An outer membrane protease of the omptin family prevents activation of the Citrobacter rodentium PhoPQ two-component system by antimicrobial peptides

Authors


*E-mail herve.le-moual@mcgill.ca; Tel. (+1) 514 398 6235; Fax (+1) 514 398 7052.

Summary

The PhoPQ two-component system of the intracellular pathogen Salmonella enterica senses and controls resistance to α-helical antimicrobial peptides (AMPs) by regulating covalent modifications of lipid A. A homologue of the phoPQ operon was found in the genome of the murine enteric extracellular pathogen, Citrobacter rodentium. Here we report that C. rodentium PhoPQ was apparently unable to mediate activation of target genes in the presence of α-helical AMPs. However, these AMPs activated C. rodentium PhoPQ expressed in a S. entericaΔphoPQ mutant. Analysis of the outer membrane (OM) fractions of the C. rodentium wild-type and ΔphoPQ strains led to the identification of an omptin family protease (CroP) that was absent in ΔphoPQ. Deletion of croP in C. rodentium resulted in higher susceptibility to α-helical AMPs, indicating a direct role of CroP in AMP resistance. CroP greatly contributed to the protection of the OM from AMP damage by actively degrading α-helical AMPs before they reach the periplasmic space. Accordingly, transcriptional activation of PhoP-regulated genes by α-helical AMPs was restored in the ΔcroP mutant. This study shows that resistance to α-helical AMPs by the extracellular pathogen C. rodentium relies primarily on the CroP OM protease.

Introduction

Antimicrobial peptides (AMPs) are important components of the innate immune system that are involved in host defence against microorganisms (Hancock and Lehrer, 1998; Zasloff, 2002; Hancock and Sahl, 2006). Although diverse in sequence and structure, AMPs are characterized by their small size, amphipathic properties and cationic net charges. There are two major classes of mammalian AMPs, the cathelicidins, such as human LL-37 and mouse CRAMP that form α-helices, and the α and β-defensins that adopt a β-sheet structure (Ganz, 2003; Zanetti, 2004). Various cell types including phagocytic cells, goblet cells and epithelial cells produce AMPs in a constitutive or inducible manner. In Gram-negative bacteria, AMPs interact first with the lipid A moiety of lipopolysaccharide (LPS) to disrupt the outer membrane (OM) and gain access to the periplasmic space. Most AMPs appear to exert their bactericidal function by disrupting the cytoplasmic membrane (Brogden, 2005; Melo et al., 2009). Bacterial pathogens have developed several mechanisms to resist killing by AMPs. These resistance mechanisms include the protease-mediated inactivation of AMPs, the export of AMPs by efflux pumps and the covalent modification of bacterial membrane components that results in a decrease of the overall negative charge and, in turn, minimize membrane interaction with cationic AMPs (Miller et al., 2005; Peschel and Sahl, 2006).

The PhoPQ two-component system (TCS) is composed of the sensor kinase PhoQ and the cognate response regulator PhoP. PhoPQ is found in a variety of Gram-negative pathogens, including the facultative intracellular Salmonella enterica. PhoPQ of S. enterica, which is by far the best-characterized PhoPQ system, is a master regulator of virulence that is critical for bacterial survival within macrophage phagosomes (Groisman et al., 1989; Miller et al., 1989). PhoQ, like most sensor kinases, possesses both kinase and phosphatase activities that play opposite roles in controlling the phosphorylation level of PhoP. These activities are regulated by several environmental cues that are sensed by the PhoQ periplasmic domain. Limiting concentrations of the divalent cations Mg2+ and Ca2+ were shown to promote PhoQ autophosphorylation and, in turn, PhoP phosphorylation (Garcia Vescovi et al., 1996; Castelli et al., 2000; Montagne et al., 2001; Sanowar and Le Moual, 2005). Recent studies revealed that PhoQ also responds to α-helical AMPs and acidic pH, leading to the activation of the PhoPQ signalling pathway (Bader et al., 2005; Prost et al., 2007). Acidic pH and α-helical AMPs are likely the physiologically relevant ligands of S. enterica PhoQ, since the vacuolar environment of phagosomes is characterized by the presence of AMPs, a pH in the range of 5.0–6.5 and a constant concentration of Mg2+ around 1 mM (Alpuche Aranda et al., 1992; Martin-Orozco et al., 2006; Prost and Miller, 2008). In agreement, PhoPQ has been shown to be critical for S. enterica resistance to the antimicrobial activity of CRAMP in murine macrophages (Rosenberger et al., 2004). PhoQ-activating signals, such as α-helical AMPs, promote structural modifications of the lipid A moiety of LPS (Miller et al., 2005). PhoP directly regulates the expression of the lipid A-modifying enzymes lpxO, pagL and pagP. The PmrAB TCS regulates expression of other lipid A-modifying enzymes such as the pmrC gene and the pmrHFIJKLM operon (renamed the arn operon). In S. enterica, PhoPQ promotes activation of PmrA-regulated genes by transcriptionally activating the expression of the PmrD protein, which in turn prevents PmrA dephosphorylation (Kox et al., 2000; Kato and Groisman, 2004).

The involvement of PhoPQ in resistance to various AMPs has also been reported for other pathogens of the family Enterobacteriaceae. These include the mammalian pathogens Yersinia pseudotuberculosis and Yersinia pestis (Marceau et al., 2004; Winfield et al., 2005), Shigella flexneri (Moss et al., 2000), as well as the insect pathogen Photorhabdus luminescens (Derzelle et al., 2004) and the plant pathogen Erwinia chrysanthemi (Llama-Palacios et al., 2003). In contrast, Pseudomonas aeruginosa PhoQ is not activated by AMPs but responds to both divalent cations and acidic pH (Prost et al., 2008). This discrepancy was attributed to the absence of a cluster of acidic amino acids from the P. aeruginosa PhoQ periplasmic domain. Thus, a PhoQ response to AMPs can be directly correlated with the presence of this acidic cluster in the PhoQ periplasmic domain (Bader et al., 2005; Cho et al., 2006; Prost and Miller, 2008).

The role of OM proteases in bacterial resistance to AMPs is less well established than that of lipid A modifications. OM proteases of the omptin family are found in various Gram-negative pathogens including S. enterica (PgtE), Y. pestis (Pla), S. flexneri (SopA) and Escherichia coli (OmpT and OmpP) (Kukkonen and Korhonen, 2004; Hritonenko and Stathopoulos, 2007; Haiko et al., 2009a). Although S. enterica PgtE cleaves AMPs such as the synthetic α-helical peptide l-C18G in vitro, its contribution to AMP resistance appears to be marginal when expressed from a single chromosomal copy (Guina et al., 2000). In vivo, Y. pestis Pla activity is essential to tissue invasion (Sodeinde et al., 1992; Lathem et al., 2007). Pla promotes degradation of fibrin clots by activating plasminogen into plasmin and inactivating the α2-antiplasmin inhibitor (Kukkonen and Korhonen, 2004). A recent study revealed that Pla was also able to inactivate AMPs such as LL-37 and CRAMP, in vitro (Galvan et al., 2008). Furthermore, E. coli OmpT was reported to efficiently degrade the AMP protamine (Stumpe et al., 1998).

Citrobacter rodentium is an extracellular enteric pathogen that causes transmissible murine colonic hyperplasia (Schauer and Falkow, 1993). Similar virulence factors make C. rodentium an excellent model organism for the study of the human pathogens enteropathogenic E. coli (EPEC) and enterohaemorrhagic E. coli (EHEC). Colonization of the intestine by these food-borne pathogens produces attaching and effacing (A/E) lesions (Mundy et al., 2005), which are characterized by intimate bacterial attachment to intestinal epithelial cells, formation of actin-rich pedestals underneath adherent bacteria and destruction of brush border microvilli. Host defences against these pathogens include the secretion of AMPs such as β-defensins and cathelicidins (O'Neil et al., 1999; Hase et al., 2002; Schauber et al., 2003). During infection, C. rodentium resides in close proximity to colonic epithelial cells expressing CRAMP, which has antimicrobial activity against this pathogen (Iimura et al., 2005).

Having identified a homologue of the phoPQ operon in the C. rodentium genome sequence, we examined whether C. rodentium PhoQ recognizes α-helical AMPs and responds accordingly by promoting resistance. We present evidence that C. rodentium PhoQ is not activated by α-helical AMPs because they are degraded by CroP, an OM protease of the omptin family, before they reach the periplasmic space. This study shows that C. rodentium and S. enterica use different strategies to resist the deleterious effect of α-helical AMPs.

Results

C. rodentium PhoQ responds to Mg2+ and pH acidification but not to α-helical AMPs

Open reading frames (ORFs) with high identities to the amino acid sequences of S. enterica PhoP (82%) and PhoQ (78%) were identified from the C. rodentium genome sequence (http://www.sanger.ac.uk/Projects/C_rodentium). As in the S. enterica phoPQ operon, the two ORFs overlapped by one nucleotide, suggesting that they are part of the same transcriptional unit. To determine whether the C. rodentium and S. enterica PhoQ sensor kinases respond to the same environmental cues, we constructed chromosomal β-galactosidase fusions with the mgtA gene in both the wild-type and ΔphoPQ C. rodentium strains. As in S. enterica and E. coli, the mgtA gene, which codes for a Mg2+ transporter, harbours a PhoP box (TGTATAxxxxxCGTTTA) that contains all the essential nucleotides (underlined) for PhoP recognition (Yamamoto et al., 2002). To examine the PhoQ response to external Mg2+, the mgtA::lacZ reporter strains were grown under PhoQ-activating (8 μM MgCl2) and PhoQ-repressing (10 mM MgCl2) conditions. The β-galactosidase activity produced by the wild-type strain grown in the presence of 8 μM MgCl2 was 22-fold higher than that of cells grown in the presence of 10 mM MgCl2 (Fig. 1A). Mg2+-mediated regulation of the mgtA::lacZ reporter was not observed in the ΔphoPQ strain. Western blot analysis of whole-cell lysates from wild-type C. rodentium grown in media containing either 8 μM or 10 mM MgCl2 is shown in Fig. 1B. The PhoPQ-inducing condition (8 μM MgCl2) produced elevated levels of PhoP (lane 1), while the expression of PhoP under the PhoPQ-repressing condition (10 mM MgCl2) was drastically repressed (lane 2). As expected, no expression was observed in the ΔphoPQ mutant under either growth condition (lanes 3 and 4). As described previously for S. enterica and E. coli, these data indicate that C. rodentium PhoPQ responds to external Mg2+ and suggest that C. rodentium PhoP autoregulates its own promoter (Soncini et al., 1995; Garcia Vescovi et al., 1996; Kato et al., 1999). Accordingly, a consensus PhoP box that consists of the two half-sites GGTTTA and TGTTTA separated by five nucleotides was identified 58 nucleotides upstream of the C. rodentium phoP gene start codon.

Figure 1.

C. rodentium PhoPQ responds to Mg2+ deprivation and acidic pH but not AMPs.
A. Regulation of the mgtA::lacZ transcriptional fusion in response to external Mg2+. Overnight cultures of the C. rodentium wild-type and ΔphoPQ mutant were diluted 1:100 in N-minimal medium (pH 7.5) containing 10 mM MgCl2 and grown at 37°C. At an OD600 of 0.4, cells were washed twice in N-minimal medium, re-suspended in fresh media containing either 8 μM or 10 mM MgCl2 and incubated at 37°C for an additional 90 min. Values are mean ± standard deviation of cultures grown in triplicate and are representative of at least three independent trials.
B. Expression of C. rodentium PhoP is regulated by external Mg2+. Anti-PhoP immunoblot of crude bacterial extracts isolated from wild-type or ΔphoPQ cells grown in N-minimal medium containing either 8 μM or 10 mM MgCl2.
C. pH acidification upregulates mgtA::lacZ transcriptional activity. Cultures were grown to an OD600 of 0.4 in N-minimal medium (pH 7.5) supplemented with 1 mM MgCl2. Cells were washed and re-suspended in fresh media supplemented with 1 mM MgCl2 and buffered with 100 mM Bis-Tris at either pH 7.5 or pH 5.5. Cells were grown for an additional 90 min at 37°C. No pH fluctuations were observed over the course of the experiment.
D. The mgtA::lacZ transcriptional activity is unresponsive to α-helical AMPs. Bacterial strains were grown in N-minimal medium (pH 7.5) supplemented with 1 mM MgCl2. At an OD600 of 0.4, 2 μM l-C18G or 10 μM CRAMP was added and cultures were incubated at 37°C for an additional 60 min.

Acidic pH was tested as a PhoQ-activating cue, since S. enterica PhoQ has been shown to respond to pH acidification. C. rodentium strains were grown in N-minimal medium containing 1 mM MgCl2, as described previously for S. enterica (Prost et al., 2007). A 3.7-fold increase in mgtA::lacZ gene transcription was observed when wild-type cells were transferred from a medium buffered with Bis-Tris at pH 7.5 to one at pH 5.5 (Fig. 1C). A similar response to pH acidification was obtained using media buffered with MOPS (2.5-fold increase in mgtA::lacZ activation) (data not shown). With both buffer systems, pH was stable over the course of the experiment. Acidic pH did not induce reporter activation in the ΔphoPQ mutant (Fig. 1C), indicating that C. rodentium PhoQ responds to acidic pH, as does its S. enterica homologue. The effect of the α-helical AMPs l-C18G and CRAMP on the expression of the mgtA::lacZ fusions was measured, as S. enterica PhoQ was previously shown to sense these AMPs (Bader et al., 2005). Strikingly, the addition of sublethal concentrations of l-C18G or CRAMP did not affect expression of the mgtA::lacZ fusion in C. rodentium wild-type (Fig. 1D). Expression of a phoP::lacZ transcriptional fusion in C. rodentium wild-type in the absence and presence of these AMPs confirmed this result (data not shown). The apparent unresponsiveness of C. rodentium PhoPQ to α-helical AMPs is in sharp contrast to what was observed for S. enterica PhoPQ under similar experimental conditions (Fig. 3A) (Bader et al., 2005) and might suggest that C. rodentium PhoQ is unable to recognize AMPs.

Figure 3.

C. rodentium PhoQ is activated by α-helical AMPs when expressed in S. enterica. β-Galactosidase activity produced by S. enterica (A) and C. rodentium (B) strains harbouring a chromosomal mgtA::lacZ fusion and complemented or not with the indicated plasmids. Cultures were grown in N-minimal medium containing 1 mM MgCl2 without AMP (black bars) or with 2 μM l-C18G (grey bars) and 10 μM CRAMP (white bars). Values are mean ± standard deviation of cultures grown in triplicate and are representative of at least three independent trials.

C. rodentium PhoPQ is involved in the adaptive response to α-helical AMPs

To further investigate the role of C. rodentium PhoPQ in AMP resistance, we compared the susceptibility of C. rodentium wild-type and ΔphoPQ to the α-helical AMPs l-C18G and CRAMP, and the cyclic lipopeptide polymyxin B (PMB) by determining minimum inhibitory concentration (MIC) values. As shown in Table 1, the C. rodentiumΔphoPQ mutant was more susceptible than wild-type to l-C18G or CRAMP, while complementation of the ΔphoPQ mutant with pCRphoPQ restored, at least partly, resistance to both peptides. Interestingly, we observed that the C. rodentium wild-type and ΔphoPQ mutant strains were equally susceptible to PMB (Table 1), indicating that, in contrast to S. enterica PhoPQ, C. rodentium PhoPQ is not involved in resistance to PMB. Similar results were obtained by disk diffusion assays using increasing amounts of l-C18G, CRAMP and PMB (data not shown). Together, these results indicate that C. rodentium PhoPQ is important for the adaptive response to α-helical AMPs, although C. rodentium PhoQ does not appear to sense these AMPs.

Table 1.  Minimum inhibitory concentrations (MICs) of AMPs for C. rodentium strains.
StrainMIC (μg ml−1)
l-C18Gd-C18GCRAMPPMB
  1. n/d, not determined.

C. rodentium wild-type3241280.5
C. rodentiumΔphoPQ44320.5
C. rodentiumΔphoPQ-pCRphoPQ16n/d128n/d
C. rodentiumΔcroP44320.5
C. rodentiumΔcroP-pCRcroP128n/d> 128n/d

The C. rodentium PhoQ periplasmic domain binds l -C18G

The sequence of the S. enterica PhoQ periplasmic domain (PhoQPeri) contains a cluster of acidic amino acids (EDDDDAE) that has been proposed to be involved in the recognition of both divalent cations and AMPs (Bader et al., 2005; Cho et al., 2006). Alignment of the amino acid sequences of the C. rodentium and S. enterica PhoQ periplasmic domains showed that this acidic cluster is strictly conserved (Fig. 2A). A construct of C. rodentium PhoQPeri, corresponding to residues 45–190 of the periplasmic sensor domain, was expressed in E. coli and purified to homogeneity. PhoQPeri was used to measure the binding of the dansylated derivative of l-C18G (dC18G) by monitoring fluorescence emission, as described previously (Bader et al., 2005). As shown in Fig. 2B, dC18G fluoresces with a λmax of 550 nm (line b). The fluorescence intensity of dC18G increased by approximately threefold upon addition of equimolar amounts of C. rodentium PhoQPeri (line c). In addition, the λmax was blue-shifted from 550 to 520 nm (lines b and c). These changes in fluorescence reflect movement of the dansyl group to a more hydrophobic environment and illustrate that dC18G binds to C. rodentium PhoQPeri. The addition of 10 mM MgCl2 caused a reduction in the fluorescence intensity and a red-shift of the λmax to 537 nm (line e), confirming that Mg2+ effectively competes with dC18G for binding to PhoQPeri. These results are very similar to those obtained previously using the S. enterica PhoQPeri protein (Bader et al., 2005). The ability of C. rodentium PhoQPeri to bind dC18G in combination with its sequence identity to S. enterica strongly suggests that C. rodentium PhoQ has the potential to recognize AMPs.

Figure 2.

The C. rodentium PhoQ periplasmic domain has features required for AMP recognition.
A. Amino acid sequence alignment of the C. rodentium (CR) and S. enterica (ST) PhoQ periplasmic domains. Stars indicate residues that have been proposed to be important for recognition of both divalent cations and AMPs in S. enterica PhoQ.
B. dC18G interacts with the C. rodentium PhoQ periplasmic domain. dC18G (1 μM) was excited at 340 nm and emission spectra were recorded (b). PhoQPeri (1 μM) was incubated with dC18G in the absence (c) or in the presence of MgCl2 at a concentration of 1 mM (d) or 10 mM (e). The emission spectrum of the protein control (1 μM PhoQPeri) is shown as (a). AU, arbitrary units.

C. rodentium PhoPQ responds to AMPs when expressed in S. enterica

To further demonstrate that C. rodentium PhoQ can respond to AMPs, we complemented the ΔphoPQ mutant strains of S. enterica and C. rodentium with plasmids pCRphoPQ and pSTphoPQ respectively. The activation of PhoPQ was measured by monitoring the β-galactosidase activity of the mgtA::lacZ fusion. Interestingly, in the context of the S. entericaΔphoPQ strain, complementation with C. rodentium phoPQ mediated a 2.4- and 1.8-fold increase in β-galactosidase activity in the presence of l-C18G and CRAMP respectively (Fig. 3A). This level of activation was comparable to that obtained by complementation of the S. entericaΔphoPQ strain with its native phoPQ operon (Fig. 3A). Conversely, S. enterica phoPQ introduced in the C. rodentiumΔphoPQ mutant was unable to activate transcription of the mgtA::lacZ fusion in response to either l-C18G or CRAMP (Fig. 3B). These data clearly show that the differential regulation of mgtA::lacZ in response to α-helical AMPs between C. rodentium and S. enterica is not due to differences in the respective PhoPQ systems.

Disruption of the C. rodentium OM by AMPs

Initially, AMPs interact with LPS and then penetrate the OM by self-promoted uptake to access the cytoplasmic membrane (Hancock and Lehrer, 1998). NPN (1-N-phenylnaphthylamine), which fluoresces in the hydrophobic environment of damaged membranes, was used as a probe to measure OM disruption induced by l-C18G, CRAMP or PMB (Loh et al., 1984). Addition of PMB to wild-type C. rodentium caused a rapid increase in fluorescence intensity that reached a plateau after 30 s (Fig. 4A). In contrast, both l-C18G and CRAMP had a moderate effect on the integrity of the C. rodentium OM. l-C18G caused a biphasic disruption of the C. rodentium OM characterized by a slight burst followed by a sustained decrease in fluorescence intensity (Fig. 4A). In contrast, the OM of the C. rodentiumΔphoPQ mutant was disrupted to a much greater extent than that of wild-type upon addition of either l-C18G or CRAMP (Fig. 4B). There was no significant difference between the C. rodentium wild-type and ΔphoPQ strains in the OM disruption produced by PMB (Fig. 4A and B). Together, these results are in good agreement with the MIC values obtained for the C. rodentium wild-type and ΔphoPQ strains (Table 1). These data show that α-helical AMPs barely disrupt the OM of C. rodentium wild-type, and suggest that these AMPs may not be able to reach the periplasmic space and activate PhoQ. In addition, they confirm that C. rodentium PhoPQ controls resistance to α-helical AMPs.

Figure 4.

PhoPQ protects the C. rodentium OM from disruption by α-helical AMPs. C. rodentium wild-type (A) and ΔphoPQ (B) cells were exposed to either 2 μM l-C18G, 10 μM CRAMP or 0.7 μM PMB. The OM uptake of NPN was monitored by measuring fluorescence over time. NPN and AMPs were added 30 s and 2 min, respectively, after the beginning of the experiment. Each experiment is representative of at least three independent trials. AU, arbitrary units.

Identification of a C. rodentium OM protease involved in AMP resistance

The OM protein fractions of the C. rodentium wild-type and ΔphoPQ strains were analysed by SDS-PAGE. As shown in Fig. 5A, the protein profiles were essentially similar with the exception of one band having an approximate molecular weight of 33 kDa that was absent from the ΔphoPQ mutant. Coomassie-stained bands were cut from lanes 1 and 2, trypsin-digested and submitted to liquid chromatography electrospray ionization tandem mass spectrometry (LC-MS/MS). Mass fingerprinting identified the protein migrating at 33 kDa in the wild-type strain as a homologue of the E. coli OM protease OmpP. A blast search of the C. rodentium genome identified an ORF coding for an OM protease of the omptin family (Accession No. ROD20151), hereafter named CroP (C. rodentiumouter-membrane protease). The protein sequence of the croP gene shared 73% and 74% amino acid sequence identity with E. coli OmpP and OmpT, respectively, and 40% sequence identity with S. enterica PgtE. The CroP sequence contained the conserved residue pairs (Asp-103–Asp-105 and Asp-230–His-232) that constitute the omptin active site (Hritonenko and Stathopoulos, 2007).

Figure 5.

Identification and stereospecificity of the CroP OM protease.
A. SDS-PAGE of OM protein fractions isolated from C. rodentium wild-type (lane 1) and ΔphoPQ (lane 2) cultures grown in LB for 6 h. The arrow indicates the band corresponding to the CroP OM protease identified by mass fingerprinting.
B. Disruption of the C. rodentium OM in response to l- and d-C18G. Wild-type C. rodentium cells were exposed to 2 μM of either l- or d-C18G and NPN fluorescence was measured over time. Each experiment is representative of at least three independent trials. AU, arbitrary units.

We hypothesized that if CroP was a functional OM protease degrading α-helical AMPs, it would display stereospecificity with respect to its substrates. To test our hypothesis, the enantiomer of l-C18G (d-C18G) was synthesized from all d-amino acids. As shown in Table 1, wild-type C. rodentium was eightfold more susceptible to d-C18G than l-C18G, suggesting that d-C18G is not degraded by CroP. Furthermore, the ΔphoPQ mutant exhibited similar MIC values for l-C18G and d-C18G, suggesting that PhoPQ controls croP expression and/or CroP activity. In addition, d-C18G caused a rapid increase in NPN fluorescence, indicating extensive damage to the wild-type C. rodentium OM (Fig. 5B). Altogether, our results are indicative of a stereospecific mechanism of AMP resistance that is consistent with the enzymatic activity of the CroP OM protease. A ΔcroP deletion mutant was generated to better define the role of CroP in resistance to α-helical AMPs.

PhoPQ regulates croP expression and the presence of CroP at the OM

To determine whether PhoP regulates expression of the croP gene at the transcriptional level, RNA was isolated from the C. rodentium wild-type, ΔphoPQ and ΔcroP strains grown under PhoPQ-inducing conditions. Real-time quantitative PCR (qPCR) was performed using the 16S rRNA gene as the reference gene for normalization. Compared with the wild-type strain, transcription of croP was reduced by 184- and 33 000-fold in the ΔphoPQ and ΔcroP strains respectively. These results indicate that croP expression is partly, but not entirely, under the control of PhoP. The absence of a prototypical PhoP box in the croP promoter may indicate that PhoPQ-mediated regulation of croP is indirect. Subsequently, the presence of the CroP protein at the OM was investigated by isolating OM protein fractions from strains grown under PhoPQ-inducing or PhoPQ-repressing conditions. Wild-type and complemented ΔcroP mutant strains grown in the presence of 8 μM MgCl2 showed a Coomassie-stained band corresponding to CroP (Fig. 6). This band was undetectable for cells grown in the presence of 10 mM MgCl2. The presence of active CroP at the OM was further demonstrated by the appearance of an OmpA degradation product (OmpA*) that was identified by mass fingerprinting (Fig. 6). Both the CroP and OmpA* bands were absent from OM protein fractions isolated from the ΔcroP mutant (Fig. 6). Additional bands missing from OM fractions of cells grown under PhoPQ-repressing conditions may correspond to CroP conformational isoforms or autoprocessed forms (Haiko et al., 2009b). Together these data suggest that PhoPQ regulates both expression of croP at the transcriptional level and the presence of CroP at the OM at a post-transcriptional level. The latter regulation level is consistent with the previous observation that localization of S. enterica PgtE to the OM is PhoPQ-dependent (Guina et al., 2000).

Figure 6.

External Mg2+ regulates the presence of CroP at the OM. OM protein fractions of C. rodentium wild-type, ΔcroP and ΔcroP complemented strains grown under PhoPQ-inducing (8 μM MgCl2) or PhoPQ-repressing conditions (10 mM MgCl2) were analysed by SDS-PAGE. Bands corresponding to the CroP and OmpA proteins are indicated. OmpA* corresponds to an OmpA degradation product, as determined by mass fingerprinting.

CroP contributes to the protection of the OM from AMP damage

Minimum inhibitory concentration values were determined for the C. rodentiumΔcroP mutant, which was found to be as susceptible as the ΔphoPQ mutant to l-C18G and CRAMP (Table 1). Complementation of the ΔcroP mutant with plasmid pCRcroP restored resistance to l-C18G and CRAMP to a greater extent than wild-type (Table 1). Similar results were obtained by disk diffusion assays (data not shown). In agreement with these results, l-C18G and CRAMP were able to disrupt the OM of the ΔcroP mutant, as illustrated by the increase in NPN fluorescence that follows the addition of these AMPs (Fig. 7A). Interestingly, the addition of l-C18G or CRAMP to the ΔcroP mutant complemented with pCRcroP produced the same baseline signal as the control without AMP, suggesting complete proteolytic inactivation of AMPs (Fig. 7B). As shown in Fig. 7A and B, the addition of PMB to both the ΔcroP and the pCRcroP-complemented strains caused the same rapid increase in fluorescence intensity as the wild-type strain (Fig. 4A), indicating that PMB is not a CroP substrate. Taken together, these results show that CroP plays a crucial role in resistance to α-helical AMPs, likely through its proteolytic activity.

Figure 7.

Involvement of the CroP OM protease in resistance to α-helical AMPs. OM disruption of the C. rodentiumΔcroP (A) and ΔcroP complemented (B) strains in the presence of AMPs. Bacterial cells were exposed to either 2 μM l-C18G, 10 μM CRAMP or 0.7 μM PMB and NPN fluorescence was measured over time. Each experiment is representative of at least three independent trials.

CroP degrades AMPs

To verify that CroP was directly responsible for the proteolytic inactivation of AMPs, cleavage of l-C18G and CRAMP was assayed using the wild-type, ΔphoPQ, ΔcroP and complemented ΔcroP strains. As shown in Fig. 8, l-C18G was completely degraded when incubated for 30 min with wild-type C. rodentium cells (lane 2). In sharp contrast, no degradation was observed when l-C18G was incubated with either ΔphoPQ or ΔcroP mutant cells (lanes 3 and 4). As expected, complementation of the ΔcroP mutant with pCRcroP led to the complete degradation of l-C18G (lane 5). Similar results were obtained using CRAMP as a CroP substrate (Fig. 8B). The results shown in Fig. 5B and Table 1 correlate with d-C18G not being degraded when incubated with C. rodentium wild-type cells (Fig. 8C). These data clearly show that the proteolytic activity of CroP, present at the C. rodentium OM, is responsible for the degradation of α-helical AMPs.

Figure 8.

Proteolytic degradation of α-helical AMPs by C. rodentium CroP.
A and B. Degradation of l-C18G (A) and CRAMP (B) by C. rodentium strains. l-C18G or CRAMP (10 μg) was incubated for 30 min at 37°C in the absence of bacterial cells (lanes 1) or in the presence of the C. rodentium wild-type (lanes 2), ΔphoPQ (lanes 3), ΔcroP (lanes 4) and ΔcroP complemented (lanes 5) strains. Aliquots were analysed by Tris-Tricine SDS-PAGE and subsequently stained with Coomassie blue.
C. d-C18G is not degraded by CroP. l-C18G or d-C18G (10 μg) was incubated for 30 min at 37°C in the absence of bacterial cells (lanes 1) or in the presence of C. rodentium wild-type cells (lanes 2).

CroP inhibits PhoQ recognition of AMPs

To ascertain whether AMPs could activate PhoPQ in the context of the C. rodentiumΔcroP mutant, we assayed the expression of mgtA::lacZ and phoP::lacZ in the presence of l-C18G and CRAMP. As shown in Fig. 9, the addition of l-C18G and CRAMP induced a 2.5- and 2.1-fold increase in mgtA::lacZ activity and a 1.8- and 1.9-fold increase in phoP::lacZ activity respectively. To further confirm this result, qPCR was performed on RNA samples isolated from the C. rodentiumΔcroP mutant exposed or not to l-C18G for 15 min. The presence of l-C18G enhanced expression of the mgtA and phoP genes by 12- and 5-fold respectively. Thus, α-helical AMPs activate C. rodentium PhoQ in the absence of CroP. These data demonstrate that the presence of CroP at the C. rodentium OM is responsible for the unresponsiveness of PhoQ to α-helical AMPs, which are degraded before they reach the periplasmic space and activate PhoQ.

Figure 9.

C. rodentium PhoPQ responds to α-helical AMPs in the ΔcroP mutant. β-Galactosidase activity from chromosomal mgtA::lacZ (A) and phoP::lacZ (B) transcriptional fusions expressed by C. rodentiumΔcroP strains. Cells were grown in N-minimal medium supplemented with 1 mM MgCl2 in the absence or presence of either 2 μM C18G or 10 μM CRAMP. Values are mean ± standard deviation of cultures grown in triplicate and are representative of at least three independent trials.

Discussion

Antimicrobial peptides are an important means of defence against bacterial pathogens. The mouse colonic environment inhabited by C. rodentium contains CRAMP, an α-helical AMP that has been shown to have antimicrobial activity against this extracellular pathogen (Iimura et al., 2005). PhoQ is a chief sensor of the host environment that responds to divalent cations, pH and AMPs in S. enterica. It remains unclear whether all PhoQ homologues respond to the same cues. To determine whether C. rodentium PhoPQ plays a role in resistance to AMPs, including CRAMP, we characterized this TCS and compared it with its homologue in S. enterica. The activities of both C. rodentium and S. enterica PhoPQ systems were repressed by millimolar concentrations of Mg2+ and activated by acidic pH. In addition, both ΔphoPQ mutants showed increased susceptibility to α-helical AMPs. However, our study revealed a striking difference between the C. rodentium and S. enterica PhoPQ systems. The C. rodentium PhoQ sensor was apparently unresponsive to α-helical AMPs such as CRAMP and l-C18G, whereas these AMPs were shown to activate the S. enterica PhoPQ system upon recognition by PhoQ (Bader et al., 2005).

We have found that sublethal concentrations of α-helical AMPs do not activate PhoPQ in C. rodentium wild-type. However, this study provides several lines of evidence showing that C. rodentium PhoQ has all the features needed for recognizing and being activated by AMPs. First, C. rodentium PhoQ possesses the acidic cluster that participates in the binding of both divalent cations and AMPs in S. enterica PhoQ (Fig. 2A) (Bader et al., 2005; Cho et al., 2006). The isolated PhoQ periplasmic sensor domain bound dC18G, much like its S. enterica homologue (Fig. 2B) (Bader et al., 2005). AMPs activated PhoPQ when the C. rodentium phoPQ operon was expressed in a S. entericaΔphoPQ mutant (Fig. 3A). Finally, C. rodentium PhoPQ was activated by AMPs in a ΔcroP mutant (Fig. 9). Thus, the observed difference in response to α-helical AMPs between the C. rodentium and S. enterica PhoQ proteins is attributable to the CroP OM protease in C. rodentium. Our study also provides compelling evidence that CroP is a key player in resistance to α-helical AMPs. First, susceptibility to α-helical AMPs was largely increased in the ΔcroP mutant (Table 1). NPN fluorescence experiments showed that CroP prevents α-helical AMPs from crossing the OM (Fig. 7). Lastly, l-C18G and CRAMP were readily degraded when incubated with bacteria expressing CroP at their OM (Fig. 8). Thus, CroP protects the C. rodentium OM from disruption induced by α-helical AMPs by degrading them before they reach the periplasmic space and activate PhoQ.

The PgtE OM protease of S. enterica shares 40% amino acid sequence identity with CroP, but did not prevent AMPs from reaching the periplasmic space and activating PhoPQ (Fig. 3A). In the study by Guina et al. PgtE appeared to have a minor effect on S. enterica survival in the presence of l-C18G when expressed from a single chromosomal copy. Only when pgtE was overexpressed on a high-copy-number plasmid was susceptibility of S. enterica to l-C18G decreased (Guina et al., 2000). It is a possibility that PgtE is far less efficient than CroP at degrading l-C18G for several reasons. First, the croP and pgtE genes may be expressed at different levels when their respective bacteria are grown in laboratory media. In support of this, previous studies have shown that pgtE expression and PgtE activity were enhanced during the intracellular growth of S. enterica in macrophages (Eriksson et al., 2003; Lahteenmaki et al., 2005). In addition, the highly divergent promoter regions of the croP and pgtE genes suggest that transcription could be regulated by different factors. A second explanation is that the CroP and PgtE OM proteases exhibit different substrate specificities and/or catalytic efficiencies. The fact that the CroP amino acid sequence is only 40% identical to that of PgtE is consistent with the possibility that CroP degrades primarily biologically active AMPs, whereas PgtE preferentially cleaves larger protein substrates (A. Portt, unpublished). Interestingly, our data indicate that CroP does not degrade all AMPs, as the cyclic lipopeptide PMB causes OM disruption regardless of the presence of CroP (Fig. 7). The resistance of PMB to enzymatic proteolysis may be due to peptide cyclization and/or the presence of the unusual amino acid analogue 2,4-diaminobutyric acid.

The important contribution of S. enterica lipid A modifications to AMP resistance has been extensively documented. Strikingly, many genes that encode LPS-modifying enzymes in S. enterica are absent from the C. rodentium genome. Of the S. enterica PhoP-regulated genes lpxO, pagL and pagP, only pagP is present in the C. rodentium genome. The PhoP-regulated pmrD gene, whose gene product connects the S. enterica PhoPQ and PmrAB regulatory pathways, is absent from the C. rodentium genome. In addition, the PmrA-regulated pmrHFIJKLM operon is missing from the C. rodentium genome. This operon is responsible for the addition of 4-aminoarabinose to lipid A, a modification that is essential to PMB resistance (Gunn et al., 1998). These findings are consistent with PhoPQ not playing a role in PMB resistance (Table 1). Although it is possible that LPS-modifying genes are harboured by C. rodentium virulence plasmids, it appears that C. rodentium possesses only the pagP and pmrC genes that are responsible for the transfer of palmitate and the addition of phosphoethanolamine to lipid A respectively. This limited number of lipid A modifications may not allow C. rodentium to strengthen the permeability barrier of its OM to an extent compatible with AMP resistance. These observations suggest that C. rodentium may rely primarily on the CroP OM protease to resist AMPs and that LPS modifications play a secondary role. Thus, C. rodentium and S. enterica appear to use different strategies to resist AMPs. Nonetheless, PhoPQ remains a central regulator of AMP resistance for both organisms.

In summary, this work provides an alternative mechanism by which the extracellular enteric pathogen, C. rodentium, resists AMPs. This mechanism relies upon the proteolytic inactivation of AMPs by an OM protease of the omptin family.

Experimental procedures

Bacterial strains and growth conditions

Bacterial strains and plasmids used in this study are listed in Table S1. Bacterial cultures were grown at 37°C in Luria–Bertani (LB) broth, Tryptic Soy Broth (TSB) or N-minimal medium supplemented with 0.2% glucose (Nelson and Kennedy, 1971). When appropriate, media were supplemented with the following antibiotics: ampicillin (100 μg ml−1), chloramphenicol (30 μg ml−1), kanamycin (50 μg ml−1), streptomycin (30 μg ml−1) and gentamicin (20 μg ml−1).

Construction of C. rodentium deletion mutants

Plasmid purification, cloning and transformation were performed according to standard procedures (Sambrook et al., 1989). All oligonucleotide primers are listed in Table S2. The C. rodentiumΔphoPQ and ΔcroP deletion mutants were generated by sacB gene-based allelic exchange (Donnenberg and Kaper, 1991). The upstream and downstream sequences of the phoPQ operon were PCR-amplified from C. rodentium genomic DNA using primers CR400, CR401, CR402 and CR403. To generate plasmid pCR001, a three-way ligation was performed with the two digested fragments (XbaI–NdeI and NdeI–SacI) and plasmid pRE118 previously cleaved with XbaI and SacI. The constructed plasmid pCR001 was confirmed by sequencing and transformed by electroporation into C. rodentium DBS100. Transformants were grown on LB agar with kanamycin to select for plasmid insertion into the chromosome of C. rodentium. The 5′ and 3′ flanking regions of croP were amplified using primers CR636, CR637, CR638 and CR639. A two-step ligation was used to insert the digested fragments (XbaI–EcoRV and EcoRV–SacI) into the XbaI and SacI sites of pRE112. The resulting plasmid, pCR002, was transformed into the E. coli donor strain χ7213 and conjugated into the C. rodentium wild-type strain. Plasmid insertion was selected for using LB agar supplemented with chloramphenicol. Clones of each deletion mutant were grown on peptone agar containing 2% sucrose to select for sucrose-resistant colonies that were either Kans or Cms, indicating that allelic exchange resulted in the loss of the wild-type copy of phoPQ and croP, respectively, along with the plasmid vehicle. Insertions were confirmed by PCR and sequencing.

Plasmid construction

Plasmid pCRphoPQ was constructed by amplifying the phoPQ operon and its promoter from C. rodentium genomic DNA using Pfx DNA polymerase (Invitrogen) and primers CR541 and CR430. The resulting PCR product was cloned into the XbaI and BamHI restriction sites of plasmid pWSK129. Similarly, pSTphoPQ was obtained by cloning the PCR-amplified S. enterica phoPQ operon and its promoter (primers ST565 and ST564) into pWSK129 previously digested with XbaI and BamHI. The C. rodentium croP gene and its promoter were PCR-amplified using primers CR634 and CR639 and cloned into the XbaI and SacI sites of pWSK129, generating plasmid pCRcroP.

Construction of chromosomal lacZ transcriptional fusions and β-galactosidase assay

Chromosomal transcriptional fusions between the mgtA or phoP promoters and the lacZ reporter gene were generated in C. rodentium and S. enterica using the suicide vector pFUSE (Baümler et al., 1996). The mgtA::lacZ fusion was constructed by PCR amplifying the mgtA promoter using C. rodentium genomic DNA and primers CR542 and CR543. The PCR product was digested with XbaI and SmaI and cloned into the corresponding sites of pFUSE. The resulting construct was transferred into the C. rodentium wild-type, ΔphoPQ and ΔcroP strains by conjugation and integrated by homologous recombination, as previously described (Daigle et al., 2001). A similar strategy was used to integrate the phoP::lacZ fusion (primers CR540 and CR541) into the C. rodentium wild-type and ΔcroP strains. The S. enterica mgtA::lacZ fusion (primers ST459 and ST458) was integrated into the chromosome of both S. enterica wild-type and ΔphoPQ. β-Galactosidase activity assays were performed in triplicate, as previously described (Miller, 1972).

MIC determination and disk diffusion assay

CRAMP (GLLRKGGEKIGEKLKKIGQKIKNFFQKLVPQPEQ) and l-C18G (ALYKKLLKKLLKSAKKLG) were synthesized with a purity of > 85% (BioChemia). d-C18G, the all d-amino acid enantiomer of l-C18G, was synthesized at the Sheldon Biotechnology Centre, McGill University. Dansylated C18G was a gift from S.I. Miller (University of Washington). PMB was purchased from Sigma. MICs were determined by the broth microdilution method in 96-well microtitre plates, as previously described (Wiegand et al., 2008). Bacterial cultures were grown in TSB and diluted to 5 × 105 cfu ml−1 in N-minimal medium containing 1 mM MgCl2. Serial dilutions of AMPs were added and plates were incubated at 37°C for 24 h. The lowest concentration of AMP that completely inhibited growth was identified as the MIC. For disk diffusion assays, aliquots (80 μl) of overnight cultures were inoculated into 1% agarose (20 ml) and poured into 15 cm Petri dishes containing LB agar. Disks containing increasing amounts of AMP were layered on top of the agarose and incubated overnight at 37°C.

Cloning, expression and purification of the C. rodentium PhoQ periplasmic domain

The C. rodentium PhoQ periplasmic domain (PhoQPeri) was amplified by PCR using primers CR562 and CR563 from genomic DNA. The PCR fragment was digested with BamHI and XbaI and ligated into plasmid pET11a digested with the same enzymes to generate plasmid pCRQPeri. The C. rodentium PhoQPeri protein was expressed in E. coli ArcticExpress™ (DE3)RIL cells (Stratagene) according to the manufacturer's instructions. Cells were harvested and re-suspended in 20 mM sodium phosphate buffer (pH 7.5) containing 500 mM NaCl, 20 mM imidazole and PMSF (17 μg ml−1). Cells were disrupted by sonication, centrifuged at 216 000 g for 30 min and the supernatant was applied to a Ni2+-NTA affinity chromatography column (GE Healthcare). Purified protein was dialysed against 20 mM Tris-HCl (pH 7.5) containing 150 mM NaCl. Protein concentration was determined using the BCA protein assay (Pierce).

Fluorescence spectroscopy

Peptide binding assays were performed as previously described (Bader et al., 2005). PhoQPeri at a final concentration of 1 μM was incubated with 1 μM dansylated C18G (dC18G) in the absence or presence of MgCl2 for 20 min at room temperature. Excitation of dC18G was at 340 nm and fluorescence emission spectra were recorded from 400 to 650 nm using 10 nm slit widths. Each spectrum was the mean of 10 consecutive scans.

OM disruption assay

Bacterial cells were grown to an OD600 of 0.5–0.6 in N-minimal medium containing 1 mM MgCl2. Cells were diluted to an OD600 of 0.37 with 5 mM HEPES (pH 7.5) and transferred into a quartz cuvette equipped with a stir bar. NPN (Sigma) was added 30 s after the start of the experiment at a concentration of 5 μM and AMPs were added 90 s later. Samples were excited at 350 nm and emitted fluorescence was recorded over time at 420 nm using 5 nm slit widths.

OM protein extraction

Outer membrane protein fractions were isolated as previously described (Hernandez-Alles et al., 1999). Briefly, cells were collected by centrifugation, re-suspended in 10 mM Tris-HCl (pH 7.5) containing 5 mM MgCl2 and disrupted by sonication. Unbroken cells were removed by centrifugation at 3300 g for 10 min. Cell membranes were collected by high-speed centrifugation at 100 000 g for 1 h. The pellet was re-suspended in 10 mM Tris-HCl (pH 7.5), 5 mM MgCl2 containing 2% sodium lauryl sarcosinate and incubated for 30 min at 25°C. OM proteins were collected by centrifugation at 100 000 g for 1 h. Following a second treatment with sodium lauryl sarcosinate, OM proteins were re-suspended in SDS sample buffer before being resolved on a 12.5% SDS-PAGE gel and stained with Coomassie blue. In-gel trypsin digestion, mass spectrometry and analysis by MASCOT software were carried out at the McGill University and Génome Québec Innovation Centre, Montreal, Canada.

Real-time quantitative RT-PCR (qPCR)

Bacterial strains were grown to an OD600 of 0.5 in N-minimal medium supplemented with 1 mM MgCl2. Total RNA was isolated using TRIzol reagents (Invitrogen) and treated with the DNA-free kit (Ambion) to remove any trace of DNA. The absence of contaminating DNA was confirmed by qPCR using primers CR712 and CR713 (Table S2). RNA (1 μg) was reverse-transcribed using Superscript II (Invitrogen) with 0.5 μg of random hexamer primers (Sigma). As a negative control, a reaction without Superscript II was also included (NRT). qPCR reactions were performed in a Rotor-Gene 3000 thermal cycler (Corbett Research) by using the QuantiTect SYBR Green PCR kit (Qiagen), according to manufacturer's instructions. Primers used are listed in Table S2. The transcriptional level of the genes of interest under each condition was normalized against the reference gene (16S rRNA) and analysed by applying the 2−ΔΔCT method (Livak and Schmittgen, 2001). For each condition, reverse transcription was performed three times independently, and the NRT sample was used as a negative control.

Proteolytic cleavage of AMPs by CroP

Bacterial cells were grown to an OD600 of 0.5–0.6 in N-minimal medium with 1 mM MgCl2. Culture aliquots (20 μl) were incubated with 10 μg of AMP for 30 min at 37°C in a total volume of 25 μl. Bacterial cells were pelleted by centrifugation. The supernatant was removed and an equal volume of Tricine sample buffer (2×) was added. Aliquots (10 μl) were analysed by Tris-Tricine SDS-PAGE (10–20% acrylamide, Bio-Rad) and Coomassie staining.

Western blotting

Wild-type C. rodentium cells were grown at 37°C to an OD600 of 0.8 in N-minimal medium (pH 7.5) containing either 8 μM or 10 mM MgCl2. Cells were harvested and re-suspended in 100 mM sodium phosphate buffer (pH 7.5), 5 mM EDTA and 10% glycerol. Cells were lysed by sonication and cell debris was removed by centrifugation at 100 000 g for 30 min. Equal amounts of whole cell lysates were run on a 10% SDS-PAGE gel and transferred to a nitrocellulose membrane. The blot was developed with an antiserum against S. enterica PhoP (1:1000), an anti-rabbit IgG horseradish peroxidase-linked antibody (1:5000) and the Immobilon Western reagent (Millipore).

Acknowledgements

We are grateful to Samuel I. Miller (University of Washington Medical School, Seattle, WA, USA) for providing dansylated l-C18G. This work was supported by the Canadian Institutes of Health Research (CIHR, MOP-15551) and the Natural Sciences and Engineering Research Council (NSERC, 217482). A.P. was the recipient of an Ontario–Quebec exchange fellowship and a F.C. Harrison fellowship. C.V. was supported by a studentship from Fonds de la Recherche en Santé du Québec (FRSQ).

Ancillary