The Salmonella pathogenicity island 1 secretion system directs cellular cholesterol redistribution during mammalian cell entry and intracellular trafficking

Authors

  • Matthew J. Garner,

    1. University of Cambridge, Department of Pathology, Tennis Court Road, Cambridge, CB2 1QP, UK.
    Search for more papers by this author
    • †The first two authors contributed equally to this study.

  • Richard D. Hayward,

    1. University of Cambridge, Department of Pathology, Tennis Court Road, Cambridge, CB2 1QP, UK.
    Search for more papers by this author
    • †The first two authors contributed equally to this study.

  • Vassilis Koronakis

    1. University of Cambridge, Department of Pathology, Tennis Court Road, Cambridge, CB2 1QP, UK.
    Search for more papers by this author

*For correspondence. E-mail vk103@mole.bio.cam.ac.uk; Tel. + 44 (0)1223 333715; Fax + 44 (0)1223 333327.

Summary

The bacterial pathogen Salmonella triggers its own uptake into non-phagocytic mammalian cells. Entry is induced by the delivery of bacterial effector pro-teins that subvert signalling and promote cytoskeletal rearrangement, although the molecular mechanisms that co-ordinate initial pathogen-host cell recognition remain poorly characterized. Here we show that cholesterol is essential for Salmonella uptake. Depletion and chelation of plasma membrane cholesterol specifically inhibited bacterial internalization but not adherence. Cholesterol accumulated at bacterial entry sites in cultured cells, and was retained by Salmonella-containing vacuoles following pathogen internalization. Cellular cholesterol redistribution required bacterial effector protein delivery mediated by the Salmonella pathogenicity island (SPI) 1 type III secretion system, but was independent of the SPI2-encoded system.

Introduction

Salmonella species are pathogens of man and animals, most commonly causing severe gastroenteritis. Bacterial entry into non-phagocytic intestinal epithelial cells is an essential early event in pathogenesis (Finlay and Falkow, 1997), a process that can be approximated in vitro using cultured cells (Finlay and Falkow, 1990). In contrast to the comparatively well-characterized ‘zipper’ uptake mechanism utilized by Listeria and Yersinia, that involves binary interactions between a bacterial surface protein and a distinct eukaryotic receptor, Salmonella employs the complex multifactorial ‘trigger’ mechanism (Finlay and Cossart, 1997). The Salmonella pathogenicity island 1 (SPI1) is required for this process. It encodes several bacterial invasion effector proteins, which upon contact with the host cell, undergo ‘type III’ export from the bacterial cytosol and are directly delivered into the eukaryotic cell plasma membrane or cytosol (Galán and Collmer, 1999). These effectors subvert cell signalling and modulate actin dynamics both directly and indirectly to induce membrane deformation (‘ruffling’) at the point of bacterial contact (McGhie et al., 2001; Zhou et al., 2000; Norris et al., 1998; Hardt et al., 1998). This ultimately drives pathogen internalization into a membrane-bound vacuole. However, while the function of some effectors has been established, the determinants that direct initial pathogen-host recog-nition and define the location of entry foci remain elusive. Although receptors have been proposed (Pace et al., 1993; Pier et al., 1998), these are not expressed by all the mammalian cell types entered by Salmonella species (Jones et al., 1993). It is therefore likely that more ubiquitous proteinacious or lipidic factor(s) exist in the target cell plasma membrane.

Dynamic cholesterol-glycosphingolipid enriched membrane microdomains (rafts) have emerged as central regulators of diverse cellular processes (Simons and Ikonen, 1997; Brown and London, 1998). These domains act as concentration and partitioning platforms for receptors and signal transduction molecules that link the plasma membrane, the cytoskeleton and intracellular signalling pathways (Oliferenko et al., 1999; Simons and Toomre, 2000). Many bacterial and viral pathogens utilize these regions to enter eukaryotic cells, and to subvert trafficking pathways to promote their intracellular replication (Rosenberger et al., 2000; van der Goot and Harder, 2001). Bacterial examples include: pathogenic FimH expressing E. coli strains which bind CD48 in micro-domains to trigger their internalization into vacuoles (Baorto et al., 1997; Shin et al., 2000); Mycobacterium bovis BCG which directly bind cholesterol, a function essential for entry into and survival within macrophages (Ferrari et al., 1999; Gatfield and Pieters, 2000); and Campylobacter jejuni and Chlamydia trachomatis which apparently require cholesterol for entry into cultured cells (Wooldridge et al., 1996; Norkin et al., 2001).

Salmonella-induced membrane ruffling is associated with submembranous actin polymerization and the activation of cellular GTPases. Provocatively, micro-domains are preferred platforms for membrane-linked actin polymerization (Rozelle et al., 2000) and concentration of Rho GTPases (Gingras et al., 1998; Johnson, 1999; Michaely et al., 1999; Castellano et al., 2000), and microdomain-associated α5β1 integrins, CD44 and MHC class I cluster at Salmonella entry foci (Garcia-Del Portillo et al., 1994). We therefore examined the role of cholesterol in Salmonella entry into cultured cells.

Results

Cholesterol is required for Salmonella entry into cultured mammalian cells

To initially investigate whether cholesterol contributes to the Salmonella entry mechanism, cellular cholesterol levels were reduced using methyl-β-cyclodextrin (MβCD), a water-soluble cyclic heptasaccharide that forms soluble inclusion complexes with cholesterol (Ohtani et al., 1989; Kilsdonk et al., 1995) and has been shown to efficiently extract plasma membrane cholesterol from cultured cells (Klein et al., 1995). MβCD-treated and control cultured HeLa cells were infected with wild-type S. typhimurium, and relative bacterial uptake was com-pared. The results revealed a dramatic (>90%) reduction in Salmonella entry into cholesterol-depleted cells (Fig. 1A, upper left). To confirm visually that depleted cells were compromised in their ability to internalize bacteria, control and MβCD-treated HeLa cells were infected with S. typhimurium constitutively expressing green fluorescent protein (GFP) and examined by fluorescence confocal microscopy. Consistent with the gentamicin protection assays, few bacteria were observed in association with, or internalized within, cholesterol-depleted cells, whereas many bacteria were associated with actin-rich ‘membrane ruffles’ or were internalized in control cells (compare MβCD to control, Fig. 1A, right). To determine whether bacterial adherence, internalization or both were inhibited as a result of cellular cholesterol-depletion, internalized and adherent (bound extracellular) bacteria were discriminated directly from reconstructed, confocal optical xz-sections, and the number of adherent bacteria was scored (Fig. 1B, right). Comparison of cholesterol-depleted and control cells showed that MβCD-treatment did not significantly inhibit Salmonella adherence to cultured cells (Fig. 1B, left).

Figure 1.

Cholesterol is required for Salmonella entry into cultured HeLa cells.

A. Effect of cholesterol depletion on Salmonella entry. Upper left: Entry of S. typhimurium SJW1103 into control (open bars) or cholesterol-depleted (filled) cultured HeLa cells as a function of invasion time. Invasion is expressed as a percentage of wild-type bacteria internalized by control cells after 60 min. Data shown is from a single trial, and representative of five independent experiments. Right: Fixed control and depleted (MβCD) HeLa cells 60 min after infection with wild-type S. typhimurium SJW1103 expressing green fluorescent protein (GFP) stained with Texas Red-conjugated phalloidin to visualize F-actin, viewed by confocal microscopy. Scale bar, 30 μm. Inset shows co-localization (yellow immunofluorescence) of bacteria and F-actin at entry foci. Scale bar, 10 μm.

B. Effect of cholesterol depletion on Salmonella adherence. Left: Mean bacterial adherence to control or cholesterol-depleted (MβCD) cultured HeLa cells fixed after 60 min. Adherent bacteria were scored from confocal analysis of 20 random fields of view under each condition. Right: Examples of internalized (in) and adherent (extracellular, ex) bacteria, distinguished using reconstituted confocal optical xz-sections.

C. Effect of MβCD-treatment on cellular cholesterol levels and distribution. Left: Quantification of cellular cholesterol following MβCD-treatment. Percentage cellular 3H-cholesterol remaining as a function of MβCD-treatment time. Inset shows mean number of viable (open bars) and non-viable (filled) MβCD-treated (MβCD, 10 mM, 60 min) and control HeLa cells per field of view (Control 10.9 ± 0.82 viable, 0.18 ± 0.12 non-viable; MβCD 12.0 ± 1.01, 0.2 ± 0.08, n = 20 fields >400 cells) Right: Distribution of cellular cholesterol in fixed control and depleted (MβCD) cultured HeLa cells stained with the cholesterol-binding compound filipin and Texas Red-conjugated phalloidin to visualize F-actin. Scale bar, 35 μm.

In agreement with previous findings (Francis et al., 1999; Yancy et al., 1996), radiolabelling assays verified that MβCD-treatment extracted ~70% of total cellular cholesterol after 60 min (Fig. 1C, left). When cholesterol was directly visualized with the cholesterol-binding compound filipin using low-dosage ultraviolet fluorescence confocal microscopy (Bornig and Geyer, 1974; Drabikowski et al., 1973), it was dispersed throughout the HeLa cell plasma membrane, was enriched in intracellular stores and concentrated at focal adhesions in control cells. In contrast, cholesterol was located predominantly in intracellular stores after depletion with MβCD (Fig. 1C, right). This is consistent with the ability of MβCD to efficiently extract cholesterol from the plasma membrane (Klein et al., 1995). It has been established that MβCD-treatment does not arrest all phagocytic events (Rodal et al., 1999), block the entry of BSA-coated latex beads into HeLa cells (Norkin et al., 2001) or inhibit SPI1-independent Salmonella engulfment by cultured macrophages (Gatfield and Pieters, 2000).

Salmonella entry was not artificially reduced as a consequence of cell loss or death, as MβCD-treatment had no significant effect on cell number or viability (Fig. 1C, left inset and Fig. 1A, right), although after depletion cells exhibited a ‘rounder’ morphology (Fig. 1A, right and Fig. 1C, right). If Salmonella internalization was inhibited specifically as a consequence of cellular cholesterol depletion, cholesterol recovery should restore bacterial entry. The detailed kinetics of recovery in a range of cultured cell lines has been previously examined (Rodal et al., 1999; Francis et al., 1999; Gustavsson et al., 1999). Membrane cholesterol can be restored to depleted cells directly by the addition of serum or cholesterol-cyclodextrin complexes to the medium, or as a result of cholesterol biosynthesis under serum-free conditions. In both cases, similar recovery kinetics are observed (Rodal et al., 1999). We thus investigated whether MβCD-directed inhibition of Salmonella entry was reversible. Cultured cells were depleted as previously and recovered in medium with or without serum prior to Salmonella infection. Under both conditions, inhibition of Salmonella entry was partially overcome with 4–8 h recovery, and was completely reversed after 12–22 h, when mean bacterial entry was restored to 90% of control values.

In parallel experiments, HeLa cells were also treated with filipin which chelates rather than extracts cholesterol. Filipin-treated and control HeLa cells were infected with wild-type S. typhimurium, and relative bacterial uptake compared. As with MβCD-treatment, an 88% reduction in mean bacterial entry (after 60 min) into filipin-treated cells was observed. As previously, these results correlated with the number of internalized bacteria observed when filipin-treated and control cells were visualized directly by fluorescence confocal microscopy (not shown). Taken together, the combined data show that cholesterol is required for Salmonella entry into cultured HeLa cells, although it is not essential for bacterial adherence to target cells.

SPI1-dependent cholesterol relocation and accumulation at Salmonella entry foci

To further analyze the role of cholesterol in the Salmonella entry process, the distribution of cellular cholesterol was examined during bacterial uptake. HeLa cells were infected with S. typhimurium expressing GFP, and cholesterol and filamentous actin (F-actin) visualized simultaneously with filipin and Texas-Red conjugated phalloidin respectively. When Salmonella-infected cells were ex-amined by low-dosage ultra violet fluorescence confocal microscopy, cholesterol relocated and accumulated at the sites of bacterial entry in every infected cell (Fig. 2A). Parallel analysis of reconstructed confocal xz and yz optical sections allowed the various stages of the bacterial entry process to be distinguished (as shown Fig. 1B, right). This more detailed approach revealed cholesterol to specifically accumulate and co-localize with the dramatic F-actin rearrangements which form the entry foci that envelop invading bacteria (Fig. 2A, indicated with arrow inv; co-localization appears as magenta immunofluorescence in merge). Cholesterol also surrounded and co-localized with internalised bacteria (Fig. 2A, indicated with arrow int), and frequently aggregated beneath adherent (extra-cellular) bacteria prior to the initiation of cytoskeletal rearrangements (Fig. 2A, indicated with arrow adh). These observations suggest that Salmonella utilizes cholesterol-rich regions of the plasma membrane to enter cultured HeLa cells.

Figure 2.

SPI1-dependent cholesterol redistribution during Salmonella entry into cultured HeLa cells.

A. Cholesterol aggregation and redistribution at Salmonella entry foci. Cultured HeLa cell fixed after infection with wild-type S. typhimurium SJW1103 expressing GFP. Cells were stained with filipin (blue fluorescence) and Texas Red-conjugated phalloidin to visualize cholesterol and F-actin, respectively, and viewed by low-dosage ultraviolet fluorescence confocal microscopy. Arrows indicate representative adherent (adh), invading (inv) and internalized (int) bacteria identified from reconstructed optical sections (as shown Fig. 1B, right). For clarity, filipin signal is also shown in greyscale. Frame is representative of >100 cells observed in three independent experiments. Scale bar, 10 μm.

B. Cholesterol redistribution is not a direct consequence of generalized membrane accumulation around invading Salmonella. Typical cultured HeLa cells fixed after infection with wild-type S. typhimurium SJW1103 expressing GFP. Cells were stained with TMA-DPH and Texas Red-conjugated phalloidin to visualize membranes and F-actin, respectively, and viewed by low-dosage ultraviolet fluorescence confocal microscopy. Scale bar, 15 μm. The inset shows reconstituted confocal optical xz-sections demonstrating that under these experimental conditions TMA-DPH localizes predominantly to the HeLa cell plasma membrane.

C. Cholesterol aggregation and redistribution requires the SPI1 type III secretion system. Cultured HeLa cell fixed after infection with S. typhimurium invG mutant. Cells were stained with filipin (blue fluorescence) and bacteria with anti-Salmonella antibody/AlexaFluor 594-conjugated antirabbit IgG (red immunofluorescence). Arrows indicate positions of adherent bacteria. For clarity, filipin signal is also shown in greyscale. Frame is representative of >100 cells observed in three independent experiments. Identical results were obtained with S. typhimurium sipB mutants. Scale bar, 15 μm.

However, the observed filipin staining may result either from active redistribution of cholesterol to the entry foci, or as an indirect consequence of generalized membrane accumulation around invading Salmonella. To distinguish these possibilities, we used the fluorescent lipid analogue 1-[4-(trimethylamino)phenyl]-6-phenylhexa-1,3,5-triene (TMA-DPH), a commonly employed membrane-selective probe, to label the plasma membrane. TMA-DPH partitions uniformly without tropism for particular lipids in vitro (Florine-Casteel and Feigenson, 1988). Labelling of the HeLa cell plasma membrane was generally uniform under the conditions used (Fig. 2B and inset), although occasional local variations in intensity were observed. An increase in TMA-DPH fluorescence occurred in infected cells, consistent with the formation of membrane ruffles and the associated increase in membrane surface area (Fig. 2B). However, TMA-DPH labelling did not specifically coincide with bacterial entry foci, or reach the extent observed previously with filipin (compare Fig. 2A,B). The difference between the filipin and TMA-DPH staining patterns indicates that general membrane accumulation does not account for the specific cholesterol enrichment at entry foci.

Since cellular cholesterol re-localizes and accumulates at bacterial entry foci [compare control uniform distri-bution (Fig. 1C, right) to that during bacterial infection (Fig. 2A)], we investigated whether this is entirely a cellular response, for example as a result of bacterial adherence or exposure to lipopolysaccharide, or requires active cross-talk between invasive Salmonella and its target cell. It is widely established that such active interplay is directed by bacterial effector proteins delivered to the host cell via the SPI1 type III secretion system (Galán and Collmer, 1999). To distinguish SPI1-dependent and –independent responses, HeLa cells were infected with a S. typhimurium invG deletion mutant, which lacks the outer membrane exit channel of the SPI1 type III export apparatus (Crago and Koronakis, 1998). Bacteria lacking invG retain the ability to adhere to, but are unable to enter, host cells (Kaniga et al., 1994). Cellular cholesterol was never observed to aggregate or accumulate in the vicinity of these mutant bacteria (Fig. 2C), which as expected could not induce host cell cytoskeletal rearrangements. Identical results were obtained with a S. typhimurium sipB deletion mutant, which although export competent, is unable to deliver effectors into the target cell membrane or cytosol (Collazo and Galán, 1997). These combined data suggest that the observed alterations in HeLa cell cholesterol distribution are not a pathogen-independent cellular response, but require cross-talk mediated by the SPI1 type III secretion system.

Cholesterol-rich cellular protrusions are induced during Salmonella entry into NIH3T3 fibroblasts

Although HeLa cells are frequently used to study Salmonella host–cell interactions (Garcia-Del Portillo et al., 1994; Finlay et al., 1991; Méresse et al., 1999), it was important to establish whether entry-associated cholesterol redistribution was cell line-specific or a more general phenomenon. NIH3T3 fibroblasts have been used to study Salmonella entry (Jones et al., 1993), cellular responses to bacterial effector proteins (McGhie et al., 2001; Tran Van Nhieu et al., 1999) and Rho GTPase signalling (Ridley and Hall, 1992). Initial assays con-firmed that Salmonella entered NIH3T3 and HeLa cells with similar efficiency, and that bacterial entry into NIH3T3 cells was also inhibited by MβCD treatment (not shown). Examination of Salmonella-infected NIH3T3 cells revealed remarkable cholesterol-rich entry-associated surface protrusions, which extended beyond the level of the cell surface (Fig. 3, upper panels). Confocal z-planes were used to reconstruct a three-dimensional model of these novel structures (Fig. 3, centre and lower panels). Their cholesterol-rich exterior enveloped a central ‘cup’ of F-actin that in turn surrounded the invading bacteria.

Figure 3.

Cholesterol redistribution during Salmonella entry into cultured NIH3T3 cells.

Upper: Typical cultured NIH3T3 fibroblast fixed after infection with wild-type S. typhimurium expressing GFP. Cells were stained with filipin (blue fluorescence) and Texas Red-conjugated phalloidin to visualize cholesterol and F-actin, respectively, and viewed by low-dosage ultraviolet fluorescence confocal microscopy. Left panel: single merged horizontal (xy) and vertical (yz-plane, left; xz-plane, bottom) confocal sections. Right panel: filipin signal in greyscale (for clarity) with the position of the nucleus (N), xz-, and yz-sections (left) superimposed. The area selected for further image analysis is indicated (box). Scale bar, 20 μm.

Lower: Three dimensional reconstruction of the cholesterol-actin rich entry-associated surface protrusion. Image (centre) was assembled from confocal xy sections taken in different z-planes ranging from the top (1) to the base of the structure (12). Upper panels correspond to z-plane 8; cellular actin occurs in z-plane 12. Scale bar, 10 μm.

These structures could also be visualized by scanning electron microscopy (SEM). NIH3T3 cells infected with wild-type bacteria exhibited dramatic membrane ruffling, and entry-associated surface protrusions were frequently observed (Fig. 4, upper panel indicated by arrow). The structures were typically enclosed, obscuring the internalized bacteria visualized previously by immunofluorescence. As suggested by the confocal z-planes (Fig. 3, lower panels 7–11) and the three-dimensional image reconstruction (Fig. 3, centre panel), they were indeed composed of densely folded membrane extensions. Consistent with the HeLa cell immunofluorescence data, no protrusions were observed by SEM on NIH3T3 cells infected with invG (or sipB, not shown) mutants (Fig. 4, centre panel), or on uninfected controls (Fig. 4, lower panel). The identification and characterization of these cell entry structures indicates that SPI1-mediated cholesterol redistribution is likely to be a general hallmark of the Salmonella entry process.

Figure 4.

Formation of entry-associated surface protrusions on NIH3T3 fibroblasts requires the SPI1 type III secretion system. Cultured NIH3T3 fibroblasts viewed by scanning electron microscopy following infection with S. typhimurium wild-type (wt) or invG mutants (invG). Uninfected cells are shown for comparison (control). Arrow indicates an entry-associated surface protrusion similar to that visualized by confocal microscopy (Fig. 3). Bacteria and cultured cells are pseudocoloured green and blue respectively. Scale bar, 5 μm.

Salmonella-containing vacuoles are enriched with cholesterol

Internalized S. typhimurium reside and replicate within a unique cytoplasmic organelle termed the Salmonella-containing vacuole (SCV). SCVs initially acquire markers characteristic of early and recycling endosomes, which are subsequently lost and replaced with lysosomal glycoproteins and the vacuolar ATPase (Steele-Mortimer et al., 1999). SCVs subvert cellular trafficking by avoiding fusion with late endosomes or prelysosomes (Méresse et al., 1999). Although SPI1 effectors are secreted after bacterial internalization (Collazo and Galán, 1997; Mukherjee et al., 2001), an additional type III secretion system and associated effector proteins encoded on a separate pathogenicity island (SPI2) are the major determinants of intracellular survival (Hensel, 2000).

As SCVs apparently retained cholesterol immediately after HeLa cell entry (Fig. 2A, indicated with arrow int), we investigated whether this high cholesterol content was subsequently maintained. After infection with S. typhimurium expressing GFP, extracellular bacteria were removed from cell monolayers by repeated washing and a chase in gentamicin-supplemented medium. Incubation was then continued (2–6 h), and cholesterol and F-actin visualized by fluorescence confocal microscopy. By 2 h post invasion, internalized bacteria had migrated from the cell periphery and as expected were no longer associated with condensed F-actin characteristic of early entry foci (Figs 5, 2 h). Bacteria-proximal membranes were efficiently and specifically labelled with filipin, indicating that the SCVs retained high levels of cholesterol. These structures no longer co-localized with F-actin and were distinct from the F-actin-cholesterol rich regions at cell boundaries (Fig. 5, 2 h). Early endosomal and plasma membranes are suggested to share similar overall lipid composition (Kobayashi et al., 1998), we therefore ex-pected a change in the SCV cholesterol composition upon interaction with late endocytic compartments (>90 min; Brumell et al., 2001a). However, after 4 h, cholesterol continued to localize around and frequently co-localized with replicating internalized bacteria (Fig. 5, 4 h), and by 6 h bacterial microcolonies were apparently enclosed within a cholesterol-rich cellular compartment (Fig. 5, 6 h and 6 h xz). As previously, this was a general phenotype, as cholesterol-rich SCVs also formed after Salmonella entry into NIH3T3 cells (not shown).

Figure 5.

Salmonella-containing vacuoles are enriched with cholesterol.

Upper: Cultured HeLa cells fixed after infection with S. typhimurium expressing GFP for 1 h and subsequent incubation for the times indicated (2–6 h). Cells were stained with filipin (blue fluorescence) and Texas Red-conjugated phalloidin to visualize cholesterol and F-actin, respectively, and viewed by low-dosage ultraviolet fluorescence confocal microscopy. Scale bar, 10 μm. The inset shows a typical cholesterol-rich cellular compartment (blue fluorescence) 6 h after bacterial entry.

Lower: Greyscale images of the filipin channels from upper panels, for clarity. Arrows indicate cholesterol sequestered around internalized bacteria.

When cholesterol-depleted HeLa and NIH3T3 cells were identically infected, only a small percentage of the bacterial inoculum entered (Fig. 1). Nevertheless, although fewer in number, the SCVs present also labelled with filipin 2 h post internalization (Fig. 6). Although discernible in HeLa cells, this was more evident in NIH3T3 cells which react more dramatically in the absence of MβCD-treatment (compare Fig. 6, 3T3 and HeLa). These data indicate that Salmonella also utilize the available cholesterol in depleted cells, and thus do not enter via some atypical mechanism.

Figure 6.

Salmonella-containing vacuoles are enriched with cholesterol in depleted cells. Cultured cholesterol-depleted NIH3T3 fibroblasts (upper) and HeLa cells (lower) fixed after infection with S. typhimurium expressing GFP for 60 min and a subsequent 2 h incubation. Cells were stained with filipin (blue fluorescence) and Texas Red-conjugated phalloidin to visualize cholesterol and F-actin, respectively, and viewed by low-dosage ultraviolet fluorescence confocal microscopy. Arrows indicate sequestered cholesterol also shown for clarity in greyscale (insets). Scale bars, 5 μm (upper) and 2 μm (lower).

SPI2 effectors do not influence the cholesterol content of Salmonella-containing vacuoles

Effectors delivered via the SPI2 type III secretion system have been implicated in the modification of the SCV and the subversion of intracellular trafficking following bac-terial internalization (Beuzón et al., 2000; Vazquez-Torres et al., 2000; Uchiya et al., 1999). To establish whether SPI2 effectors influenced the cholesterol content of SCVs, cultured HeLa cells were infected with a S. typhimurium ssaV deletion mutant, which lacks an essential component of the SPI2 type III export apparatus, but remains invasion competent (Shea et al., 1999). After infection with S. typhimurium ssaV expressing GFP, extracellular bacteria were removed from cell monolayers by repeated washing and a chase in gentamicin-supplemented medium as previously. Incubation was then continued (2–6 h), and cholesterol visualized by fluorescence confocal microscopy. As with the wild-type bacteria, bacteria-proximal membranes were again efficiently and specifically labelled with filipin throughout the time course (Fig. 7), demonstrating that the SPI2 type III secretion system is not required to maintain the high cholesterol content of SCVs. The combined findings suggest that the cholesterol central to the biogenesis of entry foci is subsequently retained by SCVs, although this could be additionally supplemented with cholesterol from internal cellular stores as a result of intracellular trafficking.

Figure 7.

The SPI2 type III secretion system is not required to maintain the high cholesterol content of Salmonella-containing vacuoles. Cultured HeLa cells fixed after infection for 60 min with S. typhimurium ssaV (SPI2) mutant expressing GFP and subsequent incubation for the times indicated (2–6 h). Cells were stained with filipin (blue fluorescence) to visualize cholesterol, and viewed by low-dosage ultraviolet fluorescence confocal microscopy. Arrows indicate sequestered cholesterol, also shown for clarity in greyscale. Scale bar, 5 μm.

Discussion

We have employed immunofluorescence microscopy and agents that sequester or chelate cholesterol to investigate the possible role of plasma membrane cholesterol in the Salmonella entry mechanism. These studies revealed that profound cholesterol redistribution occurs during Salmonella entry into non-phagocytic cells, and suggest that this may be a requirement for the process. This effect is evidently pathogen-induced, as cholesterol redistribution requires bacterial-target cell cross-talk mediated by effector proteins delivered into the host cell membrane or cytosol via the SPI1 type III secretion system. The following evidence supports these conclusions: (i) the defect in Salmonella cell entry following depletion or chelation of plasma membrane cholesterol; (ii) the rescue of the entry defect upon recovery of cellular cholesterol levels; (iii) the physical redistribution and aggregation of cholesterol at bacterial entry foci in two cultured cell models; (iv) the finding that Salmonella SPI1 mutants do not induce cholesterol accumulation.

Our observations that SCVs are enriched with cholesterol are in agreement with the recent data of Brumell et al. (2001a), but extend their findings. The time-course from 2 to 6 h after entry (Fig. 5) suggests that cholesterol-rich SCVs derive directly from the plasma membrane at the entry foci and subsequently maintain their high cholesterol content. Initial cholesterol redistribution is clearly dependent upon SPI1 effector proteins (Figs 2 and 4), which continue to be translocated up to four hours after bacterial internalization (M. Garner and R. Hayward, unpublished; Collazo and Galán, 1997). SPI2 effectors have been proposed to direct SCV biogenesis (Brumell et al., 2001a,b; Beuzón et al., 2000; Uchiya et al., 1999), although the relative contribution of SPI1 and SPI2 effectors to the complex SCV maturation process remains unknown. Our data (Fig. 7) show that the high cholesterol content of SCVs can be maintained independently of SPI2 function, possibly by as yet unidentified SPI1 effector(s) which remain functional after bacterial internalization. This might support the view that the functions of some SPI1 effectors may not be limited to the initial entry event as previously supposed, but may also influence early SCV trafficking (Mukherjee et al., 2001), while the activities of both SPI1 and SPI2 effectors are required for SCV biogenesis.

How might cholesterol participate in Salmonella cell entry? Cholesterol redistribution in response to bacterial stimuli may trigger specific cellular signalling molecules to relocate, generating a membrane platform from which bacterial subversion of cellular functions can continue. Pathogen tropism for particular regions of the target cell plasma membrane may ensure that a discrete set of cellular proteins are rapidly recruited to bacterial entry foci, which may be essential downstream targets for delivered bacterial effector proteins. Conversely, removal of membrane cholesterol may indirectly disrupt the function of other unknown cellular components that participate in the entry process. Nevertheless, a number of putative mechanisms can be proposed by which SPI1 effectors may initiate cholesterol redistribution. Effectors that are ini-tially inserted into the target cell plasma membrane may promote lipid phase changes, initiating microdomain aggregation and transient curved membrane configurations (Epand, 1998). This may also facilitate the formation of the dramatic membrane protrusions associated with filopodial and lamellipodial structures. Effector proteins delivered into the cell cytosol may generate secondary messengers or initiate cytoskeletal rearrangements, signals known to influence the movement and composition of cholesterol-rich microdomains (Simons and Toomre, 2000). Cholesterol rearrangements may not be triggered by any single effector, especially since it has emerged that entry-associated actin polymerization is induced by the co-ordinated direct and indirect activities of at least four SPI1 effectors (Hayward and Koronakis, 2002).

Shigella also uses the ‘trigger’ mechanism to enter non-phagocytic cells (Finlay and Cossart, 1997), a process that similarly requires the delivery of invasion plasmid antigens (Ipas) via a type III secretion system (Tran Van Nhieu and Sansonetti, 1999). IpaB has been proposed to interact with microdomain-associated CD44 and α5β1 integrins (Watarai et al., 1996; Skoudy et al., 2000), which also concentrate at Salmonella entry foci (Garcia-Del Portillo et al., 1994). It is not known whether cholesterol is similarly required for Shigella entry. Receptor aggregation may therefore either occur directly due to interaction with a specific effector, or indirectly as a result of pathogen-induced cholesterol redistribution. Whether these cellular proteins are ‘receptors’ or ‘bystanders’ therefore remains controversial.

Further analysis will address whether Salmonella specifically subverts subsets of cholesterol-rich micro-domains during internalization, identify the effector pro-tein(s) involved and address the role of cholesterol in the entry process. Such studies will not only allow further insights into the complex bacterial internalization mechanism, but also advance our understanding of cellular microdomain dynamics and the early events that underlie the regulation of phagocytosis.

Experimental procedures

Bacterial strains and mammalian cell culture

S. typhimurium SJW1103 (Yamaguchi et al., 1984), TNP-5 invG::TnphoA and its isogenic wild-type TNP-5 (Crago and Koronakis, 1998), CS469 ΔsspB and its isogenic wild-type ATCC 14028 (gifts from Samuel I. Miller), and HH109 ssaV [gift from David Holden (isogenic wild-type ATCC14028)], were transformed with pFPV25.1 (gift from Stanley Falkow) for constitutive green fluorescent protein (GFP) expression where appropriate. Bacteria were maintained on Luria-Bertani agar or in tryptone yeast (TY) medium with additional ampicillin (50 μg l–1) and/or chloramphenicol (20 μg l–1) when appropriate. HeLa and NIH3T3 cells were routinely cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) fetal calf serum (FCS), L-glutamine and penicillin/streptomycin (Sigma or GibcoBRL) at 37°C with 5% CO2.

Cholesterol depletion and recovery

Cultured HeLa or 3T3 cells were washed with phosphate-buffered saline (PBS) and incubated in DMEM containing 10 mM methyl-β-cyclodextrin (Sigma, MβCD; 1 h, 37°C, 5% CO2). To quantify the rate and extent of cholesterol depletion, 1 × 104 cells were labelled with 0.5 μCi [1α,2α(n)-3H]cholesterol (Amersham; 20 h, 37°C, 5% CO2). Cells were washed three times in PBS, incubated with DMEM containing 10 mM MβCD as previously for 15, 30, 45 and 60 min, and lyzed (2% (v/v) NP40, 0.2% (v/v) SDS in distilled water). Samples were clarified by centrifugation (13 000 g, 10 min, 16°C), thoroughly mixed with excess scintillant (Optiphase, Hisafe 2, Wallac), and counted in a scintillation counter. Alternatively, cultured HeLa cells were washed with phosphate-buffered saline (PBS) and incubated in DMEM containing 5 μgl–1 filipin (Sigma; 15 min, 37°C, 5% CO2). Filipin was also added to infection medium to chelate cholesterol subsequently trafficked to the plasma membrane. Filipin-treatment has no effect on the entry of BSA-coated microspheres, regardless of whether they were preincubated with anti-BSA antibodies (Norkin et al., 2001).

For recovery experiments, cells were MβCD-treated, washed and incubated 2–22 h (37°C, 5% CO2) in DMEM with or without 10% (v/v) FCS prior to invasion assays. To assess cell number after depletion and/or recovery, cells were scored directly by microscopy. To confirm these data, adherent cells were also released with trypsin and counted using a hemocytometer. Cell viability was assessed in parallel with a LIVE/DEAD viability/ cytotoxicity kit (Molecular Probes).

Invasion assay

The ability of Salmonella strains to invade cultured mammalian cells was assessed using gentamicin protection assays. Appropriate S. typhimurium stationary phase cultures were diluted 1:500 in TY medium and incubated for 6 h (37°C, 225 rpm) to maximize invasive efficiency. Washed bacteria were added at multiplicity of infection (moi) of 40 to serum-starved cultured HeLa or NIH3T3 cells (1 × 104 in triplicate wells, pretreated to deplete cholesterol when appropriate) in DMEM supplemented with L-glutamine. After incubation (37°C, 5% CO2, 60 min), cells were washed repeatedly with PBS, and remaining extracellular bacteria were killed by addition of 100 μgl–1 gentamicin (37°C, 5% CO2, 60 min). Cells were washed again with PBS, incubated further for time-course experiments, and lyzed in 10 mM Tris-Cl, pH 7.4, 0.5% (v/v) triton-X-100. Serial dilutions of cell lysates were plated on Luria-Bertani agar and the percentage of intracellular bacteria compared with the original inoculum. For depleted cells, equivalent results were obtained when invasions were performed in DMEM with L-glutamine or when supplemented with additional 10 mM MβCD.

To confirm that the observed decrease in bacterial entry into depleted cells did not result from an increase in cell membrane permeability to gentamicin as a consequence of MβCD pretreatment (i.e. induced killing of intracellular bacteria), cultured HeLa cells were incubated with bacteria for 60 min to allow internalization as previously and treated with cyclodextrin (10 mM, 60 min) then gentamicin (100 μgl–1, 1 h). The invasion rate was compared to control cells treated only with gentamicin. Results from three independent experiments verified that there was no significant difference (P = 0.05) in entry rate (bacteria recovered), demonstrating that cyclodextrin treatment does not induce killing of intracellular bacteria by gentamicin.

Confocal microscopy

1 × 104 HeLa or 3T3 cells seeded onto 13 mm glass coverslips (BDH) (48 h, 37°C, 5% CO2), depleted where appropriate, were infected as for invasion assays and fixed with 3.7% formaldehyde. Fixed samples were incubated with Texas Red-conjugated phalloidin (Molecular Probes) and a 1:3000 dilution of filipin stock [50 mg ml–1 (Sigma) in 95% (v/v) ethanol] in PBS (30 min, RT) to visualize the actin cytoskeleton and cholesterol respectively. Bacteria were either stained with anti-Salmonella rabbit polyclonal (Biodesign International, Inc.) and AlexaFluor 594 antirabbit IgG (Molecular Probes) according to the recommended protocols, or observed directly when constitutively expressing GFP. To visualize membranes, living cells were labelled with the fluorescent lipid analogue 1-[-4-(trimethylamino)phenyl]-6-phenylhexa-1, 3,5-triene (TMA-DPH, 1 μm, 10 min; Molecular Probes), then washed with DMEM prior to the addition of bacteria when appropriate. TMA-DPH distribution was not affected by subsequent fixation (R. Hayward, unpublished). Coverslips were mounted onto slides using ProLong Antifade reagent (Molecular Probes) and visualized using a fluorescence microscope (Leica DM IRBE) and ultraviolet confocal laser scanning microscopy. Confocal optical sections were analyzed using Leica TCS NT software and processed and assembled using OpenLab 2.0 software (Improvision) and Photoshop 5.5 (Adobe). Three dimensional image was reconstructed from confocal optical z-sections and rendered using VOLOCITY™ software (Improvision).

Scanning electron microscopy

1 × 104 HeLa or 3T3 cells seeded onto 13 mm glass coverslips (BDH) (48 h, 37°C, 5% CO2) were infected as for invasion assays, and fixed with 4% glutaraldehyde (16 h, 4°C). Coverslips were washed in PBS, with several changes (6 h, 4°C) and placed in secondary fixative (1% osmium tetroxide, 1 h). Samples were washed repeatedly in buffer (4°C), dehydrated in a graduated series of alcohols (50%, 70%, 95%, 100%; each 2 × 10 min incubation) and loaded into a Critical Point Drier (Polaron E 3000 CPD). Samples were mounted onto SEM stubs with silver dag and sputter-coated with AuPd (Polaron E5000 Sputter Coater) to a thickness of 5 nm. Mounted samples were viewed using a Philips XL 30 FEG scanning electron microscope equipped to capture digital images.

Acknowledgements

We thank Dr Jeremy Skepper and Tony Burgess (Multi-Imaging Centre, Department of Anatomy, University of Cambridge) for technical advice and assistance with confocal microscopy and scanning electron microscopy, and Samuel Miller, Stanley Falkow and David Holden for gifts of materials. We also thank Colin Hughes, Donald Tipper and Eva Koronakis for critical discussion of this manuscript. This work was supported by Wellcome Trust project grants to V.K and a Biotechnology and Biological Sciences Research Council studentship to M.J.G.

Ancillary