Plasticity in Escherichia coli cell wall metabolism promotes fitness and mediates intrinsic antibiotic resistance across environmental conditions

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 PBP1 a and PBP1 b; 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 differential activity across pH environments significantly alters intrinsic resistance to cell wall active antibiotics. Together, our findings reveal previously thought to be redundant enzymes are instead specialized for distinct environmental niches, thereby ensuring robust growth and cell wall integrity in a wide range of conditions.


INTRODUCTION 47
The growth and survival of single-celled organisms relies on their ability to adapt to rapidly 48 changing environmental conditions. A commensal, pathogen, and environmental 49 contaminant, Escherichia coli occupies and grows in diverse environmental niches, 50 including the gastrointestinal tract, urinary bladder, freshwater, and soil. In the laboratory, 51 the bacterium's flexibility in growth requirements is reflected in robust proliferation across 52 a wide range of temperature, salt, osmotic, pH, oxygenation, and nutrient conditions (1). 53

54
The physiological adaptations that permit growth and survival across environmental 55 conditions are not yet well understood, particularly for extracytoplasmic processes. Due 56 to the discrepancy in permeability between the plasma and outer membrane (2), the 57 periplasmic space of Gram negative bacteria is highly sensitive to chemical and physical 58 perturbations, including changes in salt, ionic strength, osmolality, and pH. Notably, upon 59 mild environmental acidification, the periplasm assumes the pH of the extracellular media, 60 while the cytoplasmic pH remains minimally disrupted (3,4). Although mechanisms that 61 contribute to cytoplasmic pH homeostasis have been described (5,6), little is known about 62 the quality control mechanisms that preserve proper folding, stability, and activity of key 63 proteins in the periplasm in the absence of a homeostatic control system. 64

65
The peptidoglycan (PG) cell wall and its synthetic machinery are among the key 66 constituents of the periplasm that must be preserved across growth conditions. Essential 67 for viability among most bacteria, PG is composed of glycan strands of repeating N-68 acetylglucosamine and N-acetylmuramic acid disaccharide units crosslinked at peptide 69 In light of these findings, we hypothesized that loss of an enzyme specialized for a 116 particular environmental niche may produce a condition-specific growth defect through 117 impaired cell wall integrity, allowing us to take a systems-level approach to identifying 118 enzymes with differential roles in growth in vivo. In screening 32 mutants across six 119 classes of nonessential periplasmic cell wall enzymes, we determined that a subset of 120 these enzymes is differentially required for fitness across pH environments. We find that 121 disruptions in in the activity of cell wall synthases PBP1a and PBP1b conferred fitness 122 defects in opposing pH ranges that can be attributed in part to pH-dependent differences 123 in activity. Concerningly, synthase specialization has consequences for intrinsic 124 resistance to b-lactam antibiotics in nonstandard growth conditions. 125 126 RESULTS 127

Identification of pH specialist cell wall synthases and hydrolases 128
To determine the contribution of individual cell wall enzymes to pH-dependent growth, we 129 cultured strains harboring insertional deletions in each of three aPBPs, six L,D-130 transpeptidases, five carboxypeptidases, four amidases, nine lytic glycosyltransferases, 131 and six endopeptidases to mid-exponential phase (OD600 ~0.2-0.6) in buffered LB media 132 (pH 6.9) then sub-cultured them into fresh LB buffered to pH 4.8, 6.9, or 8.2 for growth 133 rate analysis. These pH values were chosen as representative, physiologically relevant 134 conditions E. coli cells encounter in the lower GI tract (pH 5-9) or urine (pH 4.5-8) (26,27). 135 Classification as a pH-sensitive mutant required a significant, > 5% decrease in early 136 exponential phase mass doubling time (unpaired t-test, p < 0.01) in at least one pH 137 condition compared to the parental strain. 138 139 Collectively, seven mutants displayed significant reductions in mass doubling times 140 (MDT) at one or more pH values. We observed both acid-sensitive and alkaline sensitive 141 mutants across three enzymatic classes. Strikingly, loss of the bifunctional synthase 142 PBP1b (mrcB) abolished growth at pH 4.8 but maintained wild type MDT at neutral and 143 alkaline pH (Fig. 1A). Four additional mutants exhibited a distinct, albeit less severe, We also identified two alkaline-sensitive mutants. Loss of the bifunctional synthase 151 PBP1a (mrcA) and the lytic transglycosylase MltG (yceG) impaired, but did not abolish, 152 growth at pH 6.9 (∆mrcA, 5% decrease in MDT; ∆yceG, 10% decrease in MDT) and pH 153 8.2 (∆mrcA, 11% decrease in MDT; ∆yceG, 22% decrease in MDT). Both mutants 154 exhibited wild type growth rates at pH 4.8 ( Figure 1A we elected to focus further efforts on understanding the contribution of the bifunctional 163 aPBPs PBP1a and PBP1b to growth across a range of pH conditions. An accumulating 164 body of evidence suggests the aPBPs play overlapping, and potentially redundant, roles 165 in cytoskeletal-independent PG synthesis during growth in standard culture conditions 166 (i.e. nutrient rich, neutral pH growth media aerated at 37°C) (12,29). Indeed, PBP1a and 167 PBP1b are a synthetic lethal pair in E. coli in these conditions (20). 168

169
Based on their disparate pH-dependent growth defects, we hypothesized that PBP1a and 170 PBP1b are specialized synthases whose activity is optimized for growth under distinct pH 171 environments. Consistent with this model, cells defective in PBP1a (DmrcA) and PBP1b 172 (DmrcB) displayed defects in MDT at discrete, non-overlapping pH ranges. Between pH 173 5.1-5.9, loss of PBP1b resulted in a 5-22% reduction in MDT compared to wild type cells 174 (p < 0.01), and growth was abolished at pH values below 4.8 ( Figure 2A). The growth of 175 the parental strain was not prevented until pH 4.2, over half a pH unit lower than the 176 mutant ( Figure 2B). At pH values at or above 6.2, MDT of this mutant was 177 indistinguishable from wild type cells. Importantly, pre-conditioning the mutant in acidic 178 media (pH 5.5) did not abrogate the growth rate defect ( Figure 2B), indicating that steady-179 state pH-rather than pH shock-underlies the defect in MDT. In contrast to loss of 180 PBP1b, the MDT of cells defective for PBP1a was equivalent to the parental strain during 181 growth in acidic media (pH range 4.8-5.9), yet this mutant reproducibly exhibited between 182 a 4-15% decrease in MDT from pH 6.2 to 8.2 (p < 0.05). The growth rates of mutant and 183 wild type strains were not statistically distinguishable at pH 8.4 (Figure 2A). Loss of PBP1c 184 We next sought to test whether aPBP transpeptidase and/or glycosyltransferase activity 188 were required for fitness across pH conditions, as opposed to an indirect, structural role 189 for these enzymes in the formation of cell wall synthesis complexes (31,32). To test this, 190 we took advantage of two sets of mutants: 1) insertional deletions in lpoA and lpoB-191 genes encoding outer membrane lipoprotein cofactors required for activity, but not 192 expression or stability, of PBP1a and PBP1b, respectively (33-36) and 2) point mutations 193 that inactivate PBP1b transpeptidase and/or glycosyltransferase activity but do not impact 194 stability (37).  Analogous to cells defective for PBP1b, deletion of lpoB prevented growth at pH 4.8. 199 Likewise, loss of PBP1a's cofactor LpoA led to a growth rate defect at pH 6.9 and pH 8.2, 200 although this mutant also exhibited a slight, but statistically significant, defect in MDT at 201 pH 4.8 ( Figure 2C). Complementation analysis of PBP1b variants at acidic pH further 202 bolstered the conclusion that activity is required for pH-dependent growth. As expected, 203 production of wild-type PBP1b in trans restored growth of the ∆mrcB mutant at pH 4.8; 204 however, production of PBP1b variants rendering the transpeptidase (S510A, TP*), 205 glycosyltransferase (E233Q, GT*), or both enzymatic activities inactive (TP*GT*) failed to 206 complement growth ( Figure 2D). It should be noted that consistent with observations that 207 aPBP transpeptidase activity cannot be assayed in vitro in the absence of functional 208 glycosyltransferase activity, the mutation in the glycosyltransferase active site (E233Q) 209 previously has been shown to attenuate transpeptidase activity by 90% (35,38-40); thus, 210 although our data demonstrate that transpeptidase activity is critical for pH-dependent 211 growth, we cannot discern whether glycosyltransferase activity alone is required. 212 213 aPBP activity promotes cell wall integrity across pH environments 214 Although these findings establish PBP1a and PBP1b activity as essential for optimal 215 fitness across a wide pH range, it remained unclear whether these mutants' pH-216 dependent defects in MDT in bulk culture were due to reduction in growth across the 217 population (i.e. decreased rate of mass accumulation and cell expansion) or lysis of a 218 fraction of cell in the population. To differentiate between these two mechanisms, we 219 inoculated early exponential phase (OD600 ~ 0.05-0.1) wild type or mutant cells pre-220 grown at pH 6.9 on to agarose pads buffered to pH 4.5 or pH 8.0 and examined cells for 221 incorporation of the dye propidium iodine (PI), which permeates cells with compromised 222 membranes, by microscopy. 223

224
The pH-dependent growth defects were lytic in origin: extensive cell death was observed 225 at one hour post-shift for PBP1b and PBP1a mutants that underwent acid (pH 6.9 to pH 226 4.5) or alkaline (pH 6.9 to pH 8.0) shock, respectively ( Figure 3A). Time-lapse imaging of 227 cells following pH shift shed light on lysis kinetics: upon pH downshift, ∆mrcB cells began 228 to incorporate PI by 30 minutes (~15% cells labeled), with ~95-100% of cells labeled by 229 two hours post-shift. Negligible cell death was observed for the parental strain or for cells 230 defective for PBP1a during equivalent acid shock. Conversely, up to 15% of ∆mrcA cells 231 underwent lysis an hour following alkaline shift (pH 6.9 to pH 8.0), comparable to the 10-232 15% growth defect observed in bulk culture at similar pH shifts ( Figure 3B; Figure 2A). 233 ∆mrcB and wild type cells exhibited minimal (< 5%) or negligible cell death, respectively, 234 in response to alkaline shift. Interestingly, recovery of the ∆mrcA mutant was observed 235 90 minutes post-alkaline shock, suggesting cell damage may be sensed and initiate a 236 feedback mechanism to restore cell wall integrity and fitness. 237

238
In addition to displaying differential lysis kinetics under their respective non-permissive 239 pH conditions, PBP1a and PBP1b mutants also differed in apparent lysis mechanism.   Although our data support a model in which aPBP activity is differentially required for 258 fitness across pH environments, the mechanistic basis for pH specialization remained 259 unknown. To interrogate this, we compared the production, localization, and substrate indicating that PBP1a has reduced activity in this condition. 295 296 pH-dependent PBP1b activity alters intrinsic resistance to PBP2 and PBP3 specific 297 b-lactam antibiotics 298 PBP1b activity has previously been implicated in intrinsic resistance to b-lactam 299 antibiotics, specifically to compounds that target the elongation specific transpeptidase 300 PBP2 and the division specific transpeptidase PBP3. Strains defective for PBP1b or 301 lipoprotein activator LpoB show enhanced susceptibility to PBP2/PBP3-specific 302 compounds through rapid lysis (42,45,46), and elevated PBP1b activity was recently 303 shown to protect against PBP2 specific antibiotic mecillinam (15). 304 305 Given our finding that low pH enhances PBP1b activity and previous reports of b-lactam 306 tolerance in acidic media (47), we predicted that growth in acidified media may decrease 307 susceptibility to antibiotics that specifically inactivate PBP2 and PBP3, and that the 308 observed resistance should require PBP1b activity. If true, condition-dependent intrinsic 309 resistance may have important implications for treatment of E. coli infections in host 310 niches with variable pH (26,27). To test this model, we exposed our wild type strain 311 In support of our hypothesis, we observed a 2 to 64-fold increase in MIC to compounds 315 that selectively inactivate PBP2 and PBP3 (48) at pH values < 6.0 ( Figure 6A). Acidic pH 316 appeared to confer a protective effect on the elongation and division machinery: cells 317 cultured in low pH media retained near-normal morphology in the presence of 318 concentrations of the compounds that led to either filamentation (cephalexin, CEX) or cell 319 rounding (mecillinam, MEC) at pH 7.0 ( Figure 6B, C). In contrast, susceptibility to a non-320 specific b-lactam (ampicillin, AMP), an aPBP-targeting compound (cefsulodin, CFS) (48), 321 or a protein synthesis inhibitor (chloramphenicol, CH) was not strongly pH-dependent. 322 We additionally confirmed that low pH-dependent resistance could not be attributed to 323 stability of the antibiotics, loss of the proton motive force, or b-lactamase production 324 ( Figure 6-Figure Supplement 1, 2). Changes in outer membrane permeability, another 325 common mechanism for b-lactam resistance, would be expected to affect all b-lactams 326 equally and thus is unlikely to underlie pH-dependent resistance. Importantly, resistance 327 was not limited to our laboratory strain, as uropathogenic E. coli isolate UTI89 (49) 328 exhibited a comparable pH-dependent change in MIC to both cephalexin (CEX) and 329 mecillinam (MEC) at low pH during growth in both broth culture and in urine ( Figure 6D). Apart from activity, the spatial preferences of the enzymes are likely to contribute to 360 specialization as well: for example, PBP1a peripheral localization may lead to unfilled 361 gaps in mid-cell PG and thus contribute to the septal lysis phenotype of the DmrcB mutant 362 ( Figure 5). Importantly, as few differences in peptidoglycan composition are observed in 363 E. coli cells grown in pH 7.5 and pH 5.0 media, it seems likely that the differential 364 requirement for the aPBPs reflects a change in pH-dependent changes in enzymatic 365 activity rather than altered cell wall structure (23). 366 367 Interestingly, apart from displaying depressed capacity for substrate binding in alkaline 368 conditions, PBP1b also displays increased affinity for substrate in acidic conditions, where 369 our data support it is the predominant aPBP. Furthering bolstering the enhanced activity 370 of the enzyme in this environment, we found PBP1b mediates low pH-dependent 371 resistance to PBP2 and PBP3-targeting b-lactam antibiotics ( Figure 6E). Although these 372 findings appear in conflict with biochemical data demonstrating PBP1b 373 glycosyltransferase activity is reduced under acidic conditions in vitro (35), we reason that 374 this discrepancy may indicate that pH-dependent changes to PBP1b activity may be 375 mediated through a regulator rather than through a direct effect on enzymatic activity 376 itself. Although the outer membrane activator LpoB is an attractive candidate, it is unlikely 377 the sole activator, as it fails to significantly enhance PBP1b activity in acidic conditions 378 were not completely interchangeable. Beyond possessing distinct interaction partners 390 and subcellular localization profiles (31,32,53), mutants display differential susceptibility 391 to antibiotic treatment, osmotic shifts, and mechanical stress (29,42,54). We anticipate 392 further study of these synthases-as with other "redundant" cell wall autolysins-under 393 nonstandard culture conditions will continue to reveal unique roles for these enzymes in

Apparent redundancy ensures fitness across environmental conditions 431
Although clearly of consequence for antibiotic resistance, environmental specialization of 432 cell wall enzymes likely functions more broadly as a key adaptation allowing E. coli to 433 thrive across an unusually wide pH range (pH ~4-9) and even tolerate extreme pH shocks, 434 such as during transient exposure to gastric acid (pH ~2) (61). This physiological 435 adaptation to preserve essential processes in the periplasm likely works in concert with 436 the organism's ability to modify extracellular pH through the export of acidic and alkaline 437 substrates (62,63). In this context, pH-specialist cell wall enzymes may function in part to 438 maintain cell wall integrity until the extracellular media reaches a growth-permissive pH. 439 440 Among other pH tolerant organisms, strategies employed to expand growth across wide 441 pH ranges are likely to vary-even between closely related species. Instead of differential 442 aPBP activity underling pH tolerance, S. enterica serovar Typhimurium encodes a PBP3 443 paralog, termed PBP3SAL, that is active during septation during growth in acidic 444 environments, including in the organism's intracellular lifecycle (24). As PBP3SAL is 445 restricted to Salmonella, Enterobacter, and Citrobacter spp., alternative mechanisms to 446 cope with changing pH environments must exist. Elucidating the requirements for pH-  Table 1 and Supplemental Table 2

Growth rate measurements 498
Strains were grown from single colonies in glass culture tubes in LB + MMT buffer (pH 499 6.9) to mid-log phase (OD600 ~0.2-0.6), pelleted, and re-suspended to an OD600 1.0 (~1 x 500 10 9 CFU/mL). Cells were diluted and inoculated into fresh LB + MMT buffer at various pH 501 values in 96-well plates (150 µl final volume) at 1 x 10 3 CFU/mL. Plates were grown at 502 37°C shaking for 20 hours in a BioTek Eon microtiter plate reader, measuring the OD600 503 of each well every ten minutes. Doublings per hour was calculated by least squares fitting 504 of early exponential growth (OD600 0.005-0.1). 505

Microscopy and time lapse imaging 507
For time lapse imaging experiments, cells were grown from a single colony in LB + MMT 508 buffer (pH 6.9) to early exponential phase (OD600 ~0.05-0.1) then mounted onto 1.0% 509 agarose pads at pH 4.5, 6.9, or 8.0. Where indicated, propidium iodine was added to the 510 agarose pad at a final concentration of 1.5 µM. Cells were allowed to dry on pads 10 511 minutes prior to imaging. All phase contrast and fluorescence images were acquired on NaPO4, pH 7.4, and 100 μL of fixative (fixative = 1 mL 16% paraformaldehyde + 6.25 μL 526 8% glutaraldehyde). Samples were incubated at room temperature for 15 min, then on 527 ice for 30 min. Fixed cells were pelleted, washed three times in 1 mL 1X PBS, pH 7.4, 528 then resuspended in GTE buffer (glucose-tris-EDTA) and stored at 4°C. 529 530

Scanning electron microscopy 531
Wild type and DmrcB cells were grown to mid-exponential phase in MMT buffered pH 6.9 532 LB media and back-diluted to an OD600 = 0.1 into either pH 6.9 or 4.5 media. Cells were 533 allowed to grow for an additional hour, fixed as described above, and applied to poly-534 lysine coated coverslips. Post fixation, samples were rinsed in PBS 3 times for 10 minutes 535 each followed by a secondary fixation in 1% OsO4 in PBS for 60 minutes in the dark. The 536 coverslips were then rinsed 3 times in ultrapure water for 10 minutes each and dehydrated 537 in a graded ethanol series (50%, 70%, 90%, 100% x2) for 10 minutes each step. Once 538 dehydrated, coverslips were then loaded into a critical point drier (Leica EM CPD 300, 539 Vienna, Austria) which was set to perform 12 CO2 exchanges at the slowest speed. Once

SDS-PAGE and immunoblotting 564
Strains were grown from a single colony in LB prepared at pH 5.5, 7, or 8 to mid-log phase 565 (OD600 ~ 0.2-0.6), back-diluted, and grown to an OD600 ~ 0.4 to achieve balanced growth. 566 An aliquot of each culture was sampled to ensure the pH of the culture remained 567 unchanged from the starting pH value. Samples were pelleted, re-suspended in 2x 568 Laemmli buffer to an OD600 ~ 20, and boiled for ten minutes. Samples (10 µl) were 569 separated on 12% SDS-PAGE gels by standard electrophoresis, transferred to 570 nitrocellulose membranes, and probed with PBP1b antisera. PBP1b antisera (rabbit) was 571 used at a 1:1000 dilution, and an HRP-conjugated secondary antibody (goat α-rabbit) was Cells from minimum inhibitory concentration assays were spotted (5 µL) onto 1.0% 593 agarose pads 20 hours post-treatment and imaged by phase contrast microscopy to track 594 cell morphology in response to antibiotic treatment across pH values. Growth rate was 595 monitored by OD600 in the BioTek Eon plate reader to confirm all cells examined were in 596 the same growth phase and at approximately the same optical density prior to imaging. 597 598

Quantification and statistical analysis 599
A minimum of three biological replicates were performed for each experimental condition. 600 Data are expressed as means +/-standard deviation or standard error of the mean. 601 Where appropriate, a student's two-tailed unpaired t-test was applied with a significance 602 threshold set to p < 0.05 or < 0.01 (Figure 1). Asterisks indicate significance difference as 603 We thank Waldemar Vollmer (plasmids, PBP1b anti-serum) and Tom Bernhardt (strains) 620 for kind gifts critical for completion of this work. We thank Pam Brown and Michelle 621 Williams for advice on the Bocillin-FL labeling procedure, the Goldberg lab for use of their 622 gel imager, and Abbygail Iken for technical assistance with the UTI89 MIC assays. We depicting two biological replicates of PBP1b production at pH 5.5, 7.0, and 8.0 with mean 938 intensity +/-standard deviation (normalized to mean value at pH 7.0) shown to the right. 939 The uncropped gel and the Ponceau staining for total protein levels can be viewed in 940