Macrophage-dependent induction of the Salmonella pathogenicity island 2 type III secretion system and its role in intracellular survival


  • Daniela Maria Cirillo,

    1. Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94305, USA.
    Search for more papers by this author
    • Present address: D. M. Cirillo, Dip. di Microbiologia Clinica, Ospedale San Giovanni Battista, 10126 Torino, Italy

    • These authors contributed equally to this study

  • Raphael H. Valdivia,

    1. Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94305, USA.
    Search for more papers by this author
    • Present address: R. H. Valdivia Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, CA 94720, USA

    • These authors contributed equally to this study

  • Denise M. Monack,

    1. Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94305, USA.
    Search for more papers by this author
  • Stanley Falkow

    1. Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94305, USA.
    Search for more papers by this author

Stanley Falkow Tel. (650) 723 9187; Fax (650) 723 7282.


Salmonella pathogenicity island 2 (SPI-2) encodes a putative type III secretion system necessary for systemic infection in animals. We have investigated the transcriptional organization and regulation of SPI-2 by creating gfp fusions throughout the entire gene cluster. These gfp fusions demonstrated that SPI-2 genes encoding structural, regulatory and previously uncharacterized putative secreted proteins are preferentially expressed in the intracellular environment of the host macrophage. Furthermore, the transcription of these genes within host cells was dependent on the two-component regulatory system SsrA/SsrB and an acidic phagosomal environment. Most SPI-2 mutants failed to replicate to the same level as wild-type strains in murine macrophages and human epithelial cells. In orally infected mice, SPI-2 mutants colonized the Peyer's patches but did not progress to the mesenteric lymph nodes. We conclude that SPI-2 genes are specifically expressed upon entry into mammalian cells and are required for intracellular growth in host cells in vivo and in vitro.


Salmonella typhimurium (Salmonella enterica serovar Typhimurium) is a facultative intracellular pathogen that infects its hosts by the faecal–oral route. In infected mice, S. typhimurium colonizes the Peyer's patches, migrates to the mesenteric lymph nodes and disseminates to the spleen and liver (reviewed in Gulig, 1996; Jones and Falkow, 1996). Pharmacological and immunofluorescence microscopy experiments indicate that S. typhimurium resides within intracellular compartments in these organs (Dunlap et al., 1991; Richter-Dahlfors et al., 1997). The ability of S. typhimurium to survive and replicate in host cells, including macrophages, is essential for virulence because mutants that fail to replicate in cultured cell lines are avirulent in animals (Fields et al., 1986; Leung and Finlay, 1991). Bacterial survival within macrophages is dependent on the two-component regulatory system PhoP/PhoQ (Groisman et al., 1989; Miller et al., 1989; Belden and Miller, 1994). PhoP/PhoQ-regulated genes encode proteins with functions as diverse as lipopolysaccharide (LPS) modification, resistance to antimicrobial peptides and nutrient scavenging (Garcia-Vescovi et al., 1996; Gunn and Miller, 1996; Soncini and Groisman, 1996; Guo et al., 1997). Other macrophage resistance factors are involved in responses to stress, detoxification of reactive oxygen radicals and DNA repair (Buchmeier and Heffron, 1990; Abshire and Neidhardt, 1993; Buchmeier et al., 1993).

Recently, two laboratories have identified a cluster of virulence genes at 30.7 centisomes on the S. typhimurium chromosome that encodes the structural components of a putative type III secretion system (Ochman et al., 1996; Shea et al., 1996). S. typhimurium is equipped with two independent type III secretion systems encoded by two large virulence gene clusters: Salmonella pathogenicity island 1 (SPI-1), which is required for the invasion of mammalian cells, and Salmonella pathogenicity island 2 (SPI-2), which is necessary for bacterial survival in animals (reviewed in Finlay and Falkow, 1997; Groisman and Ochman, 1997). Type III secretion systems are used by many bacterial pathogens to deliver into host cells virulence proteins that interfere with the host cell's normal signalling pathways (Galan, 1996; Mecsas and Strauss, 1996; He, 1997). For example, the secreted molecules encoded by the Shigella spp. virulence plasmid and Salmonella SPI-1 induce cytoskeletal rearrangements in the host cell that lead to bacterial internalization, whereas the secreted molecules of Yersinia spp. (Yop) prevent bacterial uptake by interfering with signalling required for receptor-mediated endocytosis and phagocytosis (Galan, 1996; Galan and Bliska, 1996; Cornelis and Wolf-Watz, 1997). Less is known about the role of SPI-2's type III secretion system because the genes encoding secreted effector proteins have not been described. Furthermore, the role of SPI-2 in Salmonella pathogenesis has remained unclear because previous reports have disagreed about the phenotype of SPI-2 mutants in macrophages (Hensel et al., 1997a; Ochman et al., 1996).

We have identified recently a promoter within SPI-2 that is preferentially expressed in the intracellular environment of mammalian cells (Valdivia and Falkow, 1997). To study the transcriptional regulation of SPI-2 within infected cells, we cloned overlapping SPI-2 DNA fragments into three plasmids and generated transcriptional gfp fusions to all putative open reading frames (ORFs). As a result, we sequenced, characterized and generated mutations in a region of SPI-2 encoding proteins homologous to factors secreted by type III secretion systems in other bacterial pathogens. We report here that these SPI-2 mutants were unable replicate within mammalian cells and failed to disseminate through the lymphatic system during oral infections. These results suggest an essential role for SPI-2 in bacterial survival in mammalian cells during in vivo infection.


Isolation and characterization of SPI-2 genes encoding a putative type III secretion system

We have identified previously a SPI-2 gene, ssaH, that is preferentially expressed within infected macrophages (Valdivia and Falkow, 1997). To determine whether the expression of genes encoding other SPI-2 components is induced in the intracellular environment of host cells, we isolated two overlapping SPI-2 fragments that span the regions 5′ and 3′ of ssaH (see Experimental procedures). pMIC-10 A contains a 7.1 kb fragment with the type III secretion structural genes ssaHI (Valdivia, and Falkow, 1997), ssaJKLMV (Hensel et al., 1997a) (Fig. 1). Hensel et al. (1997b) have shown that ssaKLMV is transcribed as part of a large operon that encodes some of the structural components of the type III secretion pore. pMIC10B contains a 7.2 kb fragment that covers the 3′ end of spiB (Ochman et al., 1996) and the 5′ end of ssaH. This fragment contains 10 novel ORFs encoding putative proteins with significant homology to chaperones and secreted proteins from other type III secretory systems. For consistency with previous work (Hensel et al., 1997a), SPI-2 genes encoding the putative secretion apparatus will be referred to as ssa (secretion system apparatus), those encoding putative secreted proteins will be referred as sse (secretion system effector) and those encoding chaperones will be referred as ssc (secretion system chaperone). The arrangement of these genes is shown in Fig. 1. Sequence analysis of sse and ssc revealed the following:

Figure 1.

. Schematic representation of SPI-2 genes cloned in pMIC10A-C plasmids. Three different DNA fragments containing genes encoding structural (MIC 10A and MIC 10C), secretory (MIC 10B) and regulatory (MIC 10C) components of SPI-2 type III secretion system were cloned in plasmids. mTn5gfp insertion (arrows) in pMIC10A-C plasmids were mapped, transformed into S. typhimurium and tested for fluorescence induction within infected RAW 264.7 macrophages. The direction of the arrows indicate polarity of the mTn5gfp insertion. Black and white arrows represent insertions in genes with macrophage-dependent and macrophage-independent activity respectively. Grey boxes indicate genes encoding structural components of the secretion apparatus, black boxes indicate genes with regulatory function, white boxes indicate genes encoding putative secreted proteins and striped boxes indicate genes encoding putative chaperones.

ssaE encodes a putative 80-amino-acid (aa) polypeptide with 33% identity to YscE from Yersinia spp. YscE is required for Yop secretion from Yersinia spp., although it is not clear whether YscE itself is secreted (Allaoui et al., 1995). The amino terminus of SsaE is also homologous to an internal region of IpaD from S. flexneri (31% aa identity over 32 aa span) that is secreted by the type III secretion system of S. flexneri. IpaD and IpaB associate with the bacterial membrane and modulate protein secretion (Menard et al., 1996).

sseA encodes a putative 108 aa polypeptide. There are no proteins with significant homologies to SseA in the available protein databases. sseB encodes a 196 aa polypeptide with 50% similarity and 29% identity to EspA from enteropathogenic Escherichia coli (EPEC). EspA is secreted by EPEC to form extracellular appendages on the bacterial surface and is necessary for bacterial-mediated induction of host cell signalling (Knutton et al., 1998).

sscA encodes a 157 aa polypeptide with 46% similarity and 26% identity to LcrH (SycD) from Yersinia spp. that modulates the expression and secretion of YopBD (Price and Straley, 1989; Wattiau et al., 1994). SscC is also homologous to PcrH (Pseudomonas aeruginosa), SicA (S. typhimurium) and IpgC (S. flexneri ), which have been postulated to function as chaperones of secreted proteins in other type III secretion systems (Wattiau et al., 1996; Yahr et al., 1997).

sseC encodes a 484 aa protein with three putative transmembrane regions. This protein is homologous to YopB from Yersinia spp. (26% identity over 133 aa) and to EspD from EPEC (27% identity over 97 aa). YopB and EspD are secreted proteins, which are required for Yop delivery into host cells and EPEC attachment and effacement respectively (Hakansson et al., 1993; Foubister et al., 1994; Lai et al., 1997).

sseD encodes a 195 aa polypeptide with significant homology to EspB from attaching and effacing E. coli. The N-terminus of SseD is 56% similar (26% identical) to EspB, whereas the C-terminus is less conserved (22% identity, 45% similarity). EspB is secreted into mammalian cells and is required for host cell signalling (Foubister et al., 1994; Wolff et al., 1998).

sseE encodes a 138 aa protein. The N-terminus of SseE is homologous to LcrR from Yersinia spp. (29% identity and 46% similarity over a 47 aa region). LcrR regulates protein secretion in Yersinia spp. (Barve and Straley, 1990).

sscB encodes a 144 aa polypeptide with homology to secreted protein chaperones such as PcrH, LcrH, SicA and IpgC (see above). SscB and IpgC are 45% similar (23% identical) over a 114 aa acid span. SscA and SscB, the two putative chaperones in SPI-2, are 29% similar over their entire length.

sseF and sseG encode a 260 aa polypeptide and a 229 aa polypeptide respectively. SseF is 31% similar (24% identical) to SseG over a 157 aa span. Neither protein has any significant homologues in the protein databases.

To obtain the SPI-2 region 5′ of pMIC-10B, a 2.2 kb DNA fragment containing the promoter regions for the genes encoding the putative two-component regulatory system SsrA/SsrB (SpiR) and the structural genes spiCAB (Ochman et al., 1996) was amplified from S. typhimurium DNA by PCR and inserted into pBR322 to produce plasmid pMIC-10C (Fig. 1C). Although this region of SPI-2 has been previously described by two independent groups (Ochman et al., 1996; Shea et al., 1996), little is known about the operon structure and regulation of this region.

Isolation of transcriptional gfp fusions to SPI-2 genes

We monitored the expression of different SPI-2 genes within infected cells using random transcriptional gfp gene fusions (gfpmut3 allele) (Cormack et al., 1996) to SPI-2 genes cloned in plasmids pMIC10A-C. In brief, a suicide plasmid containing a mini Tn5gfp (KanR) cassette was conjugated into E. coli DH12S (StrR) bearing plasmids pMIC-10 A, pMIC-10B or pMIC-10C. mTn5gfp insertions in the pMIC plasmids were mapped and the insertion juncture sequenced with a primer specific for the 5′ end of gfp. We isolated 13 mTn5gfp insertions in pMIC-10A and eight mTn5gfp insertions in MIC-10C, creating gfp fusions to ssaJ, ssaK, ssaL, ssaV, ssrA, spiC and spiA. As the transcriptional arrangement of SPI-2 genes encoding putative targets of the secretion apparatus has not been previously described, we isolated 27 non-overlapping mTn5gfp insertions in pMIC-10B. Figure 1 shows the relative map location of the mTn5gfp insertions in pMIC10A-C.

SPI-2 genes are preferentially expressed within macrophages

Plasmids bearing different gfp fusions to SPI-2 genes were mobilized into S. typhimurium strain SL1344 and the derivatives used to infect RAW 264.7 macrophages. Induction of the gfp gene fusions was monitored by flow-cytometric measurement of fluorescence from intracellular bacteria and compared with fluorescence from extracellular bacteria (see Experimental procedures). Intracellular-dependent induction of gfp fusions in pMIC10A-C suggested that SPI-2 genes are arranged in at least four different operons. The genes in these operons cluster according to their predicted function and thus will be referred to as regulatory, structural I, secretory and structural II operons. The SPI-2 region in pMIC-10A codes for structural (structural II) components of the secretion apparatus, accessory factors for pore assembly and ATP-dependent secretion (Hensel et al., 1997a). mTn5gfp insertions in ssaJ, ssaK, ssaL and ssaV were induced within host cells, with pMIC-10A64 (ssaJ::gfp) and pMIC-10A41 (ssaV:: gfp) showing a 100- and fivefold induction respectively. In contrast, strains bearing sseG::gfp fusions, pMIC-10A53 and pMIC-10A133, which map 5′ of ssaJ, were not induced intracellularly (Fig. 1A). An ssaH::gfp fusions has been reported to have a ≈400-fold induction in macrophages (Valdivia and Falkow, 1997). This suggests that the promoter that activates the transcription of ssaHIJKLMV maps upstream of ssaH but downstream of sseG.

mTn5gfp insertions in pMIC-10B, which encodes putative secreted proteins (secretory), also indicated the presence of at least one macrophage-inducible promoter that drives the transcription of a large operon (Fig. 1). This promoter maps between ssaE (pMIC-10B216) and the 3′ end of spiB (pMIC-10B26 and pMIC-10B67). These spiB::gfp fusions were not induced within infected cells presumably because they have been dissociated from the promoter that drives transcription of spiCAB (see below). At the present time, we cannot establish whether ssaE is part of the spiCAB operon. However, the partial homology of SsaE to IpaD suggests that ssaE could be part of an operon encoding putative secreted proteins, rather than the structural proteins SpiC, SpiA and SpiB. Further promoter mapping will be required to distinguish between these two possibilities. The last gene fusion that defines the transcriptional unit encoding putative secreted proteins is a mTn5gfp insertion in sseG (pMIC-10B72). This particular fusion is only moderately induced within macrophages (threefold), therefore it is unlikely that any promoters upstream of sseG contribute to the high intracellular induction seen for mTn5gfp insertions in ssaHIJKMLV (Fig. 1).

The SPI-2 region in pMIC-10C contains the promoter regions that drives the expression of structural I (spiCAB) and regulatory (ssrAB) operons. mTn5gfp insertions in these operons indicated that their expression is induced within mammalian cells (Fig. 1C). mTn5gfp insertions in genes encoding the structural components SpiCA (pMIC-10C20) and the regulatory component SsrA/SpiR (pMIC-10C32) showed a macrophage-dependent induction of ≈40-fold and approximately eightfold respectively.

To study the kinetics of expression of different SPI-2 components, we monitored bacterial fluorescence in RAW 264.7 cells infected with SL1344 bearing mTn5gfp insertions in representative SPI-2 genes. To minimize variability in fluorescence from intracellular bacteria at early time points, infections were synchronized by spinning down bacteria onto chilled macrophage monolayers, followed by a 15 min incubation at 37°C. At determined time points, fluorescence from intracellular bacteria released by detergent-lysed macrophages was measured by flow cytometry (see Experimental procedures). These time-course studies (Fig. 2) indicated that gfp fusions to the regulatory operon, pMIC-10C32 (ssrA::gfp), were induced early after entry into cells and reached maximal inductions by ≈2 h. mTn5gfp insertions in structural operons I and II, represented by pMIC-10A64 (ssaJ::gfp) and pMIC-10C20 (spiA::gfp), were highly induced (100-fold and 40-fold respectively) by 6.5 h of infection. mTn5gfp insertions in the 5′ end of the secretory region (pMIC-10B216 and pMIC-10B7) reached a maximal induction of 18-fold and mTn5gfp fusions near the 3′ end of the same region (pMIC-10B19 and pMIC-10B111) were sixfold induced. The decrease in macrophage-dependent induction of gfp fusions distal to the promoter within ssaE could be explained by the attenuation characteristic of the transcription of large operons. Alternatively, it is possible that two or more operons are present within this locus. Further promoter mapping will be required to distinguish between these two possibilities.

Figure 2.

. Kinetics of SPI-2 gene expression. RAW 264.7 cells were infected with S. typhimurium bearing representative gfp fusions (arrows) to genes encoding structural, regulatory and secretory/chaperone components. Bacterial fluorescence was measured using flow cytometry at different times after infection. The relative map location of the gfp fusions tested is shown (not to scale). Fluorescence induction was determined as the ratio of fluorescence of intracellular bacteria to fluorescence of bacteria exposed to DMEM (+10% FCS) alone (see Experimental procedures).

Regulation of SPI-2 gene expression within host cells

Previous work on Yersinia spp. type III secretion systems indicates complex regulatory cascades and feedback loops in which the transcription of genes encoding secreted proteins is dependent on an intact secretion apparatus (Pettersson et al., 1996). We have studied the regulation of SPI-2 genes in mammalian cells, by measuring the macrophage-dependent induction of SPI-2 gfp gene fusions in S. typhimurium strains bearing different mutations in SPI-2 (Fig. 3). Reporter plasmids for SPI-2's structural (pMIC-10A64, pMIC-10A40 and pMIC-10C20), regulatory (pMIC-10C32) and secretory genes (pMIC-10B9, pMIC-10B15 and pMIC-10B7) were mobilized into SL1344 bearing mTn5 Km insertions in either ssaJ (P11D10), ssaN (P9B7), spiA (P7G2), ssrA/spiR (P3F4), or sseB (RVY6). Intracellular induction of gfp fusions to structural and secreted components was abolished only when the assays were performed in strain P3F4 (ssrA::mTn5 Km). The reporter plasmid pMIC-10C32 (ssrA::gfp) did not require a functional SsrA/SsrB for intracellular induction because a fourfold induction was observed in a P3F4 strain background, indicating that although SsrA/SsrB can induce ssrAB expression additional regulatory systems exist. Macrophage-dependent induction of SPI-2 gfp fusions tested in strains P11D10, P9B7, P7G2 and RVY6 was similar to that observed from the wild-type strain SL1344 (Fig. 3, and data not shown). This suggests that structural and secretory components of the type III secretion system do not regulate SPI-2 gene expression.

Figure 3.

. SPI-2 gene expression is dependent on SsrA/SsrB. Plasmids bearing mTn5gfp insertion in representative SPI-2 genes (spiA::gfp, sscB::gfp and ssrA::gfp) were transformed into S. typhimurium strains SL1344 (wild type), RVY6 (sseB::Km) and P3F4 (ssrA::Km). Histograms show bacterial fluorescence from extracellular bacteria (shaded) and intracellular bacteria (solid lines) after a 5 h infection of RAW 264.7 macrophages.

Macrophage-dependent induction of SPI-2 genes requires an acidic phagosomal environment

Our observations indicated that the two-component regulatory system SsrA/SsrB is essential for the regulation of all genes encoding SPI-2's type III secretion system. We have previously shown that expression of an ssaH::gfp fusion was largely independent of the PhoP/PhoQ regulon (Valdivia and Falkow, 1997), suggesting that the stimuli recognized by PhoP/PhoQ and SsrA/SsrB are potentially different. Rathman et al. (1996) reported that S. typhimurium survival in macrophages decreased when the pH of phagosomes was neutralized, indicating that low pH within the vacuolar compartment is important for the activation of genes involved in intracellular survival. To test the possibility that SPI-2 expression is linked to vacuolar pH, we monitored fluorescence induction from intracellular bacteria bearing different SPI-2 gfp fusions in bafilomycin A1-treated RAW 264.7 macrophages. Bafilomycin A1 inhibits vacuolar proton ATPases and thus prevents the acidification of endosomal and lysosomal compartments (Bowman et al., 1988). Expression from pMIC-10A64 (ssaJ::gfp), pMIC-10B7 (sseB::gfp) and pMIC-10C20 (spiA::gfp) was greatly diminished in bafilomycin-treated cells compared with untreated cells (Fig. 4A and not shown). In contrast, pMIC-10C32 (ssrA::gfp) had only a two- to threefold reduction in intracellular induction (Fig. 4B). This is consistent with the observation that ssrAB appears to be under the control of additional regulatory mechanisms (see above). As controls for inhibitor activity, we measured intracellular induction from a PhoP/PhoQ-dependent gfp fusion (pagA::gfp) (Valdivia and Falkow, 1996), which was unaffected by bafilomycin A1 treatment (Fig. 4C), and a previously characterized acid-inducible gfp fusion (aas::gfp) whose expression was impaired in the presence of the drug (Fig. 4D) (Rathman, 1996; Valdivia and Falkow, 1996).

Figure 4.

. SPI-2 gene expression in bafilomycin A1-treated RAW 264.7 cells. Bafilomycin A1-treated and -untreated macrophages were infected with S. typhimurium transformed with reporter plasmids bearing gfp fusions to ssaJ (A), ssrA (B), the pH-dependent aas (C) and the PhoP/PhoQ-dependent pagA (D). After a 5 h infection, intracellular bacteria were released by 1% Triton X-100 treatment and fluorescence was measured by flow cytometry. Red and green lines indicate bacterial fluorescence from bafilomycin A1-treated and -untreated macrophages. Shaded curves (green) indicate fluorescence of extracellular bacteria.

SPI2 mutants are deficient in intracellular growth

As the ability to survive and replicate in macrophages is a crucial feature of Salmonella pathogenesis, we investigated the replication and survival in RAW 264.7 cells of SL1344 strains containing mutations in genes encoding the secretion apparatus (P11D10, P9B7, P7G2) (Hensel et al., 1997a), regulatory components (P3F4) (Hensel et al., 1995) and putative secreted proteins (RVY6, RVY7, RVY8) (Fig. 5). The last three mutants were created by allelic exchange between mTn5gfp insertion in pMIC-10B and the S. typhimurium chromosome. As a control, a known macrophage-sensitive strain of S. typhimurium (phoP::Tn10) was included in the assays (Rathman et al., 1996). Bacterial replication and survival within infected cells was determined by the gentamicin protection assay (see Experimental procedures). To avoid S. typhimurium-mediated cytotoxicity, the gentamicin protection assays were performed with bacteria grown to stationary phase, which do not induce apoptosis in macrophages (Monack et al., 1996). The bacteria were opsonized with 20% normal mouse serum to facilitate the uptake by the macrophages. Macrophages were monitored for viability, and no cytotoxic effects were observed during infections under these condition (data not shown). All strains bearing insertions in structural and regulatory components of SPI2 were impaired in intracellular replication and survival as compared with SL1344 (Fig. 5B). Two strains with mutations in the secretory region, RVY6 and RVY7, were also impaired in intracellular growth. In contrast, RVY8, which bears a mTn5 insertion in the 3′ end of the secretory region, showed a minor defect in macrophage survival.

Figure 5.

. Effect of SPI-2 mutations on S. typhimurium survival and replication in mammalian cells. A. Schematic representation of the map location of SPI-2 mutants tested. B. Survival and replication of SPI2 mutants in RAW 264.7 macrophages. RAW 264.7 cells were infected with SL1344, SL1344 bearing a phoP::Tn10 insertion and the indicated SPI2 mutants before addition of gentamicin. Bacteria were opsonized with 20% normal mouse serum. The graph shows the number of gentamicin protected cfu at 2 h (black bars) and at 20 h (grey bars) after infection. C. Survival and replication of SPI-2 mutants in epithelial cells. Hep-2 cells were infected with bacteria grown for optimal invasion (see Experimental procedures). The graph shows the number of gentamicin protected cfu at 2 h (black) and at 20 h (grey). Replication and survival assays were performed in triplicate for each strain during gentamicin-protection assays.

We also tested the ability of strains P3F4, RVY6, RVY7 and RVY8 to replicate in epithelial cells (Hep-2). As observed with macrophage cell lines, strains P3F4, RVY6 and RVY7 were impaired in their ability to replicate within Hep-2 cells, whereas the RVY8 mutant showed only a moderate growth defect (Fig. 5C).

SPI-2 mutants colonize mice but do not spread beyond the Peyer's patches

SPI-2 mutants have been reported to be significantly attenuated after oral or intraperitoneal inoculation of mice (Shea et al., 1996). We investigated the role of SPI-2 at different stages of the infection by inoculating mice intragastrically with strains SL1344, P3F4 and RYV6 (Fig. 6). Stools were collected from each mouse 8 h after infection to monitor the bacterial passage and survival through the gastrointestinal tract. We observed similar numbers of bacteria in the stools of all mice (data not shown). At day 1 and day 3 after infection, caecum, Peyer's patches, mesenteric lymph nodes, spleens and livers were collected. These tissues were homogenized and plated on selective medium to determine the bacterial load within these organs. As expected, bacteria were recovered from the caecal content of all animals in each group. On day 1, comparable number of colony-forming units (cfu) were recovered from the Peyer's patches of mice infected with SL1344 and SPI-2 mutants. On day 3, however, cfu from the Peyer's patches of mice infected with SL1344 were 2–3 log higher than those recovered from SPI-2 mutants. Furthermore, SPI-2 mutants were virtually undetectable in the mesenteric lymph nodes and were not recovered from spleen and liver homogenates of infected mice.

Figure 6.

. Mice colonization by SL1344 and SPI2 mutants after oral inoculation. Groups of four mice were infected by intragastric delivery of SL1344 (□), P3F4 (•) and RVY6 (▵). The colonization of the GALT and internal organs was determined at day 1 (A) and day 3 (B) after inoculation by plating homogenized tissues (caecum, C; Peyer's patch, PP; mesenteric lymph nodes, LN; spleen, S; and liver, L) for viable cfu. No cfu were found on LN, S or L at day 1.


In this paper we describe the transcriptional organization and regulation of SPI-2 genes encoding a putative type III secretion system. We created transcriptional gfp fusions to genes encoding regulatory, secretory and structural components of this secretion apparatus and studied their expression within infected macrophages. Transcriptional units, as defined by mTn5gfp insertions, indicated the presence of an intracellularly induced promoter upstream of ssaH. This promoter is probably responsible for the intracellular-dependent expression observed for ssaJ::gfp, ssaK::gfp and ssaV::gfp fusions. Previous work based on RT-PCR from Luria–Bertani (LB) broth-grown S. typhimurium suggested that ssaJ and ssaK are transcriptionally unlinked because no discernible mRNA transcript was observed that spanned ssaJ and ssaK (Hensel et al., 1997a). Our experimental evidence suggests that, during intracellular growth, ssaHIJKLMV is transcribed as part of a large operon controlled by a promoter immediately upstream of ssaH. Interestingly, in Yersinia spp., the genes encoding homologues of SsaE and SsaH (yscE and YscF ) are adjacent to each other and are transcribed as part of the yscABCDEFGHILJK operon (Michiels et al., 1991). Haddix and Straley (1992) have reported the presence of a strong Ca2+-regulated promoter within yscF, which drives the expression of the yscFGHIJKLM suboperon. ssaH, ssaJ and ssaK show similar gene arrangement to their Yersinia spp. homologues (yscF, yscJ and yscL), supporting the notion that ssaHIJKLMV may be transcribed as part of an operon. Further experiments will be required to determine whether additional macrophage-inducible promoters are present downstream of ssaH. A mTn5gfp insertion in spiA (encoding a putative outer membrane protein) was also preferentially induced in the host macrophage (40-fold induction). This suggests that there are at least two operons encoding structural components of the secretion apparatus: the first one (structural I) includes spiCAB (Ochman et al., 1996) and potentially ssaE, and the second one (structural II) includes ssaJKLMVN-U (Hensel et al., 1997a). Both these operons are rapidly induced upon bacterial entry into the host cell and are continuously expressed throughout infection.

mTn5gfp insertions in pMIC10B permitted the identification of a large transcriptional unit in a previously uncharacterized region of SPI-2. Macrophage-dependent induction from mTn5gfp insertions suggested that these ORFs are transcribed as a large nine-gene operon. The ORFs in this region encode proteins with homology to proteins secreted by type III secretion systems in EPEC, P. aeruginosa and Yersinia spp. (Galan and Bliska, 1996; Mecsas and Strauss, 1996; Donnenberg et al., 1997; Yahr et al., 1997). These secreted proteins often associate with chaperones that prevent incorrect protein folding and degradation before delivery into the host cell (Wattiau et al., 1996). The ‘secretory’ region of SPI-2 encodes two such proteins, SscA and SscB, with homology to the LcrH/IppI/SicA family of type III secretion chaperones. mTn5gfp insertions in the secretory region indicate that genes encoding putative chaperones (ssc) and secreted protein (sse) are preferentially transcribed within macrophages. Although these proteins are required for survival within macrophages, their molecular function is unclear. Given that SseB, SseC, SscD and SseE are homologous to Yersinia spp. and EPEC proteins known to associate with host cells, it is likely that these Salmonella proteins are involved in the formation of extracellular structures and/or are delivered into the host cell. Furthermore, the predicted amino acid composition of SseC and SseD indicates the presence of transmembrane domains, suggesting that these proteins, by analogy to their homologues, may associate with host membranes (Hakansson et al., 1996). As secreted proteins from EPEC (Esp) and Yersinia spp. (Yop) are known to interfere with host cell signal transduction pathways, we speculate that Salmonella delivers proteins into the bacteria-containing vacuole, which may interfere with normal endosomal processes.

In some type III secretion systems, such as the Yersinia spp. Yop delivery system and S. typhimurium flagellar assembly, transcription of genes encoding secreted proteins is dependent on an intact secretion apparatus (Hughes et al., 1993; Pettersson et al., 1996). For example, in Y. enterocolitica, yopE expression is repressed by LcrQ, which is itself a secreted protein. In the absence of physical contact with host cells, YopN blocks the secretion pore, preventing the secretion of LcrQ and, hence, YopE synthesis (Pettersson et al., 1996). This provides the basis through which gene expression, protein synthesis and secretion are coupled to a physical stimulus such as binding to a host cell surface. Our experimental evidence suggests that this is not the case for SPI-2 because mutations in structural and secretory operons did not alter the macrophage-dependent induction of SPI-2 genes. SPI-2 gene expression, however, required the two-component regulatory system SsrA/SsrB (Fig. 3). It can be argued that SPI-2 secretion does not require the sophisticated coupling of gene expression and protein secretion observed in other type III secretion systems, because ‘contact-dependent’ secretion (Galan, 1996) may not be necessary in the vacuolar environment.

How does S. typhimurium sense the vacuolar environment of the host cell? The expression of SPI-2 genes encoding structural and secreted proteins was disrupted in macrophages in which vacuolar acidification was blocked (Fig. 4). We have not been able to reproduce SPI-2 gene induction by exposure to low pH in vitro (data not shown), suggesting that other factors within an acidic vacuole, and not low pH per se, provide the inducing conditions through which SsrA/SsrB activate SPI-2 gene expression. We have observed that bafilomycin A1 treatment of macrophages does not prevent expression of PhoP/PhoQ-dependent genes (data not shown), indicating that SPI-2 genes are independent of this regulatory system. These observations are consistent with experiments reported by Rathman et al. (1996), which show that S. typhimurium loses viability in bafilomycin-treated cells independently of PhoP/PhoQ.

Hensel et al. (1997a) proposed that SPI-2 is not required for survival in macrophages. A potential explanation for the discrepancy between their study and the work presented here, is that Hensel et al. (1997a) performed gentamicin protection experiments with S. typhimurium strains grown under conditions that maximized invasion and cytotoxicity (Monack et al., 1996). Therefore, differences in the intracellular growth between SPI-2 mutants and wild-type strains were probably masked by macrophage death and subsequent exposure of released bacteria to gentamicin. Gentamicin protection assays performed with bacteria grown to stationary phase, which minimizes S. typhimurium cytotoxic effects, showed marked survival differences in mammalian cells between SPI-2 mutants and wild-type S. typhimurium. These results are in agreement with the findings of Ochman et al. (1996), who reported that S. typhimurium bearing a spiA::mTn5 Km mutation did not survive in J774.1 macrophages.

Not all SPI-2 genes were found to be essential for macrophage survival. For example, a strain bearing a mutation in the 3′ end of the secretory region, RVY8, was only partially attenuated. Similar results were obtained when survival assays were performed in Hep-2 cells. Given the observation that ssaH is expressed in a variety of cells types (Valdivia and Falkow, 1997), SPI-2's role in intracellular survival is probably not macrophage specific.

During infection, S. typhimurium first encounters macrophages during colonization of the gastrointestinal-associated lymphoid tissue (GALT) of the host. We hypothesized that if SPI-2 is required for intracellular survival, secretory mutants would be unable to proceed past the GALT during natural infection. Indeed, mouse GALT colonization experiments indicated that SPI-2 mutants associated with Peyer's patches but not the mesenteric lymph nodes. In contrast, wild-type strains rapidly colonized the GALT and subsequently disseminated into the spleens and livers of infected animals.

Interestingly, studies on the phylogenetic distribution of SPI-2 among Salmonella spp. indicate that this region of DNA is only present in S. enterica serovars that cross the intestinal epithelium and survive in the internal organs of their natural hosts (Ochman and Groisman, 1996; Hensel et al., 1997b). S. bongori, which does not possess SPI-2, is unable to survive in macrophages (Groisman and Ochman, 1997). Therefore, we conclude that the acquisition of SPI-2 permitted the expansion of S. enterica pathogenic niche to the intracellular environment of host cells, as deficiencies in the ability of this pathogen to survive intracellularly lead to abortive infection early after initial colonization of the host.

Experimental procedures

Genetic and molecular techniques

Recombinant DNA and genetic manipulations were performed as previously described (Ausubel et al., 1992). Plasmids were routinely mobilized into S. typhimurium by either electroporation, conjugation or P22-HT-mediated transduction. Bacteria were grown in LB broth at 37°C on a rotating wheel. Antibiotics were used at the following concentrations: ampicillin (Amp) 100 μg ml−1, streptomycin (Str) 200 μg ml−1, kanamycin (Kan) 50 μg ml−1. A list of strains and recombinant plasmids used in this study is shown in Table 1.

Table 1. . Bacterial strains and plasmids (Hosieth and Stocker (1981)).Thumbnail image of

pMIC plasmid construction

The isolation of plasmids containing portions of SPI-2 was achieved by recombinational cloning (Valdivia and Falkow, 1997). A 1.1 kb DNA fragment bearing the ssaH promoter (Valdivia and Falkow, 1997) was isolated by PCR, digested with Sau3A and inserted into the BamHI site of pFPV25 (Valdivia and Falkow, 1996). Recombinant plasmids were screened for DNA inserts present in both orientations, resulting in plasmids pFMI-10 and pFMI-10R. These plasmids were used to transform SM10 and conjugated into S. typhimurium strain SL4702R (polA strR) (MacPhee and Beazer, 1975). Exconjugants were selected on Amp and Str. As colE1 plasmids do not replicate in polA backgrounds, selection for AmpR/StrR strains required single cross-over integrations at homologous sites. Total DNA from SL4702R::pFMI-10 and SL4702R::pFMI-10R was isolated, digested with either HindIII or SphI, ligated and used to transform DH12S to AmpR. As these restriction enzymes cut only once within the integrated plasmid, ligation of the restriction fragment ‘captures’ all chromosomal DNA downstream of the integration point to the next restriction site, although deleting gfp. In this way, a 7.1 kb fragment of DNA downstream (pMIC-10A) and a 7.2 kb fragment of DNA upstream of the ssaH promoter were isolated (pMIC-10B). These two fragments cover the region of SPI-2 coding for putative structural and secreted proteins. The region of SPI-2 containing the promoter region of ssrAB and spiCAB (Ochman et al., 1996) was amplified from S. typhimurium DNA by PCR (5′-AAA GGA TCC AGC CAT ACT CGA ATA and 5′-TTT GAA TTC TGT GAT GAG TTT CCG) and the resulting 2.2 kb fragment was inserted into pBR322 as a BamHI–EcoRI fragment to create pMIC-10C.

Construction of gfp fusions

A mini Tn5 Km cassette containing gfpmut3 allele (Cormack et al., 1996) (kind gift from B. Julien, Stanford University) was inserted into the suicide vector pGP704. This pmTn5gfp (KanR AmpR) construct was transformed into strain SM10λpir. Mutagenesis of cloned SPI-2 genes was performed as follows: 0.5 ml of stationary-phase cultures of SM10(pmTn5gfp) (KanR) and DH12S (StrR) bearing the target pMIC plasmid (AmpR) were pelleted, resuspended in 50 μl of LB broth, and spotted (10 μl) on dry L-plates. After 12 h, spots were swabbed onto LB plates supplemented with Amp, Str and Kan. Colonies were pooled and grown in LB-broth for 3–4 h. Plasmid DNA was then isolated and used to transform DH12S to Ampr/Kanr. Insertions in the cloned SPI-2 regions were mapped either by PCR or by restriction enzyme analysis. DNA sequence of the insertion junctures was determined with a primer specific for the 5′ end of gfp (5′-TTC TTC TCC TTT ACT CAT ATG).

Mutant construction by allelic exchange

Mutations in the structural and regulatory regions of SPI-2 were obtained from D. Holden (RMC, London). Mutations in the region of SPI-2 coding for putative secreted proteins were created by allelic exchange of mTn5gfp insertions in pMIC10B with the S. typhimurium chromosome. In brief, P22HT lysates of SL1344 bearing pMIC10B7, pMIC10B111 and pMIC10B19 were used to transduce the plasmids into SL4702R. The resulting strains bearing single cross-over integrated plasmids were grown overnight in the absence of antibiotics, plated to yield single colonies and tested for spontaneous loss of the Ampr marker. Loss of plasmid sequences as a result of a second cross-over event was confirmed by Southern blot analysis. The resulting SPI-2 mutations were then transduced into SL1344 for virulence and gene expression studies. All these insertional mutants are likely to have polar effects, and thus reflect the effect from lack of synthesis of gene products of the entire operon.

Gentamicin protection assays

RAW 264.7 cells (ATTC TIB71) were maintained in DMEM with 10% fetal calf serum (FCS) in a 37°C, 5% CO2 incubator. The gentamicin protection assay was performed as described by Buchmeier and Heffron (1989). Briefly, bacterial strains were grown in L-broth to stationary phase, diluted in phosphate-buffered saline (PBS), opsonized for 30 min in PBS containing 20% normal mouse serum and added to 5 × 105 RAW 264.7 cells seeded in 24-well tissue culture plates at a multiplicity of infection of 5:1 for 45 min. Infected monolayers were then treated with gentamicin (100 μg ml−1) for 2 h and either lysed or further incubated for 20 h in the presence of 10 μg ml−1 of gentamicin. Infected cells were then washed with PBS and lysed with 1% Triton X-100 in PBS and plated on solid media for cfu determination.

Hep-2 (ATCC CCL23) epithelial cells were maintained in RPMI medium supplemented with 10% FCS. Gentamicin protection assays were performed as described above, but infections were performed with bacteria grown under invasive conditions (overnight standing cultures grown at 37°C in L-broth supplemented with 0.3 M NaCl).

Measurement of intracellular gene expression

Flow-cytometric analysis of bacterial gene expression within mammalian cells was performed as previously described (Valdivia and Falkow, 1996). In brief, 105 RAW 264.7 macrophages were infected for 45 min with 5 × 105 SL1344 bearing test reporter plasmids. Excess bacteria were washed off with PBS, and the infected cells incubated with DMEM (10% FCS) for 3–6 h. At this point the supernatants were collected and the infected monolayers were washed four times with PBS. Bacteria were released from infected macrophages by treatment with 1% Triton X-100 for 10 min. Light scatter and fluorescence intensity of 5 × 103 bacterial-sized particles was acquired for the supernatant and Triton X-100-treated fractions. Non-fluorescent particles arising from detergent-lysed macrophages are often observed as a secondary peak in flow-cytometry scans, which can not be resolved from non-fluorescent bacteria (Valdivia and Falkow, 1996). Intracellular induction experiments were repeated at least three times and confirmed by immunofluorescence microscopy (data not shown).

Flow cytometry

All fluorescence measurements were performed with a FACScalibur (Becton Dickinson) cytometer. Bacteria were detected as previously described (Valdivia et al., 1996). Data analysis was performed with the program CELLQUEST (Becton Dickinson).

DNA sequencing

DNA sequences were determined with a 373A DNA Sequencer from Applied Biosystems (Perkin-Elmer). Sequencing reactions were performed with ABI PRISM Dye Terminator Cycle Sequencing Kit. ORFs, deduced amino acid sequence alignments and protein motifs were determined with programs from the Wisconsin GCG package, BLOCK structural motif searches and ALION ( DNA homologies with sequences in the available databases were determined with the program BLAST (National Center for Biotechnology Information at the National Library of Medicine) or FASTA (Wisconsin GCG package). Sequences described here are available from GenBank under accession number AF020808.

Bafilomycin A1 treatment of cells

Host cell vacuolar proton ATPase inhibition was accomplished by addition of 100 nM bafilomycin A1 (Sigma) to RAW 264.7 macrophages 15 min before infection as previously shown (Rathman et al., 1996).

Animal experiments

Eight-week-old female Balb/c mice were used for all animal infection experiments. Mice were not fed 12 h before intragastric inoculation. Bacterial strains were grown to stationary phase, washed in PBS and resuspended at a concentration of 5 × 108 cfu in 200 μl of PBS. On day 1 and day 3 after inoculation, animals were killed. For each animal, Peyer's patches, mesenteric lymph nodes, spleen and liver were removed, homogenized and plated to determine cfu g−1 of tissue. The animal work presented here was approved by the Department of Laboratory Medicine.


  1. Present address: D. M. Cirillo, Dip. di Microbiologia Clinica, Ospedale San Giovanni Battista, 10126 Torino, Italy

  2. Present address: R. H. Valdivia Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, CA 94720, USA

  3. These authors contributed equally to this study


We would like to acknowledge J. Shea, M. Hensel and D. Holden for sharing strains and data before publication (see accompanying manuscript by Hensel et al.). We are grateful to C. Johnson for DNA sequencing support and L. Ramakrishnan, J. Mecsas, N. Salama and T. McDaniel for critical review of this manuscript. This project was supported by NIH grant AI26195 and a contract from Protein Design Labs (Mountain View, CA, USA).