Peptidoglycan precursor synthesis along the sidewall of pole-growing mycobacteria

1Model, rod-shaped organisms such as Escherichia coli and Bacillus subtilis elongate across a broad swath of the cell. Mycobacterial cells, by contrast, extend from narrower polar regions. Circumscription of growth to discrete zones poses spatial challenges to the bacterial cell. For example, if polar growth and division are the only sites of cell wall synthesis in mycobacteria, the entire lateral surface of the cell must be inert. Such an expanse of non-renewable surface could leave the cell vulnerable to environmental or immune insults.

2Because cell wall peptidoglycan synthesis is critical for bacterial replication, it is often used to localize the sites of growth and division. Intriguingly, d-amino acid probes, which in other species have been shown to incorporate into peptidoglycan, label both the poles and sidewall of mycobacteria. The localization of these molecules is supported by the detection of peptidoglycan synthetic enzymes at the mycobacterial cell tips and periphery. However, both intracellular and extracellular incorporation pathways have been characterized or hypothesized for d-amino acid probes, complicating the interpretation of labeling patterns. Intracellular uptake implies that the probe enters the biosynthetic pathway at an early stage, and therefore marks nascent cell wall. Extracellular incorporation, on the other hand, suggests that the probe enters the pathway at a later stage and/or is part of enzymatic remodeling of the macromolecule in question. The extent to which peptidoglycan synthesis and remodeling are linked is not clear and may vary with species and external milieu. In Mycobacterium tuberculosis, for example, there is indirect but abundant data that suggest that there is a substantial cell envelope remodeling during infection when growth and peptidoglycan synthesis are presumed to be slow or nonexistent.

3An intracellular metabolic tagging method for the cell wall would be an ideal tool for determining whether tip-extending mycobacteria can synthesize peptidoglycan along their lateral surfaces. At least two pieces of evidence suggest that d-alanine-d-alanine dipeptide probes are incorporated into peptidoglycan via the cytoplasmic MurF ligase. First, derivatives of d-alanine-d-alanine rescue the growth of Chlamydia trachomatis treated with d-cycloserine, an antibiotic that inhibits peptidoglycan synthesis by inhibiting the production and self-ligation of d-alanine in the cytoplasm. Second, B. subtilis cells stripped of mature peptidoglycan by lysozyme treatment retain a small amount of dipeptide-derived fluorescence. While these data are suggestive, formal demonstration of intracellular incorporation requires direct evidence that the probe labels peptidoglycan precursors. More broadly, better characterization of the metabolic fate of probes would increase the precision of conclusions that can be drawn from labeling experiments.

4Here, we sought to determine how d-amino acid probes incorporate into the mycobacterial cell wall. Monopeptide d-amino acid probes chiefly reported peptidoglycan remodeling by l,d-transpeptidases while dipeptides marked lipid-linked peptidoglycan precursors. All the probes tested labeled the poles and sidewall of mycobacteria, indicating that cell wall metabolism in these regions comprises both synthetic and remodeling reactions. While peptidoglycan assembly along the mycobacterial periphery did not support obvious surface expansion, it was greatly enhanced by cell wall damage. Such activity may allow editing of a complex, essential structure at timescales faster than those permitted by polar growth.

5Mycobacteria have been shown to expand from their poles but published micrographs suggest that d-amino acid probes may label both the poles and sidewall of these organisms. Metabolic labeling can be achieved by a one-step process, in which the fluorophore is directly appended to the probe, or a two-step process in which a small chemical tag on the d-amino acid is detected by subsequent reaction with a fluorescent reactive partner (, Figure 1A). We first reexamined the localization of various d-amino acid probes reported in the literature, including RADA, which is directly conjugated to 5-carboxytetramethylrhodamine, and the two-step alkyne-d-alanine [alkDA or EDA,], and alkyne-d-alanine-d-alanine (alkDADA or EDA-DA,; we use the metabolic labeling nomenclature originally adopted in) which we detected by copper-catalyzed azide-alkyne cycloaddition (CuAAC) after fixation (Figure 1B and C). After M. smegmatis cells were incubated for ~10% generation in probe, high-resolution and quantitative microscopy revealed that they had asymmetric, bidirectional gradients of fluorescence that emanated from the poles and continued along the sidewall (Figure 2A). Polar gradients of dipeptide labeling were also apparent in live cells when we detected azido-d-alanine-d-alanine (azDADA or ADA-DA,) incorporation by either CuAAC (using low copper, bio-friendly reaction conditions (, Figure 2—figure supplement 1A) or by copper-free, strain-promoted azide-alkyne cycloaddition (SPAAC, Figure 2—figure supplement 1B).

6The mycobacterial cell envelope is comprised of covalently bound peptidoglycan, arabinogalactan and mycolic acids, as well as intercalated glycolipids and a thick capsule (, Figure 1B). Assembly of the envelope layers has long been presumed to be spatially coincident. This is largely based on biochemical data suggesting that ligation of arabinogalactan to peptidoglycan occurs concurrently with crosslinking of the latter by transpeptidases. We and others have found that cytoplasmic enzymes that mediate arabinogalactan and mycomembrane synthesis are enriched at the poles but also present along the periphery of the cell, as is metabolic labeling by OalkTMM and NalkTMM (, Figure 2A). OalkTMM and NalkTMM are trehalose monomycolate derivatives that predominantly mark covalent mycolates and trehalose dimycolate, respectively, in the mycomembrane (Figure 1B,). The azido and alkynyl groups on the different probes are not orthogonal to each other (Figure 1C) so we opted to compare peptidoglycan and mycomembrane labeling patterns by using RADA as a fiducial marker. The cell pole with brighter RADA fluorescence also had more alkDADA or OalkTMM labeling (Figure 2B), suggesting that the polar orientation of peptidoglycan and mycomembrane metabolism is coincident. We then compared the fluorescence intensity profiles of cells that had been individually labeled with the probes, and found similar, average distributions of RADA, alkDADA and OalkTMM at the poles and peripheries of the cells (Figure 2C).

7We next sought to address whether the cell envelope of the related M. tuberculosis is also labeled in polar gradients. We previously showed that alkDA incorporates into the cell surface of the organism but were unable to stain the entire population of bacteria. To investigate the origin of labeling heterogeneity, we first tested whether the structure of fluorophore (Figure 1B) influenced probe incorporation by incubating M. tuberculosis in HADA, NADA or RADA and assessing population fluorescence by flow cytometry. HADA and NADA incubation yielded well-defined fluorescent populations (Figure 2—figure supplement 2A). RADA also labeled the entire M. tuberculosis population, albeit with greater cell-to-cell variability in fluorescence intensity. Given that M. tuberculosis incorporates fluorescent probes HADA and NADA relatively evenly across the population, we hypothesized that the apparent heterogeneity that we previously observed for alkDA labeling was the result of an inefficient CuAAC ligation. We obtained very modest improvements by changing the reaction conditions, more specifically, by swapping the BTTP ligand for the TBTA ligand, altering the ratio of ligand: Cu(I) and increasing the azide label concentration. We also switched our detection moiety to an azide appended to hydroxycoumarin, the same small, uncharged fluorophore as the one-step HADA probe (Figure 1B). Under our optimized conditions we detected azido-coumarin fluorescence from ~5 to 10% of cells that had been incubated in alkDA, alkDADA or OalkTMM (Figure 2—figure supplement 2B and C).

8Although unable to achieve homogenous M. tuberculosis labeling with two-step envelope probes, we decided to test whether the sites of envelope labeling in the limited fluorescent subpopulation resemble those of M. smegmatis. HADA, alkDADA and OalkTMM tagging all produced cells that had a mixture of sidewall and polar fluorescence (Figure 2—figure supplement 2C) but exhibited a higher degree of cell-to-cell variability compared to M. smegmatis. Quantitation of HADA fluorescence showed that peptidoglycan metabolism comprised asymmetric polar gradients when averaged across the population (Figure 2D). We next asked whether there was a cell-wide difference in labeling distribution in M. tuberculosis compared to M. smegmatis. We arbitrarily defined the dimmer polar region as the first 15% of the cell length, the sidewall as the middle 70%, and the brighter polar region as the final 15%. As HADA labeled a large proportion of M. tuberculosis (Figure 2—figure supplement 2A) and was more resistant to photobleaching than NADA, we compared these ratios for HADA in septating and non-septating M. smegmatis and M. tuberculosis labeled for ~10% generation time (Figure 2D; under our growth conditions, M. smegmatis and M. tuberculosis generation times are 2.5–3 hr and 18–20 hr, respectively). Approximately 70–75% of HADA labeling in M. tuberculosis localized to the sidewall compared to 63–64% for M. smegmatis (Figure 2D). Thus, a greater proportion of peptidoglycan metabolism in M. tuberculosis likely occurs along the cell periphery than in M. smegmatis, although we cannot rule out a differential contribution of cyan autofluorescence in the two species.

9As the cell periphery is not known to support surface expansion in mycobacteria, we sought to characterize the molecular processes that underlie d-amino acid labeling patterns. OalkTMM and NalkTMM are inserted directly by the extracellular Antigen 85 complex into the mycomembrane. However, there are three potential pathways by which d-amino acid probes might incorporate into mycobacterial peptidoglycan: an intracellular, biosynthetic route and two extracellular routes, mediated by d,d-transpeptidase or l,d-transpeptidase remodeling enzymes (Figure 3A). Peptidoglycan remodeling, particularly by the l,d-transpeptidases abundantly encoded in the mycobacterial genome, may not strictly correlate with synthesis of the biopolymer. Therefore, we sought to distinguish the different routes of incorporation for d-amino acid probes in mycobacteria.

10It seemed possible that the chemical structure of the derivative (Figure 1B) and/or number of labeling steps (Figure 1A) might influence probe uptake, so we first tested the labeling sensitivity of a panel of d-amino acid derivatives to antibiotics that inhibit potential incorporation routes (Figure 3A). We also assessed OalkTMM and the d-alanine-d-alanine dipeptide probe, the latter of which has been proposed to tag the peptidoglycan of other species via the cytoplasmic MurF ligase [Figure 3A, Figure 3—figure supplement 1,]. d-cycloserine is a cyclic analog of d-alanine that inhibits the d-alanine racemase (Alr) and ligase (DdlA) in mycobacteria. Together with the β-lactamase inhibitor clavulanate, β-lactams like ampicillin block d,d-transpeptidases and d,d-carboxypeptidases. Broader spectrum carbapenems such as imipenem additionally inhibit l,d-transpeptidases. We also included vancomycin, an antibiotic that interferes with transpeptidation and transglycosylation by steric occlusion, to control for general defects in periplasmic peptidoglycan assembly. We empirically determined a time frame of drug treatment that did not compromise M. smegmatis viability (Figure 3—figure supplement 2). Within this time frame, labeling for all single residue d-amino acid probes decreased significantly in response to imipenem or d-cycloserine treatment (Figure 3B). By contrast, labeling by the dipeptide probe was relatively resistant to these antibiotics. Neither vancomycin nor ampicillin had a clear effect on labeling by any of the probes, indicating that cell death, disruption of peptidoglycan polymerization, and abrogration of d,d-transpeptidation do not explain the metabolic incorporation differences. OalkTMM labeling was sensitive to d-cycloserine, but not to imipenem nor ampicillin, suggesting that transfer of mycolates to arabinogalactan may require peptidoglycan precursor synthesis but not transpeptidation.

11To test whether distinct mechanisms of probe incorporation were inhibited by d-cycloserine and imipenem, we performed a chemical epistasis experiment. We first examined the effect of different drug concentrations on alkDA incorporation (Figure 3C). Treatment with either d-cycloserine or imipenem, but not ampicillin, resulted in dose-dependent inhibition of the probe-derived fluorescence that plateaued at approximately half of untreated levels. We then assessed the combined effect of d-cycloserine and imipenem by Bliss independence, a commonly used reference model for predicting drug-drug interactions based on the dose response for the individual drugs. This method is most appropriate when dual inhibition proceeds via distinct mechanisms for example activity against different molecules, enzymes or pathways. The effects of d-cycloserine and imipenem on alkDA incorporation were the same or slightly greater than the potencies predicted by the antibiotics individually (Figure 3D). The primarily additive nature of these antibiotics in the Bliss independence model is consistent with the idea that d-cycloserine and imipenem block distinct pathways of alkDA incorporation, and therefore that the probe incorporates into mycobacterial peptidoglycan via both cytoplasmic and l,d-transpeptidase routes.

12RADA and NADA labeling were more sensitive to imipenem than HADA and alkDA (Figure 3B), at multiple concentrations of drug (Figure 3E). To test whether the probes were differentially incorporated by l,d-transpeptidases, we knocked out three of the six enzymes encoded in the M. smegmatis genome. We chose to focus on LdtA, LdtB and LdtE because the M. tuberculosis homologs (LdtMT1, LdtMT2, and LdtMT4, respectively) have been shown in vitro to have both cross-linking and d-amino acid exchange activity and to be inhibited by imipenem. RADA and NADA fluorescence decreased by ~80–85% the absence of ldtABE, and this effect was complemented by the expression of ldtA alone (Figure 3F). We observed a more moderate effect on HADA and alkDA labeling, which were decreased by ~40–60%. Collectively these data suggest that: 1. RADA and NADA probes primarily report l,d-transpeptidase activity, tetrapeptide substrate, or both, and 2. HADA and alkDA likely label via both cytoplasmic and l,d-transpeptidase routes.

13As RADA labeling in mycobacteria is largely indicative of l,d-transpeptidase activity (Figure 3) and occurs at the poles and along the sidewall (Figure 2), we surmise that peptidoglycan is remodeled at both of these locations. We wished to determine whether remodeling was coincident with biopolymer synthesis but were limited by the multiple incorporation routes of alkDA and HADA (Figure 3). Dipeptide d-amino acid probes have been proposed to report peptidoglycan synthesis in other species via a cytoplasmic, MurF-dependent pathway [Figure 3A, Figure 3—figure supplement 1,]. Consistent with this notion, and in contrast to the monopeptide probes, alkDADA labeling was relatively stable to imipenem treatment (Figure 3B and E) and inefficiently incorporated (Figure 4—figure supplement 1). Thus we were surprised to observe that overall labeling by alkDADA decreased in the absence of LdtA, LdtB, LdtE or combinations thereof (Figure 4—figure supplement 2A). The reduction in signal occurred primarily at the poles (Figure 4—figure supplement 2B) yet loss of the enzymes did not impair bacterial growth (Figure 4—figure supplement 3). As we do not yet understand the mechanistic basis for this observation, we sought to directly test the hypothesis that dipeptide probes incorporate into peptidoglycan precursors.

14d-alanine is produced in mycobacteria by Alr, the d-alanine racemase (Figure 3—figure supplement 1). The molecule is linked to a second d-alanine by DdlA, the d-alanine ligase, and the resulting dipeptide is added to the UDP-MurNAc-tripeptide by MurF. If the alkDADA probe is able to label mycobacterial peptidoglycan via MurF, addition of the molecule to the growth medium should rescue a mutant that is unable to make d-alanine-d-alanine. We first constructed an alr deletion mutant in M. smegmatis and verified that growth is rescued by exogenous d-alanine but not by alkDA (Figure 4A). Although our antibiotic data suggest that alkDA is incorporated into peptidoglycan in part via a cytoplasmic pathway (Figure 3), the inability of this probe to rescue growth was not surprising given the substrate specificities of Ddl and MurF and inefficient synthesis of UDP-MurNAc-pentapeptide with d-amino acids other than alanine. In contrast to the alkDA results, we were able to rescue alr with either d-alanine-d-alanine or its alkynyl derivative (Figure 4A).

15We considered the possibility that alkDADA may be digested by a d,d-carboxypeptidase prior to incorporation. This would result in the release of both unlabeled d-alanine and alkDA, the first of which could account for alr growth rescue (Figure 4A). We reasoned that a more precise gauge for MurF-dependent incorporation of the intact probe would be whether it could support the replication of a strain unable to ligate d-alanine to itself. Therefore, we confirmed that alkDADA rescues the growth of a temperature-sensitive ddlA mutant at the non-permissive temperature (Figure 4B).

16Our genetic data supported a MurF-dependent pathway of alkDADA incorporation into peptidoglycan. If true, the probe should be present in precursors such as lipid I and lipid II (Figure 3—figure supplement 1). To test this hypothesis, we first optimized for mycobacteria a recently reported protocol for detecting lipid-linked precursors. We extracted lipidic species from M. smegmatis and exchanged endogenous d-alanines for biotin-d-lysine (BDL) in vitro using purified Staphylococcus aureus enzyme PBP4, a promiscuous d,d-transpeptidase. Biotinylated species were separated by SDS-PAGE and detected by horseradish peroxidase-conjugated streptavidin. We detected a biotin-linked, low-molecular-weight band that is reduced upon d-cycloserine treatment and accumulates when the lipid II flippase MurJ (MviN,) is depleted or vancomycin is added (Figure 5A), conditions that have been shown to dramatically enhance precursor detection in other species. These data strongly suggest that the BDL-marked species are lipid-linked peptidoglycan precursors. We next turned our attention to detecting lipid I/II from M. smegmatis incubated with alkDADA. Initially we were unable to identify alkDADA-labeled species from organic extracts of wildtype M. smegmatis that had been subjected to CuAAC ligation with picolyl azide biotin (Figure 5B). We reasoned that the proportion of labeled precursors might be below our limit of detection. Accordingly, we repeated the experiment in the Δalr background and found that we could clearly detect a low molecular-weight species band that accumulated with vancomycin treatment (Figure 5B, top) and that ran at the same size as BDL-labeled material (Figure 5B, bottom). We were also able to identify a low-molecular-weight species from Δalr incubated with azDADA that was revealed by either CuAAC or SPAAC ligation to alkyne- or cyclooctyne-biotin, respectively (Figure 5C). Taken together, our genetic and biochemical experiments show that the alkDADA and azDADA probes insert into mycobacterial peptidoglycan precursors by a MurF-dependent route.

17The final, lipid-linked peptidoglycan precursor lipid II is synthesized by MurG on the cytoplasmic side of the plasma membrane then flipped to the periplasm and polymerized [Figure 3—figure supplement 1,]. We previously showed that MurG fused to two different fluorescent proteins and expressed under two different promoters is present at both the poles and periphery of M. smegmatis. Labeling by alkDADA marks similar subcellular locations even with pulses as short as ~1% generation time (Figure 5—figure supplement 1). These data suggest that lipid-linked peptidoglycan precursors are synthesized at lateral sites in addition to their expected localization at the poles. However, our standard experimental protocol for detecting envelope labeling is to perform CuAAC on fixed cells. Because formaldehyde fixation can permeabilize the plasma membrane to small molecules, labeled material may be intracellular, extracellular or both. Dipeptide labeling could therefore read out lipid I/II on the cytoplasmic face of the plasma membrane, uncrosslinked lipid II on the periplasmic side, or polymerized peptidoglycan.

18To shed light on the potential fate(s) of peptidoglycan precursors made at different subcellular sites, we first stained live mycobacterial cells with fluorescent vancomycin. This reagent binds uncrosslinked peptidoglycan pentapeptides and does not normally cross the plasma membrane. Pentapeptide monomers are a low abundant species in M. tuberculosis, M. abscessus and M. leprae peptidoglycan, suggesting that fluorescent vancomycin primarily reports extracellular, lipid-linked precursors in this genus. Labeling of M. smegmatis with this probe revealed both polar and lateral patches (Figure 5—figure supplement 2) as previously noted. This observation suggests that at least some of the peptidoglycan precursors present along the periphery of the mycobacterial cell are flipped to the periplasm.

19We next sought to address whether these molecules could be used to build the peptidoglycan polymer. Transglycosylases from both the PBP (penicillin-binding proteins) and SEDS (shape, elongation, division, and sporulation) families stitch peptidoglycan precursors into the existing meshwork (Figure 3—figure supplement 1,). If peptidoglycan precursors are polymerized along the lateral surface of the mycobacterial cell, at least a subset of these periplasmic enzymes must be present at the sidewall to assemble the biopolymer. Two conserved PBPs in mycobacteria are likely responsible for most of the peptidoglycan polymerization required for cell viability, PonA1 and PonA2 (7, 19, 59). Published images of PonA1-mRFP and PonA1-mCherry localization suggested that the fusion proteins might decorate the mycobacterial sidewall in addition to the cell tips, but the resolution of the micrographs did not allow for definitive assignment. Therefore, we first verified the localization of PonA1-mRFP. We found that a subset of this fusion protein indeed homes to the lateral cell surface (Figure 5—figure supplement 3A).

20We were concerned that overexpression of PonA1-mRFP causes aberrant polar morphology and is toxic to M. smegmatis and about the propensity of mCherry to cluster. Because our attempts to produce PonA1 fusions with different fluorescent proteins were unsuccessful, we opted to take a complementary, activity-based approach. Fluorescent derivatives of β-lactam antibiotics bind specifically and covalently to PBPs, and therefore have been used to image active enzyme in both protein gels and intact cells. Our images of whole cells labeled with Bocillin, a BODIPY conjugate of penicillin, were in agreement with those from a previous publication, and seemed to indicate that Bocillin binds both the poles and sidewall of M. smegmatis (Figure 5—figure supplement 3B and C). However, given the hydrophobicity of the BODIPY dye, we considered the possibility that Bocillin might nonspecifically associate with the greasy mycomembrane. Fluorescence across the cell surface was diminished by pre-treating cells with the β-lactam ampicillin, which prevents peptidoglycan assembly by binding to PBPs, but not d-cycloserine, which inhibits peptidoglycan synthesis in a PBP-independent manner (Figure 5—figure supplement 3B and C). These experiments suggest that at least some of the sidewall labeling of Bocillin is specific, and therefore, that PBPs are present and active in these locations.

21Our data indicate that peptidoglycan precursors are made and likely polymerized both at the poles and sidewall. Peptidoglycan synthesis is often presumed to mark sites of bacterial cell growth. However, dispersed elongation has not been reported in mycobacteria. Accordingly we performed a pulse chase experiment to test whether cell expansion correlates with sites of metabolic labeling. After marking peptidoglycan with RADA, we tracked labeled and unlabeled cell surface during 15 min (~10% generation time) of outgrowth. While we cannot rule out sidewall expansion below our limit of detection, the fluorescence dilution in this experiment was consistent with previous reports and restricted to the mycobacterial poles (Figure 5—figure supplement 4).

22What is the function of peptidoglycan assembly that does not directly contribute to physical expansion of the cell? We hypothesized that one role of growth-independent cell wall synthesis might be repair. More specifically, we reasoned that insertion of peptidoglycan building blocks directly along the cell periphery would enable a real-time, comprehensive response to damage (Figure 6A). Cell wall repair that is restricted to sites of mycobacterial growth, by contrast, would be confined to the poles and renew the cell surface only after several generations. Extended incubation of M. smegmatis (~48 hr) with the peptidoglycan-degrading enzyme lysozyme substantially decreases colony-forming units. We have also shown that spheroplasts generated by combined glycine and lysozyme treatment lack peptidoglycan. Together these data indicate that the enzyme is able to access and damage peptidoglycan in intact cells. We challenged M. smegmatis for 30 min in a mixture of lysozyme and mutanolysin, another enzyme that has been extensively used for in vitro digestion of peptidoglycan. After washing away the enzyme, we assessed the sites of peptidoglycan synthesis by alkDADA labeling. Pre-treatment by the muramidases clearly shifted the fluorescence from the brighter pole towards the sidewall (Figure 6B). These data indicate that mycobacteria reallocate peptidoglycan assembly away from the faster growing pole and toward the periphery upon damage to the cell wall (Figure 6A).

23In this work, we aimed to address the seemingly discrepant observations that, on the one hand, mycobacteria expand from their tips (Figure 5—figure supplement 4,), and on the other, metabolically labeled cell wall and synthetic enzymes are detectable at both the poles and along the sidewall (Figure 5—figure supplement 3). The first step to resolving this conundrum was to unambiguously identify sites of peptidoglycan synthesis. Although the d-amino acid probes that we and others have developed for peptidoglycan labeling have been extensively used for marking the cell wall, in most cases it has not been clear whether they report the location(s) of cytoplasmic synthesis, periplasmic exchange, or a combination of processes. Here, we show that in M. smegmatis the metabolic fate of monopeptide probes is partially dependent on the substituent; whereas NBD and TAMRA-conjugated d-amino acids are primarily exchanged into mycobacterial peptidoglycan by l,d-transpeptidases, their alkyne and coumarin-conjugated counterparts appear to incorporate by both extracellular and intracellular pathways (Figure 3). In the future, biochemical analysis of peptidoglycan composition will allow better quantitation of probe incorporation via different uptake pathways.

24We show that dipeptide probes rescue the growth of a DdlA mutant (Figure 4B) and incorporate into lipid-linked peptidoglycan precursors (Figure 5B and C). To our knowledge, this is the first direct demonstration that peptidoglycan precursors can be metabolically labeled in vivo without radioactivity. Labeling by alkDADA unexpectedly decreased in the absence of l,d-transpeptidases (Figure 4—figure supplement 2). Unlike the monopeptide probes, however, alkDADA-derived fluorescence was stable to pre-treatment with imipenem, an antibiotic that targets this class of enzymes (Figure 3,). These data suggest that the dipeptide is unlikely to be a direct substrate for l,d-transpeptidases. It is possible that d,d-carboxypeptidases cleave a small proportion of alkDADA prior to incorporation and release d-alanine and alkDA. In this scenario, apparent alkDADA labeling in M. smegmatis may be a combination of intracellular alkDADA incorporation via MurF (Figures 4 and 5), extracellular alkDA incorporation by l,d-transpeptidases (Figure 3), and more limited intracellular alkDA incorporation by DdlA/MurF (Figure 3). This model is consistent with the promiscuous targeting of mycobacterial d,d-carboxypeptidases by carbapenems and the ~2000 fold more efficient incorporation of alkDA compared to alkDADA (Figure 4—figure supplement 1). Alternatively or additionally, alkDADA may be degraded abiotically or by other enzymes such as l,d-carboxypeptidases. Finally, loss of l,d-transpeptidases may introduce global alterations in peptidoglycan metabolism. Importantly, despite differences in overall fluorescence between dipeptide-labeled wild-type and ΔldtABE (Figure 4—figure supplement 1A), sidewall fluorescence was preserved in the mutant (Figure 4—figure supplement 1B). In aggregate our data strongly suggest that the mycobacterial periphery is a site of active peptidoglycan synthesis.

25The lateral surface of mycobacteria does not appear to contribute to cell elongation under normal growth conditions (Figure 5—figure supplement 4,) but nevertheless hosts a substantial portion of envelope synthesis and remodeling. The intracellular difference in signal between the poles reflected relative elongation rates, as the RADA-bright cell tip, which coincides with the alkDADA- and OalkTMM-bright cell tip (Figure 2B) grows faster than the RADA-dim cell tip (Figure 5—figure supplement 4). The twofold ratio of fast/bright:slow/dim pole fluorescence (Figures 2C, D and 6B) roughly corresponds to previous estimates of intracellular differences in polar elongation. Compared to M. smegmatis, the distribution of HADA labeling in M. tuberculosis is shifted away from the fast pole toward the periphery (Figure 2D). Diminished polarity and asymmetry is also apparent in the sub-population of M. tuberculosis that is labeled by alkDADA and OalkTMM (Figure 2—figure supplement 2C). Our data are in agreement with the heterogeneity in polar dominance observed previously for M. tuberculosis. Although we cannot rule out a contribution from cyan autofluorescence, these experiments also suggest that sidewall envelope metabolism may be even more prominent in M. tuberculosis than in M. smegmatis, comprising 70–75% of the total cell output.

26It is possible that peptidoglycan assembly along the lateral surface of the mycobacterial cell is simply a byproduct of synthetic enzymes that are en route to the polar elongasome or the divisome. Having active enzymes at the ready could enable efficient coordination between cell growth and septation. We think that this model is less likely, however, given the energetic cost of producing complex macromolecules and the known limits on the steady-state pools of lipid-linked peptidoglycan precursors. Instead we propose that cell wall synthesis along the periphery could allow mycobacteria to edit what would otherwise be an inert surface (Figure 6A). Peptidoglycan and mycomembrane metabolism in this region may thicken or fill in the gaps of envelope that was initially deposited at the poles or the septum and enable the bacterium to correct stochastic defects and repair damage. In support of this model, we find that cell wall synthesis along the sidewall is enhanced upon exposure to peptidoglycan-degrading enzymes (Figure 6B). More broadly, the ability to tailor the entire cell surface, not just the ends, should enable rapid adaptation to external stimuli. Such activity may be particularly important for M. tuberculosis, a slow-growing organism that must survive a hostile, nutrient-poor environment.

27mc2155 M. smegmatis and ΔRD1 ΔpanCD M. tuberculosis were grown at 37°C in Middlebrook 7H9 growth medium (BD Difco, Franklin Lakes, NJ) supplemented with glycerol, Tween 80 and ADC (M. smegmatis) or OADC and 50 µg/ml pantothenic acid (M. tuberculosis). The ∆alr strain was further supplemented with 1 mM d-alanine. The ddlAts strain was grown in 7H9-ADC at 30°C or 37°C as specified.

28Probes used in this study include (i) fluorescent d-amino acids HADA, NADA and RADA (Tocris Bioscience, Bristol, UK) (ii) alkDA (R-propargylglycine, Thermo Fisher, Waltham, MA) (iii) alkDADA and azDADA (Chemical Synthesis Core Facility, Albert Einstein College of Medicine, Bronx, NY) and (iv) OalkTMM and NalkTMM, synthesized as described previously. The labeling procedures were performed with modifications from. Unless otherwise indicated, mid-log M. smegmatis or M. tuberculosis were labeled with 500 μM HADA, 25 μM NADA or RADA, 50 μM alkDA, 1 or 2 mM alkDADA, 50 μM OalkTMM or 250 μM NalkTMM for 15 min or 2 hr, respectively. In some cases M. smegmatis was preincubated ± antibiotics at different fold-MIC (MICs = 80 μg/mL d-cycloserine, 8 μg/mL ampicillin (with 5 μg/mL clavulanate), 0.5 μg/mL imipenem (with 5 μg/mL clavulanate), 6 μg/mL vancomycin) or ± 500 µg/mL lysozyme and 500 U/mL mutanolysin for 30 min then grown for an additional 15 min in the presence of the probes. Cultures were then centrifuged at 4°C, washed in pre-chilled PBS containing 0.05% Tween 80% and 0.01% BSA (PBSTB) and fixed for 10 min in 2% formaldehyde at room temperature (RT). After two washes in PBSTB, the bacterial pellets were resuspended in half of the original volume of freshly prepared CuAAC solution in PBSTB containing either AF488 picolyl azide or carboxyrhodamine 110 (CR110) azide (Click Chemistry Tools, Scottsdale, AZ) or 3-azido-7-c (Jena Biosciences, Jena, Germany). For M. tuberculosis experiments and in Figure 2—figure supplement 1A, we used a modified, low-copper CuAAC reaction protocol: 200 μM CuSO4 and 800 μM BTTP (Chemical Synthesis Core Facility, Albert Einstein College of Medicine, Bronx, NY) were pre-mixed then added to PBSTB. Immediately before resuspending the pellets, 2.5 mM freshly prepared sodium ascorbate and 300 μM 3-azido-7-hydroxycoumarin were added to the mixture. After 30–60 min gentle agitation at RT, cultures were washed once with PBSTB, once with PBS and resuspended in PBS for imaging or flow cytometry analysis (FITC, BV510, and Texas Red channels on a BD DUAL LSRFortessa, UMass Amherst Flow Cytometry Core Facility). NHS ester dye labeling and fluorescent vancomycin labeling were performed as described. M. tuberculosis was post-fixed with 4% formaldehyde overnight at RT prior to removing from the biosafety cabinet. For Bocillin labeling experiments, 500 μL of mid-log M. smegmatis was washed once in PBST and resuspended in PBST containing 5 μg/mL clavulanate. Bacteria were then pre-incubated or not with 50 μg/mL ampicillin or d-cycloserine at RT with gentle agitation. After 15 min, 50 μg/mL Bocillin FL (Thermo Fisher) was added and cultures were incubated for an additional 30 min. They were then washed three times in PBST and imaged live on agar pads.

29The Δalr M. smegmatis strain was generated using standard recombineering methods. 500 bp up- and downstream of alr were cloned on either side of the hygR cassette flanked with loxP sites. After induction of recET, transformation and subsequent selection on hygromycin and 1 mM d-alanine, PCR was used to confirm the presence of the correct insert. Strains were then cured of the hygR cassette by transformation with an episomal plasmid carrying the Cre recombinase and a sucrose negative selection marker. Strains were cured of this plasmid by repeated passaging in the presence of sucrose.

30M. smegmatis lacking ldtA, ldtB and/or ldtE were generously provided by Dr. Kasia Baranowski, Dr. Eric Rubin and Dr. Hesper Rego and are described in bioRxiv https://doi.org/10.1101/291823. Briefly, strains were constructed by recombineering to replace the endogenous copies with zeocin or hygromycin resistance cassettes flanked by loxP sites as previously described above. Once the knock-outs were verified by PCR, the antibiotic resistance cassettes were removed by the expression of Cre recombinase. To complement ΔldtABE, a copy of ldtA was under the constitutive TetO promoter on a kanamycin marked vector (CT94) that integrates at the L5 phage integration site of the chromosome.

31Fixed bacteria were imaged either by conventional fluorescence microscopy (Nikon Eclipse E600, Nikon Eclipse Ti or Zeiss Axioscope A1 with 100x objectives) or by structured illumination microscopy (Nikon SIM-E/A1R with SR Apo TIRF 100x objective).

32Images were processed using FIJI and cells were outlined and segmented using Oufti. Fluorescence signals of each cell were detected using Oufti and analyzed using custom-written MATLAB codes. The fluorescence intensities that we report here have been normalized by cell area. We distinguished septating from non-septating cells using the probe fluorescence profile along the long cell axis. We used the peakfinderprogram to identify peaks in the labeling profile. Because our probes label both the cell poles as well as the septum, septating cells were those that had three peaks in their labeling profile, with the middle peak positioned between 30% and 70% along the normalized long cell axis. Non-septating cells were identified as having only two peaks.

33To detect endogenous lipid-linked peptidoglycan precursors, we adopted the assay developed in with some modifications. M. smegmatis was inoculated in 100 mL of 7H9 medium and grown to mid-log phase at 37°C. Where applicable, MurJ was depleted by 8 hr of anhydrotetracycline-induced protein degradation as described. The bacteria were then divided into 25 mL cultures that were subjected or not to freshly prepared 80 µg/mL vancomycin and/or 10 µg/mL d-cycloserine. After 1 hr of incubation at 37°C, bacteria were collected by centrifugation and cell pellets were normalized by wet weight. 200–300 mg wet pellet was resuspended in 500 µL 1% glacial acetic acid in water. 500 µL of the resuspended pellet mixture was transferred into a vial containing 500 µL of chloroform and 1 mL of methanol and kept at room temperature (RT) for 1–2 hr with occasional vortexing. The mixture was then centrifuged at 21,000x g for 10 min at RT and the supernatant was transferred into a vial containing 500 µL of 1% glacial acetic acid in water and 500 µL chloroform, and vortexed for 1 min. After centrifugation at 900x g for 2 min at RT, three phases were distinguishable: aqueous, an interface, and organic. We collected the lipids from the organic phase and from the interface and concentrated it under nitrogen. Organic extracts were resuspended in 12 µL of DMSO and then incubated with purified S. aureus PBP4 and biotin-d-lysine (BDL; deprotected from Fmoc-D-Lys(biotinyl)-OH; Chem-Impex International) as described. Upon completion of the BDL exchange reaction, 10 µL of 2X loading buffer was added to the vials. Contents were boiled at 95°C for 5 min then run on an 18% SDS polyacrylamide gel. Biotinylated species were transferred to a PVDF membrane, blotted with streptavidin-HRP (diluted 1:10,000, Thermo-Fisher) and visualized in an ImageQuant system (GE Healthcare).

34To detect lipid-linked precursors that had been metabolically labeled with alkDADA or azDADA, growth of the Δalr strain was initially supported by the inclusion of 2 mM d-alanine-d-alanine (Sigma-Aldrich) in the 7H9 medium. Δalr bacteria were harvested and washed twice in sterile PBST prior to resuspension in 100 mL of pre-warmed medium. Both wildtype and Δalr M. smegmatis were incubated in 0.5–1 mM of alkDADA or azDADA for 1 hr then harvested as described above. Organic extracts from metabolically labeled cultures were subjected to CuAAC reaction by adding, in order and in a non-stick vial: 2 µL of PBST, 1 µL of 5 mM CuSO4, 1 µL of 20 mM BTTP, 1 µL of freshly-prepared 50 mM sodium ascorbate, 3 µL of 10 mM picolyl azide biotin or alkyne biotin (Click Chemistry Tools), and 2 µL of the organic extract. The reaction was incubated for 1 hr at RT with gentle shaking prior to detecting as above.