In the disease course of bacillary dysentery, pathogenic Shigella flexneri invade colonic epithelial cells and spread both within and between host cells. The ability to spread intercellularly allows the organism to infect an entire epithelial layer without significant contact with the extracellular milieu. Using fluorescence activated cell sorter (FACS)-based technology, we developed a rapid and powerful selection strategy for the isolation of S. flexneri mutants that are unable to spread from cell to cell. The majority of mutants identified using this strategy harbour mutations that affect the structure of their lipopolysaccharide or the ability of the bacteria to move intracellularly via actin-based motility; both factors have previously been shown to be essential for cell-to-cell spread. However, using a modified strategy that eliminated both of these types of mutants, we identified several mutants that provide us with evidence that bacterial proteins of the type III secretion system, which are essential for bacterial entry into host cells, also play a role in cell-to-cell spread.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Shigella species are the causative agents of bacillary dysentery, an enteric disease that is endemic throughout the world and is particularly prevalent in developing countries. Transmission of the organism is fecal–oral from person to person and is generally via direct contact. After ingestion, the bacteria pass the gastric acid barrier, invade the colonic mucosa and trigger an acute inflammatory response accompanied by mucosal ulceration and abscess formation (LaBrec et al., 1964; Takeuchi et al., 1968).
Cultured cell lines have been used to demonstrate more thoroughly that Shigella infection of eukaryotic host cells is a multistep process directed by a number of bacterial products that are secreted by a dedicated type III secretion system (for reviews, see Mecsas and Strauss, 1996; Anderson and Schneewind, 1999) upon bacterial contact with host cells (Ménard et al., 1994a; Watarai et al., 1995). The 30-plus genes that encode the type III secretion machinery as well as the bacterial effector proteins for entry are located on a 30 kb fragment of the Shigella virulence plasmid (Maurelli et al., 1985; Sasakawa et al., 1988; Kato et al., 1989). The major effectors of bacterial entry into epithelial cells appear to be IpaB, IpaC and IpaD proteins that, upon Shigella contact with host cells, are secreted by the bacteria into the extracellular milieu. IpaB and IpaC have been shown to form a complex in the extracellular medium (Ménard et al., 1994b) and the interaction of this complex with the mammalian cells is believed to be responsible for the induction of the eukaryotic cell cytoskeletal rearrangements that are required for bacterial uptake (Clerc and Sansonetti, 1987; Adam et al., 1995). This hypothesis is supported by the finding that latex beads coated with the IpaB–IpaC protein complex are taken up by epithelial cells (Ménard et al., 1996) and that microinjection of IpaC alone is capable of inducing actin polymerization in host cells (Tran Van Nhieu et al., 1999). Furthermore, recent work from several laboratories has shown that IpaB and IpaC insert into host cell membranes (Blocker et al., 1999; V. Cabiaux, personal communication) and this event has been proposed to play a role in the formation of a pore or translocation structure that facilitates the injection of additional bacterial effector proteins.
Following uptake by the host cell, Shigella lyse the phagocytic vacuole and spread throughout the cytoplasm either by inducing actin polymerization at one end of the bacterial surface, which serves to propel the bacteria through the cytoplasm (Prévost et al., 1992), or by migrating along pre-existing host cell cytoskeletal filaments (Vasselon et al., 1991, 1992). With time, the bacteria induce the formation of finger-like protrusions that extend from the infected cell and are eventually endocytosed by neighbouring cells (Bernardini et al., 1989; Kadurugamuwa et al., 1991). These bacteria-containing protrusions form predominantly at the perijunctional region of the host cells and appear as extensions of the adherens junctions; cadherins, α-catenin, β-catenin, α-actinin and vinculin, the major components of the adherens structure, have been observed to co-localize with the tail of the protrusion (Sansonetti et al., 1994). This phenomenon requires host cell components including actin, constituents of the adherens junctions and, we believe, signal transduction molecules.
Although the phenomenon of Shigella invasion has been fairly well characterized at the molecular level, not as much is known about bacterial factors essential for spread from cell to cell. Factors encoded by the virulence plasmid as well as the chromosome play a role in this phenomenon. The Shigella icsA gene, located on the virulence plasmid, has been shown to encode a 120 kDa outer-membrane protein that is responsible for the induction of actin polymerization at the pole of intracellular bacteria (Makino et al., 1986; Bernardini et al., 1989; Lett et al., 1989). As this protein is required for intracellular spread, it is also essential for spread from cell to cell. Furthermore, any defect in expression, stability or localization of IcsA also results in a deficiency in intercellular spread. Thus, mutations in the genes encoding VirF, a transcriptional activator required for icsA expression (Sakai et al., 1988), and VirK, a protein that is involved in the regulation of post-transcriptional levels of expressed IcsA (Nakata et al., 1992), result in a deficiency in bacterial spread from cell to cell. Furthermore, mutations in a number of genes that are involved in the synthesis of the lipopolysaccharide (LPS) inhibit intercellular dissemination (Okada et al., 1991; Rajakumar et al., 1994; Sandlin et al., 1995; Bosch et al., 1997; Hong and Payne, 1997). This inhibition has been attributed to the fact that removal of the O antigen side-chain of the LPS structure results in aberrant polar localization of IcsA (Sandlin et al., 1995; Bosch et al., 1997). Two additional Shigella flexneri genes have been reported to be specifically involved in cell-to-cell spread: the plasmid encoded icsB (Allaoui et al., 1992) and a chromosomal gene vacJ (Susuki et al., 1994). The proteins encoded by these two genes have been suggested to play a role in lysis of the two cellular membranes that surround the bacteria as it pushes from an infected cell into a neighbouring cell.
Given the complexity of intercellular spread, we believe that there is a variety of yet to be identified bacterial factors that are involved in this phenomenon – factors that may play a role in targeting the bacteria to the adherens junctions or perhaps that activate host cell processes that are necessary for the formation and endocytosis of bacterial protrusions. To identify these factors and the genes that encode them, we developed a rapid and powerful fluorescence activated cell sorter (FACS)-based selection strategy for the isolation of S. flexneri mutants that are unable to spread from cell to cell.
Development of a FACS-based strategy for the isolation of S. flexneri mutants that are deficient in cell-to-cell spread
Soon after entry into host cells, wild-type S. flexneri begin to multiply and spread both intracellularly and intercellularly. In contrast to disseminating wild-type organisms, mutants that are deficient in cell-to-cell spread are confined to the initially infected cells, which eventually become packed with bacteria. Dissemination of the bacteria in epithelial monolayers can easily be visualized by immunofluorescence by using Shigella strains that constitutively express the green fluorescent protein (GFP). Such an analysis reveals that the fluorescence signal emanating from cells infected with an icsA S. flexneri mutant that is unable to spread from cell to cell (Fig. 1B) is clearly of a greater magnitude than that of cells containing disseminating wild-type organisms (Fig. 1A). Thus, we wished to take advantage of this discrepancy in order to isolate cells containing S. flexneri cell-to-cell spread mutants via FACS-based technology. To that end, we transformed M90T and an icsA− organism with pFPV25.1, a pBR322-based plasmid containing GFP under the control of a constitutively active promoter. M90T/pFPV25.1 was shown to replicate, invade eukaryotic cells and spread both intracellularly and intercellularly to the same degree as the wild-type strain (data not shown). In order to test whether or not the FACS could be used to separate mutants deficient in cell-to-cell spread from wild-type organisms, we infected flasks of Caco-2 cells with M90T/pFPV25.1, icsA–/pFPV25.1, or a mixed inoculum of M90T/pFPV25.1 and icsA–/pFPV25.1 at a ratio of 1:1. After a 2 h infection, the cells were washed and then incubated at 37°C in the presence of gentamicin to kill extracellular bacteria. After a 4 h incubation period, cells were collected and analysed by flow cytometry. Using parameters set to discriminate on the basis of fluorescence intensity and forward scatter, infected cells (Fig. 1D and E) were readily distinguishable from uninfected cells (Fig. 1C). Fluorescent profiles were first obtained for M90T/pFPV25.1- and icsA-/pFPV25.1-infected cells (Fig. 1D and E respectively). These profiles revealed a substantial degree of heterogeneity in the fluorescence signal emanating from infected cells. Nevertheless, a subpopulation of highly fluorescent cells was indeed detected in the sample of icsA-/pFPV25.1-infected cells. The cells infected with the mixed population of the wild-type and mutant bacteria were then analysed and those cells displaying the highest fluorescence intensity were collected. This population represented 0.1% of the total cell population. The percentage of each strain recovered from this procedure could be calculated by plating the bacteria on TS agar plates containing either Ap and Sp (on which only icsA–/pFPV25.1 will grow) or Ap alone (to obtain the total number of bacteria recovered). Such an analysis revealed that 79.4 ± 18% of the recovered bacteria were icsA–/pFPV25.1.
These initial studies indicated that a FACS-based selection was indeed a feasible means of rapidly selecting for S. flexneri mutants deficient in intercellular spread. Thus, the subsequent stage of the project was to create a library of mutants to be tested in this system. Using the M90T/pFPV25.1 strain as our background strain, we created a bank of GFP-positive transposon mutants using a mini-Tn10 transposon derivative that was engineered to contain a lacZ reporter gene. A library of ≈ 10 000 independent mutants was thus created. Semipolarized monolayers of Caco-2 cells were infected with subsets of the bank of transposon mutants and, after a 4 h infection period, analysed using a flow cytometer. Again, fluorescent profiles obtained for M90T/pFPV25.1- and icsA–/pFPV25.1-infected cells were used to set the parameters for each of the subsequent sorts by FACS. The fluorescence profile of cells infected with subpopulations of the GFP-positive Tn10-lacZ library was obtained and the FACS gating parameters were set so as to collect only those cells displaying the highest fluorescence intensity. This population represented anywhere from 1% to 10% of the total cell population. Bacteria derived from this initial sort were either frozen until further analysis or used to infect a fresh monolayer of Caco-2 cells. The FACS procedure was thus repeated two more times. With each subsequent FACS sort, the population of highly fluorescent infected Caco-2 cells could be seen to increase in number (Fig. 2).
Three pools, each representing ≈ 2000 independent transposon insertion mutants, were individually put through the system, and anywhere from 30 to 75 individual clones isolated from each of the three serial FACS sorts were subsequently analysed by standard plaque assays to determine the efficiency of mutant selection. Clones that failed to form plaques or that formed unusually small plaques were scored as intercellular spread mutants (Fig. 3). After one FACS sort, ≈ 25% (32 out of 126) of the recovered bacteria were identified as cell-to-cell spread mutants; after two FACS sorts, approximately half (61 out of 125) of the recovered bacteria displayed a mutant phenotype; and, after three FACS sorts, all recovered bacteria (225 out of 225) were identified as deficient in cell-to-cell spread. Thus, the FACS-based strategy represents a remarkably powerful system for the isolation of bacterial mutants that are deficient in intercellular spread.
Characterization of mutants deficient in intercellular spread
The 225 clones recovered after three sequential FACS sorts and identified as deficient in cell-to-cell spread via the plaque assay were analysed by Southern hybridization using a probe that recognizes an internal fragment of the transposon. This analysis indicated that 13 different transposon mutants had been isolated, with each mutant being represented anywhere from 1 to 30 times. Using a variety of tissue culture assays, we were able to characterize more thoroughly the intracellular phenotype of these 13 mutants. Two of the 13 mutants, observed to be completely deficient in plaque formation (Fig. 3), were deficient for IcsA expression, as determined by Western blot analysis (data not shown). The remaining 11 mutants had a similar phenotype in that all formed very small plaques and appeared to be hyperinvasive (Fig. 3). This hyperinvasive phenotype suggested that these mutants may bear an altered LPS structure (Okamura et al., 1983; M. Bernardini, personal communication). Thus, we conducted a number of assays to examine the LPS of these mutants. In contrast to the wild-type strain, all 11 mutants failed to aggregate in the presence of antibody specific for the LPS structure, and Western blot analysis of crude LPS preparations indicated that all 11 mutants lack the O antigen side-chain. The site of transposon insertion was determined for one of these mutants (MR101) and was found to be located 112 nucleotides upstream of the ATG site of a gene that is homologous (99% identity observed for the 525 nucleotides sequenced) to the Escherichia coli rfe gene (Alexander and Valvano, 1994; Rick et al., 1994). The LPS of mutant MR101 was subsequently shown to display a normal core but to be completely deficient in O antigen (Fig. 4A).
As mentioned above, an additional 32 and 61 mutants recovered from a single or two sequential FACS sorts, respectively, had also been identified as deficient for cell-to-cell spread. All but one of these mutants proved to be deficient in LPS O antigen (69 out of 93) or displayed an IcsA phenotype (23 out of 93) (data not shown). The site of transposon insertion in the remaining mutant MR49 was determined to be located within a chromosomal region that has not yet been fully characterized in S. flexneri. The sequenced region exhibited 72% identity with the E. coli waaP gene over an internal region of 355 nucleotides and 76% identity with the Salmonella typhimurium waaP gene over an internal region of 413 nucleotides. Furthermore, a DNA region ≈ 2 kb downstream of the waaP transposon insertion site was also sequenced and showed significant homology to the E. coli and S. typhimurium waaY genes, suggesting some degree of conservation of this chromosomal region.
The waa chromosomal regions of E. coli and S. typhimurium contain the major core oligosaccharide assembly operons (for a review, see Heinrichs et al., 1998) and the waaP open reading frame (ORF) encodes a protein believed to be involved in the attachment of a phosphoryl substituent of the heptose region of the inner core. LPS was extracted from whole cell lysates of wild-type S. flexneri and mutant MR49 and the LPS obtained from the mutant was indeed observed to exhibit an altered core structure and displayed reduced amounts of the O antigen side-chains (Fig. 4A). Furthermore, mutant MR49, in contrast to the wild-type organism or LPS mutants displaying a normal core structure but lacking the O antigen, was sensitive to growth on agar plates containing 10–20 µg ml−1 novobiocin or 0.1% deoxycholate. In addition, MR49 displays a 35–50% reduction in the amount of OmpC present in the outer membrane compared with M90TpFPV25.1 and MR101 (Fig. 4B). These results are consistent with the observation that E. coli and Salmonella waaP mutants are sensitive to detergents and hydrophobic antibiotics and display a reduction in outer-membrane proteins (reviewed in Schnaitman and Klena, 1993), suggesting that the S. flexneri WaaP protein shares a similar function to that described for its homologues in E. coli and S. typhimurium.
The dissemination deficiency of S. flexneri mutants lacking the LPS O antigen has been attributed to the fact that such mutants display aberrant localization of the IcsA protein (Sandlin et al., 1995; Bosch et al., 1997) and, in turn, are deficient in actin tail formation (Rajakumar et al., 1994; Sandlin et al., 1995; Bosch et al., 1997; Table 1). Likewise, MR49 displayed a similar deficiency in polymerization of actin in infected HeLa cells; actin clouds were detected around the bacteria, but very few intracellular organisms (7%) were observed in association with a well-formed actin comet tail (Table 1). Using similar experimental conditions, 41% of wild-type organisms were found to promote comet tail formation. The observed cell-to-cell spreading defect of MR49 therefore is likely to be due to this defect in intracellular movement.
Table 1. Association of S. flexneri with polymerized actin.a
Absence of polymerized actin
Actin nucleation at one or both bacterial poles
Well-formed actin tail
HeLa cells were infected for 30 min, washed and incubated for 1 h. The cells were then fixed and stained with rhodamine–phalloidin to detect actin polymerization. The experiment was repeated three times and the results presented here are from one representative experiment.
M90T (wild type)
Modification of the FACS-based selection strategy for elimination of mutants deficient in the LPS O antigen or expression of IcsA
The isolation of a variety of clones displaying either an altered LPS structure or an IcsA-phenotype validated the power of the FACS-based selection strategy. However, the preponderance of LPS and IcsA mutants isolated significantly affected our ability to identify novel cell-to-cell spreading mutants. We therefore modified the experimental strategy (Fig. 5) so as to isolate only those mutants displaying a wild-type O antigen and to eliminate icsA mutants. Selection of mutants displaying the wild-type O antigen was accomplished by first incubating the recovered bacteria with a rabbit antiserum that recognizes the wild-type O antigen, and then using magnetic beads conjugated to antibodies specific for the rabbit IgG immunoglobulin to isolate the LPS-positive bacteria. Control experiments were conducted with mixed populations of S. flexneri LPS-positive and LPS-negative strains at a ratio of 1:10 and this particular selection step was determined to confer a 50-fold enrichment of LPS-positive strains. Portions of the Tn10-lacZ S. flexneri mutant library were thus sorted according to the FACS protocol already described and the collected bacteria were then subjugated to LPS selection. These selected bacteria were used to infect fresh monolayers of Caco-2 cells and one additional FACS sort was conducted. LPS selection was once again performed and individual clones were then tested in the plaque assay. Those clones that failed to form plaques were tested for expression of IcsA using a rapid immunoblotting procedure performed on plate-grown bacteria. Only those clones that expressed IcsA were further characterized. Two pools, representing ≈ 4000 independent transposon insertion mutants, were individually put through this modified selection strategy, and 100–125 individual clones isolated from each pool were subsequently analysed by standard plaque assays. The selection strategy failed to identify any intercellularly spread mutants from the first pool of potential mutants, but 36% (44 out of 123) of the bacteria recovered from the second pool were deficient in plaque formation. Of these 44 candidates, five failed to express IcsA, as determined by immunoblotting, and LPS analysis of the remaining 39 candidates indicated that all displayed normal LPS O antigen (data not shown).
Characterization of cell-to-cell spread mutants bearing a normal LPS structure
The 39 mutants bearing a normal LPS structure were analysed by Southern hybridization and this analysis indicated that four different transposon mutants had been isolated. The site of transposon insertion was determined for each of these mutants. One of the mutants, MR146, was found to have an insertion in the intergenic region between the divergently transcribed virA and icsA genes. S. flexneri mutants in which the virA gene itself has been inactivated have previously been demonstrated to exhibit a reduced ability to spread from cell to cell. This deficiency is believed to be due to a cis-acting effect on the icsA promoter because the phenotype cannot be complemented by a plasmid carrying a wild-type copy of virA (Uchiya et al., 1995; Demers et al., 1998). Although we were able to detect production of IcsA in MR146 by colony blot analysis, the level of IcsA expression in this clone was determined to be significantly reduced compared with that of the wild-type strain when whole bacterial lysates from broth-grown cultures were subjected to anti-IcsA Western blot analysis (Fig. 6). Hence, the observed dissemination defect of MR146 is likely to be due to this decrease in IcsA protein expression.
The second cell-to-cell spread mutant analysed, MR161, was found to contain a transposon insertion within the coding sequence (75 nucleotides downstream of the ATG translational start site) of icsB, a plasmid-encoded gene located just upstream of the ipa genes. An icsB mutant had been previously constructed by replacing an internal 250 bp AccI fragment with a 2 kb fragment carrying the interposon omega (Allaoui et al., 1992). This mutant, called SF126, is able to invade cells, lyse the phagocytic vacuole and form protrusions at the surface of infected cells, but is deficient in lysis of the double membrane that forms when the bacteria-containing protrusions extend into neighbouring cells (Allaoui et al., 1992). Immunoblot analysis using antisera specific for the various Ipa proteins showed that both MR161 and SF126 display decreased expression levels of IpaA, IpaB, IpaC and IpaD (Fig. 7A), indicating that the transposon and the interposon omega, respectively, exert a polar effect on the downstream ipa genes and that neither mutant can be used to assess properly the role of IcsB in Shigella intercellular spread. For this reason, a non-polar icsB mutant was constructed and shown to express normal levels of IpaA, B, C and D (Fig. 7B). This mutant was tested in the plaque assay and found to form plaques indistinguishable from those of the wild-type strain (data not shown). These data indicate that IcsB does not play an essential role in S. flexneri dissemination and that the spreading defect of MR161 and SF126 is attributable to the decreased expression of the downstream ipa genes.
Sequence analysis of the transposon insertion sites of the final mutants MR6133 and MR141 indicated that both contain an insertion within the plasmid-encoded ipgC gene. The insertion of MR141 is located directly upstream of the ATG translational start site whereas the insertion of MR6133 is 33 nucleotides upstream from the stop codon. Immunoblot analysis was conducted to evaluate the Ipa protein profiles of the mutants (Fig. 8A). As the IpgC protein acts as a molecular chaperone for IpaB and IpaC, preventing premature IpaB/IpaC association and protein degradation (Ménard et al., 1994b), it was not surprising that both MR6133 and MR141 display decreased expression levels of IpaB and IpaC. The transposon insertions do not disrupt normal transcriptional regulation of the ipa genes as expression remains temperature dependent (Fig. 8A). Furthermore, normal expression levels of IpaA and IpgD (Fig. 8A), proteins encoded by genes located both downstream and upstream of the insertion site, indicate that the transposon does not exert polar or cis-acting effects in either of the ipgC mutants isolated here.
Immunofluorescence analysis of infected Caco-2 cells indicated that mutants MR6133 and MR141 are invasive, can lyse the phagocytic vacuole and readily replicate in the cytoplasm (Fig. 8C). The invasive phenotype, although in accordance with the fact that the mutants were isolated using the FACS-based strategy, was unexpected because a previously constructed ipgC mutant SF619 was reported to be non-invasive (Ménard et al., 1994b). When tested in standard gentamicin protection invasion assays, however, a small percentage of SF619 bacteria could be recovered from either infected Caco-2 (Fig. 8B) or HeLa cells (data not shown). Using immunofluorescence microscopy to differentiate intracellular and extracellular bacteria, we confirmed that, in fact, some SF619 bacteria are internalized and can be observed replicating in the cytoplasm (data not shown). The number of intracellular SF619 organisms, however, is greatly reduced (≈10-fold) compared with the number of wild-type organisms internalized under the same experimental conditions (Fig. 8B). In contrast, mutants MR6133 and MR141 appeared to be almost as invasive as the wild-type organism, displaying less than a twofold decrease in invasiveness (Fig. 8B).
As demonstrated here, all three of the ipgC mutants, SF619, MR6133 and MR141, display similar Ipa protein expression profiles (Fig. 8A) and, thus, the discrepancy in invasion efficiency does not appear to be due to altered protein production and/or stability. Southern analysis indicated that both MR6133 and MR141 bear a single transposon insertion (data not shown). Furthermore, SF619 and MR141 were found to be fully complemented with a plasmid encoding a single copy of the wild-type ipgC gene; the complemented strains invade host cells (Fig. 8B;Ménard et al., 1994b) and form plaques indistinguishable from those of the wild-type organism (data not shown). We performed phage transduction to transfer the kanamycin (Kn) insertion of SF619 as well as the transposon insertion of MR141 into a wild-type background. Transduced SF619 (T-SF619) performed like its parent strain, displaying a deficiency in invasion as well as plaque formation in monolayers of host cells, whereas T-MR141 exhibited the ability to invade cells but, like the original MR141 mutant, was inhibited in plaque formation in infected Caco-2 cell monolayers (data not shown). Thus, the phenotypes of the ipgC mutants can be complemented and are indeed linked to the insertional elements, suggesting that the differences in the phenotypes of the ipgC mutants is due to the nature of the mutations rather than secondary genetic events. In addition to these studies, we also tested the ability of the two plasmids bearing the subcloned ipgC region of mutants MR6133 and MR141 to complement the inefficient invasion phenotype of SF619 (Fig. 8B). Whereas pMR6133 failed to complement the invasion deficiency of SF619, SF619pMR141, interestingly, was as invasive as MR141 itself.
Although we have been unable to identify precisely the reason for the differences in the invasive capacity of the ipgC mutants generated to date, the mutants isolated in this work are, nevertheless, interesting in that they are not only able to enter into host cells almost as efficiently as the wild-type bacteria but are also capable of lysing the primary phagocytic vacuole. This latter conclusion is drawn from observations that both MR6133 and MR141 form normal actin tails (Fig. 8C, right panel; data not shown) and, after replication, fill the cytoplasmic compartment of the infected cell (Fig. 8C, middle panel).
To understand better why mutants expressing decreased quantities of the Ipa proteins are unable to spread from cell to cell, electron microscopy analysis was conducted on Caco-2 cells that had been infected with the wild-type strain or the ipgC mutants for several hours. In accordance with the results obtained using immunofluorescence microscopy, ipgC mutants, in contrast to the disseminating wild-type strain, were often observed to fill the cytoplasm of their host cells (data not shown). Additional phenotypic differences were observed when special attention was paid to protrusion formation. Shown in Fig. 9A is an example of the protrusions formed by wild-type S. flexneri. The double membrane of the protrusion, which is eventually lysed, can still be seen in this image. In contrast, the ipgC mutants, when observed in association with protrusions, were seen either (i) surrounded by as many as six membranes (Fig. 9C and D) as if the bacteria were circling at the site of cell–cell contact and, with each new attempt to invade the neighbouring cell, accumulated another membrane or (ii) trapped but nevertheless replicating within a membrane-bound vacuole in the newly infected cell (Fig. 9B). Thus, the cell-to-cell spreading deficiency of mutants MR6133 and MR141 appears to be due, at least in part, to the inability to lyse the double membrane that forms as the bacteria-containing protrusion extends into the neighbouring cell.
We have reported here the development of a FACS-based selection strategy that allows for the rapid isolation of bacterial mutants that are unable to spread in a monolayer of host cells. This selection scheme permits not only an efficient enrichment of mutants deficient in dissemination but also selectively eliminates those candidates that are compromised for invasion or intracellular multiplication. This aspect of the mutant isolation strategy is appreciated when working with a pathogen in which numerous genetic factors are involved in both invasion and normal intracellular multiplication and renders the strategy preferable to traditional methods of bacterial mutant isolation which largely rely upon the cumbersome testing of individual candidates in a plaque assay.
Just as with any other strategy that has been developed to isolate S. flexneri dissemination mutants, our initial efforts resulted in the recovery of a preponderant number of clones deficient in expression of IcsA or the LPS O antigen. Such mutants were expected as they dramatically affect the spreading ability of this organism. Mutants deficient in IcsA expression are completely inhibited in intercellular and intracellular dissemination and, in fact, are unable to even move from that part of the cell that was initially invaded. The dissemination deficiency of LPS mutants has been ascribed to the fact that they have an altered polar expression of IcsA and are less efficient in actin tail formation. It remains formally possible, however, that the surface expression of additional outer-membrane proteins is also altered in such mutants and that this too influences the bacterium’s ability to spread from cell to cell. This certainly seems the case for one of the mutants, MR49, isolated in this work. As the transposon insertion in the waaP gene of MR49 renders the strain hypersensitive to hydrophobic compounds and results in the expression of an altered LPS core, the Shigella WaaP protein is likely to perform the same task as its counterparts in E. coli and Salmonella: phosphorylation of the heptose region of the core region. Loss of phosphorylated heptose is known to give rise to significant compositional and structural changes in the outer membrane of deficient bacteria (Heinrichs et al., 1998), and MR49 does indeed display a reduction in one of the outer-membrane porins, OmpC. Interestingly, ompC mutants are impeded for intercellular dissemination (Bernardini et al., 1993). It seems plausible that, in addition to OmpC, other outer-membrane proteins may also be altered in their expression or localization in the LPS core mutant isolated here. Whereas LPS mutants lacking the O antigen, in comparison with the wild-type strain, form small plaques on monolayers of epithelial cells, MR49 forms plaques that are further reduced in size and, in fact, are barely detectable (data not shown), suggesting that this clone is indeed impeded by pleotrophic affects resulting from an alteration in the outer membrane.
Southern analysis of a portion of the LPS mutants isolated in this study indicated that these clones represent a number of mutants containing insertions in different genes. As there are more than 50 genes involved in LPS synthesis of a Gram-negative organism, it is not surprising that several of these genes may have been ‘hit’ during the transposon mutagenesis. Furthermore, the increase in the invasive capacity of the LPS mutants may have also contributed to their isolation using this FACS-based strategy; the higher the number of fluorescent organisms that initially enter the cell, the higher the detectable fluorescence intensity after several rounds of bacterial replication. Thus, if several hyperinvasive mutants enter a cell, the fluorescent signal emanating from that cell several hours later will be of greater magnitude than that of a cell infected by a single wild-type organism. This phenomenon is true regardless of whether or not the organisms efficiently spread from cell to cell. Our results support this observation in that we isolated several S. flexneri clones that are significantly more invasive than the wild-type strain but, nevertheless, have a normal LPS structure and spread normally from cell to cell (data not shown). These clones are currently being analysed and lend credence to the possibility that the strategy described here could also be applied to the isolation of hyperinvasive pathogens.
The isolation of icsA and LPS mutants deficient in cell-to-cell spread validated our selection strategy but greatly reduced our chances of recovering additional mutants of interest. The incorporation of a selection scheme for clones expressing the wild-type O antigen and a screening procedure of IcsA expression resulted in the elimination of such mutants. The screening procedure used to detect IcsA is not foolproof as clones expressing decreased levels of the protein cannot be differentiated from those expressing normal protein quantities. Three mutants isolated from this modified selection strategy, however, were shown to have a wild-type LPS structure and normal levels of IcsA expression. A common characteristic shared by these three mutants is a decrease in Ipa protein expression. Mutants MR6133 and MR141 contain transposon insertions within the gene encoding IpgC, the molecular chaperone of IpaB and IpaC, and show reduced expression of IpaB and IpaC. These mutants are, nevertheless, able to enter into host cells almost as efficiently as the wild-type bacteria, indicating that the bacteria do not require the full complement of Ipa proteins to enter into host cells. This conclusion is supported by the fact that MR161 is also able to enter into host cells, albeit at a reduced efficiency, despite its decreased levels of IpaA, IpaB, IpaC and IpaD.
The original ipgC mutant, created by replacing an internal region of the gene with a Kn resistance cassette, enters into host cells very inefficiently and this phenotype was assumed to be due to decreased IpaB and IpaC expression. This does not appear to be the case based on our recent findings. Furthermore, the discrepancy in invasion efficiencies of the original ipgC mutant SF619 and the mutants isolated here appears to be due to the type of mutation rather than to secondary genetic events because the mutants retain their respective phenotypes after phage transduction into a wild-type background. The discrepancy in the phenotypes of the various IpgC mutants may be explained by several hypotheses. SF619 was created by replacing an internal region of the gene with an in frame insertion of a Kn resistance cassette. Thus, it is possible that the remaining N-terminal region of the protein is translated and that this product elicits a dominant negative effect on the proteins it binds to, namely IpaB and IpaC. This could arise, for example, if the translated N-terminal product bound to IpaB and IpaC but did not release the proteins, thereby preventing IpaB/IpaC dimerization in the extracellular environment. The fact that a plasmid bearing the ipgC region, including the transposon insertion, of MR141 can partially complement the invasion deficiency of SF619 suggests an alternative hypothesis, which may or may not be exclusive of the first. The transposon insertion of MR141, located directly upstream of the ATG translational start site, may have significantly decreased but not completely obliterated expression of a functional IpgC protein which is not present in SF619. If such a protein exists, however, it is either produced at such low levels that it is beyond the limits of detection or it is no longer recognized by the available antibodies as we were unable to detect any such protein in MR141 or MR6133 by Western blot analysis.
After entry into the host cell, the ipgC mutants isolated in this work, as well as the polar icsB mutants, are capable of lysing the primary phagocytic vacuole as they are observed to readily multiply throughout the host cell cytoplasm and to form actin comet tails. These mutants do not efficiently spread intercellularly, however, and are observed to accumulate in the initially infected host cells. This accumulation may be attributable to multiplication of bacteria that did indeed form a protrusion, but, owing to an inability to lyse the protrusion membrane, were reabsorbed by the initially infected cell. It is formally possible that the mutants also suffer from additional dissemination defects, such as a decreased efficiency in protrusion formation, for example, which is difficult to evaluate by electron microscopy. Using non-confluent HeLa cells, mutants MR6133 and MR141 can be observed in the tips of membrane-enclosed protrusions that extend into the extracellular milieu (data not shown). We are cautious in interpreting these results, however, as protrusion formation in non-confluent cells may simply reflect the bacterium’s ability to be propelled forward by actin comet tail formation and does not necessarily imply the ability to form efficient protrusions at the adherens junction of polarized epithelial cells. What is clear, however, is that the Ipa proteins are involved in lysis of the protrusion once it has formed. The data presented in this work do not permit us to assign this task to any one specific Ipa protein. Expression of the entire ipa operon is reduced in the polar icsB mutants and the ipgC mutants display not only reduced levels of IpaB and IpaC proteins, as shown here, but also IpaD (C. Parsot, personal communication; data not shown), the cause of which is presently unknown. Furthermore, we have found that ipgC mutants, like ipaB and ipaD mutants, have an unregulated type III secretion system; proteins are secreted by the secretion system in the absence of bacterial contact with host cells or other factors known to activate the secretion machinery (data not shown). This low level of constitutive secretion is probably due to the reduced levels of cytoplasmic IpaB and IpaD proteins, which are known to block the secretion machinery in the absence of inducing signals (Ménard et al., 1994a) and which makes it difficult to attribute lysis of the double membrane to a specific Ipa protein or proteins. Nevertheless, recent results from our laboratory have provided more definitive proof that IpaB and IpaC are involved in this process; using strains in which the expression of ipgC alone or ipgC in conjunction with ipaB or ipaC is under the control of an IPTG-inducible promoter, our co-workers have shown that protrusion lysis and dissemination of these strains is indeed dependent upon expression of IpgC, IpaB and IpaC (Page et al., 1999). Finally, our analysis of a non-polar icsB mutant indicates that IcsB, in fact, does not play a role in lysis of the double membrane nor in any other step of the dissemination process. The reassignment of this role to the Ipa proteins opens the door to speculation as to what role IcsB actually does play in Shigella virulence, a role that is strongly suggested by the gene’s position between the type III secretion genes (mxi/spa) and their effectors (ipa).
An important implication of the findings reported in this work and in the work of Page et al. (1999) is that bacterial proteins that are essential for the bacterial entry event and lysis of the vacuole continue to play a role in the infection process, notably in cell-to-cell spread. As the various mutants generated here are capable of lysing the primary vacuole but not the double membrane of the protrusion, it is possible that different quantities of Ipa proteins are required to accomplish these two different tasks. In other words, the quantity of proteins that adequately supports lysis of the vacuole formed shortly after uptake is not adequate for lysis of the double membrane formed during dissemination. Alternatively, intracellular bacteria may simply synthesize and secrete the Ipa proteins less efficiently compared with extracellular bacteria, and, as a result, mutants that are already handicapped by decreased Ipa protein levels are no longer able to secrete the minimal amount of proteins required for intracellular spreading events.
In conclusion, our findings suggest that the type III secretion system of S. flexneri continues to operate once the bacteria are inside host cells and that one of the roles of the secreted effector proteins is to facilitate membrane lysis as the bacteria spread throughout a monolayer of cells. Bacterial spread from cell to cell is not detected until 1 or 2 h after the entry event, and it seems likely that in order to lyse the double membrane the bacteria must orchestrate a localized secretion of the effector proteins. It is tempting to speculate that this secretion is activated upon bacteria interaction with the adherens junction; future experiments will be conducted to test this hypothesis.
Bacterial strains and growth conditions
The Escherichia coli strains used in this work are derivatives of E. coli K-12: DH5α and DH5αλpir were used for plasmid construction and SM10λ was used for plasmid transfer to S. flexneri. The S. flexneri strains used in this study were M90T, the wild-type, fully virulent serotype 5 strain and its derivatives, including SC557, an icsA mutant generated by random TnphoA insertional mutagenesis (Bernardini et al., 1989); SF619, the original ipgC mutant constructed by allelic exchange of the wild-type ipgC gene with a mutated gene containing the aphA-3 cassette (Ménard et al., 1993); and SF126, an icsB mutant constructed by insertion of the interposon omega (Allaoui et al., 1992). E. coli and S. flexneri strains were routinely grown in Luria–Bertani (LB) medium or trypticase soy (TCS) broth respectively. Antibiotics were used at the following concentrations: ampicillin (Ap), 100 µg ml−1; kanamycin (Kn), 50 µg ml−1; and spectinomycin (Sp), 200 µg ml−1.
Creation of the GFP-expressing S. flexneri transposon bank
A wild-type S. flexneri GFP-expressing strain was constructed by transforming M90T with pFPV25.1, a pBR322-based plasmid containing GFPmut3 under the control of a constitutively active S. typhimurium rpsM promoter (Valdivia and Falkow, 1996), and selecting for Ap resistance. Using the M90T/pFPV25.1 strain as our background strain, we created a bank of GFP-positive transposon mutants using a mini-Tn10 transposon derivative containing a lacZ reporter gene. This transposon derivative is useful in that it serves not only as a means of insertional mutagenesis but can eventually also be used as a transcriptional reporter of gene activity. The Tn10-lacZ derivative was constructed by ligating a BamHI–KpnI fragment bearing the lacZ gene from pGP704 (gift from Dr Olivera Francetic, Institut Pasteur, France) into the multicloning site (MCS) of pBSL180, a suicide R6K vector that carries a mini-Tn10 derivative conferring resistance to Kn (Alexeyev and Shokolenko, 1995). Using SM10λpir, the Tn10-lacZ transposon was conjugally transferred to M90T/pFPV25.1 and transconjugates were selected for by plating on media supplemented with Kn and Ap. A library of ≈ 10 000 independent mutants was thus created and divided into 16 different pools, which were then frozen at −80°C until further use. Random insertion of the transposon was verified by Southern hybridization using internal fragments of the transposon, containing either the Kn resistance cassette or the lacZ gene, as a DNA probe. In addition, a series of assays for β-galactosidase activity (Sambrook et al., 1989) after growth of individual mutants at 37°C were conducted; a variety of levels of transcriptional activation were observed, verifying that the transposon had inserted into different genomic regions. Recombinant DNA manipulations were carried out according to previously described protocols (Sambrook et al., 1989).
Individual mutants determined to be deficient in plaque formation on monolayers of Caco-2 cells (as described below) were further characterized using Southern hybridization to verify single transposon insertion sites as well as to determine the number of different mutants isolated. Total genomic DNA was isolated and digested with either EcoRV or a combination of HindIII and EcoRI and then separated on an agarose gel. The DNA was blotted onto nylon membranes and detected with a DNA probe specific for either the Kn resistance cassette or the lacZ gene of the Tn10lacZ construct.
To determine the precise site of transposon insertion of select mutants, HindIII–EcoRI or ClaI fragments of total DNA preparations were ligated into pBluescript (Stratagene) and DH5α was then transformed with the resulting constructs. Plasmid preparations were prepared from selected Kn-resistant colonies and the insert size of the cloned fragment was verified on an agarose gel. The nucleotide sequence of the DNA flanking the transposon was determined using primers that hybridize to the T3 and T7 promoter sequences of the cloning vector or that hybridize to an 18-nucleotide fragment located 6 bp from the end of the Kn resistance-encoding cassette.
A non-polar icsB mutant was constructed according to the same strategy used to construct SF126 (Allaoui et al., 1992) with the exception that an aphA-3 cassette (Ménard et al., 1993) was used in place of the interposon omega. The inserted aphA-3 cassette introduces an early translational stop but directs in frame translation of the remaining 3′ end of the icsB gene, thereby avoiding putative polar effects on transcription of downstream genes. Southern blot analysis and PCR amplification confirmed insertion of the aphA-3 cassette in the icsB gene of the newly constructed S. flexneri mutant, which in this work is referred to as icsB– (non-polar).
Complementation of MR141 and MR6133 with p17A (ipgC+) (Ménard et al., 1994c) required that the mutants be cured of the GFP-bearing plasmid as both plasmids bear an Ap resistance cassette. The mutants were grown to stationary phase in TCS broth containing Kn but not Ap. The culture was subcultured 1:100, again in Kn-containing TCS broth, and grown to stationary phase. This procedure was repeated one more time. Single colonies were plated on TCS agar plates containing Kn and, using a fluorescence microscope equipped with a filter specific for FITC (WIBA; Olympus), the plates were screened for GFP-negative colonies. MR141 GFP-negative colonies were successfully obtained and verified to be Ap-sensitive but still positive for binding to Congo Red on TCS agar plates. One of these clones was then transformed with p17A and transformants were selected for on agar plates containing Kn and Ap. Complementation was verified by Western immunoblot analysis.
The Tn10-lacZ insertions of the various ipgC mutants were moved to a clean background by P1 transduction (Sprott et al., 1994a).
FACS selection of S. flexneri mutants that are deficient in cell-to-cell spread
Caco-2 cells were seeded at 1 × 106 cells per 75 cm2 flask in Dulbeco’s modified Eagle medium (DMEM) containing 20% FCS and 1% penicillin/streptomycin and incubated in 10% CO2 for 4–5 days. One or two pools of the GFP-positive Tn10-lacZ library were used to inoculate broth supplemented with Ap and Kn and the culture was grown overnight with aeration at 37°C. The next day, the culture was diluted 1:40 in fresh broth, incubated with aeration at 37°C for 1 h, and added to flasks of Caco-2 cells in the absence of serum (multiplicity of infection of ≈ 50:1). M90T/pFPV25.1- and IcsA/pFPV25.1-infected cells were similarly prepared. After a 2 h infection at 37°C, the cells were washed and incubated in DMEM/10% FCS containing 50 µg ml−1 gentamicin to kill extracellular bacteria. Four hours later, the cells were washed and then lifted off the bottom of the culture flask by incubation in the presence of 4 ml 0.05% trypsin/0.02% EDTA for 10 min at room temperature. The trypsin solution was diluted out with 8 ml PBS/0.05 M EDTA and 100 µl FCS and the cells were then centrifuged at 1300 r.p.m. for 10 min. Finally, the cells were resuspended in 4 ml cold PBS/0.05 M EDTA and kept on ice. The infected cell suspensions were analysed and sorted using a Elite Coulter-Bekmann flow cytometer equipped with a Coherent 95.05 laser emitting at 488 nm. Infected cells were detected using parameters set to discriminate on the basis of fluorescence intensity and forward scatter. Fluorescent profiles were first obtained for M90T/pFPV25.1- and IcsA−/pFPV25.1-infected cells. Information from these initial scans was used to set the parameters for each FACS sort of cells infected with subpopulations of the GFP-positive Tn10-lacZ library. Infected cells displaying the highest fluorescence intensity were collected and added directly to 7 ml TCS broth (inducing the eventual lysis of the eukaryotic cells). The next day, portions of these cultures were either frozen at −80°C in TCS containing 50% glycerol or directly used to inoculate a fresh culture of TCS broth. The subcultured bacteria were used to infect a fresh monolayer of Caco-2 cells and the FACS procedure was repeated two more times. After the third FACS sort, the recovered bacteria were plated on TCS agar containing 0.01% Congo Red. Only those clones that were positive for Congo Red binding were further characterized, thus allowing us to eliminate any clones that, in the process of manipulation, had lost the virulence plasmid.
In order to select for S. flexneri clones bearing an intact LPS structure, the FACS strategy was modified as follows. Flasks of Caco-2 cells were infected, collected and fluorescently sorted as described above. After the primary FACS sort, the collected cells were incubated with 1.0% Triton X-100 for 10 min on ice to lyse the eukaryotic cells. The sample was centrifuged and the pellet resuspended in PBS/0.1% BSA containing 3.0% sodium citrate. This solution was incubated with 5 µl of rabbit anti-S. flexneri 5a serum for 30 min with agitation at 4°C. The samples were once again centrifuged and resuspended in the PBS/BSA/sodium citrate solution and then incubated with 50 µl of sheep anti-rabbit IgG Dynabeads M-280 (Dynal) for 30–60 min with agitation at 4°C. The beads were isolated using a magnetic particle concentrator (Dynal MPC) and washed five times with PBS/BSA to remove non-adherent bacteria. Finally, the beads were resuspended in 500 µl TCS and used to inoculate a fresh culture of TCS. This culture was used to infect a fresh batch of Caco-2 cells and the FACS procedure and LPS-positive selection was repeated. After the second sort and LPS selection, the bacteria bound to the Dynabeads were plated on TCS agar plates containing 0.01% Congo Red and, again, only those clones that were positive for Congo Red binding were further characterized.
Immunoblotting for IcsA expression by plate grown S. flexneri
A derivation of the methods of Bhuna et al. (1991) was used to detect IcsA expression by plate-grown bacteria. Individual Congo Red-positive colonies that had been FACS sorted and selected for wild-type expression of LPS were streaked on LB agar plates. After an incubation of 6 h at 37°C, an Immobilon P membrane was placed on top of the agar and the plate returned to 37°C overnight. The next day, the membrane was air dried for 30 min (all processing of the membrane was conducted at room temperature) and then incubated with 1% hydrogen peroxide/methanol for 30 min to eliminate endogenous bacterial peroxidases. The membrane was rinsed with PBS and blocked with 1.0% casein/PBS for 1 h. A monoclonal antibody to IcsA was then added to the blocking solution (1:1000 dilution) and incubated with the membrane for 1 h. The membrane was washed, incubated for 1 h with horseradish peroxidase-labelled goat anti-mouse secondary antibodies in PBS, and the detected proteins were visualized using enhanced chemiluminescence (Amersham Life Science).
Tissue culture assays
Plaque assays were performed according to previously described methods (Oaks et al., 1985). Briefly, confluent Caco-2 monolayers in 12-well Falcon dishes were infected in the absence of FCS for 2 h with ≈ 2 × 106 bacteria grown to mid-exponential phase. The cells were then washed and overlaid with fresh medium containing 10% FCS, 50 µg ml−1 gentamicin and 0.5% agarose. Plaques were analysed after an incubation period of 2–3 days at 37°C with 10% CO2.
Quantification of the efficiency of bacterial entry into host cells was conducted using bacteria grown to exponential phase. The bacteria were added (multiplicity of infection of 20:1) to monolayers of Caco-2 cells at ≈ 85% confluence. The bacteria were spun onto the monolayers at 2000 g for 10 min and the cells were then incubated at 37°C for 1 h. After infection, the cells were washed three times and incubated with DMEM containing FCS and gentamicin for an additional 1 h. Finally, the cells were washed and lysed with 1.0% Triton X-100 and the intracellular bacteria were plated on TCS agar plates. Results are presented as the percentage of the initial inoculum that was recovered.
To analyse bacterially infected Caco-2 cells by immunofluorescence microscopy, cells were grown on coverslips and then infected as described above. After a 1 h infection and a 4–5 h incubation with gentamicin, the cells were fixed with 3.7% paraformaldehyde and permeabilized with 0.1% Triton X-100/2% BSA for 10 min. To delineate the host cells, F-actin was labelled using phalloidin conjugated to rhodamine (Molecular Probes). Samples were mounted on slides and observed with a fluorescence microscope (BH2-RFCA; Olympus Optical) equipped with a dual band filter (FITC and TRITC; Olympus). Similar procedures were used to analyse actin tail formation with the exception that HeLa cells were used instead of Caco-2 cells and the post-infection incubation period was restricted to 1 h. These assay conditions were determined to be optimal for qualitative and quantitative analyses of actin polymerization.
Slide agglutination tests were performed using bacteria taken from agar plates suspended in S. flexneri group 5-specific antiserum (gift from Dr Armelle Phalipon, Institut Pasteur, France). LPS was extracted, as described by Hitchcock and Brown (1983), from whole cell lysates of equal numbers of bacteria. Electrophoresis of the LPS preparations in 16% polyacrylamide gels in the presence of sodium dodecyl sulphate (SDS–PAGE) was performed according to the methods of Laemmli (1970) and the gels were silver-stained as has been previously described (Sprott et al., 1994b).
Analysis of bacterial outer-membrane and Ipa proteins
Bacterial outer-membrane proteins were isolated from bacterial cultures that had been grown overnight at 37°C in the presence or absence of 0.3 N NaCl. Cultures were centrifuged and resuspended in sonication buffer containing 50 mM tris-HCl and 2 mM EDTA. The bacteria were sonicated three times for 15 s each with 15 s rest periods between each pulse. Unbroken intact cells were removed by centrifugation at 3000 g for 10 min and whole membranes were then pelleted at 14 000 g for 30 min. The membrane pellets were subsequently resuspended in sonication buffer containing 1% sodium N-lauryl sarcosinate for 30 min at room temperature. The samples were once again centrifuged at 14 000 g for 60 min and then resuspended in buffer containing 1% SDS. After electrophoresis in 12% SDS, proteins were detected using Coomassie brilliant blue. Scanning densitometry was used to quantify the amount of proteins detected; the prominent lower protein band between 29 and 43 kDa did not appear to vary significantly in the analysed samples and was thus used as an internal control for protein loading.
IcsA expression in S. flexneri strains of interest was determined by immunoblotting whole bacteria lysates with antisera specific for IcsA (gift from Dr Coumaran Egile, Institut Pasteur, France). Briefly, 1 ml of bacterial cultures grown overnight at 37°C was centrifuged and the pellet resuspended in 100 µl 2× sample buffer. Ten microlitres of each sample was loaded into a well of a 12% polyacrylamide gel. After electrophoresis in the presence of SDS, proteins were transferred to nitrocellulose membranes and immunoblotting was carried out with rabbit polyclonal antibodies directed against IcsA. Horseradish peroxidase-labelled goat anti-rabbit antibody was used as a secondary antibody and the detected proteins were visualized by enhanced chemiluminescence (Amersham Life Science).
To analyse the Ipa protein profile of various bacterial clones, overnight bacterial cultures grown at 30°C or 37°C were diluted to an optical density of 0.35 and 4 ml of this diluted culture was centrifuged and resuspended in 100 µl of SDS-containing buffer. Fifteen microlitres of each sample was loaded into a well of a 10–12% polyacrylamide gel. After electrophoresis in the presence of SDS, proteins were transferred to nitrocellulose membranes and immunoblotting was carried out with mouse monoclonal antibodies directed against IpaB (Bârzu et al., 1993), IpaC (Phalipon et al., 1992) and IpgD (gift from Dr Kirsten Niebuhr, Institut Pasteur, France) and with rabbit polyclonal antibodies raised against IpgC (Ménard et al., 1994c), IpaA (Tran Van Nhieu et al., 1997), and IpaD (Ménard et al., 1993). Horseradish peroxidase-labelled goat anti-mouse or anti-rabbit antibodies were used as secondary antibodies and the detected proteins were visualized by enhanced chemiluminescence.
We would like to thank Philippe Metezeau and Hélène Biasizzo-Kiefer of the Laboratoire de Cytométrie Analytique et Préparative, Institut Pasteur, France, for their expert technical assistance in performing the FACS analysis/sorting. We are grateful to Hélène Ohayon of the Station Centrale de Microscopie Electronique, Institut Pasteur, for her expert preparation and analysis of the electron microscopy samples. We are grateful to Dr Dana Philpott for critical reading of the manuscript. M.R. is supported by the Cancer Research Fund of the Damon Runyon–Walter Winchell Foundation Fellowship, DRG-1432. N.J. and A.A. are supported by a grant from the Belgium Fonds National de la Recherche Scientifique Médicale, convention 3.4611.99. A.A. also receives support from Actions de Recherche Concertées 98/03-224 de la Direction Générale de la Recherche Scientifique-Communauté Française de Belgique.