We used flow cytometry and confocal immunofluorescence microscopy to study the localization of Salmonella typhimurium in spleens of infected mice. Animals were inoculated intragastrically or intraperitoneally with S. typhimurium strains, constitutively expressing green fluorescent protein. Independently of the route of inoculation, most bacteria were found in intracellular locations 3 days after inoculation. Using a panel of antibodies that bound to cells of different lineages, including mononuclear phagocyte subsets, we have shown that the vast majority of S. typhimurium bacteria reside within macrophages. Bacteria were located in red pulp and marginal zone macrophages, but very few were found in the marginal metallophilic macrophage population. We have demonstrated that the Salmonella SPI-2 type III secretion system is required for replication within splenic macrophages, and that sifA− mutant bacteria are found within the cytosol of these cells. These results confirm that SifA and SPI-2 are involved in maintenance of the vacuolar membrane and intracellular replication in vivo.
Salmonella typhimurium causes a systemic typhoid-like illness in mice. In susceptible (Itys) host strains, less than 10 wild-type bacteria administered by the subcutaneous, intravenous or intraperitoneal route are sufficient to cause a fatal infection (Plant and Glynne, 1974; Shea et al., 1999). After oral inoculation, translocation of S. typhimurium across the gut epithelium into the bloodstream occurs by invasion of M cells in ileal Peyer's patches (Carter and Collins, 1974; Jones et al., 1994), and also via transmigrating CD18-expressing phagocytes (Vazquez-Torres et al., 1999). Regardless of the route of inoculation, a transient bacteraemia is followed over the course of several days by the accumulation of large numbers of bacteria within the spleen and liver, leading to a second, fatal bacteraemia (Carter and Collins, 1974; Shea et al., 1999).
Despite widespread use of this murine model of systemic infection, the localization of S. typhimurium in vivo has been contentious (Wang et al., 1988; Conlan and North, 1992; Dunlap et al., 1992; Matsui et al., 2000). In confocal microscopic analysis of liver sections, bacteria were shown clearly to be present within CD18-expressing cells presumed to be macrophages (Richter-Dahlfors et al., 1997). However, data regarding the localization of Salmonella within the spleen, which is the other major site of S. typhimurium replication in vivo, remain inconclusive. Previous studies have suggested different locations, ranging from extracellular spaces to polymorphonuclear cells (Wang et al., 1988; Dunlap et al., 1992). S. typhimurium has also been shown to reside in CD11b-positive cells, which include macrophages, monocytes, polymorphonuclear cells and even some subsets of lymphocytes (Matsui et al., 2000).
In vitro studies, using different host cell types, have shown that S. typhimurium replicates intracellularly within Salmonella-containing vacuoles (SCVs) apparently protected from host cell antimicrobial activities (Buchmeier and Heffron, 1991; Rathman et al., 1997; Vazquez-Torres et al., 2000). Genetic studies have led to the identification of a large number of genes necessary for intracellular survival. These include the phoP/Q genes, encoding a two-component regulatory system, which regulates the expression of over 40 genes (Miller and Mekalanos, 1990), the spv locus on the large S. typhimurium plasmid (Gulig et al., 1998) and the Salmonella pathogenicity island-2 (SPI-2), which encodes a type III secretion system (TTSS) needed for bacterial replication within cultured macrophages (Ochman et al., 1996; Cirillo et al., 1998). The fact that strains carrying mutations in genes required for intracellular growth in vitro are typically avirulent in mice (Fields et al., 1986) suggests that S. typhimurium replicates intracellularly in vivo.
In this study, we have used flow cytometry and confocal microscopy to investigate the localization of S. typhimurium in the mouse spleen. Anatomically, the spleen can be divided into the white pulp and red pulp compartments with an intermediate marginal zone. Blood and microrganisms will flow through the marginal zone and the red pulp reticular meshwork where they come into contact with immune effectors. The marginal zone is predominantly composed of B cells and macrophages, with two major macrophage populations: marginal metallophilic macrophages (MMMs) between the white pulp and the marginal sinus; and the marginal zone macrophages (MZMs) in the red pulp side of the sinus (Kopp, 1990; Kraal, 1992). The localization and characterization of spleen macrophages harbouring Leishmania donovani was reported recently (Lang et al., 2000).
We show that after intragastric (i.g.) or intraperitoneal (i.p.) inoculation, most bacteria are found within mononuclear phagocyte subsets known to be present in the red pulp and in the marginal zone domain contacting the red pulp. We also demonstrate that the SPI-2 TTSS is required for intramacrophage replication in vivo and that SifA is involved in maintaining the integrity of the vacuolar membrane surrounding wild-type bacteria in vivo (Beuzón et al., 2000).
S. typhimurium is predominantly intracellular in the mouse spleen
To identify bacterial cells, we used a wild-type strain of S. typhimurium (12023s), transformed with a plasmid encoding green fluorescent protein (GFP) (Valdivia and Falkow, 1997). The virulence of bacteria carrying this plasmid was indistinguishable from that of the wild-type strain, in terms of lethality to mice and bacterial colony forming units (cfu) recovered from infected animals. The stability of the plasmid was shown to be maintained in bacteria during infection (data not shown).
We were unable to quantitate accurately GFP-S. typhimurium in relation to its cellular location by confocal analysis of spleen sections because of high background autofluorescence and a relatively low number of total splenocytes associated with bacteria, even in heavily infected animals. Therefore, splenocyte suspensions were prepared and analysed using flow cytometry. Irrespective of the dose or route of inoculation, a consistent proportion of splenocytes associated with GFP was detected at late stages of infection (Fig. 1A). After i.p. inoculation of 102 cfu, < 0.5% of splenocytes were GFP-positive by 3 days, increasing to 5.8 ± 2.0% after a further 48 h. A similar load of infected splenocytes (6.45 ± 3.4%) was achieved 3 days after i.p. inoculation of 105 cfu, shortly before mice would have succumbed to the infection. After i.g. inoculation of 108 cfu, 6.7 ± 5.0% of splenocytes were associated with fluorescent bacteria by 9 days.
To quantify the proportions of intracellular and extracellular bacteria, spleen cell suspensions were obtained after 3 days from mice inoculated i.p. with 105 cfu of GFP-S. typhimurium. Infected spleens were digested enzymatically and fixed immediately for fluorescence microscopy. Extracellular bacteria were stained without permeabilizing host cells, using an anti-Salmonella primary and a Texas red-conjugated secondary antibody. When images from the red and green channels were merged, extracellular bacteria appeared yellow, whereas intracellular bacteria were green. Of over 500 bacteria examined in this way, 72 ± 3.0% were intracellular, similar to estimates made using confocal analysis of liver sections (Richter-Dahlfors et al., 1997). The remaining bacteria may represent extracellularly replicating S. typhimurium, bacteria released by dying host cells or intracellular bacteria released during spleen cell preparation. Confocal microscopy was also used to confirm the intracellular location of bacteria in these splenocyte preparations. A z-stack was re-sliced horizontally and vertically to obtain the projections of perpendicular views, confirming that bacteria were intracellular (Fig. 1B).
Together, these experiments confirm that S. typhimurium has a predominantly intracellular location in the spleen, independent of the dose or route of inoculation.
S. typhimurium is located within splenic macrophages
S. typhimurium has been found in CD18-positive phagocytes in the blood and liver (Richter-Dahlfors et al., 1997; Vazquez-Torres et al., 1999) and splenic CD11b-expressing cells (Matsui et al., 2000), which could include granulocytes, monocytes and NK cells, as well as subsets of B (e.g. CD5-positive) and T (mainly CD8-positive) lymphocytes. We used flow cytometry to differentiate cell types within the spleen by both light scatter characteristics and immunophenotyping in order to determine which cells are predominantly infected by GFP-S. typhimurium in vivo. The forward and side scatter characteristics of lymphocytes and myeloid cells are well established, and allow each population to be analysed separately. In Fig. 2A, the R1 region selects lymphoid cells and the R2 region selects predominantly myeloid cells, including cells of monocyte/macrophage lineage. Three days after i.p. inoculation of mice with 105 GFP-S. typhimurium, these populations were analysed for GFP-positive cells. In the R1 region, less than 1% of cells were positive for GFP, compared with 30–40% of splenocytes in the R2 region.
To investigate in more detail the identity of host cells harbouring S. typhimurium, splenic T cells, B cells and macrophages from infected mice were identified using specific antibodies, and analysed for green fluorescence. Anti-CD3 and anti-CD19 antibodies were used to identify T and B cells respectively. Three monoclonal antibodies raised in the same host (F4/80, anti-scavenger receptor/2F8 and anti-sialoadhesin/CD169) were pooled to provide a pan-macrophage marker. The F4/80 antibody recognizes mature macrophages mainly in the red pulp of the spleen (Austyn and Gordon, 1981; Hume et al., 1983), whereas the 2F8 and anti-sialoadhesin antibodies recognize different subsets of spleen marginal zone macrophages (Crocker and Gordon, 1989; Crocker et al., 1991; Fraser et al., 1993; Hughes et al., 1995). In all experiments, appropriate isotype controls were included for gate setting. As expected, low percentages of CD3-positive cells (1.4 ± 0.7%), and of CD19-positive cells (0.9 ± 0.8%) were associated with GFP-S. typhimurium(Fig. 2B). The combination of the three macrophage markers stained 10–12% of total spleen cells, which is consistent with the expected number of macrophages in the uninfected spleen (Lang et al., 2000). In contrast to lymphocytes, 36.0 ± 3.5% of macrophages detected by these antibodies were associated with GFP-S. typhimurium(Fig. 2B). Furthermore, virtually the entire population of GFP-positive splenocytes was stained by these macrophage markers, compared with the isotype controls (Fig. 2C), indicating that macrophages account for the vast majority of infected cells.
Association of S. typhimurium with subsets of splenic macrophages
The majority of macrophages in the spleen are located in the red pulp and marginal zones. In the marginal zone, two different macrophage populations have been described (Kraal, 1992). Cells bearing the class A macrophage scavenger receptor (MSR-A) recognized by the 2F8 antibody, are highly phagocytic (Hughes et al., 1995). In contrast, marginal metallophilic macrophages labelled by the anti-sialoadhesin antibody (CD169) are relatively non-phagocytic (Hughes et al., 1995; Kraal, 1992). Red pulp macrophages can be identified by the F4/80 antibody, but may also have low level expression of MSR-A labelled by 2F8 (Hughes et al., 1995).
Mice were inoculated i.g. with 108 cfu or i.p. with 105 cfu, and spleens harvested after 9 and 3 days respectively. Splenic cell suspensions were fixed and mounted on coverslips, then stained for each macrophage marker individually and analysed by confocal immunofluorescence microscopy. For each route of inoculation, at least 200 cells containing GFP-S. typhimurium were examined, and the percentage of infected cells positive for each antibody was calculated after subtracting low level non-specific staining revealed by the appropriate isotype controls. Only those cells that stained strongly were included in this analysis, so the true overall percentage of infected cells positive for a given antibody is underestimated. Significantly greater proportions of infected cells were positive for 2F8 and F4/80 than for the anti-sialoadhesin antibody (Fig. 3A). Similar results were observed 2 days after i.p. inoculation (Fig. 3B). The total proportion of each macrophage subset was determined using flow cytometry (Fig. 3C). Similar numbers of splenocytes were positive for the 2F8 and CD169 antibodies (approximately 3% and 2% respectively) and a higher proportion of cells were positive for the F4/80 antibody (approximately 5%). Together, our results suggest that the distribution of S. typhimurium within the red pulp and marginal zone is not random: bacteria occupy a low proportion of marginal metallophilic macrophages and a relatively high proportion of phagocytic marginal zone macrophages. Figure 3D illustrates typical staining patterns of the three anti-macrophage antibodies.
A SPI-2 mutant strain has an intracellular replication defect in vivo
SPI-2 is a pathogenicity island that encodes a TTSS required for systemic growth of S. typhimurium (Shea et al., 1999). SPI-2 mutants display replication defects within cultured macrophages, but this has not been demonstrated previously in vivo. Therefore, we compared bacterial numbers in spleen cells from mice infected with wild-type GFP-S. typhimurium or a strain carrying a mutation in ssaV, which encodes a structural component of the SPI-2 TTSS required for secretion (Beuzón et al., 2000). Splenocytes were prepared from mice inoculated with 105 wild-type GFP-S. typhimurium or with 106 GFP-ssaV− strain. A higher inoculum of the mutant strain was necessary to provide a sufficient number of infected cells for analysis. Cell suspensions were fixed on coverslips, and at least 200 infected cells, representing four mice for each strain, were analysed using confocal microscopy. Despite the greater inoculum, the number of host cells containing mutant bacteria was much less than the number containing wild-type bacteria (Fig. 4A). The absolute median corresponding to four mice for each bacterial strain was one bacterium per cell for the ssaV− strain, compared with five bacteria per cell for the wild-type strain. In mice infected with the ssaV− strain, less than 5% of infected splenocytes contained more than five bacteria. In contrast, 44.5% of infected splenocytes contained more than five bacteria (Fig. 4B) in mice infected with wild-type S. typhimurium. Indeed, some cells were infected with over 50 wild-type bacteria, but such heavily infected cells were never observed in the case of the ssaV− strain. The numbers of splenocytes infected with wild-type or ssaV− mutant bacteria were also examined over a time course using flow cytometry. In contrast to a rapid increase in the numbers of splenocytes infected with the wild-type strain, there was a very slight increase in the percentage of splenocytes infected with the ssaV− mutant strain (Fig. 4C). These results are consistent with previous work from our laboratory, showing that SPI-2 mutant bacteria are capable of limited growth in vivo (Shea et al., 1999), and provide direct evidence that SPI-2 mutants have an intracellular replication defect in vivo.
SifA is involved in maintenance of the vacuolar membrane enclosing wild-type Salmonella within spleen macrophages
Many studies using cultured host cells have shown that intracellular S. typhimurium occupies a specialized vacuole apparently protected from normal bactericidal/bacteriostatic activities of the host cell (Buchmeier and Heffron, 1991; Rathman et al., 1997). Maintenance of the vacuolar membrane surrounding S. typhimurium cells is dependent on a bacterial protein called SifA. In epithelial cells and RAW 264.7 macrophages, sifA− mutant bacteria progressively lose their vacuolar membrane (Beuzón et al., 2000). This process can be monitored by microscopy, using antibodies against lysosomal membrane glycoproteins (lgps) that are present in the Salmonella vacuolar membrane (Beuzón et al., 2000). To determine if the sifA− mutant has the same phenotype in vivo, mice were inoculated i.p. with either 105 cfu of wild-type GFP-S. typhimurium or 106 cfu GFP-ssaV− or GFP-sifA− strains. After 3 days, splenocytes were isolated, fixed and stained for the lysosomal membrane glycoprotein LAMP-1, and infected macrophages were examined using confocal microscopy. For each strain, over 100 intracellular bacteria were examined. 30.9 ± 6% of wild-type bacteria and 65.7 ± 6.7% of ssaV− bacteria were associated with LAMP-1. However, only 4.8 ± 2.6% of sifA− bacteria were associated with this marker (Fig. 5). These results are in broad agreement with the association of the same strains with LAMP-1 in RAW macrophages (Beuzón et al., 2000), indicating that sifA− bacteria are likely to be released from vacuoles in vivo.
To investigate this directly, infected splenocytes were permeabilized with streptolysin-O, a pore-forming toxin from Streptococcus pyogenes. After permeabilizing the plasma membrane, bacteria present in the cytosol will be accessible to an anti-Salmonella LPS antibody, whereas bacteria enclosed in a vacuole will be protected. Splenocytes, recovered 3 days after i.p. inoculation of mice with either 105 cfu of wild-type GFP-S. typhimurium or 106 cfu GFP-sifA− strain, were treated with streptolysin-O, fixed and stained for immunofluorescence in the absence of detergent. Most sifA− mutant bacteria were accessible to the anti-LPS antibody in contrast to the wild-type strain, with approximately 65% of bacterial cells protected from the antibody (Fig. 6). In all experiments, an anti-LAMP-1 antibody, detected by Cy5-conjugated secondary, was used as a control for permeabilization (not shown). These results indicate that the sifA− mutant bacteria are found in the cytosol of spleen macrophages, whereas the majority of wild-type bacteria are enclosed within a vacuole, strongly suggesting that SifA is involved in maintenance of the intracellular vacuolar membrane in vivo.
Despite intensive use of the S. typhimurium–mouse interaction as a model of human typhoid, the localization of replicating bacteria, particularly in the spleen, has been the subject of considerable debate. The importance of macrophages for S. typhimurium replication in vivo has been emphasized indirectly by macrophage depletion (Gulig et al., 1998; Wijburg et al., 2000). Microscopic studies have shown that S. typhimurium is present in CD18-positive liver cells (Richter-Dahlfors et al., 1997) and CD11b-positive splenocytes (Matsui et al., 2000). Both of these markers are expressed by, but are not restricted to, macrophages. The scavenger receptor and sialoadhesin markers detected here are apparently restricted to macrophages (Hughes et al., 1995; Crocker et al., 1991), but immature splenic dendritic cells can also express F4/80 (Leenen et al., 1998). Our work provides direct evidence showing that the majority of S. typhimurium bacteria reside within splenic macrophages, but it is also possible that a proportion of dendritic cells may be colonized. Indeed, a recent report demonstrates a role for dendritic cells in bacterial uptake by mucosal tissues as an alternative route to M cell invasion by S. typhimurium (Rescigno et al., 2001). It is possible that bacteria reach the spleen through dendritic cells, before colonizing resident marginal zone and red pulp macrophages. Flow cytometric analysis shows clearly that the vast majority of intracellular bacteria are accounted for by host cells expressing the three macrophage markers. Furthermore, as many as 35% of total spleen macrophages were infected with bacteria. We cannot exclude the possibility that a small percentage of blood-derived neutrophils and monocytes may be infected; however, the vast majority of bacteria are found within splenic tissue macrophages.
In addition to different anatomical distributions and different cell surface markers, there are also functional differences between splenic macrophages, particularly with respect to their phagocytic activity (Kraal, 1992; Hughes et al., 1995). In this regard, it is interesting that very few S. typhimurium bacteria were found within sialoadhesin-positive marginal metallophilic macrophages that are relatively non-phagocytic, and that a disproportionally high number were found in scavenger receptor/2F8-positive macrophages (Kraal, 1992; Hughes et al., 1995). Bacterial uptake by macrophages in vivo may therefore be as a result of the phagocytic capacity of the macrophage rather than the invasive potential of S. typhimurium. Alternatively, bacteria might not come into contact with marginal metallophilic macrophages as frequently as with others, or perhaps these cells do not provide an environment conducive to intracellular replication.
Confocal microscopic analysis of infected host cells revealed that a significant proportion carried more than 10 bacterial cells, and some were found with at least 50 bacteria per host cell. On the other hand, very few macrophages contained more than five SPI-2 mutant bacterial cells. This confirms that S. typhimurium undergoes intracellular replication in the spleen, and that this is dependent on the SPI-2 secretion system. Subsequent release of bacteria from infected splenic macrophages could be mediated by SPI-2-dependent apoptosis (Rathman et al., 1997; Adrianus et al., 2000). These bacteria could then be taken up by other cells resulting in further intracellular replication.
In addition to providing information on the ability of different strains to replicate in infected cells in vivo, we have also been able to exploit confocal microscopy to study the subcellular localization of S. typhimurium strains, as has been performed recently for Leishmania donovani (Lang et al., 2000). Our results confirm that the vacuolar membrane enclosing wild-type bacteria contains the lgp LAMP-1, and that SifA has a critical role in maintenance of this membrane in vivo.
Bacterial strains and growth conditions
In this study, we used S. typhimurium 12023s (wild type) and the following mutant strains: HH109 (ssaV::aphT in 12023s) and P3H6 (sifA::mTn5 in 12023s). The plasmid pFPV25.1 carrying gfpmut3A under the control of a constitutive promoter was introduced into bacterial strains for green fluorescence visualization (Valdivia and Falkow, 1997). Bacteria were grown in Luria–Bertani (LB) medium, supplemented with ampicillin (50 μg ml−1) for plasmid-containing strains.
Female BALB/c mice (18–20 g) (B and K Universal) were used for all infection studies. Mice were inoculated i.g. or i.p. with 0.1 ml and 0.2 ml of bacterial cells, respectively, in physiological saline, as described previously (Beuzón et al., 2000). The number of bacterial cfu in all inocula was confirmed using serial dilution. Different bacterial doses were used to inoculate mice, which were then sacrificed at the timepoints indicated in the Results. Between four and 10 mice were used for each experiment.
Preparation of spleen-derived cell suspensions
Spleens were removed aseptically and placed in 2 ml of ice-cold phosphate-buffered saline (PBS). Cell suspensions were obtained either by gentle physical disruption of the spleen using a bent needle, or by enzymatic digestion. For the latter, 2 ml of Hanks' balanced salt solution (HBSS) with 0.5% collagenase A (Boehringer Mannheim) and 0.002% DNase I (Boehringer Mannheim) were injected into the spleen using a 25G needle, which was then incubated for 1 h at 37°C with shaking (150 r.p.m.). Cell suspensions were filtered through a 70-μm nylon cell strainer (Becton Dickinson) and centrifuged at 400 g for 5–10 min. Red blood cells were subjected to ammonium chloride lysis and remaining cells were fixed in 1% paraformaldehyde (PFA) for 10 min on ice, washed twice and resuspended in PBS.
Antibodies and reagents
For lymphocytes, Phycoerythrin (PE)-conjugated anti-mouse CD3 (PharMingen) diluted 1:100, and undiluted PE-conjugated anti-mouse CD19 (Serotec) were used. The primary macrophage monoclonal antibodies used were the rat anti-mouse macrophage scavenger receptor 2F8, rat anti-mouse sialoadhesin/CD169 (Serotec) and rat anti-mouse F4/80 antigen (Serotec). These were used at a dilution of 1:50. Corresponding rat isotypes IgG2a and IgG2b (Sigma) were used at appropriate concentrations. A rabbit anti-LAMP-1 polyclonal antibody was used at a dilution of 1:1000. Goat anti-Salmonella polyclonal antibody (Kirkegaard and Perry Laboratories) was used at a dilution of 1:400. Biotin conjugated F(ab)2 anti-rat IgG (Jackson Immunoresearch Laboratories) was used at a dilution of 1:500. PE-conjugated Streptavidin (Jackson Immunoresearch Laboratories) was used at a dilution of 1:100, and secondary Texas red (TR)-conjugated anti-rabbit and anti-goat antibodies (Jackson Immunoresearch Laboratories) were used at a dilution of 1:400 and secondary cyanine (Cy5)-conjugated anti-rabbit was used at a dilution of 1:400.
Flow cytometry and immunofluorescence microscopy
Isolated splenocytes (105−106 cells) were stained immediately for flow cytometric analysis. Spleens infected with S. typhimurium 12023s (without the GFP-expressing plasmid) were used as controls for autofluorescence in all experiments. To block non-specific staining, cells were incubated with either 10% donkey serum in PBS for the antimacrophage antibodies or 10% rat serum for the antilymphocyte antibodies. Sequential incubations (30 min on ice) were carried out in blocking buffer with the macrophage primary antibodies or appropriate isotypes, a biotin conjugated F(ab)2 and finally PE-conjugated streptavidin. Cells were washed in cold PBS between each step and prior to analysis. For each sample, between 10 000 and 50 000 cells were analysed on a FACScalibur cytometer (Becton Dickinson). Cellular fluorescence was excited at 488 nm with an Argon laser. GFP was detected at 525 nm in the FL1 channel and PE fluorescence at 585 nm in the FL2 channel. Data were analysed with WinMDI software version 2.8.
To prepare coverslips for immunofluorescence microscopy, 5 μl of spleen cells were placed on poly l-lysine coated coverslips and left to air-dry. Coverslips were stained at room temperature as above, mounted with Prolong Antifade reagent (Molecular Probes) and analysed using a confocal laser scanning microscope (LSM510, Zeiss).
An SCV was considered positive for anti-LAMP1 if it was detected as a ring or partial ring around green fluorescent bacteria, and if the marker was concentrated in this area compared with the immediate surroundings.
Quantification of intracellular and extracellular number of bacteria
For quantification of intracellular and extracellular bacteria, coverslips were prepared after enzymatic digestion of spleens from infected mice. The resulting cell suspension was immediately fixed in 1% PFA without any centrifugation steps. These coverslips were incubated with anti-Salmonella primary antibody and then with TR-conjugated secondary antibody. Antibodies were diluted in 10% horse serum in PBS. Control experiments were carried out to confirm that washes did not remove any detectable extracellular bacteria.
Streptolysin-O permeabilization of splenocytes
Splenocytes were recovered from infected mice by physical disruption of the spleen, in 2 ml of ice-cold serum-free Dulbecco's modified Eagle medium (DMEM), and the volume was adjusted to 25 ml. The suspension was centrifuged at 400 g for 5 min and the pellet resuspended in serum-free DMEM to a concentration of approximately 107 cells ml−1. One millilitre of the suspension was aliquoted into microfuge tubes. Next, 50 μl of preactivated (incubated at 37°C for 10 min in the presence of 4 mM DTT) Streptolysin-O (SLO; Corgenix) stock was added to a final concentration of 1 U ml−1. Tubes were inverted twice, and incubated on ice for 10 min. To remove unbound SLO, cells were washed twice with 1 ml of ice-cold serum-free DMEM by centrifuging at 400 g for 3 min at 4°C. Cells were resuspended in 1 ml of warm intracellular transport (ICT) buffer [50 mM HEPES-KOH pH 7.1, 4 mM MgCl2, 10 mM EGTA, 8.4 mM CaCl2, 78 mM KCl, 1 mM DTT, 1 mg ml−1 bovine serum albumin (BSA)] for 5 min at 37°C to initiate pore formation. Splenocytes were washed with fresh ICT buffer, fixed in 3% PFA for 15 min, washed for three times with PBS and added to poly-l-lysin coated coverslips and left to air-dry. Coverslips were treated for immunofluorescence as described above by triple labelling, in the absence of detergent. Bacteria were detected by GFP expression. Goat anti-Salmonella polyclonal antibody was used to detect bacteria in the cytosol, and anti-LAMP-1 antibody was used as control for plasma membrane permeabilization.
We thank Professor Siamon Gordon for providing the 2F8 antibody and Dr Stéphane Méresse for the anti-LAMP-1 antibody. We are grateful to members of our laboratory for critically reviewing the manuscript. We thank Hans Stauss for help with flow cytometric analysis. This work was supported by a Fellowship from the Portuguese Foundation for Science and Technology to S.S. and grants from the MRC (UK) to D.W.H and M.N.