The UPEC Pore Forming Toxin α-Hemolysin Triggers Proteolysis of Host Proteins to Disrupt Cell Adhesion, Inflammatory and Survival Pathways

SUMMARY

Uropathogenic Escherichia coli (UPEC), which are the leading cause of both acute and chronic urinary tract infections, often secrete a labile pore-forming toxin known as α-hemolysin (HlyA). We show that stable insertion of HlyA into epithelial cell and macrophage membranes triggers degradation of the cytoskeletal scaffolding protein paxillin and other host regulatory proteins, as well as components of the proinflammatory NFκB signaling cascade. Proteolysis of these factors requires host serine proteases, and paxillin degradation specifically involves the serine protease mesotrypsin. The induced activation of mesotrypsin by HlyA is preceded by redistribution of mesotrypsin precursors from the cytosol into foci along microtubules and within nuclei. HlyA intoxication also stimulated caspase activation, which occurred independently of effects on host serine proteases. HlyA-induced proteolysis of host proteins likely allows UPEC to not only modulate epithelial cell functions, but also disable macrophages and suppress inflammatory responses.

INTRODUCTION

The uroepithelial cells that comprise the stratified layers of the bladder mucosa have an exceptionally slow turnover rate, but will often exfoliate in response to infection with uropathogenic Escherichia coli (UPEC) (Bower et al., 2005). These bacteria are able to bind and invade host cells, where they can replicate to high levels, forming intracellular biofilm-like communities (Anderson et al., 2003; Mulvey et al., 2001). Alternatively, internalized UPEC can enter a more quiescent state that likely facilitates long-term bacterial persistence within the urinary tract (Blango and Mulvey, 2010; Eto et al., 2006; Mulvey et al., 2001). The exfoliation of uroepithelial cells can function to clear adherent and intracellular bacteria, but may also promote the dissemination of UPEC both within and outside of the host. Consequently, the shedding of uroepithelial cells can be viewed as a double-edged sword, potentially benefitting both host and pathogen. Exfoliation in response to infection or injury triggers a rapid regenerative response in the bladder mucosa and underlying stromal cells (Mysorekar et al., 2009; Mysorekar et al., 2002; Shin et al., 2011). While the regeneration of bladder urothelial cells has been defined over the past few years with increasing clarity, the host and bacterial factors that modulate bladder cell exfoliation during the course of a urinary tract infection (UTI) remain poorly understood.

In general, host cell exfoliation entails degradation of cell-cell and cell-matrix contacts, presumably requiring disassembly of focal adhesions and other functionally related structures. The manipulation of host cell exfoliation and turnover rates within mucosal barriers likely impacts the establishment, dissemination, and persistence of diverse bacterial pathogens both within and outside of the urinary tract. For example, the dental pathogen Porphyromonas gingivalis secretes proteases that degrade focal adhesion-associated components like focal adhesion kinase (FAK) and the cytoskeletal scaffolding protein paxillin, stimulating the exfoliation of oral keratinocytes and thereby creating a niche for bacterial expansion (Nakagawa et al., 2006). In contrast, in the intestinal tract where the continuous shedding and rapid turnover of epithelial cells is the norm, pathogens may benefit by hindering host cell exfoliation. Shigella flexneri and several other enteric pathogens accomplish this by stabilizing focal adhesions via secretion of a protein complex (OspE) that interacts with integrin linked kinase and consequently suppresses phosphorylation of FAK and paxillin (Kim et al., 2009). Likewise, Neisseria gonorrhoeae can attenuate normal host cell exfoliation within the vaginal mucosa by altering the composition of focal adhesions via engagement of specific carcinoembryonic antigen-related cell adhesion molecule (CEACAM) host receptors (Muenzner et al., 2010). Pathogens like Haemophilus influenzae, Neisseria meningitidis, and Moraxella catarrhalis may act similarly at other sites within the host.

In light of these and related observations, we sought to uncover possible mechanisms by which UPEC might disrupt focal adhesion-associated components like paxillin and thereby prime bladder epithelial cells for exfoliation. Our results indicate that sublytic concentrations of the UPEC-associated pore-forming toxin α-hemolysin (HlyA) can trigger rapid degradation of paxillin as well as a subset of other host proteins involved in cell-cell and cell-matrix interactions. We show that HlyA also induces proteolysis of key host cell signaling factors, including those central to host inflammatory responses. HlyA-induced proteolysis of these proteins is indirect, occurring in parallel with caspase activation via a pathway involving an atypical host serine protease known as mesotrypsin. Finally, we demonstrate that HlyA can similarly affect macrophages, highlighting unexpected roles for HlyA as a modulator of both phagocyte and epithelial cell functionality and survival.

RESULTS

UPEC triggers proteolysis of paxillin

Paxillin is a 68 kDa multi-domain scaffold protein that can modulate the dynamics of cytoskeletal rearrangements and host cell adhesion via coordinated interactions with numerous structural and regulatory proteins (Deakin and Turner, 2008). The stability of paxillin in human bladder epithelial cells (BECs) infected with the reference UPEC isolate UTI89 was monitored by Western blot analysis. Infection of BECs with UTI89 resulted in decreased levels of paxillin within about 1.5 h, and by 4 h the protein became undetectable (Figure 1A). UTI89-mediated degradation of paxillin was confirmed in BECs made to express eGFP (enhanced green fluorescent protein) fused to the N-terminus of paxillin. In these transfected cells, degradation of GFP-paxillin in response to infection with UTI89 was observed to occur in a stepwise fashion, resulting first in the release of GFP fused to a ~10 kDa N-terminal fragment of paxillin followed by release of intact, full-length GFP (Figure 1B). GFP itself was not degraded in these assays. Additional work showed that lipopolysaccharide (LPS), the type 1 pilus-associated adhesin FimH, and host cell invasion by UPEC or other invasive pathogens were not sufficient to trigger paxillin proteolysis within infected BECs (Figure S1).

Figure 1. Paxillin is degraded in UPEC-infected bladder cells.

(A) Lysates from uninfected (UI) BECs and BECs that were infected with UTI89 for the indicated time periods were probed by Western blot using antibodies specific for paxillin. Staining for β-actin levels in each sample served as a loading control.

(B) Lysates from BECs transfected with the plasmids pEGFP or pEGFP-paxillin and then infected with UTI89 for 0, 2 or 4 h were analyzed by Western blot using antibodies specific for GFP (top), paxillin (middle), or actin (bottom).

See also Figure S1.

To gain further insight into what factor(s) elicit paxillin degradation in our infection assays, and to determine how widespread this phenomenon was, we utilized a panel of other distinct bacterial isolates to infect BECs. Of 20 UPEC isolates examined, 14 (60%) triggered proteolysis of paxillin (Figure 2A). In contrast, paxillin remained stable in BECs infected with any of several non-UPEC isolates tested. The non-E. coli uropathogens Staphylococcus saprophyticus, S. aureus, Enterococcus faecium and Klebsiella pneumoniae also failed to induce paxillin degradation. With the exception of S. aureus, we noted that each strain that promoted loss of paxillin in these assays was also hemolytic on blood agar plates.

Figure 2. HlyA triggers paxillin degradation in BECs along with the disruption of host actin and microtubule networks.

(A) Western blots show levels of paxillin and actin in 5637 BECs following 4 h infections with UPEC isolates, non-UPEC E. coli strains, or non-E. coli uropathogenic bacteria. Uninfected (UI) control cells were treated with M9 medium without bacteria. Bacterial strains were scored as hemolytic (+) based on their ability to lyse erythrocytes on blood agar plates.

(B - D) Western blots show paxillin and actin levels present in BECs at the indicated time points after infection with wild type UTI89, UTI89ΔhlyA, UTI89ΔhlyA/pSF4000, UTI89Δcnf1, UTI89Δcnf1/pHLK102, WAM783 (+hlyABD ΔhlyC), or WAM582 (+hlyCABD).

(E) Western blots show paxillin and actin levels in 5637 BECs following 0 to 4 h incubations with preparations of crude, bacteria-free non-acylated HlyA (HlyA⁷⁸³) or wild type HlyA (HlyA^783/pHlyC).

(F) Phase contrast microscopy of BECs infected with UTI89 or UTI89ΔhlyA for the indicated times. Scale bar, 50 μm.

(G) Confocal images of BECs infected for 4 h with UTI89 or UTI89ΔhlyA and then stained to visualize microtubules or F-actin. Scale bar, 20 μm.

(H - I) BECs were infected with UTI89 for the indicated times in the presence of MβCD (5 mM), (B) EGTA (4 mM), the intracellular Ca²⁺ chelator BAPTA-AM (10 μM), or vehicle alone (DMSO, control). Paxillin and actin levels were assessed by Western blots.

See also Figure S2 and Table S1.

Degradation of paxillin is initiated by HlyA

Hemolytic activity associated with UPEC isolates is generally attributable to the pore-forming exotoxin α-hemolysin (HlyA), a prototypical member of the RTX (repeats-in-toxin) protein family encoded within the hlyCABD operon (Cavalieri et al., 1984; Linhartova et al., 2010; Marrs et al., 2005). We found that targeted deletion of hlyA in UTI89 rendered this strain non-hemolytic and unable to stimulate paxillin degradation in BECs (Figure 2B). Complementation of the hlyA mutant with a plasmid (pSF4000) encoding the complete hly operon restored the wild type phenotypes. In many UPEC isolates, including UTI89, the hly operon is genetically linked to a locus encoding another toxin known as cytotoxic necrotizing factor-1 (CNF1) (Marrs et al., 2005). Both HlyA and CNF1 have the potential to influence host cell death and tissue turnover rates within the bladder (Doye et al., 2002; Mills et al., 2000; Smith et al., 2008), but in our assays neither deletion of the cnf1 gene nor expression of recombinant CNF1 affected the ability of UTI89 to stimulate paxillin degradation (Figure 2C).

The hly operon also encodes the acyltransferase HlyC, which is responsible for acylating two conserved lysine residues within maturing HlyA toxin molecules. In the absence of acylation mediated by HlyC, HlyA is secreted and interacts with host cell membranes, but fails to oligomerize to form functional pores (Herlax et al., 2009; Stanley et al., 1998). Infection of BECs with WAM582, a K-12 strain that carries the complete wild type hly operon, promoted rapid degradation of paxillin similar to hly-positive UPEC isolates (Figure 2D). In contrast, paxillin remained stable in BECs following infection with the isogenic mutant WAM783, which encodes the hly operon minus hlyC. Of note, both WAM783 and WAM582 express similar amounts of HlyA as determined by Western blot (data not shown). Complementation of WAM783 with a plasmid encoding HlyC made this strain (WAM783/pHlyC) hemolytic and able to induce paxillin degradation (Figure 2D). Crude, bacteria-free preparations of acylated HlyA recovered from WAM783/pHlyC culture supernatants also triggered paxillin degradation, while non-acylated HlyA preparations isolated from WAM783 cultures had no effect on the stability of paxillin (Figure 2E).

Cumulatively, these results indicate that the insertion of acylated HlyA into host membranes can trigger paxillin degradation, but do not rule out possible involvement of additional bacterial factors common to both K-12 and UPEC isolates. This process did not require lytic concentrations of HlyA, as BECs intoxicated with HlyA in these experiments remained for the most part intact and viable as determined by trypan blue exclusion assays. However, bacterial strains expressing wild type HlyA did cause substantial rounding and eventual lifting of the BECs (Figure 2F and S2A-B), coincident with massive disruption of host microtubule and actin networks (Figure 2G).

Effects of HlyA membrane insertion and downstream signaling events

In addition to acylation by HlyC, the lytic activity of HlyA is also dependent upon extracellular Ca²⁺ (Bauer and Welch, 1996; Ostolaza and Goni, 1995). This cation binds C-terminal GGXGXD repeats within HlyA and promotes proper folding and insertion of the toxin into the host cell plasma membrane where it likely becomes concentrated and oligomerizes within cholesterol-rich detergent resistant microdomains (Bakas et al., 1998; Herlax et al., 2009; Ludwig et al., 1988). In our assays, depletion of cholesterol (using methyl β-cyclodextran, MβCD) or chelation of extracellular Ca²⁺ (using EGTA) prevented the stable association of HlyA with BECs (Figure S3G), completely inhibiting both the cytotoxic effects of HlyA and paxillin degradation within UTI89-infected BECs (Figures 2H and I, S3H and data not shown).

PFTs like HlyA can stimulate intracellular Ca²⁺ fluxes, the release of cellular K⁺ ions, and activation of mitogen-activated protein (MAP) kinases (Bhakdi et al., 1986; Koschinski et al., 2006; Porta et al., 2011). However, neither intracellular Ca²⁺ fluxes nor K⁺ efflux affected the loss of paxillin within UTI89-infected BECs, and inhibition of MAP kinase signaling caused at best only a modest delay in paxillin degradation (Figures 2I, S2C-F). In addition, two other PFTs (α-toxin from S. aureus and aerolysin from Aeromonas hydrophila) with the ability to alter cellular cation fluxes and MAP kinase signaling failed to trigger the proteolysis of paxillin (Figure S2G). We also ruled out the involvement of host purinergic P2X receptors and pannexin1 channels, which were recently proposed to amplify the hemolytic effects of HlyA (Figure S3) (Skals et al., 2009). Interestingly, three P2X antagonists – suramin, Evans Blue, and Brilliant Blue G – did effectively block HlyA-induced paxillin degradation in our assays, but appeared to do so by non-specifically inhibiting the stable association of HlyA with BECs, similar to MβCD and EGTA (Figure S3G). Thus, while our data refute the involvement of P2X receptors and pannexin1 as necessary facilitators of HlyA-induced paxillin degradation, they substantiate the idea that loss of paxillin within infected BECs requires the stable insertion of HlyA into target host membranes.

HlyA-dependent activation of caspases and host serine proteases

HlyA is not known to have any inherent proteolytic activity, suggesting that it may stimulate paxillin degradation by activation of one or more host proteases. Using pharmacological inhibitors, we tested the possible involvement of several of the better-known proteolytic pathways in eukaryotic cells. Our results indicated that the proteosome, calpains, and lysosomal proteases like cathepsins are likely not critical for paxillin degradation in response to HlyA (Figure S4A-C).

We also examined possible involvement of caspases, a family of conserved aspartate-specific cysteine proteases (caspases, CASP1-12) that have central roles in the initiation and execution of apoptosis (Chowdhury et al., 2008). Using fluorochrome-labeled inhibitor of caspases (FLICA) reagents (Bedner et al., 2000), we found that wild type UTI89, but not the isogenic hlyA mutant, stimulates rapid activation of the executioner caspases CASP3 and CASP7 in BECs (Figure 3A). The addition of suramin, which interferes with the effects of HlyA on BECs (see Figure S3), prevented caspase activation by UTI89 (Figure 3B). These data indicate that HlyA expression by UPEC activates caspases within BECs, in agreement with previous studies showing that HlyA can promote apoptosis in a variety of host cell types (Chen et al., 2006; Jonas et al., 1993; Mobley et al., 1990; Russo et al., 2005). HlyA-induced caspase activation in our assays correlated with the degradation of paxillin, suggesting a functional link between these events. However, the treatment of UTI89-infected BECs with the CASP3/7-specific inhibitor Z-DEVD-FMK, the pan-caspase inhibitor Z-VAD-FMK, or the CASP1/4 inhibitor Z-YVAD-FMK failed to prevent loss of paxillin downstream of HlyA intoxication (Figure 3C). Thus, while HlyA stimulates caspase activation within infected BECs, these proteases are not integral to paxillin degradation in our experiments.

Figure 3. Activation of caspases and TLCK-sensitive serine proteases by HlyA.

(A) Caspase-3/7 activation over time in 5637 BECs infected with UTI89 or UTI89ΔhlyA, as determined by fluorometry of FLICA-treated samples. Data are expressed as the mean ± SEM of three independent experiments carried out in duplicate or triplicate. *p ≤ 0.0001, as determined by Student’s t-test.

(B) Effects of suramin (1 mM) and TLCK (100 μM) on caspase-3/7 activation in BECs infected for 4 h with UTI89, as quantified by fluorometry of FLICA-treated samples. Control UTI89-infected samples were treated with vehicle alone. Data are expressed as the mean ± SEM of four independent experiments performed in triplicate. *p ≤ 0.0001, versus control values, as determined by Student’s t-test.

(C) Western blots showing levels of paxillin and actin in BECs infected with UTI89 for the indicated times in the presence of 100 μM of the CASP3/7-specific inhibitor Z-DEVD-FMK, the pan-caspase inhibitor Z-VAD-FMK, the CASP1/4 inhibitor Z-YVAD-FMK, or the negative control Z-FA-FMK.

(D) Paxillin and actin levels in BECs infected with UTI89 for the indicated times in the presence or absence of TLCK (100 μM), as determined by Western blots.

(E) Phase contrast (left) and corresponding fluorescent images (right) of BECs infected with UTI89 or UTI89ΔhlyA for 4 h in presence of the fluorescent probe FSLCK (5 μM). Scale bar, 20 μm.

(F) Fluorometric quantification of the levels of FSLCK-labeled, active serine proteases over time in BECs infected with UTI89 or UTI89ΔhlyA. Data are expressed as the mean ± SEM of three independent experiments carried out in triplicate. *p ≤ 0.0001, as determined by Student’s t-test.

(G) Immunoblots probed using anti-FITC antibody to detect FSLCK-tagged proteins recovered from uninfected BECs, BECs infected with UTI89ΔhlyA, or BECs infected with UTI89 ± 1 mM suramin.

(H) Levels of activated FSLCK-tagged serine proteases present in the bladder mucosa of mice 5 h after infection with UTI89 or UTI89ΔhlyA were quantified by fluorometry. Data are expressed as the mean ± SEM from at least 5 mice. *p = 0.0017, as determined by Student’s t-test.

(I) Activated serine proteases present in the bladder mucosa of uninfected mice or mice infected with UTI89 or UTI89ΔhlyA for 5 h were detected, following treatment with FSLCK, by Western blots probed with anti-FITC antibody. Further analysis by mass spectroscopy identified the 23-kDa band (arrow) as mesotrypsin.

See also Figure S3.

Caspases often act in concert with other proteolytic enzymes, including serine proteases (Moffitt et al., 2007). These enzymes make up more than one-third of all known proteases and function in wide-ranging biological processes (Page and Di Cera, 2008). A role for serine proteases as mediators of paxillin degradation in response to HlyA was assessed using the chymotrypsin-like serine protease inhibitor tosyl-L-phenylalanine-chloromethyl ketone (TPCK) and the trypsin-like serine protease inhibitor tosyl-L-lysine-chloromethyl ketone (TLCK). TPCK had no notable effect on paxillin stability in UTI89-infected BECs (data not shown), whereas TLCK completely inhibited HlyA-induced paxillin proteolysis (Figure 3D). Of note, TLCK did not interfere with the stable association of HlyA with BECs (Figure S3G), and also had no effect on either HlyA-induced rounding of target host cells (Figure S3H) or HlyA-dependent activation of caspases (Figure 3B). The latter observation, and the inability of caspase inhibitors to block paxillin degradation in our assays, indicates that serine proteases are activated in response to HlyA intoxication in parallel with, but independent of, caspases.

Carboxyfluorescein-spacer-leucine chloromethyl ketone (FSLCK) covalently binds and fluorescently tags activated serine proteases (Grabarek and Darzynkiewicz, 2002). Using FSLCK as a probe, we observed by fluorescent microscopy high levels of activated serine proteases in UTI89-infected BECs, versus relatively low levels in host cells infected with the isogenic hlyA mutant (Figure 3E). Fluorometric quantitation of BECs probed with FSLCK confirmed that wild type UTI89 caused significant activation of serine proteases within 2 h of infection, while the hlyA mutant elicited almost no changes above background (Figure 3F). By Western blot analysis, using antibody specific for the carboxyfluorescein fluorochrome within FSLCK, multiple FSLCK-tagged protein bands were detected in lysates from BECs following a 3 h infection with wild type UTI89 (Figure 3G). The number and intensities of the bands were markedly reduced in lysates from BECs infected with the hlyA mutant, and nearly absent in lysates from uninfected BECs or BECs that were infected with UTI89 in the presence of suramin. In total, these results indicate that HlyA can trigger the robust activation of TLCK-sensitive serine proteases within BECs, one or more of which mediate paxillin degradation.

Mesotrypsin promotes paxillin degradation downstream of HlyA

To examine the ability of HlyA to influence activation of serine proteases in vivo within the bladder, we utilized FSLCK in conjunction with a well-established mouse UTI model. By fluorometry, significantly higher levels of activated serine proteases were detected within the bladder mucosa of mice following infection with wild type UTI89 versus the hlyA mutant (Figure 3H). Western blots of bladder mucosa lysates from both infected and uninfected mice revealed only two FSLCK-tagged protein bands with apparent molecular weights of about 23 and 70 kDa (Figure 3I). Similarly sized FSLCK-tagged protein bands were also detected, among many others, in UTI89-infected BECs in our cell culture-based assays (arrowheads, Figure 3G). FSLCK labeling of the 23-kDa protein band was consistently more intense following infection of either cultured BECs or mouse bladders with wild type UTI89 relative to UTI89ΔhlyA-infected samples and uninfected controls.

Using an immunoprecipitation approach coupled with mass spectroscopy, we identified the 23-kDa FSLCK-tagged protein in lysates from UTI89-infected mouse bladders as a TLCK-sensitive host serine protease known as mesotrypsin. This enzyme is generated via processing of three zymogens, known as trypsinogens 3, 4, and 5, that are produced as splice variants of the gene PRSS3 (Nakanishi et al., 2010). These trypsinogen isoforms differ only in their N-terminal region, which is cleaved at conserved DDDDK-I sequences by the type II transmembrane serine protease enteropeptidase to produce active mesotrypsin (Light and Janska, 1989). The mesotrypsin enzyme is highly similar to trypsins 1 and 2, major pancreatic digestive proteases derived from the zymogenic precursors trypsinogens 1 and 2, respectively (Nakanishi et al., 2010). However, relative to trypsins 1 and 2, mesotrypsin has distinct substrate specificity and a higher level of resistance to many natural, endogenous protease inhibitors (Nakanishi et al., 2010; Sahin-Toth, 2005).

Among trypsinogens 1 through 5, only trypsinogen 4 transcripts were detected by RT-PCR in the human BECs used in our cell culture-based assays (Figure 4A). Transcription of the activating enzyme enteropeptidase in these host cells was also observed. Using siRNA to silence expression of trypsinogen 4 (PRSS3, Figure 4B), we were able to significantly delay the degradation of paxillin in UTI89-infected BECs (Figure 4C). These data indicate that the active form of trypsinogen 4, mesotrypsin, is at least partially responsible for loss of paxillin in HlyA-intoxicated BECs. Interestingly, we found that trypsinogen 4 undergoes radical redistribution in response to HlyA, as inferred using immunofluorescence microscopy with antibody directed against the enteropeptidase cleavage site DDDDK (anti-ECS antibody, Figure 4D). Use of the anti-ECS antibody serves as a reasonable surrogate for direct immunodetection of trypsinogen 4, as none of the other known trypsinogens appear to be expressed by the BECs used in our assays (see Figure 4A) (Nakanishi et al., 2010). In the presence of HlyA, trypsinogen 4 moves from a diffuse, primarily cytosolic localization to a more reticulated and punctate distribution at the cell periphery, along microtubules, and within and around nuclei (Figures 4D and S4D). This process is initiated within 1 h of infection with wild type UTI89, preceding the loss of paxillin.

Figure 4. HlyA induced paxillin degradation is facilitated by mesotrypsin.

(A) 5637 BECs express enteropeptdase and trypsinogen 4, but not other trypsinogens, as determined by RT-PCR.

(B) Semi-quantitative RT-PCR shows levels of PRSS3 (encoding trypsinogen 4) and GAPDH transcripts in BECS 72 h after transfection of 5637 BECs with either control nonspecific siRNA or PRSS3-specific siRNA detected.

(C) Western blots show levels of paxillin and actin present in lysates recovered from BECs treated with control, nonspecific siRNA or PRSS3-specific siRNA and infected with UTI89 for the indicated times.

(D) Confocal micrographs of uninfected BECs or BECs that were infected with UTI89 or UTI89ΔhlyA for the indicated times and then stained to visualize trypsinogens (using ECS antibody, green), LAMP1 (red), and nuclei (blue). The boxed area in each micrograph is enlarged and placed directly below the corresponding image to show details. Scale bars, 20 μm (first and third rows) and 5 μm (second and fourth rows).

See also Figure S4.

TLCK-sensitive, HlyA-dependent proteolysis of multiple host proteins

Induced activation of mesotrypsin and, perhaps, other TLCK-sensitive proteases by HlyA likely has effects extending beyond paxillin degradation. To examine this possibility, lysates from BECs that were infected with wild type UTI89 in the presence or absence of TLCK, or with UTI89ΔhlyA, were assayed by Western blot analyses using antibodies specific for a panel of 24 host proteins. Nearly half of these were degraded in a TLCK-sensitive fashion within 4 h of infection with wild type UTI89, but not the hlyA mutant (Figures 5A and B). Although these assays were too restricted to discern any definitive pattern in the types of host proteins that are degraded in response to HlyA intoxication, they do demonstrate the ability of HlyA to stimulate the loss of many proteins with key roles in mediating host cell adherence and signal transduction. These include the paxillin homologue Hic-5, the docking proteins HEF-1 and p130-CAS, and the adherens junction component and signaling factor β-catenin.

Figure 5. Widespread effects of HlyA on host protein stability and NFκB signaling.

(A - B) Levels of the host proteins indicated were assessed by Western blot using lysates from 5637 BECs infected for 0 to 4 h with UTI89ΔhlyA or wild type UTI89 ± TLCK (100 μM).

(C) 5637 BECs were infected with UTI89ΔhlyA or wild type UTI89 in the presence of LPS (5 μg/ml) or incubated with LPS alone, and 4 h later IL-6 levels in the cell culture supernatants were measured by ELISA. Data are expressed as the mean ± SEM of results from at least three independent experiments performed in triplicate; p values determined by Student’s t-test.

In addition to its role as a mediator of cell adhesion, β-catenin also functions in developmental and oncogenic cascades within the Wnt signaling pathway and as a negative regulator of the proinflammatory transcription factor NFκB (Deng et al., 2002). Interestingly, the TLCK-sensitive degradation of β-catenin in UTI89-infected BECs occurs in step with loss of the prototypical inhibitor of NFκB signaling, IκBα (Figure 5B). Usually, IκBα is ubiquitinated and subsequently degraded by the proteasome in response to proinflammatory signals (Rahman and McFadden, 2011). However, the rapid loss of IκBα in our assays required HlyA and one or more TLCK-sensitive protease, indicating a proteasome-independent mechanism for IκBα degradation in response to HlyA intoxication. Coincident with loss of IκBα and β-catenin, we also observed HlyA-dependent and TLCK-sensitive degradation of the NFκB subunit RelA (also known as p65) (Figure 5B). This correlated with significantly reduced expression the NFκB-regulated cytokine IL-6 by HlyA-intoxicated BECs in response to added lipopolysaccharide (LPS, Figure 5C). In total, these data indicate that HlyA can indirectly disrupt various aspects of host cell signaling, including key components of the NFκB pathway, via effects on serine proteases.

HlyA-induced proteolysis of macrophage proteins

We next asked if the effects of HlyA on the stability of select host proteins could be observed in host cell types other than BECs. Using both the RAW264.7 macrophage cell line and mouse peritoneal macrophages, we found that wild type UTI89, but not the hlyA mutant, triggered the degradation of paxillin as well as IκBα and RelA (Figures 6A - C). Labeling of activated serine proteases with the FSLCK reagent was also more intense in macrophages infected with wild type UTI89 versus UTI89ΔhlyA (Figure 6D). Thus, the capacity of HlyA to activate serine proteases and promote the loss of select host proteins is not limited to BECs.

Figure 6. HlyA triggers the activation of serine proteases and degradation of host proteins within macrophages.

(A- B) Western blots showing levels of paxillin and actin in either RAW264.7 or mouse peritoneal macrophages following infection with either UTI89 or UTI89ΔhlyA for the indicated times.

(C) Western blots showing levels of IκBα, RelA, and actin in RAW264.7 cells following infection with UTI89 or UTI89ΔhlyA for the times indicated.

(D) Phase contrast (left) and corresponding fluorescent images (right) of peritoneal macrophages infected with UTI89 or UTI89ΔhlyA for 3 h in presence FSLCK (5 μM) to label activated serine proteases. Scale bar, 50 μm.

DISCUSSION

In recent years, sub-lytic concentrations of several bacterial PFTs have been shown to modulate important aspects of host cell physiology, including MAP kinase and Akt signaling, inflammasome activation, plasma membrane repair systems, histone methylation, the unfolded protein response, and mitochondrial function (Aroian and van der Goot, 2007; Bischof et al., 2008; Gonzalez et al., 2011; Hamon et al., 2007; Kao et al., 2011; Los et al., 2011; Porta et al., 2011; Stavru et al., 2011; Wiles et al., 2008). The effects of PFTs on target host cells are often attributable to toxin-induced fluctuations in intracellular K⁺ and/or Ca²⁺ levels. In our efforts to identify UPEC-associated factors that influence BEC detachment, using paxillin degradation as a readout, we found that the PFT HlyA can trigger activation of TLCK-sensitive serine proteases independent of K⁺ and Ca²⁺ fluxes. This process required stable interaction between HlyA and the host cells, likely involving oligomerization and integration of the acylated, Ca²⁺-bound toxin into cholesterol-rich microdomains within the host plasma membrane. Two other bacterial PFTs, α-toxin and aerolysin, had no effect on the stability of paxillin, even though they, like HlyA, can stimulate K⁺ and Ca²⁺ fluxes (Krause et al., 1998; Seeger et al., 1984). These results indicate that the capacity of HlyA to promote loss of paxillin is not shared by all PFTs.

Degradation of paxillin in response to HlyA intoxication is due at least in part to activation of mesotrypsin. This protease has recently been shown to regulate the differentiation and desquamation of keratinocytes within the outermost layer of the skin epidermis (Nakanishi et al., 2010). We suggest that mesotrypsin might serve a similar function within the bladder mucosa. Unlike its close relatives, trypsins 1 and 2, mesotypsin is more widely expressed and is resistant to many naturally occurring trypsin inhibitors (Nakanishi et al., 2010; Sahin-Toth, 2005). In some cases, mesotrypsin can also degrade endogenous protease inhibitors, meaning it has the capacity to regulate the functionality of other proteolytic enzymes (Sahin-Toth, 2005). In our assays, silencing the expression of PRSS3 significantly delayed HlyA-induced degradation of paxillin in BECs. Eventual loss of paxillin in these assays likely results from our inability to completely extinguish expression of PRSS3 and/or the involvement of other host, or perhaps even bacterial, trypsin-like serine proteases. The proteolysis of paxillin and other host proteins in response to HlyA intoxication correlated with the redistribution of mesotrypsin precursors from the cytosol to punctate structures situated primarily along microtubules, both at the cell periphery and around and within host nuclei. The means by which HlyA triggers these positional changes and actually signals mesotrypsin activation within BECs require further investigation.

UPEC strains that express HlyA cause more extensive tissue damage within the urinary tract, correlating with more severe clinical outcomes (Johnson, 1991; Marrs et al., 2005; Smith et al., 2008). In both murine and tissue culture model systems, hlyA-positive isolates stimulate more rapid and extensive shedding of BECs than isogenic hlyA-negative mutants (see Figure S2A-B) (Smith et al., 2006; Smith et al., 2008). HlyA is also an important virulence factor associated with other closely related E. coli strains that can cause pneumonia, peritonitis, and septicemia (May et al., 2000; O’Hanley et al., 1991; Russo et al., 2005; Welch et al., 1981; Wiles et al., 2009). In the gut, HlyA mediates the formation of focal leaks within colonic epithelial cells and can thereby promote the paracellular translocation of bacteria (Troeger et al., 2007). HlyA can also attenuate the viability, chemotaxis, and effector functions of host phagocytes like neutrophils and macrophages (Cavalieri and Snyder, 1982a, b; Gadeberg et al., 1989; Russo et al., 2005).

Results presented here suggest molecular processes by which sub-lytic levels of HlyA can modulate the functionality and viability of diverse host cell types. The ability of HlyA to elicit degradation of paxillin as well as other cytoskeletal regulatory and structural components may not only promote bladder cell exfoliation, but can also act to cripple phagocytes. Coordinated activation of both serine proteases and caspases in response to HlyA intoxication represents an additional mechanism by which this PFT can compromise host cell viability. The capacity of HlyA to stimulate degradation of β-catenin, IκBα, and the NFκB subunit RelA may enable UPEC and related strains to promote host cell death while also attenuating host inflammatory responses. The TLCK-sensitive nature of these degradative processes suggests a role for HlyA-activated serine proteases like mesotrypsin in the regulation of NFκB signaling. These results add to the growing list of mechanisms used by microbial pathogens to disturb NFκB-dependent proinflammatory pathways (Rahman and McFadden, 2011).

Our observations that HlyA disrupts NFκB signaling and cytokine expression complement previous findings showing that HlyA can inactivate Akt, another key regulator of host cell survival and inflammation (Wiles et al., 2008). Together, these data explain in large part the capacity of some UPEC isolates to suppress host inflammatory responses, as reported by others (Billips et al., 2007; Hilbert et al., 2008; Hunstad et al., 2005; Loughman and Hunstad, 2011). Interestingly, in some of these earlier studies mutations that alter the ability of UPEC to dampen host cytokine expression levels were mapped to genes that affect the bacterial cell envelope and, potentially, the secretion of toxins like HlyA. In conclusion, our data reveal an unexpected process by which UPEC and related pathogens can manipulate host survival and inflammatory pathways via HlyA-dependent activation of host proteolytic cascades.

EXPERIMENTAL PROCEDURES

Host cells and bacterial strains

Human bladder epithelial cells, designated 5637 cells (ATCC HTB-9), were obtained from the American Type Culture Collection and maintained in RPMI 1640 supplemented with heat inactivated 10% fetal bovine serum (Hyclone, Logan, UT) at 37°C in the presence of 5% CO₂. Peritoneal and RAW264.7 macrophages were used as described in Supplemental Methods. Bacterial strains employed in this study are listed in Table S1. All E. coli strains were grown static in M9 minimal medium at 37°C for 48 h prior to use (Wiles et al., 2008). S. flexneri (ATCC 12022), S. enterica ser. Typhimurium (SL1344), S. saprophyticus, S. aureus, E. faecium and K. pneumoniae were grown shaking in Luria-Bertani (LB) broth (Difco) overnight. Bacteria were diluted in M9 medium (or LB broth for the non-E. coli isolates) to OD₆₀₀ ~ 0.5 prior to use. Antibodies, purified toxins and other reagents used are noted in Supplemental Information.

Infection assays, Western blots, and gene expression analysis

Confluent BEC monolayers grown in 6-well plates were serum starved overnight in 1 ml of RPMI prior to infection or drug treatments. When appropriate, BECs were treated with the indicated compounds or vehicle alone for 1 h prior to infection with UPEC. Host cells were infected with 100 μl of diluted bacterial cultures, corresponding to a multiplicity of infection of about 25. Fresh drugs, vehicle, or other compounds were added at this time and maintained for the duration of each experiment as needed. Suramin was also added fresh again at 2 h after infection to maintain its effectiveness. An aliquot of M9 medium was added as a control to the uninfected samples. Plates were centrifuged at 600 X g for 5 min to facilitate and synchronize bacterial contact with the host cells. At the specified times, BECs were washed with PBS²⁺ (PBS supplemented Ca²⁺ and Mg²⁺) prior to lysis in RIPA buffer supplemented with 1X complete protease inhibitors (Roche Applied Science) and 1 mM phenylmethanesulfonylfluoride (Sigma-Aldrich). Equivalent protein amounts (typically 50 μg of each sample, as determined by BCA assays; Pierce) were resolved by SDS-PAGE, transferred to Immobilon PVDF-FL membrane (Millipore), and processed for Western blot analysis as previously described (Eto et al., 2007; Wiles et al., 2008). Gene cloning, silencing, semi-quantitative RT-PCR, and the isolation of crude HlyA preparations were carried out as detailed in Supplemental Methods.

Detection and quantification of activated caspases and serine proteases

Confluent, serum-starved BEC monolayers in 24-well plates were treated with the FLICA reagents FAM-DEVD-FMK (1X), SR-DEVD-FMK (1X) or FSLCK (5 μM, ImmunoChemistry Technologies; Bloomington, MN) and then infected with UTI89 or UTI89ΔhlyA ± suramin (1 mM) or TLCK (100 μM) as indicated. At the specified times, BECs were washed 3X with PBS²⁺ and lysed in TTD buffer containing 0.25% Triton X-100, 10 mM Tris (pH 7.5), and 1 mM dithiothreitol. Fluorescence measurements were obtained using a LS-5 Fluorescence spectrophotometer (Perkin-Elmer; Waltham, MA) and readings were normalized to the total protein content in each sample, as determined by BCA assays. Alternatively, equal amounts of protein from the FSLCK-treated samples were resolved by SDS-PAGE and probed by Western blot analysis using anti-FITC and anti-actin antibodies.

To assay activation of serine proteases in vivo within the bladder mucosa, adult female CBA/J mice (Jackson Laboratories; Bar Harbor, ME) were inoculated via transurethral catheterization with ~10⁸ CFU of UTI89 or UTI89ΔhlyA as described (Mulvey et al., 1998). Mice were sacrificed at 5 h post-inoculation and bladders were aseptically removed, bisected, splayed and pinned down luminal side up in mammalian Ringer solution (pH 7.4) containing 5 μM FSLCK. After a 20-min incubation at 37°C, the splayed bladders were washed 5X with Ringer solution and the mucosal layer of each bladder was gently peeled off and lysed in TTD lysis buffer. FSLCK-tagged serine proteases in each sample were quantified by fluorometry or visualized by Western blot analysis as described above.

To identify the 23-kDa active serine protease present within the bladders of UTI89-infected mice, mucosal layers isolated from 4 infected, FSLCK-treated bladders were pooled and lysed in RIPA buffer. FSLCK-tagged proteins were then immunoprecipitated using rabbit anti-FITC antibody immobilized on Protein A Sepharose 4B beads, resolved by SDS-PAGE, and stained using GelCode Blue (Thermo Fisher Scientific). The 23-kDa band was then excised, subjected to in-gel trypsin digestion and identified using liquid chromatography/mass spectrometry at the University of Utah core facility.

All mouse experiments were performed with approval from the local Institutional Animal Care and Use Committee.

Microscopy

Phase contrast and fluorescence microscopic images of live BECs and macrophages were acquired using an Olympus CK40 microscope equipped with 10x (Olympus A10PL NA 0.25) and 40X (Olympus CDPlan40FPL NA 0.55) objectives and a Canon PowerShot A640 camera. For confocal microscopy, BECs grown on 12-mm-diameter glass coverslips were fixed, stained, and imaged using an Olympus FV1000 microscope as previously described (Dhakal and Mulvey, 2009).

HIGHLIGHTS.

• HlyA, but not other pore-forming toxins, triggers proteolysis of host proteins.

• Target proteins include those involved in the cytoskeleton, inflammation, and apoptosis.

• Proteolysis requires stable interactions between HlyA and host membranes.

• Proteolysis involves activation and redistribution of mesotrypsin precursors.