The invasion-associated type III secretion system of Salmonella enterica serovar Typhimurium is necessary for intracellular proliferation and vacuole biogenesis in epithelial cells


  • Present address: Laboratory of Intracellular Parasites, NIAID, NIH, Rocky Mountain Laboratory, Hamilton, MT 59840, USA.

*For correspondence. E-mail; Tel. (+1) 406 363 9292; Fax (+1) 406 363 9380.


Type III secretion systems (TTSS) are used by Gram-negative pathogens to translocate proteins into eukaryotic host cells. Salmonella enterica serovar Typhimurium (S. Typhimurium) has two of these specialized systems, which are encoded on separate Salmonellapathogenicity islands (SPI-1 and SPI-2) and translocate unique sets of effectors. The specific roles of these systems in Salmonella pathogenesis remain undefined, although SPI-1 is required for bacterial invasion of epithelial cells and SPI-2 for survival/replication in phagocytic cells. However, because SPI-1 TTSS mutants are invasion-incompetent, the role of this TTSS in post-invasion processes has not been investigated. In this study, we have used two distinct methods to internalize a non-invasive SPI-1 TTSS mutant (invA) into cultured epithelial cells: (i) co-internalization with wild-type S. Typhimurium (SPI-1-dependent) and (ii) complementation with the Yersinia pseudotuberculosis invasin (inv) gene (SPI-1-independent). In both cases, internalized invA mutants were unable to replicate intracellularly, indicating that SPI-1 effectors are essential for this process and cannot be complemented by wild-type bacteria in the same cell. Analysis of the biogenesis of SCVs showed that vacuoles containing mutant bacteria displayed abnormal maturation that was dependent on the mechanism of entry. Manipulation of Salmonella-containing vacuole (SCV) biogenesis by pharmacologically perturbing membrane trafficking in the host cell increased intracellular replication of wild-type but not mutant S. Typhimurium This demonstrates a previously unknown role for SPI-1 in vacuole biogenesis and intracellular survival in non-phagocytic cells.


Following oral ingestion, Salmonella must cross the intestinal epithelium, and pathogenicity of these bacteria depends on their ability to invade epithelial cells; (Galan and Curtiss, 1989; Clark et al., 1998). Our understanding of this process is largely from in vitro studies using cultured epithelial cells and Salmonella enterica serovar Typhimurium (S. Typhimurium), a common cause of enteritis in humans. Such studies have shown that in non-phagocytic cells, Salmonella direct their own uptake by inducing dramatic membrane ruffles on the surface of the host cell (Francis et al., 1993), a process that is distinct from receptor-mediated phagocytosis (Brumell et al., 1999). Induction of ruffling and bacterial invasion is absolutely dependent upon a type III secretion system (TTSS) encoded by the Salmonella pathogenicity island-1 (SPI-1). A second TTSS, encoded on SPI-2, is induced intracellularly (Cirillo et al., 1998; Pfeifer et al., 1999; Lee et al., 2000) and has been implicated in intracellular survival and replication in phagocytic cells (Ochman et al., 1996; Cirillo et al., 1998; Hensel et al., 1998).

Inside epithelial cells, Salmonella survive and replicate within a unique phagosome, the Salmonella-containing vacuole (SCV). Intracellular replication occurs after a lag of several hours and is coincident with the appearance of dramatic tubular elements called Salmonella-induced filaments (Sifs), which protrude from the main body of the SCV (Garcia-del Portillo et al., 1993). SCV biogenesis is dependent on selective interactions with the endocytic pathway such that certain proteins characteristic of early endosomes, the recycling pathway and lysosomes are acquired sequentially (Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999a; Steele-Mortimer et al., 1999). Thus, biogenesis is characterized by the rapid loss of the early endocytic proteins, EEA1 and transferrin receptor, and the concurrent acquisition of proteins associated with later stages of the endocytic pathway, such as the vacuolar ATPase and lysosomal glycoproteins (lgp) (Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999a; Steele-Mortimer et al., 1999). Delivery of lgp has been shown to be dependent on the small GTP-binding protein rab7 and is not a result of direct fusion with preformed lysosomes (Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999a). In addition, the cation-independent mannose 6-phosphate receptor (CI-MPR) is effectively excluded from the SCV (Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999a; Steele-Mortimer et al., 1999). Hence, although SCV biogenesis has similarities with the endocytic and phagocytic pathways, it is clearly a distinct process. It appears that the intracellular bacteria are able to block specific membrane fusion events thus determining the unique characteristics of the SCV.

The unique trafficking pattern of the SCV suggests a mediatory role for bacterial effectors and indeed several SPI-2 effectors have been implicated. The SPI-2-encoded SpiC prevents SCV fusion with lysosomes, apparently by interfering with normal membrane trafficking in the host cell (Uchiya et al., 1999). SifA, originally identified for its role in Sif formation, is suggested to promote fusion of the SCV with a novel endosomal compartment, and is required for maintenance of SCV membrane integrity (Beuzon et al., 2000; Brumell et al., 2001). Recently, two other SPI-2-encoded putative effectors, SseF and SseG, have been shown to be required for Sif formation in epithelial cells (Guy et al., 2000). Although these findings are consistent with a role for SPI-2 in modulating SCV maturation, we have recently determined that mutants of this TTSS can replicate normally in some epithelial cell lines (Brumell et al, 2001). This suggests that other factors, possibly SPI-1 effectors, are required for replication in non-phagocytic cells. This idea is supported by two published findings: first, that some SPI-1 effectors may be induced upon invasion of host cells (Pfeifer et al., 1999) and second, SopE (a SPI-1TTSS-translocated effector) interacts with the small GTP-binding protein rab5, to regulate early endosome fusion (Mukherjee et al., 2000).

In this study, we have investigated the role of SPI-1 in intracellular survival/replication and SCV biogenesis using a S. Typhimurium invA mutant, which is unable to invade epithelial cells in vitro (Galan et al., 1992) and is attenuated for oral infection of mice (Galan and Curtiss, 1989). InvA is a putative inner membrane component of the SPI-1TTSS apparatus. The mutant does not translocate SPI-1 effectors although they are synthesized (Collazo and Galan, 1997). We used two different mechanisms to internalize the invA mutant into HeLa cells. One method utilized an invA strain with a plasmid encoding the invasin (inv) gene of Yersinia pseudotuberculosis (Pace et al., 1993). Invasin-mediated uptake is a classic example of the ‘zipper’ mechanism in which a bacterial surface protein (Invasin) binds tightly to a receptor (β1-integrin) on the host cell surface (Isberg and Leong, 1990). This binding leads to tight association of the plasma membrane with the bacteria and, eventually, engulfment in a close-fitting vacuole. In contrast, SPI-1-mediated invasion by Salmonella is characterized by dramatic ruffles in the plasma membrane and uptake of the bacteria into a loose-fitting or spacious vacuole (Brumell et al., 1999). To internalize the invA mutant in a SPI-1-dependent mechanism, we took advantage of the ability of Salmonella-induced ruffles to internalize non-invasive bacteria (Ginocchio et al., 1992) and co-internalized invA with wild-type S. Typhimurium.

Our results show that the invA mutant internalized by either method is replication incompetent, indicating that SPI-1 is essential for intracellular replication. Furthermore, effectors translocated by wild-type bacteria within the same host cell cannot rescue the invA mutant in trans. We also found that the biogenesis of vacuoles containing S. Typhimurium internalized via the Invasin-mediated pathway were more similar to Y. pseudotuberculosis vacuoles than SCVs. This demonstrates that the method of bacterial uptake into host cells has a determining role in the biogenesis of bacteria-containing vacuoles.


Internalization of S. Typhimurium SPI-1 mutants

To investigate whether there is a requirement for SPI-1 effectors in intracellular survival of Salmonella, we used a well characterized invasion deficient SPI-1 mutant, invA. Two different techniques were used to promote internalization of the invA mutant (Fig. 1A). The first method involved transforming invA bacteria with a plasmid (pRI203) encoding the Invasin protein of Y. pseudotuberculosis, which promotes bacterial uptake by binding to β1-integrin molecules on the host cell surface (Isberg and Leong, 1990). Alternatively, the invA mutant was internalized by SPI-1-dependent co-internalization with wild-type S. Typhimurium.

Figure 1.

Internalization of non-invasive invA mutant S. Typhimurium.

A. Schematic representation of bacterial internalization. Wild-type Salmonella are internalized via SPI-1-induced membrane ruffling that does not involve tight association of the bacteria to the host cell (left). In contrast, Yersinia or invA/pRI203 uptake is mediated by Invasin binding to β1 integrins on the plasma membrane, which leads to intimate association of bacteria with the host cell (centre). Ruffles induced by wild-type S. Typhimurium can co-internalize the invA mutant (right).

B. Internalization efficiency of the different methods compared with invasion by wild-type S. Typhimurium. HeLa cells were infected with bacteria for 10 min, followed by a 15 min chase in the presence of gentamicin. Solubilization and estimation of intracellular cfu is described in Experimental procedures.

C. Sensitivity of bacterial internalization to wortmannin. HeLa cells were infected for 15 min, after which invasion was quantified as above. Where indicated, cells were pretreated for 30 min with 100 nM wortmannin, and the drug was present during all subsequent steps (grey bars). Efficiency is expressed as percentage of the control (–WTM, black bars) for each condition. Results are the mean ± SD of three separate experiments.

To assay bacterial internalization, we used a gentamicin resistance assay (Finlay and Falkow, 1988). HeLa cells were infected for 10 min with wild-type or mutant bacteria and then incubated in the presence of gentamicin to kill extracellular bacteria. Intracellular (gentamicin-resistant) bacteria were then estimated by solubilization of cells, followed by plating of appropriate dilutions on to Luria–Bertani (LB) agar. The results demonstrate that invA can be internalized either by complementation with Invasin or by co-internalization with wild-type S. Typhimurium. Thus, whereas invA alone is internalized at less than 0.5% of the efficiency of wild-type bacteria (not shown, and Galan and Curtiss, 1991; Galan et al., 1992), the efficiency is increased to 10 ± 1% by expression of Invasin (invA/pRI203) when equivalent numbers of bacteria are added (Fig. 1B). In comparison, co-internalization of invA with wild-type bacteria (1:1 ratio) is significantly more efficient (44 ± 17%). The highest internalization efficiency was obtained by co-internalization of invA/pRI203 with wild-type S. Typhimurium (78 ± 15%). Wild-type induced uptake of invA/pRI203 may be more efficient than that of invA because of increased attachment of bacteria that can then be readily internalized via ruffles (i.e. SPI-1-mediated) or because both Invasin-mediated and SPI-1-mediated internalization occurs simultaneously.

To test the fidelity of these internalization mechanisms, we used pharmacological inhibition of phosphoinositide 3-kinase (PI3-K), which is required for Invasin-mediated but not SPI-1-mediated invasion (Mecsas et al., 1998). The PI3-K inhibitor wortmannin (WTM) has no effect on SPI-1-mediated invasion of S. Typhimurium but almost completely inhibits (92 ± 5% inhibition compared with wild type) Invasin-mediated uptake of Y. pseudotuberculosis (Fig. 1C) (Mecsas et al., 1998; Schulte et al., 1998). The effect of WTM on Invasin-mediated uptake of invA/pRI203 was also significant (50 ± 11%). WTM partially inhibited internalization of the invA/pRI203 strain in the presence of wild-type S. Typhimurium (40 ± 12%), confirming that Invasin-mediated uptake is occurring in concert with SPI-1-driven uptake (Fig. 1C). In contrast, internalization of the invA mutant in the presence of wild-type S. Typhimurium was not inhibited by WTM, confirming that this uptake is indeed solely SPI-1-mediated. These results show that the SPI-1 invA mutant can be internalized either via Invasin- or SPI-1-mediated mechanisms, and that the efficiency of uptake is directly dependent on the mechanism involved.

SPI-1 is necessary for intracellular replication of S. Typhimurium

We next asked whether internalized invA mutants could survive and/or replicate within epithelial cells. Gentamicin resistance assays were performed, in which incubation in the presence of gentamicin was continued for up to 6 h to allow sufficient time for intracellular replication; (Leung and Finlay, 1991; Garcia-del Portillo et al., 1993). Following invasion, wild-type S. Typhimurium initially undergo a lag phase but begin to replicate by 4.5 h post invasion and, by 6 h, the numbers of intracellular viable bacteria has increased by four to eightfold (Fig. 2A). In contrast, irrespective of the method used for internalization, intracellular invA did not increase intracellularly (Fig. 2B–D). This deficiency was not strain-dependent, as similar results were obtained using other strains (not shown). Thus, translocated SPI-1 effectors are necessary for intracellular replication by S. Typhimurium. Furthermore, the results indicate that SPI-1 effectors translocated by wild-type bacteria are unable to rescue the intracellular replication deficiency of invA.

Figure 2.

Intracellular survival and proliferation of internalized invA mutants. HeLa cells were infected with wild-type S. Typhimurium (▪, □) and/or invA/pRI203 (•, ○) or invA (◆, ◊). At the indicated times, intracellular cfu were isolated and estimated as described in Experimental procedures. Where indicated (open symbols), invasion and incubations were carried out in the presence of 100 nM wortmannin. Cells were infected with the wild-type S. Typhimurium alone (A), invA/pRI203 alone (B), wild-type and invA/pRI203 strains simultaneously (C) or wild-type and invA strains simultaneously (D). Results are shown as the fold increase in gentamicin-resistant bacteria compared with the number at 1.5 h. Values are the mean ± SD of three separate experiments.

One possibility for the inability of invA to replicate is that it is unable to inhibit fusion of the SCV with lysosomes. To investigate this possibility, we perturbed host cell membrane trafficking so as to alter SCV biogenesis and, perhaps, intracellular survival of S. Typhimurium (Garciadel Portillo and Finlay, 1995; Méresse et al., 1999a; Rathman et al., 1996; Steele-Mortimer et al., 2000a). PI3-K plays an integral role in many trafficking events and the effects of inhibition by WTM are well described, and include changes in the morphology of endocytic compartments and inhibition of endocytic transport (Clague et al., 1995; Davidson, 1995; Jones and Clague, 1995; Li et al., 1995; Reaves et al., 1996; Shpetner et al., 1996; Patki et al., 1997; Gillooly et al., 1999; Prior and Clague, 1999; Tuma et al., 1999). The use of WTM, rather than other inhibitors of membrane traffic, was also practical as its effects on internalization of invA and wild-type S. Typhimurium had already been characterized.

WTM causes a significant increase (more than twofold compared with untreated cells) in the numbers of intracellular wild-type S. Typhimurium, but only following the initiation of intracellular replication at 4.5–6 h p.i. (Fig. 2A, C and D). This increase is not as a result of a direct effect of WTM on bacterial replication, as growth in cell-free media is unaffected (not shown). To ensure that the differences in bacterial numbers was not because of decreased release of bacteria from cells or detachment of infected cells, we adapted the gentamicin assay to measure both intracellular and extracellular bacteria (including those in detached cells). In this assay, gentamicin is present for only 45 min after bacterial internalization, ensuring that all bacteria not initially internalized are killed but those released from cells later are not affected. The growth media, which is normally discarded, was retained and also treated identically to the cell monolayer. Both samples were then diluted and plated for estimation of colony-forming units (cfu). As shown in Fig. 3, less than 10% of total bacteria were extracellular, even at 6 h, and WTM had no significant effect on this percentage. Thus, the observed increase in intracellular bacteria seen with WTM treatment (Fig. 2A) is not a result of changes in cell attachment or bacterial release. Surprisingly, unlike for wild-type bacteria, WTM treatment did not increase intracellular numbers of the invA mutant at 6 h p.i. whether they were internalized via Invasion-mediated invasion or by co-internalization with wild type (Fig. 2B, C and D). Thus, perturbation of host cell membrane trafficking by PI3-K inhibition is unable to overcome the replication defect of SPI-1TTSS-deficient bacteria.

Figure 3.

Wortmannin treatment does not significantly affect the numbers of extracellular bacteria. HeLa cells were infected with S. Typhimurium in the presence (▪) or absence (□) of 100 nM wortmannin. At the indicated times, both extracellular (including bacteria in detached cells) and intracellular cfus were isolated and estimated as described in Experimental procedures. Results are the mean ± SD of three separate experiments.

SCV biogenesis is SPI-1-dependent

As the perturbation of membrane traffic did not rescue the replication defect of intracellular invA, we next investigated whether there were detectable differences in SCV biogenesis compared with wild-type bacteria. Previously, we have shown that specific eukaryotic proteins appear sequentially on the SCV membrane (Steele-Mortimer et al., 1999). For example, the early endosome specific protein, EEA1, is detectable only for the first 10–20 min after bacterial internalization, whereas the lgp Lamp1 is acquired after loss of EEA1. Here, we have used the sequential appearance of EEA1 and Lamp1 to monitor normal SCV biogenesis. HeLa cells were incubated with bacteria for 10 min, extracellular bacteria were removed by washing and the cells were then incubated for up to 75 min as indicated. At the indicated times, the cells were fixed and stained for immunofluorescence analysis with antibodies recognizing lipopolysaccharide (LPS), EEA1 or Lamp1. As expected, the percentage of EEA1+ SCVs decreases from approximately 75% at 15 min to less than 1% at 75 min (Fig. 4A). In contrast, Lamp1 is found on less than 1% of SCVs at 15 min p.i. compared with almost 100% of SCVs at 75 min (Fig. 4B). According to our previous studies, almost all SCVs are Lamp1+ by this time-point (Steele-Mortimer et al., 1999). The trafficking pattern for vacuoles containing invA internalized via the Invasin-mediated pathway (invA/pRI203) was strikingly different. At 15 min, approximately 30% of SCVs are EEA1+ and this does not change significantly over time (Fig. 4C). Furthermore, Lamp1 is acquired with much slower kinetics compared with SCVs containing wild-type S. Typhimurium, such that less than 50% of the vacuoles are Lamp1+ at 75 min (Fig. 4D). To determine if this was as a result of the mechanism of internalization rather than the lack of SPI-1 effectors, we also studied Y. pseudotuberculosis-containing vacuoles (YCV). We found that the kinetics of acquisition of EEA1 and Lamp1 on these vacuoles were almost identical to invA/pR1203 SCVs, implying that the mechanism of bacterial entry has a profound effect on the biogenesis of the vacuole (Fig. 4E and F).

Figure 4.

Acquisition of EEA1 and Lamp1 by vacuoles containing Salmonella or Yersinia. Cells infected with wild-type S. Typhimurium (A and B), invA/pRI203 (C and D), or Y. pseudotuberculosis (E and F) were fixed, permeabilized and incubated with a rabbit anti-LPS antibody (all panels) and either a human anti-EEA1 antibody (A, C, E) or a mouse anti-Lamp1 antibody (B, D, F). Secondary antibodies were Alexa 488-conjugated goat anti-rabbit and Alexa 594-conjugated goat anti-mouse or anti-human. For each time-point, at least 100 vacuoles were scored as positive or negative for each marker protein. Results are the mean ± SD of three separate experiments.

Mannose 6-phosphate receptor is transiently acquired by invA/pRI203-containing vacuoles

An unusual characteristic of SCVs is the exclusion of CI-MPR, a host cell protein normally present in late endosomes (Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999a; Steele-Mortimer et al., 1999). As the invA/pRI203 mutant did not replicate within HeLa cells, and SCVs containing this strain had abnormal EEA1 and Lamp1 acquisition characteristics, we considered whether they might acquire CI-MPR. Immunofluorescence experiments using antibodies against CI-MPR showed that, indeed, this receptor could be detected on significant numbers of invA/pRI203 vacuoles at 40 min p.i. (18% versus 7% of wild-type SCVs; Fig. 5). This acquisition was transient, decreasing to less than 5% at 50 min p.i., which indicates that the vacuole is undergoing further modification and is consistent with the increase in Lamp1 acquisition after 40 min (Fig. 4D). This result is further evidence that the mechanism of uptake, invasin- versus SPI-1-mediated, has profound effects on vacuole biogenesis.

Figure 5.

CI-MPR is transiently acquired by SCVs containing invA/pRI203 S. Typhimurium. HeLa cells were infected with wild-type S. Typhimurium or the invA/pRI203 strain as described, and then fixed at the indicated times and processed for immunofluorescence using rabbit anti-CI-MPR and mouse anti-LPS antibodies, followed by Alexa 488-conjugated goat anti-mouse and Alexa 594-conjugated goat anti-rabbit. The results are from two independent experiments.

WTM perturbation of SCV biogenesis

Previously, it has been shown that WTM causes dissociation of EEA1 from endosomal membranes (Patki et al., 1997). However, under these experimental conditions, EEA1 is not completely removed from endomembranes (Fratti et al., 2001), although tubular EEA1+ structures were induced (Fig. 6E and G). These tubules were transferrin receptor-positive (not shown) and are thus similar to the WTM-induced, transferrin-containing tubules described previously (Shpetner et al., 1996). WTM also causes swelling of Lamp1-containing compartments (Davidson, 1995; Reaves et al., 1996), although this change is less dramatic than that induced in early endosomes (Fig. 6F and H). WTM had no significant effect on the removal of EEA1 from SCVs containing wild-type S. Typhimurium (Fig. 4A). However, the acquisition of Lamp1 by SCVs was dramatically reduced such that at 75 min p.i., Lamp1 was detected on less than 30% of SCVs (Fig. 4A and B). At later time-points, the difference is even more dramatic, with less than 20% of SCVs being Lamp1+ at 6 h p.i. (Fig. 7). Thus, for WTM-treated cells, there is an inverse correlation between the observed increase in numbers of intracellular wild-type bacteria and a decreased delivery of Lamp1 to these SCVs. This decreased delivery is actually associated with the release of intracellular wild-type S. Typhimurium into the host cell cytosol (J. H. Brumell, P. Tang and B. B. Finlay, submitted for publication). Strikingly, no effect was observed on the delivery of Lamp1 to vacuoles containing bacteria internalized by Invasin-mediated uptake (Fig. 4D and F).

Figure 6.

Localization of EEA1 and Lamp1 to SCVs. Cells were infected with wild-type S. Typhimurium in the absence (A–D) or presence (E–H) of wortmannin and then fixed at 5 (A, B, E, F) or 60 min (C, D, G, H) p.i. Immunostaining was performed using rabbit anti-LPS with human anti-EEA1 (A, C, E, G) or mouse anti-Lamp1 (B, D, F, H). Secondary antibodies were Alexa 488-conjugated goat anti-rabbit and Alexa 594-conjugated goat anti-mouse or anti-human. Insets are enlargements of the bacteria shown in each image, indicated by large arrows. Arrowheads indicate large EEA1+ macropinosomes formed in infected cells (A). Small arrows indicate WTM-induced EEA1+ tubules (E and G). Bar equals 10 μm.

Figure 7.

Acquisition of Lamp1 is delayed in vacuoles containing the invA mutant. Cells were co-infected with GFP-expressing wild-type S. Typhimurium (▪, □) and the invA mutant (SB103) (•, ○) in the absence (closed symbols) or presence (open symbols)) of wortmannin. At the indicated times, cells were fixed and processed for immunofluorescence as described in Experimental procedures. Total intracellular bacteria were revealed using rabbit anti-LPS and Cy5-conjugated donkey anti-rabbit secondary antibody. Lamp1 was revealed using mouse monoclonal anti-Lamp1 followed by Alexa 568-conjugated goat anti-mouse antibody. The results are for at least 100 SCVs for each condition from two separate experiments.

The kinetics of EEA1 and Lamp1 acquisition by SCVs containing invA co-internalized with wild-type S. Typhimurium, in the presence and absence of WTM treatment was also analysed. Compared with wild-type bacteria, the acquisition of EEA1 for invA was not changed (not shown). However, acquisition of Lamp1 was significantly delayed (Fig. 7). Only 44 ± 12% of SCVs containing invA were Lamp1+ at 1 h, compared with 75 ± 4% of SCVs containing wild-type S. Typhimurium. By later time-points (6 h p.i.), the numbers of Lamp1+ SCVs were similar for both wild-type and mutant. As for wild-type bacteria, WTM delayed Lamp1 acquisition by invA SCVs, although to a lesser extent (Fig. 7).

Co-internalized invA mutant and wild-type S. Typhimurium are in separate vacuoles

Co-internalization of invA with wild-type S. Typhimurium could potentially result in vacuoles containing both wild-type and mutant bacteria. To investigate this possibility we carried out immunofluorescence staining of cells in which invA had been co-internalized with wild-type S. Typhimurium expressing green fluorescent protein (GFP-S. Typhimurium). In this way, mutant and wild-type bacteria can be identified separately. Cells were infected for 15 min and then either fixed and processed immediately or incubated for a further 30 min in the presence of gentamicin. Immunofluorescence analysis was carried out using anti-LPS, to stain all cell-associated bacteria and anti-Lamp1 antibodies. As shown in Fig. 8, internalized invA mutants were contained within individual vacuoles that did not contain wild-type bacteria. At the time-points shown, intracellular bacteria have not begun to replicate and, thus, SCVs contain either single bacteria or, sometimes, two bacteria when an internalized bacterium has divided once. Rarely (less than 1% of infected cells), the mutant was found in cells containing no wild-type S. Typhimurium (not shown).

Figure 8.

Co-internalized invA mutant bacteria are in discrete vacuoles. Cells were co-infected with invA mutant and GFP-S. Typhimurium for 15 min and then either fixed immediately for immunostaining (A–D) or incubated for a further 30 min in the absence of extracellular bacteria (E–H). Immunostaining was performed using rabbit anti-LPS to reveal total bacteria (A and E) and mouse anti-Lamp1 (C, G). Wild-type S. Typhimurium are detected by anti-LPS (A and E) as well as expression of GFP (B and C). Secondary antibodies were Alexa 568-conjugated anti-mouse for Lamp1 (C, G) or Cy5-conjugated anti-rabbit for LPS (A and E). Individual channels are shown as well as overlays (D and H) in which the colours have been reassigned to optimize the visibility (red, total bacteria; blue, GFP-S. Typhimurium; green, Lamp1).


The Salmonella TTSS, encoded by SPI-1 and SPI-2, are required to translocate effectors into host cells in which they then interact with their target eukaryotic proteins. It has been proposed that SPI-1 is required to transduce signals from outside the host cell whereas SPI-2, which is induced intracellularly, transduces signals from within the SCV (Hensel, 2000). Indeed, it is well documented that SPI-1 is essential for invasion of non-phagocytic cells whereas SPI-2 is required for intracellular survival and proliferation in phagocytes (Marcus et al. 2000). However, some SPI-1 effectors are induced upon invasion into both phagocytic and non-phagocytic cells, suggesting that they may also be required post invasion (Pfeifer et al., 1999). Moreover, it takes a number of hours after bacterial invasion of host cells for SPI-2 gene induction, so it is likely that SPI-1 effectors are important at least for the early stages of host cell infection, when Salmonella must establish its intracellular niche. In this study, we have investigated the ability of an invA mutant, which does not translocate SPI-1 effectors, to survive and proliferate in cultured epithelial cells. To overcome the invasion deficiency of the invA mutant, we used two different internalization mechanisms; (i) co-internalization with wild-type S. Typhimurium (SPI-1-dependent) and (ii) Invasin-mediated uptake (SPI-1-independent). Both of these mechanisms resulted in internalization of the non-invasive invA mutant.

These two internalization systems were then used to compare the ability of intracellular SPI-1TTSS mutants to survive and replicate, and also to investigate vacuole biogenesis. Both mechanisms resulted in invA bacteria being internalized within discrete vacuoles. However, in both cases, no intracellular proliferation was detected. Thus, in addition to the many other demonstrated roles for SPI-1 secreted effectors, it appears that these effectors are required for intracellular proliferation in non-phagocytic cells. We considered that endocytic trafficking plays an important role in the intracellular survival of S. Typhimurium, as biogenesis of the SCV, within which the bacteria survive and replicate, involves dynamic interactions with the endocytic pathway (Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999a; Steele-Mortimer et al., 1999). Significantly, fusion with lysosomes, and the subsequent delivery of lysosomal hydrolases, is apparently disrupted by S. Typhimurium (Garcia-del Portillo and Finlay, 1995). Therefore, we analysed whether pharmacological inhibition of endocytic trafficking would affect the intracellular proliferation of S. Typhimurium. The PI3-K inhibitor WTM affects the endocytic pathway at multiple steps, including delivery to lysosomes (Martys et al., 1996; Reaves et al., 1996; Shpetner et al., 1996). Whereas WTM had no effect on invasion by wild-type S. Typhimurium (Fig. 1; Ireton et al., 1996; Mecsas et al., 1998), we found that intracellular proliferation was increased by WTM treatment. However, inhibition of membrane trafficking alone is not sufficient for intracellular survival, as the replication defect of the invA mutant was not rescued by WTM treatment.

As it is clear that S. Typhimurium modulates phagosome maturation to survive within the vacuole, we also compared the maturation of vacuoles containing either wild-type S. Typhimurium or the SPI-1 TTSS mutant. Our data show that early vacuole biogenesis is dependent on the mechanism of entry, as SCVs containing SPI-1 TTSS mutants co-internalized with wild-type S. Typhimurium (SPI-1-dependent) show no defect in EEA1 kinetics. In contrast, vacuoles containing either Y. Pseudotuberculosis or invA/pRI203 (Invasin-dependent internalization) had delayed EEA1 acquisition. Furthermore, CI-MPR is acquired transiently by vacuoles containing bacteria internalized via Invasin-mediated uptake but not by wild-type S. Typhimurium (this work, and Garcia-del Portillo and Finlay, 1995; Méresse et al., 1999b; Mills and Finlay, 1998).

A spectrum of effects has been reported for SPI-1 effectors, including increased intracellular Ca2+ (Ginocchio et al., 1992) and inositol 1,4,5,6-tetrakisphosphate levels (Eckmann et al., 1997), regulation of the small GTP-binding proteins Cdc42 and Rac1 (Hardt et al., 1998; Fu and Galan, 1999), activation of the protein kinases Jnk, p38 and Akt (Hobbie et al., 1997; Steele-Mortimer et al., 2000b), stimulation of the transcription factor NF-κB and release of pro-inflammatory cytokines (Hobbie et al., 1997) and induction of apoptosis in macrophages (Hersh et al., 1999; Jesenberger et al., 2000). Our data introduce another role for SPI-1-mediating intracellular proliferation in epithelial cells. Indeed, this is consistent with an essential role for SPI-1 in the early stages of gastrointestinal disease, including intestinal colonization and the induction of fluid secretion (Wallis and Galyov, 2000). Further studies are required to identify the specific SPI-1 effectors, and their mode of action, that allow the successful colonization of epithelial cells by Salmonella.

Experimental procedures


Wortmannin (Calbiochem) stock was made up at a concentration of 100 mM in DMSO (Sigma) and stored at −20°C. Anti-LPS (Salmonella O antiserum group B factors 1, 4, 5 and 12) was from Difco Laboratories. Human anti-EEA1 was a gift from Dr Ban-Hock Toh, Monash Medical School, Melbourne, Australia. Monoclonal anti-human Lamp1, developed by J. T. August, was obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Studies, Iowa City. Rabbit anti-CI-MPR was a gift from Dr Bernard Hoflack, IBL, Lille, France.

Cell culture and bacterial strains

HeLa (human cervical adenocarcinoma cells; ATCC CCL2) were cultured in growth media (GM), consisting of Eagle’s Minimal essential medium (MEM) (Gibco BRL) supplemented with 10% heat-inactivated fetal calf serum (FCS) (Gibco BRL), at 37°C in 5% CO2. Cells were passaged no more than 20 times.

Salmonella Typhimurium SL1344, invA (SB111), invA (SB103) and invA/pRI203, and Y. pseudotuberculosis YIII (plasmid cured) strains have been described previously (Hoiseth and Stocker, 1981; Pace et al., 1993; Garcia-del Portillo et al., 1994; Steele-Mortimer et al., 2000b). Bacteria were grown in Luria–Bertani (LB) broth or on LB agar plates. When appropriate, kanamycin (50 μg ml−1) or ampicillin (100 μg ml−1) was added. GFP-expressing bacteria were as described (Méresse et al., 1999a; 2001).

Gentamicin protection assays

Invasion assays were carried out essentially as described previously (Steele-Mortimer et al., 1999). Where indicated, cells were treated with WTM (100 nM) for 30 min before the addition of bacteria and continued throughout the experiment. Briefly, S. Typhimurium were grown in LB broth overnight at 37°C with shaking, and then subcultured (300 μl in 10 ml of LB) for a further 3 h. The culture was centrifuged at 10 000 g for 2 min at room temperature, and then directly resuspended in phosphate-buffered saline (PBS). Yersinia pseudotuberculosis were grown overnight in 2 ml of brain–heart infusion (BHI) media at 26°C. This suspension was diluted 1:100 in growth medium and used for invasion. Invasion was initiated by the addition of diluted bacteria directly to subconfluent HeLa cells grown in 24-well plates, and proceeded at 37°C, in 5% CO2, for 5 or 10 min Extracellular bacteria were then removed by washing with PBS and the cells were incubated in GM. In experiments requiring longer infection times, GM was supplemented with 50 μg ml−1 of gentamicin (Sigma), at 20 min post infection, to kill extracellular bacteria. At the indicated times, monolayers were washed twice with PBS and solubilized in 1 ml of 1% Triton X-100, 0.1% SDS in PBS for 5 min at room temperature. The solute was serially diluted in PBS and plated on to duplicate LB agar plates. LB plates were supplemented with 25 μg ml−1 of kanamycin for the invA mutant S. Typhimurium, and with 25 μg ml−1 of kanamycin and 25 μg ml−1 chloramphenicol for invA/pRI203. Plates were incubated at 37°C and the colonies were counted the next day.

To assay extracellular bacteria, gentamicin was present for the first 45 min only. Thereafter, the monolayers were incubated in antibiotic free media. At the indicated times the supernatant was harvested and brought to 1% Triton X-100, 0.1% SDS by the addition of 10× stock solution. Incubation and plating was then carried out as above.


HeLa cells on glass coverslips were infected with bacteria, fixed and processed for immunofluorescence as described previously (Steele-Mortimer et al., 1999). Fixation was in 2.5% formaldehyde followed by permeabilization in PBS containing 10% (v/v) normal goat serum (Gibco BRL) and 0.1% (w/v) Saponin (SS-PBS) for 10 min. For CI-MPR labelling, an additional extraction with 0.1% Triton X-100 for 4 min was performed. Primary antibodies were diluted in SS-PBS as follows; anti-LPS at 1:300, anti-CI-MPR at 1:250, anti-EEA at 1:1000 and anti-Lamp1 at 1:100. Secondary antibodies, Cy5-conjugated donkey anti-rabbit IgG, Alexa 488-conjugated goat anti-rabbit IgG, Alexa 594-conjugated goat anti-mouse or goat anti-human IgG were diluted 1:600 in SS-PBS. After each antibody incubation (30 min each), the monolayers were washed extensively with PBS containing 0.05% Saponin. Coverslips were mounted on to glass slides using Mowiol (Calbiochem). Cells were examined using a Zeiss Axiophot microscope using a 63× oil immersion objective. For each time-point, at least 50 intracellular bacteria were scored for co-localization with Lamp1, EEA1 or CI-MPR. Micrographs were taken using TMAX 400 film (Eastman Kodak) at ASA 1600. Images were processed using ADOBEPHOTOSHOP.


O.S.-M. was the recipient of a postdoctoral fellowship from the European Molecular Biology Organization (EMBO). J.H.B. is supported by a Medical Research Council of Canada postdoctoral fellowship and is an honorary fellow of the Izaac Walter Killam Memorial foundation. This work was supported by operating grants from the Medical Research Council of Canada (B.B.F) and a Howard Hughes International Research Scholar Award (B.B.F.). B.B.F. is a CIHR Distinguished Investigator.