Salmonella and Enteropathogenic Escherichia coli Interactions with Host Cells: Signaling Pathways

DANIKA L. GOOSNEY,^1,2§ SONYA L. KUJAT CHOY,^1§ AND B. BRETT FINLAY^1*

[SECTION EDITORS: FERRIC FANG AND MARTIN KAGNOFF]

January 5, 2006

Department of Microbiology and Immunology, University of British Columbia, 6174 University Blvd., Vancouver, BC, Canada V6T 1Z3,¹ and Inimex Pharmaceuticals, 6660 Northwest Marine Drive, Vancouver, BC, Canada V6T 1X2²

*Corresponding author. Phone: (604) 822-2210, Fax: (604) 822-9830, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

§These authors contributed equally to the chapter.

INTRODUCTION

The interaction between pathogen and host involves a myriad of initiations and responses from both sides. Bacterial pathogens like enteropathogenic Escherichia coli (EPEC) and Salmonella enterica have numerous virulence factors that interact with and alter signaling components of the host cell to initiate responses that are beneficial to pathogen survival and persistence. The study of Salmonella and EPEC infection reveals intricate connections between host signal transduction, cytoskeletal architecture, membrane trafficking, and cytokine gene expression. The emerging picture includes elements of molecular mimicry by bacterial effectors and bacterial subversion of typical host events, with the result that EPEC is able to survive and persist in an extracellular milieu, while Salmonella establishes an intracellular niche and is able to spread systemically throughout the host. This review will focus on recent advances in our understanding of the signaling events stemming from the host-pathogen interactions specific to Salmonella and EPEC.

SALMONELLA AND HOST SIGNALING PATHWAYS

Salmonella is a facultative intracellular pathogen that co-opts host signaling pathways as part of its virulence strategy. Key Salmonella virulence determinants are the type III secretion systems (TTSS), which are specialized and dedicated protein secretion systems that translocate bacterial proteins, termed effectors, directly into the host cell. Although other bacterial constituents such as lipopolysaccharide (LPS) and porins are potent activators of host signaling cascades (33; for reviews, see reference 90), this chapter will focus on the contribution of the Salmonella TTSS and cognate effectors. The TTSS encoded in SPI-1 is important for invasion (see section 8.4 in EcoSal); the SPI-1-translocated effectors directly and indirectly modulate host cell actin dynamics, causing characteristic membrane ruffling at the sites of bacterial attachment and leading to engulfment of Salmonella through a process resembling macropinocytosis. This stage of infection is also associated with stimulation of host inflammatory responses, including interleukin production and release, as well as Salmonella-induced cell death.

Following Salmonella engulfment, the bacteria reside within a specialized endocytic compartment termed the Salmonella-containing vacuole (SCV).The trafficking of the SCV is distinct from the usual endosomal pathway, and its interaction/fusion with various lysosomal vesicles is controlled or avoided through the actions of translocated effectors of the SPI-2-encoded TTSS. Thus, the phenotype of SPI-2 TTSS mutant Salmonella is defective intracellular replication and survival and reduced ability to persist and proliferate within macrophages and cause systemic infection. A number of SPI-2-dependent phenomena have been observed in infected cells, including formation of Salmonella-induced filaments (SIFs) and bacterial filamentation (3, 77, 86). A SPI-2-dependent apoptotic effect has also been reported (69).

SPI-1 Effector Interactions with the Host Cytoskeleton

Salmonella binds to the surface of epithelial cells of the intestine, primarily interacting with the lumenal surface. Access to the basolateral surface of enterocytes may occur either after bacterial transport through M cells in Peyer’s patches (50) or following epithelial destruction. Entry into nonphagocytic cells occurs as an active invasion process whereby the pathogens co-opt the actin cytoskeleton and induce their own uptake. The formation of characteristic membrane ruffles and bacterial engulfment within the SCV are promoted by the SPI-1 TTSS-mediated delivery of a repertoire of translocated Salmonella effectors (see section 8.4 in EcoSal for a detailed discussion). The effectors SipC and SipA are actin-binding proteins (ABPs) that directly interact with the actin cytoskeleton and modulate the activity of eukaryotic factors such as ADF/cofilin and gelsolin (41, 62). Cytoskeletal rearrangements are also accomplished indirectly through the overlapping functions of the effectors SopB (also known as SigD), SopE, and SopE2 that activate Rho GTPases such as Cdc42 and Rac1 (38, 68, 87). Salmonella then actively down-modulates the activation of Rho GTPases via the effector SptP (32, 52).

SipC and SipA—Direct Interactions with the Actin Cytoskeleton and a Model for the Formation of Membrane Ruffles.

The actin cytoskeleton is modulated by eukaryotic actin-binding proteins (ABPs) that function in the various roles of binding to G-actin (e.g., profilin), bundling (e.g., plastin), severing (e.g., ADF/cofilin), or capping (e.g., gelsolin) of F-actin filaments. The balance between G-actin (globular, monomeric form) and F-actin (filamentous, multimeric form) is carefully controlled to maintain the integrity of the cytoskeleton despite the dynamics of constant polymerization and depolymerization. This allows the cell to respond immediately to environmental cues.

The SPI-1-translocated proteins SipC and SipA interact directly with the actin cytoskeleton and associated components, thus mimicking the function of eukaryotic ABPs during invasion. The study of SipC has been complicated by its essential role in the translocation of other effectors (15), such that sipC mutants are attenuated for invasion (51). In vitro and in cultured HeLa cells, purified SipC protein is able to direct the nucleation of F-actin filaments (41). The SipC protein has three domains: a central hydrophobic region of 80 amino acids is thought to anchor SipC within the host cell membrane, such that the N and C termini lie within the host cytoplasm. The N-terminal 120 amino acids form a domain that is important for F-actin pairing and bundling, while the 209 amino acids that form the C-terminal domain are sufficient to nucleate actin polymerization. During infection the N- and C-terminal domains are predicted to nucleate the polymerization of actin directly beneath the invading Salmonella, followed by filament bundling and condensation into actin cables. Thus SipC acts directly on actin, in striking contrast to the indirect effect of other bacterial surface proteins such as ActA from Listeria or IcsA from Shigella, which simply sequester eukaryotic factors like the Arp2/3 actin-nucleating complex.

In contrast to the attenuated invasion phenotype of sipC mutants, mutants lacking SipA are able to nucleate actin but yield more diffuse actin condensation (103) and smoother membrane ruffles than wild-type Salmonella (44). Biochemical analysis indicates that purified SipA alone is able to bind to F-actin and can stimulate spontaneous assembly of filaments by reducing the critical concentration of G-actin (103). More importantly, SipA in combination with SipC, significantly enhances actin condensation and filament extension, both in vitro as well as in cell culture (62). Thus, SipA potentiates the actin-modulating abilities of SipC, and these two proteins "collaborate to nucleate actin polymerization efficiently" (62). More recently, McGhie and colleagues (61) have used a reconstituted bacterial actin-based motility assay to assess the effect of SipA on actin dynamics and more closely model the in vivo role of this effector. This ingenious methodology revealed that through binding to F-actin, SipA is able to inhibit actin depolymerization at the distal end of elongating actin filaments while exerting little effect on polymerization. Furthermore, the association of SipA with F-actin protects filaments from normal ADF/cofilin-mediated disassembly and from the actin-severing activity of gelsolin. (61). Commenting on these in situ effects of SipA, Le Clainche and Drubin (54) observed a paradox, namely, that SipA is able to arrest the dynamic unidirectional flow of actin subunits within filaments, yet it also contributes to SipC-dependent actin nucleation. This could be reconciled by partitioning or restricting these two incompatible activities to different areas of the host cell. If the actin nucleation effects of SipA were colocalized with SipC at the plasma membrane, where the Arp2/3 complex becomes activated, this area would contain the newly assembled, ATP-bound actin filaments. The "older" filaments, which are bound with ADP, are the preferential substrate for ADF/cofilin (89), so the ADF/cofilin-displacing activity of SipA could occur in the region of the cell with the older filaments. This model also offers an explanation of the membrane ruffling that is characteristic of Salmonella invasion. If SipA were localized at concentrations high enough to outcompete ADF/cofilin and gelsolin, thereby inhibiting "treadmilling" in a defined area, this might funnel the treadmilling factors to adjacent areas, which in turn could presumably protrude faster. "It could be, for example, that the area of the cell surface subjacent to a bacterium would have SipA-coated actin filaments that are not dynamic, while the area surrounding the bacterium would dynamically extend protrusions, facilitating internalization" (54). It remains to be shown whether delivery of the effectors occur with the correct timing, in sufficient concentration, and at the appropriate localization to fulfill these predictions.

SopE and SopE2 as Nucleotide Exchange Factors for Rho Family GTPases.

In addition to ABPs, the actin cytoskeleton is indirectly modulated by signal transduction pathways. Rho GTPases, polyphosphoinositides, and various adaptor proteins regulate the activity of WASP (Wiskott-Aldrich syndrome protein) family proteins, which in turn bind and stimulate the actin nucleation activity of the Arp2/3 complex. WASP proteins are intrinsically autoinhibited by intramolecular protein interactions that can be interrupted through alternative binding of the WASP GTPase-binding domain (GBD) to GTP-Cdc42 (PIP2). This enables the WASP family proteins to bring together an actin monomer and the Arp2/3 complex, which otherwise has a low intrinsic actin nucleation activity. WASP activity is also modulated as a result of interactions with the src-homology-3 (SH3) domains of adaptor proteins such as Nck and Grb2 (10, 76; for a recent review, see reference 2). The SH2 domains of these adaptor proteins simultaneously bind to tyrosine-phosphorylated sites on yet other proteins, thus creating local concentrations of signaling proteins that can influence the activity of the Arp2/3 complex. In response to Salmonella infection, the downstream signaling effects of Rho GTPases may also include transient phosphorylation of the ADF/cofilin complex (22).

The Rho family of proteins comprises small Ras-related GTP-binding proteins that cycle between active (GTP-bound) and inactive (GDP-bound) states. Since the spontaneous dissociation of the guanine nucleotide is slow, a family of eukaryotic guanine nucleotide exchange factors (GEFs) act to increase the rate of cycling between active and inactive states. This normal host function has been co-opted by Salmonella through the SPI-1-translocated effectors SopE and SopE2. SopE and SopE2 are 70% identical at the level of amino acid sequence (39) and apparently bind to the same regions of Rho GTPases as the eukaryotic GEFs despite lacking the typical eukaryotic GTPase-binding motif (5). Crystallization studies, using a C-terminally truncated Cdc42 and the catalytic domain of SopE, suggest that SopE binds Cdc42 in the switch I and II regions, locking Cdc42 into a conformation that promotes the release of the guanine nucleotide (5). Cdc42 is the sole substrate for SopE2 (31, 87). In contrast, SopE is able to activate several different members of the Rho family in vitro, including Rac1, Cdc42, RhoA, and RhoG (31, 38). It is not clear whether this differential substrate specificity provides a molecular basis for differential host specificity and virulence (72, 73, 100). Since SopE is encoded in the bacteriophage SopEΦ and is not present in all Salmonella strains (39), the overlapping function of SopE2 has been postulated to ensure activation of Cdc42 and promote efficient bacterial internalization (70).

In nonpolarized cells, Salmonella invasion is modulated by Cdc42 and characterized by Arp2/3 recruitment to membrane ruffles (14), Salmonella invasion of polarized epithelial cells similarly requires de novo actin polymerization and correlates with activation and recruitment of the Arp2/3 complex to membrane ruffles, but recent evidence indicates that different host signaling pathways are involved. While SopE-dependent Rac1 activation is important for invasion at the apical surface, Cdc42 is apparently not involved (19). At the basolateral surface, bacterial entry is SopE dependent and does not require activation of any of the four putative SopE substrates, although it results in activation of both Cdc42 and Rac1. Instead, it has been hypothesized that actin polymerization is stimulated through the activity of other toxin B-sensitive Rho family GTPase(s) in a SopE-dependent manner (19, 20). A downstream effect of these (as yet unidentified) Rho GTPases is recruitment of the Arp2/3 complex to membrane ruffles and its activation via Scar1, a WASP family protein (20).

Modulation of Cellular Phosphoinositides through the Activity of SopB/SigD.

Together with the direct activation of Rho GTPases by SopE and SopE2, SopB indirectly modulates these signaling pathways via changes in cellular phosphoinositides, which are intracellular signaling molecules (for a more comprehensive recent review, see reference 4 and section 8.4 in EcoSal). Thus, SopB, SopE, and SopE2 act synergistically to promote Salmonella invasion. SopB possesses inositol polyphosphate phosphatase activity (68) and is membrane associated (60), facilitating its localization in proximity to phosphoinositide substrates. In vitro SopB acts on phosphoinositol 3,4,5-triphosphate and inositol 1,3,4,5,6-pentakisphosphate; since hydrolysis of these phospholipids would promote an increase in calcium-dependent chloride secretion, this supports a role for SopB in the development of diarrhea (68). Further support comes from recent data indicating that Salmonella counters the epidermal growth factor (EGF)-mediated inhibition of chloride ion secretion that normally proceeds via phosphatidylinositol 3-kinase (PI3-K), and this antagonism requires SopB (1). Overexpression of SopB in tissue culture cells can cause a depletion of inositol hexakisphosphate and inositol 1,3,4,5,6-pentakisphosphate (28); hydrolysis of inositol 1,3,4,5,6-pentakisphosphate yields inositol 1,4,5,6-tetrakisphosphate, which activates the Rho GTPase Cdc42 (102). The disappearance of phosphoinositide 4,5-diphosphate from the invaginating regions of Salmonella-induced membrane ruffles also depends on SopB (92) and is important for membrane fission to complete bacterial internalization and formation of the SCV. An accessory effect of SopB on antiapoptotic signaling cascades is suggested by the activation of Akt, a protein kinase that has been implicated in cell survival, which may allow greater intracellular replication of Salmonella within epithelial cells (85).

The Dual Functions of SptP.

The reorganization of the actin cytoskeleton initiated via SopB, SopE, and SopE2 is reversed shortly after bacterial internalization, allowing infected cells to regain their normal architecture. The bacteria themselves participate in this process, through the activity of the SPI-1-translocated effector SptP. SptP is a modular protein with two domains that possess distinct independent biochemical activities: the N-terminal domain has activity similar to eukaryotic GTPase-activating proteins (GAPs) (32), whereas the C terminus possesses tyrosine phosphatase activity (52). The C terminus is apparently not important for the reversal of cytoskeletal rearrangements and will be discussed in more detail below. SptP contains the same arginine finger catalytic motif as eukaryotic GAPs and causes a reduction in Rac and Cdc42 activation by increasing the rate of GTP hydrolysis (32). Recent work has shown that the timing of the reversal is not a function of sequential delivery of bacterial effectors, because SptP and SopE are delivered in similar amounts early during infection. Instead, the overall effect on cellular signaling leading to actin cytoskeleton rearrangements is significantly influenced by differences in effector protein stability, mediated by their TTSS translocation domains (53). Proteosome-dependent degradation of SopE occurs more rapidly than the degradation of SptP, so that the actin rearrangements initiated by SopE are reversed by the longer lasting effect of SptP (53). Subtle differences in subcellular localization of these effectors within the host cell (as suggested above for SipA) might also enable SopE to exert its effects on signaling molecules even in the presence of SptP.

In addition to GAP activity, SptP is involved in down-regulating mitogen-activated protein (MAP) kinase signaling pathways through its tyrosine phosphatase activity (66). During infection, Salmonella actively down-modulates these pathways as evidenced by a return to basal levels of Jnk activation (32) and a decrease in activation of extracellular regulated kinase (ERK) (66), both of which depend on SptP. Salmonella-induced activation of ERK appears to proceed from Cdc42 via the ACK-1 and ACK-2 tyrosine kinases. In the typical pathway leading to ERK activation, Ras is activated by a number of converging signals and, in turn, activates Raf-1, whose activity depends on serine and tyrosine phosphorylation and localization to the plasma membrane. Raf then phosphorylates the MEK1 and MEK2 kinases, which, in turn, act on ERK. By cotransfecting SptP with constitutively active forms of various components of the MAP kinase pathway, Lin and colleagues (57) demonstrated that SptP interferes with activation of Raf-1 by Ras (see Fig. 1). In particular, the SptP tyrosine phosphatase domain interferes with localization of Raf to the plasma membrane, while the GAP domain interferes with the subsequent phosphorylation of Raf. Thus, although the N- and C-terminal domains have distinct activities, both actions of SptP are important for restoration of host processes following the initial response to Salmonella infection.

Fig. 1Proposed model of SptP inteference with Ras-mediated activation of ERK. (Step 1) Ras induces membrane translocation of Raf-1. SptP tyrosine phosphatase activity interferes with this step. (Step 2) After being activated by CDC42/Rac, Pak1 and 3 phosphorylate Raf at Ser338. SptP inhibits this step through its inhibitory GAP activity at Cdc42/Rac. (Step 3) Stc phosphorylates Raf at tyrosines 340/341, resulting in complete activation of Raf. Activated Raf then phosphorylates MEK1/MEK2, which phosphorylate and activate Erk1/Erk2. (Reprinted from reference 64 with permission from Blackwell Publishing Limited, Oxford, England.)

Salmonella -Induced Inflammatory Responses in Intestinal Epithelial Cells

As described above, the activation of Rho GTPases leads to cytoskeletal rearrangements associated with membrane ruffling and internalization of Salmonella. Concurrent with these physical changes in the cytoskeleton are signaling events that culminate in host inflammatory responses in the intestinal epithelium. Infection of epithelial cells with Salmonella induces activation of the MAP kinases Erk, Jnk, and p38 (45). These signaling pathways involve activation of the transcription factor NF-κB as well as increases in the level of the c-Jun subunit of the activator protein-1 (AP-1) complex (45), leading to the production of interleukin 8 (IL-8), which is often used as a read-out for the inflammatory response. Mek-1 and Mek-2 (the activator kinases of Erk-1 and -2) are also implicated since Mek-specific inhibitors cause a significant decrease in Salmonella-induced IL-8 production (66). Chen and colleagues (13) have determined that p21-activated kinase (PAK) is activated downstream of Cdc42 and mediates the SopE-dependent activation of Jnk. Since Cdc42 activation mediates both actin rearrangements and the inflammatory response, this work demonstrates that these two effects are distinct, and the PAK-mediated signal transduction pathways specifically lead to nuclear responses without altering actin dynamics.

Salmonella actively down-modulates the initial host inflammatory response as it choreographs changes in the actin cytoskeleton. This includes down-regulation of IL-8 production, mediated by SptP and the SPI-1 TTSS effector SspH1 (37). SspH1 localizes to the mammalian cell nucleus where it inhibits NF-κB-dependent gene expression (37). Furthermore, SptP appears to down-regulate secretion of the proinflammatory cytokine tumor necrosis factor alpha (TNFα) by macrophages (57).

Inflammatory Responses: the SPI-2 Effector SpiC and IL-10 Release

The SPI-2-encoded TTSS, in particular, the effector SpiC, appears to be involved in a signal transduction pathway that culminates in cytokine expression in infected macrophages (95). Compared with infection with wild-type Salmonella, a spiC mutant specifically induces lower expression of the IL-10 mRNA and production of IL-10; this effect is not seen for the proinflammatory cytokines IL-α, IL-6, or TNFα. The signaling pathway leading to production of IL-10 includes protein kinase A (PKA) but not Erk, protein kinase C (PKC), or Jnk. PKA activity decreases to uninfected levels in macrophages infected with spiC mutant bacteria, implying that activation of PKA requires a functional SpiC protein. There is also circumstantial evidence that camp-responsive element-binding protein (CREB) acts as an intermediate between PKA and SpiC-dependent IL-10 production. Activated PKA can phosphorylate CREB, regulating its ability to activate transcription of genes such as IL-10, whose promoter contains a binding site for CREB. In support of this hypothesis, serine phosphorylation of CREB was lower following infection with spiC mutant compared with wild-type Salmonella. Finally, in activated macrophages, PKA-dependent IL-10 production may inhibit cytotoxic mechanisms and increase survival of Salmonella (95). However, an important caveat is that the localization of SpiC remains controversial. There is evidence that SpiC is translocated into the host cytosol (81, 94), where it may interact with the host proteins Hook3 (81) and TassC (55) to play a role in the selective trafficking of the SCV. On the other hand, if SpiC is a structural component of the TTSS as has been proposed by others, the phenotype of a spiC mutant may be attributable to deficient secretion of the translocon component SseB and consequently global failure to translocate other SPI-2 effectors (30, 98).

Salmonella and Host Signal Cascades following Infection of Macrophages

The signal transduction pathways induced in macrophages differ significantly from those in epithelial cells following Salmonella infection. The activation of these pathways is independent of SPI-1 and not tied to activation of Rho GTPases and actin rearrangements (described above). Activation of MEK and ERK occurs via the intermediates PI3-K, phospholipase D (PLD), and PKC. The early induction of these pathways is similar following invasion with Salmonella or treatment with LPS, suggesting that the latter may be the primary activating signal (74). Furthermore, Jnk and SEK1 (stress and extracellular activated kinase-1) activation occur independently of each other—the former is mediated through PI3-K, tyrosine kinases, and PKC, while the latter occurs via phosphatidylcholine-phospholipase C and acidic sphingomyelinase, which generate the lipid second messenger ceramide. The substrates of these two kinases remain unknown (75). Salmonella infection of macrophages also correlates with an increase in activation of Raf1. Independently, there is an increase in proteosome-mediated degradation of Raf1, which occurs as a result of induction of apoptotic signaling pathways and depends on SipB and caspase-1 (discussed below). Analysis of a Raf1 mutant suggests that it causes a decrease in macrophage cytotoxicity in response to Salmonella by antagonizing the caspase-1 pathway, while the MEK/ERK and NF-κB antiapoptotic pathways are unaffected (48).

There have been few studies of macrophage signaling pathways at later time points of Salmonella infection. Recent studies have linked MAP kinases to Salmonella replication within the SCV by assessing Salmonella filamentation at 24 h postinfection of macrophages. Filamentation is a bacterial phenotype that corresponds with decreased replication within the SCV and stress-induced failure to complete cell division (77). Salmonella enterica serovar Typhimurium infection of interferon gamma (IFN-γ)-primed RAW264.7 cells results in increased activity of MEK kinase, with a corresponding increase in phosphorylation of the MEK substrate ERK, which would be expected to yield increased ERK translocation into the nucleus and subsequent alterations in gene expression. The decrease in bacterial filamentation obtained by inhibition of MEK was also seen following treatment with inhibitors of the phagocyte NADPH oxidase, even when the inhibitors were added 4 to 8 h postinfection, indicating that these signaling pathways exert their effects following bacterial invasion. The MEK and phagocyte NADPH oxidase (PhoX) pathways are operating in parallel, rather than activating each other, to mediate a bacteriostatic effect on Salmonella (77). This study is unique because it emphasizes the reciprocal effect of host signaling pathways on Salmonella.

Signaling Cascades Leading to Apoptosis

It is apparent that host signaling molecules such as the Rho GTPases, Jnk, p38, PKB/Akt, and Raf-1 are induced by and important for Salmonella infection in both epithelial and phagocytic cells. These molecules are also key regulators of apoptosis. Following infection, Salmonella co-opts various host signaling pathways to promote their own intramacrophage survival but also initiates signaling cascades leading to apoptosis, which may then facilitate their release and spread. There is evidence for at least two distinct pathways by which Salmonella causes apoptosis (for a recent review, see reference 47 and Chapter Host Cell Death in EcoSal). A rapid, SPI-1-dependent mechanism involves activation of caspases-1 and -2 (43). In addition, there is a delayed cytotoxic effect requiring plasmid virulence gene SpvB, transcriptional regulator OmpR, and the SPI-2 TTSS, but not SPI-1 (78). As with the carefully orchestrated actin rearrangements and alterations in cytokine production discussed above, it appears that Salmonella modulates the balance between pro- and antiapoptotic signaling pathways to actively determine the fate of infected phagocytic cells.

The initial interaction between Salmonella and the macrophage appears to generate a signal essential for apoptosis, because active invasion selectively activates proapoptotic pathways leading to cell death (29). Recent work using human monocytic cells confirms that active bacterial invasion induces apoptosis, mediated by activation of Cdc42 and Rac1, and independent of tyrosine kinases and PI3-K. In contrast, bacterial internalization via receptor-mediated phagocytosis does not induce apoptosis, despite the activation of Cdc42 and Rac1. The likely explanation for this difference is that during phagocytosis, tyrosine kinases and PI3-K are required for induction of Cdc42 and Rac, and these kinases lead to coactivation of Akt-dependent survival pathways. Since the level of serovar Typhimurium-induced cytotoxicity is lower in human cell lines than in murine cell lines, it is tempting to speculate about a link between cytotoxicity and the ability of particular S. enterica serovars to spread systemically within specific hosts (29).

One of the key Salmonella proteins required for rapid induction of cytotoxicity is SipB, translocated into the host via the SPI-1 TTSS. SipB binds and activates caspase-1, leading to apoptosis and the release of proinflammatory cytokines, and this interaction with caspase-1 is important for colonization of mice (43, 64). Ectopic expression of SipB in cultured cells induces formation of autophagosome-like membranous structures containing mitochondrial markers; during infection, translocated SipB localizes to and disrupts mitochondria and has been suggested to play a role in the simultaneous accumulation of autophagosomes (42). In response to invasion by Salmonella at a high multiplicity of infection, the activation of caspase-1 is aided by the simultaneous independent activation of caspase-2 signaling pathways leading to apoptosis (49). The activation of caspase-3, -6, and -8 is also independent of caspase-1 induction (49). Most of this research has examined serovar Typhimurium infection of murine macrophages and macrophage cell lines, and similar results have been obtained using bovine monocyte-derived macrophages (78), suggesting that caspase activation is a common response to Salmonella infection rather than a result of host-specific interactions.

The use of dendritic cells is gaining popularity, since these phagocytic cells are likely to be the first immune cells to encounter Salmonella in the intestinal mucosa. Similar to other cell types, infection of dendritic cells with serovar Typhimurium is an active invasion process, requiring actin polymerization in the host and a functional SPI-1 TTSS in the bacteria. As in macrophages, activation of caspase-1 is observed; treatment with caspase-1 inhibitors or use of a sipB mutant reduced Salmonella cytotoxicity for dendritic cells (96). Dreher and colleagues (26) have shown that caspase-1-dependent activation and release of IL-18 by human dendritic cells also depends on sipB. Further discussion of host signaling processes relative to innate immunity is beyond the scope of this chapter.

The SPI-2 Translocated Effector SifA, Host Signaling, and Membrane Trafficking

Rabs are a large and diverse family of Ras-like small G proteins that cycle between a GTP-bound active state and a GDP-bound inactive state, modulated by GAPs and GEFs as described above. Rabs are important for the signaling events that occur during organelle biogenesis and regulate multiple steps in vesicular membrane trafficking. They exert their effects through direct interaction with and indirect regulation of G-protein-coupled receptors, which initiate signal transduction cascades and the production of intracellular second messengers (reviewed in reference 80). Rab7 normally controls the fusion of lysosomes with late endosomes, through an interaction with RILP (Rab7-interacting lysosomal protein). In addition to the domain that binds Rab7-GTP, RILP interacts with the dynein/dynactin complex, thus bringing together Rab7-positive endosomal vesicles and the microtubule motor complex. Subsequent movement of endosomes toward the microtubule-organizing center (MTOC) may promote their fusion with lysosomes, which are concentrated at the MTOC. This has implications for Salmonella because the SCV is a specialized endosome that has altered trafficking so as to avoid degradation and bacterial killing. Recently, Harrison and colleagues (40) provided evidence that interaction between Rab7 and the SPI-2 effector SifA can interfere with the usual association of Rab7 and RILP, thereby uncoupling the SCV and Sifs from centripetal cell motors. Their work also supports a hypothesis that Sifs extend away from the MTOC as a result of being uncoupled from dynein, perhaps extending toward the periphery of the cell through the action of kinesins (centrifugal motors) (40).

SipA: Can Signaling Pathways Be Influenced by a TTSS Effector at the External Surface of the Host Cell?

TTSS effectors are conventionally believed to exert their effects on host cellular processes following translocation, that is, internal to the cell rather than at the external surface. However, recent work on SipA is challenging this idea with evidence obtained by using a model-polarized epithelial monolayer (82). In this model system, apical addition of Salmonella induces polymorphonuclear leukocyte (PMN) transmigration across the monolayer. The intervening signal transduction pathway depends on the activity of ARNO, a GEF, to activate ADP-ribosylation factor 6 (ARF6), which cycles between GTP- and GDP-bound states (21). ARNO and ARF6 show increased recruitment to the apical surface, specifically to the sites of bacterium-host contact in membrane ruffles (19). ARF6 in turn regulates PKC activation and relocation to the apical plasma membrane, apparently through the effector PLD. Finally, there is an increase in pathogen-elicited epithelial chemoattractant (PEEC) release at the apical side of the polarized cells, which induces PMN transmigration (21).

While this signal transduction pathway was dissected following apical addition of whole Salmonella, activation was independent of bacterial invasion and could be mimicked by the addition of purified SipA to the monolayer. SipA alone can induce PMN transmigration (56), as well as the subcellular redistribution of PKC from cytosol into the apical pole, apical membrane, and subapical cytoplasmic domain of host epithelial cells (82). In biochemical assays, the addition of SipA induced PKC phosphorylation; similarly, treatment of an epithelial monolayer with SipA induced activation of three specific isoforms of PKC—namely the δ, ε, and α isoforms—although only the α isoform was involved during SipA-dependent PMN transmigration (82). In summary, SipA apparently activates the signal transduction pathway that includes ARF-6 and PLD, leading to activation of PKC and resulting in release of PEEC at the apical surface of a polarized epithelium (Fig. 2). Ultimately, this induces PMN transmigration through the epithelium, resulting in loss of intestinal epithelial barrier function. Whether SipA acts extracellularly during natural infection remains to be established. A mutation in sipB, which abolishes effector protein translocation into host cells without interfering with secretion (15), dramatically reduces the severity of experimental Salmonella enteritis (101), suggesting that this may not be the case.

Fig. 2Model of serovar Typhimurium-induced signaling in epithelial cells leading to PMN transmigration. Interaction of serovar Typhimurium and its type III secreted effector protein SipA with the apical domain of polarized epithelial cells leads to activation of ARF6 at the apical membrane, presumably through the mammalian GEF ARNO. This leads to an increase in PLD activity and local production of phosphatidic acid (PA), which is speculated to be metabolized to diacylglycerol (DAG) by PA phosphohydrolase (PAP). Generation of DAG recruits PKC to the apical membrane. Activation of PKC at this site (PKC*) is necessary for the apical release of the chemokine PEEC and subsequent basolateral-to-apical PMN transmigration. (Reprinted from reference 19 with permission from the Journal of Biological Chemistry.)

It is clear from the foregoing discussion that, although our understanding of the early signaling events following Salmonella infection is becoming more substantial, many questions remain with regard to the signal transduction pathways important at later time points. These deserve attention since they presumably influence the outcome of infection: productive infection in which the host eventually succumbs, bacterial clearance leading to host recovery, or establishment of a chronic disease state.

EPEC EFFECTS ON THE HOST CELL

EPEC is an extracellular gram-negative bacterial pathogen that colonizes and infects human intestinal epithelium. It is a leading cause of infantile diarrhea in developing countries, with occasional outbreaks in developed countries as well. EPEC forms a characteristic lesion, termed the attaching and effacing (A/E) lesion, which is typified by a loss of microvillar structure at the site of bacterial attachment and the formation of a long, actin-rich stalk directly beneath the pathogen, called the pedestal. EPEC uses a TTSS to deliver proteins into the host cell, which trigger events ranging from immediate plasma membrane and cytoskeletal rearrangements at the apical surface to tight junction disruption, transduction of signals to the nucleus, and ion imbalances leading to diarrhea (Fig. 3).

Fig. 3EPEC effects on the host cell. EPEC attaches to the intestinal epithelial cell and remains extracellular, where it mediates numerous effects on the host. Signaling events include lipid raft recruitment upon initial bacterial attachment, rearrangement of the host actin cytoskeleton resulting in the formation of the pedestal, alteration in the ionic balance of the intestinal epithelium, disruption of tight junction integrity, and signaling to the nucleus via NF-κB to induce IL-8-mediated immune responses, including PMN recruitment and transmigration.

Pedestal Formation

The EPEC A/E lesion, or pedestal, is perhaps the best known and most characterized cellular event in EPEC infection. The secretion of EPEC virulence proteins leads to the loss of microvillar structure and the reorganization of underlying actin cytoskeleton into the pedestal. These structures can extend up to 10 μm in length beneath the pathogen (61). EPEC secretes several virulence factors into the host cell by using a TTSS. One of the key secreted bacterial factors is the translocated intimin receptor (Tir), which functions as a receptor, allowing the intimate attachment of EPEC to the host cell via the outer membrane ligand intimin. Binding of Tir to intimin clusters Tir beneath EPEC (93), analogous to integrin clustering by extracellular matrix in focal adhesion formation. Pedestal formation can commence only after Tir has been clustered by intimin. Tir that has been delivered to the host cell but has not clustered (due to the use of an intimin mutant) still recruits cytoskeletal proteins, but the recruitment is fuzzy and unfocused, appearing as a diffuse cloud surrounding adherent bacteria.

Tir has two transmembrane domains that allow both the amino (N) (amino acids 1 to 234) and carboxyl (C) (amino acids 392 to 550) termini to be oriented within the host cell cytosol and create a loop region between the two transmembrane domains (amino acids 272 to 362) that binds intimin (23). The C terminus of EPEC Tir contains a tyrosine residue (Y474) that is phosphorylated upon transfer to the host cell and is absolutely critical for actin polymerization to occur beneath the bacterium. It has recently been demonstrated that c-fyn, a member of the src family kinases, is responsible for direct Tir phosphorylation on Y474 (71). Two additional tyrosine kinases, Arg and Abl, have recently been shown to be sufficient for pedestal formation in the absence of other kinase activity but are not necessary (91).

Despite the importance of Tir tyrosine-474 phosphorylation, this modification alone does not account for the shift in molecular weight that Tir undergoes upon insertion into the host cell membrane. The molecular shift is due, in part, to phosphorylation of Tir serine residues S434 and S463 by PKA (97). It has been proposed that this serine modification may play two roles in EPEC infection: supplying the energy required to insert of Tir into the membrane and aiding in pedestal elongation.

Following Tir insertion, clustering by intimin, and phosphorylation, the adaptor protein Nck binds to the phosphotyrosine residue (6). Binding of Nck to Tir is both necessary and sufficient for pedestal formation. Indeed, only a small 12-residue peptide of Tir surrounding the phosphorylated tyrosine is required to induce full pedestal formation (7). Tir protein produced by a close relative of EPEC, enterohemorrhagic E. coli (EHEC), does not undergo tyrosine phosphorylation and in turn does not recruit Nck, suggesting it uses an entirely different mechanism to build its pedestal.

Once Nck is bound to phosphorylated Tir, the recruitment of other key cytoskeletal components occurs. Most notably, an actin-nucleation-promoting factor, N-WASP, is brought to the site through the actions of Nck, either directly or indirectly (36). Research by Kalman and coworkers (50a) had suggested that the GBD of Nck is both necessary and sufficient for recruitment to the site of EPEC attachment. This finding, however, was observed only for EHEC and not EPEC in a more recent study (58). Although the means by which the pedestal is targeted is not clear-cut, it is obvious that the acidic C terminus of WASP plays a critical role in pedestal formation by recruiting the Arp2/3 complex (58). The Arp2/3 complex is a group of proteins that initiate actin nucleation, actin pointed-end capping, and branching of actin filaments (65). Removal of the C terminus still allows WASP to be recruited beneath EPEC but prevents Arp2/3 recruitment, thereby preventing actin nucleation. The Arp2/3 complex localizes along the length of the pedestal, suggesting that it is incorporated into the structure as the pedestal elongates. It is at the point of WASP recruitment that the signaling events between EPEC and EHEC pedestal formation converge. Although Nck is required for EPEC but not EHEC pedestal formation, N-WASP is essential for both.

Actin polymerization leading to pedestal formation is initiated through a small domain on the C terminus, but additional cytoskeletal rearrangements occur via the N terminus of Tir. It has been demonstrated that the N terminus of Tir can directly interact with components of focal adhesions, specifically talin, vinculin, and α-actinin (9).

In addition to the actin-associated proteins discussed in detail above, numerous actin-associated proteins are also localized to the site of attachment that have not been as well characterized in pedestal formation (Table 1). These include cortactin, fimbrin, ezrin, villin, myosin light chain (MLC), CD44, calpactin, and VASP (8). Most of these proteins are recruited in a phosphoTir-dependent manner. Several, including α-actinin, talin, gelsolin, ezrin, and tropomyosin, are localized to the site of attachment in a Tir-dependent but phosphotyrosine-independent manner. Only two proteins, CD44 and calpactin, are recruited in an entirely Tir-independent manner (34). The differential regulation of cytoskeletal recruitment suggests that Tir may initiate several signaling pathways in the intestinal epithelial cell, with pedestal formation being only one outcome of this signaling.

Table 1Host proteins characterized in E. coli pedestals

Disruption of Tight Junctions

Tight junctions play two key roles in the intestinal epithelium: restricting movement of small particles and ions across the epithelium and separating apical and basolateral surfaces of the polarized cell (88). Infection of polarized intestinal cells by EPEC has been shown to disrupt both functions of tight junctions. This disruption in tight junction integrity occurs independently of A/E formation (63). Barrier function disruption depends on several key signaling pathways. First, MLC, a 20-kDa protein involved in the contractility of the cytoskeleton underlying the tight junction, is phosphorylated upon EPEC infection. MLC initially resides in the cytosol, where it is phosphorylated on the threonine residue by PKC (2), which is activated upon EPEC infection (17). Following the initial phosphorylation event, serine phosphorylation of MLC occurs, most likely by MLC kinase (MLCK) (99), and the protein becomes associated with the host cytoskeleton. These effects on MLC ultimately lead to a disruption of tight junction integrity and an increase in paracellular permeability. Second, the cytoskeletal protein ezrin has been shown to play a role in tight junction disruption. EPEC infection leads to activation of ezrin, as marked by threonine phosphorylation (which opens the molecule into its active state) and tyrosine phosphorylation (which maintains the molecule in its active state and induces further signaling events) (83). Expression of dominant-negative ezrin prevents the drop in TER normally seen during EPEC infection. Third, direct effects on tight junction proteins themselves have been observed. EPEC infection induces a shift in the tight junction protein occludin from its normal location to an intracellular compartment (84). Further evidence of cadherin junction disruption has recently been shown by Malladi and colleagues (59), who demonstrated that β-catenins normally found in association with membrane-bound cadherins dissociate and are redistributed in the cytoplasm upon EPEC infection.

In addition to disrupting barrier function, EPEC infection leads to a disruption in the maintenance of cell polarity (67). EPEC infection of polarized T84 cells results in a redistribution of basolateral proteins, β-1 integrins and Na⁺/K⁺ ATPase to a more apical location. The initial relocation event depends on Tir. Another EPEC-secreted protein, EspF, has also been demonstrated to be involved in the disruption of tight junctions through a currently uncharacterized mechanism (27).

Immune Response to EPEC Infection

In addition to A/E lesions and the disruption of barrier function, EPEC induces inflammation, which has obvious pathophysiological consequences. EPEC infection of intestinal epithelial cells results in the transepithelial migration of neutrophils. This increase in transmigration is due in part to the secretion of the chemokine IL-8 from the basolateral side of polarized intestinal epithelial cells. EPEC-induced IL-8 expression is regulated by NF-κB, a transcription factor found in the cytoplasm bound to inhibitory proteins in its resting state. EPEC stimulates NF-κB activation by phosphorylating and degrading the NF-κB inhibitory molecule, IκBα (79). This degradation occurs in a TNFα/IL-1β receptor-independent manner and does not rely on Ca²⁺ signaling. Instead, degradation is regulated by EPEC activation of the ERK1/2 signaling pathway (79). EPEC induction of this inflammatory response is distinct from signaling pathways involved in the disruption of tight junctions.

Experiments conducted by de Grado and co-workers (24) have demonstrated the involvement of additional immune regulators up-regulated by EPEC infection. Most strikingly, the up-regulation of egr-1, an immediate-early gene induced in response to changes in the local cell environment, was observed in response to EPEC infection in vitro or during Citrobacter rodentium infection of mice in vivo. Up-regulation was not due to EPEC LPS but instead depended on EPEC-secreted protein(s) delivered via the TTSS. This up-regulation occurs early within infection and depends on activation of the MEK/ERK signaling pathway.

Involvement of Lipid Rafts in EPEC Infection

Lipid rafts are small (approximately 40 nm in diameter), dynamic "patches" in the plasma membrane with a high cholesterol and glycosphingolipid content. Upon external stimulus, lipid rafts aggregate to form larger platforms, from which a variety of signaling events can occur. Many signaling molecules are found to be associated with lipid rafts, and several have been demonstrated to be recruited to the site of EPEC attachment to the epithelial cell. For example, the membrane receptor for hyaluronic acid, CD44, and a membrane fusion and trafficking protein, calpactin (p11), colocalize in lipid rafts and are both recruited to the site of EPEC attachment in a Tir-independent manner (34). Calpactin acts together with annexin II at the plasma membrane to function in membrane fusion and host cell exocytosis. Annexin II is also localized to the site of EPEC attachment in a Tir-independent manner. Further evidence of raft-clustering beneath EPEC comes from the concentration of glycosyl phosphatidylinositol-anchored proteins and cholesterol at sites of bacterial contact. It is currently hypothesized that EPEC may combine these rafts during initial attachment as a means of stabilizing itself and gathering components of the actin cytoskeleton to form pedestals.

EPEC Interactions with Phagocytic Cells

In addition to forming pedestals on intestinal epithelial cells, a rabbit strain of EPEC, RDEC-1, has been shown to form pedestals on the surface of microfold (M) cells (46). M cells are located in Peyer’s patches, where they actively sample contents of the intestinal lumen and facilitate antigen presentation in the underlying lymphoid tissue for the initiation of adaptive mucosal immune responses. The ability of pedestal-forming pathogens to prevent their own uptake into phagocytic cells has also been observed in vitro. EPEC interacts with cultured J774A.1 and RAW macrophages to prevent uptake by these cells (12, 35). This antiphagocytic phenotype requires a functional TTSS but is Tir independent. Celli and coworkers (12) demonstrated that the antiphagocytic phenotype was due to an inhibition of PI3-K-dependent F-actin rearrangements that are normally required for bacterial uptake. EPEC can inhibit FCγR-mediated phagocytosis, which is PI3-K dependent, but cannot inhibit complement receptor CR3-mediated phagocytosis, which is independent of PI3-K (11, 12). It is suggested that EPEC-mediated antiphagocytosis may play several roles during EPEC pathogenesis, including blocking uptake by M cells, which would delay the development of adaptive immune responses to EPEC, and blocking uptake by macrophages and neutrophils, which might otherwise destroy bacteria and present EPEC antigens to mucosal lymphocytes.

Pathogenesis of EPEC-Associated Diarrhea

Several mechanisms have been proposed for EPEC-mediated watery diarrhea. Chronic diarrhea may result in part from the effacement of microvillar structure, leading to a loss of absorptive surfaces and enzymes crucial to proper absorption. This is not the only factor involved, however, since brush-border disruptions are specifically localized at the site of bacterial attachment and do not affect large areas of the intestine. Furthermore, diarrhea appears too quickly in infected volunteers to account for this mechanism (25). Pedestal formation following the degradation of microvilli may also contribute to the continued production of diarrhea, since recruitment of various cytoskeletal proteins to the site of bacterial attachment and effects on MLC may contribute to a disruption of tight junction integrity, leading to increased paracellular permeability and reduced transepithelial resistance. Increasing tight junction permeability could alter normal ion transport mechanisms as well, allowing electrochemical gradients to reach equilibrium. Indeed, EPEC does alter net ion secretion in polarized epithelial cells. In fact, EPEC inhibits net ion secretion stimulated by classic secretagogues, resulting in a decreased short-circuit current (Isc) (46, 80). In one study, attenuation was shown to depend partly on HCO₃^– secretion and less on Cl^– secretion (46). Conflicting data from Collington and colleagues (16) suggests that EPEC can induce a transient, rapid increase in Isc that is at least partially Cl^– dependent. Differences in methodology and cell lines may account for these discrepancies.

The onset of diarrhea occurs as soon as 4 h post-ingestion of EPEC in human volunteers (25). A report by Crane et al. (18) demonstrates one possible mechanism by which watery diarrhea could be induced within this time frame. During EPEC infection of polarized epithelial cells, ATP is released from the apical surface into the supernatant. Extracellular ATP is quickly broken down into adenosine and phosphorylated nucleotides, which can then trigger chloride release. The influx of Na⁺ and possibly amino acids into the cell and the secretion of HCO₃^– out of the cell may also account for some of the ion imbalance induced by infection (16). Any of these disruptions might conceivably trigger the initial stages of watery diarrhea observed during EPEC infection.

SUMMARY

The linkage between signal transduction cascades, the cytoskeleton, and membrane dynamics and trafficking becomes apparent in studies of host-pathogen interactions. For example, the link between membrane trafficking and signal transduction pathways is apparent in the interaction between Rab7 and SifA during SIF formation in Salmonella infection. The movement of PKC into Triton X-100-insoluble domains during Salmonella infection and the recruitment of CD44, annexin II, and glycosylphosphatidyl inositol (GPI) proteins upon EPEC attachment to the epithelial cell are hallmarks of lipid rafts, which concentrate signaling molecules, phosphoinositides, cytoskeletal components, and membrane receptors. The dual functions of bacterial effector proteins, like EPEC Tir and Salmonella SptP, underscore the importance of the host cell’s interconnected signaling pathways and the need for pathogens to evolve numerous ways to subvert them. It becomes clear from these observations that bacterial pathogens like EPEC and Salmonella provide extremely useful tools to demonstrate the connections and interplay among structural, membrane, and nuclear components of signal transduction systems.