MurA escape mutations uncouple peptidoglycan biosynthesis from PrkA signaling

Gram-positive bacteria are protected by a thick mesh of peptidoglycan (PG) completely engulfing their cells. This PG network is the main component of the bacterial cell wall, it provides rigidity and acts as foundation for the attachment of other surface molecules. Biosynthesis of PG consumes a high amount of cellular resources and therefore requires careful adjustments to environmental conditions. An important switch in the control of PG biosynthesis of Listeria monocytogenes, a Gram-positive pathogen with a high infection fatality rate, is the serine/threonine protein kinase PrkA. A key substrate of this kinase is the small cytosolic protein ReoM. We have shown previously that ReoM phosphorylation regulates PG formation through control of MurA stability. MurA catalyzes the first step in PG biosynthesis and the current model suggests that phosphorylated ReoM prevents MurA degradation by the ClpCP protease. In contrast, conditions leading to ReoM dephosphorylation stimulate MurA degradation. How ReoM controls degradation of MurA and potential other substrates is not understood. Also, the individual contribution of the ~20 other known PrkA targets to PG biosynthesis regulation is unknown. We here present murA mutants which escape proteolytic degradation. The release of MurA from ClpCP-dependent proteolysis was able to activate PG biosynthesis and further enhanced the intrinsic cephalosporin resistance of L. monocytogenes. This latter effect required the RodA3/PBP B3 transglycosylase/transpeptidase pair. One murA escape mutation not only fully rescued an otherwise non-viable prkA mutant during growth in batch culture and inside macrophages but also overcompensated cephalosporin hypersensitivity. Our data collectively indicate that the main purpose of PrkA-mediated signaling in L. monocytogenes is control of MurA stability during standard laboratory growth conditions and intracellular growth in macrophages. These findings have important implications for the understanding of PG biosynthesis regulation and β-lactam resistance of L. monocytogenes and related Gram-positive bacteria.


Introduction
The cell wall of Gram-positive bacteria is built up from peptidoglycan (PG) strands that are crosslinked with each other to form a three-dimensional network called the sacculus. The sacculus engulfs the entire cell, determines cell shape and provides the key matrix for the attachment of other surface molecules such as teichoic acids, proteins and polymeric carbohydrates. PG biosynthesis is a target of many commonly used antibiotics that can lead to lysis of bacterial cells by weakening the integrity of the sacculus that then gives way to the high internal turgor pressure [1][2][3]. PG accounts for approximately 20% of the weight of a Gram-positive cell [4], illustrating the high demand PG biosynthesis imposes on biosynthetic and energy supplying pathways. Thus, bacteria need to tightly adjust PG production to changing environmental and growth conditions to prevent unnecessary losses of building blocks and energy.
The first step in PG biosynthesis is mediated by MurA, which initiates a series of eight cytoplasmic reactions that sequentially convert UDP-N-acetyl-glucosamine (UDP-GlcNAc) into a lipid-linked disaccharide carrying a pentapeptide side chain (lipid II). Lipid II is flipped to the other side of the membrane by Amj-or MurJ-like flippases [5][6][7], where the disaccharide unit is incorporated into the growing PG chain and the pentapeptides crosslinked either through bifunctional penicillin binding proteins (PBPs) providing transglycosylase and transpeptidase activity [8,9] or through FtsW/RodA-like transglycosylases that act in concert with monofunctional PBPs, which are mere transpeptidases [10][11][12][13].
Among the various regulatory mechanisms controlling PG production, PASTA (PBP and serine/threonine kinase associated) domain-containing eukaryotic-like serine/threonine protein kinases (PASTA-eSTKs) have an outstanding role in PG biosynthesis regulation of firmicutes and actinobacteria, two major types of Gram-positives [14][15][16]. PASTA-eSTKs sense cell wall damaging conditions by interaction of lipid II with their extracellular PASTA domains, and this leads to the activation of their cytosolic kinase domain [17,18]. In actinobacteria, many of the kinase substrates directly participate in PG biosynthesis, such as GlmU, important for biosynthesis of UDP-GlcNAc, the flippase MurJ (MviN) or the bifunctional PBP PonA1

Results
Novel gpsB suppressor mutations in murA and prpC An L. monocytogenes ΔgpsB mutant has a temperature sensitive growth phenotype and cannot multiply at 42˚C [26]. However, this mutant readily picks up shg suppressor mutations (suppression of heat-sensitive growth) correcting this growth defect and known shg suppressor mutations map to clpC, murZ, reoM and reoY, all involved in regulation of MurA stability by ClpCP-dependent proteolysis [24,27]. In order to identify further mutations suppressing the gpsB phenotype, the L. monocytogenes ΔgpsB mutant strain LMJR19 was streaked on BHI plates and incubated at 42˚C. After two days of incubation, 50 suppressors were isolated and streaked to single colonies. The clpC, murZ, reoM and reoY genes of these 50 suppressors were amplified by PCR and sequenced by Sanger sequencing. Seven suppressors had wild type alleles of the four known gpsB suppressor genes and thus must have acquired novel shg suppressor mutations somewhere else on the chromosome. Genome sequencing of these seven shg suppressors revealed unique mutations mapping either to murA or to prpC. While the murA gene was affected by mutations introducing amino acid substitutions only (shg19: murA S262L, shg21: murA N197D), the shg mutations in prpC introduced amino acid exchanges (shg24: prpC N83S L125F, shg47: prpC G39S) as well as premature stop codons (shg32: prpC 1-essentiality of murA [27], but conflicts with our previous observation that prpC could not be removed from the L. monocytogenes genome [24]. Suppressor strains shg19 (ΔgpsB murA S262L), shg21 (ΔgpsB murA N197D) and shg55 (ΔgpsB prpC  were chosen for further experiments and growth was recorded at 37˚C and 42˚C. All three suppressor strains grew almost like wild type at 37˚C and 42˚C, whereas the ΔgpsB mutant grew with delayed rate at 37˚C and did not grow at 42˚C (Fig 1A and 1B).
Suppression in shg suppressors can involve proteolytic stabilization of MurA leading to stimulation of PG biosynthesis. In order to test, whether MurA is stabilized in the novel shg suppressors, their cellular MurA levels were determined by Western blotting. This revealed a mild increase of the MurA level in shg19 (2.1±0.6-fold compared to wild type) and shg21 (1.8 ±0.1-fold) and a strong increase in shg55 (6.6±1.7-fold), which is almost as strong as observed in a ΔclpC mutant (9.1±2.7-fold, Fig 1C). Taken together, we have identified novel shg suppressor mutations in murA and prpC affecting the cellular levels of MurA.

Deletion of prpC is tolerated in the ΔgpsB background
Previous results demonstrated that prpC could not be removed from the chromosome unless compensatory prkA mutations reduced the activity of the cognate protein kinase [24]. As this suggested that prpC represents an essential gene in L. monocytogenes, we were surprised to see that prpC repeatedly was inactivated by the introduction of premature stop codons in the gpsB suppressor mutants shg32, shg42 and shg55. To test whether prpC becomes dispensable in the absence of gpsB, we tried to delete prpC in the background of the ΔgpsB mutant using integration/re-excision of a temperature sensitive plasmid designed to remove the prpC gene during excision from the chromosome. Plasmid excision is supposed to generate a 50:50 mixture of wild type and deletion mutant clones [32]. Unlike in previous prpC deletion attempts in wild type background [24], prpC could be readily deleted in the ΔgpsB mutant. Furthermore, it even seemed that prpC deletion is favored in the absence of gpsB, as 19 out of 20 tested clones had factually lost prpC. Unlike the ΔgpsB single mutant, the resulting ΔgpsB ΔprpC double mutant could grow at 42˚C, even though it did not fully reach normal wild type growth (S1A Fig). Moreover, MurA levels increased 2.8±0.5-fold in the ΔgpsB ΔprpC double mutant compared to wild type (S1B Fig). These data confirm that inactivation of prpC suppresses the ΔgpsB phenotype and that suppression is likely explained by accumulation of MurA.

Effect of N197D and S262L mutations on MurA activity
MurA proteins consist of two globular domains connected to each other. The UDP-GlcNAc binding site and the catalytic center are located at the interface between these two domains [33]. The MurA residues replaced in the two shg suppressors are located outside the active site in helical regions exposed at the surface of the N-terminal (N197D) and the C-terminal (S262L) domain (Fig 2A). We assumed that these mutations would improve MurA activity to exert a similar suppressing effect on the ΔgpsB mutant as increased MurA levels. To test this, we generated strains that carry IPTG-inducible murA genes (wild type murA as well as S262L and N197D variants) at the attB plasmid integration site of their chromosomes but have their endogenous murA genes removed. Consistent with the essentiality of murA [27], all three strains required IPTG for growth in BHI broth (S2A Fig), when their pre-cultures were cultivated in the absence of IPTG. However, growth of strains LMSW140 (imurA S262L) and LMSW141 (imurA N197D) was similar to strain LMJR123 (imurA) in the presence of 1 mM IPTG (S2A Fig) and also at lower IPTG concentrations. Next, we measured the susceptibility of these strains to fosfomycin, a known inhibitor of MurA [34]. Fosfomycin sensitivity was measured in a disc diffusion assay on BHI agar plates, which allowed growth of the inducible murA strains even in the absence of IPTG due to background murA expression. Sensitivity of these strains towards fosfomycin was increased under these conditions compared to wild type due to MurA depletion, but effects of the N197D and S262L mutations did not become evident (S2B Fig). To further test the possibility that the two murA mutations altered enzyme activity, wild type MurA, MurA N197D and MurA S262L were purified as Strep-tagged proteins to near homogeneity (S2C Fig) and their enzymatic activity was tested. This showed that the MurA N197D protein was as active as wild type MurA, whereas the activity of MurA S262L was reduced (S2D Fig). Taken together, these results rule out the possibility that increased MurA activities explain suppression of the ΔgpsB phenotype.

The MurA N197D and S262L variants escape proteolytic degradation
Next, we wondered whether the two murA mutations affect proteolytic degradation of MurA. To begin with, this idea was tested in an experiment in which we titrated the expression level of murA and its N197D/S262L variants necessary for full suppression of the ΔgpsB growth defect. If any of these MurA variants is less prone to degradation and thus stabilized, then lower induction levels should be sufficient for suppression. In agreement with previous results [27], overexpression of wild type murA rescued growth of the ΔgpsB strain LMJR117 (S3A and S3B Fig), however, 0.5 mM was the minimal IPTG concentration leading to suppression. In contrast, ten-fold lower IPTG concentrations were sufficient for full suppression in the ΔgpsB strains LMS306 and LMS307 expressing IPTG-inducible murA N197D and murA S262L alleles, respectively, from their attB sites (S3C and S3D Fig). As this observation supports the idea of affected protein degradation, stability of MurA and its two variants was directly measured in vivo. To this end, the gpsB deletion in the two suppressor strains shg19 (ΔgpsB murA S262L) and shg21 (ΔgpsB murA N197D) was repaired by re-introduction of the native gpsB allele at the original locus to generate strains that carry the two murA mutations as the sole genetic changes. MurA degradation in the resulting strains LMSW155 (murA S262L) and LMSW156 (murA N197D) was then compared to wild type and the ΔclpC mutant. These strains were grown to an OD 600 of 1.0 and 100 μg/ml chloramphenicol was added to block protein biosynthesis. MurA levels over time were then determined by Western blotting. As reported previously [24], MurA levels rapidly declined over time in wild type cells, but stayed stable in a ΔclpC mutant (Fig 2B and 2C), reflecting ClpCP-dependent proteolytic degradation of MurA. However, degradation of both MurA mutant proteins was significantly delayed ( Fig  2B and 2C), which is in good agreement with their increased levels in the original shg19 and shg21 suppressors (Fig 1C). This shows that the MurA N197D and S262L variants escape proteolytic degradation by ClpCP.

Effect of murA escape mutations on formation of MurA-ReoM complexes
Previously published bacterial two hybrid data showed that MurA interacts with ReoM, which-together with ReoY-could present MurA to ClpCP [24]. We wondered whether the MurA N197D and S262L mutants have lost the ability to form the complex with ReoM and tested this in the bacterial two hybrid system. As published before [24], B2H detects an interaction of MurA with ReoM ( Fig 3A). However, the MurA N197D and S262L proteins have both (ΔgpsB ΔclpC) were included as controls. Equal amounts of total cellular proteins were loaded. Irrelevant lanes were removed (dashed line). MurA signals from three independent experiments were quantified by densitometry and average values and standard deviations are shown. Asterisks indicated statistically significant differences compared to wild type ( � ) or compared to the ΔgpsB mutant ( �� , P<0.05, t-test with Bonferroni-Holm correction).

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations lost their interaction with ReoM in several permutations, even though their interaction with wt MurA was unchanged (Fig 3A). To further validate this observation, we generated murA wt, N197D and S262L strains that produced his-tagged ReoM as a bait for pull-down analysis. We also deleted clpC in these strains to facilitate MurA detection after pull-down. The resulting strains were treated with formaldehyde to covalently crosslink interacting proteins prior to ReoM-His purification. Western blotting detected the presence of MurA in the eluates, indicating that MurA copurified with ReoM-His. Remarkably, MurA N197D (13±6% of wt MurA level) and S262L (19±11%) copurified to a lesser extent with ReoM-His (Fig 3B), even though their expression levels were similar to wt MurA in the ΔclpC background (S4 Fig). This confirms the idea that the two MurA variants could be stabilized because they do not longer bind to ReoM in vivo.
To study the interaction of ReoM with MurA even further, we switched to surface plasmon resonance. Binding was analysed in both orientations, i.e. with ReoM immobilized to the surface of the sensor chip and MurA as analyte in solution and vice versa. Specific binding was observed only when MurA was immobilized and ReoM was injected in solution ( Fig 3C), but not in the opposite orientation. The interaction was characterized by a very slow association rate k a of 5±1 × 10 2 M -1 s -1 , but was relatively stable with a dissociation rate constant k d of 4 ±2 × 10 −4 s -1 leading to an overall affinity of 0.9±0.5 μM. Similar results were observed with MurA mutants N197D (K D of 0.7±0.3 μM) and S262L (K D of 0.8±0.4 μM, S5A, S5B, and S5C Fig). This shows that ReoM can bind MurA directly. However, this direct interaction is not sensitive to the two escape mutations. We assume that the interaction measured by SPR in vitro does not fully reflect the spectrum of interactions between ReoM with MurA that occurs in vivo.

Escape mutations in murA affect peptidoglycan biosynthesis and growth
MurA was shown to be a key switch in regulation of PG biosynthesis and MurA levels strictly correlated with PG production and ceftriaxone resistance in L. monocytogenes [24,27]. Recent work in B. subtilis further confirmed this central role of MurA in PG biosynthesis regulation [35]. In order to test whether the two murA escape mutations influence PG production, their ceftriaxone resistance was tested and the ΔclpC and the imurA mutants were included as control. Ceftriaxone resistance was more than ten-fold higher in the ΔclpC mutant and inducerdependent in the IPTG-controlled imurA strain as shown previously [24]. However, ceftriaxone resistance was also 3-4-fold higher in the two murA escape mutants, suggesting that stabilization of MurA in these two strains indeed caused stimulation of PG biosynthesis (Fig 4A). Fluorescence microscopy of nile red-stained cells also revealed changes in cellular morphology, as cells of the two murA escape and ΔclpC mutants were clearly thinner (Fig 4B), whereas cell length was not affected. Apparently, MurA stabilization also impairs maintenance of normal cell width.
Differences in growth of the murA escape mutants were not detected in BHI broth at 37˚C, but a characteristic pattern that correlates well with the anticipated MurA expression levels became evident at 42˚C. Here, the ΔclpC mutant grew to a higher optical density in stationary phase and a congruent growth curve was observed with the imurA strain in the presence of IPTG (Fig 4C). We have shown earlier that both strains highly overexpress MurA under these points to analyze MurA levels. (C) MurA signals were quantified using ImageJ and average values and standard deviations are shown (n = 3). MurA levels of each strain before chloramphenicol addition (t = 0 min) were arbitrarily set to 1 and the remaining signal intensities at later time points are shown as relative values. Statistically significant differences (compared to wild type) are marked with asterisks (P<0.05, t-test with Bonferroni-Holm correction). https://doi.org/10.1371/journal.ppat.1010406.g002

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations conditions [27]. Remarkably, both murA escape mutants also lead to higher stationary phase optical densities at 42˚C (Fig 4C), suggesting that proteolytic degradation of MurA becomes important at higher temperature to limit bacterial growth. Analysis of the ultrastructure of the PG layer under these conditions revealed thicker PG layers in murA N197D, murA S262L and This suggests that control of PG production through proteolytic stabilization of MurA controls listerial growth and influences PG ultrastructure, at least at higher temperature.
That the two murA escape mutations with their positive effects on growth and antibiotic resistance have not succeeded during evolution suggests that they are also associated with evolutionary disadvantages. Accordingly, the murA N197D and S262L mutants were found to grow slower in the presence of salt (S7A Fig) and were more sensitive against lysozyme (S7B Fig). Apparently, increased biomass yields that are caused by MurA overproduction comes at the price of increased sensitivity against different stressors.

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations

The murA N197D escape mutation uncouples viability and PG biosynthesis from PrkA signaling
The prkA gene is essential in strain EGD-e because phosphorylated ReoM keeps ClpCP in control to prevent unregulated proteolysis of its essential substrate MurA [24]. Following with this model, MurA variants that escape proteolytic degradation through ClpCP should release the cell from PrkA dependency since ClpCP (even when overactive in the absence of PrkA and phosphorylated ReoM) does not longer accept such MurA variants as substrates. In order to test this hypothesis, we tried to delete prkA in murA N197D and murA S262L backgrounds, which never has been possible in the wild type background of L. monocytogenes EGD-e [24]. Similarly, the prkA gene could not be removed in the murA S262L background, presumably because of the reduced activity of MurA S262L. In contrast, prkA could be easily deleted in the murA N197D strain. The resulting ΔprkA murA N197D mutant was viable and grew like the wild type, whereas PrkA-depleted cells could not grow at all (Fig 5A). Moreover, the ΔprkA murA N197D mutant was viable inside macrophages and even grew intercellularly with a wild type-like growth rate. This was in complete contrast to PrkA-depleted cells (Fig 5B), suggesting that even the intracellular growth defect upon PrkA depletion may entirely be MurA-dependent.

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations We then tested whether overexpression of MurA would also rescue ΔprkA lethality. In fact, deletion of prkA turned out to be possible in strain LMJR116, in which a second copy of murA was expressed from an IPTG-dependent promoter. Growth of the resulting strain (LMS272) was not IPTG-dependent at 37˚C, probably due to leakiness of the promoter controlling expression of the second murA copy. However, IPTG was required to support growth of this strain at 42˚C (Fig 5C). The clpC gene is strongly induced at 42˚C [36], leading to stimulated MurA degradation. This explains why IPTG-dependency of this strain was temperature dependent. Importantly, this experiment demonstrated that overexpression of MurA is sufficient to suppress the essentiality of prkA, which further supports the idea that control of MurA stability is the main purpose of PrkA signaling under the tested conditions.
If MurA is indeed the relevant subject of PrkA:ReoM signaling, then the murA N197D mutation should also overcome the toxicity of the phospho-ablative reoM T7A variant. We have shown in previous work that expression of reoM T7A is toxic to L. monocytogenes because MurA is rapidly degraded when ReoM cannot be phosphorylated [24]. To test this idea, a murA N197D strain was generated, in which expression of reoM T7A can be controlled by IPTG (strain LMS273) and susceptibility of this strain against IPTG was tested in a disc diffusion assay. While IPTG was clearly toxic for strain LMSW52 (ireoM T7A), strain LMS273 fully tolerated IPTG (Fig 5D). This further confirms that the main role of the PrkA:ReoM signaling axis is control of MurA stability, at least under the growth condition tested here.
Interestingly, the extreme ceftriaxone sensitivity of PrkA-depleted cells did also not develop in the ΔprkA murA N197D strain ( Fig 6A). Likewise, deletion of the reoM, reoY and murZ genes, which are all involved in control of MurA degradation, suppressed lethality and ceftriaxone susceptibility of the ΔprkA deletion-albeit to different degrees (S8 Fig). Moreover, enhanced bacitracin and penicillin G susceptibilities associated with PrkA depletion were also suppressed by the murA N197D mutation (Fig 6A). Since all these observations indicate restoration/activation of PG biosynthesis upon MurA stabilization, PG biosynthesis was analyzed directly by cell staining with the fluorescent D-alanine and PG precursor HADA [37]. As expected, PrkA-depleted cells showed reduced staining intensity compared to wildtype, particularly at the cell poles, and this effect was abolished in the ΔprkA murAN197D mutant ( Fig  6B). Thus, PrkA not only becomes non-essential for viability and virulence, but also for PG biosynthesis in mutants, in which MurA escapes proteolytic degradation by ClpCP. Together with the results of Kelliher et al. [28], this suggests that PG biosynthesis is constitutively activated upon MurA stabilization.
To further verify our claims, we re-created the murA N197D mutation in the clean background of strain EGD-e. The resulting mutant (LMS308) showed the same ceftriaxone resistance as the gpsB-repaired murA N197D strain LMSW156 ( This extends previous conclusions on the relevance of ReoM for PrkA signaling [24,28] as it suggests that ReoM is specific for MurA and likely does not control the degradation of additional essential substrates in L. monocytogenes in a similar stringent manner.

ClpC-dependent activation of PG biosynthesis requires PBP B3 and RodA3
Our data together with the data of Kelliher et al. [28] suggest that stabilization of MurA enhances PG formation. Moreover, they show that ceftriaxone resistance levels tightly correlate with the degree of MurA accumulation [24,27], suggesting that ceftriaxone resistance can

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations be used as a proxy for PG biosynthesis. While most enzymes necessary for the cytoplasmic steps of PG biosynthesis are non-redundant, there is a high degree of redundancy in L. monocytogenes genes encoding PBPs and FtsW/RodA-like transglycosylases, required for the final steps outside the cell [38,39]. Inactivation of the class A PBPs had only mild effects on cephalosporin resistance in previous experiments [40], and were not further considered here. However, PBP B3, one out of three class B PBPs, substantially contributed to cephalosporin resistance [40,41]. Likewise, deletion of rodA1 and rodA2 only slightly affected cephalosporin resistance, whereas deletion of rodA3 had a greater impact [39]. Thus, it seemed likely that PBP B3 and RodA3 cooperate to ensure activation of PG biosynthesis and ceftriaxone resistance when MurA accumulates.
In order to test this hypothesis, we determined the effect of pbpB3 and rodA3 mutations on the ceftriaxone resistance level of the ΔclpC mutant (MIC 96±0 μg/ml), which is strongly increased compared to wildtype (9.3±2.3 μg/ml, Fig 7A). As deletion of pbpB3 was not possible in the ΔclpC background, clpC was deleted in an IPTG-inducible ipbpB3 mutant. Depletion of   Fig 7A). Remarkably, reduction of ceftriaxone resistance almost down to the level of a ΔpbpB3 mutant was also observed in the ipbpB3 ΔclpC strain, when grown in the absence of IPTG (1.0±0 μg/ ml). MurA overexpression in the wild type background yielded the same ceftriaxone resistance level (MIC 96±0 μg/ml) as observed with the ΔclpC mutant and deletion of pbpB3 also fully abrogated this effect (Fig 7B). This demonstrates that the increased ceftriaxone resistance levels of the ΔclpC mutant and after MurA overexpression are PBP B3-dependent.
Next, the rodA2-rodA1 gene pair and the rodA3 gene were inactivated in the ΔclpC mutant. While ceftriaxone resistance was only slightly affected by the ΔrodA2-rodA1 deletion (MIC 6.7 ±1.2 μg/ml) compared to wildtype (10.7±2.4 μg/ml in this experiment), inactivation of rodA3 had a somewhat stronger effect (5.3±1.2 μg/ml). More interestingly, the increased ceftriaxone resistance level of the ΔclpC mutant (96±0 μg/ml) was not affected by the ΔrodA2-rodA1 deletion (74.7±18.5 μg/ml), but was reduced back almost to wildtype level when rodA3 was inactivated (13.3±2.3 μg/ml, Fig 7C). All in all, this shows that PBP B3 and RodA3 collectively determine the increased ceftriaxone resistance of the ΔclpC mutant. This in turn suggests that PBP-FtsW and RodA could form a cognate pair and are of special importance for PG biosynthesis and ceftriaxone resistance, also under conditions leading to proteolytic stabilization of MurA.

Discussion
Many substrates of PASTA-eSTKs have been identified in the different Gram-positive bacteria. Among these kinase substrates are proteins from various cellular pathways, but some functional classes are overrepresented: (i) Proteins acting in carbon and cell wall metabolism such as the HPr and YvcK proteins of B. subtilis and L. monocytogenes [28,[42][43][44], (ii) regulatory proteins such as the WalR and GraR response regulators of B. subtilis and Staphylococcus aureus, respectively [45,46], and (iii) proteins acting in cell division, among which GpsB is one of the proteins consistently found as a PASTA-eSTK substrate in B. subtilis, L. monocytogenes and Streptococcus pneumoniae [28,47,48]. This diversity of substrates and their own-often pleiotropic-functions have impeded the identification of primary and the discrimination of less important kinase substrates. We previously have identified ReoM as a substrate of PrkA in L. monocytogenes EGD-e and could show that deletion of reoM suppresses loss of viability that results from PrkA depletion [24]. Moreover, expression of a phospho-ablative reoM T7A allele was toxic on its own and thus phenocopied the absence of PrkA [24]. These tight geno-and phenotypic relations between the PrkA kinase and ReoM, which is only one out of 23 known PrkA substrates in L. monocytogenes [28], suggested that ReoM must be a relevant kinase target. We here further support this conclusion by our observation that prkA even can be deleted in a ΔreoM background, again indicating that the phosphorylation of ReoM is the particular PrkA-dependent phosphorylation event that is crucial for viability. Remarkably, Kelliher et al. reported that prkA is non-essential in L. monocytogenes strain 10403S [28] and discuss strainspecific differences as an explanation. In fact, L. monocytogenes strains EGD-e and 10403S are relatively distinct with respect to the population structure of the species L. monocytogenes and even belong to different molecular serogroups (EGD-e: IIc, 10403S: IIa) [49]. Thus, differences in the stringency of MurA degradation by ClpCP, in expression of the genes acting in the LMS268 (ΔclpC rodA3´), LMSH67 (ΔrodA2-rodA1) and LMS270 (ΔclpC ΔrodA2-rodA1). MICs were determined using E-tests and average values and standard deviations calculated from three repetitions are shown. Asterisks indicate statistically significant differences compared to the ΔclpC mutant (P<0.01, t-test with Bonferroni-Holm correction). https://doi.org/10.1371/journal.ppat.1010406.g007

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations ReoM/ClpCP/MurA axis or even in the relative expression of murA and its paralogue murZ may account for this difference. Despite this discrepancy and consistent with our findings, reoM deletion suppressed all prkA associated phenotypes also in strain 10403S [28].
How ReoM stimulates MurA degradation is presently not known, but it may act as a ClpCP activator or as an adaptor, which presents substrates to the protease complex. ReoM could either be specific for MurA or could also control the stability of other proteins. Our observation that MurA variants exist that escape proteolytic degradation, further supports the idea that MurA is a protease substrate even though this has never been confirmed in vitro [24,27,30]. Moreover, the observations that their amino acid substitutions are located on the MurA surface and impair the interaction with ReoM would be consistent with the adaptor hypothesis. An interesting aspect is the observation that the escape mutations disturb formation of ReoM:MurA complexes in vivo (even in a heterologous system) but not in vitro. We cannot explain this at the moment, but speculate that effects that occur during translation of MurA, for which SPR is insensitive, could be a possible reason for this discrepancy.
A central result of our study is the finding that one of the murA escape mutations (N197D) suppressed prkA essentiality under standard growth conditions and during intracellular growth in macrophages, to which prkA mutants are specifically sensitive [44]. This particular murA mutation does not alter enzymatic activity and only rescues MurA from proteolytic degradation. It also overcomes the lethality of the reoM T7A allele, the expression of which normally would lead to rapid MurA degradation. These findings provide a genetic answer to the question whether other L. monocytogenes proteins exist, the stability of which depends on ReoM phosphorylation: As uncoupling of MurA from PrkA-dependent control of ReoMmediated ClpCP activation restores otherwise lethal prkA and reoM phenotypes, we have to conclude that MurA is not only the main substrate or target of ReoM, but also that control of MurA stability is the main purpose of PrkA-mediated signaling in L. monocytogenes during normal laboratory growth and inside macrophages. This conclusion is further substantiated by the observation that artificial MurA overexpression also rescues the lethal ΔprkA phenotype. According to our present model, PrkA gets activated during cell wall stress and phosphorylation of ReoM limits MurA degradation (Fig 8). Our data suggest that PG biosynthesis is activated when MurA accumulates. How PG biosynthesis activation under these conditions alters the chemical structure of the sacculus is not known, but a thicker cell wall is produced, especially at the poles, and this occurs concomitantly with an increase of ceftriaxone resistance. L. monocytogenes PrkA presumably localizes to the cell division septum as reported for the PASTA kinases of B. subtilis, S. pneumoniae and S. aureus [17,47,50]. Such a possible septal localization of PrkA could be related to the polar PG thickening in murA (this work), reoM and reoY mutants [24]. In all these mutants, MurA escapes degradation and if this occurs primarily at those subcellular sites, at which PrkA is enriched (i. e. the septum), then thicker PG cross walls and later thicker polar PG layers might be a possible consequence. That the murA escape mutants are more susceptible to lysozyme despite their thicker PG might be explained by imbalances between increased PG biosynthesis and unaltered PG modification. PG modifying reactions such as N-deacetylation and O-acetylation are the main determinants of lysozyme resistance in L. monocytogenes [51,52] and increased lysozyme resistance could occur when PG modification does not keep pace with PG biosynthesis.
As a second important result of our study we could show that the high level of ceftriaxone resistance upon MurA stabilization depends on RodA3 and PBP B3, which may constitute a cognate glycosyltransferase/transpeptidase pair (Fig 8). In our current model, the PrkA!ReoM/ReoY!MurA/ClpCP!RodA3/PBP B3 cascade would represent a closed homeostatic circuit that senses lipid II (by PrkA) to control its production (through control of MurA stability) and consumption (by RodA3/PBP B3) (Fig 8). The genes in this pathway collectively determine the intrinsic resistance of L. monocytogenes to cephalosporins [24,[39][40][41]53] and are conserved in several other Gram-positive bacteria (S10 Fig). In Enterococcus faecalis, most of the corresponding homologues were shown to maintain the high intrinsic cephalosporin resistance level of this organism [29,[54][55][56][57], and deletion of pbpC, which encodes the PBP B3 homologue of B. subtilis (S10 Fig), was also necessary for cephalosporin resistance [58]. Moreover, inactivation of Stk1, the PASTA-eSTK of S. aureus, sensitizes methicillin resistant S. aureus (MRSA) against β-lactams including oxacillin and this is overcome by reoM and reoY deletions [28,59]. Interestingly, MecA, which represents the major methicillin resistance determinant of MRSA [60], is the equivalent homologue of L. monocytogenes PBP B3 (S10 Fig) [8], suggesting that methicillin resistance in MRSA is regulated in an analogous way. Lastly, the PrkA signaling cascade lacks ReoY and a homologue of PBP B3 in S. pneumoniae, which is typically susceptible to penicillins. This raises the interesting possibility that this cascade generally supports a specific type of PG biosynthesis or repair, which confers a higher resistance against cephalosporins and other β-lactams. In which way PG biosynthesis is altered under such conditions is not known, but PG formation at specific subcellular sites or generation of PG with particular crosslinking rates or with a certain tertiary structure would be  [24,27]. Image with modifications from [74]. https://doi.org/10.1371/journal.ppat.1010406.g008

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations conceivable options. Further studies, also including other bacterial species, would have to be performed to test such hypotheses. Table 1 lists all strains used in this study. L. monocytogenes strains were routinely cultivated in brain heart infusion (BHI) broth or on BHI agar plates. Growth curves were recorded in 96 well microplate readers. Antibiotics and supplements were added when required at the following concentrations: erythromycin (5 μg/ml), kanamycin (50 μg/ml), X-Gal (100 μg/ml) and IPTG (as indicated). Escherichia coli TOP10 was used as standard host for all cloning procedures [61].

General methods, manipulation of DNA and oligonucleotide primers
Standard methods were used for transformation of E. coli, isolation of plasmid DNA and chromosomal DNA isolation [61]. Transformation of L. monocytogenes was carried out by electroporation as described by others [62]. The manufacturer´s instructions were followed for restriction and ligation of DNA. All primer sequences are listed in Table 2.

Construction of plasmids and strains
Plasmid pSW52 was constructed for overexpression of Strep-tagged MurA in E. coli. To this end, the murA open reading frame was amplified with primers SW130/SW131 and cloned into pET11a using SpeI/XhoI. The S262L and N197D mutations were then introduced into pSW52 by quikchange mutagenesis using the primer pairs SW150/SW151 and SW152/SW153, respectively. The same primers were used to introduce both murA mutations into plasmid pJR82 by quikchange mutagenesis, yielding plasmids pSW63 and pSW64 for inducible expression of murA S262L and murA N197D, respectively, in L. monocytogenes. Likewise, these primers were used to introduce both mutations into the murA gene of the bacterial two hybrid vectors pJR116 and pJR117.
Plasmid pSH571 was constructed for introduction of the murA N197D mutation into the genome. For this purpose, murA fragments up-and downstream of N197 were amplified with primers SHW948/SHW951 and SHW950/SHW949, respectively, and joined by splicing by overlapping extension PCR (SHW950 and SHW951 introduced the N197D mutation). The resulting fragment was inserted into pMAD by restriction free cloning.
Plasmid pJD16 was generated for reoM-his expression in L. monocytogenes. To this end, reoM was amplified using JD13/JD14 as the primers (the reverse primer introduced a His-10 tag) and cloned into pIMK2 using PstI/SalI. An unwanted sequence that arose during cloning was then removed from the reoM N-terminus by restriction free cloning using the oligonucleotides JD25/JD26.
Derivatives of pIMK2 and pIMK3 were introduced into L. monocytogenes strains by electroporation and clones were selected on BHI agar plates containing kanamycin. Plasmid insertion at the attB site of the tRNA Arg locus was verified by PCR. Plasmid derivatives of pMAD were transformed into the respective L. monocytogenes recipient strains and genes were deleted as described elsewhere [32]. All gene deletions and insertions were confirmed by PCR. Plasmid pSAH67, designed for insertional disruption of rodA3, was transformed into L. monocytogenes and transformants were selected on BHI agar plates containing erythromycin at 30˚C. Plasmid insertion into the chromosomal rodA3 gene was enforced by streaking the transformants to single colonies on BHI agar plates containing erythromycin at 40˚C. Disruption of rodA3 was confirmed by PCR.

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations

Antibiotic susceptibility testing
Minimal inhibitory concentrations were determined as described previously using E-test strips with a ceftriaxone or bacitracin concentration range of 0.016-256 μg/ml or a penicillin G concentration range of 0.002-32 μg/ml (all from Bestbion dx , Germany) [38]. Susceptibility against fosfomycin was tested in a disc diffusion assay using filter discs loaded with 8 μl of a 20 mg/ml fosfomycin solution. For this purpose, L. monocytogenes colonies, grown on BHI agar plates, were resuspended in BHI broth and used to swab-inoculate BHI agar plates. Filter discs were spotted on BHI agar plates and incubated over night at 37˚C to measure the inhibition zone diameter the next day.

Genome sequencing
Chromosomal DNA for genome sequencing was isolated using the GenElute Bacterial Genomic DNA Kit (Sigma-Aldrich). Libraries were generated from 1 ng genomic DNA by using the Nextera XT DNA Library Prep Kit (Illumina). Sequencing was performed using a MiSeq Reagent Kit v3 cartridge (600-cycle kit) on a MiSeq benchtop sequencer in paired-end mode (2 x 300 bp). Reads were mapped to the L. monocytogenes EGD-e reference genome (NC_003210.1) [65] in Geneious (Biomatters Ltd.) to identify single nucleotide polymorphisms using the Geneious SNP finder tool. Genome sequences of gpsB suppressor strains were deposited at ENA under project accession number PRJEB47255.

Isolation of cellular proteins and Western blotting
For protein isolation, cells from a 20 ml culture volume were harvested by centrifugation, washed with ZAP buffer (10 mM Tris/HCl pH7.5, 200 mM NaCl), resuspended in 1 ml ZAP buffer also containing 1 mM PMSF and disrupted by sonication. Cellular debris was removed

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations by centrifugation and the supernatant was considered as total cellular protein extract. Aliquots of these protein samples were separated by SDS polyacrylamide gel electrophoresis (PAGE) and transferred onto positively charged polyvinylidene fluoride membranes employing a semi-dry transfer unit. DivIVA and MurA were detected by polyclonal rabbit antisera recognizing B. subtilis MurAA [66] and DivIVA [67] as the primary antibodies, respectively. An anti-rabbit immunoglobulin G conjugated to horseradish peroxidase was used as the secondary antibody and detected using the ECL chemiluminescence detection system (Thermo Scientific) in a chemiluminescence imager (Vilber Lourmat).

In vivo formaldehyde crosslinking and pull-down
In vivo crosslinking of protein complexes using formaldehyde, purification of His-tagged bait proteins using MagneHis magnetic beads and protein decrosslinking by heat treatment were performed according to a previously published protocol [68]. Decrosslinked samples were separated by SDS-PAGE and either stained using a colloidal Coomassie stain or analyzed by Western blotting.

Protein purification
Proteins were overproduced in E. coli BL21. Overexpression strains were cultivated in 1 l LB broth (containing 100 μg/ml ampicillin) at 37˚C and 250 rpm. Protein expression was induced by addition of 1 mM IPTG at OD 600 = 0.5. Cultivation was continued over night at 16˚C and 200 rpm. Cells were harvested by centrifugation (11.325 x g, 10 min, 4˚C) and the cell pellet was washed once in 20 ml ZAP buffer. Afterwards, the cell pellet was resuspended in 20 ml ZAP buffer containing 1 mM PMSF. Cells were disrupted using an EmulsiFlex C3 homogenizer (Avestin Europe GmbH), cell debris was removed by centrifugation (4.581 x g, 10 min, 4˚C) and the lysate was cleared in an additional passage through a filter (0.45 μm pore size).
Strep-tagged proteins were purified using affinity chromatography and Strep-Tactin Sepharose (IBA Lifesciences, Germany) according to the manufacturer's instructions. Fractions containing purified proteins were pooled, aliquoted and stored at -20˚C.

MurA activity assay
An assay monitoring phosphate release from phosphoenol-pyruvate (PEP) was used for determination of MurA activity [69]. For this purpose, 5 μg of purified MurA protein were mixed with 10 mM uridine 5 0 -diphospho-N-acetylglucosamine (UDP-GlcNAc, Sigma-Aldrich) in a reaction buffer containing 100 mM Tris/HCl pH8.0 and 150 mM NaCl (final volume: 50 μl) and preincubated at 37˚C for 15 min. The reaction was started by addition of 5 μl 10 mM PEP (Sigma-Aldrich) and stopped by addition of 800 μl staining solution after different time intervals. The staining solution was freshly prepared from 10 ml ammonium molybdate solution (4.2 g in 100 ml 4 M HCl), 30 ml malachite green solution (225 mg malachite green in 500 ml H 2 O) and 10 μl Triton X-100. Absorption was measured at λ = 660 nm, corrected for background in the absence of UDP-GlcNAc and used to calculate the amount of released phosphate using a standard curve generated with solutions with different phosphate concentrations.

Surface plasmon resonance
Binding kinetics between ReoM and MurA were analysed by surface plasmon resonance (SPR) measurements at 25˚C using HBS-EP+ as running buffer (10 mM HEPES, 150 mM NaCl, pH 7.4, 3 mM EDTA, 0.05% Tween 20) on a Biacore T200 unit. For ligand immobilisation Series S sensor chips CM5 and covalent amine coupling by using an Amine Coupling Kit

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations were employed (all Cytiva, Freiburg, Germany). In initial experiments, either recombinant ReoM (20 μg/ml, diluted in 10 mM acetate buffer, pH 4.5) or recombinant MurA wt (20 μg/ ml, diluted in 10 mM acetate buffer, pH 4.5) were immobilized on the surface of a sensor chip at high surface densities by injecting the diluted proteins over EDC/NHS-activated CM5 chips for 7 minutes at a flow rate of 10 μl per minute before blocking with 1.0 M ethanolamine-HCl pH 8.5. As a negative control, bovine serum albumin was coupled on the respective reference flow cells. Subsequently, in an initial binding experiment, either a 1:3 dilution series starting at 1 μM (46 μg/ml) MurA wt was injected over immobilized ReoM (3032 RU final response) and BSA (7715 RU) for 60 s at a flow rate of 30 μl/min followed by 300 s injections of HBS-EP+ for 300 s or, by switching orientation between immobilized ligand and analyte in solution, a 1:3 dilution series starting at 10 μM ReoM (210 μg/ml) was injected over immobilized MurA wt (5359 RU) and BSA (7520 RU) in a similar fashion. Optimum regeneration conditions to remove bound ReoM from immobilized MurA were determined by a regeneration scouting.
Here, 10 mM glycin at pH 1.5 injected for 120 s at a flow rate of 10 μl/min completely removed ReoM as analyte while retaining the binding activity of immobilized MurA. For kinetic binding analysis, the immobilisation levels of BSA on the control flow cell 1 (265 RU) and MurA N197D (flow cell 2, 512 RU), MurA S262L (flow cell 3, 754 RU) and MurA wt (flow cell 4, 525 RU) were lowered to a target immobilisation level of 500 resonance units (RU) to avoid mass transport limitation and avidity effects for binding of bivalent ReoM. ReoM was injected as analyte in solution for 120 s to monitor binding association followed by 600 s injection of HBS-EP+ to monitor complex dissociation. The highest concentration of ReoM was 280 μg/ml (13.3 μM) followed by six 1:3 dilutions with duplicate measurements of 4.3 μM. Regeneration was performed as determined in the regeneration scouting, all measurements were performed in two independent experiments. Double referenced binding curves [70] were fit to a 1:1: Langmuir interaction model using the Biacore Evaluation software (Cytiva, version 3.2) to determine association rate constants k a , dissociation rate constants k d and the equilibrium dissociation constant K D .

Lysis assays
Lysis assays were performed as described previously [71] with minor modifications. L. monocytogenes strains were grown in BHI broth at 37˚C until an optical density of around OD 600 1.0. Cells were collected by centrifugation (6000 x g, 5 min, 4˚C) and the cell pellet was resuspended in 50 mM Tris/HCl pH8.0 to an optical density of OD 600 = 1.0. Lysozyme was added to a final concentration of 2.5 μg /ml (where indicated) and the cells were shaken at 37˚C. Lysis was followed by measuring optical density every 15 min in a spectrophotometer.

Macrophage infection assay
Experiments to measure intracellular growth of L. monocytogenes strains inside J774 mouse macrophages were essentially carried out as described earlier [72,73].

Microscopy
Cell membranes were stained through addition of 1 μl of nile red solution (100 μg ml -1 in DMSO) to 100 μl of exponentially growing bacteria. Images were taken with a Nikon Eclipse Ti microscope coupled to a Nikon DS-MBWc CCD camera and processed using the NIS elements AR software package (Nikon). Cell widths were determined using the tools for distance measurements provided by NIS elements AR.

PLOS PATHOGENS
Release of cell wall biosynthesis from PrkA control by murA mutations hydrochloride (HADA) and grown overnight at 37˚C. Cells were washed two times with 500 μl BHI broth and subjected to fluorescence microcopy. ImageJ was used to determine the fluorescence intensity maxima at the cell poles by densitometry. All values were corrected for background before further analysis.
Ultrathin section transmission electron microscopy was performed essentially as described earlier using a Tecnai 12 transmission electron microscope (Thermo Fisher/ FEI) operated at 120 kV [38].