Salmonella type III effectors PipB and PipB2 are targeted to detergent-resistant microdomains on internal host cell membranes

Authors


E-mail lknodler@niaid.nih.gov; Tel. (+1) 406 363 9332; Fax (+1) 406 363 9380.

Summary

The intracellular pathogen, Salmonella enterica, translocates type III effectors across its vacuolar membrane into host cells. Herein we describe a new Salmonella effector, PipB2, which has sequence similarity to another type III effector, PipB. In phagocytic cells, PipB2 localizes to the Salmonella-containing vacuole (SCV) and tubular extensions from the SCV, Salmonella-induced filaments (Sifs). We used the specific targeting of PipB2 in macrophages to characterize Sifs in phagocytic cells for the first time. In epithelial cells, PipB2 has a unique localization pattern, localizing to SCVs and Sifs and additionally to vesicles at the periphery of infected cells. We further show that the N-terminal 225-amino-acid residues of PipB2 are sufficient for type III translocation and association with SCVs and Sifs, but not peripheral vesicles. Subcellular fractionation demonstrated that both PipB and PipB2 associate with host cell membranes and resist extraction by high salt, high pH and to a significant extent, non-ionic detergent. Furthermore, PipB and PipB2 are enriched in detergent-resistant microdomains (DRMs), also known as lipid rafts, present on membranes of SCVs and Sifs. The enrichment of Salmonella effectors in DRMs on these intracellular membranes probably permits specific interactions with host cell molecules that are concentrated in these signalling platforms.

Introduction

The intracellular pathogen Salmonella enterica infects a wide variety of warm-blooded hosts, with the manifestation of infection being dictated by both the infecting serovar and the host (Kingsley and Baumler, 2000). As a successful pathogen, S. enterica serovar Typhimurium (S. Typhimurium) has acquired many genetic elements by horizontal transfer that enable it to interact with and exploit host cells. In particular, two type III secretion systems (TTSSs) encoded on large pathogenicity islands enable the delivery of numerous bacterial proteins, termed effectors, into the host cell where they can directly manipulate host cell functions (Marcus et al., 2000; Hansen-Wester and Hensel, 2001). Importantly these TTSSs are regulated to act at distinct stages of Salmonella pathogenesis. The TTSS encoded on Salmonella pathogenicity island-1 (SPI-1) translocates effectors that act cooperatively to promote the invasion of non-phagocytic cells (Galan, 2001). Biogenesis of the Salmonella-containing vacuole (SCV) (Steele-Mortimer et al., 2002), induction of an early onset apoptotic cascade in macrophages (van der Velden et al., 2000) and initiation of inflammatory responses (Lee et al., 2000) have also been attributed to the actions of SPI-1 TTSS effectors. In contrast, the SPI-2-encoded TTSS is activated once Salmonella are intracellular (Valdivia and Falkow, 1997; Cirillo et al., 1998). SPI-2 TTSS effectors are translocated across the SCV membrane and are required for many functions associated with intracellular pathogenesis including bacterial replication in both phagocytic and non-phagocytic cells (Ochman et al., 1996; Cirillo et al., 1998; Hensel et al., 1998), systemic spread of bacteria in murine hosts (Cirillo et al., 1998; Hensel et al., 1998), biogenesis of the SCV (Gallois et al., 2001; Vazquez-Torres et al., 2000) and induction of a delayed onset apoptosis in macrophages (Monack et al., 2001). While each TTSS is dedicated to the translocation of a subset of effectors, some effectors can be translocated via both TTSSs (Miao and Miller, 2000). It is likely that both TTSSs are functionally active in the time period shortly after bacterial invasion, although this has yet to be conclusively demonstrated.

While the individual functions of several SPI-1 TTSS effectors have been elucidated, comparatively little is known about the mechanisms of action of SPI-2 TTSS effectors. SifA is required for maintenance of the SCV membrane (Beuzon et al., 2000) and the aggregation of host cell membranes in epithelial cells into tubular structures called Salmonella-induced filaments (Sifs) (Stein et al., 1996) and localizes to these structures in infected cells (Brumell et al., 2002). Although the role of Sifs in Salmonella pathogenesis is unclear, SifA mutants are severely attenuated in virulence studies (Beuzon et al., 2000). Another SPI-2 TTSS effector, SseJ, also regulates SCV membrane dynamics (Ruiz-Albert et al., 2002) and associates with SCVs and Sifs (Kuhle and Hensel, 2002; Freeman et al., 2003). Likewise, SseF and SseG are translocated to the SCV and Sifs (Kuhle and Hensel, 2002) and modulate the aggregation of host cell endosomes (Guy et al., 2000; Kuhle and Hensel, 2002). SpiC has been described as an effector that inhibits interactions between the SCV and late endosomes/lysosomes (Uchiya et al., 1999) and as being essential for the secretion of other SPI-2 effectors (Yu et al., 2002), most probably via translocon assembly (Freeman et al., 2002). Little is known about the other dedicated SPI-2 TTSS effectors SspH 2, SseI/SrfH (Miao and Miller, 2000; Worley et al., 2000), PipB (Knodler et al., 2002), SifB (Freeman et al., 2003) and SopD2 (Brumell et al., 2003) or SlrP and SspH 1, effectors that are translocated by both TTSSs (Miao and Miller, 2000). Interestingly, for most of these SPI-2 TTSS effectors, mutations result in little or no effect on intracellular proliferation or virulence in animal models (Ruiz-Albert et al., 2002; Freeman et al., 2003), whereas mutants in components of the SPI-2 TTSS, which are defective in the translocation of all SPI-2 TTSS-associated effectors, are avirulent (Cirillo et al., 1998; Hensel et al., 1998). This suggests a significant level of redundancy or overlapping function for SPI-2 TTSS effectors, as has been established for SPI-1 TTSS effectors (Friebel et al., 2001; Zhou and Galan 2001).

Herein we describe a new SPI-2 TTSS effector, PipB2. PipB2 shares significant sequence similarity with the recently described Salmonella type III effector, PipB (Knodler et al., 2002). Despite being encoded outside of SPI-2, pipB2 is part of the SPI-2 regulon and is a virulence factor as assessed by mouse infection studies. Using an epitope tagging approach, we show that in epithelial cells PipB2 localizes to SCVs and Sifs and also to vesicles at the cell periphery. In macrophages, PipB2 localizes to the SCV and lysosomal glycoprotein (lgp)-rich tubular structures extending from the SCV, establishing for the first time that Sifs are also present in phagocytic cells. Both PipB and PipB2 associate with host cell membranes. More specifically, they are enriched in detergent-resistant microdomains (DRMs) on the membranes of SCVs and Sifs. Our results demonstrate for the first time that bacterial type III effectors can target DRMs present on internal membranes. Because DRMs serve as clustering sites for proteins and play a critical role in the regulation of cell signalling, membrane trafficking and cytoskeleton (Brown and London, 1998), the partitioning of bacterial effectors to these subdomains may allow intracellular pathogens to readily interface with DRM-dependent host cell processes.

Results

PipB2, a homologue of PipB, is part of the ssrAB regulon

We recently identified PipB to be a Salmonella type III effector encoded within SPI-5 and translocated by the SPI-2 TTSS (Knodler et al., 2002). With the release of two Salmonella genome sequences, a gene with significant sequence similarity to pipB, which we have designated pipB2, was identified in the S. Typhi genome (designated STY2897) and the S. Typhimurium LT2 genome (designated STM2780) at 60.3 centisomes. A recent report of the complete transcriptional profile of intracellular Salmonella (Eriksson et al., 2003) demonstrated that expression of STM2780 (pipB2) was strongly induced upon infection of macrophages, a characteristic of genes comprising the SPI-2 regulon.

pipB2 encodes a polypeptide of 350 amino acids, with a predicted molecular mass of 37 350 Da. PipB and PipB2 share 29% identity and 55% similarity at the amino acid level and both proteins contain numerous tandem pentapeptide repeats in their C-terminal regions (Fig. 1). While no position is completely conserved in these repeats, the consensus sequence is A(N/D)(L/M/F)XX where X is any amino acid. PipB and PipB2 contain 22 and 30 of these pentapeptide repeats respectively. The pentapeptide repeat family of proteins is extensive, spanning both bacterial and eukaryotic kingdoms, but very little is known about the function of most of these proteins.

Figure 1.

Amino acid sequence alignment of PipB and PipB2. Alignment was performed using CLUSTAL W (Thompson et al., 1994). Identical residues are indicated by an asterisk and conserved amino acid substitutions are represented by a dot. Numbering on the right refers to the position in each amino acid sequence. The C-terminal regions of PipB and PipB2 are composed of tandem pentapeptide repeats. While no position in each repeat is absolutely conserved, the consensus of the repeats is A(N/D)(L/M/F)XX, where X is any amino acid. The middle amino acid in each pentapeptide repeat is highlighted with a grey box for clarity. Indicated by a solid line overhead or underneath are regions of PipB and PipB2, respectively, predicted by TMpred (Hofmann and Stoffel, 1993) to be membrane spanning.

Recently, we (Knodler et al., 2002; Brumell et al., 2003), and others (Kuhle and Hensel, 2002; Freeman et al., 2003), have used an epitope tagging strategy to successfully monitor the expression and translocation of Salmonella effector proteins. By fusing two haemagglutinin (HA) tags to the C-terminus of PipB, we demonstrated that this protein is a SPI-2 TTSS translocated effector (Knodler et al., 2002). Here we used this epitope-tagging approach to follow the expression of PipB2 under growth conditions that are known to repress or induce the expression of genes comprising the SPI-2 regulon. Plasmid-borne expression of PipB2-2HA was under the control of the pipB2 promoter. SL1344 ΔpipB2 ppipB2-2HA were grown in SPI-1 or SPI-2 inducing media as described in Experimental procedures and bacterial proteins were subject to immunoblotting (Fig. 2A). Expression of PipB2-2HA was barely detectable under SPI-1-inducing conditions but was greatly induced by growth in limiting nutrient media, conditions that induce SPI-2 gene expression and are thought to mimic those inside the SCV (Deiwick et al., 1999). Because SsrA/SsrB is the two-component regulatory system controlling SPI-2 gene expression (Cirillo et al., 1998; Deiwick et al., 1999), we next analysed PipB2-2HA expression in an ssrB mutant background. Under SPI-2-inducing growth conditions, little PipB2-2HA signal was detectable in an ssrB mutant (Fig. 2A). The pattern of PipB2-2HA expression during bacterial growth under these different conditions was similar to that of SseB, which is encoded within SPI-2 and functions as part of the SPI-2 translocon (Nikolaus et al., 2001). From these results we concluded that pipB2 is part of the SPI-2 regulon.

Figure 2.

PipB2 expression is dependent on SsrA/SsrB.
A. SL1344 ΔpipB2 ppipB2-2HA or SL1344 ssrB::kan ppipB2-2HA bacteria were grown in SPI-1-inducing media (overnight standing culture in LB broth) or SPI-2-inducing media (stationary phase culture in N-minimal media pH 7.4) and proteins from equal bacterial numbers were analysed by SDS–PAGE. HA-tagged PipB2, DnaK and SseB were detected by immunoblotting. Numbers on the left of each panel indicate molecular mass (kDa).
B. PipB and PipB2 are not secreted in vitro. S. Typhimurium wild type or ssaV-deficient strain P2D6 each harbouring ppipB-M45 or ppipB2-M45 were grown in PCN-P media pH 5.8 for 16 h. Bacteria were harvested by centrifugation and an aliquot was used as the total cell fraction (T). Secreted protein in the supernatant (SN) was recovered by TCA precipitation. The bacterial pellets were resuspended in PBS and protein on the bacterial surface (D) was detached by vigorous mixing. Detached protein was subsequently recovered by TCA precipitation. Proteins in the various fractions were separated by SDS–PAGE on 10% Tricine gels and transferred to nitrocellulose membranes. Aliquots corresponding to equal bacterial numbers were analysed. The presence of SseB or M45-tagged PipB or PipB2 was detected by incubation with polyclonal antiserum against SseB or a monoclonal antibody against the M45 epitope.

It has been demonstrated that several SPI-2 TTSS effectors and translocon components are secreted to the bacterial surface and into the culture supernatant in vitro when Salmonella are exposed to minimal media of acidic pH (Beuzon et al., 1999; Klein and Jones, 2001; Nikolaus et al., 2001; Hansen-Wester et al., 2002). The ‘secretion’ of effectors into culture media is distinct from the ‘translocation’ of effectors into eukaryotic cells. Specifically, the secretion of type III effectors into culture media has been monitored via C-terminal fusions to the M45 epitope, an 18-amino-acid peptide of the adenovirus protein E4-6/7 (Hansen-Wester et al., 2002). To directly compare the secretion of PipB and PipB2 with previously published in vitro results, we created C-terminal fusions of these Salmonella proteins to the M45 epitope. S. Typhimurium wild type or a ΔssaV mutant harbouring ppipB-M45 or ppipB2-M45 were grown in minimal media with limiting phosphate (PCN-P). PipB-M45 and PipB2-M45 in the bacterial pellet, secreted into the culture supernatant and associated with the bacterial cell-surface were monitored by immunoblotting with a monoclonal antibody against the M45 epitope (Fig. 2B). Surprisingly, PipB, a known SPI-2 TTSS effector (Knodler et al., 2002), was not present on the bacterial surface or secreted into the culture supernatant. Similarly, no secretion of PipB2-M45 was detected after growth in PCN-P at pH 5.8. However, as a positive control, a known SPI-2 TTSS substrate, SseB, was detected in the detached fraction and in the culture supernatant in a wild-type background, but not the SPI-2 TTSS mutant, ΔssaV. Taken together, these results indicate that PipB and PipB2 are not secreted substrate proteins in vitro.

Role of PipB2 in intracellular proliferation and virulence

SPI-2 plays an important role in the pathogenesis of Salmonella infection, contributing significantly to intracellular survival and proliferation in host cells. Having established that pipB2 is regulated by SsrA/SsrB, we next addressed whether pipB2 contributes to intracellular parasitism in phagocytic cells. RAW 264.7 macrophage-like cells were infected with opsonized bacteria as described in Experimental procedures. Monolayers were lysed and plated to enumerate colony-forming units (cfu) at 2 and 21 h post-infection (pi). Wild-type bacteria replicated almost sevenfold over this time frame whereas a ΔSPI-2::kan mutant, which is deleted of all SPI-2-associated TTSS apparatus, effectors and chaperones and transcriptional regulator encoding genes, had a dramatic replication defect (Fig. 3). Deletion of pipB2 had no apparent effect on intracellular replication in RAW264.7 cells (Fig. 3) or HeLa cells (results not shown). Similarly, we have previously shown that a ΔpipB mutant is indistinguishable from wild-type bacteria in this assay (Knodler et al., 2002).

Figure 3.

PipB2 is not required for intracellular survival and replication in phagocytic cells. RAW264.7 cells were infected with opsonized bacteria and gentamicin-protected bacteria were enumerated at 2 and 21 h pi. Values are mean ± SD of the fold-replication over this time period from at least three separate experiments.

Infection of mice with S. Typhimurium results in a systemic disease whereby bacteria replicate primarily in the spleen and liver (Kingsley and Baumler, 2000). In competitive index assays, the relative abilities of two bacterial strains to colonize the host are measured. These experiments provide a more sensitive measure of the contribution of a given Salmonella gene to virulence than the traditional time-to-death studies. To investigate if pipB2 plays a role in virulence, we compared the ability of the ΔpipB2 mutant to compete against a wild-type strain in the mouse model of infection. Mice were inoculated intraperitoneally with equal numbers of wild-type and the ΔpipB2 mutant bacteria, each carrying plasmids with different resistance markers. Forty-eight hours afterwards, infected spleens were collected and homogenized and the bacteria were differentiated on antibiotic-containing plates. SPI-2 is known to be crucial for the systemic spread of Salmonella in the mouse model of typhoid fever (Kingsley and Baumler, 2000). Accordingly, mutation of ssaR, which encodes a structural component of the SPI-2 TTSS, dramatically attenuated the ability of this strain to colonize mice (Table 1). We have previously observed that in time-to-death studies, infection with a ΔpipB mutant is indistinguishable to that by wild-type bacteria (Knodler et al., 2002). In the more sensitive C.I. assay, this pipB deletion mutant also showed no virulence attenuation – the ΔpipB bacteria competed equally well with wild-type bacteria to colonize the spleen (Table 1). In contrast, however, the pipB2 deletion mutant had a slight, but statistically significant, virulence defect (Table 1), similar to that shown previously for mutations in sseJ and srfJ (Ruiz-Albert et al., 2002), Salmonella genes that are also expressed in an SsrA/SsrB dependent manner (Miao and Miller, 2000; Worley et al., 2000). Collectively we conclude from these studies that while neither pipB nor pipB2 are required for intracellular replication in cultured cell lines, the functions of pipB2 are adequately non-redundant such that a ΔpipB2 mutant suffers a reduced virulence in mice.

Table 1.  Competitive index analysis of S. Typhimurium mutants.
Mixed infectionsCompetitive index
  • a

    . Significantly different from 1.0 (P < 0.05).

  • b

    . Significantly different from 1.0 (P < 0.001).

  • Female BALB/c mice were inoculated intraperitoneally with a mixture of two strains (5 ×104 of each strain), one strain bearing plasmid pWSK29 (ampr.) and the other bearing plasmid pWSK129 (kanr.). Infected spleens were harvested and homogenized 48 h later and bacteria plated on LB agar containing ampicillin or kanamycin to enumerate cfu. The competitive index was calculated as the output ratio of mutant:wild type (wt) divided by the input ratio of mutant/wt bacteria. Values are mean ± SD for a minimum of 13 mice from at least two separate experiments. The competitive indices were analysed using the Student's t-test.

wild type versus wild type 1.17 ±  0.11
ΔpipB versus wt 1.15 ±  0.01
ΔpipB2 versus wt 0.66 ±  0.13a
ΔssaR versus wt0.025 ±  0.004b

PipB2 is translocated to Sifs in macrophages

Because pipB2 is part of the SPI-2 regulon, contributes to Salmonella virulence and encodes a protein with significant sequence similarity to PipB, PipB2 has the characteristics of a Salmonella-translocated effector. To investigate if PipB2 was indeed translocated into host cells, RAW264.7 macrophage-like cells were infected with SL1344 wild type carrying ppipB2-2HA as described in Experimental procedures. Infected cells were fixed and processed for immunofluorescence to detect translocated PipB2-2HA. PipB2-2HA signal was detectable at time points after 6 h pi, localizing to the SCV (results not shown). By 12 h pi, PipB2-2HA was localized to both the SCV and structures that extended from the SCV (results not shown), although these were difficult to clearly discern by confocal microscopy. We found that the flattened morphology of RAW264.7 cells primed with interferon-γ (IFN-γ) prior to infection with Salmonella were more suitable for detailed inspection of these structures and chose to use these conditions for further analysis. While we acknowledge that IFN-γ priming can alter membrane trafficking (Tsang et al., 2000) and the killing ability of macrophages (Rosenberger and Finlay 2002), we also believe that Salmonella is likely to be confronted by both naïve and activated macrophages in vivo. Thus using IFN-γ-primed macrophages is a physiologically relevant model in which to study Salmonella-induced structures. At 12 h pi, we observed that PipB2-2HA associated with linear/punctate extensions from the SCV in 54 ± 7% (n = 3 expts) of the positively staining cells (Fig. 4A and D), which co-localized with the late endosomal/lysosomal marker, lysosomal associated protein-1 (LAMP-1) (Fig. 4B and E). These PipB2-positive structures were also observed in infected THP-1 cells, a human monocytic cell line, although with less frequency (results not shown). The localization pattern of PipB2-2HA in RAW264.7 cells is extremely similar to the pattern described for Sifs in epithelial cells (Garcia-del Portillo et al., 1993; Brumell et al., 2001). Both linear (Fig. 4A and C) and punctate (Fig. 4B) PipB2-2HA staining of the Sifs was observed in RAW264.7 cells. The Salmonella-induced aggregation of lgp-rich compartments in macrophages has not been studied as these cells exhibit an extensive network of lgp-positive tubuloreticular lysosomes under resting and/or uninfected conditions (Swanson et al., 1987). With the specific targeting of PipB2-2HA in RAW264.7 cells, we now have an adequate marker to study Sifs in phagocytic cells. The striking aggregation of host cell late endosomal/lysosomal compartments in cultured epithelial cells is dependent on the actions of one SPI-2 TTSS effector, SifA (Stein et al., 1996; Beuzon et al., 2000; Brumell et al., 2002). As such, we assessed whether the formation of Sifs in RAW264.7 cells also required SifA. In a ΔsifA background, PipB2-2HA labelled tubular/punctate extensions were evident in only 7 ± 3% (n = 3 expts) of the positively staining cells (Fig. 4D). The few Sifs formed by the ΔsifA mutant were much shorter (<2 µm) than those typically seen in wild-type infected cells (>20 µm). We conclude from these results that Sifs are formed in phagocytic cells in a SifA-dependent manner.

Figure 4.

PipB2 is translocated into host cells.
A. PipB2-2HA localizes to SCVs and Sifs. IFNγ-primed RAW264.7 cells were infected with opsonized SL1344 wild-type ppipB2-2HA bacteria, fixed at 12 h pi, permeabilized and immunostained for Salmonella LPS and PipB2-2HA. An overlay of the two confocal images is presented on the right. Arrowheads indicate linear PipB2-2HA staining extending from SCVs.
B. PipB2-2HA co-localizes with LAMP-1. RAW264.7 cells were infected as described in (A) except immunostaining was for LAMP-1 and PipB2-2HA. The merged image is shown on the right. Scale bar = 10 µm. Arrowheads indicate punctate PipB2-2HA staining extending from SCVs.
C. PipB2-2HA does not co-localize with cathepsin D. RAW264.7 cells were infected as described in (A) except immunostaining was for cathepsin D and PipB2-2HA. The merged image is shown on the right. Arrowheads indicate linear PipB2-2HA staining extending from SCVs.
D. Sifs are formed in a SifA-dependent manner in RAW264.7 cells. RAW264.7 cells were infected with SL1344 wild-type ppipB2-2HA or SL1344 ΔsifA ppipB2-2HA as described in (A). Fixed cells were immunostained for Salmonella LPS and PipB2-2HA. For each experiment, at least 100 PipB2-positively staining cells from two coverslips were scored for the presence of Sifs (linear or punctate PipB2–2HA staining extending from SCVs). Results are presented as the percentage of Sif-positive cells and are the mean ± SD from three separate experiments.
E. Sifs in RAW264.7 cells co-localize with LAMP-1 but not CI-M6PR, LBPA or cathepsin D. RAW264.7 cells were infected with SL1344 wild-type ppipB2-2HA as described in (A). Fixed cells were immunostained for PipB2-2HA and either LAMP-1, CI-M6PR, LBPA or cathepsin D. At least 100 PipB2-2HA staining Sifs were scored for co-localization with each host cell marker per experiment. Values are the mean ± SD from at least three separate experiments.

In epithelial cells, Sifs appear to be continuous with the SCV and are positive for host cell markers normally associated with late endosomes and lysosomes, including lysobisphosphatidic acid (LBPA), cathepsin D, rab7 (Brumell et al., 2001) and lgps such as LAMP-1 (Garcia-del Portillo et al., 1993) and CD63 (Beuzon et al., 2000). Recently, the translocation of two SPI-2 effectors, SseJ and SifB, in RAW264.7 cells ‘to LAMP-1 positive structures distant from the SCV’ was demonstrated (Freeman et al., 2003), but these structures were not further characterized. We used the targeting of PipB2-2HA to Sifs to identify the host cell marker composition of these bacterially induced structures in phagocytic cells for the first time. RAW264.7 cells were infected with SL1344 wild-type ppipB2-2HA and at 12 h pi cells were fixed, immunostained and examined by fluorescence microscopy. Results are presented in Fig. 4E. The vast majority of PipB2-2HA labelled Sifs (93 ± 3%, n = 3 expts) were positive for LAMP-1 (see also Fig. 4B). However, little association of a phospholipid component of internal membranes of late endosomes, LBPA (Kobayashi et al., 1998), with Sifs was observed. Likewise, only a small percentage of Sifs were positive for cation-independent mannose 6-phosphate receptor (CI-M6PR), a recycling receptor that traffics from the trans-Golgi network to late endosomes/lysosomes (Bleekemolen et al., 1988; Griffiths et al., 1988) or cathepsin D, a lysosomal protease (Fig. 4C). Previous studies on the maturation of SCVs in murine primary and immortalized macrophages have demonstrated that there is minimal fusion of lysosomes and late endosomes with the vacuole (Ishibashi and Arai, 1990; Buchmeier and Heffron, 1991; Rathman et al., 1997; Garvis et al., 2001). Collectively, our results detailing the host cell marker composition of Sifs are consistent with those for SCVs in macrophages, which supports the idea that Sifs are indeed contiguous with the SCV.

PipB2 is translocated to peripheral vesicles in epithelial cells

We next examined the subcellular localization of PipB2-2HA in cultured epithelial cells as these non-phagocytic cells also serve as an important replicative niche for Salmonella. SL1344 ΔpipB2 bacteria harbouring ppipB2-2HA were used to infect HeLa cells as described. At various time points pi, HeLa cells were fixed and processed for immunostaining. PipB2-2HA signal was detectable by immunofluorescence after 4 h pi, although the fluorescence signal was weak and visible in < 10% of infected cells. Thereafter signal intensity increased such that by 10 h pi, 60–75% of infected cells were positive for a PipB2-2HA signal (n = 3 expts). These intracellular expression kinetics are consistent with the induction of PipB2 by nutrient-limiting conditions (Fig. 2A), such as those found in the SCV. Furthermore, in agreement with our in vitro expression data (Fig. 2A), in an ssrB::kan mutant background, no expression (by Western blotting) or translocation (by immunofluorescence) of PipB2-2HA in infected HeLa cells was detected (results not shown). The localization pattern of PipB2-2HA in HeLa cells, to SCVs and Sifs (Fig. 5A), resembled that seen in infected macrophages. A similar localization pattern has been described for other SPI-2 effectors in epithelial cells, including PipB (Knodler et al., 2002), SseF and SseG (Kuhle and Hensel, 2002), SseJ and SifB (Freeman et al., 2003) and SifA (Brumell et al., 2002). Significantly, however, PipB2-2HA was also detected in vesicles at the periphery of infected epithelial cells (Fig. 5A), a localization pattern that has not been described for any other Salmonella effector. These peripheral vesicles appeared to be polarized to regions of membrane extensions rather than being evenly spread around the entire cell periphery (Fig. 5A).

Figure 5.

PipB2 translocation in epithelial cells.
A. PipB2 localizes to SCVs, Sifs and peripheral vesicles. HeLa cells were infected with late-log phase SL1344 ΔpipB2 ppipB2-2HA and fixed at 10 h pi. After permeabilization, immunostaining was for PipB2-HA and Salmonella LPS. An overlay of the two confocal images is presented on the right. Scale bar = 10 µm. Asterisks and arrowheads indicate PipB2-labelled peripheral vesicles and Sifs respectively. The cell periphery is outlined for clarity.
B. The N-terminal 225-amino-acid residues of PipB2 are sufficient for type III translocation and association with SCVs and Sifs, but not accumulation in peripheral vesicles. HeLa cells were infected with SL1344 ΔpipB2 ppipB2 N225-2HA and immunostained as described for (A). Arrowheads point to Sifs decorated with PipB2 N225-2HA. The cell periphery is outlined for clarity.

The signal sequence for translocation of Salmonella effectors lies within their N-terminal domain (Knodler et al., 2002; Brumell et al., 2003) and some type III effectors contain a conserved amino acid sequence motif that directs translocation (Miao and Miller, 2000). While the N-termini of PipB and PipB2 share significant sequence similarity (Fig. 1), neither possess the conserved WEK(I/M)XXFF motif (Miao and Miller, 2000). To investigate if the N-terminal domain of PipB2 was sufficient for translocation, a plasmid construct expressing the N-terminal 225-amino-acid residues of PipB2 fused to two HA tags, under the control of the pipB2 promoter (ppipB2 N225-2HA) was introduced into SL1344 ΔpipB2. HeLa cells were infected with these bacteria and processed for immunofluorescence staining at 10 h pi. Figure 5B illustrates that PipB2 N225-2HA was translocated to SCVs and Sifs. Surprisingly, unlike for the full-length construct, no accumulation of fluorescence signal was detected in peripheral vesicles for the N-terminal truncation (Fig. 5B). These results highlight two important features about PipB2 translocation. While the N-terminal 225 residues of PipB2 are sufficient for type III translocation and association with Sifs and SCVs, this domain is not sufficient for trafficking to peripheral vesicles in infected epithelial cells.

Peripheral localization of PipB2 vesicles is dependent on microtubules but not Sif formation

To gain insight into the host cell processes involved in PipB2 trafficking to peripheral vesicles, we studied the effect of various inhibitors of intracellular membrane trafficking processes on PipB2 localization. HeLa cells were infected with SL1344ΔpipB2 ppipB2-2HA bacteria as described. At 9.5 h pi, inhibitor was added for 20 or 30 min as indicated and then cells were fixed and processed for immunofluorescence. For mock-treated samples, 61.3 ± 4.5% of the PipB2-positively staining cells had a strong HA-signal accumulation in peripheral vesicles (Fig. 6A). Treatment with either brefeldin A (BFA), which causes a block in anterograde ER-to-Golgi membrane traffic (Strous et al., 1993) and in some cell lines results in the collapse of the trans-Golgi network and certain classes of endosomes onto the microtubule organizing centre (Reaves and Banting, 1992; Tooze and Hollinshead, 1992), or wortmannin (WTM), an inhibitor of phosphoinositide 3-kinase-dependent host cell processes including endocytic trafficking (Kjeken et al., 2001), did not alter the distribution of peripheral vesicles (Fig. 6A). Likewise, de-polymerization of F-actin by cytochalasin D treatment did not overtly perturb the peripheral localization of PipB2-positive vesicles (Fig. 6A).

Figure 6.

Peripheral localization of PipB2-positive vesicles is dependent on microtubules but not the actions of SifA.
A. Effect of inhibitors on peripheral vesicular staining of PipB2. HeLa cells were infected with SL1344 ΔpipB2 ppipB2-2HA for 9.5 h. Inhibitors were then added for 30 min, or 20 min in the case of cytochalasin D, followed by PFA fixation of cells. Cells were permeabilized and immunostained for Salmonella LPS and PipB2-2HA. Final inhibitor concentrations used were: 5 µg ml−1 nocodazole (NDZ), 1 µg ml−1 brefeldin A (BFA), 100 nM wortmannin (WTM) and 1 µg ml−1 cytochalasin D (cytoD). Control is mock-treated cells. At least 100 PipB2-positively staining cells were scored for the localization pattern of PipB2-2HA in each experiment. Results are presented as the percentage of HA-staining cells showing a concentration of signal in peripheral vesicles and are the mean ± SD from at least three separate experiments.
B. Redistribution of PipB2-positive peripheral vesicles after nocodazole treatment. HeLa cells were infected with ΔpipB2 ppipB2-2HA or ΔpipB ppipB-2HA bacteria as described in (A) and 5 µg ml−1 nocodazole was added 30 min prior to fixation (+NDZ). Immunostaining was as described in (A). An overlay of the two confocal images is presented on the right. For comparison, on the far right of each panel is an overlay of the confocal images for infected cells that were not treated with NDZ (–NDZ). Arrows and arrowheads (upper panel) indicate PipB2-decorated peripheral vesicles and Sifs respectively. Asterisks demonstrate perinuclear accumulation of PipB2-positive vesicles after NDZ treatment. In the lower panel, labelling of Sifs by PipB is indicated by arrowheads. The periphery of each cell has been outlined for clarity.
C. Trafficking of PipB2-positive peripheral vesicles is independent of SifA. HeLa cells were infected with SL1344 ΔsifA ppipB2-2HA for 10 h, fixed and immunostained for PipB2-2HA and LPS. An overlay of the two confocal images is shown on the right. Arrows indicate PipB2-positive peripheral vesicles.

In contrast, depolymerization of microtubules by treatment of HeLa cells with nocodazole (NDZ) resulted in a striking redistribution of PipB2-positive vesicles (Fig. 6B). After 30 min of NDZ treatment, only 9.0 ± 2.8% of cells showed peripheral vesicular PipB2 staining (Fig. 6A). This decrease in peripheral staining was associated with an increase in perinuclear staining of PipB2 (Fig. 6B; compare +NDZ and –NDZ panels). In contrast to PipB2, PipB was never detected in peripheral vesicles in HeLa cells (Knodler et al., 2002) (Fig. 6B, far right panel). Accordingly, no dramatic change in the localization pattern of PipB was observed after 30 min of NDZ treatment (Fig. 6B; compare +NDZ and –NDZ panels). Taken together, these results indicate that PipB2-positive vesicles accumulate at peripheral membrane sites independently of actin filaments, but an intact microtubule network is required for their peripheral localization.

Microtubules provide tracks for directional vesicle transport and organelle positioning, and also serve as a scaffold for Sif formation in HeLa cells (Brumell et al., 2002). Having established an essential requirement for microtubules for PipB2-2HA targeting to peripheral vesicles, we further examined if Sif formation along these microtubules is also required. HeLa cells were infected with SL1344 ΔsifA ppipB2-2HA and processed at 10 h pi for immunostaining. ΔsifA mutants do not induce the formation of Sifs (Stein et al., 1996) (Fig. 4C) and also progressively lose their vacuolar membrane in both epithelial cells and macrophages (Beuzon et al., 2000). As would be expected in the absence of Sifs, no PipB2-2HA decoration of these structures was observed in a ΔsifA mutant background (Fig. 6C). However, we found that PipB2-2HA did localize to peripheral vesicles and some SCVs (Fig. 6C). Presumably, the PipB2-2HA staining of SCVs represented the fraction of ΔsifA bacteria that were still enclosed by a membrane at 10 h pi. Our results imply that trafficking of PipB2-2HA to peripheral vesicles requires microtubules but is independent of the actions of SifA, and thus Sif formation per se.

PipB2 is not required for Sif formation or maintenance of the SCV in host cells

Several SPI-2 effectors have been shown to contribute to vacuolar membrane maintenance (Beuzon et al., 2000; Ruiz-Albert et al., 2002) and Sif formation (Stein et al., 1996; Guy et al., 2000; Kuhle and Hensel, 2002). Because PipB and PipB2, like some other SPI-2 TTSS effectors, are translocated to SCVs and Sifs, we asked whether these effectors could also modulate such events. HeLa cells were infected with wild-type bacteria or the described isogenic deletion strains and processed for immunostaining at 9 h pi. Infected cells were scored for vacuolar integrity, as assessed by the presence of LAMP-1-positive staining around intracellular bacteria, and for the presence of LAMP-1-positive Sifs. Using this method, ΔpipB, ΔpipB2 or ΔpipBΔpipB2 strains were indistinguishable from wild type with respect to the presence of LAMP-1 on the SCV and the induction of Sifs in HeLa cells (Table 2). Furthermore, no reproducible alterations in Sif morphology were observed upon infection with ΔpipB, ΔpipB2 or ΔpipBΔpipB2 strains (results not shown), unlike what has been noted for sseF and sseG mutants (Kuhle and Hensel, 2002).

Table 2.  Sif formation and integrity of the vacuolar membrane for S. Typhimurium mutants.
StrainSif-positive cells (%)LAMP-1 positive SCVs (%)
  1. HeLa cells were seeded on coverslips in 24-well plates and infected with late-log phase bacteria as described in Experimental procedures. Monolayers were fixed in PFA at 9 h pi and immunostained with anti-Salmonella LPS and anti-human LAMP-1. Infected cells (>50 per coverslip, two coverslips per experiment) were scored by fluorescent microscopy for Sif formation, as judged by the presence of linear or punctate LAMP-1 staining extending from the SCV. Results are presented as percentage of infected cells containing Sifs. The association of bacteria with LAMP-1 was used as a measure of vacuolar integrity. Bacteria (>100 per coverslip, two coverslips per experiment) were scored for co-localization with LAMP-1. All values represent the mean ± SD from three separate experiments.

Wild type69 ±  6.794 ±  3.3
ΔpipB68 ±  5.194 ±  3.8
ΔpipB258 ±  8.694 ±  4.5
ΔpipBΔpipB261 ± 8.494 ±  4.0
ΔsifA6.3 ± 2.351 ±  10
ΔpipBΔsifA3.3 ± 2.740 ±  9.8
ΔpipB2ΔsifA2.3 ± 1.544 ±  10.9
ΔpipBΔpipB2ΔsifA3.7 ± 2.746 ±  8.7

Very recently it was shown that SseJ is a SPI-2 TTSS effector that localizes to SCVs and Sifs (Kuhle and Hensel, 2002; Freeman et al., 2003) and has actions complementary to SifA – a ΔsifA mutant progressively loses its vacuolar membrane in host cells, whereas a ΔsifAΔsseJ mutant does not (Ruiz-Albert et al., 2002). Because PipB and PipB2 are targeted to the same endosomal membranes as SifA, we asked whether there also existed cross-talk between these effectors. Accordingly we constructed pipB and/or pipB2 deletion mutants in a ΔsifA background and scored these deletion strains for Sif formation and SCV association with LAMP-1 as above. However, there were no additional effects compared with the sifA deletion alone (Table 2). In conclusion, despite being targeted to SCVs and Sifs, PipB and PipB2 are not overtly required for maintenance of the SCV membrane or the formation of Sifs in epithelial cells. Moreover, PipB and PipB2 appear to have actions independent of SifA.

PipB and PipB2 are targeted to host cell membranes

Immunofluoresence analysis determined that both PipB and PipB2 are targeted to SCVs and Sifs, and PipB2 also to peripheral vesicles. However, whether these bacterial proteins were attached to these structures or physically contained within them was not assessable by immunostaining. Thus, to determine the nature of the subcellular association of PipB and PipB2, HeLa cells were infected with SL1344 ΔpipB harbouring ppipB-2HA[formerly designated pACB-2HA (Knodler et al., 2002)], which expresses HA-tagged PipB, or ΔpipB2 ppipB2-2HA as described. Infected cells were mechanically disrupted at 17–18 h pi and the post-nuclear supernatant was fractionated into cytosolic and membrane fractions. Each of the fractions was then analysed by immunoblotting for the HA-tagged effectors (Fig. 7A). As expected, both PipB-2HA and PipB2-2HA partitioned with unbroken cells, host cell nuclei and cytoskeleton and intact bacteria in the P1 fraction. PipB-2HA and PipB2-2HA were also detected in the high-speed ultracentrifugation particulate fraction (P2), which contains host cell membranes (Fig. 7A). No signal was detected in the host cell cytosol (S2 fraction) for the translocated effectors. The bacterial protein DnaK was solely found in the P1 fraction, indicating there was no bacterial contamination of the post-nuclear supernatant. As positive controls, the endoplasmic reticulum integral membrane protein calnexin was detected in the P2 fraction and the host cytosolic protein β-tubulin was detected in the S2 fraction. An N-terminal truncation of PipB2, PipB2 N225-2HA, which by immunofluorescence analysis localized to the SCVs and Sifs, but not peripheral vesicles (Fig. 5B), was detected only in the P1 and P2 fractions (Fig. 7A). A weaker signal was detected in the P2 fraction for the N-terminal truncation, suggesting that PipB2 N225-2HA is translocated less efficiently than PipB2-2HA. Collectively these results demonstrate that both PipB and PipB2 are targeted to host cell membranes and the N-terminal 225 residues of PipB2 are sufficient for this membrane association.

Figure 7.

PipB and PipB2 are targeted to host cell membranes and this association is TX-100 resistant.
A. PipB and PipB2 are associated with host cell membranes. HeLa cells were infected for 16–18 h with SL1344 ΔpipB ppipB-2HA, ΔpipB2 ppipB2-2HA or ΔpipB2 ppipB2 N225-2HA. Subcellular fractionation was as described in Experimental procedures. The three fractions obtained were P1 (host cell nuclei and cytoskeleton and intact bacteria), P2 (host cell membrane fraction) and S2 (host cell cytosol). Equal volumes of each fraction were separated by SDS–PAGE and subject to immunoblotting. PipB-2HA, PipB2-2HA and PipB2 N255-2HA were detected with anti-HA antibodies. Controls were for the bacterial cytosolic chaperone DnaK, the host cell integral membrane protein calnexin, and the host cell cytosolic protein β-tubulin.
B. PipB and PipB2 association with host cell membranes is resistant to TX-100 extraction. HeLa cells were infected with Salmonella as described for (A) and subject to subcellular fractionation as described in Experimental procedures. Host cell membranes were extracted on ice with high salt (1 M NaCl), high pH (pH 11.4) or detergent (TX-100) and the insoluble fraction (P) was separated from the soluble material (S) by ultracentrifugation. Equal volumes of each fraction were then separated by SDS–PAGE and subject to immunoblotting. PipB-2HA and PipB2-2HA were detected with anti-HA antibodies. Host cell membrane controls were calnexin and caveolin-1 (antibody detects α- and β-isoforms of caveolin-1).
C. Densitometric analysis of TX-100-resistant PipB-2HA, PipB2-2HA, calnexin and caveolin-1 signals from immunoblotting. Densitometric scanning (Alpha Innotech AlphaImager software) of the intensity of each band in the TX-100 insoluble fraction is expressed as a percentage of the total band intensity (soluble plus insoluble) for each protein. Results are mean ± SD from three independent experiments.

Membrane-associated PipB and PipB2 are resistant to TX-100 solubilization and accumulate in detergent-resistant microdomains

Both PipB and PipB2 are predicted by TMpred (Hofmann and Stoffel, 1993) to have membrane-spanning regions, so we sought to address the nature of their association with host cell membranes. The post-nuclear supernatant from Salmonella-infected HeLa cells (17–18 h pi) was collected as described and divided into four aliquots. The aliquots were ultracentrifuged to obtain the particulate fraction (P2) and the supernatant was discarded. The P2 fraction was further extracted on ice under various conditions and subject to a second ultracentrifugation to obtain soluble and insoluble (or particulate) membrane fractions. These fractions were subject to Western blot analysis with anti-HA antibodies to detect the epitope-tagged Salmonella effectors (Fig. 7B). Treatment with high salt (1 M NaCl) or alkaline pH (0.2 M Na2CO3, pH 11.4) removes proteins that are peripherally associated with membranes via electrostatic or hydrophilic interactions respectively. The association with host cell membranes of both PipB-2HA and PipB2-2HA resisted disruption by these treatments. The non-ionic detergent Triton X-100 (TX-100) solubilizes integral membrane proteins, as shown for the integral membrane protein, calnexin (Fig. 7B). Interestingly, a significant proportion of the PipB-2HA and PipB2-2HA signals were not solubilized with TX-100 (Fig. 7B). Such TX-100 insolubility at low temperatures is a biochemical characteristic of proteins associated with sphingomyelin-rich and cholesterol-rich membrane microdomains, often termed detergent-resistant microdomains (DRMs) or lipid rafts. Accordingly, caveolin-1, a major integral membrane component of caveolae (Brown and London, 1998), which are specialized DRMs, was significantly resistant to solubilization by TX-100 at 4°C (Fig. 7B). Densitometric quantification of the intensity of each signal in the insoluble fraction, expressed as a percentage of the total signal intensity (i.e. soluble plus insoluble signals) (Fig. 7C) (n = 3 expts), revealed that 30.8 ± 7.2% of the total PipB-2HA and 35 ± 6.9% of the total PipB2-2HA were resistant to TX-100 extraction at 4°C. Similarly, 57 ± 4.4% of the total caveolin-1 was found in the TX-100 insoluble fraction, whereas less than 10% of total calnexin, an integral membrane protein not known to be associated with DRMs, was TX-100 insoluble. These results suggest that PipB and PipB2 partition into DRMs.

From a biochemical point of view, DRMs are characterized by both their resistance to solubilization by certain non-ionic detergents at low temperatures and their buoyancy in sucrose density gradients (Brown and London, 1998). To demonstrate conclusively that PipB and PipB2 accumulate in these membrane microdomains, we isolated DRMs from infected HeLa cells on the basis of their insolubility in TX-100 and low buoyant density in sucrose gradients. HeLa cells were infected and harvested 17–18 h pi as described above. The cells were then lysed, solubilized in TX-100 and loaded onto the bottom of a 40/30/5% discontinuous sucrose gradient. After ultracentrifugation fractions were collected from the top and aliquots were analysed by SDS–PAGE and immunoblotting (Fig. 8). Cytosolic proteins and solubilized proteins remain in the 40% sucrose layer, whereas insoluble DRM-associated proteins float to the 30/5% sucrose interface (Chamberlain and Gould, 2002; Drevot et al., 2002). In accordance with published lipid raft isolations (Chamberlain and Gould, 2002; Drevot et al., 2002), protein was detectable only in the high-density fractions (fractions 10–12), and not the low-density fractions, using Ponceau Red staining (not shown). In fact, it has been estimated that protein recovered in raft fractions from sucrose buoyant density isolations accounts for only 0.7–2% of the total protein (Chamberlain et al., 2001; Rothenberger et al., 2002). Caveolin-1, a protein that accumulates in DRMs, was detected in fraction 4 (representing the 5/30% sucrose interface) and fractions 9, 10, 11 and 12 (30/40% sucrose interface and 40% sucrose) (Fig. 8). This confirms that fraction 4 is enriched in DRMs. Under our experimental conditions, we never found caveolin-1 to be restricted to fraction 4, which likely highlights the dynamic association of proteins with DRMs and also the sensitivity of raft isolation to TX-100/protein ratio (Parkin et al., 1999; Chamberlain and Gould, 2002) and detergent used (Roper et al., 2000). In contrast, two mammalian proteins that do not associate with DRMs, calnexin and LAMP-1, were not detected in fraction 4 (Fig. 8). In agreement with the TX-100 insolubility data presented in Fig. 7(B, C), PipB-2HA and PipB2-2HA were present in the low buoyant density fraction, fraction 4 (Fig. 8). Like for caveolin-1, some of the PipB-2HA and PipB2-2HA was not associated with rafts. Only bacterially translocated PipB-2HA and PipB2-2HA were detectable by immunoblotting in the collected fractions as the vast majority of bacteria (>99.9%) were removed prior to sucrose density gradient centrifugation (see Experimental procedures). In confirmation of this, we were unable to detect specific signal by immunoblotting for two bacterial membrane proteins, a Salmonella major outer membrane protein or the β-subunit of ATP synthase (results not shown). From these results, we conclude that PipB and PipB2 partition into DRMs after translocation into host cells.

Figure 8.

PipB and PipB2 associate with detergent-resistant microdomains. HeLa cells were infected with SL1344 ΔpipB ppipB-2HA or ΔpipB2 ppipB2-2HA and harvested 17–18 h pi. Cells were lysed and centrifuged to remove host cell nuclei and cytoskeleton and intact bacteria. The post-nuclear supernatant was subject to buoyant sucrose gradient centrifugation. Fractions (1 ml) were collected from the top (numbered 1–12) and equal volumes of each were separated by SDS–PAGE and subject to immunoblotting. Antibodies used were anti-HA to detect PipB-2HA and PipB2-2HA, anticaveolin-1, anticalnexin and anti-LAMP-1. Numbers on the left refer to molecular mass markers in kDa. Sucrose concentrations are indicated at the top.

Discussion

Herein we describe a new Salmonella effector, PipB2, which has sequence similarity to a previously identified SPI-2 TTSS effector, PipB. PipB and PipB2 are members of a family of pentapeptide repeat-containing proteins that spans both the prokaryotic and eukaryotic kingdoms. To date, they are the only type III effectors described from this family. Other members of this pentapeptide repeat family have been implicated in a diverse range of processes, including heterocyst formation in cyanobacteria (Black et al., 1995; Liu and Golden, 2002), organization of the cellular actin cytoskeleton in plants (Banno and Chua, 2000) and antibiotic resistance in bacteria (Montero et al., 2001; Tran and Jacoby, 2002). However, the function of the pentapeptide repeat motif has not been elucidated. Molecular modelling predicts that the pentapeptide repeats form a superhelical structure, with each pentapeptide repeat forming one β-strand and three repeats comprising one turn of a right-handed β-helix (Bateman et al., 1998). It is likely that this extended helical structure is involved in protein–protein interactions (Bateman et al., 1998). Our future work on PipB and PipB2 aims to elucidate the contribution of these repeat domains to protein structure and function.

Our data indicate that not all SPI-2 TTSS substrates are secreted in vitro. Bacterial growth in low phosphate minimal media at an acidic pH induces the secretion of SseB, SseC and SseD (Nikolaus et al., 2001) and SseF-M45, SseG-M45, SseJ-M45, SifA-M45 and SifB-M45 (Hansen-Wester et al., 2002). However, three known substrates of the SPI-2 TTSS, PipB, PipB2 and SseI/SrfH are not secreted to the bacterial surface or into the culture medium after growth under the same conditions (this paper; M. Hensel, unpublished results). The lack of secretion of PipB2-M45 was not due to incorrect folding as translocated PipB2-M45 could be detected in host cells by immunofluorescence (L. Knodler and O. Steele-Mortimer, unpubl. res.). Our results imply that the composition of PCN-P media is sufficient for induction of genes encoding the SPI-2 TTSS and its effectors, but it is not sufficient for complete protein secretion by the SPI-2 TTSS in vitro. This may reflect a hierarchy of secretion as has been demonstrated for Yersinia Ysc type III secretion (Wulff-Strobel et al., 2002), or that the stimuli required for secretion of PipB and PipB2 are not present in PCN-P medium at acidic pH. This illustrates that caution should be exercised when drawing conclusions from in vitro secretion data. As described herein for PipB and PipB2, a putative type III effector that is not secreted into growth media may nevertheless be translocated into host cells. Evidently we cannot yet duplicate in culture media all the environmental signals that Salmonella encounter within host cells.

PipB2 is targeted to SCVs and Sifs in host cells, a localization pattern in common with other SPI-2 TTSS effectors including PipB (Knodler et al., 2002), SseJ and SifB (Freeman et al., 2003), SseF and SseG (Kuhle and Hensel, 2002) and SifA (Brumell et al., 2002). The prevalence of Salmonella effectors associating with Sifs in vitro supports one hypothetical role of Sifs, to provide a scaffold for the attachment of bacterial proteins (Knodler et al., 2002). However, to date, the formation of Sifs has only been reported in cultured cell lines and their role in Salmonella pathogenesis remains unknown. Assuming that Sifs are indeed contiguous with the SCV, the association of Salmonella effectors with these structures highlights the importance of the vacuolar membrane as a frontline of protection against the host cell's arsenal of antimicrobial defences. By associating with the vacuolar membrane, the dividing line between host and bacteria, Salmonella effectors can interact directly with host cell signalling and also remain in ‘contact’ with intraphagosomal Salmonella. It could be predicted from their shared localization pattern that these type III effectors act cooperatively to protect intracellular bacteria. Such overlapping functions have certainly been demonstrated for SPI-1 TTSS effectors (Zhou and Galan, 2001) and may be one explanation as to why most SPI-2 TTSS effector deletion mutants have little or no phenotype in replication and virulence assays.

We also report the unique targeting of PipB2 to peripheral vesicles in epithelial cells. Trafficking of PipB2-positive vesicles to the cell periphery is dependent on microtubules but not the formation of Sifs. It is evident from their host cell marker composition that these vesicles are not equivalent to any defined host cell compartment (L. Knodler and O. Steele-Mortimer, unpubl. res.) and we are currently investigating in more detail the nature of these vesicles and their physiological significance to Salmonella infection. Interestingly, the N-terminal 225 residues of PipB2 do not appear to be sufficient for this vesicular targeting. Because this N-terminal truncation is not translocated as efficiently as the full-length construct (Fig. 7A and our unpublished results) we cannot discount that undetectable amounts of PipB2 N225-2HA decorate peripheral vesicles. However, in stark contrast to the full-length construct, there is certainly no large accumulation of peripheral staining observed for the N-terminal truncation. The observation that PipB2 is uniquely targeted to these peripheral vesicles does suggest that PipB and PipB2 may have acquired different functions. Through evolution, Salmonella has acquired several type III effector ‘duplicates’– proteins that share a significant degree of sequence similarity (e.g. PipB and PipB2, SopE and SopE2, SifA and SifB, SopD and SopD2). With time, it is likely these ‘duplicate’ effectors have diversified to provide overlapping, or even new, functions. The finely tuned RhoGTPase specificity of SopE and SopE2 is a classic example of this (Friebel et al., 2001). It is feasible that PipB and PipB2 can functionally compensate for each other, but their differential localization in the host cell suggests they may act on a different subset of host molecules. While the exact contribution of these type III effectors remains elusive, our virulence studies indicate that the functions of PipB2 are more critical than PipB to the pathogenesis of S. typhimurium in the mouse model of infection. It should be noted, however, that PipB contributes significantly to enteropathogenesis as assessed in bovine-ligated ileal loops (Wood et al., 1998). Whether PipB2 is also required for Salmonella pathogenesis in other hosts remains to be determined. In host cells, neither PipB nor PipB2 appear to be overtly involved in vacuolar membrane homeostasis, as for SifA and SseJ (Beuzon et al., 2000; Ruiz-Albert et al., 2002), or modulate endosomal fusion events, as for SifA, SseF or SseG (Stein et al., 1996; Guy et al., 2000; Kuhle and Hensel, 2002). However, it remains possible that PipB and PipB2 moderate more subtle aspects of vacuolar trafficking that we have not yet investigated. One intriguing possibility is that PipB and PipB2 modulate signal transduction events that are required for bacterial fitness in the intracellular environment. Certainly the localization of these type III effectors in lipid rafts would facilitate their interactions with such host cell signalling pathways.

Host cell membrane association is a recurring theme for bacterial effectors. In the case of Salmonella, the SPI-1 TTSS effectors SigD/SopB (Marcus et al., 2002) and SipC/SspC (Scherer et al., 2000) localize to host cell membranes. Of the known SPI-2 TTSS effectors, PipB and PipB2 (this paper), SopD2 (Brumell et al., 2003), SseF and SseJ (Kuhle and Hensel, 2002) and SifA (Brumell et al., 2002) (Boucrot et al., 2003) are all targeted to host cell membranes as determined by subcellular fractionation. For at least a subset of the Salmonella type III effectors, their N-terminal domains are not only sufficient for type III translocation, but also membrane association (this paper; L. Knodler and O. Steele-Mortimer, unpubl. res.; Brumell et al., 2003). Other examples of membrane-associated type III effectors include YpkA, a serine/threonine kinase translocated by Yersinia to the inner leaflet of the plasma membrane (Hakansson et al., 1996), the ExoS toxin from Pseudomonas aeruginosa (Pederson et al., 2000), for which membrane association is critical for its cytotoxic activity (Pederson et al., 2002), and the translocated intimin receptor (Tir) from enteropathogenic Escherichia coli (EPEC) (Kenny et al., 1997). A number of possible scenarios could explain the shared association of Salmonella effectors with membranes. For example, these bacterial proteins may bind a common protein that is an integral component of host cell membranes, namely endosomal membranes. This ligand could be either another bacterial protein, such as SseC or SseD, which are predicted to associate with the vacuolar membrane (Nikolaus et al., 2001), or a mammalian protein. Alternatively, these Salmonella effectors may be post-translationally modified within host cells. Covalent lipid modifications such as prenylation and/or fatty acylation are sufficient for membrane association of many mammalian proteins (McCabe and Berthiaume, 1999) and interestingly some type III effectors from P. syringae (Nimchuk et al., 2000). Such covalent modifications can also contribute to the association of both integral and peripheral membrane proteins with DRMs (Melkonian et al., 1999), microdomains in which Salmonella effectors accumulate (this paper). If these type III effectors are modified, then glycosylphosphatidylinositol (GPI)-anchoring can be ruled out, at least for PipB, PipB2, SopD2, SifB and SseJ. Studies of these five effectors have utilized epitope tagging at their C-terminus, but GPI-anchored proteins are cleaved at their C-terminus prior to GPI addition. Because these bacterial proteins can be detected by immunofluorescence with anti-HA antibodies on host cell membranes, a GPI-moiety has not been added to their C-termini. Recently, the C-terminal hexapeptide sequence of SifA was shown to be essential for its membrane anchoring in host cells (Boucrot et al., 2003). Interestingly, this sequence has homology to isoprenylation motifs found on rab proteins, suggesting that lipidation of SifA at its C-terminus occurs after translocation into host cells. Another explanation for the membrane association of Salmonella effectors is that they simply insert into the host cell membranes, as has been described for Tir from EPEC (Kenny et al., 1997). PipB and PipB2 are predicted by TMpred (Hofmann and Stoffel, 1993) to contain membrane-spanning regions and are not extracted from membranes by treatment with 1 M NaCl or high pH (this paper), but not all Salmonella type III effectors are predicted to be transmembrane (L. Knodler and O. Steele-Mortimer, unpubl. data). Additional studies are required to elucidate the specific sequences that contain this subcellular localization information.

DRMs allow the recruitment and concentration of molecules involved in cellular signalling which provides a spatial means to control these pathways. Thus DRMs serve as important regulatory sites within host cells. DRMs are found on both the plasma and intracellular membranes, although comparatively little is known about intracellular DRMs. In BHK cells, the presence of raft-like domains on late endosomes, which are enriched for cholesterol, GPI-anchored proteins and flotillin-1, but devoid of rab7 and LAMP-1, has been reported (Fivaz et al., 2002). DRMs have also been isolated from the Golgi complex (Gkantiragas et al., 2001), recycling endosomes (Gagescu et al., 2000) and lysosomes (Taute et al., 2002). Flotillin-1 (a classic raft domain marker)-enriched raft domains are present on latex bead phagosomes, but interestingly the intracellular protozoan Leishmania donovani actively inhibits the acquisition of lipid rafts by phagosomes (Dermine et al., 2001). Here we report that Salmonella type III effectors can associate with DRMs. Using immunofluoresence, PipB and PipB2 are only detected on endosomal membranes, and not the plasma membrane. We thus deduce that PipB and PipB2 concentrate in intracellular DRMs that are present on the membranes of SCVs and Sifs. Given that other Salmonella effectors also localize to these structures, it is possible that this observation will not be restricted to PipB and PipB2. Our data is supported by previous observations about the composition of SCVs and Sifs – SCVs have a high-cholesterol content (Garner et al., 2002) and can sequester up to 30% of the total cellular cholesterol at late stages of infection (Catron et al., 2002) and Sifs are labelled by filipin (Brumell et al., 2001), a cholesterol-binding compound. This is the first report of type III effectors interacting with intracellular DRMs although pathogens have been shown to exploit plasma membrane rafts to gain entry into host cells (Norkin, 2001; Lafont et al., 2002) and endocytosed toxins can traffic to intracellular destinations in host cells by associating with DRMs (Abrami et al., 1998; Norkin, 2001). Our results suggest that intracellular DRMs are also involved in bacterial pathogenesis, more specifically mediating Salmonella-induced signal transduction events in host cells. The partitioning of type III effectors into these regulatory sites may allow intracellular bacteria to modulate the activities of proteins located within these rafts, leading to the activation or deactivation of specific host cell signalling pathways. Future studies will reveal whether pathogenic targeting of these intracellular signalling platforms is a common phenomenon and expand on the significance of DRMs in the pathogenesis of bacterial infections.

Experimental procedures

Bacterial strains and plasmids

Bacterial strains used in this study were as described previously unless otherwise stated (Knodler et al., 2002). For construction of the ΔpipB2 mutant, the pipB2 open reading frame (ORF) flanked by approximately 900 bp of upstream region and 750 bp of downstream region was amplified from S. Typhimurium SL1344 chromosomal DNA with the oligonucleotides B2 KO-1 (5′ ACC GCA GTA GAT ATC TAC CAG 3′) and B2 KO-4 (5′ TTC AGA TCG TTA ATC ACA ACA AAC 3′). The resulting amplicon was cloned into PCR2.1 TOPO vector (Invitrogen) to generate pipB2 TOPO. Using this plasmid as a template, inverse PCR was performed with the oligonucleotides B2 KO-2 (5′-A CGC GTC GAC GTG CAT GAT AAA ATT TAT CAT ATA G-3′) and B2 KO-3 (5′-A CGC GTC GAC CAA ACA CTC TTT AAC GAA TTT-3′) and Elongase (Invitrogen). The PCR product was purified, digested with SalI (engineered restriction site underlined above in oligonucleotide sequences), and self-ligated to give ΔpipB2 TOPO. The ΔpipB2 fragment was excised with SacI/XbaI, ligated into the corresponding sites of the positive suicide vector pRE112 (Edwards et al., 1998) and transformed into E. coli SY327λpir. The ΔpipB2 deletion mutant was then constructed by allelic exchange into S. Typhimurium SL1344 wild type, SL1344 ΔpipB (Knodler et al., 2002) or SL1344 ΔsifA (Stein et al., 1996) and confirmed by PCR analysis. The ΔSPI2::kan mutant was constructed using the λ Red recombinase method (Datsenko and Wanner, 2000). Oligonucleotide primers were designed to amplify the kanamycin resistance gene from plasmid pKD4. Oligonucleotide SPI-2 KO-F anneals upstream of ssrB: 5′-GGC ACA GTT AAG TAA CTC TGT CAC TTT ATG AAC CTG TAG CTT TCT CAT CAT TGT AGG CTG GAG CTG CTT CG-3′. Oligonucleotide SPI-2 KO-R anneals downstream of ssaU: 5′- TCG GTA GAA TGC GCA TAA TCT ATC TTC ATC ACC ATA CGT AAC AAG GCT GCA ACG CAT ATG AAT ATC CTC CTT AG-3′. Purified PCR product was electroporated into S. Typhimurium SL1344 wild type carrying the λ Red recombinase expression plasmid pKD46. Kanamycin-resistant colonies were subsequently cured of pKD46, which is a temperature-sensitive plasmid, by growth at 37°C. Primers flanking the deleted region were used in PCR to verify deletion of an approximately 25 kb genomic sequence and replacement by a kanamycin resistance gene. Finally, the mutant allele was moved into fresh S. Typhimurium SL1344 background by P22 transduction.

The complementing plasmid ppipB2-2HA is a pACYC184 (New England Biolabs) derivative that encodes, under the control of the pipB2 promoter, PipB2 tagged at the C-terminus with tandem HA epitopes. The pipB2 ORF and approximately 500 bp of upstream region was amplified from SL1344 chromosomal DNA with the oligonucleotides pipB2-Sal (5′-A CGC GTC GAC ACG GCT CTA CTA CTC GAT AG-3′) and pipB2-R-Bgl (5′-GGA AGA TCT AAT ATT TTC ACT ATA AAA TTC GTT-3′). The PCR product was subsequently digested with SalI/BglII. The two HA tags were amplified from pACB-2HA (Knodler et al., 2002) with HA-F-Bgl (5′-GGA AGA TCT TTT TAT CCG TAT GAT GTG CCG-3′) and pipB-C3 (5′-AAG CTT GTT TAT AAA ATC CCT TTA TCT CGA-3′) and the resulting amplicon cloned into pCR2.1 TOPO (Invitrogen). The PCR product was released from the cloning vector by a BglII/HindIII digest and ligated with the SalI/BglII-digested pipB2 ORF into SalI/HindIII-digested pACYC184. A construct expressing only the N-terminal 225-amino-acid residues of PipB2 tagged at the C-terminus with two HA epitopes was similarly constructed. This N-terminal region was amplified with pipB2-Sal (see above) and pipB2-225R-Bgl (5′-GGA AGA TCT ATC GAG GGT AGC GCC ACA CAT-3′) to generate ppipB2 N225-2HA. The construction of plasmids for the expression of gene fusions to the M45 epitope tag was performed as previously described (Hansen-Wester et al., 2002). Primers PipB-For-EcoRI (5′-CTA GAA TTC ATT TTG CTC TGT TTG CGG G-3′) and PipB-Rev-SmaI (5′-CGA CCC GGG AAA TAT CGG ATG GGG GAA A-3′), and PipB2-For-EcoRI (5′-AGC GAA TTC ACT TTG CTG CAT CGT CA-3′) and PipB2-Rev-EcoRV (5′-GCG GAT ATC AAT ATT TTC ACT ATA AAA TTC GTT-3′) were used to amplify PipB and PipB2 respectively. The PCR products were double-digested with SmaI/EcoRI for PipB or EcoRV/EcoRI for PipB2 and subcloned in plasmid p2062 (Hansen-Wester et al., 2002) to generate 3′ fusions to the M45-tag. The inserts were transferred into the low-copy plasmid pWSK29 (Wang and Kushner, 1991) to yield p2603 and p2621, expressing pipB-M45 and pipB2-M45 respectively. Plasmids were introduced by electroporation into S. Typhimurium wild type or mutant strain P2D6 [ssaV::mTn5 (Shea et al., 1996)], deficient in the SPI2-encoded type III secretion system. All plasmids were verified by DNA sequencing.

Virulence studies in mice

Female BALB/c mice (6–8 weeks old) were obtained from Jackson Laboratories and used for all mixed infection studies. The protocols used were in direct accordance with guidelines drafted by the University of British Columbia's Animal Care Committee and the Canadian Council on the Use of Laboratory Animals. Because all of the mutants we tested in these virulence assays were non-polar deletion mutants without resistance markers, we electroporated the two strains to be competed with low-copy number plasmids that vary only in their resistance markers. pWSK29 is ampicillin resistant while pWSK129 is kanamycin resistant (Wang and Kushner, 1991). With these plasmids, selective media can be used to distinguish between the two competing strains. Bacteria harbouring these plasmids were grown shaking overnight at 37°C in 10 ml of Luria–Bertani (LB) with appropriate antibiotics to stationary phase. Wild-type and mutant bacteria were diluted in phosphate-buffered saline (PBS) and mixed in equal proportions. Serial dilutions were plated on LB agar containing ampicillin (100 µg ml−1) (to select for pWSK29-bearing bacteria) or kanamycin (50 µg ml−1) (to select for pWSK129-bearing bacteria) to determine the colony-forming units (cfu) of each input strain. A total of 1 × 105 bacteria in a volume of 0.3 ml were injected intraperitoneally into each mouse. Mice were sacrificed 48 h post-inoculation by cervical dislocation, the infected spleens were removed and homogenized in PBS. Wild-type and mutant bacteria were enumerated by serial dilutions onto LB agar containing ampicillin or kanamycin. The competitive index was calculated by dividing the ratio of mutant/wild-type strains in the output (splenic cfu) by the ratio of mutant/wild type in the input (initial inoculum cfu). All experiments were performed at least twice with a minimum total of 13 mice.

PipB2 expression and secretion i n vitro

For the analysis of PipB2 expression in culture media, ΔpipB2 ppipB2-2HA bacteria were grown under SPI-1- or SPI-2-inducing conditions. For induction of genes comprising the SPI-1 regulon, bacteria were grown standing overnight at 37°C in LB broth. Induction of the SPI-2 regulon in N-minimal media was as described previously (Knodler et al., 2002). Proteins from equivalent numbers of bacterial cells, as determined by OD600 readings, were separated on 10%, 12% or 15% SDS–PAGE gels, transferred to nitrocellulose and blocked in Tris-buffered saline containing 0.1% (v/v) Tween 20 (TBST) and 5% (w/v) powdered milk for 1–2 h at room temperature. Blots were then incubated with the following primary antibodies in TBST milk overnight at 4°C: mouse α-HA (1:2000) (Covance) to detect epitope-tagged PipB or PipB2, mouse α-DnaK (1:1000) (Stressgen) and rabbit α-SseB (1:30 000) (Beuzon et al., 1999). Secondary antibodies, goat α-mouse HRP (1:10 000) or goat α-rabbit HRP (1:10 000), were applied for 1 h at room temperature in TBST milk, followed by chemiluminescent detection (cell signalling).

The secretion of M45-tagged derivatives of PipB and PipB2 under in vitro conditions was analysed as previously described (Hansen-Wester et al., 2002). S. Typhimurium strains were grown for 16 h in 200 ml of PCN-P minimal media at pH 5.8. Total cell fractions and fractions containing precipitated protein detached from the cell surface or secreted into the culture supernatant were separated by SDS–PAGE on 10% Tricine gels and transferred to nitrocellulose. Immunoblotting was with hybridoma supernatants against the M45 epitope (courtesy of Dr P. Hearing) or polyclonal α-SseB (Beuzon et al., 1999).

Bacterial infection of mammalian cells

HeLa (human cervical adenocarcinoma cell line, ATCC CCL2) and RAW264.7 (murine macrophage-like cell line, ATCC TIB-71) cells were grown as described previously. Bacteria were grown to late-log phase prior to infection of epithelial cells (Knodler et al., 2002). RAW264.7 cells were infected with opsonized stationary-phase bacteria as described elsewhere (Knodler et al., 2002). Details of invasion and replication assays in HeLa and RAW 264.7 cells have also been described previously (Knodler et al., 2002).

Immunofluorescence analysis and treatment with inhibitors

HeLa cells and RAW264.7 cells were grown on 12 mm glass coverslips overnight in 24-well tissue culture dishes. Bacterial infection conditions for HeLa cells have been described in detail elsewhere (Knodler et al., 2002). RAW264.7 cells were grown as described above, but primed with 100 U ml−1 recombinant mouse IFN-γ (R and D Systems) for 24 h prior to infection. After the addition of opsonized bacteria, IFN-γ was no longer included in the culture medium. Cells were fixed at 10 h pi for HeLa cells or 12 h pi for RAW 264.7 cells in 2.5% paraformaldehyde (PFA) at 37°C for 10 min, followed by extensive washing in PBS. Unless otherwise stated, fixed cells were permeabilized in PBS containing 10% (v/v) goat serum and 0.1% (w/v) saponin for 30 min at room temperature. Alternatively, for CI-M6PR staining, cells were permeabilized in PBS containing 10% (v/v) goat serum and 0.05% (v/v) Triton X-100 (TX-100) for 10 min at room temperature. Primary and secondary antibodies were applied in PBS-saponin-goat serum for 45 min to 1 h at room temperature. Coverslips were mounted in Mowiol (Aldrich) onto glass slides. Primary antibodies and the dilutions used were: rabbit α-Salmonella LPS (1:500) (Difco), mouse α-human LAMP-1 (1:100) (Developmental Studies Hybridoma Bank clone H4A3), rat α-mouse LAMP-1 (1:100) (Developmental Studies Hybridoma Bank clone 1D4B), mouse α-LBPA (1:100) (courtesy of Dr Jean Gruenberg), rabbit α-CIM6PR (1:100) (courtesy of Dr Stéphane Méresse), mouse α-HA (1:1000) (Covance), rat α-HA biotin (1:50) (Roche Applied Science), rabbit α-cathepsin D (1:200) (courtesy of Dr Stuart Kornfeld). Secondary antibodies were: Alexa Fluor 488 goat α-rabbit IgG, Alexa Fluor 594 goat α-rabbit IgG, Alexa Fluor 488 goat α-mouse IgG, Alexa Fluor 594 goat α-mouse IgG or Alexa Fluor 488 goat α-rat IgG all at 1:800 dilution (Molecular Probes). Fluorescence was examined with a Zeiss Axiovert S100 microscope attached to a Bio-Rad Radiance Plus or a Zeiss Axiovert Zoom LSM510 laser scanning confocal microscope with the 63× oil objective. Images of 512 × 512 pixels or 1024 × 1024 pixels were acquired using Bio-Rad Lasersharp software or Zeiss Axiovert LSM510 software respectively. Sections were assembled into flat projections and imported into Adobe Photoshop.

For inhibitor treatments, inhibitors were added to infected HeLa cells 9.5 h pi. Monolayers were incubated for a further 30 min in the case of BFA, NDZ or WTM, or for 20 min with cytochalasin D. Stock solutions of inhibitors (all from Sigma) were diluted at least 2000-fold. Final concentrations used were: 1 µg ml−1 BFA, 5 µg ml−1 NDZ, 100 nM WTM and 1 µg ml−1 cytochalasin D. For each experiment, to assess the efficacy of inhibitor treatments, NDZ-treated cells were stained with α-alpha-tubulin (1:50) (Sigma) to visualize microtubules, cytochalasin D-treated cells for F-actin with Alexa Fluor 488 phalloidin (1:1000), WTM-treated cells for endosomal tubulation and BFA-treated cells for dispersal of the Golgi compartment.

Mechanical fractionation of infected host cells

For fractionation of infected host cells, the method was essentially as described previously (Marcus et al., 2002). Briefly, two 10 cm dishes of HeLa cells were infected with 25 µl late-log phase subculture for 10 min. Extracellular bacteria were removed by aspiration, the monolayers washed three times with PBS and then incubated in growth medium containing 50 µg ml−1 gentamicin for 90 min. This was replaced by fresh medium containing 10 µg ml−1 gentamicin for the remaining incubation time. For biochemical fractionation, cells were washed in ice-cold PBS, resuspended in 300 µl homogenization buffer (3 mM imidazole pH 7.4, 250 mM sucrose, 0.5 mM EDTA) containing Protease Cocktail III (Roche Applied Science) and mechanically disrupted by six passes through a 22 gauge needle. A low-speed centrifugation (8000 g) was used to pellet unbroken cells, host cell nuclei and cytoskeleton and bacteria (P1 fraction). After collecting the supernatant (250 µl), hot 1× SDS-sample buffer (300 µl) was added to the pellet. The supernatant was ultracentrifuged at 100 000 g to separate host cell membranes (P2 fraction) from the cytoplasm (S2 fraction). Hot 6× SDS-sample buffer (50 µl) was added to the S2 fraction and the P2 fraction was resuspended in 300 µl hot 1× SDS-sample buffer. Equal volumes of each fraction were separated on 10% or 12% SDS–PAGE gels, transferred to nitrocellulose and subject to immunoblotting as described above. Blots were then incubated with the following primary antibodies: mouse α-HA (1:2000) (Covance), rabbit α-calnexin (1:10 000) (Stressgen), mouse α-beta-tubulin (1:2000) (Sigma) or mouse α-DnaK (1:1000) (Stressgen).

Extraction studies of membrane-associated PipB and PipB2 were also performed on infected HeLa cells. Two 10 cm dishes were infected and mechanically fractionated for each experiment. After removal of the P1 fraction by a low speed centrifugation step as described above, the supernatant containing host cell membranes and cytosol was divided into four aliquots and ultracentrifuged at 50 000 g to pellet host cell membranes. The membranes were resuspended in 200 µl of one of the following extraction buffers: (i) 10 mM Tris-HCl pH 7.4, 5 mM MgCl2; (ii) 10 mM Tris-HCl pH 7.4, 5 mM MgCl2, 1 M NaCl; (iii) 0.2 M Na2CO3 pH 11.4, 5 mM MgCl2; and (iv) 10 mM Tris-HCl pH 7.4, 5 mM MgCl2, 1% (v/v) TX-100. Host cell membranes were then incubated on ice for 30 min with resuspension by repeated pipetting every 5 min and then subject to another ultracentrifugation at 100 000 g for 30 min. The soluble material was collected (200 µl) and 40 µl hot 6× SDS-sample buffer was added. The insoluble fraction was resuspended in 240 µl hot 1× SDS-sample buffer. Equal volumes of soluble and insoluble fractions were then analysed by SDS–PAGE analysis and blots treated as described above. Primary antibodies used and dilutions were as described above with the addition of rabbit α-caveolin-1 (1:5000) (Transduction Laboratories).

Isolation of detergent-resistant microdomains by sucrose gradient centrifugation

Detergent-resistant microdomains were isolated from infected HeLa cells essentially as described previously, with some modifications (Cheng et al., 2001). Briefly, 4 × 10 cm dishes of Salmonella-infected HeLa cells (see above section) were rinsed twice in PBS and gently scraped into 2 ml cold 10 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA (TNE) containing Protease Cocktail III (Roche Applied Science). Cells were collected by centrifugation at 1000 g for 10 min at 4°C and the pellet was resuspended in 1.5 ml cold TNE containing 1% (v/v) TX-100 and lysed on ice for 30 min. Nuclei, cellular debris and bacteria were pelleted by centrifugation at 4000 g for 10 min at 4°C. Bacterial enumeration on LB agar plates from samples taken before and after this centrifugation step confirmed that most of the bacteria (> 99.9%) were removed by this low speed centrifugation step. Next, 1 ml of cleared supernatant was mixed with 1 ml 80% (w/v) sucrose in TNE buffer and transferred to a Beckman 14 × 89 mm ultracentrifuge tube. The sample was successively overlaid with 6 ml 30% (w/v) sucrose in TNE and 3.5 ml 5% (w/v) sucrose in TNE. Samples were centrifuged at 200 000 g in a SW41 rotor for 16–18 h at 4°C. Afterwards, 1 ml fractions were collected from the top of the gradient (fraction 1) to the bottom (fraction 12). For fraction 12, 0.5 ml of TNE was added to the remaining 0.5 ml and the sample mixed vigorously to resuspend insoluble material from the bottom of the tube. Then, 5 ml of cold acidified acetone was added to each fraction and the samples were held overnight at −20°C. Samples were centrifuged at 3500 g at 4°C for 10 min, pellets washed twice with cold 70% ethanol and dried in a fume hood. Each pellet was finally solubilized in 200 µl hot 1× SDS-sample buffer. Aliquots (10 µl) of each fraction were subject to 10%, 12% or 15% SDS–PAGE analysis as described above. Primary antibodies used for immunoblotting were mouse α-HA, rabbit α-calnexin, rabbit α-caveolin-1 and mouse α-human LAMP-1 (1:500) (Developmental Studies Hybridoma Bank, clone H4A3).

Acknowledgements

We are grateful to Jean Celli for his critical reading of this manuscript. We kindly thank Rey Carabeo and Ted Hackstadt for their help with DRM preparations and their reading of this manuscript and Scott Grieshaber for his assistance with analysis of confocal microscopy images. We also thank Stuart Kornfeld, Jean Gruenberg, Patrick Hearing and Stéphane Méresse for providing antibodies. The monoclonal antibodies H4A3 and 1D4B developed by J. T. August were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. BAV is supported by a Canadian Digestive Disease Institute/Medical Research Council of Canada postdoctoral fellowship and is an Honorary Izaak Walton Killam Fellow. M.H. is supported by the Deutsche Forschungsgemeinschaft grants HE1964/4-3 and HE1964/8-1. B.B.F. is an International Research Scholar of the Howard Hughes Medical Institute and a Distinguished Investigator of the Canadian Institute for Health Research. B.B.F. is supported by the Canadian Institute of Health Research and the Howard Hughes Medical Institute.

Ancillary