Complex Response of the CpxAR Two-Component System to β-Lactams on Antibiotic Resistance and Envelope Homeostasis in Enterobacteriaceae

The Cpx stress response is widespread among Enterobacteriaceae. We previously reported a mutation in cpxA in a multidrug-resistant strain of Klebsiella aerogenes isolated from a patient treated with imipenem. This mutation yields a single-amino-acid substitution (Y144N) located in the periplasmic sensor domain of CpxA. In this work, we sought to characterize this mutation in Escherichia coli by using genetic and biochemical approaches.

ABSTRACT

The Cpx stress response is widespread among Enterobacteriaceae. We previously reported a mutation in cpxA in a multidrug-resistant strain of Klebsiella aerogenes isolated from a patient treated with imipenem. This mutation yields a single-amino-acid substitution (Y144N) located in the periplasmic sensor domain of CpxA. In this work, we sought to characterize this mutation in Escherichia coli by using genetic and biochemical approaches. Here, we show that cpxA^Y144N is an activated allele that confers resistance to β-lactams and aminoglycosides in a CpxR-dependent manner, by regulating the expression of the OmpF porin and the AcrD efflux pump, respectively. We also demonstrate the effect of the intimate interconnection between the Cpx system and peptidoglycan integrity on the expression of an exogenous AmpC β-lactamase by using imipenem as a cell wall-active antibiotic or by inactivating penicillin-binding proteins. Moreover, our data indicate that the Y144N substitution abrogates the interaction between CpxA and CpxP and increases phosphotransfer activity on CpxR. Because the addition of a strong AmpC inducer such as imipenem is known to cause abnormal accumulation of muropeptides (disaccharide-pentapeptide and N-acetylglucosamyl-1,6-anhydro-N-acetylmuramyl-l-alanyl-d-glutamy-meso-diaminopimelic-acid-d-alanyl-d-alanine) in the periplasmic space, we propose these molecules activate the Cpx system by displacing CpxP from the sensor domain of CpxA. Altogether, these data could explain why large perturbations to peptidoglycans caused by imipenem lead to mutational activation of the Cpx system and bacterial adaptation through multidrug resistance. These results also validate the Cpx system, in particular, the interaction between CpxA and CpxP, as a promising therapeutic target.

INTRODUCTION

The envelope of Gram-negative bacteria is a complex multilayered structure that is essential for cell viability and provides a protective barrier against environmental stresses. This structure consists of the inner and outer membranes, which are separated by the periplasmic space containing the cell wall or peptidoglycan (PG). PG is a covalently linked scaffold of glycan chains and short peptides that maintains the cell shape and resists osmotic stress (1). It also provides an oxidizing environment where extracytoplasmic proteins can be stabilized by disulfide bonds (2). PG is synthesized and modified throughout the cell cycle by penicillin-binding proteins (PBPs), including high-molecular-weight (HMW) PBPs, which polymerize and cross-link the glycan chains, and low-molecular-weight (LMW) PBPs, often referred to as PG hydrolases, which remodel existing chains (1). In Escherichia coli and closely related Enterobacteriaceae, the majority of LMW PBPs are d,d-carboxypeptidases—PBP5, PBP6, and DacD that remove the terminal d-alanine from the pentapeptide chains—and/or endopeptidases—PBP4, PBP7, and AmpH that cleave the peptide side chains and disconnect the glycan polymers (1). While HWM PBPs are essential (3), LMW PBPs are dispensable under standard laboratory conditions (4).

Bacteria have evolved sophisticated strategies to monitor and maintain the integrity of their envelope. In E. coli, several signal transduction pathways sense envelope alterations in the periplasm and control the expression of adaptive genes (5–9). One of these pathways is Cpx, a two-component system that consists of the transmembrane sensor kinase CpxA and the cytoplasmic response regulator CpxR. Although the molecular characteristics of the inducing signal(s) are unknown, activation of CpxA leads to autophosphorylation at a conserved histidine residue and phosphate transfer to a conserved aspartate in CpxR, which then remodels gene expression (8, 10–12). Divergently transcribed from cpxAR is cpxP, which encodes a periplasmic protein that negatively modulates the activity of the Cpx system by interacting with the periplasmic sensor domain (SD) of CpxA (13–15). The Cpx system was initially recognized as mediating the adaptation to protein misfolding in the periplasm, since the first identified genes of the Cpx regulon encoded envelope protein folding and degradation factors such as the protease/chaperone DegP, the peptidyl-prolyl isomerase PpiA, and the disulfide oxidase DsbA (11, 16–18). Recent studies have linked the Cpx system to a wider range of other signals and adaptations (19–24), including the detection and repair of perturbations to the PG. First, the Cpx response is activated by the simultaneous deletion of LMW PBPs, resulting in the downregulation of flagellar genes and mobility defects (23). Second, genes of the Cpx regulon are upregulated in the presence of cell wall-active antibiotics such as β-lactams (24). Third, Cpx controls the expression of PG-modifying enzymes such as the LdtD (YcbB) transpeptidase (10, 21, 22), which catalyzes unusual 3→3 cross-links and mediates resistance to β-lactams (25).

We previously reported a cpxA mutation in a clinical strain of Klebsiella aerogenes isolated from a patient treated with imipenem, a potent β-lactam antibiotic. This strain presented high-level β-lactam resistance associated with the loss of outer membrane porins and an increased β-lactamase activity (26). This mutation yields a single-amino-acid substitution (Y144N) located in CpxA-SD. In this work, we characterized CpxA^Y144N as a gain-of-function CpxA* mutant. Biochemical analyses showed that this mutation abrogates the interaction between CpxA-SD and CpxP that normally keeps the system in a resting state in the absence of an inducing stress. Accordingly, it increases phosphotransfer events between CpxA and CpxR. Phenotypic analyses showed it conferred resistance to aminoglycosides and β-lactams by acting on identified Cpx regulon members. We also found that cell wall damage is able to induce AmpC in a Cpx-dependent manner, suggesting an intimate interconnection between these two pathways.

RESULTS

cpxA^Y144N is an autoactivated cpxA allele (cpxA*) that confers antibiotic resistance to β-lactams and aminoglycosides.

In a previous study, we reported the evolution of antibiotic resistance in four K. aerogenes strains, which were sequentially isolated during the clinical course of a patient treated with imipenem. Comparative genomics of these isolates (P1 to P4) showed that P4 carried mutations in omp36 (encoding the E. coli OmpC ortholog), leading to a premature stop codon, and in cpxA, producing a Y144N substitution located in CpxA-SD. Interestingly, we observed that P4 did not produce Omp35 (the E. coli OmpF ortholog), although the gene sequence was not altered, and presented increased β-lactamase activity, both which may have contributed to high-level β-lactam resistance in this strain (26). To elucidate the specific effects of the cpxA^Y144N mutation on antibiotic resistance, wild-type cpxA and cpxA^Y144N were amplified from K. aerogenes G7 and P4, respectively, cloned into the pBAD33 plasmid under the control the P_ara arabinose-inducible promoter, and expressed in E. coli K-12 MC4100 Δara174. Several gain-of-function cpxA* mutations were previously isolated and characterized. Among these, cpxA104(R33C) and cpxA101(T253P) were generated by site-directed mutagenesis and used as positive controls (12). The effects of the CpxA mutants as well overexpression of the lipoprotein NlpE, a well-characterized Cpx-inducing cue (27), were tested on a chromosomal ppiA-lacZ fusion whose activity directly depends on the CpxAR system (16, 28). Consistent with previous results, ppiA::lacZ activity increased 2- to 3-fold in response to the overexpression of NlpE, CpxA^R33C, and CpxA^T253P. Similar activation was observed when the CpxA^Y144N but not wild-type CpxA was overexpressed, suggesting that cpxA^Y144N is a new constitutively activated cpxA* allele (Fig. 1).

FIG 1.

cpxA^Y144N is a cpxA* allele. Effects of CpxA, CpxA^Y144N, CpxA^R33C, CpxA^T252P, and NlpE on ppiA expression by assaying β-galactosidase activities of a chromosomal lacZ fusions.

Antibiotic resistance of cpxA* mutants is due to specific activation of CpxR.

Several recent works suggested that CpxAR is involved in antibiotic resistance in laboratory and clinical strains of enterobacteria (10, 21, 29–33). Null mutations of cpxA, lacking both kinase and phosphatase activities, have been shown to activate the Cpx response resulting from the phosphorylation of CpxR by small molecule phosphodonors and to protect cells against external stress caused by hydroxyurea and aminoglycosides (34, 35). Here, we took advantage of the dominant effect of the plasmid-expressed cpxA* mutations to analyze their effect on antibiotic resistance by using an efficiency-of-plating (EOP) assay in a wild-type background. Compared with that of the empty plasmid or wild-type cpxA, cpxA* mutations conferred some resistance to several antibiotic classes, including β-lactams, aminoglycosides, and fosfomycin, but not to fluoroquinolones and macrolides (Fig. 2A and see Fig. S5 in the supplemental material).

FIG 2.

cpxA* mutations confer resistance to β-lactams, aminoglycosides, and fosfomycin. (A) Efficiency-of-plating (EOP) assay on LB agar plates supplemented with 3 μg/ml amikacin (AMK), 2.5 μg/ml gentamicin (GEN), 0.5 μg/ml imipenem (IMP), 0.125 μg/ml ceftazidime (CAZ), or 4 μg/ml fosfomycin (FOF) and grown at 37°C. Susceptibility was determined for the following strains: RAM1292 transformed with pBAD33, pBAD33-cpxA, pBAD33-cpxA^Y144N, pBAD33-cpxA^R33C, or pBAD33-cpxA^T252P. (B) Deletion of cpxR in RAM1292 pBAD33-cpxA^Y144N abolished antibiotic resistance.

cpxA* mutations could cause pleiotropic effects resulting from the phosphorylation of heterologous response regulators other than CpxR. We eliminated this possibility by showing that antibiotic resistance was abolished when a cpxR null mutation was introduced in the presence the of a cpxA^Y144N allele (Fig. 2B). Altogether, these results demonstrate that antibiotic resistance conferred by the Cpx response is due to specific phosphorylation of CpxR and concomitant changes in the expression of genes that belong to the Cpx regulon.

Multiple Cpx regulon members may be responsible for antibiotic resistance.

Having clarified the relationship between the Cpx response and antibiotic resistance, we sought to identify the Cpx regulon members responsible for this phenotype by using a candidate approach. This consists of generating null mutants of genes that are known to be upregulated by Cpx or overexpressing genes that are known to be downregulated by Cpx and then looking for a decrease in resistance in the presence of cpxA*. AcrD is an inner membrane efflux pump that functions with outer membrane TolC to expel aminoglycosides (36). EOP assays showed that disruption of acrD or tolC decreased resistance to aminoglycosides in a cpxA* background, suggesting that resistance to this class of antibiotics is somehow attributable to an elevated expression of acrD (Fig. 3A). Activation of Cpx decreases ompF expression. This occurs directly at the transcriptional level through the binding of phosphorylated CpxR (CpxR∼P) to the ompF promoter and indirectly through the activation of MzrA that connects CpxAR to EnvZ-OmpR (10, 11, 28, 37). We confirmed the effect of Cpx activation on porin expression by Western blotting of whole-cell lysates after transient overproduction of NlpE or CpxA* (Fig. 3B). In addition, EOP assays showed that plasmid expression of Omp35, the OmpF ortholog in K. aerogenes, decreased resistance to β-lactams but not to aminoglycosides in a cpxA* background (Fig. 3C). Interestingly, the replacement of OmpF to OmpC is often associated with an increased resistance to β-lactams in clinical isolates of problematic enteric bacteria such as K. aerogenes and Klebsiella pneumoniae (26, 38, 39). These data suggest that OmpF is the preferred route for the uptake of β-lactams and that the Cpx-dependent inhibition of ompF expression is responsible for resistance to this class of antibiotics. Although not tested here, Kurabayashi and colleagues showed that activation of Cpx downregulated glpT and uhpT, which encode transporters for fosfomycin together with their native substrates, glycerol-3-phosphate and glucose-6-phosphate (31).

FIG 3.

The Cpx response regulates genes that impact antibiotic resistance. (A and C) EOP assays. (A) Deletion of the acrD or tolC efflux pump components decreased resistance to aminoglycosides in RAM1292 pBAD33-cpxA^Y144N. (B) OMP expression level is altered by cpxA^Y144N. RAM1292 transformed with pBAD33, pBAD33-cpxA, pBAD33-cpxA^Y144N, pBAD33-cpxA^R33C, or pBAD33-cpxA^T252P was cultured and induced for 2 h with 0.2% l-arabinose. OMPs were detected from whole-cell lysates using polyclonal antibodies directed against OmpF/C/A as described in Materials and Methods. Equivalent of 0.2 OD₆₀₀ units was loaded per well. OmpF/C levels were normalized using the OmpA intensity signal. (C) Plasmid expression of the OmpF ortholog of K. aerogenes Omp35 restores the susceptibility of RAM1292 pBAD33-cpxA^Y144N to β-lactams.

In vitro and in vivo biochemical characterization of CpxA^Y144N.

Like many bacterial sensor kinases, CpxA possesses three enzymatic activities that are critical for signal transduction: autokinase, response regulator kinase, and phosphorylated response regulator phosphatase activities (8). To compare the in vitro biochemical activities of CpxA to those of CpxA^Y144N, we used inside-out inner membrane vesicles (IMVs) enriched in CpxA-His or CpxA^Y144N-His. Autokinase activity was quantified using the ADP-Glo kinase assay. This assay is performed in two steps. After the kinase reaction in the presence of ATP, the ADP-Glo reagent is added to terminate the kinase reaction and deplete the remaining ATP; then, the kinase detection reagent is added to convert ADP into ATP. Newly synthesized ATP is measured using a luciferase/luciferin reaction. The light generated correlates to the amount of ADP generated in the kinase assay, which is indicative of the kinase activity. When the two proteins were incubated with ATP, similar levels of ADP were generated (Fig. 4A). This indicated that both proteins catalyze autophosphorylation and that the Y144N substitution does not modify the autokinase activity. Kinase activities were tested by using IMVs and purified CpxR-His. Phosphorylated CpxR was visualized by using the Phos-tag PAGE system. CpxR can be phosphorylated in vitro by small phosphodonor molecules such as acetyl phosphate (40) and used as a positive control (Fig. 4B). However, the Y144N substitution did not increase the kinase activity of CpxA on CpxR. Phosphatase activities were tested by using IMVs and phosphorylated CpxR-His. However, similar amounts of phosphate were measured by using a phosphate colorimetric assay (Fig. 4C). Altogether, results from in vitro assays suggest CpxA and CpxA^Y144N behave similarly.

FIG 4.

Enzymatic activities of CpxA^Y144N. (A) CpxA-His and CpxA^Y144N-His exhibit similar autophosphorylation activities. Autophosphorylation was tested by using the ADP-Glo kinase assay on inner membrane fractions as described in Materials and Methods. Reactions were performed with 5 μl IMVs in the presence of 100 μM or 1 mM ATP for 30 min at room temperature. The two-step assay includes the removal of the remaining ATP and then the simultaneous conversion of the ADP produced by the kinase reaction to ATP and into light by a luciferin/luciferase reaction. Bioluminescent signals are proportional to the ADP produced and the activity of the kinase. The amount of ADP produced from each reaction was calculated by using ATP-to-ADP standard curves. (B) CpxA-His and CpxA^Y144N-His exhibit similar in vitro kinase activities on CpxR. CpxA-His or CpxA^Y144N-His in enriched IMVs (5 μl) was phosphorylated under standard conditions and then incubated in the presence of 1.5 μM purified CpxR-His for 30 min. Samples were analyzed by using Phos-tag PAGE fractionation. (C) CpxA-His and CpxA^Y144N-His exhibit similar in vitro phosphatase activities on CpxR. CpxR-His was phosphorylated with acetyl-phosphate as described in Materials and Methods. Phosphorylated CpxR-His was incubated in the presence of CpxA-His- or CpxA^Y144N-His-enriched IMVs (5 μl). Phosphate standards and samples (duplicates) were brought to 200 μl with water in a microplate; 30 μl of phosphate reagent was added to each well, and mixtures were incubated at room temperature in the dark. After 30 min, absorbance at 650 nm was measured. (D) Substitution Y144N in the periplasmic sensor domain of CpxA affects binding to CpxP. Protein-protein interaction analysis was performed by using BATCH. CpxP and the sensor domain of the wild-type CpxA or mutant CpxA^Y144N (CpxA-SD and CpxA^Y144N-SD) were fused to the C- and the N-terminal ends of the T18 and T25 fragments of B. pertussis adenylate cyclase, respectively. E. coli cells DHM1 cotransformed with plasmids encoding the hybrid proteins were grown overnight, spotted on an X-Gal plates, and incubated at 30°C for 24 h. (E) In vivo detection of phosphorylated and nonphosphorylated CpxR. E. coli BL21(DE3) was cotransformed with pET24+-cpxR-His and pBAD33-cpxA or pBAD33-cpxA^Y144N. Cells were grown at 37°C to mid-log phase, and protein expression was induced with 0.2% l-arabinose for 2 h and then with 1 mM IPTG for 3 h. Bacteria were pelleted by centrifugation and lysed with formic acid; samples were rapidly fractionated by Phos-tag PAGE and blotted with anti-His antibodies.

Overproduction of the small periplasmic CpxP protein inhibits the Cpx response and prevents Cpx activation (41). This effect depends on an intact sensor domain (SD), as cpxA24, which encodes a variant of CpxA lacking 32 amino acids in its periplasmic loop, is a strong cpxA* allele (12). It has been shown that CpxP interacts with CpxA-SD in unstressed cells and that CpxP inhibits CpxA autophosphorylation in reconstituted proteoliposomes by direct interaction (13, 15). A peptide array indicated that the C-terminal region of CpxA-SD (E₁₃₈DNYQLYLIRPASSSQSDEINLLFD₁₆₂) might play an important role for the interaction with CpxP (14). Because this region contains Y144 (underlined), we investigated the impact of the Y144N substitution on the ability of CpxA-SD to interact with CpxP by using a bacterial two-hybrid assay (BACTH). BATCH is based on the functional complementation of the T18 and T25 domains of the adenylate cyclase of Bordetella pertussis, resulting in cAMP synthesis and activation of the lactose operon in an appropriate adenylate cyclase-deficient reporter strain (42). A previous study showed an efficient interaction when CpxA-SD (P₂₈-P₁₆₄) and signal sequenceless CpxP were fused to the N or the C termini of T25 and T18 domains (13). Here, E. coli DHM1 was cotransformed with CpxP-T18 and T25-CpxA-SD or T25-CpxA^Y144N-SD, and transformants were tested for their ability to metabolize 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) on supplemented LB plates. Coexpression of CpxP-T18 and T25-CpxA-SD but not T25-CpxA^Y144N-SD resulted in blue colonies (Fig. 4D). This suggests that the Y144N substitution disrupts the interaction with CpxP and likely explains the constitutive activation of cpxA^Y144N in vivo. To confirm this, we analyzed the phosphorylation status of CpxR in intact E. coli BL21(DE3) cotransformed with pBAD33-cpxA or pBAD33-cpxA^Y144N and pET24a+-cpxR-His by using the Phos-tag PAGE system and immunoblotting with anti-His antibodies. Here, phosphorylated CpxR-His was only detected in cells expressing the mutant but not the wild-type CpxA (Fig. 4E).

Interplay between the Cpx system, PBPs, and the cell wall-active antibiotic imipenem for AmpC β-lactamase induction.

Enterobacter and Klebsiella spp. but not E. coli harbor an inducible chromosomal AmpC β-lactamase (43–45). Although AmpC is produced at very low levels in wild-type strains cultured under standard laboratory conditions, its expression is highly inducible in the presence of β-lactams (AmpC inducers), such as cefoxitin and imipenem. The AmpC induction mechanism is complex and involves at least three proteins that are involved in the PG recycling pathway: AmpR (transcriptional regulator of the LysR family), AmpD (a cytoplasmic amidase), and AmpG (an inner membrane permease) (see Fig. 6) (46–52). In the current model and under noninducing conditions, muropeptides from normal PG degradation are removed from the cell wall and transported via AmpG into the cytoplasm, where they are cleaved by AmpD to generate free peptides and recycled for cell wall synthesis. Under inducing conditions, AmpD is unable to process high levels of muropeptides, which interact with AmpR, creating a conformation that activates the transcription of ampC. In clinical strains, high level resistance to β-lactams (especially to third-generation cephalosporins) is due to the derepression of ampC, mainly resulting from ampD mutations (51–54). The results described above show that expression of cpxA* in E. coli provided some resistance to β-lactams but not to the level observed in K. aerogenes P4, suggesting the presence an additional mechanism. Compared to that in strains P1 to P3, K. aerogenes P4 showed an increased (∼8.5-fold) β-lactamase activity under noninducing conditions (Fig. 5A). This activity was attributed to a derepressed AmpC, as it was only inhibited by boronic acid (inhibitor of class C β-lactamases such as AmpC) and not by a mixture of tazobactam and clavulanic acid (inhibitors of class A β-lactamases such as TEM-3) (Fig. 5B). Because a previous whole-genome analysis did not reveal mutations in ampD (26), we examined the effect of the cpxA^Y144N mutation on the expression of AmpC. The ampR-ampC operon from K. aerogenes ATCC 13048 was cloned into the pACYC184 vector and coexpressed with cpxA or cpxA^Y144N in E. coli C41(DE3). The production of CpxA or CpxA^Y144N was induced with isopropyl-β-d-thiogalactopyranoside (IPTG) for 1 h, and then the cells were treated with or without imipenem. AmpC levels were monitored by using a nitrocefin degradation assay. Although the two strains were responsive to imipenem, the expression of CpxA^Y144N resulted in a significant (∼5-fold) increase in AmpC activity in the absence of imipenem. This suggested that the Cpx system could be an alternative pathway to induce the expression of ampC. To test this hypothesis, different genetic backgrounds of E. coli K-12 MC4100 were transformed with pACYC184-ampRC and AmpC levels were analyzed as described above. As expected, AmpC activity was highly induced in the presence of imipenem in a wild-type background (∼12-fold activation) (Fig. 5D). In contrast, the absence of AmpG or CpxAR decreased AmpC induction 3.5- to 4.5-fold and the triple mutant showed no induction at all, suggesting that both the AmpG permease and the Cpx system are required for maximal induction (Fig. 5D).

FIG 6.

Schematic representation of the interplay between the Cpx system and the chromosomal AmpC β-lactamase regulatory pathway in Enterobacteriaceae. In a wild-type background and under noninducing conditions, the PG is recycled. PG degradation products (disaccharide-peptides [aD-peptides]) are generated in the periplasm by the activity of PBPs. aD-peptides (aD-tripeptides, aD-tetrapeptides, and aD-pentapeptides) are transported into the cytoplasm by the inner membrane permease AmpG and processed by the amidase AmpD into their corresponding monosaccharide-peptides (aM-peptides). aM-peptides are then transformed into UDP-MurNAc-pentapeptides and incorporated to the PG biosynthesis pathway to complete the recycling process. In the cytoplasm, these UDP-MurNAc-pentapeptides interact with AmpR, creating a conformation that represses the transcription of ampC. In addition, normal interactions between CpxP and the periplasmic sensor domain of CpxA keeps the Cpx system off. Strong AmpC inducers such as imipenem simultaneously bind to several PBPs, leading to an increase of aD-pentapeptides in the periplasm (AmpG saturation) and aM-tripeptides (AmpD saturation) in the cytoplasm. In the cytoplasm, the accumulated aM-tripeptides displace UDP-MurNAc-pentapeptides from AmpR, creating a conformation that activates the transcription of ampC. In addition, our data showed that the increased level of aD-pentapeptides in the periplasm is sensed by the Cpx system. The release of CpxP from CpxA-SD by competing interactions with aD-pentapeptides probably switches the Cpx system on. High level of CpxR∼P leads to AmpC overexpression and thus to β-lactam resistance. The proposed feedback loop involving the lytic glycosylase appears in red. It also unknown whether CpxA or CpxP interacts with the putative signaling muropeptide in the periplasm. The cpxA^Y144N mutation alters the interaction between CpxA-SD and CpxP. This activates the Cpx response and leads to ampC overexpression, even in the absence of AmpC inducers. This mutation also acts on other genes of the Cpx regulon that impact antibiotic resistance.

FIG 5.

Interplay between the Cpx system, PBPs, and the cell wall-active antibiotic imipenem for β-lactamase induction. (A to D) Nitrocefin degradation assays were performed as described in Materials and Methods. (A) K. aerogenes clinical strain P4 exhibits high β-lactamase activity. Total β-lactamase activity was assessed in a series of clinical strains of K. aerogenes (P1 to P4) sequentially isolated from a patient under treatment with IMP. (B) β-Lactamase activity in P4 is mainly due to AmpC. Inhibition of the β-lactamase activity in K. aerogenes P4 and control strains of E. coli expressing constitutive TEM-3 or AmpC from a plasmid. Tazobactam-clavulanic acid and boronic acid were used as specific inhibitors of class A (e.g., TEM-3) and class C (e.g., AmpC) β-lactamases, respectively. (C) AmpC activity is increased upon the expression of CpxA^Y144N in the absence of IMP. E. coli C41(DE3) was cotransformed with pACYC184-ampRC and pET24a+-cpxA-His or pET24a+-cpxA^Y144N-His. Expression of the CpxA proteins was induced with 1 mM IPTG for 1 h at 30°C, and then cells were exposed or not to IMP (0.32 μg/ml) for 1 h. (D) Functional AmpG and CpxAR are both required to indue AmpC activity in response to IMP exposure. (E) Loss of specific PBPs induces AmpC activity in a Cpx-dependent manner. (D and E) E. coli MC4100, CS109, and derivative mutants were transformed with pACYC184-ampRC, grown, and induced as described in Materials and Methods. **, P < 0.05; ***, P < 0.01 versus wild-type control samples; nd, not determined.

The parental strain E. coli CS109 and the derivative mutant CS448.3 were described by Evans and colleagues in a large study that surveyed the impact of PBPs on bacterial motility (23). Interestingly, the simultaneous removal of PBP4, PBP5, PBP7, and AmpH in CS448.3 compromised migration as a result of the activation of the Cpx stress response, which inhibits the production of flagellar proteins. Figure 5E shows AmpC activity was increased in CS448.3 compared to that in CS109. In addition, inactivation of cpxA in CS448.3 reduced AmpC activity to the level of the parental strain. While ampG is still essential for full AmpC induction, this indicates that activation of Cpx connects PBP loss and AmpC activation. This is also consistent with studies that previously linked these two events (55, 56). Altogether, these results demonstrate that the inactivation of PBPs, either by gene deletion or by inhibitory interactions with imipenem, can be sensed by the Cpx system in the periplasmic space and that Cpx activation provides an adapted response through ampC induction.

DISCUSSION

Cpx, together with Bae, Rcs, σ^E, and Psp, is usually considered the main envelope stress response system in E. coli (9). There is convincing evidence that these systems, especially Cpx, can sense and respond to cell wall damage, while the molecular signals remain mostly unknown. Here, we provide functional characterization of the first gain-of-function mutation of CpxA, which was identified in a multidrug-resistant clinical strain of K. aerogenes (26). We demonstrated that this mutation confers resistance to β-lactams and aminoglycosides by regulating the expression of effector genes involved in their uptake (ompF) or extrusion (acrD). Together with previous studies, these data enlighten the role of the Cpx system in antibiotic resistance in laboratory and clinical strains of enterobacteria (10, 21, 29–33). This mutation also sheds new light on how Cpx might sense and respond to PG damage induced by β-lactam treatment, through ampC regulation and a putative positive feedback loop involving the lytic murein transglycosylase Slt (Fig. 6).

β-Lactams are a potent class of antibiotics used for treating infections caused by Gram-negative bacteria. According to the most widely accepted model, cell wall damage that follows β-lactam treatment results from the drug-induced imbalance of the cell wall biosynthesis machinery, between the murein synthetases and hydrolases. β-Lactams bind to PBPs with different affinities and thus were considered to have different consequences on cell morphology and viability (57). Noteworthy, carbapenems, including imipenem and meropenem, were shown to saturate simultaneously all four essential HMW PBPs (PBP1a, PBP1b, PBP2, and PBP3) as well as LMW PBP4, -5, and -6 at low concentrations, leading to rapid cell killing. In an elegant study, Cho et al. showed that β-lactams (amdinocillin and cephalexin) induce toxic malfunctioning of the cell wall biosynthesis machinery, in which Slt acts a quality control enzyme by degrading uncross-linked glycan chains (58). Interestingly, slt belongs to the Cpx regulon (10).

Previous transcriptomic studies have revealed that β-lactams lead to the expression of genes controlled by envelope stress response systems. Indeed, β-lactams specifically targeting the essential bifunctional PBPs (PBP1a and PBP1b) (cefsulodin) and the monofunctional PBP2 (amdinocillin), used alone or in combination, increased the expression of genes regulated by the Rcs, Cpx, σ^E, and Bae systems. Interestingly, Rcs was the only response that was activated under all tested conditions, suggesting this system preferentially responds to PG damage in comparison to the responses from the others (24). A22, a drug that, like amdinocillin, targets the cell elongasome, specifically activates the Rcs response in an RcsF-dependent manner (59). In addition, amdinocillin or A22 (targeting PBP2 or MreB, both part of the elongasome) and cephalexin (targeting PBP3, part of the divisome) led to a 2-fold increase in Cpx activation (21). Endogenous signals, such as genetic blockade of a critical step in PG synthesis, can also activate envelope stress responses. In E. coli, the simultaneous deletion of specific PBPs (PBP4, -5, and -6 and AmpH) led to a reduction in motility that was dependent on the activation of both the Cpx and Rcs systems (23). Interestingly, the activation of the Rcs system was dependent on the activation of the Cpx system, but not vice versa, suggesting Cpx acts upstream of Rcs.

Many Gram-negative bacteria harbor an inducible chromosomic AmpC β-lactamase, produced in response to specific β-lactam treatments, such as imipenem or cefoxitin. In enterobacteria, expression of ampC is controlled by the transcriptional regulator AmpR (43–45, 48). Mutants defective for the cytoplasmic AmpD amidase involved in normal cell wall recycling overexpress ampC even in the absence of inducer (46, 51–54). This and other findings have led to the conclusion that β-lactam treatment generates major cell wall damage resulting in the accumulation of anhydro-disaccharide-pentapeptide in the periplasm, which is transported into the cytoplasm by the AmpG inner membrane permease and converted to anhydro-monosaccharide-pentapeptide, serving as a signal for ampC induction (46–52). Cho et al. showed that Slt is responsible for the degradation of uncross-linked nascent PG into anhydro-disaccharide-pentapeptide when cells are treated with β-lactams (58, 60). Moreover, previous studies have shown the importance of Slt on ampC induction (61, 62) and the hypersensitivity of slt-defective mutants to β-lactams (63). Bridging all these data together makes the Cpx system emerge as a central player in bacterial physiology upon β-lactam treatment. Cpx first senses PG end products arising from the concomitant inactivation of PBPs and degradation by Slt; Cpx tunes an adapted response by modulating the expression of a number of cell wall-modifying enzymes. Among these are Slt and LdtD (10, 22). LdtD is one of the five l,d-transpeptidases of E. coli that catalyze noncanonical 3→3 peptide cross-links. The Cpx-dependent or -independent expression of ldtD was later shown to substantially increase resistance to β-lactams (21, 25). Although the inactivation of ldtD in the cpxA^Y144N background did not significantly affect bacterial resistance to β-lactams in EOP assays, we cannot rule out that ldtD expression was altered. In the near future, analyses of the PG composition in untreated and imipenem-treated wild-type and Δslt cells are necessary to investigate whether β-lactams that induce large and small amounts of AmpC β-lactamase produce the same PG turnover. We have observed that imipenem is able to activate the Cpx response by using a specific ppiA-lacZ reporter fusion (see Fig. S6 in the supplemental material). Experiments are now needed to investigate a feedback mechanism in vivo by using slt-defective and -overexpressing cells. Additionally, an in vitro assay will be valuable to test Cpx activation and identify the nature of the signaling muropeptide(s).

According to the proposed membrane topology of CpxA, amino acid Y144 is located in its periplasmic SD (12). We have shown that this mutation produced constitutively phosphorylated CpxR. This most likely results from the loss of the periplasmic interaction between CpxA and CpxP, an accessory inhibitory protein, which holds the system in a resting state in the absence of stress (13–15, 41). One can suspect that imipenem therapy can select gain-of-function cpxA* mutations that mimic the presence of an inducing cue. It is also possible that activation of the Cpx system is maintained in this background via the Slt feedback loop. It has been proposed that Slt inhibitors may be used in combination therapies with β-lactams (62). Similarly, inhibiting Cpx function would be an effective way to keep AmpC β-lactamases silent and preserve the activity of potent β-lactams against clinically problematic enterobacteria.

MATERIALS AND METHODS

Strains, media, and plasmids.

Bacterial strains and plasmids used this study are listed in the table in the supplemental material. Unless otherwise specified, all assays were performed in the parental strain E. coli MC4100 (64). Single-gene-deleted strains were obtained from the KEIO collection (65) and provided by GE Healthcare Dharmacon. Deletions marked with kanamycin resistance cassettes flanked by FLP recombination target (FRT) sites were introduced into MC4100 by P1 transduction and cured by using the FLP helper plasmid pCP20 for serial deletions (66). Genomic DNA of appropriate bacterial strains was prepared with the Wizard Genomic DNA purification kit (Promega) according to the manufacturer’s protocol. Construction of plasmids was performed by using the In-Fusion HD Cloning kit (Clontech) according to the manufacturer’s instructions. Point mutations in the cpxA gene were generated with the QuikChange II XL site-directed mutagenesis kit (Stratagene) according to the manufacturer’s protocol. Oligonucleotides were provided by MWG Eurofins. Cloned DNA inserts were sequenced to confirm the presence of engineered mutations and the absence of other PCR-generated mutations (MWG Eurofins). Bacteria were routinely grown at 30 or 37°C in Luria-Bertani (LB) broth or agar (Difco). When needed, ampicillin (100 μg/ml), chloramphenicol (30 μg/ml), kanamycin (50 μg/ml), l-arabinose (0.02 to 0.2%), or IPTG (1 mM) was added. All chemicals were purchased from Sigma.

β-Galactosidase assays.

β-Galactosidase activity was assayed on late-log-phase bacterial cultures as described by Miller (67). Experiments were independently repeated at least three times.

Antibiotic sensitivity assay.

Strain susceptibility was determined by efficiency of plating (EOP). Strains containing the empty pBAD33 or cpxA derivative plasmids were first induced for 2 h at 37°C with l-arabinose and then serially diluted and plated onto LB agar plates supplemented with antibiotics. The aminoglycosides used were amikacin (AMK; 3 μg/ml) and gentamicin (GEN; 5 μg/ml), β-lactams were ceftazidime (CAZ; 0.125 μg/ml) and imipenem (IMP; 0.25 μg/ml), and others were fosfomycin (FOF), norfloxacin (NFX; 0.05 μg/ml), and erythromycin (ERY; 5 μg/ml). EOP assays were independently repeated at least three times.

Overexpression and purification of CpxR-His.

A single colony of BL21(DE3) transformed with pET24a+-cpxR-His was grown overnight at 37°C. The next day, the strain was subcultured 1:100 until reaching an optical density at 600 nm (OD₆₀₀) of 0.4, and protein expression was induced with 1 mM IPTG for 3 h at 37°C. The cells were harvested by centrifugation and lysed in 10 mM Tris-HCl (pH 8.0) with 150 mM NaCl supplemented with 20 mM imidazole by one passage through a cell disruptor (Constant Systems) at 2 × 10⁵ kPa. After removal of the cell debris, CpxR-His in the supernatant was subjected to nickel affinity chromatography by using a HiTrap chelating HP column (GE Healthcare Life Sciences) according to the manufacturer’s instructions. Purified CpxR-His protein was passed through a HiTrap desalting column (GE Healthcare Life Sciences) to remove imidazole (see Fig. S1). All protein purification steps were operated by an AKTA protein purification system (Amersham Biosciences).

Overexpression of CpxA-His and CpxA^Y144N-His and preparation of inner membrane vesicles.

Single colonies of C41(DE3) transformed with pET24a+-cpxA-His or pET24a+-cpxA^Y144N-His were grown overnight at 37°C. Strains were subcultured 1:100 at 37°C until reaching an OD₆₀₀ of 0.4, and protein expression was induced with 1 mM IPTG for 3 h at 30°C. Cells were harvested by centrifugation and lysed in 10 mM Tris-HCl (pH 8.0) with 150 mM NaCl by one passage through a cell disruptor (Constant Systems) at 2 × 10⁵ kPa. This mainly yields inside-out inner membrane vesicles (IMVs) (68). After removal of cell debris by low-speed centrifugation (7,000 × g, 20 min, 4°C), the supernatant was ultracentrifuged (100,000 × g, 60 min, 4°C) to collect whole-cell envelopes. These were separated into inner and outer membranes by ultracentrifugation through a 30% to 55% (wt/vol) sucrose density gradient as previously described (69). After SDS-PAGE, fractions corresponding to inner membranes enriched in CpxA-His or CpxA^Y144N-His were pooled, resuspended in phosphorylation buffer (50 mM Tris-HCl [pH 7.5], 10% glycerol [vol/vol], 2 mM dithiothreitol, 50 mM KCl, and 5 mM MgCl₂) after removal of sucrose (see Fig. S2), and used in phosphorylation assays immediately after preparation.

Analyzes of CpxA autokinase, kinase, and phosphatase activities in vitro.

Purified IMVs containing CpxA-His or CpxA^Y144N-His were diluted in phosphorylation buffer at a concentration of approximately 20 mg/ml. Autophosphorylation was tested by using the ADP-Glo kinase assay (Promega) according to the manufacturer’s instructions (see Fig. S3 and text in the supplemental material).

For testing kinase and phosphatase activities, CpxA-His and CpxA^Y144N-His in IMVs were allowed to autophosphorylate in the presence of ATP and then pelleted by centrifugation and washed twice in phosphorylation buffer to remove residual ATP. For kinase activity assays, CpxR-His was added at a concentration of 1.5 μM (CpxA/CpxR molar ratio of ∼1:3) and incubated for 30 min at room temperature. Purified CpxR-His was labeled with 50 μM acetyl phosphate and used as a positive control (see text in the supplemental material). All reactions were stopped by the addition of 5× Laemmli buffer. Phosphoproteins were separated by 50 μM Phos-tag (Wako Laboratory Chemicals) 12% PAGE in the presence of 100 μM MnCl₂ and detected by staining with Coomassie blue R250. For phosphatase activity assays, CpxR∼P (from acetyl phosphate labeling) was added, and mixtures were incubated for 30 min at room temperature. Released phosphate levels were measured by using the phosphate colorimetric assay kit (Sigma) according to the manufacturer’s instructions (see Fig. S4 and text in the supplemental material).

Phos-tag analysis of the CpxR phosphorylation in vivo.

The presence of phosphorylated CpxR (CpxR∼P) in lysates of BL21(DE3) cotransformed with pET24a+-cpxR-His and pBAD33-cpxA or pBAD33-cpxA^Y144N was assessed by coupling Phos-tag PAGE to immunoblotting with horseradish peroxidase (HRP)-conjugated 6×His epitope tag monoclonal antibody (1:2,000) (Sigma). For this, single bacterial colonies were grown overnight at 37°C. The next day, the strains were subcultured 1:100 until reaching an OD₆₀₀ of 0.4. CpxA expression was first induced with 0.2% l-arabinose for 2 h at 37°C, and then 1 mM IPTG was added to induce CpxR-His expression for another 3 h at 37°C; 2 × 10⁹ cells were harvested by centrifugation, and cell pellets were vigorously resuspended in 1.2 M formic acid. Whole-cell lysates were solubilized by the addition of 5× Laemmli buffer and neutralized by the addition of 5 M NaOH. Samples (10 μl, equivalent to 0.2 OD units) were loaded and separated by 25 μM Phos-tag (Wako Laboratory Chemicals) 12% PAGE in the presence of 50 μM MnCl₂. Before transfer onto a polyvinylidene difluoride (PVDF) membrane, the gel was washed with transfer buffer supplemented with 1 mM EDTA for 10 min to remove excess of MnCl₂.

Bacterial two-hybrid assay.

The interactions between CpxA-SD and CpxP were studied by using a bacterial adenylate cyclase two-hybrid (BATCH) assay (Euromedex). The target proteins CpxA and CpxP of K. aerogenes were separately fused to the C termini of the T25 and T18 domains, respectively, as previously described (13). To estimate the interaction between CpxA and CpxP, DHM1 cya cells were transformed with the recombinant plasmids pKT25-cpxA/cpxA^Y144N and pUT18C-cpxP. For detection of lactose-metabolizing clones, bacteria were grown overnight at 30°C. The next day, 10 μl of each culture was spotted onto LB agar supplemented with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal; 40 μg/ml) and 0.5 mM IPTG and incubated for 24 h at 30°C. DHM1 cells were transformed with pKT25-zip and pUT18C-zip as positive controls (blue, Lac⁺) and with empty pKT25 and pUT18C plasmids as negative controls (white, Lac⁻).

Other protein methods.

Whole-cell envelopes were prepared as described above and resuspended in 20 mM HEPES-NaOH buffer (pH 7.2). All samples were diluted in Laemmli buffer and heated for 5 min at 100°C before loading. Samples corresponding to 0.2 OD units were separated by 10% SDS-PAGE. To better resolve OmpF and OmpC, 4 M urea was added to the running gel. Proteins were either visualized after staining with Coomassie brilliant blue R250 or transferred onto PVDF blotting membranes (Bio Rad) as described previously (26). Protein quantification was performed by using a Pierce bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific) according to the manufacturer’s protocol.

Quantification of AmpC β-lactamase activity.

β-Lactamase activity was assessed by using a microtiter plate nitrocefin hydrolysis assay as previously described (70). All chemicals were from Sigma, excepting nitrocefin (Merck) and IMP (Sequoia Chemicals Ltd.).

Statistical analyses.

All assays were repeated at least three times. Results are presented on the basis of the calculated means with standard deviations. When appropriate, P values were calculated and are indicated with asterisks in the figure legends.