A Salmonella inositol polyphosphatase acts in conjunction with other bacterial effectors to promote host cell actin cytoskeleton rearrangements and bacterial internalization

Authors


Abstract

A central feature of Salmonella pathogenicity is the bacterium's ability to enter into non-phagocytic cells. Bacterial internalization is the consequence of cellular responses characterized by Cdc42- and Rac-dependent actin cytoskeleton rearrangements. These responses are triggered by the co-ordinated function of bacterial proteins delivered into the host cell by a specialized protein secretion system termed type III. We report here that SopB, a Salmonella inositol polyphosphatase delivered to the host cell by this secretion system, mediates actin cytoskeleton rearrangements and bacterial entry in a Cdc42-dependent manner. SopB exhibits overlapping functions with two other effectors of bacterial entry, the Rho family GTPase exchange factors SopE and SopE2. Thus, Salmonella strains deficient in any one of these proteins can enter into cells at high efficiency, whereas a strain lacking all three effectors is completely defective for entry. Consistent with an important role for inositol phosphate metabolism in Salmonella-induced cellular responses, a catalytically defective mutant of SopB failed to stimulate actin cytoskeleton rearrangements and bacterial entry. Furthermore, bacterial infection of intestinal cells resulted in a marked increase in Ins(1,4,5,6)P4, a consumption of InsP5 and the activation of phospholipase C. In agreement with the in vivo findings, purified SopB specifically dephosphorylated InsP5 to Ins(1,4,5,6)P4in vitro. Surprisingly, the inositol phosphate fluxes induced by Salmonella were not caused exclusively by SopB. We show that the SopB-independent inositol phosphate fluxes are the consequence of the SopE-dependent activation of an endogenous inositol phosphatase. The ability of Salmonella to stimulate Rho GTPases signalling and inositol phosphate metabolism through alternative mechanisms is an example of the remarkable ability of this bacterial pathogen to manipulate host cellular functions.

Introduction

Microbial pathogens that have sustained long-standing associations with their hosts have evolved sophisticated mechanisms to secure their replication in these environments (Galán and Bliska, 1996). Often these mechanisms involve very intimate interactions with host cells that result in the stimulation of specific and highly regulated cellular responses. An example of such a pathogen is Salmonella typhimurium. An essential feature of the pathogenic life cycle of this bacterium is its ability to enter into cells that are normally non-phagocytic, in particular those of the intestinal epithelium (Galán, 1999). The entry process is the result of carefully co-ordinated cellular responses triggered by a battery of bacterial proteins delivered into the host cell cytoplasm by a specialized organelle known as the type III secretion system (Galán and Collmer, 1999). The cellular responses triggered by Salmonella are characterized by marked actin cytoskeleton rearrangements, leading to membrane ruffling, macropinocytosis and subsequent bacterial internalization (Galán, 1999). These cellular responses are a consequence of the modulation of the activity of a subset of actin-organizing, small-molecular-weight, GTP-binding proteins of the Rho subfamily, in particular Cdc42 and Rac (Chen et al., 1996a; 1999; Hardt et al., 1998a). Salmonella triggers the activation of these Rho family GTPases by injecting a protein termed SopE, which acts as a potent exchange factor for Cdc42 and Rac (Hardt et al., 1998a). The rearrangement of the actin cytoskeleton triggered by Salmonella is further enhanced by the activity of SipA, another bacterial protein delivered into the host cell via the type III secretion system (Zhou et al., 1999a, b). SipA exerts its function by influencing different aspects of actin dynamics: it lowers the critical concentration of actin required for polymerization; it stabilizes F-actin filaments; and it enhances the bundling activity of actin-associated proteins such as T-plastin (Zhou et al., 1999a, b). The actin cytoskeleton rearrangements induced by Salmonella are reversible; therefore, shortly after bacterial infection, the cell regains its normal architecture (Takeuchi, 1967). Remarkably, this process is mediated by SptP, yet another type III-secreted bacterial protein that exerts its function by acting as a GTPase-activating protein (GAP) for Cdc42 and Rac (Fu and Galan, 1999). Thus, Salmonella has evolved the ability sequentially to activate and downmodulate the function of Rho GTPases to induce its own uptake into host cells in a remarkable ‘yin and yang’ interaction.

Although the molecular mechanisms of bacterial internalization can be explained by the sequential activity of the Salmonella proteins SopE, SipA and SptP, there is evidence that other bacterial-encoded proteins must also be involved in stimulating cellular responses that lead to bacterial uptake. For example, a Salmonella strain carrying a null mutation in sopE retains significant signalling capacity (Hardt et al., 1998b; Wood et al., 1996), arguing for the existence of a redundant effector protein(s) capable of mediating bacterial internalization. Such an effector protein(s) must exert its function through the activation of Rho GTPases, in particular Cdc42, as expression of a dominant-negative mutant of this small G protein results in the abrogation of bacterial entry (Chen et al., 1996a). Furthermore, such an effector protein must be delivered into the host cell by the type III secretion organelle because Salmonella mutant strains defective in this system are completely unable to stimulate cellular responses (Galán and Curtiss, 1989). A candidate protein for such activity is SopE2, a recently identified homologue of SopE, which has been reported in the S. typhi genome sequencing project (The Sanger Center, http://www.sanger.ac.uk/Projects/S_typi) and has been described recently in Salmonella dublin (Bakshi et al., 2000). An S. dublin strain carrying a loss-of-function mutation in sopE2 exhibited a small defect in invasion (Bakshi et al., 2000). However, the combined effect on bacterial entry of mutations in sopE and sopE2 has not been investigated.

In this paper, we show that another type III-secreted protein, SopB, stimulates actin cytoskeleton rearrangements and bacterial internalization. SopB has hitherto been thought not to be implicated in the invasion process, which is not substantially affected by deletion of the sopB gene (Galyov et al., 1997). Instead, SopB has been implicated in the downstream events that lead to fluid accumulation and enteropathogenicity (Galyov et al., 1997; Norris et al., 1998). In the current study, we show the importance of SopB in the stimulation of Cdc42-dependent rearrangements of the actin cytoskeleton that are a prerequisite of cellular invasion. This new information has previously been masked by a partial functional redundancy among the type III-secreted family of proteins. In addition, we show that the recently noted inositol phosphatase activity of SopB (Norris et al., 1998) is essential for its ability to promote cytoskeletal reorganization. The importance of inositol metabolism to bacterial-induced cellular responses is underscored by our finding that inositol phosphatase activity is also activated independently by SopE. However, SopE has no inherent phosphatase activity and, instead, activates an endogenous inositol phosphate phosphatase. This discovery represents a remarkable new feature of the ability of Salmonella to modulate host cell signalling pathways.

Results

Multiple type III-secreted effector proteins are required for S. typhimurium entry into cultured cells

A major goal of this study was to investigate the role of different type III-secreted effector proteins in promoting S. typhimurium entry into host cells. SopB is a type III-secreted protein that has not previously been considered essential for the invasion process (Galyov et al., 1997). We examined the effect of introducing a null mutation in sopB on the ability of S. typhimurium to enter into cultured cells. Intestinal Henle-407 cells were infected with different S. typhimurium strains, and the internalization levels were measured at different times after infection. An S. typhimurium strain carrying a null mutation in sopB was impaired in its ability to gain access to host cells, although such a defect was small and was overcome after longer infection times (Fig. 1; data not shown). These results suggested that SopB was capable of stimulating bacterial internalization. However, we hypothesized that the presence of SopE in this bacterial strain was masking a potentially much greater contribution of SopB to the invasion phenotype. We therefore constructed an S. typhimurium strain defective for both the SopE and the SopB proteins and evaluated its ability to gain access into cultured intestinal epithelial cells. The sopEsopB double mutant strain was severely impeded in its ability to enter into host cells (Fig. 1). The introduction of a plasmid encoding sopB into this strain restored its ability to enter into cells almost to wild-type levels (Fig. 1). These results demonstrate that SopB is capable of triggering bacterial internalization and that SopB and SopE are functionally redundant in the mediation of bacterial entry into host cells.

Figure 1.

Effect of mutations in type III-secreted effector proteins on the ability of S. typhimurium to enter into cultured intestinal epithelial cells. Intestinal Henle-407 cells were infected with wild-type S. typhimurium or different isogenic mutant derivatives, and the levels of bacterial internalization were examined 60 min after infection as indicated in Experimental procedures. Values represent the percentage of cells that have been invaded (i.e. that contain five or more internalized bacteria) by the different Salmonella. The n-values represent the number of cells examined in a minimum of three independent experiments.

SopE2 is a SopE-homologous protein detected in the S. typhi genome project (The Sanger Center, http://www.sanger.ac.uk/Projects/S_typi) and recently reported to contribute to S. dublin entry into host cells (Bakshi et al., 2000). However, the observation that the sopEsopB double mutant strain was completely defective for entry (see above) suggested that, in S. typhimurium, the sopE2 gene was either absent or not delivered in sufficient amounts to stimulate cellular responses. To investigate whether sopE2 was present in the S. typhimurium strain used in these studies we carried out a polymerase chain reaction (PCR) with primers derived from the reported sequences (The Sanger Center, http://www.sanger.ac.uk/Projects/S_typi) (Bakshi et al., 2000). A product of the expected size was obtained from this reaction, and nucleotide sequence analysis determined that such a product corresponded to the sopE-homologous gene sopE2 (data not shown). The predicted sequence of SopE2 is 69% identical to SopE; therefore, it is likely that this protein may also act as an exchange factor for Cdc42 and Rac. We hypothesized that, in S. typhimurium, SopE2 may not be delivered in sufficient amounts to overcome the antagonistic effect of the GTPase-activating protein SptP (Fu and Galan, 1999). To investigate this possibility, we constructed a sopEsopBsptP triple mutant and examined its ability to enter into cultured cells. In contrast to the sopEsopB double mutant, the strain carrying an additional mutation in sptP regained its ability to trigger bacterial internalization (Fig. 1) particularly after longer (2 h) infection times (data not shown). We hypothesized that the ability of the sopEsopBsptP triple mutant to enter host cells resulted from the activity of SopE2. To test this hypothesis, we introduced into the S. typhimurium sopEsopBsptP triple mutant a loss-of-function mutation in sopE2, resulting in a strain deficient in the four potential effectors of actin cytoskeleton responses. The resulting sopEsopE2sopBsptP quadruple mutant strain was completely defective in its ability to gain access into host cells even after longer infection times (Fig. 1; data not shown). The invasion defect of the quadruple mutant was equivalent to that of a strain completely defective in type III secretion, suggesting that no other effector protein capable of stimulating these responses is encoded by this strain of S. typhimurium.

SopB mediates bacterial entry by stimulating membrane ruffling and actin cytoskeleton rearrangements

SopE and presumably SopE2 mediate bacterial entry by stimulating Cdc42- and Rac-dependent actin cytoskeleton rearrangements and membrane ruffling. However, it is not known whether SopB mediates entry by a similar process. We therefore investigated whether SopB-mediated uptake was also dependent on the stimulation of actin cytoskeleton rearrangements and membrane ruffling. Henle-407 intestinal epithelial cells were infected with one of the following strains of S. typhimurium: wild type; the isogenic sopE, sopE2 or sopB single mutants; the sopEsopB or the sopEsopE2 double mutants; the sopEsopBsptP triple mutant; the sopEsopE2sopBsptP quadruple mutant; or the sopEsopB double mutant carrying a plasmid encoding sopB. The infected cells were stained with rhodamine-labelled phalloidin to visualize the actin cytoskeleton. Strains carrying a loss-of-function mutation in sopE, sopE2 or sopB stimulated profuse actin cytoskeleton rearrangements and membrane ruffling (Fig. 1). Consistent with their severe defect in invasion, the sopEsopB double mutant and the sopEsopE2sopBsptP quadruple mutant strains of S. typhimurium did not induce membrane ruffling in infected cells (Fig. 2). In contrast, the sopEsopBsptP triple mutant, the sopEsopE2 double mutant and the sopEsopB double mutant strain carrying a plasmid encoding sopB induced marked actin cytoskeleton rearrangements and membrane ruffling that closely resembled the changes induced by wild-type S. typhimurium (Fig. 2). These results indicate that SopB can stimulate membrane ruffling and actin cytoskeleton rearrangements and further support the functional redundancy between SopB and the Rho GTPase activators SopE and SopE2.

Figure 2.

Effect of mutations in invasion-associated effector proteins on the ability of S. typhimurium to induce actin cytoskeleton rearrangements in cultured intestinal epithelial cells. Intestinal Henle-407 cells were infected with wild-type S. typhimurium or different mutant derivatives (as indicated), and the actin cytoskeleton was visualized by staining with rhodamine-labelled phalloidin (red) 30 min after infection (inset showing cells infected with the sopBsopEsptP triple mutant strain shows cells 60 min after infection; this particular strain exhibits delay stimulation of membrane ruffling; see text). Internalized bacteria appear green, whereas extracellular bacteria appear blue (see Experimental procedures). Arrows point at ruffling events induced by the different strains. Images represent projections of confocal optical z-section obtained on a Zeiss LSM-510 confocal microscope. Bar at the top left represents 20 µm. Three-dimensional rendering of these images can be seen at the following URL: http://info.med.yale.edu/micropath/galan/Pages/galan_movies.html.

SopB exerts its function within the host cell after its translocation via the S. typhimurium type III secretion system (Galyov et al., 1997). To investigate its potential effect on the actin cytoskeleton without the interference of other effectors delivered by S. typhimurium into host cells, COS-1 cells were transiently transfected with a vector that expresses epitope-tagged SopB. Although expression of SopB could not be detected by immunofluorescence microscopy using an antibody directed to the epitope tag, cells that had been transfected by the sopB-encoding expression vector exhibited characteristic actin cytoskeleton rearrangements resembling the membrane ruffles induced by S. typhimurium (Fig. 3). Up to 15% of cells transfected with the SopB construct exhibited these actin cytoskeleton rearrangements, whereas cells transfected with the vector control did not show any noticeable change in the organization of the actin cytoskeleton (Fig. 3). The actin cytoskeleton rearrangements induced by SopB expression appeared to be more localized than those induced by the transient expression of the Salmonella Rho GTPase exchange factor SopE, which induces changes distributed throughout the cell (Fig. 3) (Hardt et al., 1998a). In this sense, the actin cytoskeleton rearrangements induced by SopB resembled more closely the changes induced by S. typhimurium infections, as bacterial-induced membrane ruffles tend to be localized at the point of bacteria–host cell contact. Taken together, these results demonstrate that SopB is capable of inducing actin cytoskeleton rearrangements and membrane ruffling, implicating this protein as an effector of similar responses induced by S. typhimurium.

Figure 3.

Transient expression of SopB in cultured cells induces phosphatase-dependent actin cytoskeleton rearrangements. Cos-1 cells were transfected with plasmids encoding SopE, SopB and the catalytic mutant SopBC462SC and subsequently stained with rhodamine-labelled phalloidin to visualize the actin cytoskeleton. Arrows point at membrane ruffles and actin cytoskeleton rearrangements induced by the expression of SopE or SopB. The quantification of transfected cells exhibiting actin cytoskeleton rearrangements is shown in the bar graph. Numbers represent the percentage of total cells undergoing membrane ruffling. Control experiments indicated that the percentage of transfected cells in these experiments range between 30% and 40%. The n-value represents the number of cells counted in at least three independent experiments.

SopB and SopE act redundantly to dephosphorylate InsP5 to Ins(1,4,5,6)P4 in the host cell

SopB has previously been shown to have inositol phosphate phosphatase activity (Norris et al., 1998). The observation that SopB and SopE were functionally redundant in their ability to stimulate actin cytoskeleton rearrangements and membrane ruffling prompted us to investigate whether inositol phosphatase activity might interface with this process. In agreement with previous reports (Eckmann et al., 1997; Norris et al., 1998), we found that infection of intestinal epithelial cells by Salmonella resulted in the depletion of the host cell InsP5 reservoir and the accumulation of a specific InsP4 isomer, namely, Ins(1,4,5,6)P4 (Table 1). A substantial dephosphorylation of InsP5 to Ins(1,4,5,6)P4 was observed as early as 10 min after infection, at which point the levels of this InsP4 were increased ≈ ninefold (Table 1). Thus, SopB acts as a specific 3-phosphatase upon InsP5in vivo. At this early time point, the total accumulation of lower inositol phosphates (InsP1−4) was ≈ 37 800 c.p.m. well−1, yet only ≈ 1900 c.p.m. well−1 was lost from the InsP5 peak. The only other known pathway for generating inositol phosphates is by the activation of phospholipase C (PLC), which must therefore accompany Salmonella infection of Henle-407 cells. By 30 and 60 min after bacterial invasion, cellular InsP5 levels were reduced by 85–90%, and this was also accompanied by an up to 50% decrease in InsP6 levels. At these time points, the level of Ins(1,4,5,6)P4 had declined from its 10 min value, presumably because of its dephosphorylation to lower inositol phosphates. The dramatic nature of these particular metabolic changes is noteworthy. No other cellular stimulus is known to induce the degradation of cellular InsP5 and InsP6 so rapidly.

Table 1. Inositol polyphosphate levels in intestinal Henle-407 cells after S. typhimurium infection.

Infection
time
Inositol
phosphate
products
Strains
UninfectedWild type sopE sopB sopB sopE
  1. 3H-inositol-labelled Henle-407 cells were infected with wild-type S. typhimurium and different mutant derivatives for the indicated times, and the levels of the different inositol polyphosphates (c.p.m.) were determined by HPLC analysis. InsP4 represents the total of both Ins(1,3,4,5)P4 and Ins (1,3,4,6)P4. The data are presented as the mean of two independent measurements. Values were normalized based on the c.p.m. of each lipid sample and then multiplied by 105 (the variation of the total lipid count between samples was less than 10%). In all cases, the standard deviation was less than 10%. In parentheses is the ratio between the levels of the different inositol phosphates in infected cells and those in uninfected control cells. The values for uninfected cells were obtained at time zero. Equivalent results were obtained in at least three repetitions of this experiment.

10 minIns1P + Ins4P41356529 (1.58)5263 (1.27)5863 (1.42)4870 (1.18)
InsP217 36824 079 (1.39)20 948 (1.21)22 045 (1.27)18 751 (1.08)
Ins(1,3,4)P3283452 (1.59)444 (1.57)473 (1.67)365 (1.29)
Ins(1,4,5)P313102106 (1.61)1759 (1.34)2427 (1.85)1462 (1.12)
InsP4198271 (1.37)160 (0.81)274 (1.39)166 (0.84)
Ins(1,4,5,6)P488772 (8.74)212 (2.40)156 (1.77)74 (0.84)
Ins(1,3,4,5,6)P540473213 (0.79)3612 (0.89)3669 (0.91)3895 (0.96)
30 minInsP610 95111 410 (1.04)11 460 (1.05)10 852 (0.99)11 175 (1.02)
Ins1P + Ins4P41357135 (1.73)5440 (1.32)7143 (1.73)5343 (1.29)
InsP217 36814 123 (0.81)17 882 (1.03)18 028 (1.04)18 622 (1.07)
Ins(1,3,4)P3283314 (1.11)376 (1.33)449 (1.58)345 (1.22)
Ins(1,4,5)P31310417 (0.32)668 (0.51)1612 (1.23)1861 (1.42)
InsP4198181 (0.91)222 (1.12)432 (2.19)125 (0.63)
Ins(1,4,5,6)P488621 (7.03)1416 (16.06)369 (4.19)91 (1.03)
Ins(1,3,4,5,6)P54047617 (0.15)1347 (0.33)3183 (0.79)3951 (0.98)
InsP610 95110 392 (0.95)10 448 (0.95)10 080 (0.92)11 059 (1.01)
60 minIns1P + Ins4P41354214 (1.02)4138 (1.00)9447 (2.28)5767 (1.39)
InsP217 3682816 (0.16)10 008 (0.58)14 163 (0.82)20 809 (1.20)
Ins(1,3,4)P328373 (0.26)171 (0.60)348 (1.23)241 (0.85)
Ins(1,4,5)P3131091 (0.07)210 (0.16)810 (0.62)2784 (2.13)
InsP4198163 (0.82)217 (1.10)357 (1.81)226 (1.14)
Ins(1,4,5,6)P48844 (0.49)222 (2.52)279 (3.16)74 (0.83)
Ins(1,3,4,5,6)P54047577 (0.14)573 (0.14)2151 (0.53)3956 (0.98)
InsP6109517704 (0.70)8999 (0.82)10 432 (0.95)11 863 (1.08)

It was thought previously that SopB alone was responsible for the ability of Salmonella to increase inositol phosphate phosphatase activity in host cells (Norris et al., 1998). In contrast, we found that the ability of Salmonella to stimulate inositol phosphate metabolism was reduced but not eliminated by the deletion of the sopB gene (Table 1). Thus, SopB plays an important but not exclusive role in mobilizing the cellular pools of InsP5 and InsP6. As our bacterial internalization assays indicated that there is a functional overlap between SopB and SopE (see above), we investigated whether SopE might contribute to the stimulation of inositol phosphate turnover. SopE has no homology with any known inositol phosphate phosphatase, and recombinant SopE did not express any InsP5 phosphatase activity in vitro, even when incubated at ninefold higher concentrations than SopB (data not shown). Nevertheless, in vivo, a sopE-defective mutant strain of S. typhimurium was less active than the wild type at hydrolysing InsP5 and InsP6 and at generating lower inositol phosphates, including those such as Ins(1,4,5)P3 that arise by PLC activation (Table 1). These data indicate that SopE must activate both PLC as well as an endogenous InsP5/InsP6 phosphatase activity. A sopBsopE double mutant strain was unable to promote any significant changes in inositol phosphate turnover in host cells, indicating that these two proteins are responsible for most of the inositol phosphate fluxes resulting from bacterial infection (Table 1).

To investigate whether the Salmonella-associated inositol phosphatase activity related to the actin cytoskeletal rearrangements, we constructed an S. typhimurium strain that expresses a phosphatase-defective SopB in which a critical active-site cysteine residue was changed to a serine (SopBC460S) (Norris et al., 1998). An S. typhimurium sopEsopB double mutant strain expressing SopBC460S was defective in its ability to stimulate actin cytoskeleton rearrangements (data not shown). To examine the effect of expression of this mutant form on cultured cells, COS-1 cells were transfected with a plasmid expressing SopBC460S, and the actin cytoskeleton was visualized by phalloidin staining. Expression of the catalytically defective SopBC460S mutant did not result in any changes in the actin cytoskeleton (Fig. 3), indicating that the phosphatase activity is required for the SopB-mediated actin cytoskeleton reorganization.

The accumulation of Ins(1,4,5,6)P4 and consumption of Ins(1,3,4,5,6)P5 were the two major changes in inositol phosphate levels that were observed immediately after Salmonella infection. However, microinjection into Ref52, Cos-1 or Henle-407 cells of a solution of up to 30 µM Ins(1,4,5,6)P4 (which results in an intracellular concentration of ≈ 10 µM) did not promote detectable changes in the actin cytoskeleton (data not shown). We also tested whether a decrease in cellular InsP5 might facilitate bacterial entry by relieving a potential inhibitory effect of this inositol polyphosphate. However, microinjection of a solution of up to 100 µM Ins(1,3,4,5,6)P5 (which results in an intracellular concentration of ≈ 30 µM) did not cause any changes in the cytoskeleton by itself, nor did it impair the cytoskeletal rearrangements induced either by wild-type Salmonella or the sopE or sopB isogenic mutants (data not shown). These observations indicate that cytoskeletal reorganization induced by Salmonella does not depend upon the change in steady-state levels of InsP5 and Ins(1,4,5,6)P4per se. Instead, it may be the actual interconversion of these two inositol phosphates that is functionally important, possibly as a molecular switch.

SopB specifically dephosphorylates InsP5 to Ins(1,4,5,6)P4in vitro

The specific dephosphorylation of Ins(1,3,4,5,6)P5 to Ins(1,4,5,6)P4 that accompanies Salmonella infection (see above; Eckmann et al., 1997; Norris et al., 1998) is not consistent with the reported in vitro activity of SopB, which was shown to cleave non-specifically nearly every phosphate from InsP5, forming several InsP4 isomers (Norris et al., 1998). In an effort to explain the inconsistency between the in vitro and in vivo observations, we re-examined the specificity of SopB phosphatase activity in vitro using purified recombinant enzyme. We found that the kinetics of [3H]-InsP5 dephosphorylation were biphasic; a rapid burst of hydrolysis, completed within 1 min, was followed by a slower sustained phase (Fig. 4). The purified recombinant SopBC460S mutant enzyme lacked any measurable phosphatase activity (Fig. 4). Contrary to what was reported previously (Norris et al., 1998), we found that SopB specifically removed the 3-phosphate from InsP5; therefore, the only InsP4 formed was Ins(1,4,5,6)P4 (Fig. 4). Thus, our in vitro data on SopB specificity towards InsP5 are in complete agreement with the in vivo findings and indicate that this bacterial enzyme exhibits a much narrower specificity than previously reported.

Figure 4.

In vitro phosphatase activity of purified SopB. The main graph shows data obtained after the incubation of GST-SopB with ≈ 10 000 d.p.m. [3H]-Ins(1,3,4,5,6)P5 for either 0 (open circles) or 60 min (closed circles), as described in Experimental procedures. Samples were quenched, neutralized and chromatographed by HPLC as described in Experimental procedures. The InsP4 region of the chromatograph is demarcated by the horizontal bar. The arrows indicate the elution positions of 14C-labelled standards of the indicated InsP4 isomer [i.e. Ins(1,3,4,6)P4, Ins(1,3,4,5)P4 and Ins(1,4,5,6)P4]. The inset shows data from other experiments, in which only 3000 d.p.m. [3H]-Ins(1,3,4,5,6)P5 was used, and the rate of the reaction was analysed on gravity-fed anion-exchange columns (filled circles). The inset also shows data (open circle) obtained with the SopBC460S catalytically inactive mutant. These data are representative of three or more experiments.

SopB-mediated bacterial internalization is dependent on Cdc42 but not Rac-1

Previous studies from our laboratory demonstrated that the S. typhimurium-induced cellular responses are dependent on the small GTP-binding proteins Cdc42 and, to a lesser extent, Rac-1 (Chen et al., 1996a). Thus, expression of a dominant-negative mutant of Cdc42 (Cdc42N17) effectively blocked S. typhimurium-induced actin cytoskeleton rearrangements and entry into host cells (Chen et al., 1996a). In contrast, expression of a dominant-negative mutant of Rac-1 (Rac-1N17) only partially blocked bacterial internalization, arguing for a less important role for Rac-1 in S. typhimurium-induced cellular responses (Chen et al., 1996a). Given the redundancy of bacterial effectors capable of stimulating these responses, these results suggest that Cdc42 is required for the responses induced by all effectors (e.g. SopB and SopE), whereas Rac-1 is required for the responses induced by only a subset of them (e.g. SopE). We tested this hypothesis by examining the effect of expression of dominant-negative mutants of Cdc42 and Rac-1 on the ability of strains carrying null mutations in either sopB or sopE to enter into cultured cells. As shown in Fig. 5 and consistent with previous observations (Chen et al., 1996a), expression of Cdc42N17 effectively blocked the internalization of wild-type as well as the sopBand sopE mutant strains of S. typhimurium (Fig. 5). In contrast, expression of Rac-1N17 did not affect the internalization of the sopEsopB+S. typhimurium mutant strain, although it significantly impaired the ability of the sopBsopE+ strain to enter into cultured host cells (Fig. 5). These results indicate that SopB-mediated entry is dependent on Cdc42 but not on Rac-1. In contrast, SopE-mediated entry requires both Cdc42 and Rac-1. The latter results are consistent with our previous observation that SopE has exchange activity towards both Cdc42 and Rac-1 (Hardt et al., 1998a). Furthermore, these results are also consistent with our previous findings that SopE can bind the nucleotide-free or GDP-bound forms of both these GTPases; therefore, expression of either Cdc42N17 or Rac-1N17 would result in the titration of this bacterial effector protein (Hardt et al., 1998a).

Figure 5.

Effect of the expression of dominant-negative mutants of Cdc42 and Rac-1 on the internalization of S. typhimurium wild-type and mutant strains. Cos-1 cells were transfected with double cistronic vectors that direct the coupled expression of different forms of Cdc42 and Rac and green fluorescent protein (GFP) as indicated. Transfected cells were infected with the different S. typhimurium strains as indicated, and the levels of bacterial internalization were examined as indicated in Experimental procedures. Bacterial invasion is indicated as the percentage of transfected cells that exhibited five or more internalized bacteria. The n-values represent the number of cells examined in a minimum of three independent experiments.

SopB is required for S. typhimurium-induced Jnk activation

In addition to actin cytoskeleton rearrangements, S. typhimurium induces nuclear responses leading to the production of pro-inflammatory cytokines in intestinal epithelial cells (Chen et al., 1996a; Hobbie et al., 1997). The stimulation of these responses is the consequence of the activation of Cdc42 and its effector, the p21-activated kinase (PAK), by effector proteins delivered to the host cell via the type III secretion system (Chen et al., 1996a; 1999; Hobbie et al., 1997). Activation of PAK leads to the stimulation of the MAP kinase pathways, in particular Jnk and p38 (Hobbie et al., 1997). We therefore tested the requirement of SopB for the S. typhimurium stimulation of nuclear responses. Cultured intestinal epithelial cells were infected with different strains of S. typhimurium, and Jnk activation was measured by Western immunoblot using an antibody specific for the phosphorylated (active) form of Jnk. As shown previously (Hobbie et al., 1997; Chen et al., 1999), wild-type S. typhimurium induced a marked activation of Jnk 30 min after infection (Fig. 6). In contrast, strains carrying a mutation in either sopE or sopB were impaired in their ability to stimulate Jnk activation (Fig. 6). A strain carrying loss-of-function mutations in both sopE and sopB was completely impaired in its ability to stimulate Jnk (Fig. 6). These results indicate that SopB is required for S. typhimurium induction of nuclear responses and that the synergistic activity of both SopE and SopB is required for the full activation of Jnk.

Figure 6.

Effect of mutations in sopB and sopE on the ability of S. typhimurium to induce JNK activation in cultured intestinal epithelial cells. Henle-407 cells were infected with wild-type S. typhimurium or various mutant strains, and the levels of phosphorylated (active) JNK were determined by Western blot analysis with an antibody specific for the phosphorylated form of JNK. The blot was stripped and reprobed with an antibody against JNK. Samples from cells infected with wild type (A), the sopE (B), sopB (C) or sopE sopB (D) mutant derivatives or the sopE sopB mutant carrying sopB complementing plasmid (E).

Discussion

Salmonella spp. use a type III secretion system to stimulate cellular responses that lead to bacterial internalization and nuclear responses (Galán, 1999). The stimulation of these cellular responses is the consequence of the co-ordinated activity of several bacterial effector proteins delivered into the host cells by this specialized protein secretion system. In this study, we have shown that SopB, a type III-secreted protein that exhibits inositol polyphosphatase activity, also functions to mediate Jnk activation, actin cytoskeleton rearrangements and bacterial internalization. This observation is surprising, as this protein was previously thought not to be required for bacterial entry but, rather, for later events leading to the stimulation of chloride secretion and enteropathogenicity (Galyov et al., 1997; Norris et al., 1998). We have shown that the contribution of SopB to the invasion phenotype has not been fully appreciated in previous studies because its actions are masked by the redundant function of the Rho GTPase activators SopE and SopE2. Thus, at least three type III-secreted effector proteins, SopE, SopE2 and SopB, act in a functionally overlapping manner to stimulate actin cytoskeleton rearrangements. Although introduction of loss-of-function mutations in each of these genes individually did not result in a significant defect in bacterial invasion, a strain carrying mutations in all three genes was severely defective for the stimulation of actin cytoskeleton rearrangements and entry into host cells. These results also showed that these three effectors account for all the capacity of Salmonella to induce membrane ruffling.

We have also shown that the contribution of the SopE-homologous protein SopE2 to the invasion phenotype of S. typhimurium can only be revealed in the absence of the GTPase-activating protein SptP, indicating that the SopE2-mediated activation of Cdc42 and Rac is not robust enough to overcome the antagonist effect of SptP. The difference in activity between SopE and SopE2 may result from different amounts delivered into host cells, different timing of delivery or different intrinsic enzymatic activity. More experiments will be required to distinguish between these possibilities.

Despite their functional redundancy, the Salmonella type III-secreted effectors of actin cytoskeleton rearrangements exert their function by different mechanisms. SopE and presumably SopE2 stimulate Rho GTPase signalling by acting as exchange factors for both Cdc42 and Rac (Hardt et al., 1998a). We have shown here that the inositol polyphosphatase activity of SopB is essential for its ability to stimulate Cdc42-dependent actin cytoskeleton rearrangements. We have also shown that both SopB and SopE promote dephosphorylation of InsP5, albeit by different mechanisms: SopB has inherent inositol phosphate phosphatase activity, whereas SopE does not. Instead, we propose that SopE activates an endogenous cellular inositol phosphate phosphatase. A good candidate for such an enzyme is the multiple inositol polyphosphate phosphatase (MIPP), which is the only cellular enzyme so far shown to be able to catalyse the 3-phosphate-specific hydrolysis of InsP5 to Ins(1,4,5,6)P4 (Nogimori et al., 1991; Craxton et al., 1997). In addition, both SopB and SopE stimulate cellular responses that lead to PLC activation, as shown by the accumulation of lower inositol phosphates in excess of that which can be accounted for by the breakdown of higher inositol phosphates. It has been recently reported that GTP-loaded Cdc42 and Rac can directly activate PLCβ2 (Illenberger et al., 1998) and bind and presumably activate PLCγ (Hong-Geller and Cerione, 2000). Therefore, the activation of PLC by these Salmonella effectors is probably a direct consequence of their activity on these Rho GTPases. The observation that Salmonella has evolved at least two independent mechanisms to both activate Rho GTPases and promote inositol phosphate metabolism argues for a central importance of both events in Salmonella pathogenesis.

How does SopB promote actin cytoskeleton reorganization and bacterial entry? The cytoskeleton rearrangements that lead to membrane ruffling are regulated by multiple, complex interactions between actin-binding and -organizing proteins that include Cdc42, Rac, exchange factors, GAPs and other effector proteins (Hall, 1998; Ridley, 1999). The proper assembly and regulation of this multimeric signalling complex relies on the appropriate protein–protein and protein–lipid interactions, particularly those mediated by pleckstrin homology (PH) domains, which are present in all the cellular GEFs that activate the Rho family of proteins (Cerione and Zheng, 1996; Shaw, 1996). The high affinity of PH domains for inositol lipids is one of the factors that localizes the signalling complex to the plasma membrane (Kavran et al., 1998), so inositol phospholipid-metabolizing enzymes that associate with membrane ruffles can help to regulate this process (Honda et al., 1999; Mochizuki and Takenawa, 1999). Another important factor is the cellular levels of inositol phosphates that can actively compete with inositol lipids for binding to certain PH domains (Takeuchi et al., 1996; Kavran et al., 1998). Thus, the nature of the ligand that is associated with a PH domain – soluble inositol phosphate or membrane-bound lipid – can determine the protein's subcellular location (Lockyer et al., 1997). In addition, inositol-based molecules can directly regulate Cdc42 activity (Zheng et al., 1996). Furthermore, the precedent that at least one inositol phosphate, namely Ins(1,3,4,5)P4, can activate Ras GAP (Cullen et al., 1995) suggests the possibility that Rho GTPase signalling may also involve a SopB-generated inositol phosphate acting upstream of Cdc42. Therefore, membrane ruffling in the host cell could be affected in multiple ways by Salmonella-induced changes in the metabolism of both inositol phospholipids and inositol phosphates.

Our results indicate that Salmonella has evolved at least two alternative mechanisms to stimulate cellular responses through Rho GTPases. One mechanism involves the direct activation of Cdc42 and Rac by SopE, which acts as a potent exchange factor for these GTPases. The other involves the inositol polyphosphatase SopB, which stimulates Cdc42-dependent signalling by unknown mechanisms. In addition, our results show that Salmonella has evolved alternative mechanisms to promote inositol phosphate metabolism: one involving the direct enzymatic activity of one of its effector proteins, SopB, and the other the activation of an endogenous inositol phosphatase, presumably through the activation of Rho GTPases by the exchange factor SopE. These constitute a further demonstration of the remarkable ability of this bacterial pathogen to manipulate host cellular functions.

Experimental procedures

Bacterial strains, plasmids and growth conditions

All bacterial strains were grown under conditions that stimulate the expression of the invasion-associated type III secretion system (Chen et al., 1996b). The wild-type S. typhimurium strain SL1344 (Hoiseth and Stocker, 1981) and its sopE (Hardt et al., 1998a) and sptP (Kaniga et al., 1996) mutant derivatives have been described previously. A sopB mutant strain (SB923) was constructed by deleting an EcoRV–NdeI DNA fragment from the sopB gene that had been amplified by PCR from the chromosome of S. typhimurium strain SL1344. The mutated allele was introduced into the chromosome of wild-type S. typhimurium by allelic exchange as described previously (Kaniga et al., 1994). The sopE2 gene was amplified by PCR with primers complementary to sequences upstream and downstream of the coding sequence. A deletion mutation was constructed by recombinant PCR. A DNA fragment containing the deletion mutation in sopE2 was cloned into the R6K-derived suicide vector pSB890 (Kaniga et al., 1994) and introduced into wild-type S. typhimurium, yielding the sopE2 mutant strain SB1300. Strains carrying different combinations of mutant alleles of sopE, sopE2, sopB and sptP were constructed by conjugation- and P22-mediated allele replacement strategies according to standard procedures (Kaniga et al., 1994). The sopB-complemented strain was made by integrating a plasmid carrying the entire sopB gene into the chromosome of the sopB sopE double mutant strain. A catalytically defective mutant of SopB was constructed by changing the cysteine at position 462 to a serine using the QuikChange site-directed mutagenesis kit according to the manufacturer's instructions (Stratagene). The mutated gene was introduced into the Salmonella chromosome by homologous recombination as described previously (Kaniga et al., 1994). Plasmids carrying the full-length cDNA encoding human wild-type Cdc42Hs (Cdc42wt), dominant-negative mutant forms of Cdc42Hs (Cdc42HsN17) or Rac1 (Rac1N17) have been described previously (Chen et al., 1996a). The eukaryotic expression vector expressing sopE has also been described (Hardt et al., 1998a). Plasmids expressing sopB and its catalytic mutant form sopBC460S were constructed by cloning PCR fragments containing the entire sopB gene or sopBC460S into the eukaryotic expressing vector pSB965 (Chen et al., 1996a).

Transfections, immunofluorescence microscopy and invasion assays

Levels of Salmonella entry into Henle-407 and COS-1 cells were measured by differential immunofluorescence staining as described previously (Chen et al., 1996a). Transfection of COS-1 cells was carried out as described previously (Chen et al., 1996a). Filamentous actin was stained with rhodamine phalloidin (Molecular Probes). Samples were visualized under a Nikon Diaphot 300 fluorescence microscope or a Zeiss LSM510 confocal microscope.

Purification of recombinant proteins and microinjection of inositol polyphosphates

GST, GST-SopB and GST-SopE proteins were purified as described elsewhere (Guan and Dixon, 1991) using glutathione Sepharose 4B (Pharmacia). Inositol polyphosphates (Matreya) were microinjected into Ref52, Cos-1 or Henle-407 cells at the indicated concentration as described previously (Fu and Galán, 1998).

JNK activation assay

JNK activation after Salmonella infection was measured by standard Western blot using monospecific antibodies to the phosphorylated form of Jnk according to the manufacturer's instructions (New England Biolabs) and as described previously (Fu and Galan, 1999).

[3H]-inositol labelling and extraction of cellular inositol polyphosphates

Henle-407 monolayers in 24-well plates were incubated for 3 days with 35 µCi well−1[3H]-inositol in inositol-free 50% DMEM, 50% Ham's F-12 medium, supplemented with 5% dialysed newborn calf serum. Cultures were washed three times with prewarmed Hank's balanced saline solution (HBSS) and infected with Salmonella strains at a multiplicity of infection (MOI) of 30. At specified time points, infection was terminated by placing the tissue culture plates on ice and washing twice with ice-cold PBS. Cells were quenched and neutralized (Shears, 1997), and the extracted inositol phosphates were analysed by high-performance liquid chromatography (HPLC) using a Synchropak Q100 column as described previously (Saiardi et al. 2000).

In vitro assays of Ins(1,3,4,5,6)P5 dephosphorylation by SopB

Recombinant GST-SopB protein was incubated for the indicated times at 30°C at a final concentration of 0.1 mg ml−1 in 25 µl of buffer containing 50 mM HEPES (pH 7.2), 10 mM dithiothreitol (DTT), 0.25 mg ml−1 bovine serum albumin (BSA), 10 µM Ins(1,3,4,5,6)P5 (CellSignals) and 3500 d.p.m. [3H]-Ins(1,3,4,5,6)P5 (synthesized as described by Saiardi et al., 2000). Assays were quenched and neutralized, and then the degree of dephosphorylation was ascertained using gravity-fed anion-exchange columns as described previously (Shears, 1997). In some experiments, ≈ 10 000 d.p.m. [3H]-Ins(1,3,4,5,6)P5 was used, and the reaction products were then identified by HPLC using 14C-labelled InsP4 standards as described previously (Saiardi et al., 2000).

Acknowledgements

This work was supported by Public Health Service Grants AI30492 and GM52543 from the National Institutes of Health to J.E.G. L.D.H. was supported by fellowship DRG-1522 from the Cancer Research Fund of the Damon Runyon–Walter Winchell Foundation.

Ancillary