CbrA Mediates Colicin M Resistance in Escherichia coli through Modification of Undecaprenyl-Phosphate-Linked Peptidoglycan Precursors

Overexpression of the chromosomal cbrA gene allows E. coli to resist colicin M (ColM), a bacteriocin specifically hydrolyzing the undecaprenyl-PP-MurNAc(-pentapeptide)-GlcNAc (lipid II) peptidoglycan precursor of targeted cells. This resistance results from a CbrA-dependent modification of the precursor structure, i.e., reduction of the α-isoprenyl bond of C₅₅-carrier lipid moiety that is proximal to ColM cleavage site. This modification, observed here for the first time in eubacteria, annihilates the ColM activity without affecting peptidoglycan biogenesis. These data, which further increase our knowledge of the substrate specificity of this colicin, highlight the capability of E. coli to generate reduced forms of C₅₅-carrier lipid and its derivatives. Whether the function of this modification is only relevant with respect to ColM resistance is now questioned.

ABSTRACT

Colicin M is an enzymatic bacteriocin produced by some Escherichia coli strains which provokes cell lysis of competitor strains by hydrolysis of the cell wall peptidoglycan undecaprenyl-PP-MurNAc(-pentapeptide)-GlcNAc (lipid II) precursor. The overexpression of a gene, cbrA (formerly yidS), was shown to protect E. coli cells from the deleterious effects of this colicin, but the underlying resistance mechanism was not established. We report here that a major structural modification of the undecaprenyl-phosphate carrier lipid and of its derivatives occurred in membranes of CbrA-overexpressing cells, which explains the acquisition of resistance toward this bacteriocin. Indeed, a main fraction of these lipids, including the lipid II peptidoglycan precursor, now displayed a saturated isoprene unit at the α-position, i.e., the unit closest to the colicin M cleavage site. Only unsaturated forms of these lipids were normally detectable in wild-type cells. In vitro and in vivo assays showed that colicin M did not hydrolyze α-saturated lipid II, clearly identifying this substrate modification as the resistance mechanism. These saturated forms of undecaprenyl-phosphate and lipid II remained substrates of the different enzymes participating in peptidoglycan biosynthesis and carrier lipid recycling, allowing this colicin M-resistance mechanism to occur without affecting this essential pathway.

IMPORTANCE Overexpression of the chromosomal cbrA gene allows E. coli to resist colicin M (ColM), a bacteriocin specifically hydrolyzing the undecaprenyl-PP-MurNAc(-pentapeptide)-GlcNAc (lipid II) peptidoglycan precursor of targeted cells. This resistance results from a CbrA-dependent modification of the precursor structure, i.e., reduction of the α-isoprenyl bond of C₅₅-carrier lipid moiety that is proximal to ColM cleavage site. This modification, observed here for the first time in eubacteria, annihilates the ColM activity without affecting peptidoglycan biogenesis. These data, which further increase our knowledge of the substrate specificity of this colicin, highlight the capability of E. coli to generate reduced forms of C₅₅-carrier lipid and its derivatives. Whether the function of this modification is only relevant with respect to ColM resistance is now questioned.

INTRODUCTION

Colicin M (ColM) is a bacteriocin that certain strains of Escherichia coli produce in order to kill competitor strains of E. coli or related species (1). This colicin, which had been known for a long time to interfere with bacterial cell wall biogenesis (2), was later identified as a hydrolase (phosphodiesterase) catalyzing the specific degradation of the peptidoglycan undecaprenyl-PP-MurNAc(-pentapeptide)-GlcNAc lipid II intermediate precursor (MurNAc, N-acetylmuramic acid; GlcNAc, N-acetylglucosamine) (3). ColM-treated E. coli cells indeed accumulated undecaprenol and 1-pyrophospho-MurNAc(-pentapeptide)-GlcNAc as lipid II degradation products, and the resulting depletion of the pool of lipid II rapidly led to the arrest of peptidoglycan synthesis and, ultimately, to cell lysis (3). The crystal structure of ColM was determined (4), and active site amino acid residues were identified using site-directed mutagenesis experiments (5). Several ColM homologues that are produced by a restricted number of strains from Pseudomonas, Burkholderia, and Pectobacterium species were subsequently identified and characterized both biochemically and structurally (6–9).

ColM is the only colicin targeting peptidoglycan metabolism, as all the other colicins identified to date exhibit either a membrane pore-forming or a nuclease activity (1, 10). Its mode of action requires several successive essential steps: (i) binding to a specific receptor, FhuA, present in the outer membrane of susceptible cells (11), (ii) subsequent internalization into the periplasmic space via the TonB/ExbB/ExbD translocation machinery (1, 10), and (iii) maturation by the bifunctional periplasmic protein FkpA exhibiting chaperone and peptidyl-prolyl-cis-trans isomerase activities (12), after which ColM can reach and degrade its target, i.e., the lipid II molecules present on the outer side of the plasma membrane. E. coli mutants (fhuA, tonB, or fkpA) defective in either of these steps are fully resistant to ColM. Naturally occurring ColM-producing E. coli isolates express the ColM gene (cma) together with a cmi immunity gene on a plasmid (13). Low-level expression of the Cmi protein in the periplasm of wild-type cells confers to them high-level resistance to ColM (14). However, the mechanism of action of the Cmi immunity protein remains unknown. More recently, the overexpression of an E. coli gene named cbrA (formerly yidS) was shown to render cells specifically resistant to ColM (15). Based on the sequence similarities it exhibits with geranylgeranyl reductases (GGR) and other flavin adenine dinucleotide (FAD)-dependent oxidoreductases, Helbig et al. suggested that the CbrA protein might function as a reductase of isoprenoid molecules. However, these authors did not detect any modification of the pools and structure of undecaprenyl-phosphate (C₅₅-P) derivative lipids in CbrA-overexpressing cells, leading them to conclude that hydrogenation of peptidoglycan lipid intermediates should not be the mechanism of CbrA-mediated ColM resistance (15). They also purified the CbrA protein and showed that it contained noncovalently bound FAD, but they failed to demonstrate any enzymatic activity with this preparation (15). In the present study, we revisited this question and demonstrated that overexpression of CbrA in E. coli cells resulted in the accumulation in the cell membranes of a new form of carrier lipid displaying a saturated α-prenyl unit. This modified dihydro-C₅₅-P lipid is shown to be accepted as a substrate both in vitro and in vivo by the MraY and MurG enzymes, and the α-saturated form of lipid II thus generated is shown to be efficiently used by peptidoglycan polymerases. However, in contrast to the natural lipid II, this α-saturated form of lipid II appeared to be totally resistant to the hydrolase activity of ColM.

RESULTS

Expression of CbrA and colicin M resistance.

The cbrA (yidS) gene of E. coli was cloned into pTrc99A and pTrcHis30 vectors that allow expression of the protein in wild-type and N-terminal His-tagged forms, respectively, under the control of the strong isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible trc promoter. The resulting plasmids pMLD391 and pMLD682 were shown to confer high-level resistance to ColM to E. coli cells, confirming previous data from Helbig et al. (15). Indeed, only faint growth inhibition zones were observed when 5 μl of an undiluted stock of ColM (1 mg/ml) was spotted on 2× yeast extract-tryptone (2YT) agar plates seeded with the BW25113(pMLD391) and BW25113(pMLD682) strains, while clear lysis zones were observed with the control strain BW25113(pTrc99A) for up to a 10⁵-fold dilution of the same ColM stock (Fig. 1A and C). This protection provided by CbrA was similarly observed when cells were grown in liquid medium (Fig. 1B). No induction by IPTG was required, showing that the basal expression of the cbrA gene from these plasmids was already sufficient to confer resistance to the highest concentration of ColM tested here. The fact that both pMLD391 and pMLD682 plasmids conferred ColM resistance demonstrated that the addition of an N-terminal His₆ tag did not affect the protein activity (Fig. 1A and C).

FIG 1.

Effect of the overexpression of wild-type and mutant CbrA proteins on E. coli cells’ susceptibility to colicin M. (A and C) Spot killing assays of the capability of wild-type and mutated CbrA proteins to protect E. coli cells against ColM. In panel A, spots (5 μl) of serial 10-fold dilutions (from top to bottom) of ColM stock (1 mg/ml) were deposited onto lawns of BW25113 strains (ca. 10⁸ cells) on 2YT agar plates, and growth inhibition zones were observed after 24 h of incubation at 37°C. (C) Five-microliter spots of nondiluted (nd; ca. 10⁷ cells) or serial 10-fold dilutions of cultures of the same strains were deposited onto 2YT agar plates containing ColM at the indicated concentrations (0, 10, and 50 ng/ml from left to right), and growth was observed after 24 h of incubation at 37°C. CbrA variants overexpressed by these strains are indicated above the lanes. Control, cells carrying the empty pTrc99A vector; WT(His₆) and WT, cells expressing CbrA with or without an N-terminal His₆ tag (pMLD682 and pMLD391 plasmids, respectively); C46A, Y202A, W204A, and K208A, cells expressing the corresponding CbrA mutants (pMLD682-derived pMLD684, pMLD690, pMLD688, and pMLD692 plasmids, respectively). In panel B, BW25113 cells carrying either the pTrc99A empty vector or the cbrA-expressing derivative plasmid pMLD391 were grown exponentially at 37°C in 2YT medium. At the time indicated by the arrow (OD, 0.4), pure ColM was added at the indicated final concentrations (in ng/ml), and growth was monitored at 600 nm. Closed and open symbols correspond to the control and cbrA-overexpressing strains, respectively. Representative data from experiments performed at least in triplicate are shown.

Effect of CbrA overproduction on the pools and structures of C₅₅-isoprenoids.

The prototroph E. coli strain FB8 was used as a genetic background for analyzing the effects of the overexpression of the cbrA gene on the pools of C₅₅-P derivatives present in cell membranes. This strain, transformed with either the empty vector pTrc99A or the cbrA-expressing pMLD391 plasmid, was grown exponentially in 2YT-ampicillin medium, and the pools of C₅₅-isoprenoids were extracted and analyzed by high-performance liquid chromatography (HPLC) as described in Materials and Methods. Figure 2A shows a typical profile obtained for a control cell extract in which the peak corresponding to C₅₅-P is indicated. The identity of this compound eluted at ca. 20 min was confirmed by comparison with an authentic standard and by matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry analysis, as described previously (16). Integration of this peak allowed us to determine a pool value of ca. 250 nmol/g of cell dry weight for this lipid in strain FB8 (Table 1). Interestingly, an additional peak was observed when the extract from the cbrA-expressing strain was analyzed under the same conditions, which eluted 2 to 3 min after the peak of C₅₅-P (Fig. 2C). This compound was collected, and its mass spectrometry analysis yielded an m/z value of 848.00 for the [M-H]^– ion, which was 2 Da higher than that of C₅₅-P (calculated m/z value, 845.66), suggesting that this new compound differed from C₅₅-P by the saturation of one prenyl unit (Fig. 3B). A sample of commercial authentic dihydro-C₅₅-P had the same retention time in HPLC and showed an m/z value of 848.10 (Fig. 3A). That the overexpression of the cbrA gene resulted in the accumulation of a hydrogenated form of C₅₅-P in the cell membranes thus demonstrated the reductase activity of the CbrA protein. Moreover, the amount of dihydro-C₅₅-P significantly increased when cbrA gene expression was further induced by adding 0.5 mM IPTG 1 h before the culture was arrested (Table 1). As reported earlier (16), applying the Kato extraction procedure to these membrane extracts also allowed us to determine the pools of the undecaprenyl-pyrophosphate (C₅₅-PP) precursor and those of its dihydro derivative (Table 1). Our data thus showed that a significant proportion of total C₅₅-isoprenoids were present in their dihydro form in the membranes of CbrA-overexpressing cells, ca. 33% and 61% in the absence and presence of IPTG, respectively.

FIG 2.

Accumulation of dihydro-C₅₅-isoprenoids in membranes of CbrA-, but not Cmi-, overexpressing cells. FB8 cells carrying either the pTrc99A vector (A and B), the cbrA-expressing plasmid pMLD391 (C and D), or the cmi-expressing plasmid pMLD232 (E) were grown exponentially at 37°C in 2YT medium. When the OD₆₀₀ reached 0.7, colicin M was added (B and D) or not added (A, C, and E) to the cultures at a concentration of 75 ng/ml. Cultures were stopped 10 min later, and lipids were extracted from the cell pellets by using the Bligh and Dyer procedure (53). Chloroform extracts were analyzed by reverse-phase HPLC on a Nucleosil 100 C₁₈ column (5 μm, 250 by 4.6 mm), using 2-propanol:methanol 1:4 (vol/vol) containing 10 mM phosphoric acid as the eluent at a flow rate of 0.6 ml/min. Peaks were detected at 210 nm. Experiments were carried out 4 times under each condition.

TABLE 1.

Pools of C₅₅-isoprenoids in E. coli cell membranes^a

Strain	IPTGb	ColMb	Pool level (nmol/g of cell dry wt)c for:
			C55-P	C55-PP	C55-OH	Dihydro-C55-P	Dihydro-C55-PP	Dihydro-C55-OH	Total
FB8(pTrc99A)	−	−	246 ± 37	137 ± 25	ND	ND	ND	ND	383 ± 44
FB8(pTrc99A)	−	+	94 ± 12	52 ± 8	102 ± 19	ND	ND	ND	248 ± 24
FB8(pMLD391)	−	−	186 ± 26	74 ± 15	ND	61 ± 15	65 ± 18	ND	386 ± 38
FB8(pMLD391)	+	−	116 ± 23	36 ± 8	ND	113 ± 23	125 ± 24	ND	390 ± 41
FB8(pMLD391)	−	+	75 ± 15	26 ± 7	47 ± 3	94 ± 19	148 ± 28	ND	390 ± 38
FB8(pMLD391)	+	+	79 ± 16	5 ± 1	41 ± 6	102 ± 18	164 ± 21	ND	391 ± 32

^aFB8 cells carrying either the empty vector pTrc99A or the cbrA-expressing plasmid pMLD391 were grown exponentially at 37°C in 2YT medium. In all cases, cultures were stopped at an OD of 1.0 and divided into two 50-ml samples which were treated using the Bligh and Dyer (53) and Kato (52) procedures, respectively, as described previously (16). The extracted isoprenoids were analyzed by HPLC as described in Materials and Methods and the legend to Fig. 2. Values are expressed as the mean ± standard deviation (SD) from 4 experiments.

^b+, addition of IPTG (0.5 mM) at an OD of 0.4 and/or of ColM (75 ng/ml) at an OD of 0.7; −, no IPTG or ColM addition.

^cND, not detected.

FIG 3.

Mass spectrometry analysis of dihydro-C₅₅-P extracted from E. coli cells. Lipid extracts from cbrA-overexpressing cells were analyzed by HPLC as described in the text and the Fig. 2 legend. The peak corresponding to dihydro-C₅₅-P was collected, and an aliquot was analyzed by MALDI-TOF mass spectrometry as described in Materials and Methods. (A) Authentic commercial dihydro-C₅₅-P standard; (B) purified lipid extracted from E. coli cell membranes. Peaks at m/z 848.10 (A) and 848.00 (B) that were assigned to be the [M-H]^– ions were observed.

We previously reported that the treatment of E. coli cells by the lipid II-degrading colicin M (ColM) resulted in the accumulation of undecaprenol (C₅₅-OH) in cell membranes, a compound that does not normally exist in E. coli and could not be reused for recycling the active form C₅₅-P of the carrier lipid (16). We confirmed here that control FB8 cells accumulated C₅₅-OH following treatment by ColM (Fig. 2B and Table 1). The overexpression of the cbrA gene greatly reduced but did not abolish the formation of C₅₅-OH resulting from ColM treatment (Fig. 2D and Table 1). Interestingly, dihydro-C₅₅-OH was not detected in these cells, although dihydro forms of C₅₅-P and C₅₅-PP were readily detected, suggesting that C₅₅-OH was not a substrate of the CbrA reductase. However, these in vivo data did not allow us to make conclusions on the substrate specificity of this reductase. Indeed, the rapid turnover of the C₅₅-P carrier lipid pool known to occur during peptidoglycan biosynthesis suggests that if such a modification was to concern only C₅₅-P, C₅₅-PP, or lipid II, it would then be quickly spread and shared by all these lipids in fine. Moreover, the absence of dihydro-C₅₅-OH in cell membranes also suggested that if produced, dihydro-lipid II molecules should be resistant to the ColM hydrolase activity. The latter hypothesis was therefore tested using appropriate in vitro assays, as described below.

Localization of the reduced isoprenyl unit.

Which one of the 11 double bonds present in C₅₅-P (Fig. 4) was reduced by CbrA was determined using mild acid hydrolysis experiments. Indeed, owing to the presence of the allyl phosphate linkage and the ready formation of an allylic carbocation, this carrier lipid and its derivatives are prone to acid hydrolysis and to further rearrangements (17, 18). As shown in Fig. S1 in the supplemental material, the C₅₅-P lipid extracted and purified from E. coli cell membranes was totally degraded following incubation in 1 M HCl for 30 min at 100°C, whereas the dihydro-C₅₅-P lipid purified from the CbrA-overexpressing cells perfectly resisted these mild acid hydrolysis conditions. This difference of behavior between the C₅₅-P allyl derivative and the dihydro-C₅₅-P alkyl derivative allowed us to conclude that the double bond targeted by CbrA was the α-isoprenyl unit.

FIG 4.

CbrA and ColM targeted bonds in C₅₅-P carrier lipid derivatives. The structure and mixed E,Z (trans/cis) stereochemistry of the C₅₅-P carrier lipid are shown. The site of cleavage of ColM in the peptidoglycan lipid I and II intermediates is indicated, as well as the α-prenyl unit that is specifically reduced by the CbrA enzyme (the nature of R remaining to be identified in that case). ColM is known to specifically hydrolyze the lipid I and lipid II peptidoglycan intermediates and to be inactive on C₅₅-P, C₅₅-PP, and UDP-MurNAc-pentapeptide (3).

Saturation of the lipid II α-isoprenyl unit abolishes ColM activity.

The above-described results thus strongly suggested that the acquired resistance to ColM of the CbrA-overexpressing cells resulted from the saturation of the α-isoprenyl unit present in the C₅₅ lipid moiety of the peptidoglycan lipid II intermediate, i.e., the ColM target. That the reduction of this bond, which is very close to the ColM cleavage site (Fig. 4), may abolish or at least reduce the activity of this enzyme was indeed highly conceivable. Evidence for this was provided in different ways. First, we made valuable use of the C₅₅-P and dihydro-C₅₅-P extracts purified by HPLC from the E. coli membranes (Fig. 2) to generate the corresponding radiolabeled peptidoglycan undecaprenyl-PP-MurNAc-pentapeptide (lipid I) intermediates with purified MraY translocase and UDP-MurNAc-[¹⁴C]pentapeptide. We had indeed previously shown that ColM cleaved both the lipid I and lipid II peptidoglycan intermediates (3) and that MraY could accept reduced as well as unreduced polyprenyl-phosphates of various sizes as substrates (19, 20). As shown in Fig. 5, lipid I generated from the “natural” unreduced C₅₅-P lipid could subsequently be cleaved by ColM, a reaction releasing 1-pyrophospho-MurNAc-pentapeptide as a product, as reported earlier (3). However, when dihydro-C₅₅-P instead of C₅₅-P was used, radiolabeled dihydro-lipid I was generated, but no subsequent cleavage by ColM occurred (Fig. 5).

FIG 5.

ColM activity on natural and dihydro forms of peptidoglycan lipid I intermediate. C₅₅-P and dihydro-C₅₅-P were extracted and purified from E. coli cell membranes as described in Materials and Methods and Fig. 2. These lipids were subsequently used for the in vitro synthesis of natural (unreduced) (A and B) and dihydro (C and D) forms of peptidoglycan lipid intermediate I (C₅₅-PP-MurNAc-pentapeptide) using purified MraY enzyme and radiolabeled UDP-MurNAc-[¹⁴C]pentapeptide (UM5) as the cosubstrate. These assays were performed in the absence (A and C) or presence (B and D) of ColM. Radiolabeled substrate (UM5) and products (lipid I and 1-PP-MurNAc-pentapeptide) were separated by TLC and detected with a radioactivity scanner, as detailed in the text.

Then, purified radiolabeled lipid IIs were also synthesized, this time using commercial sources of C₅₅-P and dihydro-C₅₅-P, purified MraY and MurG enzymes, UDP-MurNAc-pentapeptide, and UDP-[¹⁴C]GlcNAc as the radiolabeled substrate. The data shown in Fig. 6 clearly demonstrated that ColM exhibited its hydrolytic activity only on the natural, unreduced version of the lipid II, consistent with the preceding observations with lipid I. That the dihydro-lipid II was not a substrate of ColM in vitro was thus consistent with the acquired ColM-resistant phenotype of CbrA-overexpressing cells. It also explained why only undecaprenol (C₅₅-OH), but not dihydro-C₅₅-OH, was detectable in membranes from these cells following ColM treatment.

FIG 6.

ColM activity on natural and dihydro forms of peptidoglycan lipid II intermediate. Radiolabeled lipid IIs were synthesized as described in the text using purified MraY and MurG enzymes, UDP-MurNAc-pentapeptide, UDP-[¹⁴C]GlcNAc, and either C₅₅-P or dihydro-C₅₅-P as the substrate. The natural (unreduced) (A and B) and dihydro (C and D) forms of [¹⁴C]lipid II thus generated were purified and incubated in the absence (A and C) or presence (B and D) of ColM. The radiolabeled substrates (lipids II) and reaction product 1-PP-MurNAc(-pentapeptide)-GlcNAc were separated by TLC and detected with a radioactivity scanner, as detailed in the text.

Dihydro-lipid II sustains peptidoglycan synthesis.

The basal-level expression of CbrA provided by plasmid pMLD391 in the absence of IPTG was sufficient to confer ColM resistance without affecting bacterial growth (Fig. 1). This indicated that the ongoing rate of cell wall peptidoglycan synthesis might not be significantly affected under these conditions, thereby suggesting that the dihydro-lipid II molecules thus generated were accepted as substrates by the peptidoglycan polymerases (glycosyltransferases) acting immediately downstream in this pathway. It also suggested that the C₅₅-PP phosphatase activities that play an essential role in the regeneration (recycling) of the C₅₅-P carrier lipid pool could dephosphorylate and allow recycling of dihydro-C₅₅-PP molecules released during the latter polymerization steps.

As multiple peptidoglycan glycosyltransferases and C₅₅-PP phosphatases exist in E. coli, whether only some members of these protein families could accept dihydro-substrates was then questioned and tested using appropriate mutant strains.

E. coli produces several glycosyltransferases—the three class A, bifunctional penicillin-binding proteins PBP1A, PBP1B, and PBP1C (which also exhibit transpeptidase activity) and the monofunctional glycosyltransferase MtgA, which are encoded by the mrcA, mrcB, pbpC, and mtgA genes, respectively (21, 22). The SEDS family (shape, elongation, division, and sporulation) proteins FtsW and RodA were also recently shown to act as peptidoglycan transglycosylases (23, 24). The mrcA and mrcB genes are individually dispensable, but inactivation of both genes is lethal, PBP1A and PBP1B being two main peptidoglycan glycosyltransferases in this species. The mrcA and mrcB single mutants and the mrcB pbpC mtgA triple mutant expressing PBP1A as the sole glycosyltransferase (25) were transformed by the cbrA-expressing pMLD679 plasmid or the pUCP24Nco control vector, and the resulting transformants were then tested for their susceptibility to ColM. This vector conferring gentamicin resistance was preferentially used here to avoid any effect of ampicillin on the PBP activity. As shown in Fig. 7A and B, the mrcB mutant and the triple mutant appeared slightly more susceptible to ColM than the wild-type and mrcA mutant strains, a finding likely correlated with the lower peptidoglycan content and hypersusceptibility to cell wall targeting agents characterizing PBP1B-deficient strains (26). However, overexpression of CbrA conferred ColM resistance to all of these strains without yielding any apparent toxicity effect (Fig. 7). This result demonstrated that both PBP1A and PBP1B were able to catalyze peptidoglycan polymerization using dihydro-lipid II as a substrate.

FIG 7.

Susceptibility to ColM and resistance conferred by CbrA in transglycosylase and C₅₅-PP phosphatase mutants. BW25113 mutant strains carrying deletions of genes encoding peptidoglycan transglycosylases or C₅₅-PP phosphatases were transformed by the control vector pUCP24Nco or the cbrA-expressing plasmid pMLD679. These strains were grown overnight in 2YT-gentamicin liquid medium, and 5-μl spots of nondiluted (nd) or serial 10-fold dilutions of these cultures were deposited onto 2YT-agar plates containing (B, D) or not containing (A, C) ColM at a concentration of 15 ng/ml. Growth was observed after 24 h of incubation at 37°C. – and + indicate the absence (empty vector) or presence (pMLD679) of cbrA-overexpressing plasmid, respectively. The numbers 1 to 7 correspond to the following strains: 1, wild-type BW25113 parental strain; 2, ΔmrcA strain (PBP1A mutant); 3, ΔmrcB strain (PBP1B mutant); 4, ΔmrcB ΔpbpC ΔmtgA strain (PBP1B/PBP1C/MtgA triple mutant); 5, ΔpgpB ΔlpxT ΔbacA strain; 6, ΔybjG ΔlpxT ΔbacA strain; and 7, ΔybjG ΔlpxT ΔpgpB strain. YbjG, PgpB, and BacA are the only remaining C₅₅-PP phosphatases expressed in the triple mutant strains 5, 6 and 7, respectively. Representative data from experiments performed at least in triplicate are shown.

Similarly, in E. coli, four C₅₅-PP phosphatases were identified which conjointly participate in the recycling of C₅₅-PP molecules that are released during peptidoglycan polymerization steps. These integral membrane proteins belong to two protein families: BacA and PAP2 (phosphatidic acid phosphatases of type 2) proteins PgpB, YbjG, and LpxT, all of them having their active site exposed toward the periplasm (27, 28). None of them is individually essential for growth, but construction of multiple conditional mutants showed that at least one of the three proteins BacA, YbjG, and PgpB was required and sufficient to sustain cell wall peptidoglycan synthesis and ensure cell viability (29). To know whether each of these different phosphatases can dephosphorylate/recycle dihydro-C₅₅-PP, the protection provided by CbrA against ColM was assessed in mutants expressing only one of them. These deletion mutants, BW25113 ΔpgpB ΔlpxT ΔbacA (YbjG only), BW25113 ΔybjG ΔlpxT ΔbacA (PgpB only), and BW25113 ΔybjG ΔlpxT ΔpgpB (BacA only), and the parental BW25113 strain were here too transformed by the cbrA-expressing pMLD679 plasmid or the empty vector pUCP24Nco, and the susceptibility of transformants to ColM was then determined. As shown in Fig. 7C and D, all these mutants appeared to be significantly protected from ColM following CbrA overexpression. Interestingly, the strain expressing PgpB as the unique C₅₅-PP phosphatase appeared more susceptible to ColM than the other strains in the assay conditions used (ColM at 15 ng/ml), suggesting that the recycling of the C₅₅-P pool might be partially slowed down in that strain. These different results showed that the main peptidoglycan polymerases and C₅₅-P-recycling phosphatases tolerated this structural modification (reduction) of the carrier lipid structure and remained active in cbrA overexpression conditions.

That the CbrA-dependent incorporation of dihydro intermediates in the C₅₅-P carrier lipid cycle could affect the cell peptidoglycan content and structure was then questioned. BW25113 cells overexpressing (pMLD679) or not overexpressing (pUCP24Nco vector) the CbrA protein were grown exponentially and harvested at the same optical density (OD) of 1.0, and the peptidoglycan polymer was extracted and quantitated as previously described (30). The peptidoglycan contents of both strains appeared identical, ca. 2,800 ± 70 nanomoles (in terms of diaminopimelic acid content per liter of culture), indicating that the rate of peptidoglycan biosynthesis was not affected under these conditions of CbrA overexpression, a finding consistent with the demonstrated capability of peptidoglycan synthetases to utilize dihydro-C₅₅-containing substrates. The fine structure of the latter purified peptidoglycan preparations was then determined using the classical procedure consisting of the HPLC analysis of fragments (muropeptides) resulting from the digestion of this polymer by a specific muramidase (mutanolysin) (31). Here, too, no significant change in the pattern of muropeptides and in the overall ratio of dimers versus monomers (i.e., the peptidoglycan cross-linking index) was observed, whether the CbrA protein was overexpressed or not (Fig. S2).

The impact of CbrA overexpression on the cell susceptibility to bacitracin was also tested. This antibiotic is known to act by tightly binding to the pyrophosphoryl moiety of C₅₅-PP, the sequestration of this lipid preventing the recycling of the C₅₅-P carrier lipid and leading to an arrest of peptidoglycan biosynthesis. We thus analyzed whether the reduction of the α-isoprenyl unit that is proximal to the pyrophosphoryl group in this lipid could modify the cell susceptibility to this antibiotic. As shown in Fig. S3, only a slight decrease of the cell susceptibility to bacitracin was observed following CbrA overexpression, suggesting that dihydro-C₅₅-PP remains a target for this antibiotic. However, the susceptibility of E. coli and other Gram-negative species to this antibiotic being intrinsically very low, due to the presence of the outer membrane, further in vitro analyses would be required to estimate the real impact this modification has on the affinity of bacitracin for the C₅₅-PP lipid.

Site-directed mutagenesis of CbrA.

As reported by Helbig et al. (15), the CbrA protein exhibited sequence similarities with geranylgeranyl reductases and other FAD-dependent oxidoreductases of diverse functions belonging to the para-hydroxybenzoate hydroxylase (PHBH) superfamily. Sequence alignments, together with the elucidation of the crystal structures of two archeal members of this family, a geranylgeranyl reductase (GGR) from Sulfolobus acidocaldarius (32) and a digeranylgeranyl glycerophospholipid reductase (DGGR) from Thermoplasma acidophilum (33), both in complex with FAD, allowed the identification of several conserved amino acid residues and motifs likely involved in the binding of FAD and substrates or in catalysis. In particular, a characteristic YXWXFP motif located within the catalytic domain was highlighted in these studies, the aromatic residues of which point to the FAD and are suggested to maintain the lipid substrate in the appropriate position for reduction (residues 215 to 220 and 209 to 214 of S. acidocaldarius and T. acidophilum protein sequences, respectively) (Fig. 8). A widely conserved cysteine residue, Cys47 and Cys44 in the Sa_GGR and Ta_GGR proteins, respectively, was also shown to play an essential role in catalysis (32, 33). This residue was indeed present in Ec_CbrA (Cys46), but the YXWXFP motif appeared less conserved, as only two of the aromatic residues, namely, Tyr202 and Trp204, were found in the CbrA protein sequence (Fig. 8). Site-directed mutagenesis of these residues, and also of a proximal conserved Lys208 residue, was performed on the CbrA-expressing plasmid pMLD682, and the resulting variants were tested for their capability to protect E. coli cells against ColM. As shown in Fig. 1A and C, the replacement of either of the Cys46, Tyr202, Trp204, or Lys208 residues by an alanine strongly affected the CbrA activity, as judged by the loss of viability observed for these clones at ColM concentrations that normally did not affect growth of wild-type CbrA-expressing cells. Indeed, the susceptibility to ColM of transformants overexpressing the Cys46Ala and Trp204Ala CbrA mutants was similar to that of the BW25113 host strain, and a partial protection against this colicin was provided by the Tyr202Ala mutant and to a lesser extent by the Lys208Ala mutant, suggesting that the reductase activity of the Tyr202Ala and Lys208Ala mutant proteins was not totally abolished. These results, which demonstrated that the catalytic activity of CbrA was required to confer immunity against ColM, confirmed the essential role played by these different conserved residues in the reaction mechanism of this reductase, as previously reported for some members of this protein superfamily (32, 33).

FIG 8.

Sequence alignment of CbrA and geranylgeranyl reductases. Homologous geranylgeranyl reductase protein sequences were retrieved with blastp. Amino acid sequences of CbrA from E. coli (Ec_CbrA) and geranylgeranyl reductases from S. acidocaldarius (Sa_GGR) and T. acidophilum (Ta_GGR), whose structures are solved, plus other GGR annotated proteins, whose accession numbers are given, were aligned using the constraint-based multiple protein alignment tool (COBALT) from NCBI (54). Conserved residues are written in red in sequence blocks. Conserved motifs shared by members of the para-hydroxybenzoate hydroxylase family (PHBH) that are involved in nucleotide and FAD binding are indicated below the sequences in red and orange, respectively. The motif that is unique to the GGR subfamily and is involved in lipid binding is indicated in green. Secondary structure elements based on the 3D structure of Sa_GGR (PDB accession number 3ATR) are presented on top: α-helices (α) and 3₁₀-helices (η) are displayed as squiggles, β-strands (β) are displayed as arrows, and turns are represented with the letter T. The figure was generated using ESPript-3.0 (55). The GenBank accession numbers of Sa_GGR, Ec_CbrA, and Ta_GGR are WP_011277849, WP_001340434, and WP_010900941, respectively.

Purification and enzymatic assays of the CbrA protein.

A purification of the CbrA protein was performed in order to biochemically characterize this protein and identify its substrate among prenol derivatives. The N-terminally His₆-tagged CbrA protein was successfully overproduced in BW25113(pMLD682) cells (Fig. S4) and shown to be almost exclusively recovered in the membrane fraction following cell disruption, as already experienced by Helbig et al. (15). In contrast to the control, the CbrA-containing membrane extract was yellowish, likely indicating the presence of oxidized FAD cofactor. The CbrA protein mainly remained in the pellet fraction following solubilization of membranes by 4% of N-dodecyl-β-d-maltopyranoside (DDM) detergent. However, significant amounts were recovered in the supernatant fraction as the CbrA protein was subsequently successfully purified from the solubilized membrane fraction, as shown in Fig. S4. Of note, the final purified sample (yield, ca. 0.2 mg of protein per liter of culture) had apparently lost the oxidized FAD cofactor, as judged by the lack of typical absorption spectrum.

The total membrane extracts and the purified CbrA protein were used for reductase enzymatic assays as described in Materials and Methods. The C₅₅-PP and C₅₅-P lipids were extensively incubated with CbrA samples in the presence of FAD and either NADPH, NADH, or sodium dithionite as hydrogen donors. HPLC analyses of these reaction mixtures performed as described in Materials and Methods did not reveal any significant conversion of these lipids or change of their hydrogenation status (data not shown). The lipid II peptidoglycan precursor was also tested as a CbrA substrate. In that case, radiolabeled lipid II was incubated overnight with CbrA samples in the presence of FAD and the different hydrogen donors, and a treatment by ColM was subsequently performed to assess the conversion of lipid II to a ColM-resistant, dihydro-lipid II product. However, no difference in the extent of lipid II hydrolysis by ColM was observed between the control and CbrA-treated lipid II (data not shown), indicating that the α-isoprenyl bond of this substrate had not been reduced in these assay conditions.

Toxicity of CbrA overproduction.

As mentioned above, basal expression of the cbrA gene from pTrc-derivative plasmids was sufficient, i.e., did not need induction by IPTG, to allow cell protection against ColM. Whether increased production of CbrA could further improve this protection was then tested. In fact, the addition of IPTG at a 100-μM to 1-mM concentration in the growth medium turned out to be toxic for cell growth. Interestingly, this toxicity was only detected with the wild-type CbrA protein but not with the mutants whose capability to protect cells against ColM was abolished or dramatically reduced (Fig. S5). Indeed, addition of IPTG decreased by a 5- to 6-log factor the number of colonies observed after overnight growth in the case of the BW25113(pMLD682) strain expressing the wild-type CbrA protein. No significant loss of viability was observed under these conditions for cells carrying the pMLD690, pMLD688, and pMLD692 plasmids expressing the Tyr202Ala, Trp204Ala, and Lys208Ala CbrA variants, respectively (Fig. S5). Similar amounts of these different proteins were detected in cell membrane extracts following overnight growth of the latter strains in the presence of 1 mM IPTG (Fig. S6), showing that the decreased toxicity of the four mutant proteins did not result from differences in the expression and/or stability of these proteins. The correlation thus highlighted between the toxicity and the protein activity showed that this was not simply the overproduction of the CbrA protein but, rather, the increase of its specific enzymatic (reductase) activity in the cell content that was responsible for the observed loss of cell viability.

ColM resistance conferred by Cmi is not CbrA dependent.

E. coli strains are protected from the colicins they produce via the concomitant expression of an immunity protein (1), which is named Cmi (or ImM) in the case of ColM. The three-dimensional (3D) crystal structure of the Cmi protein was determined earlier, and amino acid residues that are important for its activity were identified (14). However, by which mechanism Cmi confers ColM resistance remains unknown. It was thus tempting to speculate that both Cmi and CbrA proteins could mediate ColM resistance by the same mechanism, i.e., the reduction of the lipid II C₅₅ moiety. Whether Cmi could itself be a reductase or may indirectly modulate the expression or activity of the CbrA protein was then tested. First, an analysis of the pools of C₅₅-lipids in membranes of Cmi-expressing, ColM-resistant cells (carrying the pMLD232 plasmid) did not reveal the presence of dihydro forms of these lipids (Fig. 2E). Second, we showed that Cmi expression protected cbrA mutant cells as efficiently as the wild-type cells against ColM (Fig. S7), thereby demonstrating that the Cmi activity did not rely on CbrA and that the Cmi- and CbrA-mediated resistance mechanisms were unrelated.

DISCUSSION

The inactivation of either of the genes required for the reception on the cell surface, the import in the periplasm, or the maturation of ColM confers resistance against this colicin to E. coli cells (10, 11). Overexpressing the fhuA receptor and tonB/exbBD import machinery genes was shown earlier to improve the uptake of ColM and thus to further increase the cell susceptibility to this colicin (34). To the best of our knowledge, the cbrA (yidS) gene was, to date, the only example of a gene whose overexpression resulted in ColM resistance. We demonstrated here that the CbrA-mediated resistance phenotype resulted from a modification of structure of the lipid II target that immunized it against the hydrolase activity of this colicin. This structural modification ultimately shared by all the C₅₅-P-derived lipid intermediates present in the membrane (Fig. 4) was tolerated by the main enzymes involved in peptidoglycan biosynthesis and C₅₅-P carrier lipid recycling, thus allowing ColM resistance to occur without affecting the cell viability.

Helbig et al. (15) failed earlier to demonstrate an enzymatic (reductase) activity of CbrA, using the cell-extracted mixture of undecaprenyl derivatives as substrates, and our attempts described here were, unfortunately, not more successful. Indeed, the CbrA protein was efficiently overproduced and purified in His-tagged form (see Fig. S4 in the supplemental material), but none of the potential substrates we tested (C₅₅-PP, C₅₅-P, and lipid II) appeared to be reduced by this enzyme in the in vitro assay conditions used. Maybe the functional stability of this membrane protein was only conserved in a particular membrane environment, the recognition of its substrates was affected by the presence of the detergent, or the in vitro reconstitution of CbrA activity failed because this enzyme requires a functional partner. This also raised the question of the identity of the natural substrate(s) and that of the physiological role of this enzyme. The detection of dihydro forms of C₅₅-P, C₅₅-PP, and lipid II in the membranes of CbrA-overexpressing cells did not mean that all of these lipids were substrates of CbrA. Indeed, the reduction of either of them was expected to be rapidly propagated to the others, due to the functioning of the C₅₅-P carrier lipid cycle, C₅₅-P being used by the MraY and MurG enzymes to synthesize lipid II and the C₅₅-PP released from lipid II by transglycosylases being then recycled back to C₅₅-P by the C₅₅-PP phosphatases. The finding that C₅₅-OH but not dihydro-C₅₅-OH accumulated in membranes of CbrA-overexpressing cells following treatment with ColM suggested that C₅₅-OH was not a substrate of CbrA. However, considering that C₅₅-OH is released from lipid II by ColM on the outer side of the inner membrane and that the catalytic site of CbrA is likely oriented toward the cytoplasmic side of this membrane, the absence of dihydro-C₅₅-OH could simply reflect an incapacity of C₅₅-OH, in contrast to C₅₅-P, to be translocated (flipped) back to the cytoplasm.

To date, the capability to reduce polyprenyl substrates in vitro or in vivo has been confirmed for a limited number of putative geranylgeranyl reductases (GGR) and related enzymes, most of them originating from archaeal species (35). Some reductases were specific to pyrophosphate substrates, but others accepted both alcohol and pyrophosphate forms of their isoprenoid (farnesyl, geranylgeranyl, etc.) substrate and could saturate only one or several of the double bonds present in these molecules. In some cases, the presence of the pyrophosphate group was shown to prevent reduction of the α-prenyl group (35). These enzymes are expected to use NADPH/NADH or ferredoxin as an electron donor, as demonstrated for the T. acidophilum (33) and Methanosarcina acetivorans (36) GGRs, respectively, but as is the case for the S. acidocaldarius GGR (32), the source of electrons often remains unknown.

Whether the reductase activity of CbrA toward lipid II and/or the lipid carrier demonstrated here reflects its specific physiological role or is fortuitous and observable only when this protein is overproduced to unnatural levels remains unclear. Helbig and coworkers showed earlier that the expression of the E. coli cbrA gene was under the control of the CreBC two-component regulatory system (15). An 8-fold increase of transcription of this gene (as judged by using chromosomal lacZ fusions) was indeed detected following a shift from rich medium (LB) to glucose-M9 minimal medium, which conferred partial resistance against ColM. However, no difference in the susceptibility to ColM was observed between the wild-type and cbrA and creBC mutant cells when these strains were grown in rich medium (15). As shown here, no dihydro-C₅₅-P was detected in membrane extracts of the wild-type strain, confirming that basal expression of CbrA does not provide any significant protection against ColM. That the primary function of this reductase is to protect susceptible E. coli strains against ColM-producing strains seems unlikely. One would rather imagine that the main substrate of this enzyme remains to be identified and that the modification of lipid II structure observed here represents a side activity exhibited under particular “nonphysiological” circumstances, i.e., when this protein is significantly overproduced. Interestingly, in this respect, cbrA-overexpressing cells appeared more resistant to osmotic shock than wild-type cells (15), which made Helbig and coworkers hypothesize that CbrA may impact in some way the structure of the cell envelope and possibly the outer membrane. We observed here that the high-level overproduction of the wild-type CbrA protein was toxic for E. coli cells, no overnight growth being detected on plates following IPTG induction of gene expression from the pTrc-derivative plasmids. However, this toxic effect was not observed with the Cys46Ala variant and other inactive mutants generated in the present work, demonstrating that the CbrA enzyme activity, and not simply the protein overproduction, accounted for the observed phenomena. Whether this results from a more extensive reduction of C₅₅-P derivative lipid intermediates in membranes or the modification of other putative substrates remains to be elucidated. The fact that the expression of the chromosomal cbrA gene is only induced under stress conditions would suggest a potential implication of these reduced lipids in adaptation to harsh conditions.

Polyprenyl phosphate lipids allow the translocation of sugars/glycans across cell membranes in all kingdoms of life. They are required for the synthesis of various bacterial cell wall polysaccharides (peptidoglycan, teichoic acids, O-antigen, etc.), as well as for protein glycosylation in eukaryotic, archaeal, and some bacterial species (37). In bacteria, interfering with the synthesis (or recycling) of the C₅₅-P carrier lipid results in the arrest of peptidoglycan synthesis and, ultimately, in cell lysis. This essential carrier lipid and its metabolism thus represent valuable potential targets in the search for new antibacterial agents (38, 39). However, such antibiotics should be highly specific and not toxic for mammal cells, considering that congenital protein glycosylation disorders are known to lead to severe malformations and developmental retardations with dramatic outcomes in humans (40). Interestingly, in this respect, the structure of the lipid carrier varies to some extent in living organisms, in the carbon chain length (mainly C₅₅ in bacteria, up to C₁₀₀ in eukaryotes), its stereochemistry (cis/trans configuration of double bonds), and the hydrogenation status of the terminal isoprenyl unit bound to the phosphate group (37). This α-isoprenyl unit remains unsaturated in bacteria but appears saturated in eukaryotes and archaea, suggesting the occurrence of an additional reduction step leading to the so-called dolichyl phosphate carrier lipid in the latter species. The latter reaction was shown to be catalyzed, at least partially, by the steroid 5α-reductase type 3 (SDR5A3) in humans (41), a protein that does not exhibit any significant sequence homology with CbrA (data not shown). The potential utilization of ColM and its different homologs as an alternative to antibiotics for combating multidrug-resistant pathogenic species was previously envisaged (42). The present demonstration that ColM could specifically cleave bacterial lipid II peptidoglycan precursors carrying an unsaturated C₅₅ moiety while preserving dihydro-lipid II and thus dolichyl phosphate-linked glycans from the human host thus reinforces the interest in this class of bacteriocin and supports their potential use as antibacterial agents.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The E. coli strain DH5α (Bethesda Research Laboratories) was used as the host for propagation of plasmids, and the prototroph FB8 strain was described earlier (43). The wild-type BW25113 strain was obtained from B. Wanner, and the derivative cbrA mutant strain (JW5631) was from the Keio collection (44). Other mutant strains carrying single or multiple deletions of the PBP1A (mrcA), PBP1B (mrcB), PBP1C (pbpC), and MtgA (mtgA) genes were previously described (25), as were strains carrying single or multiple deletions of the C₅₅-PP phosphatase genes (pgpB, ybjG, lpxT, and bacA) (27, 29, 45). The pTrc99A plasmid was from Amersham Biosciences, and the pTrcHis30 (46) and pUCP24Nco (47) plasmid vectors were described earlier. The pMLD232 plasmid used for expression of ColM immunity protein Cmi was previously described (14). Unless otherwise noted, cells were grown in 2YT medium (48) at 37°C. Ampicillin, kanamycin, chloramphenicol, and gentamicin were used at 100, 50, 25, and 10 μg/ml, respectively. Growth was monitored at 600 nm with a Shimadzu UV-1601 spectrophotometer.

General DNA techniques and E. coli cell transformation.

Gene amplification (PCR) was performed with a Thermocycler 60 apparatus (Bio-Med) and the Expand-Fidelity polymerase (Roche), and the resulting fragments were purified using Wizard purification kits (Promega). Standard procedures for gel electrophoresis, gene cloning, and plasmid purification were used (49). Site-directed mutagenesis of CbrA enzyme residues was performed directly on cbrA-expressing plasmids by using the Quikchange II site-directed mutagenesis kit from Agilent Technologies and the oligonucleotides shown in Table S1 in the supplemental material. Transformation of E. coli cells by plasmids was performed as described earlier by Dagert and Ehrlich (50) or by electroporation.

Construction of expression plasmids.

Plasmids allowing overexpression of the CbrA protein were constructed as follows. The cbrA gene was amplified by PCR from the chromosome of E. coli strain DH5α using PCR primers Cbra-1 and Cbra-2 that incorporated NcoI and HindIII restriction sites at the 5′ and 3′ extremities of the gene, respectively (Table S1). The resulting DNA fragment digested by these enzymes was inserted between the same sites of the pTrc99A vector, generating the plasmid pMLD391 that allows expression of the wild-type cbrA gene under the control of the strong isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible trc promoter. For construction of the pMLD682 plasmid allowing expression of CbrA with an N-terminal His₆ tag, the gene was amplified using the Cbra-2 and Cbra-3 oligonucleotides (Table S1), and the resulting fragment was cleaved by BamHI and HindIII and then inserted between the same sites of the pTrcHis30 vector. The cbrA gene-containing NcoI-HindIII fragment from pMLD391 was also cloned between the same sites of the pUCP24Nco vector (47), yielding the pMLD679 plasmid (gentamicin resistance). The pMLD684, pMLD690, pMLD688, and pMLD692 plasmids allowing expression of Cys46Ala, Tyr202Ala, Trp204Ala, and Lys208Ala mutated CbrA proteins, respectively, were generated by site-directed mutagenesis of the pMLD682 plasmid. The sequences of cloned inserts were verified by DNA sequencing (Eurofins-MWG).

Preparation of crude extracts, purification of CbrA, and enzymatic assays.

E. coli BW25113 cells carrying either the pTrcHis30 vector as a control or the pMLD682 plasmid were grown at 37°C in 2YT-ampicillin medium (1-liter cultures). When the optical density of the culture reached 0.8, IPTG was added at a final concentration of 1 mM and growth was continued overnight at 22°C. Cells were harvested and washed with 40 ml of cold 20 mM Tris-HCl buffer (pH 7.4) containing 150 mM NaCl (buffer A). The cell pellet was suspended in 6 ml of the same buffer, and cells were disrupted by sonication in the cold (Bioblock Vibracell sonicator, model 72412). The resulting suspension was centrifuged at 4°C for 30 min at 200,000 × g in a TL100 Beckman centrifuge, and the pellet consisting of membranes and associated proteins (including CbrA) was recovered (total membrane extract). The latter extract was subjected to solubilization by resuspension in 20 ml of buffer A containing 4% DDM. This mixture was incubated for 4 h at 4°C with shaking and then centrifuged as described above to separate nonsolubilized proteins (sediment) and solubilized proteins (supernatant). The supernatant obtained from CbrA-overexpressing BW25113(pMLD682) cells was used for the purification of the N-terminally His₆-tagged CbrA protein. It was incubated for 1 h at 4°C with nickel-nitrilotriacetate (Ni²⁺-NTA)–agarose preequilibrated in buffer A containing 10 mM imidazole. Then, the polymer was washed extensively with buffer A containing 0.2% DDM and 10 mM imidazole and with buffer A containing 0.2% DDM and 50 mM imidazole, and elution of proteins was obtained with buffer A containing 0.2% DDM and 200 mM imidazole. CbrA-containing fractions were dialyzed overnight against 100 volumes of buffer A containing 0.05% DDM. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis of proteins was performed as described by Laemmli and Favre (51), and protein concentrations were determined with a NanoDrop apparatus (Thermo Scientific).

Enzymatic (reductase) CbrA assays using C₅₅-PP or C₅₅-P as the substrate were performed as follows. The reaction mixture (100 μl) contained 0.5 mM lipid substrate, 20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.2% DDM, 100 μM FAD, and 100 mM NADPH, NADH, or sodium dithionite. Total membrane extracts (5 μl) or the purified CbrA protein (30 μl, about 3 μg of protein) were added to the reaction mixture, which was then incubated overnight at 30°C with shaking. The lipids were recovered by extraction with butanol, dried under vacuum, eventually submitted to Kato’s procedure (C₅₅-PP samples), and analyzed by HPLC as described in this report and previously (16) to determine their hydrogenation status. Enzymatic assays using lipid II as the substrate were performed in a reaction mixture (20 μl) containing 12 μM ¹⁴C-radiolabeled lipid II in the buffer used for ColM enzymatic activity assay (see below) supplemented with 1 mM NADPH or sodium dithionite. Then, 2 μl of CbrA-containing membrane extracts or purified CbrA protein was added to the reaction mixture, which was incubated overnight at 30°C. Then, 2 μl of purified ColM was added, the mixture was incubated for 1 h at 37°C, and the extent of lipid II hydrolysis was evaluated by thin-layer chromatography (TLC) on LK6D silica gel plates (Whatman) using 1-propanol–ammonium hydroxide–water (6/3/1, vol/vol) as the mobile phase. Radioactive spots were located and quantified with a radioactivity scanner (RITA Star; Raytest Isotopenmessgeräte GmbH, Straubenhardt, Germany).

ColM enzymatic activity assay.

The hydrolase activity of ColM was routinely tested in a reaction mixture (10 μl) containing 100 mM Tris-HCl (pH 7.5), 150 mM NaCl, 20 mM MgCl₂, 10 mM β-mercaptoethanol, 12 μM ¹⁴C-radiolabeled lipid II (140 Bq), 0.2% DDM, and purified ColM (5 μl of appropriate dilution in buffer A). Some assays used lipid I instead of lipid II as the substrate. In that case, radiolabeled lipid I was generated by using purified MraY enzyme, UDP-MurNAc-[¹⁴C]pentapeptide, and either C₅₅-P or dihydro-C₅₅-P as the lipid substrate. Reaction mixtures were incubated for 1 h at 37°C with shaking (Thermomixer; Eppendorf), and reactions were stopped by heating at 100°C for 1 min. These mixtures were analyzed by TLC, and the radioactive spots were detected and quantified as described above.

ColM cytotoxicity assays.

The cytotoxic activity of ColM was tested on 2YT-agar plates overlaid with 3 ml of soft nutrient agar containing ca. 10⁸ cells of the E. coli strains to be tested. Serial dilutions of pure ColM-containing samples were prepared in buffer A, and 5-μl aliquots were spotted on the overlay. Plates were incubated for 24 h at 37°C, and the sensitivity of strains and efficiency of ColM were judged from the clear zones of growth inhibition observed at the spot position. Alternatively, ColM was incorporated in the 2YT agar medium, and 5-μl serial dilutions of cultures of the E. coli strains to be tested were spotted on the plates. ColM cytotoxicity was also tested in 2YT liquid medium by adding ColM at different concentrations to exponentially growing cultures of E. coli.

Extraction and purification of C₅₅-P and its derivatives from E. coli membranes.

Cell pellets harvested from 100-ml cultures of the different strains to be tested were treated essentially as described previously (16). One half of the cell pellet was submitted to the procedure of Kato et al. (52) that allows conversion of C₅₅-PP into C₅₅-P, and lipids were in all cases extracted according to Bligh and Dyer (53) with minor modifications (16). Extracts were analyzed by HPLC as described earlier (16), and injections of standard commercial lipids under the same conditions allowed the calibration and the quantification of these lipids in cell extracts. Mild acid hydrolysis experiments were performed to allow identification of the reduced bond in the dihydro-C₅₅-P lipid extracted from cbrA-overexpressing cells. Samples of C₅₅-P and dihydro-C₅₅-P lipids of commercial source or purified from E. coli cells were treated with 1 M HCl for 30 min at 100°C or not treated, and the resulting mixtures were analyzed by HPLC as described above.

Matrix-assisted laser desorption–ionization time-of-flight (MALDI-TOF) mass spectrometry analysis of lipids.

MALDI-TOF mass spectra were recorded in the reflectron mode with delayed extraction on a PerSeptive Voyager-DE STR instrument (Applied Biosystems) equipped with a 337-nm laser. Aliquots were dissolved in methanol–2-propanol (1:1, vol/vol), mixed with two volumes of 100 mM terthiophene in ethyl acetate, and then deposited (1 μl) on sample plates. Spectra were recorded at an acceleration voltage of –20 kV and an extraction delay time of 200 ns. A mixture of peptidoglycan precursors (UDP-MurNAc, UDP-MurNAc-dipeptide, and UDP-MurNAc-pentapeptide) was used as an external calibrant.

Extraction and quantification of peptidoglycan.

BW25113 cells (0.8-liter cultures) carrying either the cbrA-expressing plasmid pMLD679 or the empty vector pUCP24Nco were grown exponentially at 37°C in 2YT-gentamicin medium. At an optical density at 600 nm (OD₆₀₀) of 1.0, bacteria were harvested in the cold, suspended in a few milliliters of 0.9% NaCl solution, and then injected under vigorous stirring in 30 ml of a hot (95 to 100°C) aqueous 4% SDS solution. After 1 h of this treatment, suspensions were kept overnight at room temperature and then centrifuged for 30 min at 200,000 × g. The resulting pellets containing the peptidoglycan polymer were washed several times with water, and after final resuspension in 3 ml of water, aliquots were hydrolyzed (16 h at 95°C in 6 M HCl) and analyzed with a Hitachi model 8800 amino acid analyzer (ScienceTec). The cell peptidoglycan content was expressed in terms of its characteristic components (diaminopimelic acid, muramic acid) as previously reported (30). Its fine structure was determined by using the classical procedure of Glauner, consisting of the analysis of fragments (muropeptides) released following digestion of this polymer by a muramidase (31). The purified peptidoglycan preparations were digested by mutanolysin, and muropeptides were reduced with sodium borohydride and then separated by HPLC on a 3-μm Hypersil octadecyl silane (ODS) column (4.6 by 250 mm; Thermo Scientific). A gradient of methanol (from 0% to 25% in 90 min) in 50 mM sodium phosphate, pH 4.5, was used for elution at a flow rate of 0.5 ml/min. Peaks of muropeptides were detected at 207 nm.

Chemicals.

Undecaprenyl-pyrophosphate (C₅₅-PP), undecaprenyl-phosphate (C₅₅-P), undecaprenol (C₅₅-OH), and their dihydro derivatives were provided by the Institute of Biochemistry and Biophysics of the Polish Academy of Sciences. [¹⁴C]lipid II labeled in the GlcNAc moiety was prepared as described previously (3). The C₅₅-PP-MurNAc(-l-Ala-γ-d-Glu-meso-DAP-d-Ala-d-Ala)-GlcNAc lipid II from E. coli was used in this study, where DAP represents diaminopimelic acid. N-Dodecyl-β-d-maltopyranoside (DDM) was purchased from Thermo Fisher Scientific, isopropyl-β-d-thiogalactopyranoside (IPTG) was purchased from Eurogentec, and Ni²⁺-NTA–agarose was purchased from Qiagen. Colicin M was purified as described previously (5). Mutanolysin, antibiotics, and reagents were from Sigma. Oligonucleotide synthesis and DNA sequencing were performed by Eurofins-MWG.