SifA, a Type III Secreted Effector of Salmonella typhimurium, Directs Salmonella-Induced Filament (Sif) Formation Along Microtubules

Authors

  • John H. Brumell,

    1. Biotechnology Laboratory and Departments of Biochemistry and Molecular Biology, Microbiology and Immunology, University of British Columbia, Vancouver, BC, V6T-1Z3, Canada
    Search for more papers by this author
  • Danika L. Goosney,

    1. Biotechnology Laboratory and Departments of Biochemistry and Molecular Biology, Microbiology and Immunology, University of British Columbia, Vancouver, BC, V6T-1Z3, Canada
    Search for more papers by this author
  • B. Brett Finlay

    Search for more papers by this author
    • 1 The first two authors contributed equally to this study.


*Corresponding author: B. Brett Finlay, bfinlay@interchange.ubc.ca

Abstract

A unique feature of Salmonella enterica serovar typhimurium (S. typhimurium) is its ability to enter into (invade) epithelial cells and elongate the vacuole it occupies into tubular structures called Salmonella-induced filaments (Sifs). This phenotype is dependent on SifA, a Salmonella virulence factor that requires the SPI-2-encoded type III secretion system for delivery into host cells. Previous attempts to study SifA and other type III secreted proteins have been limited by a lack of suitable reagents. We examined SifA function by expressing SifA with two internal hemagglutinin epitope tags. By employing subcellular fractionation techniques, we determined that translocated SifA was membrane associated in infected HeLa cells. Confocal microscopy revealed that SifA associated with the Salmonella vacuole and with Sifs. Our analysis also revealed that microtubules serve as a scaffold for Sifs, and that SifA colocalizes with microtubules at sites of interaction between lysosomal glycoprotein-containing vesicles and Sifs. Treatment with the microtubule inhibitor nocodazole blocked Sif formation but did not prevent SifA translocation into the Salmonella vacuole. While polymerized actin has been observed on Sifs, this phenotype was transient and did not play a role in promoting or maintaining Sif formation. Our findings demonstrate the essential role of microtubule dynamics in the formation of Sifs and the utility of this epitope tagging strategy for the study of bacterial type III secreted proteins.

Salmonella enterica serovar typhimurium (S. typhimurium) is a facultative intracellular pathogen that is a major cause of gastroenteritis in humans and causes a disease resembling typhoid fever in genetically susceptible mice (1). This pathogen occupies a vacuolar niche inside infected host cells during the course of infection, although intracellular trafficking of this vacuole appears to be unique in different cell types (2). Using epithelial cells, in vitro studies have demonstrated that S. typhimurium modulates its interactions with the host endosomal system immediately after invasion via the Salmonella pathogenicity island (SPI)-1-encoded type III secretion system. Early trafficking of the Salmonella-containing vacuole (SCV) is characterized by transient interaction with early endosomes (3), the rapid acquisition of lysosomal glycoproteins such as lysosomal-associated membrane protein (LAMP)-1 and the avoidance of late endosomes and lysosomes during the first 2–3 h post infection (4).

Several hours after internalization, when S. typhimurium have initiated intravacuolar replication, the SCV begins to elongate into tubular filaments called Salmonella-induced filaments (Sifs). Sifs are large, stable structures that can be visualized in infected epithelial cells by immunostaining lysosomal glycoproteins and other endocytic markers such as the vacuolar type proton ATPase (5,6). We have recently demonstrated that Sif formation involves a delayed fusion of late endocytic compartments with the Salmonella-containing vacuole (7). Importantly, the small molecular weight GTPase Rab7 is present on Sifs and its activity is required for their formation. While these studies demonstrated that the membrane required for Sif formation was derived, at least in part, from late endocytic compartments, the scaffold for these structures and the mechanisms that mediate their formation are currently unknown.

The Sif phenotype is dependent on the actions of sifA, a major virulence factor of S. typhimurium (8). Mutation of sifA blocks Sif formation in epithelial cells, blocks intracellular survival in macrophages (5,9,10), and attenuates virulence in the mouse typhoid model (8). Transfection of uninfected epithelial cells with a SifA–GFP fusion protein mediates swelling and tubulation of endocytic compartments in a manner similar to S. typhimurium-infected cells, suggesting that SifA acts within host cells and is necessary and sufficient to alter the host endocytic system (9). Based on its similarity to other effectors of the SPI-2 type III secretion system, and the requirement of a functional SPI-2 system for Sif formation, it is strongly suspected that SifA is an SPI-2 effector. However, conclusive evidence of this hypothesis is lacking, and the subcellular location of SifA during Sif formation is not known.

In addition to blocking Sif formation in epithelial cells, deletion of sifA also leads to loss of the Salmonella-containing vacuole in macrophages both in vitro (5) and in vivo (11). It is not known if Sif formation and maintenance of the SCV are separable functions; however, it is becoming clear that this virulence factor affects membrane trafficking in infected host cells. To determine the mechanism of SifA function, we have examined its delivery into host cells using a novel method of epitope tagging type III secreted effector proteins of a bacterial pathogen. This method allowed us, for the first time, to visualize the localization of translocated SifA in infected HeLa cells, and demonstrated its association with host membranes. SifA associated with microtubules, which we found to be the cytoskeletal structures that support Sifs. These findings demonstrate the highly localized actions of SifA in modulating membrane traffic along microtubules in infected epithelial cells.

Results

A novel method for epitope tagging type III secreted effectors

Previous attempts to study the localization of SPI-2 effectors in infected host cells using affinity-purified rabbit polyclonal antibodies have been unsuccessful. Our efforts have included the use of antibodies raised to five Secretion system effector (Sse) proteins and the putative SPI-2 effectors SifA and SifB. The inability to detect SPI-2 effectors in infected cells could be due to insufficient immunoreactivity of our polyclonal antibodies or, more likely, low abundance/instability of SPI-2 effectors upon delivery into host cells. As an alternative method of studying the localization of SifA, we generated a series of bacterial plasmids encoding SifA fused to epitope tags at the C-terminus of the protein. A 300-bp upstream region encoding the sifA promoter was sufficient to drive expression of Hsv- or Myc-tagged SifA in vitro, but we were unable to detect these proteins in infected cells.

Suspecting that the C-terminus of SifA might be unsuitable for modification with these epitopes, we generated a plasmid with two hemagglutinin (HA) epitope tags within the coding region for the SifA protein (psifA-2HA, see Materials and Methods). The use of tandem HA tags dramatically increases the immunoreactivity of tagged proteins and has been used successfully in our laboratory to label the C-terminus of a SPI-2 translocated effector (12). The two HA tags were inserted between amino acids 136 and 137 of the SifA protein (see Figure 1A). This site was chosen since it delineates the boundary between a conserved type III secretion signal at the N-terminus of SifA (13) and the C-terminal portion of the protein which has homology to a family of Gram-negative bacterial proteins (Figure 1B, see Discussion). HA-tagged SifA was efficiently expressed in S. typhimurium cultures in vitro under conditions that induce the expression of SPI-2 effectors (Figure 1C). Immunoblot analysis of a SPI-2 encoded protein, SseB, confirmed induction of the SPI-2 type III secretion system (TTSS) under these conditions. SifA expression was minimal under conditions known to induce the SPI-1 TTSS and its translocated effectors such as SigD (Figure 1C).

Figure 1.

Generation of a sifA expression plasmid with two internal hemagglutinin (HA) epitope tags. A. Site of insertion of the 2HA tags within the SifA protein, which bridges the conserved N-terminal type III secretion signal of SifA and a modular domain in the C-terminus of the protein. B. Partial multiple alignment of SifA with other Gram-negative bacterial proteins. Identical residues are in black, well-conserved residues in dark gray and least conserved in light gray. Alignments were performed with CLUSTAL W and colored with BOXSHADE through the Biology Workbench 3.2 server (http://workbench.sdsc.edu/cgi/bw.cgi). TrcA, TrcP and Orf19/MAP sequences are from enteropathogenic Escherichia coli. IpgB sequence is from Shigella flexneri. Note the invariant tryptophan residue found in all sequences (see asterisk). C. Expression of SifA in Salmonella enterica serovar typhimurium (S. typhimurium) cultures in vitro. Bacteria transformed with psifA-2HA were grown in cultures under conditions that lead to induction of either the SPI-1 or SPI-2 type III secretion systems and their translocated effectors (see Materials and Methods). Samples were then analyzed by SDS PAGE and immunoblotting. Expression of SifA-2HA was up-regulated under SPI-2-inducing conditions, as was the expression of another SPI-2 effector, SseB. Expression of SifA-2HA was minimal under SPI-1-inducing conditions, which lead to induction of the SPI-1 effector SigD. The relative amount of protein loaded was determined by blotting samples with antibodies to Dna kinase.

Membrane association of SifA in infected HeLa cells

To determine if HA-tagged SifA could be translocated into host cells, we infected HeLa cell cultures for 21 h with S. typhimurium expressing psifA-2HA. Subcellular fractionation of infected cells was then performed by mechanical disruption and differential centrifugation as previously described by our laboratory (14). Efficient separation of cytosol (C) and membrane (M) fractions was confirmed by immunoblotting with β tubulin (cytosol) antibodies and calnexin (a membrane-bound component of the endoplasmic reticulum) (Figure 2). Immunoblotting with antibodies to the HA epitope revealed that SifA was present in the membrane but not cytosolic fraction of infected HeLa cells. SifA was also detected in the insoluble pellet fraction (P) which contains bacteria and unbroken cells.

Figure 2.

Delivery of SifA-2HA into host cells and association with membranes. HeLa cells were infected with sifA mutants of Salmonella enterica serovar typhimurium (S. typhimurium) expressing psifA-2HA and subjected to mechanical lysis and fractionation of membranes by centrifugation (see Materials and Methods). Fractions were subjected to SDS PAGE and immunoblotting for the HA epitope tag, a membrane fraction marker (Calnexin) or a cytosol marker (Tubulin). As shown, SifA-2HA associated with membranes in cells infected with the sifA strain but not with the ssaR (SPI-2 mutant) strain of S. typhimurium.

In parallel experiments, HeLa cells were infected with S. typhimurium mutants in ssaR, an essential component of the SPI-2 type III secretion apparatus needed for Sif formation (9) and virulence in mice (15). HA-tagged SifA in the SPI-2 TTSS mutant was in the insoluble pellet fraction, which contained intact bacteria, indicating it is expressed under these conditions. However, delivery of SifA into either the cytosol or membrane fractions of HeLa cells was not observed. These findings demonstrate that HA-tagged SifA is expressed and translocated into host cells in an SPI-2-dependent manner where it associates with cellular membranes.

SifA directs Sif formation along microtubules

Expression of HA-tagged SifA complemented loss of the chromosomal sifA gene in S. typhimurium-infected HeLa cells by forming Sifs. As shown in Figure 3(B), immunostaining of LAMP-2 revealed elongation and tubulation of the SCV in sifA/psifA-2HA-infected cells that was indistinguishable from Sifs present in wild-type infected HeLa cells. Immunostaining with pre-absorbed antibodies to the HA epitope (see Materials and Methods) revealed that SifA-2HA colocalizes with lysosomal glycoproteins at the Salmonella-containing vacuole and along the length of the Sif tubules (Figure 3A). Interestingly, SifA-2HA did not fully colocalize with the LAMP-2-containing vesicles associated with Sifs that give a ‘bead on a string’ appearance to these structures (see inset of Figure 3C). Instead, SifA-2HA localized to the sites where these vesicles were in contact with Sifs. This suggests that SifA might modulate the fusion/fission machinery of late endosomes/lysosomes as they interact with Sifs. Costaining with antibodies to S. typhimurium lipopolysaccharide (LPS) confirmed that the localization of HA-tagged SifA is not limited to regions of the SCV immediately surrounding intracellular bacteria, but extends along Sifs throughout the cell (data not shown). The HA antibody had a moderate level of background staining; however, no specific intracellular signal was detected in ssaR (SPI-2 TTSS defective)-infected cells (Figure 3D–F). Thus, delivery of HA-tagged SifA requires the SPI-2 TTSS. These findings validate the use of this epitope-tagging method of studying SifA, and demonstrate the highly localized nature of its action during Sif formation following SPI-2 secretion.

Figure 3.

Functional complementation of Sif formation by psifA-2HA expression. HeLa cells were infected with either sifA (A- C) or ssaR (D- F) mutants of Salmonella enterica serovar typhimurium (S. typhimurium) SL1344 expressing psifA-2HA and incubated for 18 h. Infected cells were then fixed and immunostained for LAMP-2 (B,E, red in merged image) or the HA epitope (A,D, green in merged image) and analyzed by confocal microscopy (see Materials and Methods). Cells infected with sifApsifA-2HA bacteria were capable of forming Sifs that resembled those in wild-type infected cells. The SifA-2HA protein localized to Salmonella-containing vacuoles and was observed along the length of Sif tubules (see arrow in insets of upper panels). SifA also localized to sites where LAMP2+ vesicles were associated with Sifs (see arrowhead in insets of upper panels). No specific signal for the HA epitope was observed in ssaRpsifA-2HA infected cells. Size bar indicates 5 μm.

Sifs elongate from the Salmonella-containing vacuole in a linear pattern, suggesting they require a cytoskeletal scaffold. Co-immunostaining with LAMP-2 and tubulin demonstrated that Sifs are associated with microtubules (Figure 4A–C). Triple immunostaining with bacterial lipopolysaccharide (LPS), tubulin and HA-tagged SifA confirmed that SifA-2HA colocalizes with microtubules along the length of Sifs (data not shown). Treatment of infected cells with the microtubule inhibitor nocodazole disrupted Sifs (Figure 4D–F), consistent with a role for microtubules in Sif formation and maintenance (6). However, nocodazole did not prevent the delivery of SifA-2HA to the Salmonella-containing vacuole (Figure 4G–I). This demonstrates that the ability of SifA to alter host cell endocytic trafficking, in particular the elongation and tubulation of the SCV, requires an intact microtubular network.

Figure 4.

Microtubules are the host-cell scaffold for Sifs. HeLa cells were infected with wild-type Salmonella enterica serovar typhimurium (S. typhimurium) and incubated for 18 h. Infected cells were then fixed and immunostained for LAMP-2 (A,D, green in merged images C and F) or tubulin (B,E, red in merged images C and F) and analyzed by confocal microscopy (see Materials and Methods). Association of Sifs with microtubules can be visualized in C (see inset). Intracellular bacteria are indicated with an arrow. Where indicated, the microtubule disrupting agent nocodazole (2 μg/ml) was added to infected cells at 17 h post infection and incubated for a further 2 h. Cells were then fixed and processed for immunofluorescence as in A–C. Microtubule disruption blocked Sif formation (D–F) but did not block delivery of SifA-2HA (H, green in merged image I) into the Salmonella-containing vacuole (G–I). Size bar indicates 10 μm.

We have previously demonstrated that transfection of HeLa cells with a vector encoding SifA fused to the N-terminus of the enhanced mutant of green fluorescent protein (SifA-GFP) causes swelling and perinuclear aggregation of LAMP-1-containing compartments (9). Transfection also induces the formation of Sif-like tubules, which also colocalize with microtubules (data not shown). This suggests that SifA alone contains the necessary biochemical activity to alter host endocytic trafficking along microtubules to form Sifs. Interestingly, transfection of HeLa cells with a GFP fusion to SifB, a putative SPI-2 effector of S. typhimurium that is related (26% identical, 44% similarity) to SifA, did not lead to the formation of Sif-like tubules. Indeed, swelling and perinuclear aggregation of LAMP-1-containing compartments was seen in only 32 ± 7% of SifB-GFP-transfected cells, compared to 84 ± 2% of SifA-GFP-transfected cells and 25 ± 7% of GFP-transfected cells. Thus, while related, SifB does not appear to be functionally equivalent to SifA in host cells.

SPI-2-dependent actin rearrangements are not required for Sifs

It was recently demonstrated that polymerized actin is present on the SCV and on Sifs at 8 h post infection, a phenotype that, like Sif formation, is dependent on the SPI-2 TTSS (16). To examine a role for the actin cytoskeleton in mediating Sif formation, we infected HeLa cells for various times with S. typhimurium and stained actin filaments with Alexa 568-conjugated phalloidin (see Materials and Methods). Consistent with the findings of Méresse et al., polymerized actin was apparent on the SCV and Sifs at 8 h post infection in wild-type infected cells (Figure 5, upper panels). This phenotype required the SPI-2 TTSS, since ssaR mutants did not mediate actin polymerization on the SCV at 8 h post infection (Figure 5). As witnessed by Méresse et al., actin filaments were also visible on the SCV of sifA-infected cells, demonstrating that SifA is not required for this phenotype (data not shown). Consistent with these findings, we observed that transfection of HeLa cells with SifA-GFP had no effect on the actin cytoskeleton (data not shown). Treatment of cells with the actin-disrupting agent cytochalasin D did not block Sif formation [data not shown and (17)]. These results suggest the actin cytoskeleton does not play a role in initiating Sif formation, despite its recruitment to the SCV.

Figure 5.

SPI-2-dependent actin rearrangements. HeLa cells were infected with either wild type or ssaR (SPI-2 mutant) strains of Salmonella enterica serovar typhimurium (S. typhimurium) for either 8 or 18 h, as indicated. Fixed cells were then stained for bacterial lipopolysaccharide (blue in merged images), F-actin (using Alexa 568-conjugated phalloidin, red in merged images) or LAMP-1 (green in merged images). Size bar indicates 5 μm.

To determine if actin rearrangements play a role in the maintenance of Sifs, we examined cells infected for 16–24 h (Figure 5). At these later time points, disruption of the actin cytoskeleton in epithelial cells has been previously reported in S. typhimurium-infected MDCK epithelial cells (18). In our experiments with HeLa cells, wild-type bacteria were capable of disrupting actin stress fibers, confirming this report. However, polymerized actin was not detectable on the SCV at 18 h post infection (Figure 5), in contrast to that seen at 8 h. Focal contacts were not affected at this time-point; however, we did observe actin ruffling at the dorsal plasma membrane in these infected cells (data not shown). Interestingly, this phenotype was not witnessed in ssaR-infected cells (Figure 5, bottom panels). This suggests that SPI-2 effectors mediate both early actin polymerization on the SCV [at 8 h post infection (16)] and disruption of the actin cytoskeleton during laterstages of colonization of host cells (18). Importantly, Sif tubules were observed in approximately 45% of wild-type-infected cells and colocalized with microtubules 18 h post infection (e.g. Figure 4, upper panels). These findings demonstrate that SPI-2-mediated actin rearrangements are not required for either the formation or the maintenance of Sifs in infected HeLa cells.

Discussion

The study of translocated effectors of the SPI-2 TTSS and their intracellular targeting has been hampered by a lack of suitable reagents to follow delivery of these effectors into host cells. Here, we present the generation and use of an approach for the study of SifA, an important SPI-2 effector that modulates membrane trafficking inside host cells. HA-tagged SifA was translocated into host cells in an SPI-2 dependent manner, providing evidence that SifA is an SPI-2 effector. Furthermore, the ability of SifA-2HA to induce Sif formation demonstrates that its function was not impaired by the immunotagging strategy. We are currently testing the utility of this method for examination of the eight other proteins (9) that share a conserved N-terminal type III secretion signal with SifA.

C-terminal similarity of SifA to other Gram-negative bacterial proteins

The location for insertion of 2 HA tags in the coding sequence for SifA was chosen based on its similarity with other putative TTSS effectors. First, SifA has N-terminal similarity to a family of type III-secreted effectors from S. typhimurium that is limited to their first 130–150 amino acids (10,13,19). This region is required for type III secretion (13) and contains a number of predicted structural features that are conserved (9). Second, by performing PSI-BLAST analysis (http://www.ncbi.nlm.nih.gov/BLAST) of the C-terminus of SifA (amino acids 150–336), we observed a weak but consistent similarity between SifA and a group of proteins from other Gram-negative pathogens (see Figure 1B). The highest similarity was shared with SifB, a putative translocated effector of S. typhimurium, although its expression and delivery into host cells has not been examined. Our observation that SifB-GFP expression in HeLa cells does not alter endocytic compartments suggests that SifB performs a different function than SifA. As with the N-terminal secretion signal present on SifA, it seems likely that the C-terminus of SifA was acquired from a common ancestor and modified for a specific function, in this case modulation of host cell endocytic trafficking. That SifA retains its ability to form Sifs after insertion of two HA tags between the N- and C-terminal domains suggests that these regions fold separately, and that SifA has a modular structure. Furthermore, our identification of a putative ‘effector domain’ implies that other members of this divergent family which remain uncharacterized (e.g. IpgB) may also be TTSS effectors. Consistent with this hypothesis, Orf19/MAP was recently identified as a TTSS effector of enteropathogenic Escherichia coli (20).

A bacterial virulence factor that alters endosome traffic along microtubules

Our analysis of SifA revealed that microtubules are the cytoskeletal scaffold of Sifs in infected epithelial cells. Nocodazole treatment blocked Sif formation, and has also been previously shown to inhibit intracellular replication of S. typhimurium in epithelial cells (6). Delivery of SifA into the Salmonella-containing vacuole, however, was not impaired by nocodazole treatment. Indeed, we routinely observed an increase in the SifA-2HA signal immediately surrounding intravacuolar bacteria following drug treatment. This suggests that SifA is first delivered into the SCV via the SPI-2 type III secretion system, after which it moves along microtubules to directly modulate late endocytic trafficking and initiate Sif formation. Our observation that transfected SifA-GFP expression leads to the formation of Sif-like filaments along microtubules further supports this model. To the best of our knowledge, SifA is the first example of a bacterial virulence factor that directly alters endosome traffic along microtubules.

Actin rearrangements mediated by SPI-2 effectors

Here, we confirm a previous observation (16) that polymerized actin is present on the SCV at 8 h post infection. Since these actin rearrangements proceed in the absence of SifA [data not shown and (16)] and Sifs can be formed in the presence of agents that disrupt actin polymerization [data not shown and (17)], we conclude that actin polymerization is not required for the formation of Sifs. It is significant to note that SPI-2-dependent actin rearrangements are essential for maintenance of the SCV (16), also a function of SifA. These findings thus suggest that the mechanism by which actin rearrangements and SifA act to maintain the SCV are unique.

Our studies have also revealed the temporal nature of SPI-2-dependent actin rearrangements in epithelial cells: actin polymerization around the SCV/Sifs was maximal at 8 h post infection but was absent after 18 h. Loss of SCV-associated actin correlated with a disruption of actin stress fibers. This is in contrast to the prolonged presence of actin observed associated with the SCV during S. typhimurium infection in murine macrophages (16). Such a difference in actin accumulation may reflect a cell-type-specific adaptation of S. typhimurium trafficking (2). Disruption of the actin cytoskeleton in epithelial cells has been previously shown to be dependent on SpvB, an ADP-ribosylating enzyme encoded on the S. typhimurium virulence plasmid that utilizes actin as a substrate (18,21). Transfection of CHO cells with SpvB is sufficient to cause actin depolymerization (22). Our finding that disruption of the actin cytoskeleton occurs at late times post infection suggests that SpvB acts at later times to counteract the activity of unidentified SPI-2 effectors which initiate actin polymerization.

These findings highlight the complexity with which S. typhimurium modulates host cytoskeletal structures from within its intracellular niche using the SPI-2 encoded type III secretion system. This includes actin rearrangements on both the SCV and throughout the host cell, which occur in a temporally regulated fashion. Such a situation is not unlike actin rearrangements initiated by the SPI-1 TTSS, which are mediated by a battery of translocated SPI-1 effectors and which are down-regulated following bacterial invasion (23). Through the actions of SifA, the host-cell microtubule network is also exploited by intracellular S. typhimurium. By modulating endosome trafficking along microtubules, SifA initiates elongation of the SCV into Sifs, and promotes the integrity of this compartment. The study of intracellular pathogenesis by S. typhimurium is thus likely to provide a better understanding of mammalian cytoskeletal processes.

Materials and Methods

Cell culture and bacterial strains

HeLa (human epithelial cell line) cells were obtained from ATCC (Manassas, VA, USA). Cells were maintained in DMEM (Gibco BRL, Burlington, ON, Canada) supplemented with 10% FCS (Gibco BRL) at 37 °C in 5% CO2 without antibiotics. Cultures were used between passage numbers 10–25. Cells were seeded (5 × 104 cells/well) into 24-well plates with 12-mm sterile coverslips 18 h prior to each experiment. Where indicated, cells were incubated with either 5 μg/ml U18666A (Biomol, Plymouth Meeting, PA, USA), 2 μg/ml nocodazole (Sigma, Oakville, ON, Canada) or 5 μg/ml cytochalasin D (Sigma). Wild-type S. typhimurium SL 1344 was used for these studies. The sifA (8) and ssaR (9) mutants have been previously described.

Plasmid construction

The low-copy plasmid pACYC184 was used to express epitope-tagged SifA in S. typhimurium (NCBI accession # X06403). Primers JBO42 (5′- AGC AAG CTT ACA CGC ATC CAG GCA TGA AGT TTA TTC-3′) and JBO86 (5′-ACG TGT CGA CTT ATA AAA AAC AAC ATA AAC AGC CGC TTT G-3′) were used to amplify SifA and its upstream promoter from a genomic library of S. typhimurium SL1344. The polymerase chain reaction (PCR) product was digested with HindIII and SalI and cloned into the corresponding sites in pACYC184. Inverse PCR of the resulting plasmid was then performed with JBO87 (5′- CCG CTC GAG ATT TTA AAA TCG CAT CCA CAA ATG ACG GCC-3′) and JBO94 (5′-CCG CTC GAG CGC ATA ATC CGG CAC ATC ATA CGG ATA CGC ATA ATC CGG CAC ATC ATA CGG ATA ATC CGG GCG ATC TTT CAT TAA AAA ATA AAG-3′) using the Elongase kit (Gibco BRL), and the product of this reaction was digested with XhoI. Treatment with T4 DNA ligase (New England Brolabs, Beverly, MA, USA) yielded plasmid psifA-2HA, which encodes 2 hemagglutinin epitope (HA) tags immediately following the conserved N-terminal type III secretion signal (13) and between amino acids 136 and 137 of the SifA protein. psifA-2HA was transformed into S. typhimurium by electroporation and expression of epitope-tagged SifA from psifA-2HA was confirmed by growth in vitro in conditions that induce the expression of either SPI-1 effectors (late logarithmic growth in Luria-Bertani broth) or SPI-2 effectors (late logarithmic growth in N-minimal medium), as previously described (24,25). Samples of these cultures were sedimented, resuspended in protein sample buffer and subjected to SDS PAGE and immunoblotting on PVDF membranes.

For expression of SifA in epithelial cells, the N-terminal fusion of SifA to the enhanced GFP mutant has been described (9). For expression of SifB in epithelial cells, an N-terminal fusion to the enhanced GFP mutant was constructed by PCR amplification of sifB from a genomic library of S. typhimurium SL1344 using the following primers: forward 5′-AGC AAG CTT AGC ATG CCA ATT ACT ATC GGG AGA GGA TTT-3′; reverse 5′-TGT CGG TAC CCC ACT CTG GTG ATG AGC CTC ATT TTT TGT-3′. After digestion with HindIII and Kpn1 (sites underlined above), the PCR product was cloned into the corresponding sites in the multiple cloning region of the pEGFP-N1 N-terminal protein fusion vector from Clontech (Palo Alto, CA, USA). Plasmid DNA was purified using the Midiprep kit from Qiagen (Valencia, CA, USA) and used for transfection of cells with the FuGene 6 transfection reagent (Boerhinger Mannheim, Laval, Quebec, Canada) according to the manufacturer's instructions.

Bacterial infection of cell cultures

Late-log bacterial cultures were used for infections and prepared using a method optimized for bacterial invasion (3). Bacteria were grown for 16 h at 37 °C with shaking and then subcultured (1 : 33) in Luria-Bertani broth for 3 h. Bacterial inoculum were prepared by pelleting at 10 000 × g for 2 min and then directly resuspending in PBS, pH 7.4. The inoculum was diluted and added to cells at a multiplicity of infection of ≈?100 : 1 for 10 min at 37 °C. After infection, extracellular bacteria were removed by extensive washing with PBS and addition of growth medium containing gentamicin (50 μg/ml). For experiments in which cells were incubated for more than 2 h after infection with bacteria, the gentamicin concentration was decreased to 5 μg/ml at the 2-h time point. Mechanical fractionation of infected HeLa cells was performed essentially as described (14).

Processing cells for immunofluorescence microscopy

Cells were fixed in 2.5% paraformaldehyde (PFA) in PBS pH 7.2 for 10 min at 37 °C. Fixed cells were washed twice with PBS and permeabilized by treatment with 0.2% saponin (Calbiochem, San Diego, CA, USA) in PBS containing 10% normal goat serum (SS-PBS) for 1–16 h. Long blocking times (8–16 h) were essential for immunostaining with monoclonal antibodies to the HA epitope. Primary and secondary antibodies were overlaid on coverslips in SS-PBS for 60–90 min, followed by three washes with PBS. Coverslips were mounted onto 1-mm glass sides using Mowiol (Aldrich, Milwaukee, WI, USA). For triple labeling experiments, samples were first costained with monoclonal antibodies to the HA epitope and polyclonal antibodies to LAMP-2, followed by staining with monoclonal tubulin antibodies covalently conjugated to Cy3 and processed as above. Samples were analyzed using a Zeiss Axiovert S100 TV microscope (63 X objective) attached to a Bio-Rad Radiance Plus confocal microscope. Confocal sections were assembled into flat projections using NIH Image, imported into Adobe Photoshop in RGB format and assembled in Canvas 5 for labeling.

Antibodies

Rabbit polyclonal antibodies to SigD (generously provided by Dr Sandra Marcus) and SseB were generated by repeated immunization of New Zealand white rabbits with recombinant glutathione S-transferase (GST) fusions of each virulence factor (approximately 1 mg/injection). GST-SseB was constructed by PCR amplification of sseB from a genomic library of S. typhimurium SL1344 using the following primers: forward 5′-GCA GGA TCC ATG TCT TCA GGA AAC ATC TTA TGG-3′; reverse 5′-CGT GTC GAC TCA TGA GTA CGT TTT CTG CGC TAT-3′. After digestion with BamHI and Sal1 (sites underlined above), the PCR product was cloned into the corresponding sites in the multiple cloning region of the pGEX6P-1 protein fusion vector from Pharmacia Biotech (Baie d’Urfé, Quebec, Canada). This plasmid was transformed into the BL21 strain of E. coli for overexpression and purification using glutathione-coupled beads (Sigma). Raw antisera were affinity-purified using Sepharose beads (Pharmacia) with covalently coupled antigen. Nonspecific cross-reactivity of the antibodies was minimized by incubation with acetone powders of the BL21 strain that was used to overexpress the recombinant fusion proteins. Rabbit polyclonal antibodies to human LAMP-2 were generously provided by Dr M. Fukuda, La Jolla Cancer Research Foundation, USA (26). Murine monoclonal anti-human LAMP-1 (clone H4A3), anti-human LAMP-2 (H4B4) and monoclonal rat anti-murine LAMP-1(1D4B) were obtained from the Developmental Studies Hybridoma Bank, Iowa City, USA. Mouse monoclonal antibodies to β-tubulin, including those covalently conjugated to Cy3, were purchased from Sigma. Rabbit polyclonal antibodies to S. typhimurium LPS were obtained from Difco (Detroit, MI, USA). Mouse monoclonal antibodies to E. coli Dna Kinase (which cross-reacts with the S. typhimurium protein) and rabbit polyclonal antibodies to calnexin were purchased from Stressgen (Victoria, BC, Canada). Mouse monoclonal antibodies to the hemagglutinin epitope tag (MMS-101R; Covance) were used at a final dilution of 1 : 400 for immunofluorescence, following removal of nonspecific cross-reactivity by incubation with an acetone powder from HeLa cells. Secondary antibodies used were: Alexa 488-conjugated donkey anti-rabbit IgG, Alexa 488-conjugated goat anti-murine IgG, Alexa 568-conjugated donkey anti-rabbit IgG, Alexa 568-conjugated goat anti-murine IgG and Cy5-conjugated donkey anti-rabbit IgG, all from Molecular Probes (Eugene, OR, USA). Polymerized actin was stained with Alexa 568-conjugated phalloidin, also from Molecular Probes.

Acknowledgments

The authors wish to thank members of the Finlay lab for careful reading of this manuscript and Dr Sandra Marcus (University of Alberta) for providing affinity-purified polyclonal antibodies to SigD. We also thank Dr Elaine Humphrey of the Electron Microscopy Laboratory, University of British Columbia, for her assistance with confocal microscopy. This work was supported by grants (B.B.F.) and a postdoctoral fellowship (J.H.B.) from the Canadian Institutes of Health Research. B.B.F. is an International Research Scholar of the Howard Hughes Medical Institute and a Distinguished Investigator of the Canadian Institutes for Health Research. J.H.B. is an honorary fellow of the Izaac Walton Killam Memorial Foundation.

Ancillary