Impacts of Penicillin Binding Protein 2 Inactivation on β-Lactamase Expression and Muropeptide Profile in Stenotrophomonas maltophilia

Inducible expression of chromosomally encoded β-lactamase(s) is a key mechanism for β-lactam resistance in Enterobacter cloacae, Citrobacter freundii, Pseudomonas aeruginosa, and Stenotrophomonas maltophilia. The muropeptides produced during the peptidoglycan recycling pathway act as activator ligands for β-lactamase(s) induction. The muropeptides 1,6-anhydromuramyl pentapeptide and 1,6-anhydromuramyl tripeptide are the known activator ligands for ampC β-lactamase expression in E. cloacae. Here, we dissected the type of muropepetides for L1/L2 β-lactamase expression in an mrdA deletion mutant of S. maltophilia. Distinct from the findings with the ampC system, 1,6-anhydromuramyl tetrapeptide is the candidate for ΔmrdA-mediated β-lactamase expression in S. maltophilia. Our work extends the understanding of β-lactamase induction and provides valuable information for combating the occurrence of β-lactam resistance.

BlrAB (the homologue of CreBC in E. coli), which is involved in ␤-lactamase expression, has been identified in A. hydrophilia (12). The inhibition of the DD-CPase activity of PBP4 by ␤-lactams results in the accumulation of N-acetylglucosaminyl-N-acetylmuramyl pentapeptides (disaccharide pentapeptides) in the periplasm, which then induces BlrAB activation and ␤-lactamase gene expression (12).
S. maltophilia is an important nosocomial pathogen and is regarded as a reservoir of antibiotic resistance determinants (20). Intrinsic resistance to ␤-lactams occurs via the induction of chromosomally encoded L1 and L2 ␤-lactamases (21). In general, the L1/L2 induction in S. maltophilia is similar to that in E. cloacae and P. aeruginosa, via the ampG-ampD-nagZ-ampR regulatory circuit, with a few minor differences. First, instead of AmpG permease, an AmpN-AmpG permease system is responsible for the transport of degraded PG sacculi in S. maltophilia (22). Second, ampD I , but not ampD II , has been shown to be relevant to ␤-lactamase expression, although two ampD homologues, ampD I and ampD II , have been identified in the S. maltophilia genome (23). Third, we have demonstrated that disruption of mrcA (encoding PBP1a) results in an increase in basal L1/L2 activity in a NagZ-independent manner (24,25). Therefore, in addition to the known NagZ-dependent pathway, a NagZ-independent pathway for L1/L2 expression exists in S. maltophilia. The involvement of a TCS in PBP inactivation-mediated ␤-lactamase expression has not been reported so far in S. maltophilia.
Each bacterium has its own unique PBP system. Different ␤-lactams exhibit distinct affinities toward individual PBPs in different microorganisms. Therefore, determination of the relationship between PBP inactivation and ␤-lactamase induction potential in individual microorganisms is of great importance. Here, we further investigated the roles of other PBPs, in addition to PBP1a, in L1/L2 expression in S. maltophilia. We found that PBP2 inactivation confers increased L1/L2 basal activity in the absence of ␤-lactam challenge. Furthermore, we tried to elucidate the possible AL responsible for strain ⌬mrdA-mediated L1/L2 expression. A more complete understanding of the circuits regulating ␤-lactamase expression could lead to the identification of potential targets for controlling ␤-lactamase-mediated resistance and preservation of the antibacterial activities of the ␤-lactam class of antibiotics.

RESULTS
Analysis of putative PBP genes in S. maltophilia K279a. Genome-wide bioinformatics analysis of the S. maltophilia K279a genome (26) in comparison with the known E. coli PBPs led to the identification of gene sequences with the closest homologues in the S. maltophilia genome. These genes were Smlt3826, Smlt3681, Smlt3602, Smlt4056, Smlt0750, Smlt0462, and Smlt4050, the orthologues for the mrcA, mrcB, pbpC, mrdA, ftsI, dacB, and dacC E. coli genes, respectively. S. maltophilia lacks clear homologues for E. coli dacA, dacD, and pbpG. The predicted PBPs of S. maltophilia shared 36 to 47% identity and 53 to 63% similarity with those of E. coli (see Table S1 in the supplemental material).
The impact of each PBP inactivation on the basal and cefuroxime-induced ␤-lactamase activities was assessed (Fig. 1). S. maltophilia KJ harbors two inducible ␤-lactamases, L1 and L2, and cefuroxime has been proven to be a potent L1/L2 inducer (21). In agreement with previous data (24), the inactivation of mrcA caused an increase in basal L1/L2 activity. Likewise, the KJΔmrdA displayed a high basal ␤-lactamase activity. The basal ␤-lactamase activities of the KJΔmrcB, KJΔpbpC, KJΔdacB, and KJΔdacC mutant strains were as low as that of wild-type KJ (Fig. 1A). Furthermore, we also compared the ␤-lactamase activities of the PBP mutants with the wild-type KJ in the presence of cefuroxime. All the PBP mutants tested displayed cefuroxime-induced ␤-lactamase activities comparable to that of wild-type KJ (Fig. 1B). The subsequent study was focused on elucidating the mechanism of the increased expression of mutant strain ⌬mrdA-mediated basal ␤-lactamase.
Impact of mrdA inactivation on bacterial growth and morphology. First, we assessed whether mrdA inactivation affected bacterial growth by monitoring the optical density at 450 nm (OD 450 ). KJΔmrdA displayed a lower growth rate during exponential growth and an approximately 3-h delay before entering the stationary phase. However, the stationary-phase cells of KJ and KJΔmrdA exhibited equivalent biomass formation ( Fig. 2A).
PBP2 is required for rod morphology and cell wall elongation in E. coli (27). The KJΔmrdA mutant was therefore evaluated for morphological changes. In contrast to the rod-shaped KJ cells, KJΔmrdA cells were spherical (Fig. 2B). In view of this morphological aberration of strain KJΔmrdA, we wondered whether the difference in the OD 450 between logarithmic-phase KJ and mutant KJΔmrdA cells was caused by a growth defect or a morphological bias. To address these possibilities, the bacterial growth of strain KJ and KJΔmrdA cells at 6, 15, and 21 h was quantified by CFU enumeration, and the results supported the former possibility. The bacterial growth, evaluated based on either the OD 450 ( Fig. 2A) or CFU enumeration (Fig. 2C), was consistent with the conclusion that the KJΔmrdA deletion mutant has compromised logarithmic-phase growth. Impact of mrdA inactivation on the expression of L1 and L2. There are two active ␤-lactamases, L1 and L2, in S. maltophilia KJ; however, their inducible expression levels appear to be differentially regulated (21). Therefore, we were interested in evaluating the involvement of L1 and L2 in the strain ⌬mrdA-mediated basal ␤-lactamase activity increase. The L1 and L2 genes were individually deleted from the KJΔmrdA strain, yielding mutants KJΔmrdAΔL1 and KJΔmrdAΔL2. Individual loss of L1 and L2 decreased the basal ␤-lactamase activity of KJΔmrdA (Fig. 3A). In addition, the L1 and L2 transcripts of KJ and mutant KJΔmrdA cells were also quantified by quantitative reverse transcription-PCR (qRT-PCR). In parallel, the expression levels of L1 and L2 genes in the mutant KJΔmrdA were increased 6.2-Ϯ 0.8-fold and 4.6-Ϯ 1.0-fold (means Ϯ standard deviations) above those in KJ (Fig. 3B), highlighting that strain ⌬mrdA-mediated increased basal ␤-lactamase activity is regulated at the transcription level.
Roles of ampR, ampNG, nagZ, and ampD I in the mutant ⌬mrdA-mediated increase in basal ␤-lactamase activity. AmpR and AmpNG are essential for ␤-lactamase expression in S. maltophilia (22,28). We were interested in assessing the essentiality of AmpR and AmpNG in mutant ⌬mrdA-mediated basal L1/L2 activity. The ΔampR and ΔampG alleles were introduced into strain KJΔmrdA, yielding the double mutants KJΔmrdAΔR and KJΔmrdAΔG, respectively, and these mutants were evaluated for basal L1/L2 activity. Not surprisingly, deletion of ampR or ampG of strain KJΔmrdA resulted in basal ␤-lactamase activity at levels comparable to that of the wild-type KJ (Fig. 4), consistent with the known roles of AmpR and AmpNG.
To assess the impact of nagZ on the mutant ⌬mrdA-mediated basal L1/L2 activity, the nagZ allele was deleted from the chromosome of strain KJΔmrdA, yielding the mutant KJΔmrdAΔZ. The basal ␤-lactamase activity of KJΔmrdAΔZ reverted to nearly the same level as that of wild-type KJ (Fig. 4), indicating that strain ΔmrdA-mediated basal L1/L2 activity is NagZ dependent.
AmpD I overexpression is known to attenuate the ␤-lactam-induced and mutant strain ⌬mrcA-mediated basal ␤-lactamase activity in S. maltophilia (23,24). An ampD Icontaining plasmid (pAmpDI) was introduced into strain KJΔmrdA to assess the role of AmpD I in ⌬mrdA-mediated basal ␤-lactamase activity. The basal ␤-lactamase activity of mutant strain KJΔmrdA reverted to the level of wild-type KJ when AmpD I was overexpressed ( Fig. 4), supporting that the stem peptide is critical for the activity of the AL generated by mrdA inactivation.
Impact of mrdA inactivation on muropeptide profiles. On the basis of the known AmpC paradigm, the exact AL of 〈mpC expression involves muropeptides, and these AL precursors are transported from the periplasm to the cytosol by AmpG permease (14). Because the roles of AmpG, NagZ, AmpR, AmpD I , and CreBC in strain ⌬mrdAmediated L1/L2 expression are very similar to those of their corresponding homologues in the AmpC induction model, we hypothesized that the AL for L1/L2 expression are muropeptides, and their precursors should accumulate at high levels in the ⌬mrdA mutant. To evaluate this, we set to determine the muropeptide profile of PG in wild-type KJ and KJΔmrdA.
Two different protocols were used for muropeptide preparation to clarify the muropeptide content in different cellular locations, as described in Materials and Methods. First, muropeptides were isolated from total cell extracts via muramidase treatment, representing the total muropeptides in murein, periplasm, and cytosol. The total muropeptides and their relative abundance levels were analyzed using liquid chromatography and mass spectrometry (LC-MS) (31)(32)(33)(34). Compared to the wild-type KJ, strain KJΔmrdA showed a clear increase in two peaks in the total muropeptide ion chromatogram, labeled as peak 1 (retention time of 4.53) and peak 2 (retention time of 7.07) in Fig. 5A. The corresponding compounds for peaks 1 and 2 were identified as the reduced form of N-acetylglucosaminyl-N-acetylmuramyl-L-alanyl-D-glutamyl-meso-diaminopimelic acid-D-alanine (GlcNAc-MurNAc-tetrapeptide) (M4) and N-acetylglucosaminyl-1,6-anhydro-N-acetylmuramyl-L-alanyl-D-glutamyl-meso-diaminopimelic acid-D-alanine (GlcNAc-anhMurNAc-tetrapeptide) (M4N), respectively (see Fig. S1 and S2 for tandem mass analysis results). The ratio of muropeptide composition is summarized in Fig. 5B and Table S2. The most notable increase in strain KJΔmrdA was in M4N, which increased by ϳ10-fold, whereas M4 muropeptide showed a less significant increase in strain KJΔmrdA. Based on inferences from the known AmpC and L1/L2 induction models (9,20), we assumed that M4N is the major AL precursor, transported from the periplasm into the cytosol by the AmpN/AmpG permease system, and further processed into the anhMurNAc tetrapeptides by NagZ in an ⌬mrdA mutant. To support this assumption, we isolated the periplasmic muropeptides without muramidase treatment. Compared to that of wild-type KJ, the periplasmic muropeptides profile of KJΔmrdA had a considerable increase (ϳ5-fold) in M4N (Fig. 6), consistent with the findings for the total muropeptides profiles of strains KJ and KJΔmrdA.

DISCUSSION
PBPs, the target of ␤-lactams, are enzymes responsible for PG synthesis. HMW-PBPs are known to be the primary targets of ␤-lactam (35), but LMW-PBPs are thought to participate in sensitizing bacteria to ␤-lactams (36). Therefore, a ␤-lactam can simultaneously inhibit HMW-and LMW-PBP, resulting in a synergistic bactericidal effect.  Table S2). *, P Ͻ 0.05 (for M4N). Muropeptide symbols: M, monomer; D, dimer; T, trimer (numbers following the letters are the number of amino acid stem peptides). (G) Additional glycines at the end of stem peptides (replacing L-alanine). N, terminating anhydro-muropeptide; dAc, no acetyl group in GluNAc site. The muropeptides that have never been reported in the literature were further validated by MS/MS analysis (Fig. S4 to S6).
However, for ampR-␤-lactamase-harboring Gram-negative bacteria, the loss of PBP activity by ␤-lactam binding may produce surplus or different muropeptide intermediates involved in chromosomal ␤-lactamase expression. ␤-Lactams, which can simultaneously bind to the HMW-PBPs and LMW-PBPs, display strong ␤-lactamase induction potency and cause high ␤-lactam resistance; for example, imipenem or cefoxitin is a potent inducer of ␤-lactamase in E. cloacae and P. aeruginosa (10). Thus, an effective ␤-lactam for infection treatment can also be a potent inducer of ␤-lactam resistance. Because the extent of inactivation of each PBP by different ␤-lactams varies, ␤-lactam treatment-mediated ␤-lactamase expression is complex, and the extent of inactivation of each PBP and the individual impact of PBP inactivation on induced ␤-lactamase activity are difficult to evaluate. Thus, PBP in-frame deletion mutants obtained via genetic manipulation in the laboratory can be used as an alternative strategy, and this method has been extensively applied in numerous studies (24,29). Two reports have linked single PBP gene inactivation to increased basal ␤-lactamase activity in ampR-␤-lactamaseharboring Gram-negative bacteria, including dacB inactivation in P. aeruginosa (mutant PAΔdacB) (29) and mrcA inactivation in S. maltophilia (mutant KJΔmrcA) (24). In this study, we used a novel example, KJΔmrdA (with an mrdA in-frame deletion mutant of S. maltophilia).
Escherichia coli AmpG (AmpG EC ) shows a strict substrate restriction. The principal requirement for AmpG EC uptake is the presence of anhydrodisaccharide, GlcNAc-anhMurNAc, but the presence of peptides is not necessary. Thus, anhydrodisaccharide and anhydrodisaccharide peptides are compatible substrates for AmpG EC [for example, GlcNAc-anhMurNAc (M0N), GlcNAc-anhMurNAc tripeptide (M3N), GlcNAc-anhMurNAc tetrapeptide (M4N), and GlcNAc-anhMurNAc pentapeptide (M5N)] (14). In addition, the ALs for E. cloacae AmpC expression are anhMurNAc tripeptide and anhMurNAc pentapeptide (18,19). Therefore, for AmpC expression in E. cloacae, the precursors of ALs are transported by AmpG from the periplasm to the cytosol and further processed by NagZ in the cytosol. Sufficient cytosolic AmpD activity hydrolyzes AL precursors, prevents formation of ALs, and abolishes AmpC expression. A similar AmpC expression model was also proposed for P. aeruginosa, as the components involved in AmpC expression in P. aeruginosa are functionally the same as those in E. coli (29), although the exact structures of AL precursors and ALs remain unknown in the P. aeruginosa system. The  Table S3. An equal amount of purified N-acetylglucosaminyl-N-acetylmuramyl-L-alanyl-D-glutamic acid in its reduced form (GlcNAc-MurNAc dipeptide; M2), which was only observed in total muropeptides (Fig. 5), was spiked in both wild-type KJ and the KJΔmrdA strain as an internal standard to quantify the periplasmic muropeptides. information regarding L1/L2 expression is limited compared to the that for the AmpC model, but the following findings provide some clarification for L1/L2 expression in S. maltophilia. (i) Distinct from the AmpG in the AmpC model, the permease system for AL precursor transportation is the AmpN-AmpG system in S. maltophilia (22). In our previous study, we showed that an intact ampN-ampG operon is essential for L1/L2 expression; furthermore, AmpG EC can complement the L1/L2 expression by the ampN-ampG mutant in S. maltophilia (22). Therefore, it is reasonable to speculate that the AL precursors for L1/L2 expression are generated in the periplasm, with the structural feature of anhydrodisaccharide with tripeptides, tetrapeptides, or pentapeptides, and then transported by AmpNG permease. (ii) There are two reported PBP inactivation mutants associated with L1/L2 upregulation in the S. maltophilia model: one is the mrcA mutant (24), and the the other is the mrdA mutant in or study. A critical distinction between the two mutants is the contribution of NagZ to L1/L2 upregulation. Basal L1/L2 expression in the ⌬mrcA mutant is NagZ independent (25), but such expression in the ⌬mrdA mutant is NagZ dependent. We suggest that there are at least two different ALs, AL1 and AL2, for L1/L2 expression in S. maltophilia. AL1 is NagZ dependent, while AL2 is not (25). In our study, we further linked AL1 to anhMurNAc tetrapeptides in a mutat ⌬mrdA background, and the AL1 precursor was found to be GlcNAc-anhMurNAc tetrapeptides (M4N), which accumulate in th periplasm of strain KJΔmrdA cells. Here, we still cannot immediately rule out the possibility that anhMurNAc tripeptide and anhMurNAc pentapeptide are the functional NagZdependent ALs for L1/L2 expression of S. maltophilia, as in the AmpC model for E. cloacae (9,19). Nevertheless, we are convinced that anhMurNAc tetrapeptides, rather than anhMurNAc tripeptide and anhMurNAc pentapeptide, are the key ALs for muan ⌬mrdA-mediated L1/L2 expression.
PBP2 possesses TPase, which cross-links the adjacent stem pentapeptides from different strands and maintains the bacterial rod shape. Consistent with the deletion study findings in E. coli (37) and P. aeruginosa (38), the mrdA deletion mutant of S. maltophilia displayed a round morphology (Fig. 2B), supporting that PBP2 functions mainly in cross-linking of PG during the bacterial elongation stage. The substrate for PBP2 is the D-Ala-D-Ala moiety of pentapeptides of a PG monomer. In contrast, the terminal D-Ala of pentapeptides is also a target for LMW-PBP action, as LWM-PBP are a type of DD-CPase, which removes the terminal D-Ala from pentapeptides in PG. Because the terminal D-Ala is a substrate for both TPase and DD-CPase, PBP2 inactivation may spare more substrates for DD-CPase. This may explain why the total and periplasmic M4N muropeptides in mrdA mutants were much more abundant than those in wildtype KJ (Fig. 5).
The results of this study led to the development of a model of muropeptide turnover and L1/L2 expression in a mutant ⌬mrdA background (Fig. 7). Inactivation of PBP2 increases the accumulation of M4N in the periplasm. M4N is transported into the cytosol via the AmpN/AmpG permease system. AmpD I hydrolyzes the imported M4N into GlcNAc-anhMurNAC and free tetrapeptide, which are typically recycled into UDP-MurNAc pentapeptide for use in PG monomer biosynthesis. When AmpD I activity is saturated by M4N, the surplus M4N is further processed by NagZ to form the exact AL, anhMurNAc tetrapeptides, to induce L1/L2 expression in the presence of a functional AmpR. When excess AmpD I activity is exogenously introduced, M4Ns are completely hydrolyzed by AmpD I and no ALs are formed, explaining why no ␤-lactamase activity is detectable in strain KJΔmrdA(pAmpDI). Furthermore, the CreBC TCS of S. maltophilia shows a low response to the muropeptide of M4N. Collectively, strain ⌬mrdA-mediated L1/L2 expression depends on functional AmpG, AmpR, and NagZ, is constrained by AmpD I activity, and is less related to the CreBC TCS. This is supported by our ␤-lactamase activity determinations (Fig. 4) and muropeptide profiles (Fig. 5 and 6).

MATERIALS AND METHODS
Construction of in-frame deletion mutants. S. maltophilia in-frame deletion mutants were created using a double-crossover homologous recombination system as previously described (23). The recom-binant plasmids pΔmrcB, pΔpbpC, pΔmrdA, pΔdacB, and pΔdacC were prepared for mutant constructions. Two DNA fragments targeting the 5=-terminus and 3=-terminus of the mutated PBP gene were  (Table S4). These PCR amplicons were digested and subsequently cloned into pEX18Tc (39). The recombinant plasmid mobilization, transconjugant selection, and mutant confirmation steps were performed as described previously (23). The double mutant was constructed from the single mutant sequentially through the same procedure.
Determination of ␤-lactamase activity. ␤-Lactamase induction was studied using a published procedure (25) with nitrocefin (100 M) as the substrate. Enzyme activity was calculated by using a molar absorption coefficient for nitrocefin of 20,500 M Ϫ1 cm Ϫ1 at 486 nm. The specific activity was expressed as the nanomoles of nitrocefin hydrolyzed per minute per milligram of protein. The protein content was determined with the Bio-Rad protein assay reagent, with bovine serum albumin as a standard. All experiments were performed in triplicate.
Bacterial growth. Overnight cultures were inoculated into fresh LB broth with an initial OD 450 of 0.15. Cell growth was monitored by recording the OD 450 at an interval of 3 h. Cells from the time points of 6, 15, and 21 h were serially diluted and plated in duplicate on LB agar. After overnight incubation at 37°C, the number of colonies formed was counted and recorded.
Scanning electron microscopy. The preparation of bacterial specimens was carried out as previously described (40). Then, bacterial morphology was examined by using a high-resolution FEI Inspect S scanning electron microscope.
Preparation of muropeptides. (i) Total muropeptides. Total muropeptides were isolated as described previously with some modifications (41). A 100-ml mid-logarithmic-phase bacterial culture was harvested by centrifugation, resuspended in 3 ml of phosphate-buffered saline buffer, and dropped into a boiling 6-ml 6% SDS solution. After boiling for 15 h, the murein saccule pellets were collected by ultracentrifugation (40,000 ϫ g for 20 min at room temperature), washed free of SDS with Milli-Q water, and digested with pronase (100 g/ml in 10 mM Tris-HCl [pH 7.2], 0.06% [wt/vol] NaCl) for 4 h at 60°C. The reaction was inactivated by boiling the sample for 30 min. Samples were pelleted by ultracentrifugation (40,000 ϫ g for 20 min at room temperature), resuspended in 200 l of 50 mM sodium phosphate buffer (pH 4.9), and digested with muramidase (40 g/ml) for 16 h at 37°C. Muramidase digestion was stopped by incubation at 100°C for 10 min. The coagulated proteins were removed by centrifugation at 14,000 rpm for 10 min. The muropeptides in supernatant were reduced by using 100 mM sodium borate and 20 mg/ml sodium borohydride in 0.5 M borate buffer (pH 9.0) for 3 h. Finally, the sample was adjusted to pH 3 to 4 by using 50% (vol/vol) orthophosphoric acid.  M4N is transported from the periplasm into the cytosol by the AmpN/AmpG permease system. The imported M4N is hydrolyzed into GlcNAc-anhMurNAC and free tetrapeptide by AmpD I , and the resulting products are further recycled into UDP-MurNAc pentapeptide, which is introduced into the PG monomer biosynthetic pathway. Once the AmpD I activity is saturated, surplus M4N is processed by NagZ, generating the AL, anhMurNAc tetrapeptide. The AL binds to the transcriptional regulator AmpR and converts AmpR into an activator for L1 and L2 expression.
(ii) Periplasmic muropeptides. Periplasmic muropeptides were prepared as described previously with some modifications (42). Cells from 30-ml mid-logarithmic-phase bacterial cultures were harvested by centrifugation. The cell pellets were washed twice with phosphate-buffered saline buffer, suspended in 1 ml of hypertonic solution (40% sucrose in 10 mM sodium phosphate buffer [pH 7.4]), and then incubated on ice for 20 min. After centrifugation at 12,000 rpm for 10 min, cell pellets were resuspended into 200 l of hypotonic solution (10 mM sodium phosphate buffer [pH 7.4]) and incubated on ice for an additional 20 min. Cells were pelleted at 12,000 rpm for 10 min, and the supernatant was labeled as the periplasmic fraction. The periplasmic muropeptides were reduced by 100 mM sodium borate and 20 mg/ml sodium borohydride in 0.5 M borate buffer (pH 9.0) for 3 h. Finally, the sample was adjusted to pH 3 to 4 with 50% (vol/vol) orthophosphoric acid.
(iii) M2 muropeptides. The total muropeptides samples were pooled and subjected to LC separation/purification to collect individual muropeptide species. The Waters 600 HPLC system consists of a 996 photodiode array detector, 600 solvent pump, and 600 controller and was used to separate the muropeptides. The mobile phases A and B were ultrapure water and acetonitrile, respectively, and both were added with 0.1% (vol/vol) trifluoroacetic acid. The LC elution condition was slightly different based on the pump pressure of the HPLC system. After balancing the Unitary C18 column (100 Å, 5 m, 10 mm by 250 mm) with 5.0% mobile phase B for 30 min, a 2.0-ml sample was injected into the sample loop. The elution started with 5.0% mobile phase B for 15.0 min, which was then raised to 14.0% mobile phase B in 30.0 min, and the system was washed with 95.0% mobile phase B for 10.0 min. The fractions were collected every 30 s. The purity of each fraction was checked with a Q Exactive Plus hybrid quadrupoleorbitrap mass spectrometer (Thermo Fisher Scientific, USA). M2 was dissolved in ultrapure water and the stock was held at Ϫ80°C until use for standards for the quantitative analysis of periplasmic muropeptides.
LC-MS analysis of muropeptides. (i) Total muropeptides. The total muropeptides sample was centrifuged at 14,000 rpm for 30 min, and the supernatant was diluted 10-fold with H 2 O for subsequent LC-MS analysis.
A Dionex UltiMate 3000 UHPLC system coupled with a Q Exactive Plus hybrid quadrupole-orbitrap mass spectrometer (Thermo Fisher Scientific, USA) was used for LC-MS analysis. The mobile phases A and B were ultrapure water and acetonitrile, respectively, and both were added with 0.1% (vol/vol) formic acid. After balancing the Acquity UPLC CSH C18 column (130 Å, 1.7 m, 2.1 mm by 100 mm) with mobile phase A, 3 l of sample was injected into the sample loop. The percentage of the mobile phase B was increased to 3.0% within 1 min, then raised to 12.0% in 7 min. Within 4 min, mobile phase B was raised to 13.5% and then raised rapidly to 90.0% mobile phase B within 1 min and left for 3 min. The column temperature was controlled at 52°C in the whole analysis program. The capillary temperature of Q Exactive Plus was set at 360°C. The Thermo HESI-II probe was used as the ion source, and the spray voltage was set at 3.5 kV, with the auxiliary gas temperature set at 406°C. For MS1 acquisition, the resolution was set with 35,000, and the mass range was from 450 to 2,000 m/z at positive mode. Data-dependent tandem MS (MS/MS) was then performed for further structural validation of the peptidoglycans. A full MS spectrum (450 to 2,000 m/z, 35,000 resolution) with a 100-ms maximum injection time followed by MS1-dependent MS/MS acquisition which consisted of 10 scans in a cycle with a Δm/z 3 isolation window (17,500 resolution) with 120-ms maximum injection time. The top 10 most abundant peaks in the MS1 scan were sequentially isolated and fragmented by higher-energy collision dissociation (HCD) using nitrogen as the collision gas with a normalized collision energy of ϳ17%. All the MS spectra were recorded in profile mode.
(ii) Periplasmic muropeptides. The periplasmic muropeptide samples were centrifuged at 14,000 rpm for 30 min, and 20-l volumes of supernatants were mixed with 2 l of purified M2 as the internal standard for subsequent LC-MS analysis.
The instruments used to analyze periplasmic muropeptides sample were the same as the total muropeptides sample, whereas the elution condition for LC was different. The same CSH C18 column first balanced with 2.0% mobile phase B, and 11-l sample volumes were injected into the sample loop. The percentage of the mobile phase B was increased to 5.0% from 0.5 to 1.0 min and kept at 5.0% for 1.0 min, then raised to 8.0% in 2.0 min. It was increased to 9.5% B in the next 5.0 min, up to 11.0% B within 1.5 min, and then raised rapidly to 90.0% B within 1.5 min and left for 3 min. The capillary temperature of Q Exactive Plus was set at 253°C. Other instrument settings were the same as those for the LC-MS analysis of total muropeptides. Data processing. A reference table of muropeptides was established for peak assignments in the following manner: (i) a core m/z list of all possible monomer, dimer, and trimer muropeptides containing different numbers of stem peptides was first generated and was based on the main structures of monomers of S. maltophilia cell wall. (ii) The list was further expanded by including possible modifications of muropeptides, e.g., reduction, stem peptide cross-linking, glycosidic bond cleavage between N-acetylglucosamine and N-acetylmuramic acid, during the sample preparation. (iii) A list of m/z values for possible in-source fragmentation products was added. The theoretical m/z and isotopic patterns of all the proposed muropeptides in the reference table were determined with the ecipex package and were used to annotate the ion signals obtained by LC-MS (43).
An automatic muropeptide analysis platform based on the R programing language was developed and used to determine the relative levels of each muropeptide species. In short, MSconvert, a free software published by ProteoWizard, was implemented to convert the .raw file to an .mzXML file so as to deconvolute the Thermo RAW file in R. We utilized the mzR package to read the raw mass spectra for peak picking in each scan (44), where ions of intensities lower than 1.0 ϫ 10 5 obtained by Q Exactive Plus mass spectrometers in our experiments were neglected. Ions of the same m/z (tolerance, 0.02 m/z), similar isotope intensity (tolerance, 20%), and same charge state that continuously appeared in more than two successive scans were considered identical. Ions of different retention times were distinguished as different molecular species. Ions were annotated based on their m/z and isotopic patterns, as suggested in the reference muropeptides table. Intensities of identical muropeptides of different charge states and their products of in-source fragmentation that appeared at the same retention times were summed up. The areas under the curves (AUCs) of each muropeptide species in the ion chromatograms were calculated and exported as composition ratios. Those AUC values for reduced muropeptides were combined with AUCs for their corresponding parent muropeptides. For total muropeptide analysis, the ratios of the AUCs of each muropeptide and the total AUC of all annotated muropeptides were calculated to determine the relative compositions of each annotated species, whereas for periplasmic muropeptides analysis the ratio of the AUC of each muropeptide to the AUC of rM2 was calculated to determine the quantity of each annotated species. All the LC-MS data reported in the manuscript were manually extracted and inspected meticulously to validate the efficacy of the automatic platform.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/ mSystems.00077-17.   The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.