Although these structures (and GMCs) are unlikely to have physiological relevance, they suggested a functional link between SseJ and SifA (Ruiz-Albert et al., 2002). Of particular interest was the observation that a sifA sseJ double mutant does not lose its vacuolar membrane, implying that loss of the vacuolar membrane around a sifA mutant requires the action of SseJ (Ruiz-Albert et al., 2002). Because SseJ is predicted to be an acyl transferase, one possiblity is that its function is to modify SCV membrane lipids to facilitate budding and/or scission. In the absence of SifA, SseJ activity on the SCV membrane could result in its destabilisation and eventual loss.
Other members of the SsrA-SsrB regulon
Using a genetic screen of 20 000 mutants, Worley et al. (2000) identified numerous genes, located outside SPI-2, that are regulated by SsrB. Ten genes were characterized further, all of which appear to have been acquired by lateral transfer. Their level of ssrB-dependent activation in macrophages ranged from two- to 80-fold. The most highly regulated gene, srfH, has similarity to sspH1, sspH2 and slrP, and is located in a Gifsy-2 prophage. It was recently confirmed by Figueroa-Bossi et al., 2001) that srfH and sseI are the same gene. The product of another gene, named srfJ, shares striking amino acid similarity with the human lysosomal glucosyl ceramidase. In several systems ceramide initiates the stress-activated protein kinase (SAPK)/c-jun kinase (JNK) cascade to signal apoptosis. Therefore SrfJ might be the effector responsible for SPI-2-dependent apoptosis (van der Velden et al., 2000), but this has yet to be examined. Mutation of srfJ leads to a very mild attenuation of virulence in mice (Ruiz-Albert et al., 2002). The possible functions of most of these genes remain unknown, and indeed it has yet to be demonstrated that they are SPI-2 effectors.
Further organizational complexity of genes encoding Salmonella effectors has been revealed by the discovery that in another PAI, SPI-5, two adjacent genes encoding the SPI-1 TTSS effector SigD/SopB and a novel SPI-2 effector, PipB, are co-ordinately regulated by their respective regulatory systems (Knodler et al., 2002). An HA-tagged PipB was shown to be translocated to the SCV membrane and Sifs (Knodler et al., 2002). However, in contrast to SifA, PipB was not required for the formation or maintenance of either the Sifs or the SCV membrane. No defect in intracellular survival or virulence could be detected for a pipB mutant, perhaps reflecting functional redundancy with another effector. PipB translocation signals were localized to its N-terminal domain. However, this region does not contain the WEKI(I/M)XXFF motif present in the translocation domain of some of the other SPI-2 secreted effectors. How PipB, SpiC, SseF and SseG are recognized and translocated by the SPI-2 TTSS is unknown. It is possible that all effectors have N-terminal three dimensional similarity not apparent from primary sequence; alternatively, different mechanisms might be employed to target effector protein secretion and translocation.
Further work is required to establish if the genes described above encode the complete repertoire of SPI-2 effectors. It would be beneficial if bacterial growth conditions could be found which induce secretion of substantial amounts of SPI-2 effectors in vitro, as this would enable their identification by polyacrylamide gel electrophoresis followed by microsequencing or mass spectrometry. However, even when large culture volumes are used in conjunction with optimized secretion conditions, we have only been able to detect low amounts of secreted SPI-2 effectors by Coomassie blue or silver staining of gels (J. Garmendia and D. W. Holden, unpubl. data).
Salmonella typhimurium can induce a rapid programmed cell death in macrophages via the SPI-1 TTSS through the action of the secreted protein SipB ( Chen et al., 1996 ; Monack et al., 1996 ; Hersh et al., 1999 ). However, S. typhimurium was also found to induce a delayed apoptosis in infected macrophages beginning 12–13 h after ingestion, in the absence of a SPI-1 secretion system or SipB ( van der Velden et al., 2000 ). This process was dependent on a functional SPI-2 TTSS, but was not simply a consequence of intracellular bacterial proliferation because a mutant with an unrelated replication defect was as cytotoxic as the wild-type strain ( van der Velden et al., 2000 ). Recent reports suggests that the apoptotic-like process is in fact an unusual form of necrosis ( Brennan and Cookson, 2000 ; Watson et al., 2000 ). In vivo , this SPI-2-dependent cytotoxicity could facilitate bacterial spread within the spleen and liver. In agreement with this, dying cells containing Salmonella have been observed in livers of infected mice ( Richter-Dahlfors et al., 1997 ).
Although caspase-1 has been shown to be necessary for orally administered Salmonella to colonize murine Peyer's patches, inoculation of caspase-1–/– mice with S. typhimurium via the intraperitoneal route resulted in a level of bacterial colonization of the spleen that was no different from that found in wild-type mice (Monack et al., 2000). Experiments with bone marrow-derived macrophages showed that SPI-2-induced macrophage cytotoxicity is only partly dependent on caspase-1 (Monack et al., 2001). These results suggest either that SPI-2-dependent cytotoxicity has no significant role in promoting bacterial proliferation in the spleen in vivo, or that it does have, but by a mechanism that is independent of caspase-1. The SPI-2 secreted effector(s) responsible for this macrophage cytotoxicity have not been identified. In Salmonella dublin, the SPI-2 locus has been shown to be required for apoptosis of intestinal epithelial cells (Paesold et al., 2002).
Intracellular actin assembly
Several hours after uptake by a variety of host cells, the SPI-2 TTSS induces the assembly of F-actin in close proximity to the SCV membrane and Sifs (Méresse et al., 2001; Brumell et al., 2002; Miao et al., 2003). The G-actin sequestering agent latrunculin B prevented the accumulation of actin in Salmonella-infected cells and prevented bacterial replication in macrophages. In addition, latrunculin B treatment caused vacuolar membrane loss from a significant proportion of wild-type bacteria (Méresse et al., 2001). Therefore the physiological role of the SPI-2 dependent actin meshwork would appear to be connected with vacuolar membrane dynamics. It might stabilize the structure of the membrane, or facilitate transport of vesicles destined to fuse with the SCV. The effector(s) controlling this process could be unknown, as strains carrying mutations in either sifA, sifB, slrP, srfJ, sseE, sseF, sseG, sseI, sspH2, sseI sspH2 or sseJ were all proficient for actin assembly (Méresse et al., 2001; Yu et al., 2002; Miao et al., 2003). However, in common with actin rearrangements induced by TTSSs of other bacterial pathogens, the process is likely to be modulated by the concerted actions of several effectors. Indeed, it has recently been shown that SspH2 and SseI co-localize with the SPI-2 actin meshwork and a two-hybrid screen revealed that they interact with the actin crosslinking protein filamin through their conserved amino-terminal domains (Miao et al., 2003). SspH2 was also shown to bind the actin-binding protein profilin. In addition, another S. typhimurium virulence protein, SpvB, which has been shown to ADP-ribosylate actin (Lesnick et al., 2001; Tezcan-Merdol et al., 2001), strongly inhibited SPI-2 associated actin polymerization. However, the physiological relevance of this result is not clear, as genetic studies have showed that SPI-2 and the spv operon function independently in vivo (Shea et al., 1999).
Yet another characteristic of SCVs that appears to be SPI-2-dependent is alteration of iNOS localization in infected macrophages (Chakravortty et al., 2002). Confocal microscopic examination of infected RAW 264.7 and peritoneal macrophages revealed that the subcellular localization of iNOS was dramatically different in macrophages infected with wild-type Salmonella or SPI-2 mutants. iNOS was shown to co-localize strongly with SPI-2 mutants but not with the wild-type strain. It was suggested that increased killing of SPI-2 mutants by nitrogen-derived radicals contribute to their reduced intracellular proliferation. No difference was observed between wild-type Salmonella- and SPI-2 mutant-infected cells in terms of iNOS protein levels or nitrite production, indicating that SPI-2 is not involved in inhibiting this activity. Curiously, when the growth of SPI-2 mutant and wild-type bacteria was examined by mixed infections in spleens and livers of iNOS–/– mice, the SPI-2 mutant was strongly out-competed up to three days post inoculation, but then grew faster than the wild-type strain to reach equivalent numbers in the liver 48 h later. The reason for this is not clear, but it clearly warrants further investigation.
Cholesterol accumulation near the SCV
Recent studies have found that cholesterol is recruited to SCVs in epithelial cells and macrophages (Brumell et al., 2001b; Catron et al., 2002). In macrophages, recruitment of cholesterol and the cholesterol-associated glycosylphosphatidylinositol (GPI)-anchored protein CD55 were shown to be dependent on a functional SPI-2 TTSS (Catron et al., 2002), suggesting the involvement of a SPI-2 effector. However, other mutants with intracellular replication defects were not analysed for cholesterol recruitment, so it is conceivable that the phenotype is not a specific function of SPI-2 but a more general characteristic of replicating bacteria. In any event, the recruitment of cholesterol is striking and is consistent with the emerging theme linking SPI-2 and membrane dynamics. It is interesting to note that GMCs induced by expression of SseJ also contain cholesterol (Ruiz-Albert et al., 2002).