Introduction

The peptidoglycan (PG) cell wall, a network of glycan strands and peptide crosslinks, is a hallmark structure of bacteria1. PG encases the cytoplasmic membrane, defines cell geometry, and protects the cell against lysis due to its high internal osmotic pressure. As PG is essential for the survival of most bacteria, disrupting PG has become an effective strategy for antibiotic therapies. PG assembly requires multiple synthases, including glycosyltransferases (GTases) that polymerize glycan chains and transpeptidases (TPases) that form peptide crosslinks2. During the vegetative growth of most rod-shaped bacteria, two major enzymatic systems carry out the elongation of PG sacculi, the Rod system and class-A penicillin-binding proteins (aPBPs). The Rod system determines rod shape and accounts for most of the PG synthesis during cell growth3. The core components of the Rod system include RodA, a SEDS-family GTase, PBP2, a TPase of the class B penicillin-binding proteins (bPBPs), and MreB, a bacterial actin homolog that orchestrates the actions of RodA and PBP24,5,6,7. In contrast, aPBPs possess both GTase and TPase activities and their functions do not depend on the Rod system4,8,9,10. While aPBPs assemble PG in different environments, repair PG defects, and regulate cell diameter3,11,12, their absence rarely abolishes cell survival or rod-like morphology3,13. However, the inhibitors of aPBPs usually trigger rapid collapse of rod shapes14,15.

The insertion of new PG strands requires local hydrolysis of the existing PG network2,16,17,18. PG hydrolases, including lytic transglycosylases, amidases, and endo/carboxypeptidases, are difficult to study in most model organisms. On the one hand, these enzymes are highly redundant in most bacteria, where strains lacking single, or several hydrolases usually do not show significant growth or morphological defects. On the other hand, as uncontrolled hydrolytic activities could compromise PG integrity and cause cell lysis, it is difficult to observe highly activated PG hydrolysis during normal growth16,19. To maintain the integrity of the cell wall, PG hydrolases must coordinate with synthases in time and space16,18,19,20. However, even less is known about the mechanisms of their functional coordination.

Myxococcus xanthus is a rod-shaped, Gram-negative bacterium that possesses both the Rod system and aPBPs. In response to chemical signals, such as glycerol, vegetative cells of M. xanthus thoroughly degrade their PG and transform into spherical spores within a few hours21,22,23. When spores germinate, emerging cells rebuild rod-shaped PG without preexisting templates23. Such morphological transitions make M. xanthus an ideal organism to study PG degradation and assembly24. The Rod system is essential for the de novo establishment of rod shape during M. xanthus spore germination23,25. In contrast, spherical spores can still germinate into rods in the presence of cefsulodin and cefmetazole, the inhibitors of aPBPs. However, these newly established rods retrogress to spheres before cell division23. This observation suggests that aPBPs are important for the maintenance of rod-like morphology albeit the mechanism remains unknown.

In this work, we identified DacB, a PBP4-family D-Ala-D-Ala endo/carboxypeptidase, as a major PG hydrolase that collapses the rod shape of M. xanthus by degrading PG, especially at cell poles. We found that whereas either the inhibition or absence of aPBPs is sufficient to enrich DacB to cell poles, only the inhibition, but not absence, of aPBPs triggers the loss of rod shape in vegetative cells. We demonstrate that inhibited PBP1a2, an aPBP, promotes the binding between DacB and PG, and thus reduces the diffusion of DacB. Conversely, DacB also regulates the dynamics and distribution of aPBPs. Our results elucidate the mutual regulation between PG synthases and hydrolases, which plays central roles in the maintenance of cell integrity.

Results

The inhibition of aPBPs triggers rapid collapse of rod shape

To study how aPBPs and the Rod system support rod shape, we treated vegetative M. xanthus cells using moenomycin (8 µg/ml) and mecillinam (100 µg/ml). Whereas moenomycin is a phosphoglycolipid that specifically inhibits the GTase activity of aPBPs but does not affect RodA, the GTase of the Rod system10, mecillinam is a ß-lactam that specifically inhibits the Rod system by blocking the Tpase activity of PBP226,27. After 2 h, 72.7% (n = 967 cells) of moenomycin-treated cells became spherical. Agents that inhibit PG synthesis or disrupt PG were suspected to induce sporulation of M. xanthus28. To distinguish whether the spherical cells that resulted from moenomycin treatment are spores or vegetative cells that had lost rod shape, we treated these cells with sonication. Compared to the glycerol-induced spores that 91.7% (n = 2515 cells) are resistant to sonication23, the moenomycin-induced spheres are collapsed vegetative cells because only 5.3% (n = 1034 cells) of them remained after sonication. Although prolonged moenomycin treatment turned 100% cells (n = 1356 cells) into spheres, we did not observe mass cell lysis. Instead, cell concentration continued to increase during the treatment (Fig. 1a, b). Different from L-forms that usually generate new cells through irregular budding29, the moenomycin-induced spheres produced daughter cells of the same size (Fig. 1a), indicating that these cells proliferate through binary fission. Currently, we are investigating how M. xanthus cells divide as spheres. In contrast to moenomycin, 2-h of mecillinam treatment only caused minor bulging near the centers of cells and none of the treated cells (n = 988 cells) lost rod shape (Fig. 1a). After prolonged incubation, these bulges became more pronounced while the cells still maintained rod-shaped near their poles (Fig. 1a). It is possible that aPBPs could substitute PBP2 in the presence of mecillinam, especially near cell poles. Nevertheless, our data indicate that the inhibition of aPBPs by moenomycin is sufficient to trigger rapid collapse of rod shape in M. xanthus.

Fig. 1: The inhibition of aPBPs abolishes rod shape of M. xanthus.
figure 1

a Moenomycin (MOE, 8 µg/ml) that inhibits aPBPs induces rapid loss of rod shape of M. xanthus, whereas mecillinam (MEC, 100 µg/ml), an inhibitor of the Rod system, does not. White arrows point to the bulges induced by MEC. Scale bars, 5 µm. b Because the cell concentrations continue to increase during treatment, neither MOE nor MEC induce mass cell lysis. Cells were inoculated at OD600 0.5 and growth measured for 24 h from six biological replicates. Data are presented as mean values ± SD. Source data are provided as a Source Data file.

Full size image

Cells maintain rod shape in the absence of all aPBPs

As moenomycin inhibits aPBPs specifically, we investigated whether the cells that lack certain aPBPs still maintain rod shape. M. xanthus encodes three aPBPs, including two homologs of PBP1a (PBP1a1, ORF K1515_1180530 and PBP1a2, ORF K1515_08245) and one putative PBP1c (ORF K1515_25460). The strains that lack one, two, or all three aPBPs were all viable and only the absence of PBP1a1 caused moderate slow-down of growth in early exponential phase (Fig. 2a). All seven mutants maintained rod shape. Nevertheless, compared to wild-type, mutant cells were moderately, but significantly shorter and wider (Fig. 2b, c). Thus, while being dispensable for both growth and rod-like morphology, all aPBPs regulate cell dimensions. Compared to wild-type cells, the strains lacking aPBPs, especially PBP1a1 and PBP1a2, showed significantly enhanced resistance against moenomycin, which is evidenced by higher numbers of cells that remained rods after a 2-h treatment by 8 µg/ml moenomycin (Fig. 2d). However, even in the absence of all three aPBPs, 15.3% (n = 1217 cells) of cells still lost rod shape (Fig. 2d). This result suggests minor and nonspecific toxicity of moenomycin on M. xanthus. Nevertheless, our data indicate that it is the inhibition, but not absence of aPBPs that triggers the collapse of rod shape.

Fig. 2: Cells maintain rod shape in the absence of all aPBPs.
figure 2

a Cells lacking one, two, or all three (∆3) aPBPs are all viable and only the absence of PBP1a1 causes moderate slow-down of growth. Cells were inoculated at OD600 0.1 and growth measured for 24 h from six biological replicates. Data are presented as mean values ± SD. b–d While dispensable for the establishment of rod shape, all three aPBPs regulate the length (b) and width (c) of cells. Compared to the wild-type cells, the absence of aPBPs but not the expression of an enzymatic inactive PBP1a2 increased the numbers of rod-shaped cells after MOE treatment (8 µg/ml, 2 h) (d). p values were calculated using a one-way ANOVA test between two unweighted, independent samples. Whiskers indicate the 25th–75th percentiles and red dots the median. The total number of cells analyzed is shown on top of each plot. Source data are provided as a Source Data file.

Full size image

Moenomycin promotes PG hydrolysis by DacB

Blocked aPBP activities by moenomycin could cause the loss of rod shape. However, this hypothesis cannot explain the fact that cells completely lacking aPBPs are still rods. Alternatively, the inhibition of aPBPs could activate or alter the distribution of certain PG hydrolases, which results in mis-regulated PG degradation. In a previous attempt to label PG using a fluorescent D-amino acid, TAMRA 3-amino-D-alanine (TADA)23,31, we found that deleting dacB, a gene encoding a PBP4-family D-Ala-D-Ala endo/carboxypeptidase, improved the incorporation of TADA significantly23 (Fig. 3a, b). Thus, consistent with its peptidase activity, DacB actively removes TADA from the peptide stem of PG.

Fig. 3: Moenomycin promotes PG hydrolysis by DacB, especially at cell poles.
figure 3

a The absence of DacB enhances the efficiency of TADA incorporation. b Average fluorescence intensities of TADA in vegetative wild-type and ∆dacB cells. p values were calculated using a one-way ANOVA test between two unweighted, independent samples. Whiskers indicate the 25th–75th percentiles and red dots the median. The total number of cells analyzed is shown on top of each plot. Source data are provided as a Source Data file. c Moenomycin induces near-even loss of TADA signal in the wild-type cells while the ∆dacB cells retain TADA at their poles. Scale bars, 5 µm.

Full size image

To visualize PG degradation by DacB, we labeled the entire PG layer of cells by inducing the germination of M. xanthus spores in the presence of TADA. Both the newly generated wild-type and ΔdacB cells showed homogeneous incorporation of TADA (Fig. 3a). After 1 h of moenomycin treatment, the labeled wild-type cells lost fluorescence near evenly throughout the PG layer, whereas bright loci of TADA remained at the poles in ΔdacB cells (Fig. 3c). Taken together, moenomycin activates the hydrolytic activity of DacB, especially at cell poles.

Eliminating the peptidase activity of DacB, by either deleting the dacB gene or replacing its conserved catalytic serine residue32 with alanine (DacBS75A, Fig. S1a), resulted in moderate, but statistically significant, increase in cell width and decrease in cell length (Fig. 4a, b), similar to the phenotypes caused by the absence of aPBPs (Fig. 2b–d). These observations suggest that DacB contributes to the maintenance of rod shapes. To test if DacB participates in the degradation of rod shape in moenomycin-treated cells, we grew cells in liquid medium to OD600 1.0 and tested their response to moenomycin (8 µg/ml). After a 2-h incubation, compared to the wild-type cells only 27.8% remained rods, 61.3% of the ΔdacB cells (n = 1008 cells) and 57.5 (n = 1095 cells) dacBS75A cells retained rod shapes, indicating significantly higher resistance (Fig. 4c). We then expressed photo-activatable-mCherry (PAmCherry)-labeled DacB ectopically using a vanillate-inducible promoter33 in addition to the endogenous dacB gene. Induced by 200 µM vanillate, the cells overexpressing DacB were hyper-sensitive to moenomycin (14.1% rods, n = 1046 cells), whereas the ones harboring the empty vector showed sensitivity similar to the wild-type (25.2% rods, n = 1172 cells) (Fig. 4c, S1). Interestingly, the overexpression of DacB did not alter cell dimensions significantly (Fig. 4a–c).

Fig. 4: DacB regulates cell morphology and sensitivity against moenomycin.
figure 4

a, b DacB moderately regulates cell morphology. p values were calculated using a one-way ANOVA test between unweighted, independent samples. Whiskers indicate the 25th−75th percentiles and red dots the median. c DacB significantly regulates cell sensitivity against moenomycin. The deletion and overexpression of dacB increases and decreases the numbers of rod-shaped cells, respectively, after moenomycin treatment (8 µg/ml, 2 h). The PAmCherry label and vanillate, the inducer for dacB overexpression, does not affect moenomycin sensitivity. The total number of cells analyzed is shown on top of each plot. Source data are provided as a Source Data file.

Full size image

Moenomycin alters the spatial distribution of DacB

To investigate the effects of moenomycin on DacB, we fused a DNA sequence encoding PAmCherry to the endogenous dacB gene in wild-type M. xanthus. The resulted strain expressed DacB-PAmCherry as a full-length fusion protein that did not show detectable degradation (Fig. S1). The introduction of PAmCherry did not significantly affect cell morphology or the strain’s sensitivity to moenomycin, indicating that the tagged DacB protein is functional (Fig. 4). We used a 405-nm excitation laser to activate the fluorescence of a few labeled DacB particles randomly in each cell and quantified their localization using a 561-nm laser. Along the long axis of cells, we loosely defined a region within three pixels (480 nm) from each end of cell as a “pole” and the rest of the cell as the “nonpolar region”. By this definition, poles account for ~20% of the cell surface in a typical M. xanthus cell (~5 µm long, 1 µm wide). We found that in untreated cells, 28.0% (n = 1955 particles) of DacB particles were detected at cell poles, suggesting that DacB distributes near evenly in the membrane. Consistent with the degradation pattern of PG (Fig. 3c), DacB aggregated at poles (47.6%, n = 938 particles) in the presence of moenomycin (Fig. 5a, S2).

Fig. 5: The binding to PG decreases DacB mobility.
figure 5

Moenomycin (MOE) and glycerol-induced sporulation trigger PG degradation by DacB, especially at cell poles. a Both MOE and glycerol enrich DacB particles to cell poles. b, c MOE and glycerol reduce the mobility of DacB particles. In contrast, mecillinam (MEC) does not affect the distribution or dynamics of DacB (a–c). d Whereas the purified periplasmic domains of DacB and DacBS75A both co-precipitate with PG, DacBS75A binds to PG with higher affinity. This experiment was repeated independently three times with similar results. Uncropped gel images are provided in the Source Data file. Thus, compared to DacB-PAmCherry, DacBS75A-PAmCherry particles display lower mobility (b, c). e, f Moenomycin increases the population of immobile DacB (e) and reduces the diffusion coefficient of mobile DacB (f) throughout the whole cell. The number of particles analyzed is shown on top of each bar. Mean and standard deviation were calculated through 1000 bootstrap samples of single-particle tracks. * Significant. NS non-significant. Mean values with a difference >0.005 are considered significantly different. Source data are provided as a Source Data file.

Full size image

Moenomycin enhances PG-binding of DacB

To understand how moenomycin promotes PG degradation by DacB, it is essential to monitor its local peptidase activity in vivo, which is technically difficult. To circumvent this hurdle, we attempted to use the binding between DacB and PG to proximate instantaneous DacB activity in live cells. Compared to individual synthases and hydrolases, the existing PG meshwork is a large and relatively stationary substrate. In Escherichia coli, individual particles of aPBPs distribute between immobile and mobile populations and their inhibitors increase their mobility9. Similarly, we expect that the mobility of DacB will decrease when it binds to PG. We imaged single DacB-PAmCherry particles at 10 Hz using single-particle tracking photo-activated localization microscopy (sptPALM) under the highly inclined and laminated optical sheet (HILO) illumination25,34,35. Using this setting, only a thin section of each cell surface that was close to the coverslip was illuminated. To analyze the data, we only chose the fluorescent particles that remained in focus for 4–12 frames (0.4–1.2 s). As free PAmCherry particles diffuse extremely fast in the cytoplasm, entering and exiting the focal plane frequently, they usually appear as blurry objects that cannot be followed at 10-Hz close to the membrane25. For this reason, the noise from any potential degradation of DacB-PAmCherry was negligible. Using a 405-nm laser (0.3 kW/cm2, 0.1 s) to activate PAmCherry and a 561-nm laser for imaging, we detected 1476 DacB-PAmCherry particles from 217 cells. In contrast, we only detected 27 particles from 369 cells without activation and 25 particles from 449 unlabeled cells using the same setting. Thus, the noises from the background, and autoblinking, the spontaneous switch-on of PAmCherry, were also negligible. DacB particles displayed two dynamic patterns, immobile and diffusion. The immobile particles remained within a single pixel (160 nm × 160 nm) before photo-bleach and the mobile ones displayed typical diffusive behavior (Fig. S2). In wild-type, untreated vegetative cells, 13.1% (n = 1955 particles) of DacB particles were immobile and the diffusion coefficient of the mobile population is 2.72 × 10−2 ± 3.7 × 10−3 μm2/s (n = 1698 particles) (Fig. 5b, c).

Sporulation provides a scenario where PG hydrolysis activities are upregulated. During glycerol-induced sporulation when cells thoroughly degrade PG, the expression of the dacB gene is not regulated36. To test if DacB degrades PG during sporulation, we used the length/width ratio (L/W) of cells to quantify the sporulation progress and found that the ∆dacB cells did not degrade PG efficiently at poles and thus formed spores slowly. By contrast, the overexpression of DacB-PAmCherry significantly accelerated the rod-to-sphere transition (Fig. S3). Thus, compared to vegetative growth, DacB is activated during glycerol-induced sporulation, especially at cell poles. Consistently, DacB concentrated at poles (59.4%, n = 2290 particles) in the presence of glycerol, its immobile population increased to 25.6% (n = 2290 particles) and the diffusion coefficient of the mobile population decreased to 2.2 × 10−2 ± 4.0 × 10−3 μm2/s (n = 1704 particles) (Fig. 5a–c).

To further confirm the connection between DacB dynamics and its binding to PG, we labeled DacBS75A with PAmCherry and expressed it as the sole source of DacB using the native dacB locus and promoter. DacBS75A-PAmCherry accumulated as a full-length protein which did not show detectable degradation (Fig. S1). We expected that the mutation on its active site would prevent DacBS75A-PAmCherry from binding to PG and thus increase its mobility. Surprisingly, DacBS75A-PAmCherry particles displayed reduced mobility, where 16.0% (n = 2046 particles) of particles were immobile and the diffusion coefficient of the mobile population was 2.1 × 10−2 ± 4.9 × 10−3 μm2/s (n = 1719 particles). These data strongly suggested that despite to the loss of peptidase activity, DacBS75A could bind PG with enhanced affinity. To test this hypothesis, we expressed the periplasmic domains (amino acids 21 – 494) of DacB and DacBS75A in E. coli and purified the recombinant proteins. After a 1-h incubation with purified M. xanthus PG and a brief centrifugation, both proteins precipitated with PG. Compared to the wild-type protein, a significantly bigger population of DacBS75A enriched in the pellet, indicating increased binding to PG (Fig. 5d). The mechanism for this enhanced binding remains to be elucidated. Nevertheless, as multiple residues contribute to PG-binding in DacB32, other residues could overcompensate the substitution of the active serine. Taken together, increased immobile population and decreased diffusion coefficients of DacB particles strongly correlates with their enhanced binding to PG.

Compared to untreated cells, during moenomycin-induced PG degradation, the immobile population of DacB increased to 22.6% (n = 938 particles) and the diffusion coefficient of the mobile population decreased to 1.9 × 10−2 ± 4.9 × 10−3 μm2/s (n = 722 particles) (Fig. 5b, c). Thus, moenomycin enhances PG-binding of DacB. In contrast, mecillinam that does not trigger immediate PG degradation, did not show significant effects on either the dynamics or distribution of DacB (Fig. 5a–c).

To determine whether moenomycin promotes PG-binding by DacB throughout the whole cell or specifically at cell poles, we studied the dynamics of DacB in nonpolar regions. Like its effects on whole cells, moenomycin increased the immobile population to 27.2% (n = 484 particles) in nonpolar regions and decreased the diffusion coefficient of the mobile population to 1.7 × 10−2 ± 4.0 × 10−3 μm2/s (n = 352 particles). Thus, moenomycin promotes PG-binding by DacB throughout the whole cell (Fig. 5e, f).

Moenomycin activates DacB through PBP1a2

We hypothesized that moenomycin could activate DacB through aPBPs. To test this hypothesis, we first investigated the cellular distribution of DacB-PAmCherry particles in the strains that lack aPBPs. Like the effect of moenomycin, the absence of one aPBP is sufficient to enrich DacB to cell poles (Fig. 6a). However, none of these mutant lost rod shape (Fig. 2). Thus, although all three aPBPs regulate the distribution of DacB, the polar enrichment of DacB is not sufficient to dismantle rod shape.

Fig. 6: Moenomycin activates PG hydrolysis by DacB through PBP1a2.
figure 6

a All three aPBPs and their activities are required for the even distribution of DacB. Their absence and inhibition by MOE both enrich DacB to cell poles. b, c MOE does not reduce DacB mobility in the absence of PBP1a2 but does so in the presence of PBP1a2E97A. The number of particles analyzed is shown on top of each bar. Mean and standard deviation were calculated through 1000 bootstrap samples of single-particle tracks. * Significant. NS non-significant. Mean values with a difference >0.005 are considered significantly different. Source data are provided as a Source Data file.

Full size image

We then investigated if aPBPs regulate PG-binding by DacB. In untreated vegetative cells, the absence of each aPBP did not alter the mobility of DacB (Fig. 6b, c, S4). Thus, uninhibited aPBPs do not regulate PG-binding by DacB. In response to moenomycin, DacB mobility reduced in Δpbp1a1 and Δpbp1c cells. Strikingly, moenomycin failed to reduce DacB mobility in the absence of PBP1a2 (Fig. 6b, c). Taken together, moenomycin-inhibited PBP1a2 promotes PG-binding by DacB.

Is the inactivation of PBP1a2 sufficient to activate PG hydrolysis by DacB? To answer this question, we replaced the conserved catalytic glutamate residue26 of PBP1a2 with alanine using site-directed mutagenesis and expressed the mutant protein PBP1a2E97A as the sole source of PBP1a2 using the native pbp1a2 locus and promoter. The alanine substitution did not affect the expression or stability of PBP1a2E97A (Fig. S5). Cells expressing PBP1a2E97A and the ones lacking PBP1a2 displayed indistinguishable morphological defects, including decreased cell length and increased width. However, the loss of active site did not enhance cell resistance against moenomycin, suggesting that the mutant PBP1a2 still binds to moenomycin (Fig. 2). Unlike the Δpbp1a2 cells, the untreated cells expressing PBP1a2E97A did not enrich DacB at cell poles (Fig. 6a) or reduce DacB mobility (Fig. 6b, c), indicating that the binding between moenomycin and PBP1a2 is required to promote PG-binding by DacB (see discussion). In contrast, moenomycin reduced DacB mobility through PBP1a2E97A (Fig. 6b, c).

DacB regulates the distribution and dynamics of aPBPs

The coordination between PBP1a2 and DacB suggests that DacB could also regulate the activity and distribution of aPBPs. To test this hypothesis, we expressed PAmCherry-labeled aPBPs using their endogenous loci and promoters. All three fusions accumulated as stable proteins in M. xanthus and unlike the aPBP deletion mutants, cells expressing these fusion proteins maintained wild-type cell morphology and sensitivity to moenomycin (Fig. S6a, b), indicating that PAmCherry-labeled aPBPs are functional. Deletion of the dacB gene affected the distribution of all three aPBPs (Fig. 7a). Like DacB, aPBPs displayed two dynamic patterns, immobile and diffusion (Fig. 7b, S6c). We compared the effects of dacB deletion and moenomycin on these aPBPs. As the absence of DacB and the presence of moenomycin showed similar effects on the distribution and diffusion of PBP1a1 and PBP1c, DacB could activate both aPBPs. (Fig. 7). Because the deletion of dacB promoted PBP1a2 diffusion, (Fig. 7b, c), DacB could also activate PBP1a2 by facilitating its binding to PG. Taken together, our data suggest that DacB could activate all three aPBPs.

Fig. 7: DacB and moenomycin regulate the dynamics and distribution of aPBPs.
figure 7

The effects of the absence of DacB and the presence of moenomycin on the distribution (a) and dynamics (b, c) of individual aPBPs. The number of particles analyzed is shown on top of each bar. Mean and standard deviation were calculated through 1000 bootstrap samples of single-particle tracks. * Significant. NS, non-significant. Mean values with a difference greater than 0.005 are considered significantly different. Source data are provided as a Source Data file.

Full size image

Discussion

It has long been speculated that PG hydrolases associate with synthases in many functions that maintain the integrity of PG16,19,20. Although most of the PG synthases and hydrolases have been identified, their functional coordination is less understood. In this report, we dissected the functional coordination between PBP1a2, a bifunctional PG synthase, and DacB, a PG hydrolase. We found that when PBP1a2 binds moenomycin, it promotes PG-binding by DacB and thus activates PG hydrolysis. Why have bacteria developed such a seemingly self-destructive mechanism, and why moenomycin-inhibited PBP1a2 activates DacB but the catalytic inactive PBP1a2E97A does not? We speculate that moenomycin mimics a growing PG strand and thus locks PBP1a2 into a catalytic active conformation26. In contrast, PBP1a2E97A is in the catalytic inactive conformation. Under physiological conditions, catalytic active PBP1a2 could recruit DacB to the PG assembly sites, where DacB breaks existing peptide crosslinks and thus facilitates the insertion of newly synthesized murein strands. The fact that moenomycin-bound PBP1a2 recruits DacB to PG, rather than forming a diffusive PBP1a2-DacB complex, suggests that no matter whether these two proteins interact directly, their coordination is mediated by PG.

A major hurdle in studying the coordination between aPBPs and PG hydrolases is that both categories of enzymes are redundant, and many single mutations do not cause clear phenotypes. Furthermore, PG synthases and hydrolases may not form stable complexes, especially in the absence of PG, that are readily detected by biochemical methods. As inhibited PBP1a2 activates the hydrolytic activities of DacB and cells overexpressing DacB are more sensitive to moenomycin than the wild-type ones, our results support the hypothesis that PBP1a2 and DacB may only interact transiently in M. xanthus and that one PBP1a2 may activate multiple DacB molecules. In addition, we found that cells lacking aPBPs and DacB are viable, whereas the changes in their coordination cause cell lysis. For these reasons, genetic and biochemical approaches alone might not be sufficient for studying the complex coordination among these enzymes. This research has provided a method to study such coordination before the identification of the coupling mechanisms. If the localization and dynamics of one protein respond to the inhibitor or absence of another, it is sufficient to conclude that the functions of these proteins coordinate.

Recent research on rod-like morphology has emphasized the roles of the cylindrical section where the Rod system plays major roles in the assembly of PG2,17,37. In contrast, the contribution of cell poles, where PG remains metabolically inert after division38, is less appreciated. As moenomycin induces PG hydrolysis at cell poles, we propose that poles may lose inertness under certain stresses. Particularly, our findings provide an explanation for the lytic effects of aPBP inhibitors. As the Rod system is generally excluded from cell poles39, when aPBPs are inhibited, poles become void of PG synthases. The lack of PG repair in combination with the activation of hydrolysis, causes rapid PG degradation at poles, which results in sudden loss of rod shape. In contrast, the cylindrical surfaces of cells are maintained by both the Rod system and aPBPs. Thus, when one system is inhibited, the other is still able to repair PG damages3.

Despite exhaustive data on how antibiotics exert action on their targets, we know very little about how these drugs actually kill bacteria40. For instance, ß-lactams specifically inhibit the TPase activity of PBPs. However, rather than the blockage of PG assembly, the actual lethal effect of ß-lactams is inducing the degradation of un-crosslinked glycan strands, which are produced by the uninhibited GTases41. Similarly, blocking the GTase activities of aPBPs is not likely the direct bactericidal effect of moenomycin. Like M. xanthus, although wild-type Bacillus subtilis is sensitive to moenomycin, a strain lacking all four aPBPs is still viable and produces PG with only minor defects4,13. In this study, we provide evidence that moenomycin inhibits PBP1a2, which in turn, regulates the activity and distribution of DacB and eventually causes cell lysis. We propose that the bactericidal action of moenomycin is inducing an imbalance between aPBPs and PG hydrolases. Our results suggest that disrupting the coordination between PG synthases and hydrolases could be an effective strategy for antibacterial therapies.

Methods

Strains and growth conditions

Vegetative M. xanthus cells were grown in liquid CYE medium (10 mM MOPS pH 7.6, 1% (w/v) Bacto™ casitone (BD Biosciences), 0.5% yeast extract, and 8 mM MgSO4) at 32 °C, in 125-ml flasks with vigorous shaking, or on CYE plates that contains 1.5% agar. All genetic modifications on M. xanthus were made on chromosome. Deletion and insertion mutants were constructed by electroporating M. xanthus cells with 4 μg of plasmid DNA. Transformed cells were plated on CYE plates supplemented with 100 µg/ml sodium kanamycin sulfate or 10 µg/ml tetracycline hydrochloride. DacB, PBPs, and their variants were labeled with PAmCherry at their C-termini by fusing their genes to a DNA sequence that encodes PAmCherry through a AAGGAGTCCGGCTCCGTGTCCTCCGAGCAGCTGGCCCAGTTCCGCTCCCTGGAC (KESGSVSSEQLAQFRSLD) linker. All constructs were confirmed by PCR and DNA sequencing. The expression and stability of PAmCherry-labeled proteins were determined by immunoblotting using an anti-mCherry antibody (Rockland Immunochemicals, Inc., Lot 46705) and a Goat anti-Rabbit IgG (H+L) Secondary Antibody, HRP (Thermo Fisher Scientific, catalog # 31460). The blots were developed with Pierce™ ECL Western Blotting Substrate (Thermo Fisher Scientific REF 32109) and MINI-MED 90 processor (AFP Manufacturing). M. xanthus strains used in this study are listed in Table S1. In all imaging experiments using fluorescence microscopy, moenomycin and mecillinam were added at 4 µg/ml and 100 µg/ml, respectively.

Sporulation and TADA-labeling

To induce sporulation, glycerol was added to 1 M when liquid cell culture reaches OD600 0.1–0.2. After rigorous shaking overnight at 32 °C, the remaining vegetative cells were eliminated by sonication, and spores were purified by centrifugation (1 min, 15,000 × g and 4 °C). The pellet was washed three times with water. To prepare TADA-labeled vegetative cells, purified spores were diluted to OD600 0.5 into 1 ml of liquid germination CYE (CYE medium supplemented with 150 µM TADA, additional 0.2% casitone and 1 mM CaCl2) and incubated in an 18-mm test tube at 32 °C, with vigorous shaking.

Protein expression and purification

DNA sequences encoding amino acids 21–494 of DacB and DacBS75A were amplified by polymerase chain reaction (PCR) and inserted into the pET28a vector (Novogen) between the restriction sites of EcoRI and HindIII. The resulted plasmids were transformed into E. coli strain BL21(DE3). Transformed cells were cultured in 20 ml LB (Luria-Bertani) broth at 37 °C overnight and used to inoculate 1 L medium which contains 1.0% (w/v) Bacto Tryptone, 0.5% Bacto yeast extract (BD Biosciences), 0.5% NaCl and 1.0% glucose. Protein expression was induced by 0.1 mM IPTG (isopropyl-h-d-thiogalactopyranoside) when the culture reached an OD600 of 0.8. Cultivation was continued at 16 °C for 10 h before the cells were harvested by centrifugation at 6000 × g for 20 min. Proteins were purified using a NGC™ Chromatography System (BIO-RAD) and a 5-ml HisTrap™ column (Cytiva)42,43. Purified proteins were concentrated using Amicon™ Ultra centrifugal filter units (Millipore Sigma) with a molecular cutoff of 10 kDa and stored at −80 °C.

PG purification and DacB binding assay

M. xanthus cells were grown until mid-stationary phase and harvested by centrifugation (20 min, 6000 × g). Supernatant was discarded and the pellet was resuspended and boiled in 1x PBS with 5% SDS for 2 h. SDS was removed by repetitive wash with MilliQ water and centrifugation (21,000 × g, 10 min, 20 °C). Purified PG from 100 ml culture was suspended into 1 ml 1× PBS and stored at −20 °C44. For the binding assay, 10 μl of purified PG was mixed with 10 μl of purified protein (4 mg/ml) and incubated at 25 °C for 1 h. 5 μl of the mixture was taken out and used as “total protein”. The remaining 15-ul mixture was subjected to centrifugation (21,000 × g, 10 min, 25 °C), and the supernatant was removed. The pallet was resuspended into 15 μl 1x PBS. The total protein, pellet, and supernatant samples were mixed with equal volumes of 2X loading buffer and 5 μl of each mixture was applied to SDS PAGE. Protein bands were stained using Coomassie brilliant blue.

Microscopy analysis

For all imaging experiments, 5 μl cells were spotted on an agarose pad. For the treatments with inhibitors, inhibitors were added into both the cell suspension and agarose pads. The length and width of cells were determined from differential interference contrast (DIC) images using a MATLAB (MathWorks) script25,45. DIC and fluorescence images of cells were captured using a Hamamatsu ImagEM X2TM EM-CCD camera C9100-23B (effective pixel size 160 nm) on an inverted Nikon Eclipse-TiTM microscope with a 100x 1.49 NA TIRF objective. For sptPALM, M. xanthus cells were grown in CYE to 4 × 108 cfu/ml and PAmCherry was activated using a 405-nm laser (0.3 kW/cm2), excited and imaged using a 561-nm laser (0.2 kW/cm2). Images were acquired at 10 Hz. For each sptPALM experiment, single PAmCherry particles were localized in at least 100 individual cells from three biological replicates. sptPALM data were analyzed using a MATLAB (MathWorks) script25,45. Briefly, cells were identified using DIC images. Single PAmCherry particles inside cells were fit by a symmetric 2D Gaussian function, whose center was assumed to be the particle’s position25. Particles in consecutive frames were considered to belong to the same trajectory when they were within a user-defined distance of 320 nm (two pixels). Particles that explored areas smaller than 160 nm × 160 nm (within one pixel) in 0.4-1.2 s were considered as immobile. For the mobile particles, mean square displacements (MSDs) were calculated. Time-averaged MSD (TAMSD) of each mobile particle was calculated according to the standard method and fit to log (TAMSD) = log (4D) + α\(\,\cdot\) log (Δt), where D is the diffusion coefficient and Δt is time lapse46. Based on a previous simulation, diffusive particles on M. xanthus cell surface feature α values between 0 and 1.525. In untreated cells, most PAmCherry-labeled particles belonged to this category (93.4% for DacB-PAmCherry, n = 1995 particles; 89.7% for DacBS75A-PAmCherry, n = 2046 particles; 92.7% for PBP1a1-PAmCherry, n = 1223 particles; 94.6% for PBP1a2-PAmCherry, n = 5909 particles; and 95.1% for PBP1c-PAmCherry, n = 2667 particles). For simplicity, we considered all the mobile particles of these proteins diffused and determined their diffusion coefficient (D) from a linear fit to the first four points of the MSD using a simpler formula MSD = 4DΔt9,25,45. Mean and standard deviation were calculated through 1000 bootstrap samples of single-particle tracks. Mean values with a difference >0.005 are considered significantly different.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.