Salmonella typhimurium leucine-rich repeat proteins are targeted to the SPI1 and SPI2 type III secretion systems


Samuel I. Miller. E-mail millersi@u.washington,edu; Tel. (+1) 206 616 5107; Fax (+1) 206 616 4295.


Salmonellae encode two virulence-associated type III secretion systems (TTSS) within Salmonella pathogenicity islands 1 and 2 (SPI1 and SPI2). Two Salmonella typhimurium genes, sspH1 and sspH2, that encode proteins similar to the Shigella flexneri and Yersinia species TTSS substrates, IpaH and YopM, were identified. SspH1 and SspH2 are proteins containing leucine-rich repeats that are differentially targeted to the SPI1 and SPI2 TTSS. sspH2 transcription was induced within RAW264.7 macrophages, and was dependent upon the SPI2-encoded regulator ssrA/ssrB. In contrast, sspH1 transcription is independent of SPI2, and is not induced after bacterial phagocytosis by eukaryotic cells. Infection of eukaryotic cells with strains expressing a SspH2–CyaA fusion protein resulted in SPI2 TTSS-dependent cAMP increases. In contrast, SspH1–CyaA-mediated cAMP increases were both SPI1 and SPI2 TTSS dependent. sspH2-like sequences were found in most Salmonella serotypes examined, whereas sspH1 was detected in only one S. typhimurium isolate, indicating that the copy number of sspH genes can be variable within Salmonella serotypes. S. typhimurium deleted for both sspH1 and sspH2 was not able to cause a lethal infection in calves, indicating that these genes participate in S. typhimurium virulence for animals.


Salmonellae are enteric Gram-negative bacteria that cause gastroenteritis and enteric (typhoid) fever after oral ingestion (reviewed in Miller et al., 1995). Salmonella serotypes exhibit either a narrow or broad host range. Salmonella typhi infects only humans and causes typhoid fever. Bacteria spread after oral ingestion to the liver, spleen and bone marrow, where they persist in the phagosomal compartment of macrophages. In contrast, Salmonella typhimurium is a broad host range serotype that causes host-specific clinical manifestations. Inbred mice with a specific mutation in the Nramp allele develop a systemic illness similar to human typhoid fever after S. typhimurium infection. Other animals, including humans and cattle, develop gastroenteritis, a disease characterized by diarrhoea and intestinal neutrophil infiltration. S. typhimurium gastroenteritis is usually self-limiting. However, in young or immunocompromised individuals, it can be lethal. Despite the varying host ranges and diseases caused by Salmonella serotypes, many virulence determinants are conserved among most salmonellae, notably two type III secretion systems (TTSS) that are required for several aspects of pathogenesis (Hueck, 1998).

Virulence-associated TTSS are specialized secretion systems that permit Gram-negative bacterial pathogens to translocate effector proteins directly into the host eukaryotic cell cytoplasm (Hueck, 1998). In Salmonella, two TTSS are encoded by separate pathogenicity islands, Salmonella pathogenicity island 1 (SPI1) and Salmonella pathogenicity island 2 (SPI2). These pathogenicity islands contain genes that appear to encode independent secretory apparatuses, translocase proteins, regulatory genes and effectors (Ochman et al., 1996; Hensel et al., 1997; Hueck, 1998). Other translocated effectors (sopE, sopB/sigD) are encoded outside SPI1 and SPI2 (Wood et al., 1996; Galyov et al., 1997; Hardt et al., 1998a; Hong and Miller, 1998).

SPI1 TTSS is required for a number of virulence phenotypes, including invasion of epithelial cells by induction of macropinocytosis (Francis et al., 1993), stimulation of IL-8 secretion from and neutrophil transmigration across epithelial monolayers (McCormick et al., 1993), cytotoxicity of macrophages (Chen et al., 1996; Monack et al., 1996) and fluid efflux into ligated bovine ileal loops (Galyov et al., 1997). These functions are consistent with a role for SPI1 TTSS in the pathogenesis of gastroenteritis. In accordance with this theory, SPI1 gene expression is maximal in bacterial culture conditions that mimic the environment of the intestinal lumen (Bajaj et al., 1996), and an intact SPI1 TTSS is required for virulence in calf gastroenteritis (Tsolis et al., 1999a). Several effector proteins translocated by SPI1 TTSS have been identified, and implicated in the invasion process (Wood et al., 1996; Fu and Galán, 1998; Hardt et al., 1998b; Hong and Miller, 1998; Zhou et al., 1999) and induction of intestinal fluid efflux (Galyov et al., 1997; Jones et al., 1998; Norris et al., 1998). Although effector proteins have a variety of effects when expressed in mammalian cells, deletion of no single non-translocase effector causes a complete defect in any SPI1 TTSS related phenotypes, suggesting phenotypic redundancy among effector proteins (Hueck et al., 1995; Kaniga et al., 1996; Wood et al., 1996; Galyov et al., 1997).

Although SPI1 TTSS has only a modest effect on S. typhimurium virulence in mice, SPI2 TTSS mutation results in a severe reduction in murine virulence (Hensel et al., 1995; Ochman et al., 1996). SPI2 genes are maximally expressed only when bacteria have resided intracellularly for several hours (Cirillo et al., 1998). Consistent with a role in intracellular survival, SPI2 TTSS is required for efficient replication within host cells (Cirillo et al., 1998; Hensel et al., 1998). It is not known what molecular events are involved in SPI2-related phenotypes, in part because translocated effector proteins have not been described. Such effectors are predicted to be translocated across the vacuolar membrane to gain access to the host cell cytosol.

This work reports the identification and characterization of proteins containing leucine-rich repeats (LRR) (for a review of LRR domains, see Buchanan and Gay, 1996) that are differentially translocated by SPI1 and SPI2 TTSS.


Identification of two S. typhimurium genes predicted to encode proteins similar to the IpaH family and YopM

DNA sequence analysis of the chromosomal region surrounding a PhoP-activated gene, pagJ, suggested that this region could have been acquired by horizontal transmission (Gunn et al., 1998). Therefore, DNA sequencing of plasmid pCS01, containing the chromosomal region upstream of pagJ was performed (GenBank accession number AF013776). An open reading frame (ORF), designated sspH1 (Salmonellasecreted protein H1), that was predicted to encode a 700-amino-acid protein was identified. A potential ribosome binding site (TAGGCA) is located 8 base pairs upstream of the predicted sspH1 start codon. A diagram of the chromosomal region surrounding sspH1 is shown in Fig. 1. The predicted sspH1 gene product was of interest because it has sequence similarity to the IpaH family in Shigella flexneri (Venkatesan et al., 1991) and YopM in Yersinia species (Boland et al., 1996; Perry et al., 1998) (Fig. 1). These proteins are secreted (IpaH9.8 and YopM) and translocated (YopM) by TTSS (Boland et al., 1996; Demers et al., 1998; Skrzypek et al., 1998), and contain LPX repeats, a subtype of the leucine-rich repeat (LRR) superfamily of protein binding domains (Buchanan and Gay, 1996).

Figure 1.

. Genetic map of chromosomal regions surrounding sspH1 and sspH2 and graphical representation of LPX family members. A. Chromosomal region surrounding sspH1 and sspH2. The sspH1 sequence lies 171 bp from the rhsD-like sequence found upstream of pagJ and is divergently transcribed from pagJ (Gunn et al., 1998). Sequences located 274 bp upstream of the sspH2 start codon encode 54 amino acids, which are 52% similar to a homologue of the phage P2 tail fibre assembly protein, whereas sequences beginning 444 bp downstream of the sspH2 stop codon encode at least 195 amino acids with 49% similarity to a bacteriophage P1 ORF of unknown function (orfA in the DOC-GP10 intergenic region). Farther upstream from sspH2 (≈ 1 kb), sequences encode at least 191 amino acids with 50–70% similarity to tail fibre proteins of several bacteriophages (unpublished data). B. LPX family members (plus IpaH2.5 and IpaH1.4) are shown in a graphical representation. orf1–3 have been identified in the Y. pestis genome sequence. Textures indicate regions of sequence similarity and grey shaded boxes are LRR repeat domains (with the number of repeats indicated). Overall percentage similarity to SspH2 is listed to the right. Percent identity to corresponding SspH2 domains are listed under selected domains.

As a result of the multicopy nature of the genes encoding the IpaH protein family, Southern blot analysis of chromosomal DNA was undertaken to determine whether other genes similar to sspH1 were present in S. typhimurium 14028s. Probes derived from sspH1 detected a second hybridizing sequence, which was subsequently cloned (pEM12) and subjected to DNA sequence analysis (GenBank accession number AF160727). An ORF that is predicted to encode a 788-amino-acid protein with 69% identity to SspH1 was identified and named sspH2. A potential ribosome binding site (AAGGTT) occurs 8 bp upstream of the methionine start site of the sspH2 ORF. Sequences with similarity to phage genes are located in the vicinity of sspH2 (Fig. 1). Based on its proximity to pagJ, sspH1 is located between centisomes 23 and 25 on the S. typhimurium chromosome (Gunn et al., 1998). Using a collection of Mud-P22 bacteriophage insertions (Benson and Goldman, 1992), sspH2 was determined to map to centisome 66.8–69 on the S. typhimurium chromosome, as probes created from the sspH2 LPX repeats and C-terminal domain hybridized to phage DNA from strain TT16706 zgc-1715::MudP (data not shown). As sspH1 is absent in LT2, the strain in which the library was created, it cannot account for the hybridization signal.

Analysis of domain structure of SspH1 and SspH2

Analysis of the predicted amino acid sequence of SspH1, SspH2 and the family of related proteins permits the division of SspH1 and SspH2 into four domains (Fig. 1). Analysis of SlrP, a third LPX repeat protein that was recently identified in S. typhimurium (Tsolis et al., 1999b) is also presented. Data presented below indicate that secretion/translocation signals are present in domains A and B. The SspH2 domain A displays low sequence identity to SspH1 (28%) and SlrP (30%), which are more identical to each other (42%). Throughout the remainder of the proteins, SspH1 and SspH2 are more identical to each other than to SlrP. The SspH1/SspH2/SlrP domain B has similarity to the 24–45 residues that precede the LPX repeats in the IpaH family and YopM. Eight (SspH1), 12 (SspH2), or 11 (SlrP) leucine-rich repeats form the third domain. The majority of the LRR of SspH1 and SspH2 conform exactly to an LPX repeat structure LTSLPxLPxx LxxLxaxxNx, in which ‘a’ represents any aliphatic residue (Venkatesan et al., 1991). The SlrP LRR structure is one residue longer (LxxLPxxLPxxLxxLxaxxNx), similar to the repeat structure seen in three ORFs encoded by Yersinia pestis (The Sanger Center,, Contig 675). Although the final leucine-rich sequence in SlrP (LxxLPxxLxxFx9IxVxxNPF), and SspH1/SspH2 (LxxLPxx LxxLx5VxLxxNPL) is recognizable as belonging to the LRR superfamily (Buchanan and Gay, 1996), it does not share the LPX consensus sequence. Terminal repeat motifs similar to those seen in SspH1 and SspH2 are also present in all LPX proteins containing an IpaH-like C-terminal domain, but absent in YopM. The C-terminal domains of SspH1, SspH2 and SlrP share sequence similarity with the C-terminal domain of the IpaH family (Fig. 1).

sspH2 is conserved among most Salmonella serotypes

In order to determine whether sspH1 and sspH2 are conserved among Salmonella serotypes, Southern blot analysis was performed. Probes created from the sequences encoding the LPX and C-terminal domains of sspH1 or sspH2 hybridize to both genes, whereas probes created from domain A sequences differentiate sspH1 and sspH2 (Fig. 2 and data not shown). When several Salmonella serotypes were examined with various sspH probes, sspH2-like sequences were found in almost all serotypes, including four bands in Salmonella arizonae. In accordance with this data, the S. typhi genome sequencing project (The Sanger Center, Contig 307) has identified a gene that encodes a protein that is 97% identical to SspH2 in S. typhimurium 14028s. On the other hand, sspH1 was restricted to S. typhimurium 14028s and was absent even from four other S. typhimurium isolates examined. In addition, sspH2 domain A probes detected an additional band (Fig. 2). Southern blots, using other sspH2 or sspH1 (Fig. 2 and data not shown) probes, were not able to detect this sequence under the conditions used. All Southern blot results were confirmed with at least two other restriction enzyme digestions. These results suggest that sspH2 might be a virulence factor common to most salmonellae, whereas sspH1 has a more restricted distribution.

Figure 2.

. Phylogenetic distribution of sspH hybridizing sequences. Southern blots of various Salmonella strains were performed utilizing probes as indicated. DNA was digested with MluI (except second panel of B was digested with EcoRI) and probed with (A) sspH2 LPX repeats, (B) sspH2 domain A, and (C) sspH1 domain A. The position of sspH1 and sspH2 in S. typhimurium 14028s are indicated. The additional band hybridizing to sspH2 domain A is indicated with a (*) for S. typhimurium 14028s. Results were confirmed in experiments using at least two different restriction enzymes. Although the sspH1 band and the sspH2 domain A hybridizing band (*) migrate at similar sizes in the MluI digests shown, other Southern blots using different restriction enzymes (data not shown and second panel of B) have verified that the sspH2 domain A probe does not hybridize to sspH1.

Transcription of sspH2, but not sspH1, is induced in intracellular bacteria

Single-copy transcriptional fusions of sspH1 and sspH2 to the firefly luciferase (f-luc) gene were created in order to examine the expression of these two genes. These fusions were constructed using the pGPLFR03 suicide vector that also encodes the renilla luciferase (r-luc) gene under constitutive expression from the tet promoter to allow for standardization of f-luc activity to cell number.

The effects of various regulatory mutations were examined for their ability to alter sspH1::f-luc and sspH2::f-luc expression (Table 1). Mutations in most SPI1 or SPI2 regulatory genes (hilA, invF, phoP, ssrA) had little effect on sspH1::f-luc or sspH2::f-luc transcription in L broth (LB) (the minor effect of phoP::Tn10d-Cm seen in this experiment was not reproducible) (Table 1). A mutation in sirA caused a consistent decrease in expression between 2- and 3.4-fold for sspH1::f-luc and 1.6- and 2.3-fold for sspH2::f-luc (Table 1 and data not shown). Similar results (5.5- and 2.3-fold regulation of sspH1::f-luc and sspH2::f-luc by sirA) were obtained in micro-aerophilic cultures (data not shown). Constitutive activation of phoP (pho-24) also caused a reproducible 1.4- to 1.9-fold decrease in expression of sspH2::f-luc (Table 1 and data not shown).

Table 1. . Transcription of sspH1::f-luc and sspH2::f-luc in the presence of regulatory mutations. a. Stationary phase bacteria were diluted 1:100 in fresh LB and incubated with aeration for 4 h before luciferase enzyme activity was determined.RLU, firefly luciferase relative light units over 10 s/renilla luciferase relative light units over 10 s.Thumbnail image of

SPI2 genes are minimally expressed in rich culture media, but are induced by SsrA/SsrB after bacteria have resided in eukaryotic cells for several hours (Valdivia and Falkow, 1997; Cirillo et al., 1998). In order to assay SsrA/SsrB regulation, murine macrophage-like cells (RAW264.7) were infected with S. typhimurium strains, carrying the sspH1::f-luc or sspH2::f-luc fusions. Little or no alteration of expression was seen in non-adherent bacteria expressing either fusion after 1 h of infection (data not shown). Cell-associated bacteria displayed reduced transcription of sspH1 and sspH2 by 1.7- and 6.4-fold respectively, after 1 h of infection (Table 2). After prolonged exposure to the phagosomal/endosomal environment in RAW264.7 (Table 2, macrophages) or HeLa (data not shown) cells, sspH2::f-luc was highly induced. Expression of sspH2::f-luc in intracellular bacteria was 70-fold higher than expression in a ssrA/ssrB mutant, indicating that sspH2 is co-ordinately regulated with TTSS genes in SPI2. Conversely, sspH1 expression levels in intracellular bacteria were approximately half those seen in LB and were unaffected by mutations in ssrA/ssrB (Table 2), indicating that sspH1 is not greatly induced or repressed in intracellular bacteria.

Table 2. . Transcription of sspH1::f-luc and sspH2::f-luc in LB, RAW264.7 macrophages and in low-magnesium media. a. Stationary phase bacteria were diluted 1:100 in fresh LB and incubated with aeration for 4 h before luciferase enzyme activity was determined. RLU, firefly luciferase relative light units over 30 s/renilla luciferase relative light units over 10 s.b. RAW264.7 cells were infected with stationary phase bacteria for 1 h and lysed, or treated five additional hours with gentamycin before lysis. Luciferase enzyme activity was determined in cell-associated bacteria.c. Stationary phase bacteria were diluted 1:100 in 8 μM MgCl2 N-minimal media and incubated with aeration for 16 h before luciferase enzyme activity was determined.d. Less than or equal to the limit of detection listed.Thumbnail image of

A recent paper (Deiwick et al., 1999) has described in vitro conditions that induce SPI2 gene expression. In agreement with the published data and our data above, we found that when bacteria were grown in low- magnesium minimal media (8 μM MgCl2 N-minimal media) sspH2::f-luc expression was induced dependent on ssrA/ssrB, whereas sspH1::f-luc expression was not induced or dependent on ssrA/ssrB (Table 2).

SspH1 is secreted and translocated by the SPI1 TTSS

In order to examine secretion of SspH1 by the SPI1 TTSS, polyclonal antibodies were raised against a GST–SspH1 fusion protein. When SspH1 was expressed from a low copy expression vector, it was readily detected in the supernatant fractions of wild-type S. typhimurium, whereas SPI1 mutants (ΔprgH-K) were unable to secrete SspH1 (Fig. 3A). Analysis of sspH2 mutants or sspH2 overexpressing strains indicate that none of the hybridization signal seen can be attributed to cross-reaction of the GST–SspH1 antisera with SspH2 (not shown). This antisera was unable to detect SspH1 protein in wild-type bacteria (Fig. 3A).

Figure 3.

. Secretion of SspH1 by the SPI1 TTSS. Bacterial supernatant proteins from overnight cultures (upper panels) or cellular proteins (lower panels) from late logarithmic (3 h) or late stationary (overnight) phase cultures were separated by SDS–PAGE, transferred to nitrocellulose and probed with antibodies. A. Supernatant and cellular proteins from wild-type bacteria (CS401) or bacteria expressing SspH1 in wild-type (EM163) and ΔprgH-K (EM186) were probed with antiserum raised against GST–SspH1. B. Supernatant and cellular proteins from strains expressing SspH1–CyaA (EM113–115) or SspH2–CyaA (EM117–119) were probed with monoclonal antibody 3D1, which recognizes CyaA. Cytoplasmic protein contamination was low and equal in all supernatant preparations and cannot account for the differences seen between the strains. Molecular weight markers (kDa) are shown to the right of each gel.

In order to determine whether SspH1 and SspH2 are translocated into the cytoplasm of eukaryotic cells, protein fusions to the catalytic domain of the adenylate cyclase toxin (CyaA) from Bordetella pertussis were created. This system has been used extensively to examine translocation (Sory and Cornelis, 1994; Sory et al., 1995; Boland et al., 1996; Schesser et al., 1996; Jones et al., 1998; Wolff et al., 1998). Briefly, the toxin efficiently converts ATP to cAMP only in the presence of calmodulin, which is found in the cytoplasm of eukaryotic cells. As S. typhimurium that do not express CyaA fusions do not alter host cell cAMP levels (data not shown), increases in cellular cAMP in infected cells are indicative of the introduction of the cyclase fusion into the eukaryotic cell cytosol.

Domains A and B (208/214 codons) of either sspH1 or sspH2 were translationally fused to the catalytic domain of cyaA and expressed from a constitutive (lac) promoter in pWSK29, a low copy expression vector (Wang and Kushner, 1991). Similar adenylate cyclase activity (not shown) and protein content (Fig. 3C) can be found in bacteria expressing SspH1–CyaA and SspH2–CyaA, indicating comparable expression and activity levels of the two constructs. SspH1–CyaA was secreted into culture supernatants by wild-type S. typhimurium, but not by prgH::Tn5-lacZY mutants (Fig. 3B). SspH2–CyaA was secreted into bacterial culture supernatants at very low levels compared with SspH1–CyaA by wild type but not by prgH::Tn5-lacZY mutants (Fig. 3B). An SPI2 TTSS apparatus mutant (ssaT::mTn5) did not alter the ability of S. typhimurium to secrete SspH1–CyaA or SspH2–CyaA under these conditions (Fig. 3B)

When wild type or a SPI2 apparatus mutant (ssaT::mTn5) expressing SspH1–CyaA were used to infect eukaryotic cells, increases in intracellular cAMP were marked 1 h post infection in several cell lines examined, including HeLa cells, bone marrow-derived macrophages (BMM, data not shown) and RAW264.7 macrophages (Fig. 4A). SspH1–CyaA expressing strains, carrying mutations in genes encoding the SPI1 secretory apparatus (ΔprgH-K, data not shown) or translocase (ΔsspB, ΔsspC or ΔsspD; Fig. 4A and data not shown), were unable to cause cellular cAMP levels to increase. Strains expressing the SspH2–CyaA fusions, on the other hand, did not produce any increase in cellular cAMP levels at early time points in RAW cells (Fig. 4A), BMM or HeLa cells (data not shown). Even at a multiplicity of infection (MOI) of 100, SspH2–CyaA expressing strains were not able to increase the level of cAMP in HeLa cells after 1 h of infection (data not shown).

Figure 4.

. cAMP increases mediated by SspH–CyaA fusion proteins. Measurement of cAMP increases mediated by SspH1–CyaA or SspH2–CyaA inRAW264.7 macrophages. Macrophages were infected at an MOI of 5 with (A) SPI1 induced (late logarithmic) bacteria for 1 h or (B) SPI1 non-translocating (stationary phase) bacteria for 1 h plus 5 h gentamycin treatment. Error bars represent ± 1 standard deviation. All strains were determined to be equally invasive and persistent within RAW264.7 macrophages and negligibly cytotoxic under the conditions used in these experiments. Protein content was determined using the Bradford assay.

Published data have shown that the ssaT::mTn5 (P9B7) mutation causes defects in SPI1 TTSS-related phenotypes, such as epithelial cell invasion and protein secretion into bacterial culture media (Hensel et al., 1997; Deiwick et al., 1998). When transduced into S. typhimurium 14028s and its derivatives, this mutation did not have any effect on epithelial cell invasion, macrophage cytotoxicity (data not shown), profiles of secreted proteins (Fig. 3B and data not shown) or SPI1 TTSS translocation efficiency (Fig. 4A), under the conditions used in our laboratory. Thus, we were able to use this mutation to eliminate SPI2 TTSS without interfering with the function of SPI1 TTSS.

SspH1–CyaA and SspH2–CyaA are translocated by SPI2 TTSS

SPI2 TTSS is expressed by intracellular bacteria and we thus speculate that SPI2 translocates proteins across the phagosomal membrane. The CyaA fusion protein strategy was employed to examine SPI2 TTSS-dependent translocation. Stationary phase bacteria did not demonstrate SPI1 TTSS-dependent translocation of SspH1–CyaA into RAW264.7 cells (Fig. 4B), and were thus utilized to examine SPI2 TTSS-dependent translocation in the absence of SPI1 TTSS activity. Stationary phase bacteria, expressing either SspH1–CyaA or SspH2–CyaA, were unable to mediate increases in intracellular cAMP after 1 h of infection (data not shown), but were able to cause cAMP level elevation after 6 h of infection (Fig. 4B). A mutation in the SPI2 (ssaT::mTn5) but not in the SPI1 (ΔsspC) TTSS eliminated the ability of SspH1–CyaA and SspH2–CyaA fusion-expressing strains to cause cAMP levels to rise in infected cells (Fig. 4B). These results suggest that both SspH2 and SspH1 contain secretion/translocation signals recognized by SPI2 TTSS. These increases in cAMP are consistent with translocation of the fusion protein across the phagosomal membrane.

cAMP elevations could also result from increased permeability of the phagosomal membrane, which would permit previously secreted proteins to traffic to the cytoplasm. In order to investigate this possibility, the SspH1–CyaA expressing strains were grown under SPI1 inducing conditions (late logarithmic growth) and were used to infect macrophages for 6 h. cAMP increases were detectable in the presence of either SPI1 (ΔsspC) or SPI2 (ssaT::mTn5) individual mutations (data not shown), consistent with SspH1–CyaA being translocated by both SPI1 and SPI2 TTSS under these growth conditions. However, no increases in cAMP were detected when macrophages were infected with the ΔsspC, ssaT::mTn5 double mutant (data not shown). This strain should be able to secrete SspH1–CyaA into the vacuole via the SPI1 TTSS, but lacks the SPI1 translocase (ΔsspC). A second experiment to address this issue in the presence of functional SPI1 and SPI2 TTSS was performed. Overnight culture supernatants of wild-type bacteria expressing SspH1–CyaA were concentrated and added to an infection of wild-type bacteria not expressing CyaA fusions, in an attempt to provide SspH1–CyaA to the vacuolar compartment after S. typhimurium-induced macropinocytosis. The concentrated supernatant of 100 bacteria was added for every one wild type in the infection, which did not affect macrophage cytotoxicity. After 1 h or 6 h of infection, no increases in cAMP were detected (data not shown). In order to ensure that the SspH1–CyaA fusion protein was present in the vacuolar compartment, infected cells were treated with trypsin to degrade extracellular protein, washed and lysed in non-denaturing conditions. Adenylate cyclase activity was recovered after 1 h of infection, consistent with the presence of SspH1–CyaA in the vacuolar compartment. No activity was recovered after washing and five additional hours of gentamycin treatment, probably due to degradation of phagocytized/macropinocytosed protein (data not shown). In addition, co-infection of wild-type bacteria not expressing a CyaA fusion protein (MOI = 5) and ΔsspC, ssaT::mTn5 expressing SspH1–CyaA (MOI = 50) did not cause host cell cAMP levels to rise, suggesting that the translocation system cannot be supplied by co-infecting bacteria (data not shown). These results are consistent with the hypothesis that proteins in the vacuole are not released to the cytosol independently of the TTSS. These controls have not completely ruled out the possibility that the SPI2 TTSS causes the phagosomal membrane to lose integrity and permits traffic between the cytoplasm and the vacuolar contents. If this is the case, then any secreted protein will have access to the cytosol.

sspH1 and sspH2 contribute to virulence in calves

In order to analyse the role of sspH1 and sspH2 in Salmonella virulence, strains deleted for sspH1, sspH2, or both sspH1 and sspH2 were created by allelic exchange (see Experimental procedures) and verified by polymerase chain reaction (PCR) amplification (data not shown) and Southern blot analysis (Fig. 2 and data not shown).

The ΔsspH1ΔsspH2 strain was tested in various assays for virulence, investigating those phenotypes for which SPI1 or SPI2 function is required. A functional SPI1 TTSS is required for invasion of HeLa cells (Pegues et al., 1995), induction of IL-8 secretion from a polarized monolayer of T-84 cells (Hobbie et al., 1997) and cytotoxicity in BMM (Chen et al., 1996; Monack et al., 1996). A SPI2 TTSS is required for efficient replication in HeLa cells and RAW264.7 macrophages (Cirillo et al., 1998; Hensel et al., 1998), as well as mouse virulence (Hensel et al., 1995) and competition with wild-type strains for colonization of the spleen (Shea et al., 1999). The ΔsspH1ΔsspH2 mutant strain was phenotypically similar to its wild-type parent in all of these assays (data not shown).

Oral inoculation of calves with S. typhimurium 14028s results in severe diarrhoea, weight loss, dehydration and 100% mortality at doses of 1010 cfu per animal or above. Infection of calves can thus be used as a model system to study gastroenteritis (Tsolis et al., 1999a; Galyov et al., 1997). SPI1 mutants display severely reduced virulence in calves (decreased diarrhoea and 100% survival at 1010 cfu per calf), whereas SPI2 mutants are only modestly attenuated (less than 15-fold) and can cause a lethal infection at 1010 cfu per calf (Tsolis et al., 1999a). The ability of wild type, ΔsspB and ΔsspH1ΔsspH2 to colonize the bovine intestine was determined in competitive infection experiments performed with wild-type strain IR715 (Table 3). The ΔsspH1ΔsspH2 mutant colonized intestinal tissues at levels similar to that of the parent strain (P > 0.05). In contrast, a ΔsspB mutant was defective for colonization of intestinal tissues (P < 0.01), as has been shown previously for Salmonella dublin (Galyov et al., 1997). We conclude that mutations in sspH1 and sspH2 do not reduce the ability of S. typhimurium to invade and colonize the intestinal mucosa of calves.

Table 3. . Infection of calves with Salmonella. NT, not testeda. Results of two experiments are shown. In each experiment, two calves were inoculated at 1010 cfu per calf. Wild type, ΔsspH1ΔsspH2 and ΔsspB were inoculated in one experiment and wild type, ΔsspH1,and ΔsspH2 were inoculated in a second experiment.b. Faeces were inspected daily and consistency was scored qualitatively. Calves that developed severe diarrhoea (faecal scores consistently 3–4) in the course of the infection are listed as (+), those that did not (faecal scores consistently 1–2) are considered (–).c. Competitive colonization with strain IR715 (wild type) inoculated at 1 × 109 and at a ratio of 1:1 with strain listed. Bacteria were recovered from infected organs 4 days post infection. Data shown are the mean from four calves (cfu strain/cfu IR715). ML, mesenteric lymph node; PP, ileal Peyer's patches.d. The ratio for ΔsspH1ΔsspH2 is not significantly different (P > 0.05) from that of the wild type, whereas the ratio for ΔsspB is significantly different (P < 0.01) from that of the wild type. Significance of differences between mutant and wild-type strains were determined using Student's t-test after logarithmic transformation of the data.Thumbnail image of

In oral virulence studies, calves inoculated with either wild type, ΔsspH1, or ΔsspH2 strains developed severe diarrhoea and died 1–4 days post infection (Table 3). In contrast, infection with the ΔsspH1ΔsspH2 mutant did not result in mortality, despite the development of severe diarrhoea. These results suggest either synergistic or redundant effects of sspH1 and sspH2 on S. typhimurium infection in calves. As a control, infection with the ΔsspB mutant was determined to cause markedly reduced diarrhoea and was not lethal (Table 3). For calves that succumbed to the infection, gross pathology was recorded at necropsy, and histological sections of intestinal lesions were examined. Calves infected with wild type and ΔsspH1 exhibited acute profound fibrinopurulent necrotizing enteritis in ileal Peyer's patches and ileal villous intestine (data not shown). In contrast, calves infected with ΔsspH2 showed a reduced severity of intestinal lesions, with moderate or patchy fibrinopurulent ileitis in ileal Peyer's patches and the terminal ileum (data not shown). Previous experiments (Tsolis et al., 1999a) indicate that intestinal sections of calves infected at this dose with wild-type S. typhimurium consistently exhibit acute profound pathology at necropsy. Mutations in the SPI2 TTSS, on the other hand, reduced the severity of intestinal lesions (Tsolis et al., 1999a) to a level similar to that observed for the ΔsspH2 mutant. These results indicate that sspH1 and sspH2 contribute to virulence of S. typhimurium 14028s in calves.


The LPX repeat family of type III secreted proteins is an intriguing group of proteins because of the presence of the LRR domain and because of the multicopy nature of these genes within Salmonella, Yersinia and Shigella. S. flexneri encodes five copies of ipaH genes on the virulence plasmid (Venkatesan et al., 1991) and may contain five additional genes with similarity to ipaH on the chromosome (Buysse et al., 1995). One of the IpaH proteins has been shown to be secreted by the mxi/spa TTSS, and expression of all five ipaH genes is induced to varying extents after contact with host cells (Demers et al., 1998). YopM, encoded by the Yersinia virulence plasmid, is required for virulence (Leung et al., 1990) and recent work has provided evidence that YopM has access to the nucleus after translocation by the ysc TTSS (Skrzypek et al., 1998). Although YopM was the first LPX protein identified, it appears to be an outlier in the family because it lacks the C-terminal domain present in most other LPX gene products. In addition to yopM (Boland et al., 1996; Perry et al., 1998), Y. pestis contains at least three more LPX repeat genes (Fig. 1), which are present in an operon-like structure in the chromosome (The Sanger Center). Another LPX repeat gene, y4fR, has also been found in Rhizobium located near genes encoding TTSS elements (Freiberg et al., 1997). This work presents SspH1 and SspH2, two LPX repeat proteins in Salmonella typhimurium 14028s.

Because of the multicopy nature of LPX genes, we examined the distribution of sspH-like sequences in different Salmonella serotypes and found that, whereas most Salmonella serotypes contain only one sspH hybridizing sequence, S. arizonae has at least four hybridizing bands. The fact that sequences similar to sspH2 are conserved in almost all Salmonella serotypes tested suggests that this gene serves a central role in the virulence strategies of salmonellae. The presence of sspH1 in only one isolate of Salmonella typhimurium indicates that this gene may be a recent acquisition and may serve a more specialized function. This is one of the few virulence genes that is not widely conserved among different S. typhimurium isolates, and it will be interesting to determine whether sspH1 confers a novel phenotype on S. typhimurium 14028s. A second band hybridizing to probes created from the sspH2 domain A was detected in the chromosomal DNA of several Salmonella serotypes. Because the data presented here indicate that the N- terminus of sspH2 encodes the secretion/translocation signal, this sspH2 domain A hybridizing band may represent another SPI2-translocated effector protein. Preliminary sequencing data indicate that this sequence encodes a 322 codon ORF that is 92% identical to residues 1–132 of SspH2, but has no similarity to LPX repeats or the C-terminal domain. SlrP, another LPX family member in S. typhimurium, has also recently been identified by signature-tagged mutagenesis in a screen for murine virulence factors (Tsolis et al., 1999b). Hybridization data indicate that slrP and the sspH genes do not cross- hybridize (data not shown). As slrP participates in murine virulence, whereas sspH1 and sspH2 are involved in bovine virulence, these proteins may serve different functions in Salmonella pathogenesis. It seems likely that SlrP, similarly to SspH1 and SspH2, is a putative type III-translocated effector protein.

The virulence defect of the ΔsspH1ΔsspH2 mutant in calves indicates that these two genes are virulence factors, but does not provide insight into the function of the two proteins. As the ΔsspH1 and ΔsspH2 strains are both virulent in calves, the two genes may serve additive or redundant functions. Although the ΔsspH2 mutant caused a lethal infection, the associated intestinal lesions were not as severe as those seen in wild type-infected calves, and more closely resemble those caused by SPI2 mutants. This suggests that the minor virulence defect of SPI2 mutants in the calf infection model is attributable in part to an inability to translocate SspH2. Calves infected with ΔsspH1ΔsspH2 developed severe diarrhoea, similar to calves infected with wild-type S. typhimurium, but were able to recover from the disease. This suggests that sspH1 and sspH2 may be important for permitting S. typhimurium to persist in the host, rather than causing fluid efflux and diarrhoea.

In order to determine whether SspH1 and SspH2 are translocated proteins of the SPI1 or SPI2 TTSS, regulation of transcription and the presence of translocation signals were examined. Transcription of sspH1 was maximal in the late logarithmic phase of growth, concomitant with maximal SPI1 gene expression. This expression was partially dependent upon sirA but was hilA-independent. sspH1 expression was not greatly repressed or induced after exposure to the intracellular environment. This indicates that sspH1 expression levels are similar in the intracellular and extracellular environments, and that SspH1 is available for translocation by both the SPI1 and SPI2 TTSS. In agreement with this hypothesis, results of the experiments with SspH1–CyaA are consistent with the presence of secretion/translocation signals that are recognized by both the SPI1 and SPI2 TTSS. Thus, SspH1 is a translocated protein of both the SPI1 and the SPI2 TTSS. It would be interesting to determine if other SPI1 TTSS-translocated proteins are also targets of the SPI2 TTSS. In contrast, sspH2 transcription is induced in the intracellular environment dependent upon ssrA/ssrB, the two-component regulatory system required for SPI2 gene expression. Likewise, the results of experiments utilizing the SspH2–CyaA fusion proteins suggest that translocation signals are only efficiently recognized by the SPI2 TTSS, making SspH2 a SPI2-specific translocated protein. The low level of SPI1 TTSS-dependent SspH2–CyaA secretion is probably not physiologically relevant because no significant SPI1 TTSS-dependent translocation was detected, but does indicate that SspH2 contains secretion signals that can be recognized by the SPI1 TTSS, albeit at much lower efficiency than the secretion signals present in SspH1. It should be noted that secretion in culture by the SPI2 TTSS has not yet been demonstrated, and these experiments are consistent with the absence of SPI2 TTSS secretion under these experimental conditions.

The secretion/translocation signals have been mapped in many translocated proteins to the amino terminal regions of the proteins (Sory et al., 1995; Schesser et al., 1996). This roughly corresponds to the more divergent domain A in SspH1 and SspH2, and it is likely that the differences in their translocation profiles are attributable to the differences found in this domain. The dichotomy between the translocation and expression profiles of SspH1 and SspH2 is interesting. It indicates that some secretion/translocation signals are recognized by both TTSS, whereas others are more specifically targeted to only one of the two TTSS. It also may indicate that the two proteins serve different functions, or at the very least, that they serve their respective functions at different times during the course of the bacterial life cycle (intracellular versus extracellular).

The sequence of the proteins allows speculation about how SspH1 and SspH2 may function to promote virulence. The presence of the LPX repeat domain in SspH1 and SspH2 should prove highly significant, as it is probably that, like other LRRs, it functions as a protein-binding motif. The C-terminal domain is the second striking feature of the LPX family of type III secreted proteins. Although YopM lacks this domain, it is conserved in the IpaH and SspH families as well as SlrP, y4fR and two of the three LPX-containing ORFs detected in the Y. pestis genome sequence. In addition, two of the five sequenced ipaH genes lack LPX repeats and only encode the C-terminal domain. We hypothesize that the C-terminal domain has a function that can be directed to specific target proteins or subcellular locations determined by the affinity of the LPX repeat domain for its ligand. This may account for the presence of multiple copies of similar LPX repeat genes in Salmonella, Shigella and Yersinia. The different copies, with only slight modification of the side chains presented in the LPX repeats, could interact with different host proteins and thus serve completely different functions in pathogenesis. Indeed, mutation of a single side chain in the LRR of polygalacturonase-inhibiting protein has been shown to alter the ligand specificity (Leckie et al., 1999).

In conclusion, we have identified two SPI2-translocated effector proteins. SspH2 is translocated by and co- ordinately regulated with the SPI2 TTSS, whereas SspH1 is expressed intracellularly and extracellularly and can be translocated by both the SPI1 and the SPI2 TTSS. The importance of these genes in Salmonella virulence is indicated by the virulence defect of ΔsspH1ΔsspH2 in calves. The sheer number of LPX proteins that are being identified highlights the importance and probable versatility of this class of proteins.

Experimental procedures

Bacterial strains, eukaryotic cell lines and growth conditions

Bacterial strains and plasmids used are listed in Table 4. Bacteria were grown and HeLa cells were cultured as has been previously described (Rakeman et al., 1999). RAW264.7 cells, a murine macrophage-like cell line, were grown in RPMI 1640 + 10% fetal calf serum.

Infection of cultured eukaryotic cells

Cultured eukaryotic cells were infected with stationary phase bacteria (overnight culture) or late logarithmic phase bacteria (overnight bacteria back diluted 1:50 or 1:100 and grown for 3 or 4 h). All strains used in luciferase and cyclase assays entered and persisted at levels similar to the wild type. Cytotoxicity was negligible in all experiments. RAW264.7 cells were plated in 24-well plates at 5 × 105 cells well−1 (cyclase assays) or six-well plates at 1.25 × 107 cells well−1 (luciferase assays) and incubated at 37°C in 5% CO2 overnight. They were then infected with bacteria in RPMI 1640 + 10% FBS for 1 h, washed three times with PBS and either lysed or treated for 5 h with 15 μg ml−1 gentamycin, washed with PBS and lysed. Lysis was performed in appropriate buffers as described below. Cytotoxicity was detected by measuring lactate dehydrogenase activity in the culture supernatant, using the Cytotox 96 Assay from Promega.

DNA techniques

Chromosomal DNA was isolated as has been described without the final phenol extraction step (Rakeman et al., 1999). Southern blots were performed as has been described (Johnston et al., 1996). PCR was performed according to the protocol given by New England Biolabs and Stratagene for Vent or Pfu DNA polymerases. DNA sequence analysis was performed as has been described (Gunn et al., 1998) or with a Perkin-Elmer ABI Prism 377 Automated DNA Sequencer and the sequencher 3.0 program. sspH1 was sequenced from pCS01. sspH2 was sequenced from pEM12 and various subclones. Chromosomal map positions were determined using Southern blot hybridization of bacteriophage DNA isolated from a set of Mud-P22 insertions as has been previously described (Benson and Goldman, 1992).

Protein analysis

Secreted proteins were purified as has been described (Pegues et al., 1995). Late logarithmic (3 h growth of 1:50 dilution of overnight culture) or late stationary (overnight culture) phase bacteria were washed with saline, resuspended in sample buffer and boiled for cellular protein fractions. SDS–PAGE and Western blot techniques were performed as has been described (Pegues et al., 1995). GST–SspH1 was purified by binding bacterial lysates to glutathione agarose beads (50% slurry in PBS), washing with PBS and eluting (5 mM glutathione, 50 mM Tris pH 8). Eluted protein was gel purified and injected into rabbits at Pocono Rabbit Farm and Laboratories. Antisera was blot-affinity purified with GST–SspH1 as has been described (Tang, 1993). Monoclonal antibody 3D1, specific for CyaA, was kindly provided by E. Hewlett.

Plasmid and strain construction

Bacterial strains were constructed by P22HT int transduction as has been described (Davis et al., 1980). pCS01 was constructed by digesting pBB05BJ with BglII and SacI, and cloning into pBSKS digested with SacI and BamHI. pEM12 was created by inserting a 5.8 kb size selected HindIII digestion of S. typhimurium chromosomal DNA into the HindIII site of pBSIISK+ and screening for inserts that hybridize to sspH1 probes in Southern blots. pGPLFR03 was constructed by inserting the renilla luciferase gene from pRL-null (Promega) under expression from the tet promoter (from pBR322) into the blunted BamH1 site closest to the origin of pGPL01. Transcription from the tet promoter was terminated with the rrnBT1T2 transcriptional terminator from pBAD18 (Guzman et al., 1995) downstream from the renilla luciferase gene. DNA encoding the N-terminal 360 amino acids of SspH1 and DNA 718 bp upstream from the sspH1 ATG was PCR amplified and cloned into pGPLFR03 in order to create pEM47, the sspH1::f-luc transcriptional fusion vector. The sspH2::f-luc transcriptional fusion vector pEM48 was created using PCR amplification of DNA spanning the HincII site ≈ 1.2 kb upstream of sspH2 to the second codon of sspH2, and ligation of this fragment into pGPLFR03. These pir-dependent plasmids were integrated into the CS019 chromosome. The SspH1 expression plasmid, pEM41, was created by PCR amplification of sequences encoding the entire sspH1 ORF along with the putative ribosome binding site, and ligation of the product into pWSK29 such that it is transcribed from the lac promoter. The GST–SspH1 fusion protein was produced from plasmid pEM02, which was created by PCR amplification of DNA sequences containing codons 190–700 and ligation of the product into pGEX2T (Pharmacia). Plasmids pEM25 and pEM30 express protein fusions of sspH1 and sspH2 to the catalytic domain of cyaA. pMJH20 is the parent vector for pEM25 and pEM30. DNA sequences containing codons 2–406 of cyaA were PCR amplified from pACT7 (Sebo et al., 1991) with primers engineered to create a 5′SmaI site and 3′ stop codon and EcoRI site. This product was ligated into pWSK29 digested with SmaI and EcoRI to create pMJH20. pEM25 and pEM30 were created using PCR amplification of DNA encoding the putative ribosome binding site, start codon and N-terminal 208 or 214 codons of sspH1 or sspH2, and inserting the products into pMJH20 digested with SmaI and SacI. Resulting SspH–CyaA fusion proteins are expressed from the lac promoter and the sspH native ribosome binding site.

Luciferase enzyme assays

Bacteria or RAW264.7 macrophage cells infected with bacteria at an MOI of 5 were suspended in 500 μl of luciferase buffer (Gunn and Miller, 1996), frozen, thawed and lysed by sonication for 20 s. Ten microlitres (Table 1) or 20 μl (Table 2) of this lysate was assayed for luciferase activity, using the Dual Luciferase Reporter Assay System from Promega. Firefly and renilla luciferase units were recorded sequentially for 10 s (Table 1, f-luc) or 30 s (Table 2, f-luc) and 10 s (r-luc) respectively, in a Berthold LB9501 luminometer (100 μl injector for f-luc readings, 100 μl manual injection for r-luc readings). Firefly luciferase relative light units were divided by renilla luciferase relative light units to normalize for cell lysis and cell number. Assays were performed in triplicate wells and repeated at least two times. For low- magnesium growth, bacteria were grown overnight in LB, washed with 8 μM MgCl2 N-minimal media with glycerol as a carbon source, then back diluted 1:100 in the same media in triplicate and grown on a roller drum at 37°C for 16 h before samples were taken for luciferase enzyme assays.

cAMP assays

RAW264.7 cells were infected with either late logarithmic or stationary phase bacteria at an MOI of 5. Lysis of infected eukaryotic cells in denaturing conditions was performed in lysis buffer 1 (supplied with cAMP EIA kit) supplemented with 100 mM HCl, after which samples were boiled and neutralized with NaOH before cAMP quantification, using the cAMP Enzyme Immunoassay System (Amersham Pharmacia Biotech). Assays were performed in triplicate wells and repeated at least two times. SspH1–CyaA from EM176 was concentrated and added to infections as follows. Twenty-four millilitres of overnight culture were centrifuged in a Beckman L8–60 M ultracentrifuge (SW41 rotor) at 30 000 r.p.m. for 1 h and 50 min to pellet bacteria. Twenty millilitres of the supernatant was concentrated using a Millipore ultrafree spin concentrator with a 5 kDa molecular weight cut-off, washed with 15 ml of Dulbecco's modified Eagle medium (DMEM) and concentrated again to a volume of ≈ 500 μl. A volume of concentrated supernatant equivalent to the volume of culture containing 100 cfu was added for every one wild-type (CS401) bacteria (MOI 5) in infections of eukaryotic cells. In order to assay the presence of adenylate cyclase activity, infected cells were washed with PBS, treated with typsin (Gibco-BRL) for 10 min at 37°C to degrade extracellular CyaA, washed with PBS again, resuspended in 100 μl of water and sonicated briefly to lyse the macrophages. This lysate was tested for adenylate cyclase activity by the addition of an equal volume of 2 × CyaA reaction buffer [100 mM Tris, 4 mM ATP, 12 mM MgCl2, 200 μg ml−1 BSA (bovine serum albumin), 0.24 mM CaCl2, 0.2 μM calmodulin (Sigma)]. Protein content was determined using the Coomassie Plus Protein Assay Reagent (Pierce).

Construction of strains deleted for sspH1 and/or sspH2

In frame deletions of greater than 95% of the coding sequence of both genes were created using PCR amplification of flanking DNA with Vent, and subsequently cloned into the allelic exchange vector, pKAS32 (Skorupski and Taylor, 1996). pEM28 contains the ΔsspH1 construct and pEM34 contains the ΔsspH2 construct. Allelic exchange was performed in strain CS401 as has been described (Rakeman et al., 1999). Strains carrying ΔsspH1, ΔsspH2 or both ΔsspH1 and ΔsspH2 were verified using PCR and Southern blot analysis.

Calf virulence assessment

Calf infections were performed as has been described (Tsolis et al., 1999a). Briefly, calves were orally infected at 1010 cfu per calf for virulence assays. Calves that were unable to stand or feed were euthanized. Fecal scores were determined daily and scored from 1 to 4 as follows: 1, normal faeces; 2, soft faeces with loss of distinct conformation; 3, loose faeces with reduced solid matter; 4, running or watery faeces with markedly reduced or little solid matter, blood or shreds of fibrinous matter. Competitive infections with IR715 were performed using oral co-infection of four calves with ≈ 1 × 109 cfu of each strain per calf. After four days calves were sacrificed, tissues were homogenized in PBS and colony forming units from tissues were obtained by dilutional plating with appropriate antibiotics. A Student's t-test to determine whether input and output ratios differed significantly was performed after logarithmic transformation of these data.


We gratefully thank the following people: John S. Gunn and Bill Belden for initial sequencing of sspH1; Michael J. Hantman for construction of the CyaA fusion plasmid, pMJH20, and for critical review of this manuscript; John S. Gunn and Tina Guina for assistance in the construction of pGPLFR03; Andrew Gewirtz and James Madara for examining the induction of IL-8 secretion by sspH mutants; Erik Hewlett for the gift of monoclonal antibodies recognizing CyaA and David Holden for providing SPI2 mutant strains. This work was supported by the Poncin Scholarship Fund (E.M.), RO1 AI30479 (S.M.) and AI09312 (C.S.) from the National Institutes of Health, grants 9800465 (A.B. and L.G.A.) and 9702568 (R.T.) from the USDA/NRICGP, and grant AI40124 (A.B.) from the Public Health Service.