1. Microbiology and Infectious Disease
Download icon

Plasticity of Escherichia coli cell wall metabolism promotes fitness and antibiotic resistance across environmental conditions

  1. Elizabeth A Mueller
  2. Alexander JF Egan
  3. Eefjan Breukink
  4. Waldemar Vollmer
  5. Petra Anne Levin  Is a corresponding author
  1. Washington University in St. Louis, United States
  2. Newcastle University, United Kingdom
  3. Utrecht University, Netherlands
  • Cited 1
  • Views 1,675
  • Annotations
Cite this article as: eLife 2019;8:e40754 doi: 10.7554/eLife.40754

Abstract

Although the peptidoglycan cell wall is an essential structural and morphological feature of most bacterial cells, the extracytoplasmic enzymes involved in its synthesis are frequently dispensable under standard culture conditions. By modulating a single growth parameter—extracellular pH—we discovered a subset of these so-called ‘redundant’ enzymes in Escherichia coli are required for maximal fitness across pH environments. Among these pH specialists are the class A penicillin binding proteins PBP1a and PBP1b; defects in these enzymes attenuate growth in alkaline and acidic conditions, respectively. Genetic, biochemical, and cytological studies demonstrate that synthase activity is required for cell wall integrity across a wide pH range and influences pH-dependent changes in resistance to cell wall active antibiotics. Altogether, our findings reveal previously thought to be redundant enzymes are instead specialized for distinct environmental niches. This specialization may ensure robust growth and cell wall integrity in a wide range of conditions.

Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed (see decision letter).

https://doi.org/10.7554/eLife.40754.001

Introduction

The growth and survival of single-celled organisms relies on their ability to adapt to rapidly changing environmental conditions. A commensal, pathogen, and environmental contaminant, Escherichia coli occupies and grows in diverse environmental niches, including the gastrointestinal tract, bladder, freshwater, and soil. In the laboratory, the bacterium’s flexibility in growth requirements is reflected in robust proliferation across a wide range of temperature, salt, osmotic, pH, oxygenation, and nutrient conditions (Ingraham and Marr, 1996).

The physiological adaptations that permit growth and survival across environmental conditions are not yet well understood, particularly for extracytoplasmic processes. Due to the discrepancy in permeability between the plasma and outer membrane (Rosenbusch, 1990), the periplasmic space of Gram-negative bacteria is sensitive to chemical and physical perturbations, including changes in salt, ionic strength, osmolality, and pH. Notably, upon mild environmental acidification, the periplasm assumes the pH of the extracellular media (Slonczewski et al., 1981; Wilks and Slonczewski, 2007). Although mechanisms that contribute to cytoplasmic pH homeostasis have been described in detail (Castanie-Cornet et al., 1999; Castanié-Cornet et al., 2010), comparatively little is known about the quality control mechanisms that preserve proper folding, stability, and activity of key proteins in the periplasm.

The peptidoglycan (PG) cell wall and its synthetic machinery are among the fundamental constituents of the periplasm that must be preserved across growth conditions. Essential for viability among most bacteria, PG is composed of glycan strands of repeating N-acetylglucosamine and N-acetylmuramic acid disaccharide units crosslinked at peptide stems (Vollmer et al., 2008). Beyond providing a force necessary to resist turgor pressure, the cell wall maintains cell shape, and components of the cell envelope serve as a major interface for cell-cell and cell-host interactions (Typas et al., 2012; McDonald et al., 2005). As an essential process, PG synthesis is also the principle target of several classes of antibacterial agents, including β-lactam (e.g. penicillin) and glycopeptide (e.g. vancomycin) antibiotics.

PG precursors are assembled in the cytosol and translocated across the inner membrane into the periplasm, where cell wall synthases construct the PG network through a series of glycosyltransferase (glycan polymerizing) and transpeptidase (peptide crosslinking) reactions. PG synthases include bifunctional class A penicillin binding proteins (PBPs), as well as monofunctional transpeptidases (class B PBPs) and monofunctional glycosyltransferases of the shape, elongation, division, and sporulation (SEDS) protein families (Sauvage et al., 2008; Meeske et al., 2016; Cho et al., 2016; Taguchi et al., 2019). LD-transpeptidases synthesize non-canonical LD-crosslinks between peptide stems. They are predominately active during PG remodeling during stationary phase growth in E. coli (Pisabarro et al., 1985; Magnet et al., 2007) and are required under severe envelope stress (Morè et al., 2019). In addition to synthases, a series of periplasmic cell wall hydrolases and autolysins—including DD-carboxypeptidases, DD and LD-endopeptidases, lytic transglycosylases, and amidases—are required to accommodate nascent strand insertion for expansion of the PG network, create substrate binding sites, and separate cells during the final stages of cytokinesis (Typas et al., 2012). These enzymes may also play a role activating synthases to ensure cell wall integrity (Lai et al., 2017).

The periplasmic steps of PG synthesis and remodeling exhibit high levels of enzymatic redundancy, the function of which remains unclear. While the cytoplasmic steps of PG precursor biogenesis in E. coli have nearly a 1:1 stochiometric ratio between reactions and enzymes (12 reactions: 14 enzymes), over 36 enzymes can carry out the nine reactions that take place in the periplasm (Pazos et al., 2017). Moreover, with the exception of the SEDS glycosyltransferase/bPBP pairs RodA/PBP2 and FtsW/PBP3 required for lateral expansion of the cell wall and cell division, respectively (Meeske et al., 2016; Cho et al., 2016), the remaining periplasmic cell wall enzymes appear to be nonessential. Inactivation of an individual enzyme—and in some cases, even multiple enzymes in the same class—often fails to confer discernable growth or morphological phenotypes under standard culture conditions (Singh et al., 2012; Nelson and Young, 2000; Heidrich et al., 2002; Suzuki et al., 1978; Heidrich et al., 2001). Although technological breakthroughs have aided in the identification of cell wall metabolic genes encoding proteins with similar functions (e.g. Peters et al., 2016a), elucidating the potential fitness benefit to redundancy has proven challenging.

One model to account for the apparent redundancy of periplasmic cell wall proteins is that enzymes within a given class may be specialists for distinct environmental niches, thereby allowing bacteria to cope with the diverse chemical and physical properties that might affect protein stability and function in this compartment (Pazos et al., 2017). In support of this hypothesis, several groups have identified cell wall enzymes that have increased activity in acidic media. Peters and colleagues demonstrated that E. coli carboxypeptidase PBP6b plays a key role in maintenance of cell morphology during growth at pH 5.0 (Peters et al., 2016b), while Castanheira et al. identified a PBP3 homolog in Salmonella Typhimurium that is preferentially involved in septation at low pH (Castanheira et al., 2017). Similarly, the lytic transglycosylase MltA exhibits maximal activity in acidic conditions in vitro (van Straaten et al., 2005), although whether this property is relevant in vivo remains unknown.

In light of these findings, we hypothesized that loss of an enzyme specialized for a particular environmental niche may produce a condition-specific growth defect through impaired cell wall integrity, allowing us to take a systems-level approach to identifying enzymes with differential roles in growth in vivo. In screening 32 mutants across six classes of nonessential periplasmic cell wall enzymes, we determined that a subset of these enzymes is differentially required for fitness across pH environments. We find that disruptions in the activity of cell wall synthases PBP1a and PBP1b conferred fitness defects in opposing pH ranges that can be attributed in part to pH-dependent differences in enzymatic activity. We further demonstrate that synthase specialization has consequences for intrinsic resistance to β-lactam antibiotics in nonstandard growth conditions.

Results

Identification of pH specialist cell wall synthases and hydrolases

To determine the contribution of individual cell wall enzymes to pH-dependent growth, we cultured strains harboring deletions in genes encoding each of three class A PBPs, six LD-transpeptidases, five carboxypeptidases, four amidases, nine lytic glycosyltransferases, and six endopeptidases to mid-exponential phase (OD600 ~0.2–0.6) in buffered LB media (pH 6.9) then sub-cultured them into fresh LB buffered to pH 4.8, 6.9, or 8.2 for growth rate analysis. These pH values were chosen as representative, physiologically relevant conditions E. coli cells encounter in the lower GI tract (pH 5–9) or urine (pH 4.5–8) (Watson et al., 1972; Henderson and Palmer, 1912). Preliminary hits were identified by a significant (>5%) decrease in early exponential phase (OD6000.005–0.1) mass doublings per hour (DPH) in at least one pH condition compared to the parental strain. Representative growth curves and fits are presented in Figure 1—figure supplement 1. Mutants exhibiting significant growth defects at one or more pH values were re-tested across an expanded set of pH conditions for validation; those that displayed consistent growth defects across a discrete range of pH values were classified as pH-sensitive mutants (Figure 1—figure supplement 2; Figure 2; Supplementary file 3).

Figure 1 with 3 supplements see all
Identification of pH specialist cell wall enzymes.

Mutants in genes encoding for non-essential class A PBPs (A), endopeptidases (B), and lytic transglycosylases (C) were screened for growth defects compared to the parental strain in LB media buffered to pH 4.8, 6.9, or 8.2. Bars depict mean mass doublings per hour ±SD of three independent biological replicates. NG denotes ‘no growth’ observed throughout the course of the experiment (20 hr). Cartoons depict enzymatic activity of the indicated enzyme class with GTase and DD-TPase referring to glycosyltransferase and DD-transpeptidase activity, respectively. Asterisks denote a significant >5% growth defect as determined by a one-way ANOVA corrected for multiple comparisons as follows: **, p<0.01; ***, p<0.001; ****p<0.0001. Mean ± SD values for mutants in this figure and all additional mutants tested can be viewed in Supplementary file 3. Representative growth curves, fits, and source data from the class A PBP mutants can be viewed in Figure 1—figure supplement 1 and Figure 1—source datas 13.

https://doi.org/10.7554/eLife.40754.002
Figure 2 with 2 supplements see all
pH-dependent growth requires class A PBP activity.

(A–C) Mean mass doublings per hour and transformed percent parental growth for ∆mrcA (PBP1a; EAM543) and ∆mrcB (PBP1b; EAM546) deletions compared to parental strain (MG1655) in LB media buffered from pH 4.8–8.4 Significance was determined by an unpaired t-test corrected for multiple comparisons using the Holm-Sidak method. Error bars represent SD from six independent biological replicates. (D) Growth rate analysis of cells defective for LpoA (EAM657) and LpoB (EAM659) cultured in buffered LB at pH 4.8, 6.9, or 8.0. Bars represent mean mass doublings per hour ± SD from three independent biological replicates. Asterisks denote significance as determined by a one-way ANOVA corrected for multiple comparisons as follows: ****, p<0.0001. Growth of these mutants across an expanded set of pH conditions can be viewed in Figure 2—figure supplement 2. (D) Complementation analysis of PBP1b variants synthesized from a plasmid (pUM1Bα, pUM1Bα*, pUM1BTG*α, or pUM1BTG*α*) and induced with 5 μM IPTG in the ΔmrcB (EAM696) background in buffered LB at pH 4.8 and 6.9. Bars represent mean mass doublings per hour ± SD from three independent biological replicates. NG denotes ‘no growth’ observed throughout the course of the experiment (20 hr).

https://doi.org/10.7554/eLife.40754.009

Collectively, five mutants met these stringent criteria and displayed significant pH-dependent reductions in DPH. We observed both acid-sensitive and alkaline-sensitive mutants across three enzymatic classes. Strikingly, loss of the bifunctional synthase PBP1b (mrcB) abolished growth at pH 4.8—over half a pH unit lower than the growth restrictive condition for the parental strain (Figure 1A; Figure 1—figure supplement 3A). Growth of the ΔmrcB mutant was also significantly attenuated (10–25% defect in DPH) at pH values between 5.1–5.9 but was indistinguishable from the parental strain in neutral and alkaline pH (Figure 2A,C). Pre-conditioning the mutant in acidic media (pH 5.5) did not abrogate the growth rate defect (Figure 1—figure supplement 3A), indicating that steady-state pH—rather than pH shock—underlies the defect in DPH. Mutants defective in production of lytic transglycosylase MltA and endopeptidase MepS (spr) also exhibited a specific, albeit less severe, defect in DPH in acidic media: their growth was attenuated by 5–15% compared to wild type cells at pH values at or below 6.2 (Figure 1B,C; Figure 1—figure supplement 2). Consistent with a role in acid tolerance, MltA was previously shown to have elevated enzymatic activity in acidic conditions in vitro (van Straaten et al., 2005), and cells defective for PBP1b or MepS production exhibit reduced colony growth on acidic agarose plates (Nichols et al., 2011).

We also identified two alkaline-sensitive mutants. Loss of the bifunctional synthase PBP1a (mrcA) and the lytic transglycosylase MltG (yceG) impaired, but did not abolish, growth specifically in neutral and alkaline media (Figure 1A,C; Figure 1—figure supplement 2; Figure 2A,C). Loss of MltG was associated with a greater magnitude and range of growth impairment (pH 6.2–8.4 compared to pH 6.5–8.2 for ΔmrcA). Both mutants’ growth was restored to wild-type levels in acidic media (pH <6.0) and at pH 9.0 (Figure 1—figure supplement 3B).

Individual deletions in the genes encoding for the six LD-transpeptidases, five carboxypeptidases, and four amidases failed to confer any pH-dependent defects in DPH at any pH tested, consistent with either a limited role of these enzymes in exponential phase growth (Pisabarro et al., 1985; Magnet et al., 2007) or additional layers of redundancy (Supplementary file 3).

Class A PBP activity ensures fitness across a wide pH range

Given their opposing impact on DPH under acidic and alkaline conditions, we elected to focus further efforts on understanding the contribution of the bifunctional class A PBPs PBP1a and PBP1b to growth across a range of pH conditions. An accumulating body of evidence suggests the class A PBPs play overlapping, and potentially redundant, roles in PG synthesis during growth in standard culture conditions (i.e. nutrient rich, neutral pH growth medium aerated at 37°C) (Cho et al., 2016; Yousif et al., 1985). Indeed, E. coli requires at least one of these enzymes for viability during growth in standard culture conditions (Suzuki et al., 1978).

Based on their disparate pH-dependent growth defects, we hypothesized that PBP1a and PBP1b are specialized synthases whose activity is essential for maximal growth in distinct pH environments. Consistent with this model, cells defective in PBP1a (ΔmrcA) and PBP1b (ΔmrcB) displayed defects in DPH at discrete, non-overlapping pH ranges (Figure 2A–C). Loss of PBP1c (pbpC), a third class A PBP with an unclear role in cell wall metabolism (Schiffer and Höltje, 1999), did not result in a defect in DPH at any pH tested alone or combination with cells defective for PBP1a or PBP1b (Figure 2—figure supplement 1), indicating this enzyme does not play a role in pH-dependent growth under the conditions tested here.

We next sought to test whether PBP transpeptidase and/or glycosyltransferase activity were required for fitness across pH conditions, as opposed to an indirect, structural role for these enzymes in the formation of cell wall synthesis complexes (Müller et al., 2007; Bertsche et al., 2006). To test this, we took advantage of two sets of mutants: 1) deletions in lpoA and lpoB—genes encoding outer membrane lipoprotein cofactors required for activity, but not expression or stability, of PBP1a and PBP1b, respectively (Typas et al., 2010; Paradis-Bleau et al., 2010; Egan et al., 2014; Lupoli et al., 2014), and 2) point mutations that inactivate PBP1b transpeptidase and/or glycosyltransferase activity but do not impact stability (Meisel et al., 2003).

Implicating PBP activity in growth across pH environments, loss of the cofactors LpoA and LpoB mimicked the pH-dependent growth defects of loss of the enzymes themselves. Analogous to cells defective for PBP1b, deletion of lpoB prevented growth at pH 4.8. Likewise, loss of PBP1a’s cofactor LpoA led to a significant defect in DPH between pH values 5.9–8.2 (Figure 2D; Figure 2—figure supplement 2). Interestingly, loss of either Lpo protein did not perfectly recapitulate loss of its cognate class A PBP (Figure 2—figure supplement 2; Figure 2A–C), suggesting the presence of additional relevant regulators in vivo. Complementation analysis of PBP1b variants at acidic pH further bolstered the conclusion that activity is required for pH-dependent growth. As expected, production of wild-type PBP1b in trans restored growth of the ∆mrcB mutant at pH 4.8; however, production of PBP1b variants rendering the transpeptidase (S510A, TP*), glycosyltransferase (E233Q, GT*), or both enzymatic activities inactive (TP*GT*) failed to complement growth (Figure 2E). It should be noted that the mutation in the glycosyltransferase active site (E233Q) previously has been shown to attenuate transpeptidase activity by 90%, consistent with observations that PBP transpeptidase activity cannot be assayed in vitro in the absence of functional glycosyltransferase activity (Egan et al., 2014; Bertsche et al., 2005; Terrak et al., 1999; Born et al., 2006). Thus, although our data demonstrate that transpeptidase activity is critical for pH-dependent growth, we cannot discern whether glycosyltransferase activity alone is required.

Class A PBP activity promotes cell wall integrity across pH environments

Although these findings establish PBP1a and PBP1b activity as essential for optimal fitness across a wide pH range, it remained unclear whether these mutants’ pH-dependent defects in DPH in bulk culture were due to reduction in growth across the population (i.e. decreased rate of mass accumulation and cell expansion) or lysis of a fraction of cells in the population. To differentiate between these two mechanisms, we inoculated early exponential phase (OD600 ~0.05–0.1) wild type or mutant cells cultured at pH 6.9 on to agarose pads buffered to pH 4.5 or pH 8.0 and examined cells for incorporation of the dye propidium iodide (PI), which permeates cells with compromised membranes, by microscopy.

Consistent with a lytic origin, extensive PI incorporation was observed at one hour post-shift for PBP1b and PBP1a defective cells that underwent acid (pH 6.9 to pH 4.5) or alkaline (pH 6.9 to pH 8.0) shock, respectively (Figure 3C,D). To confirm bona fide lysis, we transformed a plasmid expressing gfp in to the mutants and measured PI incorporation and loss of cytoplasmic GFP concurrently. Without exception, PI+ cells lacked cytoplasmic GFP signal (Figure 3—figure supplement 1).

Figure 3 with 1 supplement see all
Cells defective for class A PBPs lyse upon exposure to non-permissive pH conditions.

(A–B) Single cell elongation rates for wild-type (MG1655), ∆mrcA (EAM543), and ∆mrcB (EAM546) cells during growth on agarose pads at pH 4.5 (A; n = 134, 81, and 155 cells) and pH 8.0 (B; n = 386, 257, and 246 cells). Rates were determined in the MATLAB-based program SuperSegger as described in the methods. Error bars represent SD. (C–D) Micrographs depicting representative images of propidium iodide incorporation in ∆mrcA (D, left) and ∆mrcB (C, left) mutants at t = 0 or 60 min post indicated pH shift. Scale bar represents 5 μm. Cell viability curves for wild-type, ∆mrcA (PBP1a), and ∆mrcB (PBP1b) strains after acidic (D, right) or alkaline pH (C, right) shift as indicated. Cell death was determined by uniform cytoplasmic staining with propidium iodide. Markers indicate mean percent viability ± SD of three biological replicates. Greater than 100 cells were analyzed for each strain at each time point per replicate.

https://doi.org/10.7554/eLife.40754.012

Time lapse imaging of cells following pH shift shed light on lysis kinetics: upon pH downshift, ∆mrcB cells began to incorporate PI by 30 min (~15% cells labeled), with ~95–100% of cells labeled by two hours post-shift. Negligible cell death was observed for the parental strain or for cells defective for PBP1a during equivalent acid shock (Figure 3C). Conversely, up to 15% of ∆mrcA cells underwent lysis an hour following alkaline shift to pH 8.0 (Figure 3D). Although we did observe reduced rates of lysis for the ∆mrcA mutant at later time points, this recovery was not recapitulated in liquid culture and thus was not investigated further (Figure 3—figure supplement 1D). ∆mrcB and wild type cells exhibited minimal (<5%) or negligible cell death, respectively, in response to alkaline shift on agarose pads. Prior to cell lysis, single cell growth rate of both mutants and the parental strain were similar at each pH condition (Figure 3A,B), indicating lysis is the sole determinant of decreased DPH observed in bulk culture.

In addition to displaying differential lysis kinetics under their respective non-permissive pH conditions, PBP1a and PBP1b mutants also differed in apparent lysis mechanism. Time lapse imaging during acid shock revealed a high fraction of ∆mrcB cells lysed during division, often from a bulge emanating at the septum (Figure 4A, white arrows; Figure 4—video 1). Scanning electron microscopy confirmed the bulges were coincident with the septum in this mutant (Figure 4C). To quantitate the lytic phenotype of the mutants, we categorized the lysis mechanism into three groups: septal bulge, non-septal bulge (including polar and peripheral bulging), and lysis not associated with visible bulging. Septal bulging was determined based on association of the bulge origin with the visible constriction site by phase contrast microscopy. Indeed, this analysis confirmed our observation: 55% of ∆mrcB mutants lysed at the septum following pH downshift with the remaining fraction associated with a non-septal bulge (17%) and no bulge (28%) (Figure 4D). In contrast, lysis of the ∆mrcA mutant during growth in alkaline pH was not associated with division, and instead, lysis of ~60% of the cells was coincident with a non-septal bulge, typically emanating from the periphery (Figure 4B,D; Figure 4—video 2). These differences may reflect distinct localization preferences of the enzymes (Bertsche et al., 2006; Banzhaf et al., 2012) or disparate weak regions in the PG across pH conditions.

Figure 4 with 2 supplements see all
Distinct lytic phenotypes for cells defective for PBP1a and PBP1b upon pH shift.

(A–B) Representative phase contrast frames of time lapse imaging of ∆mrcB (PBP1b; EAM546) and ∆mrcA (PBP1a; EAM543) mutants upon acidic (A) or alkaline (B) pH shift, respectively, as compared to the parental strain (MG1655). White arrows indicate membrane bulges. Scale bar denotes 5 μm. Full videos can be viewed in Figure 4—videos 1 and 2. (C) Representative scanning electron microscopy micrographs for ∆mrcB (PBP1b; EAM546) mutant shifted to either pH 6.9 or pH 4.5 for two hours prior to fixation. Scale bar represents 1 μm. (D) Quantification of lysis phenotype between mutants. Lytic terminal phenotype was categorized into three groups: lysis via septal bulge, non-septal bulge, or no bulge. Determination of lytic phenotype was based on the frames preceding propidium iodide incorporation (time step = 3 min). Micrographs (top to bottom) depict representative images of no bulge, septal bulge, and non-septal bulge, respectively with arrows (scale bar = 2 μm). At least 50 cells across at least two independent biological replicates were assessed (ΔmrcA, n = 128; ΔmrcB, n = 278 cells). Bars are subdivided based on percent lytic phenotype in each mutant.

https://doi.org/10.7554/eLife.40754.014

PBP1a localization and activity are impaired at low pH

Although our data support a model in which class A PBP activity is differentially required for cell wall integrity across pH environments, the mechanistic basis for pH specialization remained unclear. To interrogate this, we compared the production, localization, and biochemical activity of PBP1a and PBP1b as a function of pH.

We predicted differential PBP production across pH conditions may contribute to the enzymes’ specialization, as has previously been shown for acid-specialist carboxypeptidase PBP6b (Peters et al., 2016b). Consistent with previous proteomic mass spectrometry data (Schmidt et al., 2016), bulk protein levels of both class A PBPs were modestly reduced in acidic media (Figure 5C; Figure 5—figure supplement 1). However, pH-dependent differences in production did not suggest a correlation between either class A PBP’s levels and its contribution to fitness in a particular pH environment.

Figure 5 with 4 supplements see all
pH-dependence of aPBP localization, production and activity.

(A, B) Representative micrographs illustrating aPBP localization in strains expressing Plac::gfp-mrcA (EAM707) and Plac::gfp-mrcB (EAM718) grown in AB minimal media supplemented with 0.2% maltose and 250 μM IPTG at pH 5.0, 7.0 and 8.0. Scale bar indicates 2 μm. (C) Western blot depicting representative biological replicates of PBP1a, PBP1b, and FtsZ levels in wild-type cells (MG1655) cultured at pH 4.8, 6.9, and 8.2. Percent aPBP level (using FtsZ levels as an internal loading control and normalized to pH 6.9 values) across pH conditions is shown to the right. Ponceau staining for total protein levels can be viewed in Figure 5—figure supplement 1. (D) Representative HPLC chromatograms of muropeptide analysis. Peak 1 (black arrows), Penta-P (stems from remaining lipid II/glycan chain ends); peak 2; Tetra (GT and CPase product); peak 3, Penta (GTase product); peak 4, TetraTetra (GTase, TPase and CPase product); peak 5, TetraPenta (GTase and TPase product), peak 6, TetraTetraTetra (GTase, TPase and CPase product); peak 7, TetraTetraPenta (GTase and TPase product); peak 8, TetraTetraTetraTetra (GTase, TPase and CPase product). (E) Quantification of TPase domain activity (sum of TPase and CPase products) of PBP1A + LpoA and PBP1B + LpoB and FtsN at pH 4.8, 6.9 and 8.2. Data is the mean ± range of two replicates. Corresponding representative HPLC chromatograms are shown in D.

https://doi.org/10.7554/eLife.40754.017

We next turned to examining PBP localization across pH conditions, using functional GFP fusions to PBP1a and PBP1b produced from the attHK locus and under IPTG inducible control (Paradis-Bleau et al., 2010). Similar to previous reports (Bertsche et al., 2006; Paradis-Bleau et al., 2014), the fusion proteins exhibited discrete localization profiles at neutral pH: GFP-PBP1a localized predominantly to the cell periphery, while GFP-PBP1b was present at both the periphery and the septum. Although GFP-PBP1b localization did not noticeably differ across pH conditions, GFP-PBP1a peripheral signal was reduced at pH 5.0, and the fusion adopted an irregular clustered distribution throughout the cell body (Figure 5A,B; Figure 5—figure supplement 2). The physiological significance of this phenotype remains unclear and requires additional investigation.

To examine the effect of pH on PBP synthase activity, we first utilized an end-point assay that concurrently measures glycosyltransferase activity and transpeptidase domain activity, which are coupled in the bifunctional class A PBPs (Bertsche et al., 2005; Born et al., 2006; Egan et al., 2018). Although purified PBP1a and PBP1b exhibit biochemical activity alone in vitro, we chose to test the influence of pH on the class A PBPs in the context of their key cellular activators, including LpoA for PBP1a and LpoB and FtsN for PBP1b. The outer membrane lipoprotein activators LpoA and LpoB are required for the function of their cognate class A PBP in vivo (Typas et al., 2012; Paradis-Bleau et al., 2010), and the essential division protein FtsN can act synergistically with LpoB to enhance PBP1b glycosyltransferase activity up to 16-fold in vitro (Egan et al., 2015).

Briefly, purified enzymes and their cognate activators were solubilized into detergent micelles and incubated with [14C]lipid II precursor in buffer at pH 4.8, 6.9, and 8.2. After one hour, the resulting PG was digested into muropeptides and resolved by high performance liquid chromatography. Glycosyltransferase activity is reflected qualitatively in the proportion of [14C]lipid II utilized, resulting in a decrease in peak 1. Transpeptidase activity—including both crosslinking activity and carboxypeptidase activity—is quantified by the fraction of muropeptides with modified peptides (peaks 5–9) (Egan et al., 2015). Strikingly, at pH 4.8 PBP1a + LpoA exhibited little glycosyltransferase or transpeptidase domain activity; in this condition the majority of [14C]lipid II precursor was not polymerized (Figure 5D, peak 1). In contrast, PBP1b + LpoB + FtsN maintained similar end-point activity across all tested pH conditions (Figure 5D,E). Low PBP1a glycosyltransferase activity at pH 4.8 was confirmed in a continuous fluorescence assay, which measures the polymerization rate of a Dansyl-labeled Lipid II substrate. Consistent with previous work on PBP1b (Egan et al., 2014), in the absence of their cognate activators, both PBP1a and PBP1b exhibited reduced rates of glycosyltransferase activity in acidic conditions. However, co-incubation of PBP1b with LpoB and division protein FtsN significantly increased its polymerization rate under all pH conditions, while PBP1a exhibited poor activity even in the presence of LpoA (Figure 5—figure supplement 3).

To test whether pH-dependent changes in PBP activity reflect a change in affinity of the enzymes for their cognate lipoprotein activators, we performed surface plasmon resonance experiments in which PBP1a and PBP1b were immobilized to chips and exposed to LpoA or LpoB at various concentrations. PBP1a-LpoA and PBP1b-LpoB bound at KD values of 520 ± 49 nM and 213 ± 22 nM at pH 6.9, respectively. Both KD values were ~2 fold higher at pH 8.2 (Figure 5—figure supplement 4), but affinity values could not be determined at pH 4.8 due to significant non-specific binding of both LpoA and LpoB to the chip. Altogether, our data support a model in which PBP1a activity is reduced in acidic conditions, rendering the cell reliant on PBP1b for cell wall integrity and viability.

Low pH promotes intrinsic resistance to PBP2 and PBP3-specific β-lactams

Combined with the recent findings of other groups (Peters et al., 2016b; Castanheira et al., 2017; Montón Silva et al., 2018), our results suggest that the active cell wall synthesis machinery varies across pH environments. We hypothesized that one potential consequence of environmental plasticity in the cell wall synthesis machinery may be changes in intrinsic resistance to cell wall active antibiotics. If true, condition-dependent intrinsic resistance may have important implications for treatment of E. coli infections in host niches with variable pH (Watson et al., 1972; Henderson and Palmer, 1912). To test this model, we measured the minimum inhibitory concentration (MIC) of a panel of compounds during growth of E. coli strain MG1655 in a range of physiologically relevant pH conditions (pH 4.5–8.0). We focused on the β-lactam class of antibiotics, which often target specific PBPs at a drug’s MIC (Kocaoglu and Carlson, 2015).

In support of our hypothesis, we observed a 4 to 32-fold increase in MIC to a subset of the tested β-lactams at pH values < 6.0 (Figure 6A; Supplementary file 4). In particular, cells displayed an increase in intrinsic resistance to compounds that specifically target PBP2 and PBP3, class B PBPs that are essential for cell elongation and division, respectively (Kocaoglu and Carlson, 2015). Consistent with acidic pH conferring a protective effect on the elongation and division machinery and previous reports (Goodell et al., 1976), cells cultured in low pH media retained near-normal morphology in the presence of concentrations of the compounds that led to either filamentation (cephalexin, CEX) or cell rounding (mecillinam, MEC) at pH 7.0 (Figure 6B,C; Figure 6—figure supplement 2). The pH-dependent change in intrinsic resistance was not limited to our laboratory strain: uropathogenic E. coli isolate UTI89 (Chen et al., 2006) exhibited a comparable change in MIC to both cephalexin (CEX) and mecillinam (MEC) at low pH during growth in both broth culture and in urine (Figure 6D; Supplementary file 5). In contrast, susceptibility to non-specific β-lactams (ampicillin, AMP; amoxicillin, AMX), a class A PBP-targeting compound (cefsulodin, CFS) (Kocaoglu and Carlson, 2015), or a protein synthesis inhibitor (chloramphenicol, CH) was not strongly pH-dependent. To rule out alternative causes of antibiotic resistance, we confirmed that differences in drug susceptibility could not be attributed to pH-dependent changes in antibiotic stability, proton motive force, β-lactamase production, or outer membrane permeability (Figure 6—figure supplements 1 and 3). These findings suggest that pH-mediated plasticity in the cell wall synthesis machinery influences intrinsic β-lactam sensitivity.

Figure 6 with 4 supplements see all
Intrinsic resistance to PBP2 and PBP3-targeting β-lactams at low pH.

(A) Heat map summarizing fold change in minimum inhibitory concentrations (MIC) of antibiotics for strain MG1655 cultured in LB (pH 4.5–8.0) after 20 hr. Cells in heat map are colored based on median fold change (FC) in MIC at indicated pH compared to pH 7.0 from at least three biological replicates. Fold change values of >8 are indicated in black inside relevant cell. Untransformed median MIC values can be viewed in Supplementary file 4. Abbreviations for antibiotic names are as follows: AMP, ampicillin; AMX, amoxicillin; CFS, cefsulodin; MEC, mecillinam; DOR, doripenem; MEM, meropenem; CEX, cephalexin; AZT, aztreonam; PIP, piperacillin; CH, chloramphenicol. Predominant cellular PBP target is indicated to the left. (B) Representative micrographs of cells treated with PBP3 inhibitor cephalexin (CEX) at pH 7.0 and pH 5.5. Distribution of cell lengths at sub-MIC concentration at pH 7.0 and 5.5 (n = 303 and 250) are shown to the right. (B) Representative micrographs of cells treated with PBP2 inhibitor mecillinam (MEC) cultured at pH 7.0 or pH 5.0. Distribution of cell aspect ratios (width/length) at sub-MIC concentration at pH 7.0 and 5.0 (n = 250 and 250) are shown to the right. For both panels A and B, scale bar indicates 3 μm, and NG denotes ‘no growth’ observed at the indicated concentration of antibiotic. Error bars represent SD. Significance was assessed by a Kruskal-Wallis test with asterisks denoting significance as follows: ****, p<0.0001. (D) Fold change in minimum inhibitory concentration of E. coli strain UTI89 to cephalexin (CEX) and mecillinam (MEC) grown at pH 5.0 compared to pH 7.0 in broth culture and in urine. Untransformed MIC values can be viewed in Supplementary file 4. (E) Fold change in minimum inhibitory concentration to cephalexin (CEX) for indicated strains grown at pH 5.5 compared to pH 7.0. EAM696 (ΔmrcB) derivatives producing PBP1b variants were grown in the presence of 10 μM IPTG. Untransformed MIC values can be viewed in Supplementary file 5. Bars represent mean fold change in minimum inhibitory concentration ± SD across at least three biological replicates.

https://doi.org/10.7554/eLife.40754.022

PBP1b is required for low pH-dependent β-lactam resistance

We next sought to identify which, if any, cell wall enzymes were required for resistance to PBP2 and PBP3-targeting compounds in acidic media. We narrowed our focus to three classes of enzymes: 1) non-essential transpeptidases, including the class A PBPs and the LD-transpeptidases, which have been implicated in β-lactam resistance (Lai et al., 2017; Hugonnet et al., 2016; Peters et al., 2018); 2) pH specialist autolysins identified in this work; and 3) PBP6b, a carboxypeptidase required for proper morphology—but not growth—in acidic media (Peters et al., 2016b). Mutants defective for production of each enzyme were tested for loss of resistance to CEX at pH 5.5 compared to pH 7.0.

Genetic analysis suggests that PBP1b activity is specifically required for pH-dependent resistance to β-lactam antibiotics. Strains defective for production of PBP1a, PBP1c, LdtD, LdtE, MepS, MltA, MltG, and PBP6b exhibited a similar increase in resistance to CEX at pH 5.5 as the parental strain (Figure 6—figure supplement 4A). In contrast, loss of PBP1b abolished resistance at pH 5.5 and in fact, slightly increased susceptibility to CEX (Figure 6E). This phenotype was not specific to CEX: resistance to other PBP2 and PBP3 targeting compounds was also eliminated or significantly reduced in cells defective for PBP1b (Figure 6—figure supplement 4C). PBP1b enzymatic activity was required for resistance. Loss of the enzyme’s cognate outer membrane lipoprotein LpoB or inactivation of its catalytic activity eliminated CEX resistance at low pH (Figure 6E; Figure 6—figure supplement 3B; Supplementary file 6). Importantly, a mutant with a comparable growth rate defect at pH 5.5 (ΔmrcB 1.91 ± 0.02 DPH; ΔtolA 1.40 ± 0.04 DPH) still displayed the same fold change in resistance to CEX at pH 5.5 as the parental strain. Likewise, a mutant in PBP5 (ΔdacA) with increased sensitivity to CEX even at neutral pH, similar to ΔmrcB (Schmidt et al., 1981; García del Portillo and de Pedro, 1990), also retained the resistance phenotype in acidic growth conditions (Sarkar et al., 2010) (Figure 6—figure supplement 3D). In sum, our findings point to a specific role for PBP1b in intrinsic β-lactam resistance in acidic media.

Discussion

Class A PBPs protect cell wall integrity across environmental conditions

By varying a single growth parameter—extracellular pH—we uncovered specialized roles for a subset of nonessential cell wall synthases and autolysins in E. coli that previously had been classified as redundant for growth. Of the pH specialist enzymes identified, we focused on the bifunctional synthases PBP1a and PBP1b, which we find are required for cell wall integrity in distinct pH environments. Failure to produce PBP1b in acidic media (pH <5.9) or PBP1a in more alkaline conditions (pH 6.5–8.2) reduced fitness and led to cell lysis (Figure 2 and Figure 3). This lytic death is characteristic of class A PBP-deficient cells (Suzuki et al., 1978) and is consistent with their essential role in PG synthesis, perhaps by filling gaps in the PG foundation (Cho et al., 2016). Importantly, a recent study failed to observe major differences in global PG composition in E. coli cells grown in pH 7.5 and pH 5.0 (Peters et al., 2016b). Hence, the differential requirement for the class A PBPs in growth across pH conditions is unlikely to be a consequence of altered cell wall structure or distinct enzymatic activity of the PBPs.

Instead, our data suggest that pH-dependent differences in class A PBP activity underlie their contribution to maximal fitness in distinct pH environments. Even in the presence of its activator LpoA, PBP1a exhibits little glycosyltransferase and transpeptidase activity in vitro in acidic conditions, rendering the cell reliant on PBP1b to provide the essential class A PBP activity (Figure 5D,E; Figure 5—figure supplement 3). At present, the precise mechanism for reduced activity of PBP1a in acidic media remains unclear; possible causes include pH-dependent changes in structure, substrate binding, or catalysis, as well as reduced affinity to LpoA in acidic conditions. We attempted to test the latter model using surface plasmon resonance (SPR), but technical limitations prevented us from drawing firm conclusions (Figure 5—figure supplement 4). Although further work is needed, the depletion of PBP1a from the cell periphery in acidic media may also reflect an inactive state of the protein in vivo (Figure 5A).

Unlike PBP1a, PBP1b activity is largely invariant in vitro across the pH conditions tested when assayed in the presence of its key regulators (Figure 5D,E; Figure 5—figure supplement 3). Although this finding is consistent with our observation that PBP1b can partially compensate for loss of PBP1a in alkaline media (Figure 2), it remains unclear why PBP1a is required for maximal fitness in alkaline media. Additional factors in vivo, such as novel class A PBP regulators, may be responsible for the cell’s preference for PBP1a in alkaline conditions that are not recapitulated in our in vitro assays. In support of this model, PBP1b exhibits reduced binding to radiolabeled penicillin in membrane extracts incubated in alkaline buffer relative to neutral or acidic conditions (Amaral et al., 1986). Alternatively, differences in the enzymes’ localization and/or interaction partners may also play a role in their pH specialization. For example, PBP1a and PBP1b preferentially associate with specialized cell wall synthesis complexes essential for cell elongation and cell division, respectively (Müller et al., 2007; Bertsche et al., 2006; Banzhaf et al., 2012). Our current study was limited to interrogating the pH specialization of nonessential cell wall enzymes. Future work investigating the influence of pH on the essential components of the elongation and division machinery is thus necessary to determine whether the activity of these complexes affects the cell’s class A PBP preference.

Although PBP1a and PBP1b share an essential role in growth and cell wall integrity under standard culture conditions (i.e. nutrient rich, neutral pH growth medium) (Suzuki et al., 1978), there had been previous hints these enzymes were not interchangeable. As previously mentioned, each class A PBP possess unique interaction partners and subcellular localization profiles (Müller et al., 2007; Bertsche et al., 2006; Banzhaf et al., 2012; Gray et al., 2015; Leclercq et al., 2017). Mutants defective for each enzyme also display differential susceptibility to antibiotic treatment (discussed below), osmotic shifts, and mechanical stress (Yousif et al., 1985; Paradis-Bleau et al., 2014; Auer et al., 2016) and have different roles in de novo regeneration of rod shape (Ranjit et al., 2017). We anticipate further study of these synthases—as with other ‘redundant’ cell wall autolysins—under nonstandard culture conditions will continue to reveal unique roles for these enzymes in cell wall biogenesis.

Plasticity in cell wall metabolism influences intrinsic resistance to cell wall active antibiotics

Analogous to alternative PBP usage in methicillin resistant Staphylococcus aureus (Berger-Bächi, 1999; Chan et al., 2016), our results suggest that environment-driven plasticity in E. coli PG synthesis may have consequences for intrinsic resistance to β-lactam antibiotics with a narrow target specificity (Figure 6). Strikingly, we find that culturing E. coli in low pH media is sufficient increase intrinsic resistance to PBP2 and PBP3-targeting β-lactams up to 32-fold. Considering E. coli encounters a wide range of pH environments across host niches (Watson et al., 1972; Bilobrov et al., 1990), our observation reinforces the importance of conducting antibiotic susceptibility testing under physiologically relevant conditions (Ersoy et al., 2017; Thulin et al., 2017).

Mechanistically, our data demonstrate low pH-dependent resistance specifically requires PBP1b activity. PBP1b has previously been implicated in intrinsic resistance to β-lactam antibiotics: cells defective for PBP1b or LpoB are hypersensitive to a wide variety of β-lactams (Nichols et al., 2011; Paradis-Bleau et al., 2014; García del Portillo and de Pedro, 1990). Likewise, elevated PBP1b activity protects cells from the lethality of the PBP2 specific antibiotic mecillinam (Lai et al., 2017). At present, the precise role for PBP1b in β-lactam protection remains unclear. It appears to be a function specifically endowed to PBP1b, for cells defective for PBP1a production do not display differential β-lactam susceptibility (Nichols et al., 2011; García del Portillo and de Pedro, 1990), highlighting another possible source of specialization between the class A PBPs. Interestingly, we and others have observed cells cultured in acidic media exhibit near normal morphology in the presence of concentrations of PBP2 or PBP3 inhibitors that lead to cell rounding or filamentation at neutral pH (Figure 6B,C) (Goodell et al., 1976). We thus speculate that PBP1b may play a role in preserving the normal functions of the elongation and division machinery in low pH environments. PBP1b may substitute for the essential function of PBP2 and PBP3 in these complexes (Modell et al., 2014) or may indirectly support growth by serving a quality control function (Morè et al., 2019). Clarifying PBP1b’s role in intrinsic resistance to β-lactams will shed light on the mechanics of cell wall biogenesis and inform the design and use of novel antibiotic therapies.

Apparent redundancy ensures fitness across environmental conditions

Apart from influencing antibiotic resistance, environmental specialization of cell wall enzymes is likely a key adaptation that allows E. coli to thrive across an unusually wide pH range (pH ~4–9) and even tolerate extreme pH shocks, such as during transient exposure to gastric acid (pH ~2) (Jordan et al., 1999). Plasticity in the cell wall synthesis machinery likely works in concert with the organism’s ability to modify extracellular pH through the export of acidic and alkaline substrates (Lu et al., 2013; Krulwich et al., 2011). In this context, pH-specialist cell wall enzymes may function in part to maintain cell wall integrity until the extracellular media reaches a growth-permissive pH.

Among other pH tolerant organisms, strategies employed to expand growth across wide pH ranges are likely to vary, even between closely related species. S. enterica serovar Typhimurium encodes a PBP3 paralog, termed PBP3SAL, that is active during growth in acidic environments, including intracellularly in the phagolysosome (Castanheira et al., 2017). As PBP3SAL is restricted to Salmonella, Enterobacter, and Citrobacter spp., alternative mechanisms to cope with changing pH environments must exist. Elucidating the requirements for pH-dependent growth in organisms outside the Enterobacteriaceae will shed light on whether class A PBPs, which are broadly conserved across bacteria (Typas et al., 2010), play a central role in the process.

Apart from pH, we anticipate enzyme specialists exist across environmental conditions; redundancy in cell wall enzymes is present throughout bacterial species, even among those that only grow at a narrow pH range (Pazos et al., 2017). B. subtilis, for example, encodes 16 PBPs, yet its growth is restricted to pH 6.0–9.0 (Wilks et al., 2009). Ionic strength, osmolality, and temperature also vary across the diverse habitats bacteria occupy. Like pH, these factors may have significant impacts on the chemical and physical properties of the periplasm and the extracytoplasmic space of Gram-positive bacteria. In support of this idea, PBP2 from Caulobacter crescentus displays differential localization patterns as a function of extracellular osmolality (Hocking et al., 2012), and the lytic transglycosylase MltA from E. coli is ~10 times more active at 30°C than at 37°C in vitro (van Straaten et al., 2005; Lommatzsch et al., 1997). We expect environmental specialization may also underlie the high levels of redundancy in other periplasmic protein classes, including sugar transporters, efflux pump adapter proteins, and chaperones (Jensen et al., 2002; Smith and Blair, 2014; Rizzitello et al., 2001). Nevertheless, it is clear that improved understanding of the contribution of many enzymes to bacterial fitness in nature demands a departure from standard growth conditions used to study bacterial physiology in the lab.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or referenceIdentifiersAdditional information
Gene (Escherichia coli)mrcBNAEcoCyc:EG10605
Gene (E. coli)mrcANAEcoCyc:EG10748
Gene (E. coli)lpoANAEcoCyc:G7642
Gene (E. coli)lpoBNAEcoCyc:G6565
Strain, strain
background (E. coli)
MG1655; wild-typePMID:6271456
Strain, strain
background (E. coli)
UTI89PMID:11402001
Antibodyanti-PBP1aGift of Waldemar Vollmer(1:5000)
Antibodyanti-PBP1bGift of Waldemar Vollmer(1:1000)
Antibodyanti-FtsZGift of David Weiss(1:5000)
Recombinant DNA reagentmrcB::kan; ∆mrcBPMID:16738554; Coli Genetic
Stock Center
CGSC:JW0145-1
Recombinant DNA reagentmrcA::kan; ∆mrcAPMID:16738554; Coli Genetic
Stock Center
CGSC:JW3359-1
Recombinant DNA reagentlpoA::kan; ∆lpoAPMID:16738554; Coli Genetic
Stock Center
CGSC:JW3116-1
Recombinant DNA reagentlpoB::kan; ∆lpoBPMID:16738554; Coli Genetic
Stock Center
CGSC:JW5157-1
Recombinant DNA reagentpCP20PMID:10829079
Recombinant DNA reagentpUM1BαPMID:12949085
Recombinant DNA reagentpUM1BγPMID:12949085
Recombinant DNA reagentpUM1Bα*PMID:12949085
Recombinant DNA reagentpUM1BTG*αPMID:12949085
Recombinant DNA reagentpUM1BTG*α*PMID:12949085
Chemical compound, drugAmpicillin Sodium SaltSigma AldrichCatalog:A9518
Chemical compound, drugAmoxicillinSigma AldrichCatalog:A8523
Chemical compound, drugCefsulodin Sodium Salt HydrateSigma AldrichCatalog:C8145
Chemical compound, drugMecillinamSigma AldrichCatalog:33447
Chemical compound, drugDoripenem HydrateSigma AldrichCatalog:SML1220
Chemical compound, drugMeropenem HydrateSigma AldrichCatalog:M2574
Chemical compound, drugCephalexinSigma AldrichCatalog:33989
Chemical compound, drugAztreonamSigma AldrichCatalog:A6848
Chemical compound, drugPiperacillin Sodium SaltSigma AldrichCatalog:P8396
Software, algorithmSuperSeggerPMID:27569113
Software, algorithmFIJIPMID:22743772
OtherPropidium iodideSigma AldrichCatalog:81845

Bacterial strains, plasmids, and growth conditions

Request a detailed protocol

Unless otherwise indicated, all chemicals, media components, and antibiotics were purchased from Sigma Aldrich (St. Louis, MO). Bacterial strains and plasmids used in this study are listed in Supplementary file 1 and Supplementary file 2, respectively. All deletion alleles were originally provided by the Coli Genetic Stock (Baba et al., 2006) and transduced into E. coli strain MG1655. For the hits identified in the growth rate screen, the expected mutation was confirmed by diagnostic PCR with Taq polymerase. Unless otherwise indicated, strains were grown in lysogeny broth (LB) media (1% tryptone, 1% NaCl, 0.5% yeast extract) supplemented with 1:10 MMT buffer (1:2:2 molar ratio of D,L-malic acid, MES, and Tris base) to vary media pH values between pH 4–9. AB defined media (Clark and Maaløe, 1967) was fixed to indicated pH values with addition of 5M HCl or 5M NaOH. Uropathogenic E. coli strain UTI89 was cultured in urine provided by a healthy donor and supplemented with MMT buffer to fix the pH. When selection was necessary, cultures were supplemented with 50 µg/mL kanamycin (Kan) and 25–100 µg/mL ampicillin (Amp). Cells were grown at 37°C either in 96-well microtiter plates shaking at 567 cpm or in glass culture tubes shaking at 200 rpm for aeration.

Growth rate measurements

Request a detailed protocol

Strains were grown from single colonies in glass culture tubes in LB +MMT buffer (pH 6.9) to mid-log phase (OD600 ~0.2–0.6), pelleted, and re-suspended to an OD600 of 1.0 (~1×109 CFU/mL). Cells were diluted and inoculated into fresh LB +MMT buffer at various pH values in 96-well plates (150 µl final volume) at 1 × 103 CFU/mL. Uncovered plates sealed with gas permeable membrane strips were grown at 37°C shaking for 20 hr in a BioTek Eon microtiter plate reader, measuring the OD600 of each well every ten minutes. Mass doublings per hour (DPH) was calculated by least squares fitting of early exponential growth (OD600 0.005–0.1) in R. Best fit lines with an R2 value below 0.95 were excluded from further analysis. Examples of growth curves and fit lines are presented in Figure 1—figure supplement 1, along with a sample script (Supplementary file 7) and representative source data (Figure 1—source datas 13). To allow for direct comparison between wild type and mutant growth, some panels present % wild type growth, which represents the DPHMutant/DPHWT × 100.

Microscopy and time lapse imaging

Request a detailed protocol

For time lapse imaging experiments, cells were grown from a single colony in LB +MMT buffer (pH 6.9) to early exponential phase (OD600 ~0.05–0.1) then mounted onto 1.0% agarose pads at pH 4.5, 6.9, or 8.0. Where indicated, propidium iodide was added to the agarose pad at a final concentration of 1.5 µM. Cells were allowed to dry on pads 10 min prior to imaging. All phase contrast and fluorescence images were acquired on a Nikon Ti-E inverted microscope (Nikon Instruments, Inc) equipped with a 100X Plan N (N.A. = 1.45) Ph3 objective, X-Cite 120 LED light source (Lumen Dynamics), and an OrcaERG CCD camera (Hammamatsu Photonics, Bridgewater, N.J.). Filter sets were purchased from Chroma Technology Corporation. The objective was pre-heated to 37°C using an objective heater. Image capture and analysis was performed using Nikon Elements Advanced Research software. Cell death quantification was determined by cells uniformly stained with propidium iodide, and terminal lytic phenotype of cells was determined by assessment of the frames immediately preceding propidium iodide incorporation. Single cell elongation rate, defined as k = ΔL/Δt/L, was determined in the MATLAB-based program SuperSegger (Stylianidou et al., 2016). Cells that lysed during the movie, indicated by negative k values, were filtered out prior to analysis.

For class A PBP localization studies, cells were grown in AB minimal media supplemented with 0.2% maltose and 250 µM IPTG overnight then sub-cultured into fresh media the following morning. Cells were grown to OD600 0.1–0.2 at 37°C then fixed by adding 20 μL of 1M NaPO4, pH 7.4, and 100 μL of fixative (fixative = 1 mL 16% paraformaldehyde + 6.25 μL 8% glutaraldehyde). Samples were incubated at room temperature for 15 min, then on ice for 30 min. Fixed cells were pelleted, washed three times in 1 mL 1X PBS, pH 7.4, then resuspended in GTE buffer (glucose-tris-EDTA) and stored at 4°C. Quantification shown in Figure 5—figure supplement 2 was performed in FIJI (Schindelin et al., 2012). Briefly, an intensity profile was generated for each cell by drawing a line across the midline of the cell from pole to pole. Maximum length and intensity were normalized to 100%. Cells enriched for class A PBP localization at the cell periphery would be expected to have increased intensity at the cell poles (0% and 100% of cell length).

Scanning electron microscopy

Request a detailed protocol

Wild type and ΔmrcB cells were grown to mid-exponential phase in MMT buffered pH 6.9 LB media and back-diluted to an OD600 = 0.1 into either pH 6.9 or 4.5 media. Cells were allowed to grow for an additional hour, fixed as described above, and applied to poly-lysine coated coverslips. Post fixation, samples were rinsed in PBS 3 times for 10 min each followed by a secondary fixation in 1% OsO4 in PBS for 60 min in the dark. The coverslips were then rinsed three times in ultrapure water for 10 min each and dehydrated in a graded ethanol series (50%, 70%, 90%, 100% x2) for 10 min each step. Once dehydrated, coverslips were then loaded into a critical point drier (Leica EM CPD 300, Vienna, Austria) which was set to perform 12 CO2 exchanges at the slowest speed. Once dried, coverslips were then mounted on aluminum stubs with carbon adhesive tabs and sputter coated with 6 nm of iridium (Leica ACE 600, Vienna, Austria). After coating, the samples were then loaded into a FE-SEM (Zeiss Merlin, Oberkochen, Germany) imaged at 3 KeV with a probe current of 178 pA using the Everhart Thornley secondary electron detector.

SDS-PAGE and immunoblotting

Request a detailed protocol

Strains were grown from a single colony in LB + 1:10 MMT at pH 4.8, 6.9, or 8.2 to mid-log phase (OD600 ~0.2–0.6), back-diluted to 0.005 in 5 mL of media, and grown to an OD600 between 0.2–0.3. Samples were pelleted, re-suspended in 2x Laemmli buffer to an OD600 ~20, and boiled for ten minutes. Samples (10 µl) were separated on 12% SDS-PAGE gels by standard electrophoresis and transferred to nitrocellulose membranes. Blots were probed with PBP1b (1:1000), PBP1a (1:5000), and FtsZ rabbit antisera (1:5000) and HRP-conjugated secondary antibody (1:2000-1:10000; goat α-rabbit). Blots were imaged on a LiCor Odyssey imager. Quantitation was determined in ImageJ and normalized to FtsZ levels as an internal loading control.

In vitro protein materials and interaction and activity assays

Request a detailed protocol

Lipid II versions were prepared as previously described (Bertsche et al., 2005; Breukink et al., 2003). The following proteins were prepared as previously described; PBP1B (Bertsche et al., 2006), LpoB (Egan et al., 2014), PBP1A (Born et al., 2006), and LpoA (Jean et al., 2014). Antisera against PBP1A and PBP1B were obtained from Eurogentec (Liege, Belgium) and purified over an antigen column as described previously (Bertsche et al., 2006).

SPR experiments were performed as previously described (Egan et al., 2014). LpoA and LpoB samples were prepared for injection over the PBP surface by 1:1 serial dilution from 10 μM to 19.5 nM. Assays were performed in triplicate at 25 ˚C, at a flow rate of 75 μL/min and with an injection time of 5 min. The running buffers consisted of 20 mM of either; sodium acetate pH 4.8, HEPES/NaOH pH 6.9, or Tris/HCl pH 8.2 with 150 mM NaCl, and 0.05% Triton X-100. The dissociation constant (KD) was calculated by non-linear regression using SigmaPlot 13 software (Systat Software Inc). Continuous fluorescence GTase assays were performed as described previously (Egan and Vollmer, 2016) with modification. Final buffer composition was 20 mM of either; sodium acetate pH 4.8, HEPES/NaOH pH 6.9, or Tris/HCl pH 8.2 plus 150 mM NaCl, 10 mM MgCl2, and 0.05% Triton X-100. Enzymes were assayed alone at 1 μM at 37°C, and at 0.2 μM in the presence of 0.5 μM regulator(s) at 25°C. The muramidase usually included in the assay samples to digest newly synthesized glycans, thereby improving fluorescence signal, was omitted to avoid indirect pH effects on observations. The slopes of the resulting curves correlate with the GTase rate and were calculated at their fastest point using linear regression in Excel 2016 (Microsoft). When presented in Figure 5—figure supplement 3, the values are inverted from negative for simplicity. Measurement of total PG synthesis activity using radiolabelled lipid II substrate was performed as previously described (Biboy et al., 2013) using enzyme (0.5 μM) and regulator(s) (2 μM) in the same three buffers indicated for the GTase assay. Total TPase activity was calculated as the percentage of muropeptide products known to be produced by this domain’s function, including peptide cross-linking and DD-carboxypeptidase activity.

Antibiotic susceptibility testing

Request a detailed protocol

For determination of minimum inhibitory concentrations, cells were grown from a single colony in LB media at the indicated pH to mid-exponential phase (OD600 ~0.2–0.6) at 37°C with aeration and then inoculated at 1 × 105 CFU/mL into LB media of the same pH in sterile 96-well plates with a range of two-fold dilutions of the indicated antimicrobial agent (final volume, 150 µL). Plates were incubated at 37°C shaking for 20 hr before determination of the well with the lowest concentration of the antibiotic that had prevented growth by visual inspection.

Antibiotic stability testing

Request a detailed protocol

Antibiotics were incubated for 20 hr in LB media at pH 4.5, 7.0, or 8.0 and then diluted into 96-well plates containing LB media (pH 7) and 1 × 105 CFU/mL MG1655. Plates were then incubated at 37°C shaking for 20 hr before determination of the compound’s minimum inhibitory concentration.

Terminal phenotype assessment

Request a detailed protocol

Cells from minimum inhibitory concentration assays were spotted (5 µL) onto 1.0% agarose pads 20 hr post-treatment and imaged by phase contrast microscopy to track cell morphology in response to antibiotic treatment across pH values. Growth rate was monitored by OD600 in the BioTek Eon plate reader to confirm all cells examined were in the same growth phase and at approximately the same optical density prior to imaging. Cell dimensions were quantified in the MATLAB-based program SuperSegger (Stylianidou et al., 2016).

Quantification and statistical analysis

Request a detailed protocol

A minimum of three biological replicates were performed for each experimental condition unless otherwise indicated. Data are expressed as means ± standard deviation (SD) or standard error of the mean. Statistical tests employed are indicated in the text and corresponding figure legend. Analysis was performed in R or GraphPad Prism. Asterisks indicate significance as follows: *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    Control of acid resistance in Escherichia coli
    1. MP Castanie-Cornet
    2. TA Penfound
    3. D Smith
    4. JF Elliott
    5. JW Foster
    (1999)
    Journal of Bacteriology 181:3525–3535.
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
    Activities and regulation of peptidoglycan synthases
    1. AJF Egan
    2. J Biboy
    3. I van't Veer
    4. E Breukink
    5. W Vollmer
    (2015)
    Philosophical Transactions of the Royal Society B: Biological Sciences 370:20150031.
    https://doi.org/10.1098/rstb.2015.0031
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
    On the intensity of urinary acidity in normal and pathological conditions
    1. LJ Henderson
    2. WW Palmer
    (1912)
    The Journal of Biological Chemistry.
  30. 30
  31. 31
  32. 32
    Effect of temperature, pressure, pH, and osmotic stress on growth
    1. JL Ingraham
    2. AJ Marr
    (1996)
    Escherichia Coli And Salmonella Typhimurium: Cellular and Molecular Biology.
  33. 33
  34. 34
  35. 35
    Survival of low-pH stress by Escherichia coli O157:h7: correlation between alterations in the cell envelope and increased acid tolerance
    1. KN Jordan
    2. L Oxford
    3. CP O'Byrne
    (1999)
    Applied and Environmental Microbiology 65:3048–3055.
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
    Structural modifications in the peptidoglycan of Escherichia coli associated with changes in the state of growth of the culture
    1. AG Pisabarro
    2. MA de Pedro
    3. D Vázquez
    (1985)
    Journal of Bacteriology 161:238–242.
  60. 60
  61. 61
  62. 62
    Structural and functional properties of porin channels in E. coli outer membranes
    1. JP Rosenbusch
    (1990)
    Experientia 46:167–173.
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
    Effects of furazlocillin, a beta-lactam antibiotic which binds selectively to penicillin-binding protein 3, on Escherichia coli mutants deficient in other penicillin-binding proteins
    1. LS Schmidt
    2. G Botta
    3. JT Park
    (1981)
    Journal of Bacteriology 145:632–637.
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84

Decision letter

  1. Michael T Laub
    Reviewing Editor; Massachusetts Institute of Technology, United States
  2. Gisela Storz
    Senior Editor; National Institute of Child Health and Human Development, United States
  3. Michael T Laub
    Reviewer; Massachusetts Institute of Technology, United States

In the interests of transparency, eLife includes the editorial decision letter, peer reviews, and accompanying author responses.

[Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed.]

Evaluation of resubmission:

The authors have now responded to each of the issues raised by the three reviewers. In general, the responses are thorough and appropriate. The changes made to the text as well as the additional experiments performed have now strengthened many of the conclusions and claims made in the paper. In particular, the evidence supporting the pH-specialization of PBP1a and PBP1b has been improved. This includes new in vitro studies, done collaboratively with the Vollmer group, showing that PBP1a activity is significantly reduced at pH 4.8, potentially explaining why PBP1b activity becomes required at lower pH. Precisely how pH affects PBP1a activity still is not totally clear - an attempt was made to measure stimulation by LpoA but technical difficulties prevented a clear assessment of this. The authors have also addressed a key concern from one reviewer regarding the increased resistance to a sub-class of beta-lactam antibiotics at low pH showing that the effects are likely not due to non-specific changes in permeability. All told, the changes made to the paper have improved it and strengthened the major conclusions.

Decision letter after peer review:

Thank you for submitting your article "Plasticity of E. coli cell wall metabolism promotes fitness and antibiotic resistance across environmental conditions" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Michael T Laub as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Gisela Storz as the Senior Editor.

The Reviewing Editor has highlighted the concerns that require revision and/or responses, and we have included the separate reviews below for your consideration. If you have any questions, please do not hesitate to contact us.

Summary of major concerns:

This paper presents data to support the intriguing conclusion that the apparent redundancy of cell wall synthetic proteins may stem from their specialization to different environments. The reviewers agreed that the paper was well-written and contained a substantial body of data to demonstrate that PBP1b, in particular, assumes a pivotal role in low pH environments. There were, however, a number of major concerns raised, detailed in the individual reviews below. Most notably, the reviewers were concerned about whether PBP1a is really "specialized" for alkaline pH and hence whether PBP1a/b redundancy has evolved to ensure optimal activities in opposing pH niches. They also thought the paper would benefit from additional data and insights into the notion that PBP1b may replace PBP2/3 in the elongation and division machineries at low pH. There were also concerns about the statistics and number of replicates in several figures.

You can see below that reviewer 2 initially was quite positive. However to be complete, we want to give you a sense of the subsequent discussion where that reviewer stated:

"I was obviously more enthusiastic about the story than both of you, although I too was surprised to see it in eLife. I think that you both brought up important points and that my liking of the overarching idea of specialization made me miss some of the issues."

"So, thinking about everyone's comments and my new perspective, I believe that the paper is not at par for eLife. It presents a very interesting model to explain the so-called redundancy and lack of phenotypes. I like the idea that the lack of phenotypes reflects that we are looking under the wrong conditions (which the paper illustrates) because those enzymes have been selected for in the real world for billions of years. I also like the warning about antibiotic sensitivity being affected by environmental conditions like pH. However, the problem is that the authors overinterpret some of their results. It seems that the story boils down to PBP1A specifically not being able to work at low pH for an unknown mechanism. PBP1B seems to be able to work under all conditions, although slightly worse than 1A at high pH, but the data do not support the perfect picture of Figure 7. They could get more data in response to reviews but, most importantly, they will have to revise the model."

Separate reviews (please respond to each point):

Reviewer #1:

This paper examines the pH-dependencies of PBP1a and PBP1b, arguing that each is "specialized" for a different pH. The data are relatively clean and soundly interpreted in most cases. Overall, the paper paints a compelling picture that the apparent redundancy of cell wall enzymes may reflect specialization for different environments. Below are several issues that I think the authors should respond to before publication – in some cases, this could involve significant new experiments, but I think such experiments would substantially add to the paper's mechanistic depth and long-term impact on the field.

1) Figure 5C shows PBP1b levels – but what about PBP1a?

2) Is there any significance to the clustering of GFP-PBP1a at low pH? I was sort of surprised that this observation wasn't followed up at all.

3) PBP1b provides resistance to various antibiotics at low pH. But I don't really understand why. If the model is that PBP1a is more active and neutral and alkaline pH with PBP1b more active at low pH, why isn't PBP1a leading to comparable levels of resistance at the higher pHs? Do PBP1b and PBP1a, in fact, have different activities? I thought they were carrying out the same reaction, with one more active at low pH and the other more active at higher pHs, so the source of the difference in antibiotic susceptibility doesn't really make sense to me, but maybe I'm missing something here.

4) I really think it's a lost opportunity to not provide more insight into the molecular basis for the pH differences in PBP1a and PBP1b. It appears that PBP1b activity can be reconstituted in vitro, including with LpoB and FtsN, so testing things in vitro at different pHs so testing things in vitro at different pHs seems potentially feasible. I would encourage the authors to at least consider whether something in this regard can be done to strengthen and enhance the paper.

5) I wondered whether there are phylogenetic analyses to support the notion that PBP1a and PBP1b have evolved, at least in part, to allow cells to cope with different pH conditions. Are there strains of E. coli lacking PBP1b or even closely related species lacking it? If so, can they not tolerate low pHs and/or would their PBPs complement a PBP1a/1b double deletion only at certain pHs?

Minor Comments:

Abstract, last sentence: Awkward sentence – rephrase for clarity.

"providing a force" maybe not "providing the force" in light of recent findings from KC Huang et al.

"in in" typo.

Introduction, last sentence: Sort of a cryptic end to the Introduction – clarify why this is concerning and what the consequences are.

Subsection “Identification of pH specialist cell wall synthases and hydrolases”, first paragraph: The authors probably need to do an ANOVA test to control for multiple hypothesis testing. Even with a p < 0.01 to increase stringency, I think an ANOVA is the right choice here.

What does 'cytoskeletal-independent' mean? I don't think of either MreB or FtsZ as cytoskeletal elements anymore – reword for accuracy/precision.

"optimized" is a strong word…maybe "tuned" would be better. I just always think it's tough to show something is truly optimal.

Subsection “aPBP activity promotes cell wall integrity across pH environment”, first paragraph: typo, "cells" not "cell".

Typo "previous a proteomic".

Reviewer #2:

Most bacteria build a peptidoglycan cell wall that protects them from osmotic lysis. While enzymes required for the synthesis of the peptidoglycan precursor (lipid-linked disaccharide pentapeptide) are essential, many of the enzymes participating in the construction of the cell wall are not. In Escherichia coli, there is functional redundancy between two class a penicillin-binding proteins (PBPs) since removal of both PBP1a and PBP1b causes lethality. This has led to the idea that functional redundancy might be widespread among enzymes that synthesize and modify the cell wall, since removal of individual factors does not typically confer phenotypes. Here, Mueller and Levin challenge the view that these enzymes are simply redundant. They showed that the so-called "redundant" PBP1a and PBP1b enzymes have each specialized to function in different pH ranges. While PBP1a is dispensable in acidic conditions, PBP1b is essential; in contrast, in alkaline environments, PBP1b is dispensable, but PBP1a is required for optimal survival. This is a significant discovery and it is possible that, as the authors propose, similar specialization might apply to other enzymes involved in peptidoglycan synthesis and remodeling. Furthermore, the authors demonstrate that the intrinsic resistance to antibiotics can be different based on environmental pH.

The manuscript is well written, experiments were well done, conclusions are justified, and the work reports significant findings.

There is only one issue that should be easily addressed to strengthen one of the claims:

The authors used a propidium iodide assay to quantify lysis. This assay measures the uptake of the dye not necessary lysis. Although the images of cells in Figure 3 and Figure 4 show that some cells have lysed, the authors should track loss of cytoplasmic contents (e.g. leakage of cytoplasmic GFP or LacZ) to demonstrate and quantify lysis. It would also be nice to complement lysis data with a kill curve (determine number of cells alive by CFUs after exposure to the specific pH). The authors could still leave the propidium iodide data to reflect envelope problems (but not necessarily lysis or even death).

Minor Comments:

1) Delete repeated "in".

2) Subsection “Identification of pH specialist cell wall synthases and hydrolases”, second paragraph: I think it should state "six" mutants. Not "seven".

3) Subsection “aPBP activity ensures fitness across a wide pH range”, first paragraph: Denome et al., 1999, might not be the best reference to cite the synthetic lethality of PBP1a and 1b because the pbp1a mutant used in that study contained additional mutations, as described in Meberg et al., J Bacteriol. 183:6148-9; PMID: 11567017. It would be more appropriate to cite the first report of the synthetic lethality: Suzuki et al. (1978) PNAS 75: 664-668; https://doi.org/10.1073/pnas.75.2.664.

Reviewer #3:

Bacteria maintain a tremendous enzymatic redundancy when it comes to building up their cell wall. In this paper, the authors systematically assess the contribution of 33 enzymes (carrying 7 distinct enzymatic functions) to the growth rate of E. coli in low and high pH, both commonly encountered environments for E. coli in the human body. The authors then focus on one group, the bifunctional synthases, and provide evidence that its 2 main members, PBP1a and PBP1b are specialized for acting in high and low pH, respectively. They also link this PBP1b specificity to increased resistance to a subgroup of β-lactam inhibitors at acidic environments. Overall, the authors use this example to propose that division of labor and environment-specific action is the main reason behind the apparent redundancy of cell wall enzymes.

The strengths of this paper are that:

• it is clearly and well written;

• it systematically interrogates the relationships between a specific environment (pH) and cell wall biosynthetic enzymes;

• it makes a stronger case for environment-dependent changes in the activity of cell-wall enzymes;

• it provides strong evidence for the crucial role of PBP1b at low pH (which I think the authors should focus on);

• it provides preliminary evidence that there may be a striking reorganization of the division and elongation machineries at low pH (which the authors can provide more evidence for – see point 4).

On the other hand, there are also obvious weaknesses:

• the novelty of pH-dependency of cell wall enzymes is questionable; several papers have already provided single-case examples;

• the claim that redundancy of cell wall enzymes boils down only to environment-specialized function is not something the current data can support. Specialization may be spatiotemporal (e.g. cell cycle), allow links to distinct protein machineries, allow additional functions… Redundancy could also ensure robustness of an essential process, targeted by many natural products in nature (antibiotics, T6SS effectors);

• the data for PBP1a taking control in high pH are weak/not convincing (see point 2), which undermines the main statement about clear pH division of bifunctional synthases;

• the statistics, the experimental description and sometimes the experiment setup do not warrant for the claims made (see points 1, 2 and 5);

• there are alternative ways to interpret the antibiotic resistance at low pH, which are independent of the PBP1b increased activity (see point 3).

Major points:

1) Credit should be given to authors for their effort to systematically detect growth defects of mutants in cell wall enzymes across different pH's (Figure 1). However, since most of the effects are small, statistics and experimental setups need be stronger/more transparent to substantiate claims. n=3 (Figure 1 and Figure 1—figure supplement 1) is very low to call 5% effects and to do t-tests (which assume normal distributions). It is also unclear how many measurements and what window of OD's are used for the exponential growth fit, and how good fits are at the end. Representative examples as a supp. figure would be extremely useful, in addition to providing the raw data and fits as supp. data. Last, there is no mention if authors look for/normalize plate effects (positional biases) when measuring for growth. This is very common in microtiter assays, and could very well be driving some of the small differences.

Since there are now a number of studies published in which the fitness of E. coli genome-wide mutant libraries has been probed across many different conditions (PMID 21185072, 27355376, 29769716), the authors may want to compare their results to these studies. pH is certainly one of these conditions. This way they can also look for further evidence of more conditional-specific roles for cell wall enzymes.

2) The are several pieces of data that make me doubt about the validity of the claims about the PBP1a importance/specialization in alkaline pH. I would suggest to remove (the emphasis from) this part, unless authors provide much stronger evidence.

In growth assays, the specific defect of the mrcA mutant in alkaline pH is questionable. Effect is marginal at ph8.2 in Figure 1A (mrcA seems to also have a very mild defect in all pHs in same figure) and Figure 2A mentions no n (done once?) and has no error bars to assess the significance of differences. Besides the two figures disagree; in 2A, the biggest difference is for ph~7 (difference gone for ph>8) and in 1A is at pH=8.2. Also speaking against of specialized role of PBP1A-LpoA at alkaline conditions, the lpoA mutant has the same small growth defect in all pH's (Figure 2C – if anything smaller at alkaline pHs)!

In microscopy, the cell death of mrcA is only transient (Figure 3B), so this cannot be the reason for the change in steady state growth as authors suggest (subsection “aPBP activity promotes cell wall integrity across pH environments”, second paragraph). Such an effect would result into also a transient growth defect in batch culture, which will be invisible the way authors do their experiments (as they back dilute 10^3 cells/ml so first doublings are in ODs that are below level of detection). Only way to explain batch growth rate defect (if it holds) is with an accompanying change in single cell growth rate, which authors can measure.

The data on PBP1A activity (Figure 5E) at different pH's are again not convincing. Signal is very low (btw this is completely the opposite from Figure 5D, where PBP1a has the strongest signal!), kinetics are fast (so many points and replicates are needed) and this seems to be the only replicate for measuring rates. It is also unclear how fits are done for Kobs (what type of nonlinear regression?), but they are definitely bad for ph5 (it seems as if a linear fit is shown). Besides more replicates with more points in the Bocillin assay, an in-vitro PBP1a assay may allow to see better kinetics and would strengthen any argument. Please keep in mind that pH7 that experiments are done for Figure 5E are not alkaline but neutral.

Overall, I find little evidence that PBP1a-LpoA takes over in alkaline conditions (as implied in text and Figure 7). Even if all effects are validated, at best it means that PBP1a/LpoA have a role under these conditions, but PBP1b/LpoB can compensate to a large degree.

3) Increased resistance to a sub-class of β-lactams at low pH is due to increased PBP1B activity. Although this could very well be (effect gone in mrcB mutant and controls in Figure 6—figure supplement 3C and D provide strong evidence), everything could be also explained with some of these drugs having decreased entry at low pH. Note that β-lactams have differential uptake and efflux preferences – Aztreonam for example needs OmpF to enter cell (see PMID – 29980614). Less drug in the cell could explain also why cells treated with Cephalexin or Mecillinam need more drug to change their morphology at pH5, although presumably the targets of the drugs, PBP3 and PBP2 do not change levels.

Measuring drug intracellular concentrations would be required to exclude such a scenario. Or alternatively some of the experiments in point 4 could strengthen the PBP1B increased activity in complexes. Also including more non-specific/broad β-lactams would (Carbenicillin, Amoxicillin) would help towards this direction. Note that Imipinem is not a selective PBP2 inhibitor- it targets many PBPs in addition to LD-transpeptidases.

4) The authors propose 3 models to rationalize how PBP1b increased activity at low pH could make cells more resistance to PBP2/3 inhibitors. They favor the one which PBP1b replaces PBP2 and PBP3 in the elongasome and divisome (as increased repair could for example not explain the need for increased Cephalexin to see filamentation at low pH). This scenario has some straightforward ramifications that are rather easy to test: PBP2/3 should be less active at low pH (Bocillin assay), PBP1b/2/3 dynamics may differ at low pH, cells should tolerate more depletion of PBP2/PBP3 at low pH. Also it could be interesting to test if Salmonella shows the same higher resistance at low pH (in PBP1b-dependent manner) when PBP3Sal is knocked out but not when there (because PBP3Sal is still as sensitive to PBP3 inhibitors).

5) Quantification and statistics suffer throughout the paper. Replicates are not always there (Figure 2A, Figure 5E) and many times too little for the statistics authors want to do: standard deviation from 2 replicates (5C, 6D), t-tests from sets of 3 experiments (Figure 1). Microscopy is not always quantified (Figure 5A, 6B/C), and the most worrisome is that panels do not always agree (Figure 1-2, or Figure 5D and 5E). All this should be amended for the conclusions to be on solid ground.

Minor Comments:

1) Introduction, second paragraph: the reason for the high sensitivity of the periplasm is not the different permeability of the two membranes; please rephrase.

2) Introduction, second paragraph: careful, there is a homeostatic control system: HdeA and HdeB are periplasmic chaperones induced and required during acidic conditions; quite some things known about their function. Also there is some literature on the requirement of other housekeeping periplasmic quality control enzymes during acid shock.

3) Introduction, fifth paragraph: why curiously?

4) Introduction, last paragraph: one "in" too much.

5) Throughout the text: "insertional deletions". Mutations come either from insertions or deletions. Since these are mutants from the KEIO collection, they are deletions.

6) Subsection “Identification of pH specialist cell wall synthases and hydrolases”, first paragraph: commensal E. coli (MG1655 is not even that) are not specialized for growing in urine. UPEC strains carry ~2,000 genes more than MG1655, so this leaves quite some room for other alkaline-specialized cell wall enzymes.

7) Subsection “aPBP activity ensures fitness across a wide pH range”, first paragraph: effect is not dramatic for mrcA.

8) Subsection “aPBP activity ensures fitness across a wide pH range”, second paragraph: how is p-val calculated with no replicates here?

9) Subsection “aPBP activity ensures fitness across a wide pH range”, second paragraph (Figure 2—figure supplement 1); could be informative to test pbpC double mutants with mrcA and mrcB to check if the effect of PBP1c is masked by any of the other two.

10) Subsection “aPBP substrate binding is pH-dependent”, first paragraph/Figure 5A: please quantify.

11) Subsection “pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific β-lactam antibiotics”, first paragraph: mrcB and lpoB mutants are very sensitive also to Cefsulodin, and a number of broad-acting cell wall enzymes. In contrast mecillinam effects are mild. These facts are hard to reconcile with line of thought that follows (see the second paragraph of the aforementioned subsection).

12) Subsection “pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific β-lactam antibiotics”, third paragraph/Figure 6D: other drugs and n>2 would be helpful to show that effect is specific.

13) Subsection “pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific β-lactam antibiotics”, last paragraph: referring to Figure 6E not 6D.

14) Subsection “Specialization role for aPBPs in cell wall integrity across environmental conditions”, first paragraph: if effects are direct on enzyme activity cannot be deduced from the Bocillin assay; many other upstream effects could compromise enzyme activity.

15) Subsection “Specialization role for aPBPs in cell wall integrity across environmental conditions”, second paragraph: first time I hear about depressed enzymes…

16) Subsection “Plasticity in cell wall metabolism potentiates intrinsic resistance to cell wall active antibiotics”, first paragraph: I find this highly speculative. Slow growth may as well be playing a more active role in resistance development. Also these microbes can become equally well resistant to broad cell wall inhibitors as they do to PBP3 inhibitors.

17) Subsection “Bacterial strains, plasmids, and growth conditions”: are strains re-transduced (once)? One clone used? Mutation checked?

18) Figure 3B: would be nice to see a longer experiment (past 2hrs); also can make points/schemes smaller to see error bars.

19) Figure 5D/E: inconsistent. PBP1a has the strong signal in d, quantified as low signal in e. PBP1b effect is ~ 2-fold at 15 min at Figure 5D, but looks less in 5E.

20) Figure 6A: What happens with Mecillinam at pH>7?

21) Figure 6—figure supplement 1: Azt is unstable at higher pH, but this does not correlate with increase in MIC (Figure 6)?!

22) Figure 6B/C: quantitative data would be more convincing.

23) Figure 6—figure supplement 3 legend: Define pH's you are comparing.

https://doi.org/10.7554/eLife.40754.036

Author response

Reviewer #1:

This paper examines the pH-dependencies of PBP1a and PBP1b, arguing that each is "specialized" for a different pH. The data are relatively clean and soundly interpreted in most cases. Overall, the paper paints a compelling picture that the apparent redundancy of cell wall enzymes may reflect specialization for different environments. Below are several issues that I think the authors should respond to before publication – in some cases, this could involve significant new experiments, but I think such experiments would substantially add to the paper's mechanistic depth and long-term impact on the field.

1) Figure 5C shows PBP1b levels – but what about PBP1a?

We thank the reviewer for this suggestion. We acquired a PBP1a antibody and have updated Figure 5C to include PBP1a levels across pH conditions. Normalizing to FtsZ levels as an internal loading control, this analysis revealed that both PBP1a and PBP1b levels are modestly reduced (~2 and 4-fold, respectively) during growth at pH 5.2 compared to pH 6.9. As noted in the text, this finding is consistent with the trend observed in Schmidt et al., 2016, by proteomic mass spectrometry when comparing cells grown at pH 6.0 to pH 7.0. In contrast, PBP1a and PBP1b levels were not significantly altered during growth at pH 8.2 compared to pH 6.9. This result strengthens our conclusion that changes in aPBP levels across pH environments are unlikely to contribute to the enzymes’ pH specificity.

2) Is there any significance to the clustering of GFP-PBP1a at low pH? I was sort of surprised that this observation wasn't followed up at all.

We agree with the reviewer this is an interesting observation that warrants further investigation in future studies. In the present revision, we have quantified PBP1a subcellular localization (see Figure 5—figure supplement 2), as suggested by reviewer 3. As expected, during growth at pH 7.0, GFP-PBP1b is enriched at the cell boundaries, similar to the localization profile previously reported for GFP-PBP1a (e.g. Paradis-Bleau, 2010). However, during growth at pH 5.0, GFP-PBP1a enrichment at the cell boundaries is reduced and instead forms irregular shaped puncta throughout the cell body. It is tempting to speculate that the reduction in peripheral GFP-PBP1a may reflect a reduction in the available/active PBP1a pool at the cell envelope and thus contribute to the cell’s reliance on PBP1b to provide the essential aPBP activity during growth in acidic media. This pH-dependent change in GFP-PBP1a localization may be a product of many factors, including but not limited to improper localization of a PBP1a activator (e.g. LpoA), sequestration of PBP1a by an unknown factor, an inactive conformational state of PBP1a, degradation of PBP1a or LpoA, or an inability of PBP1a to successfully traffic to the membrane. Rigorous assessment of each of these possibilities will be investigated in future studies.

3) PBP1b provides resistance to various antibiotics at low pH. But I don't really understand why. If the model is that PBP1a is more active and neutral and alkaline pH with PBP1b more active at low pH, why isn't PBP1a leading to comparable levels of resistance at the higher pHs? Do PBP1b and PBP1a, in fact, have different activities? I thought they were carrying out the same reaction, with one more active at low pH and the other more active at higher pHs, so the source of the difference in antibiotic susceptibility doesn't really make sense to me, but maybe I'm missing something here.

PBP1a and PBP1b share the same biochemical activities in vitro(i.e. transpeptidase and glycosyltransferase activity) and appear to be interchangeable for growth in vivo. At the same time, their activity is differentially required in vivo in response to different types of cell envelope stress, including antibiotic, osmotic, and mechanical challenge. In particular, PBP1b seems to be critical in response to these stressors, and PBP1a cannot fully compensate in the absence of PBP1b.

At present, it remains unclear why PBP1b is preferentially required for resistance to cell wall stress. Several explanations have been proposed. One possibility is that differences in enzymatic activity (e.g. differences in cross linking ability) and/or enzymatic efficiency of the class A PBPs contributes to their distinct roles in stress resistance. For example, Born et al., 2006, found that under optimal in vitroconditions PBP1b synthesized PG with 2x the amount of crosslinked peptides compared to PBP1a. While the relationship between the enzymatic properties of the class A PBPs and the PBP1b mutant antibiotic hypersensitivity has yet to be rigorously interrogated, there is little evidence to suggest enzymatic differences are playing a role in the pH-dependent resistance phenotype described here: comparison of E. coli PG composition at pH 7.5 and 5.0 did not reveal any differences in overall PG structure, including the percentage of crosslinked peptides (Peters et al., 2016). An alternative possibility is that the class A PBPs play unique roles in cell wall quality control. In support of this idea, in the time since our manuscript was submitted, work from Moré and colleagues (Moré et al., 2019) identified a role for LpoB/PBP1b (but not LpoA/PBP1a) in a so-called “PG repair machine” important for survival after outer membrane stress. It is possible that PBP1b may play a similar role in responding to β-lactam induced cell wall “damage”, and PBP1a cannot compensate. This model also remains to be tested.

Although of significant interest, the cause of the disparate contribution of PBP1a and PBP1b to β-lactam protection is not the focus of this investigation, so we have refrained on extensively commenting on it in the text. We have, however, added lines to the Discussion and throughout the text more explicitly distinguish between the class A PBPs’ overlapping roles in supporting cell growth (under standard culture conditions) and in stress/antibiotic resistance.

4) I really think it's a lost opportunity to not provide more insight into the molecular basis for the pH differences in PBP1a and PBP1b. It appears that PBP1b activity can be reconstituted in vitro, including with LpoB and FtsN, so testing things in vitro at different pHs so testing things in vitro at different pHs seems potentially feasible. I would encourage the authors to at least consider whether something in this regard can be done to strengthen and enhance the paper.

We thank the reviewer for encouraging us to investigate the impact of pH on the biochemical activity of PBP1a and PBP1b in the context of their known activators. To this end, we collaborated with Waldemar Vollmer’s group, which has extensive expertise in reconstituting activity of the class A PBPs in the presence of their cognate activators in vitro. The results of these experiments are included in Figure 5D, E, Figure 5—figure supplement 2, and Figure 5—figure supplement 3 and are discussed in the subsection “PBP1a localization and activity are impaired at low pH”. Overall, they support a model in which PBP1a activity is significantly reduced at pH 4.8, rendering the cell reliant on PBP1b activity in acidic media.

Briefly, we performed two PG synthesis assays. The first is an end point assay, in which purified enzymes and activators are reconstituted in micelles, supplied radiolabeled Lipid II precursor, and allowed to react for our hour. Synthesized PG is digested into muropeptides and resolved via HPLC. The fraction of unutilized Lipid II provides a qualitative metric for end-point glycosyltransferase activity, and the end-point transpeptidase activity can by quantified by the sum of the peaks corresponding to crosslinked products. This analysis revealed that PBP1b-LpoA has little activity at pH 4.8, while PBP1b retains similar activity across all pH conditions (Figure 5D, E).

This finding is further bolstered by the results of a continuous fluorescence assay (Figure 5—figure supplement 3). This assay measures glycosyltransferase reaction rate by quantifying the rate of polymerization of Dansyl-labled Lipid II substrate. Polymerization causes a decrease in fluorescence signal. As previously shown for PBP1b, both class A PBPs had a reduced polymerization rate in acidic media when assayed in the absence of their regulators. However, the presence of LpoB and FtsN significantly stimulated polymerization rate of PBP1b in all pH conditions (with the highest fold-change observed at pH 4.8); in contrast, LpoA failed to stimulate PBP1a under the same conditions. Note that the experimental conditions (e.g. temperature and enzyme concentration) vary between panels A and B due to technical limitations in accurately quantifying the rapid activity of PBP1b in the presence of its activators at 37 °C.

In attempt to distinguish whether the lack of PBP1a activity in acidic conditions is due to a decrease in affinity for LpoA, we performed a series of SRP experiments to measure the binding affinity of PBP1a and PBP1b to their cognate Lpo at pH 4.8, 6.9, and 8.2. Although affinity measurements could be made at pH 6.9 and 8.2, the Lpos both bound to the chip nonspecifically at pH 4.8 and impeded affinity calculations. Therefore, while it is evident that LpoB can stimulate PBP1b in acidic conditions (as evidenced by the GTase assay), it remains unclear whether the LpoA stimulation of PBP1a occurs at pH 4.8.

5) I wondered whether there are phylogenetic analyses to support the notion that PBP1a and PBP1b have evolved, at least in part, to allow cells to cope with different pH conditions. Are there strains of E. coli lacking PBP1b or even closely related species lacking it? If so, can they not tolerate low pHs and/or would their PBPs complement a PBP1a/1b double deletion only at certain pHs?

This is a very interesting hypothesis that we would love to test. However, previous phylogenetic analysis (Typas et al., 2010; see Figure 6) revealed that all γ-proteobacteria, including all examined E. coli and Enterobacteriaceae genomes, possess orthologs of PBP1a and PBP1b. While α and β proteobacteria only encode PBP1a orthologs, their evolutionary distance from E. coli MG1655 would complicate any conclusions that we would hope to make in doing such an analysis.

Minor Comments:

Abstract, last sentence: Awkward sentence – rephrase for clarity.

Rewritten, as suggested.

"providing a force" maybe not "providing the force" in light of recent findings from KC Huang et al.

Modified in text in light of the recent finding by Rojas et al., 2018, that the outer membrane in Gram negative bacteria is also a load bearing element.

"in in" typo.

Modified in text.

Introduction, last sentence: Sort of a cryptic end to the Introduction – clarify why this is concerning and what the consequences are.

We agree. This sentence has been revised for clarity.

Subsection “Identification of pH specialist cell wall synthases and hydrolases”, first paragraph: The authors probably need to do an ANOVA test to control for multiple hypothesis testing. Even with a p < 0.01 to increase stringency, I think an ANOVA is the right choice here.

Agreed. We have re-analyzed our data using a one-way ANOVA with a p < 0.01 and normalized for multiple comparisons. With the exception of the ΔmltB mutant, the remaining five mutants’ growth rate defects remained significant by this analysis. Each the remaining mutants displays a consistent defect in mass doublings per hour across a discrete range of pH values (Figure 1—figure supplement 2), supporting our classification as ‘pH specialists’. In contrast, loss of MltB did not confer a consistent defect in DPH across pH conditions compared to the parental strain (see Author response image 1).

Author response image 1

What does 'cytoskeletal-independent' mean? I don't think of either MreB or FtsZ as cytoskeletal elements anymore – reword for accuracy/precision.

We agree and have removed the adjective ‘cytoskeletal-independent’ to avoid confusion.

"optimized" is a strong word…maybe "tuned" would be better. I just always think it's tough to show something is truly optimal.

We agree and have replaced ‘optimized’ with ‘tuned’ as the reviewer suggested.

Subsection “aPBP activity promotes cell wall integrity across pH environment”, first paragraph: typo, "cells" not "cell".

Corrected.

Typo "previous a proteomic".

Corrected to ‘a previous proteomic’

Reviewer #2:

Most bacteria build a peptidoglycan cell wall that protects them from osmotic lysis. While enzymes required for the synthesis of the peptidoglycan precursor (lipid-linked disaccharide pentapeptide) are essential, many of the enzymes participating in the construction of the cell wall are not. In Escherichia coli, there is functional redundancy between two class a penicillin-binding proteins (PBPs) since removal of both PBP1a and PBP1b causes lethality. This has led to the idea that functional redundancy might be widespread among enzymes that synthesize and modify the cell wall, since removal of individual factors does not typically confer phenotypes. Here, Mueller and Levin challenge the view that these enzymes are simply redundant. They showed that the so-called "redundant" PBP1a and PBP1b enzymes have each specialized to function in different pH ranges. While PBP1a is dispensable in acidic conditions, PBP1b is essential; in contrast, in alkaline environments, PBP1b is dispensable, but PBP1a is required for optimal survival. This is a significant discovery and it is possible that, as the authors propose, similar specialization might apply to other enzymes involved in peptidoglycan synthesis and remodeling. Furthermore, the authors demonstrate that the intrinsic resistance to antibiotics can be different based on environmental pH.

The manuscript is well written, experiments were well done, conclusions are justified, and the work reports significant findings.

There is only one issue that should be easily addressed to strengthen one of the claims:

The authors used a propidium iodide assay to quantify lysis. This assay measures the uptake of the dye not necessary lysis. Although the images of cells in Figure 3 and Figure 4 show that some cells have lysed, the authors should track loss of cytoplasmic contents (e.g. leakage of cytoplasmic GFP or LacZ) to demonstrate and quantify lysis. It would also be nice to complement lysis data with a kill curve (determine number of cells alive by CFUs after exposure to the specific pH). The authors could still leave the propidium iodide data to reflect envelope problems (but not necessarily lysis or even death).

We thank the reviewer for this suggestion. To complement this analysis, we transformed PBP1a and PBP1b defective cells with a plasmid that expresses gfp under an IPTG-inducible promoter and measured concurrent π staining and loss of cytoplasmic GFP signal by time lapse microscopy when cells were exposed to their respective ‘non-permissive’ pH conditions. All cells that stained PI+ (n = 419 for ΔmrcB and n = 789 for ΔmrcA) also lost cytoplasmic GFP signal within 1-2 frames (acquisitions were taken every 3 minutes). Population-level π staining/cytoplasmic GFP loss kinetics are now summarized in Figure 3—figure supplement 1. We do note that the lysis kinetics of the ΔmrcB mutant harboring the plasmid are slower compared to the untransformed strain (compare to Figure 3C), and qualitatively there appears to be a reduced rate of lysis via bulging. We speculate that this difference in phenotype may reflect a change in turgor pressure upon excess GFP production, but we did not investigate it further. Altogether, in combination with our new finding that single cell elongation rate is invariant among the mutants across pH conditions (a suggestion of reviewer 3), this new data strongly supports our model that the bulk culture growth rate defect of the mutants at their respective non-permissive pH condition is due to cell lysis.

Minor Comments:

1) Delete repeated "in".

Modified in text.

2) Subsection “Identification of pH specialist cell wall synthases and hydrolases”, second paragraph: I think it should state "six" mutants. Not "seven".

The reviewer is correct. We have modified the text to five to reflect the new number of ‘hits’ after altering our statistical analysis and conducting further validation studies (see comments to reviewers 1 and 3).

3) Subsection “aPBP activity ensures fitness across a wide pH range”, first paragraph: Denome et al., 1999, might not be the best reference to cite the synthetic lethality of PBP1a and 1b because the pbp1a mutant used in that study contained additional mutations, as described in Meberg et al., J Bacteriol. 183:6148-9; PMID: 11567017. It would be more appropriate to cite the first report of the synthetic lethality: Suzuki et al., 1978.

We thank the reviewer for catching this. We have replaced the reference with the one suggested in the text.

Reviewer #3:

[…] Major points

1) Credit should be given to authors for their effort to systematically detect growth defects of mutants in cell wall enzymes across different pH's (Figure 1). However, since most of the effects are small, statistics and experimental setups need be stronger/more transparent to substantiate claims. n=3 (Figure 1 and Figure 1—figure supplement 1) is very low to call 5% effects and to do t-tests (which assume normal distributions). It is also unclear how many measurements and what window of OD's are used for the exponential growth fit, and how good fits are at the end. Representative examples as a supp. figure would be extremely useful, in addition to providing the raw data and fits as supp. data. Last, there is no mention if authors look for/normalize plate effects (positional biases) when measuring for growth. This is very common in microtiter assays, and could very well be driving some of the small differences.

We thank the reviewer for encouraging us to provide more information on our experimental set up and data analysis, and we wholeheartedly agree that growth rate measurements can be highly sensitive to uncontrolled variables. We have rigorously designed our screen to minimize sources of uncontrolled variation. We have added additional information in the text to clarify our experimental and analysis pipeline, including subsection “Identification of pH specialist cell wall synthases and hydrolases”, in the Materials and methods, Figure 1—figure supplement 1, Figure 1—figure supplement 2, and Supplementary file 3 to address these concerns.

As we state in the original version of the manuscript, growth rates were determined by least-squares fitting of growth curves between the OD600 values of 0.005-0.1 for three independent replicates per mutant. This analysis was performed in R. For transparency, we now provide sample growth curves, fits, and fit statistics for a subset of our mutants in Figure 1—figure supplement 1. Best fit lines with R2 values of < 0.95 were excluded from further analysis. In addition to the representative curves and fits, we have included the growth rates (as measured in mass doublings per hour +/- standard deviation) for all tested mutants at pH 4.8, 6.9, and 8.2 in Supplementary file 3 (n=3 for all measurements). No positional effects were observed. As suggested by reviewer 1, we re-assessed significance of our hits using a one-way ANOVA, normalized for multiple comparisons; 5 of the 6 original hits remained significantly attenuated with this analysis.

To address the concern that some of our hits confer relatively small reductions in growth rate, we added an additional layer of validation to our analysis. We reasoned that a consistent growth defect across a discrete range of pH values would be a rigorous way to validate bona fide hits as opposed to statistical aberrations (analogous to what we had previously done for PBP1a and PBP1b-defective cells). To this end, we compared the growth rate of each mutant to the WT across a range of pH values (pH 4.8-8.4). Six biological replicates spread across at least two days were performed. Excitingly, all five of the mutants tested demonstrated consistent, statistically significant growth rate defects across a contiguous range of pH values (see Figure 1—figure supplement 2 and Figure 2). This data is now plotted in both absolute doublings per hour as well as% doublings per hour of the parental strain to allow for easier comparison between the mutants’ growth across different pH conditions (see point #2). This analysis strongly suggests each of the mutants pulled out in our screen exhibits a specific and reproducible pH sensitivity. We would like to further note that relatively small differences in mutant growth rate may lead to significant fitness defects in natural environments with competition for limited resources.

Since there are now a number of studies published in which the fitness of E. coli genome-wide mutant libraries has been probed across many different conditions (PMID 21185072, 27355376, 29769716), the authors may want to compare their results to these studies. pH is certainly one of these conditions. This way they can also look for further evidence of more conditional-specific roles for cell wall enzymes.

We thank the reviewer for making us aware of these resources. Excitingly, loss of function mutants in the genes encoding PBP1b and MepS both have acid-specific growth defects for colony formation (PMID 21185072). We have added a sentence to the text (subsection “Class A PBP activity ensures fitness across a wide pH range”) to reflect this.

2) The are several pieces of data that make me doubt about the validity of the claims about the PBP1a importance/specialization in alkaline pH. I would suggest to remove (the emphasis from) this part, unless authors provide much stronger evidence.

In growth assays, the specific defect of the mrcA mutant in alkaline pH is questionable. Effect is marginal at ph8.2 in Figure 1A (mrcA seems to also have a very mild defect in all pHs in same Fig) and Figure 2A mentions no n (done once?) and has no error bars to assess the significance of differences. Besides the two figures disagree; in 2A, the biggest difference is for ph~7 (difference gone for ph>8) and in 1A is at pH=8.2. Also speaking against of specialized role of PBP1A-LpoA at alkaline conditions, the lpoA mutant has the same small growth defect in all pH's (Figure 2C- if anything smaller at alkaline pHs)!

In microscopy, the cell death of mrcA is only transient (Figure 3B), so this cannot be the reason for the change in steady state growth as authors suggest (subsection “aPBP activity promotes cell wall integrity across pH environments”, second paragraph). Such an effect would result into also a transient growth defect in batch culture, which will be invisible the way authors do their experiments (as they back dilute 10^3 cells/ml so first doublings are in ODs that are below level of detection). Only way to explain batch growth rate defect (if it holds) is with an accompanying change in single cell growth rate, which authors can measure.

The data on PBP1A activity (Figure 5E) at different pH's are again not convincing. Signal is very low (btw this is completely the opposite from Figure 5D, where PBP1a has the strongest signal!), kinetics are fast (so many points and replicates are needed) and this seems to be the only replicate for measuring rates. It is also unclear how fits are done for Kobs (what type of nonlinear regression?), but they are definitely bad for ph5 (it seems as if a linear fit is shown). Besides more replicates with more points in the Bocillin assay, an in-vitro PBP1a assay may allow to see better kinetics and would strengthen any argument. Please keep in mind that pH7 that experiments are done for Figure 5E are not alkaline but neutral…

Overall, I find little evidence that PBP1a-LpoA takes over in alkaline conditions (as implied in text and Figure 7). Even if all effects are validated, at best it means that PBP1a/LpoA have a role under these conditions, but PBP1b/LpoB can compensate to a large degree.

While we acknowledge that the growth defect for the PBP1a-defective strain in neutral/alkaline media is modest in comparison to the complete loss of growth of the PBP1b-defective strain at pH 4.8, we stand by our assertion that PBP1a is required for maximal fitness in this pH range.

At least part of this reviewer’s concern seems to stem from misinterpretation of the data extrapolated from the figures. For example, although the reviewer claims there is a discrepancy between Figures 1A and 2A in terms of which pH PBP1a-deficient cells have the greatest maximal growth defect, in both of these figures, PBP1a-deficient cells have a greater growth defect at pH 8.2 than 6.9. To avoid confusion, we have added a table denoting the mean growth rate +/- SD for each mutant tested in our original screen (Supplementary file 3), as well as panels in Figure 2 and Figure 1—figure supplement 2 depicting the% WT growth (DPH for mutant/DPH of wild-type x 100). We hope that this representation of the data allows for more accurate comparisons among the mutants and pH conditions. We have also updated our figure legends to reflect the n value for each experiment. Figure 2A-C, for example, has an n = 6 for each strain (3 replicates were presented in the original draft of the manuscript, but the n was increased to 6 in response to this reviewer’s concern in point #1). Our findings in these experiments reveal a reproducible and significant defect for PBP1a from pH 6.9-8.2.

In this revision, we also repeated the pH sensitivity testing of the lpo mutants at an extended pH range. Cells defective for LpoA display a comparable range and magnitude of growth defect compared to PBP1a defective cells. (updated Figure 2C; Figure 2—figure supplement 2). These data provide additional support for a specific role for PBP1a in these conditions.

In response to the reviewer’s concern that the transient lytic phenotype for the mutant defective for PBP1a could not underlie the bulk culture growth defect, we examined lysis kinetics in liquid culture to more closely mimic the conditions of our initial screen. Briefly, cells were cultured to mid-exponential phase in pH 6.9 media, back-diluted to an OD600 of 0.005 at pH 8.2 and sampled at various time points for propidium iodine (PI) staining via fluorescence microscopy. As shown in Figure 3—figure supplement 1D, up to 10% of the cells lysed up to 3 hours post-alkaline shock with no observable recovery. Although we hoped to extend the experiment beyond 3 hours, at cell densities with OD600 values > 1, the buffering capacity of the media was overwhelmed, and the media began to decrease in pH. In light of this finding, we speculate that the recovery phenotype observed on the agarose pad may reflect a decrease in local pH upon increasing cell density, protecting the cells against further lysis. In tandem, we measured single cell elongation rate for each of the mutants and parental strain from our time lapse videos at pH 4.5 and 8.0 in SuperSegger, a MATLAB based cell segmentation program. As shown in Figure 3A-B, no differences in elongation rate were observed between the strains at each pH, implicating lysis as the sole source of the bulk culture growth defect.

The reviewer has valid concerns about the Bocillin assays; our signal is low and inconsistent (likely in part due to a technical limitation in using an LED-based imager instead of a fluorescence scanner with higher sensitivity, as is typical for these experiments), and these assays fail to differentiate between intrinsic and extrinsic effects on enzyme activity. In consideration of this, we have removed them from the manuscript and instead replaced them with in vitroassays of biochemical activity conducted in collaboration with Waldemar Vollmer’s group, as suggested by reviewer 1. Briefly, we find that PBP1a has negligible glycosyltransferase and transpeptidase activity at pH 4.8, which causes the cell to be reliant on PBP1b activity in this pH range (Figure 5D, E; Figure 5—figure supplement 3; Figure 5—figure supplement 4). Although we do not observe any differences in PBP1b activity that account for the preference for PBP1a in more alkaline conditions, these phenotypes may be too subtle to capture in these assays, or alternatively, differences in the aPBP’s subcellular localization and/or interactions partners may also contribute to their pH specialization. Future work will be necessary to evaluate these possibilities.

3) Increased resistance to a sub-class of β-lactams at low pH is due to increased PBP1B activity. Although this could very well be (effect gone in mrcB mutant and controls in Figure 6—figure supplement 3C and d provide strong evidence), everything could be also explained with some of these drugs having decreased entry at low pH. Note that β-lactams have differential uptake and efflux preferences – Aztreonam for example needs OmpF to enter cell (see PMID – 29980614). Less drug in the cell could explain also why cells treated with Cephalexin or Mecillinam need more drug to change their morphology at pH5, although presumably the targets of the drugs, PBP3 and PBP2 do not change levels.

Measuring drug intracellular concentrations would be required to exclude such a scenario. Or alternatively some of the experiments in point 4 could strengthen the PBP1B increased activity in complexes. Also including more non-specific/broad β-lactams would (Carbenicillin, Amoxicillin) would help towards this direction. Note that Imipinem is not a selective PBP2 inhibitor- it targets many PBPs in addition to LD-transpeptidases.

These β-lactams act on their targets in periplasmic space, not in the cytoplasm. We assume the reviewer meant to ask us to measure periplasmic concentration of the compounds. To our knowledge, there is no direct method to measure this, and to design one would require significant innovation beyond the scope of the current study.

We have added additional data to the revision to support our conclusion that changes in outer membrane permeability are unlikely to underlie the low pH-dependent resistance phenotype.

A) Additional compounds we have tested and included in this revision match our previously observed results. At the suggestion of the reviewer, we added an additional broad-spectrum antibiotic (amoxicillin) and two additional PBP2 inhibitors (meropenem and doripenem). As expected, MG1655 exhibited reduced susceptibility to the PBP2 inhibitors at pH values < 6.0. In contrast, the MIC to amoxicillin was only mildly elevated (2-fold) at pH values < 5.0. (compared to >8-fold for the majority of PBP2/PBP3-specific compounds). Importantly, while meropenem and doripenem do target the L,D-transpeptidases (similar to imipenem, as the reviewer noted), cells that lack three or six L,D-transpeptidases still exhibited a comparable increase in the drugs at pH 5.0 (Figure 6—figure supplement 4E), indicating that these enzymes do not play a role in this phenotype. As an aside, we have removed imipenem from the study because we found it to be unstable at pH 4.5.

B) As a separate means of addressing the contribution of OM permeability to our phenotype, we treated cells with sub-inhibitory concentrations of polymyxin B, a compound which forms pores in the outer membrane. We reasoned that this treatment would non-specifically compromise outer membrane integrity in cells independent of pH. Thus, if our pH-dependent changes in MIC were due to differences in outer membrane integrity, there would be no difference in CEX MIC at pH 7 and pH 5.0. Instead, we observed the same fold-change in MIC as cells not treated with polymyxin B. Importantly, the concentrations of Polymixin B used in these assays decrease the cells’ CEX MIC at both pH values, indicating comparable outer membrane disruption (Figure 6—figure supplement 3C). This data offers strong evidence that low pH dependent resistance cannot be explained solely by changes in outer membrane permeability.

4) The authors propose 3 models to rationalize how PBP1b increased activity at low pH could make cells more resistance to PBP2/3 inhibitors. They favor the one which PBP1b replaces PBP2 and PBP3 in the elongasome and divisome (as increased repair could for example not explain the need for increased Cephalexin to see filamentation at low pH). This scenario has some straightforward ramifications that are rather easy to test: PBP2/3 should be less active at low pH (Bocillin assay), PBP1b/2/3 dynamics may differ at low pH, cells should tolerate more depletion of PBP2/PBP3 at low pH. Also it could be interesting to test if Salmonella shows the same higher resistance at low pH (in PBP1b-dependent manner) when PBP3Sal is knocked out but not when there (because PBP3Sal is still as sensitive to PBP3 inhibitors).

We agree with the reviewer that this is a very interesting question and appreciate their suggestions on ways in which we can test this model in future studies. However, we believe rigorously addressing this model necessitates a thorough investigation into the effect of pH on the composition and the activity of both the elongation and division machinery, which is beyond the scope of the current manuscript. Consequently, we have significantly revised the Discussion to decrease the emphasis on this point.

5) Quantification and statistics suffer throughout the paper. Replicates are not always there (Figure 2A, Figure 5E) and many times too little for the statistics authors want to do: standard deviation from 2 replicates (5C, 6D), t-tests from sets of 3 experiments (Figure 1). Microscopy is not always quantified (Figure 5A, 6B/C), and the most worrisome is that panels do not always agree (Figure 1-2, or Figure 5D and 5E). All this should be amended for the conclusions to be on solid ground.

As the reviewer suggested, we have re-evaluated our statistical test choice (Figure 1; Figure 2D), added additional replicates (Figure 2A-C; Figure 5C; Figure 6D), and quantified our microscopy (Figure 5A, B; Figure 6B, C).

Minor Comments:

1) Introduction, second paragraph: the reason for the high sensitivity of the periplasm is not the different permeability of the two membranes; please rephrase.

It is a widely held belief that differential membrane permeability contributes to pH and osmotic sensitivity of the periplasm. The outer membrane contains porins, including OmpC and OmpF, that are permissive to ions, protons and water; these porins are not present in the inner membrane, and thus the inner membrane is often considered the major permeability barrier for Gram negative bacteria. If the reviewer can point us to references that refute this point or elaborate on why they took issue with this statement, we are happy to revise accordingly.

2) Introduction, second paragraph: careful, there is a homeostatic control system: HdeA and HdeB are periplasmic chaperones induced and required during acidic conditions; quite some things known about their function. Also there is some literature on the requirement of other housekeeping periplasmic quality control enzymes during acid shock.

We have removed the clause “… in the absence of a homeostatic control system.”

3) Introduction, fifth paragraph: why curiously?

Rewritten.

4) Introduction, last paragraph: one "in" too much

Amended.

5) Throughout the text: "insertional deletions". Mutations come either from insertions or deletions. Since these are mutants from the KEIO collection, they are deletions.

We have changed “insertional deletions” to deletions in all instances in the text.

6) Subsection “Identification of pH specialist cell wall synthases and hydrolases”, first paragraph: commensal E. coli (MG1655 is not even that) are not specialized for growing in urine. UPEC strains carry ~2,000 genes more than MG1655, so this leaves quite some room for other alkaline-specialized cell wall enzymes.

We agree with the reviewer that the pangenome of UPEC and intestinal E. coli exceeds that of MG1655, and importantly, we never claim to identify an exhaustive list of pH-specialized cell wall enzymes. While it would be interesting to investigate pH-specialized cell wall enzymes in UPEC, selecting a representative strain to conduct this analysis would be challenging for several reasons. In particular, UPEC strains exhibit remarkable genetic diversity and lack a conversed genetic signature for urovirulence (Schreiber IV et al., 2017).

7) Subsection “aPBP activity ensures fitness across a wide pH range”, first paragraph: effect is not dramatic for mrcA.

The word ‘dramatic’ has been removed.

8) Subsection “aPBP activity ensures fitness across a wide pH range”, second paragraph: how is p-val calculated with no replicates here?

As addressed above, n = 3 for this experiment.

9) Subsection “aPBP activity ensures fitness across a wide pH range”, second paragraph (Figure 2—figure supplement 1); could be informative to test pbpC double mutants with mrcA and mrcB to check if the effect of PBP1c is masked by any of the other two.

As suggested, we constructed these mutants and tested for pH-dependent growth rate defects from pH 4.8-8.4. Loss of PBP1c did not exacerbate any of the growth defects, indicating its effects are not masked by the presence of the other aPBPs (Figure 2—figure supplement 1).

10) Subsection “aPBP substrate binding is pH-dependent”, first paragraph/Figure 5A: please quantify.

As suggested, we have now quantified PBP1a localization (see Figure 5—figure supplement 2). See comments to reviewer 1 for additional commentary on localization profile.

11) Subsection “pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific β-lactam antibiotics”, first paragraph: mrcB and lpoB mutants are very sensitive also to Cefsulodin, and a number of broad-acting cell wall enzymes. In contrast mecillinam effects are mild. These facts are hard to reconcile with line of thought that follows (see the second paragraph of the aforementioned subsection).

This section has been removed from the text.

12) Subsection “pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific β-lactam antibiotics”, third paragraph/Figure 6D: other drugs and n>2 would be helpful to show that effect is specific.

We repeated this experiment with fresh urine (n=3) and results are shown in an updated version of Figure 6D.

13) Subsection “pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific β-lactam antibiotics”, last paragraph: referring to Figure 6E not 6D.

Corrected.

14) Subsection “Specialization role for aPBPs in cell wall integrity across environmental conditions”, first paragraph: if effects are direct on enzyme activity cannot be deduced from the Bocillin assay; many other upstream effects could compromise enzyme activity.

We agree. As previously mentioned, we have removed the Boc-FL binding assays from the manuscript.

15) Subsection “Specialization role for aPBPs in cell wall integrity across environmental conditions”, second paragraph: first time I hear about depressed enzymes…

Revised.

16) Subsection “Plasticity in cell wall metabolism potentiates intrinsic resistance to cell wall active antibiotics”, first paragraph: I find this highly speculative. Slow growth may as well be playing a more active role in resistance development. Also these microbes can become equally well resistant to broad cell wall inhibitors as they do to PBP3 inhibitors

We have revised this section to soften the language.

17) Subsection “Bacterial strains, plasmids, and growth conditions”: are strains re-transduced (once)? One clone used? Mutation checked?

Two transductants/clones were tested per strain. Positive hits were confirmed by diagnostic PCR. The Materials and methods section has been updated to reflect this.

18) Figure 3B: would be nice to see a longer experiment (past 2hrs); also can make points/schemes smaller to see error bars.

Unfortunately, after 2 hours cells grown at pH 8.0 become too dense on the agarose pads to accurately quantify. As previously mentioned, our liquid culture experiment with the PBP1a-defective strain extends to 3 hours.

19) Figure 5D/E: inconsistent. PBP1a has the strong signal in d, quantified as low signal in e. PBP1b effect is ~ 2-fold at 15 min at Figure 5D, but looks less in 5E.

As previously mentioned, we have removed the Boc-FL experiments that this comment refers to.

20) Figure 6A: What happens with Mecillinam at pH>7?

At present, is unclear how alkaline pH is conferring mecillinam resistance. We chose not to investigate this further as it was not conserved across other PBP2 inhibitors. However, the mechanism of high pH-dependent resistance is likely distinct from that observed in acidic conditions, as rod shape is not similarly preserved (see Figure 6—figure supplement 2).

21) Figure 6—figure supplement 1: Azt is unstable at higher pH, but this does not correlate with increase in MIC (Figure 6)?!

We anticipate that this “discrepancy” is due to the fast-acting lytic activity (<2 hours) of most β-lactams, including PBP3 inhibitors (see Fredborg et al. BCM Microbiol., 2015 for example lysis kinetics for piperacillin). Therefore, AZT treatment likely kills cells faster than it is hydrolyzed and inactivated at pH 8.0.

22) Figure 6B/C: quantitative data would be more convincing.

As suggested, we have now quantified cell length and aspect ratio for cells treated with sub-MIC levels of cephalexin and mecillinam, respectively (see revised Figure 6B/C).

23) Figure 6—figure supplement 3 legend: Define pH's you are comparing.

Amended.

https://doi.org/10.7554/eLife.40754.037

Article and author information

Author details

  1. Elizabeth A Mueller

    Department of Biology, Washington University in St. Louis, St. Louis, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5482-6551
  2. Alexander JF Egan

    Centre for Bacterial Cell Biology, Institute for Cell and Molecular Biosciences, Newcastle University, Newcastle upon Tyne, United Kingdom
    Contribution
    Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Eefjan Breukink

    Membrane Biochemistry and Biophysics, Department of Chemistry, Faculty of Science, Utrecht University, Utrecht, Netherlands
    Contribution
    Resources, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Waldemar Vollmer

    Centre for Bacterial Cell Biology, Institute for Cell and Molecular Biosciences, Newcastle University, Newcastle upon Tyne, United Kingdom
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Petra Anne Levin

    Department of Biology, Washington University in St. Louis, St. Louis, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing—review and editing
    For correspondence
    plevin@wustl.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2071-0547

Funding

National Science Foundation (DGE-1745038)

  • Elizabeth A Mueller

Wellcome (101824/Z/13/Z)

  • Waldemar Vollmer

National Institutes of Health (GM127331)

  • Petra Anne Levin

National Institutes of Health (GM64671)

  • Petra Anne Levin

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Tom Bernhardt and Joe Vogel for gifts of strains and plasmids, respectively. We appreciate sample preparation and electron microscopic imaging assistance from Matthew Joens, Daniel Geanon, Greg Strout and Dr. James Fitzpatrick from the Washington University Center for Cellular Imaging which is supported by Washington University School of Medicine, The Children’s Discovery Institute of Washington University and St. Louis Children’s Hospital (CDI-CORE-2015–505) and the Foundation for Barnes-Jewish Hospital (3770). We are indebted to members of the Levin and Zaher labs for fruitful discussions on technical and philosophical matters related to this this research, as well as Corey Westfall, Joseph Merriman, and Katharina Peters for critical reading of this manuscript.

Senior Editor

  1. Gisela Storz, National Institute of Child Health and Human Development, United States

Reviewing Editor

  1. Michael T Laub, Massachusetts Institute of Technology, United States

Reviewer

  1. Michael T Laub, Massachusetts Institute of Technology, United States

Publication history

  1. Received: August 5, 2018
  2. Accepted: March 23, 2019
  3. Version of Record published: April 9, 2019 (version 1)

Copyright

© 2019, Mueller et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,675
    Page views
  • 262
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Immunology and Inflammation
    2. Microbiology and Infectious Disease
    Lu Yu et al.
    Research Article
    1. Evolutionary Biology
    2. Microbiology and Infectious Disease
    Peter Major et al.
    Research Article Updated