Regulation of cell wall plasticity by nucleotide metabolism in Lactococcus lactis

characters Link between nucleotide ABSTRACT To ensure optimal cell growth and separation, and to adapt to environmental parameters, bacteria have to maintain a balance between cell wall (CW) rigidity and flexibility. This can be achieved by a concerted action of peptidoglycan (PG) hydrolases and PG synthesizing/modifying enzymes. In a search for new regulatory mechanisms responsible for the maintenance of this equilibrium in Lactococcus lactis , we isolated mutants that are resistant to the PG hydrolase lysozyme. We found that 14% of the causative mutations were mapped in the guaA gene, the product of which is involved in purine metabolism. Genetic and transcriptional analyses combined with PG structure determination of the guaA mutant enabled us to reveal the pivotal role of the pyrB gene in the regulation of CW rigidity. Our results indicate that conversion of L-aspartate (L-Asp) to N-carbamoyl-L-aspartate by PyrB may reduce the amount of L-Asp available for PG synthesis and thus cause the appearance of Asp/Asn-less stem peptides in PG. Such stem peptides do not form PG cross-bridges, resulting in a decrease in PG crosslinking and, consequently, reduced PG thickness and rigidity. We hypothesize that the concurrent utilization of L-Asp for pyrimidine and PG synthesis may be part of the regulatory scheme, ensuring CW flexibility during growth rigidity in stationary phase. The fact that L-Asp availability is dependent on nucleotide metabolism, which is tightly regulated in accordance with the growth rate, provides L. lactis cells the means to ensure optimal CW plasticity without the need to control the expression of PG synthesis genes.


INTRODUCTION
Peptidoglycan (PG) is the major component of the Gram-positive bacterial cell wall (CW), which envelops the cell as a multilayer sacculus. PG consists of a basic unit made up of N-acetylglucosamine-N-acetyl-muramic acid (GlcNAc-MurNAc) disaccharides bound to stem pentapeptides. Disaccharide pentapeptide units are synthesized intracellularly and transported through the cytoplasmic membrane as lipid-disaccharide pentapeptides, called lipid II. These blocks are covalently linked to the preexisting PG polymers by high-molecular-weight penicillin binding proteins, or PBPs (1). Class A PBPs contain both transglycosylation and transpeptidation domains, whereas class B PBPs are involved only in transpeptidation. Transglycosylation links the disaccharide pentapeptide to the pre-existing PG chain, while transpeptidation connects the stem pentapeptides to neighboring chains, which ensures PG cross-linking through the formation of an interpeptide bridge. Cross-linking in Lactococcus lactis involves the synthesis of an interpeptide bridge made of one D-amino acid (D-Asp or D-Asn), and, in this species, the PG cross-linking index was estimated to be 35.5% (2). Studies of Staphylococcus aureus have revealed that this PG cross-linking correlates with CW rigidity (3).
The basic PG structure is often modified, as PG glycan chains can undergo Ndeacetylation or O-acetylation, and free carboxyl-groups of amino acids of peptide chains may be amidated (4). In L. lactis, MurNAc O-acetylase is encoded by the oatA gene and is associated with resistance to peptidoglycan hydrolases (PGH) (5). N-Deacetylation of the GlcNAc present in PG is achieved by the PG-deacetylase PgdA and has also been shown to protect PG from PGH activity (6,7). The free carboxyl groups of PGforming amino acids are amidated intracellularly, before the precursors are translocated through the cytoplasmic membrane (8). In L. lactis, these amino acids include D-Glu, found in stem peptides, and D-Asp, on side chains or cross-bridges (2). Amidation of D-Asp takes place after it has been added to the PG precursor by AslA and is catalyzed by an asparagine synthase (AsnH) (9), since D-Asn is not a substrate for the aspartate ligase AslA (10). In L. lactis, the D-Asp cross-bridge is only partially (75%) amidated during the exponential phase, in contrast to other bacteria in which amidation is almost complete (2). D-Asp amidation of L. lactis PG decreases sensitivity to cationic antimicrobials such as lysozyme or nisin, which may be related to a reduction of the net negative charge inside the cell wall (10).
A strong PG sacculus is needed to counteract both high turgor pressure and cell wall stress related to environmental factors. At the same time, cleavage of PG strands is needed to allow the insertion of newly synthesized PG blocks during bacterial cell growth and for daughter cell separation after division (2,11,12). PG flexibility is especially advantageous during exponential growth. Two opposing features of PG, i.e. rigidity and flexibility, require the coordinated and balanced action of PG synthesizing and degradation enzymes. The loss of this balance may cause growth arrest and/or cell lysis. In bacteria, such a balance is achieved mostly by regulating the activities of potentially lethal autolytic enzymes such as PGHs (13). The factors that affect CW sensitivity to autolysins include i) their proteolytic degradation (14), ii) their specific localization within the cell, often at the septal region, iii) shielding of PG from PGH by secondary cell wall polymers such as teichoic acids or wall polysaccharides, iv) alanylation of (lipo-)teichoic acids, v) Oacetylation or N-de-acetylation of PG, vi) amidation of D-Glu and D-Asp in PG stem peptides, and vii) glycosylation of autolysins (2). In addition to the introduction of PG breaks by PGHs, PG flexibility in L. lactis could also result from defective PG synthesis and crosslinking due to a deficiency in PonA, one of the class-A PBPs (15). PG plasticity could also be enhanced by the incorporation of exogenously added non-canonical D-amino acids into PG (16). For example, the weakening of CWs by incorporation of the achiral amino acid glycine is exploited for the preparation of competent cells of Gram-positive bacteria (17,18). It is important to note that too-rigid PG also has deleterious consequences for bacterial growth. Overexpression of the L. lactis oatA gene leads to PG that is excessively resistant to PGH, and consequently results in growth arrest (7). The lethal effect of overly strong PG is consistent with findings that several PGHs are collectively essential for the growth of Escherichia coli, Bacillus subtilis, and Staphylococcus aureus. This lethality could be related to the need for PGH-mediated PG remodeling during key steps in the cell cycle (19)(20)(21)(22).
To identify factors that may be involved in maintenance of the balance between CW flexibility and CW rigidity during cell growth, we isolated mutants of L. lactis that were resistant to hen egg white lysozyme. Analysis of one class of resistant mutants, which had a deletion of a large chromosome fragment encompassing the guaA gene, led us to the potential mechanism that allows the adjustment of CW rigidity to bacterial growth rate requirements without the transcriptional regulation of gene expression. This mechanism is based on the utilization of L-Asp for both PG and pyrimidine synthesis. To the best of our knowledge the presented data are the first indication of a link between nucleotide metabolism and cell wall plasticity in bacteria.

RESULTS
Screening for spontaneous lysozymeresistant mutants reveals a deficiency in guanine biosynthesis -To select mutants with affected in CW structure we chose to use lysozyme for its double antibacterial activity. First, it is a muramidase and hydrolyzes the ß-1,4 glycosidic bonds between Nacetylmuramic acid and N-acetylglucosamine of PG, ultimately resulting in cell lysis. Second, lysozyme acts as a cationic antimicrobial that creates pores in the cytoplasmic membrane, leading to leakage of intracellular ions from the cell (23). We reasoned that using such "bi-cidal" antimicrobial would allow isolation of a larger spectrum of different mutants. We selected 59 independent spontaneous lysozyme-resistant mutants of L. lactis strain MG1363. Of these, 8 (14%) were unable to efficiently grow in rich M17G medium. This deficiency was corrected by the addition of guanine (20 µg/ml). The guanine deficiency was also confirmed using chemically defined medium [SA, (24), results not shown]. Since it has been reported that guanine auxotrophy could be due to inactivation of the guaA gene that encodes GMP synthase (25,26), we verified the presence of mutations in this gene in the guanine auxotrophs. The guaA gene has been shown to be involved in acid resistance in L. lactis, a phenotype that could be linked to CW modification (26). Unexpectedly, we failed to PCR-amplify the DNA fragment carrying the guaA gene using primers gua1 and gua2 from the chromosomal DNA of the obtained mutants ( Fig. 1). In the L. lactis chromosome, the large 33 893-bp DNA fragment that carries the guaA gene is flanked by two IS905 elements oriented in the same direction (27). This organization allows the deletion of this fragment by homologous intra-chromosomal recombination. Indeed, large chromosomal rearrangements due to intra-chromosomal recombination between IS905 elements were previously shown to occur in L. lactis (28). We tested this possibility by PCR with primers gua3 and gua4, located outside of the IS905 elements, and obtained a 3590-bp fragment from all eight spontaneous guanine auxotrophs. DNA nucleotide sequence determination of the amplified DNA fragment from one of these mutants, VES2824, showed that it contained one copy of IS905 DNA, thus confirming that an intra-chromosomal recombination event was the basis of the deletion. Since the deleted fragment included other genes involved in purine metabolism (Fig. 1), we constructed a deletion mutant of only the guaA gene (strain VES4883, Table 1). As was the case for VES2824, the mutant VES4883 required guanine for growth in M17G and SA media and exhibited identical lysozyme resistance. The lysozyme resistance of strains VES2824 and VES4883 decreased when guanine was added to the medium (Fig. 2).
Transcriptome analysis of the ∆guaA mutant shows down-regulation of pyrimidine biosynthetic genes -To determine which genes may be involved in the lysozyme resistance of strain VES4883 (∆guaA), we compared its transcriptional profiles with that of the control strain MG1363. In order to identify which genes are induced by cell wall stress, we included in this assay MG1363 cells treated with lysozyme (Table 2). In the ∆guaA mutant, we observed that, first, the dlt-operon genes dltA, dltB, and dltD, which are responsible for D-alanylation of teichoic acids (29), were upregulated more than three-fold. Second, expression of ponA, which encodes the PGsynthesis enzyme PBP1A (30), was upregulated four-fold in the ∆guaA mutant. Since this gene is involved in PG assembly, its increased expression could result in more cross-linked PG and thus contribute to lysozyme resistance. Third, expression of pyrR, which encodes the transcriptional regulator of the pyrimidine biosynthetic genes, was markedly decreased. This may explain the observation of strong down-regulation of other genes involved in pyrimidine biosynthesis: pyrB, pyrP, carA, pyrK, pyrDb, pyrF, pyrE, and pyrC (25). The only PG-synthesis genes that were affected in the ∆guaA mutant were murB and murC, which encode enzymes involved in the synthesis of UDP-MurNAc pentapeptide precursors (2). Notably, the cellenvelope-stress genes cesSR and spxB were not up-regulated in the ∆guaA mutant, indicating that the lysozyme resistance of this strain is not monitored by the cesSR regulon.
The cesSR and spxB genes were induced in lysozyme-treated MG1363 cells as expected. This corroborated the hypothesis that lactococci react to cell wall stress by inducing genes of the cesSR operon, which leads to resistance to PG hydrolysis (7,31). The expression of PG-synthesis genes was not affected in lysozyme-treated cells, indicating that PG synthesis is not regulated in response to lysozyme-provoked CW stress. Despite its apparent role in lysozyme resistance, dlt operon expression was also not affected under these stress conditions.
Inactivation of pyrB confers lysozyme resistance and links pyrimidine biosynthesis to peptidoglycan assembly -The genes pyrP, pyrR, pyrB, and carA, which are strongly repressed in the ∆guaA mutant (Table 2), constitute an operon in L. lactis MG1363 (32). The pyrP gene encodes a uracil permease, required for utilization of exogenous uracil. The other two genes in the operon, pyrB and carA, encode pyrimidine biosynthetic enzymes. In L. lactis, PyrB is a unique aspartate transcarbamoylase, which converts L-Asp to L-carbamoyl-L-aspartate (25). L-Asp is also used for CW biosynthesis: it is converted to D-Asp by RacD racemase, and is subsequently attached to the stem peptide of PG by AslA ligase (10) and then converted to D-Asn by AsnH (9). Simultaneous utilization of L-Asp for PG and for pyrimidine biosynthesis could make the D-Asp/Asn content in the CW dependent on pyrimidine biosynthesis. In this case, pyrB could play a pivotal role in the tradeoff between L-Asp utilization for PG or for pyrimidine biosynthesis. Therefore we focused further studies on this gene by constructing a pyrB deletion mutant of MG1363 and testing this strain for lysozyme resistance.
Since the L. lactis pyrB mutant is a uracil auxotroph (25), the lysozyme resistance test was performed in M17G medium supplemented with uracil (100 µg/ml). We also included L. lactis dltD and ponA mutants, since the expression of these genes was affected in the transcriptomics experiment described above (Table 2). We observed that a ponA mutation only slightly affected lysozyme resistance while the inactivation of dltD in the ∆guaA mutant markedly abolished lysozyme resistance (Fig. 3). This indicates that increased TA alanylation may indeed play an important role in this phenotype of the ∆guaA mutant. Inactivation of pyrB resulted in lysozyme resistance, albeit slightly weaker than that of the ∆guaA mutant and of strain VES2824, which carried the large chromosomal deletion. Inactivation of pyrB in the ∆guaA mutant did not affect the lysozyme resistance phenotype, which was consistent with the transcriptomics data showing that pyrB is already strongly repressed in the ∆guaA mutant.
Inactivation of pyrB results in a thicker and more rigid cell wall -The cell envelope of the ∆pyrB mutant, growing exponentially in M17G medium was considerably thicker (41±1.3 nm, Fig. 4B) than that of parental strain MG1363 (36±1,6 nm), as observed in transmission electron micrographs (Fig. 4A). Cell wall thickening was also observed in the ∆pyrB mutant when cells were grown in M17G supplemented with uracil (44±2,6 versus 35±1,7 nm for MG1363, Fig.  4B). As the TEM results pointed towards possible changes in cell wall rigidity in the ∆pyrB mutant, strains VES6497 (∆pyrB) and MG1363 (WT) were grown in M17G medium with and without uracil (100 µg/ml) and examined by atomic force microscopy (AFM) in order to measure cell wall rigidity (33,34). Using an innovative method for sample immobilization in micro-structured polydimethylsiloxane (PDMS) stamps (35), combined with multi-parametric imaging, we were able to image L. lactis cells collected in the exponential growth phase while simultaneously probing their nanomechanical properties (rigidity of the cell wall). Height images (Fig. 5A, B, C, and D) showed that the cell morphology was not modified by either pyrB inactivation or the addition of uracil to the culture medium. On the corresponding rigidity images (Fig. 5E, F, G, and H), each pixel corresponds to a Young's modulus (YM) value which reflects the rigidity of the cell wall. These images thus revealed that global stiffness varied between the two strains, with darker MG1363 cells being softer than the lighter ∆pyrB mutants. However, two dividing cells did not always present the same nanomechanical properties (see Fig. 5G), which indicates that exponentially growing cells are heterogeneous in their cell wall rigidity. To precisely quantify the variations in cell stiffness, local force measurements were performed in small areas (500 × 500 nm) of the cells (36). Fig. 5I shows representative force vs. indentation obtained from the two strains in the two growth conditions, and fitted with a Hertz model to extract the YM values. As was already observed in the multiparametric imaging data, indentation curves ( Fig. 5I) clearly show that MG1363 cells are softer than ∆pyrB mutant cells. Indeed, MG1363 cells display an average YM value that is eight-fold lower than that of ∆pyrB mutant cells, both in the absence (226.9 ± 160.5 kPa vs. 1707.4 ± 920.4 kPa, respectively) and the presence of uracil (95.8 ± 61.7 kPa vs. 780.1 ± 560.7 kPa, respectively) ( Fig. 5J). Despite the large standard variations, which are due to the heterogeneity found in exponentially growing cells, the differences in YM values are highly significant (p < 0.0001). This quantitative analysis also showed that the presence of uracil resulted in an approximately two-fold decrease in the rigidity of both wild type and mutant cells (Fig. 5J), which indicates that growth conditions have a slight impact on the nanomechanical properties of cell walls. Overall these nanomechanical data connected the increase in CW thickness in the ∆pyrB mutant that was observed in TEM with the rigidity of the cell wall, i.e. thicker cell walls are more rigid.
Since it has been reported (7) that CW thickening is related to severe growth arrest of L. lactis cells, we further examined the growth characteristics of the ∆pyrB mutant.
Growth defect of the ∆pyrB mutant is linked to cell wall rigidity -As reported previously (25), the ∆pyrB mutant was not able to grow in minimal SA medium without the addition of uracil (results not shown). Therefore we evaluated the growth characteristics of the ∆pyrB mutant in rich M17G medium. In this medium, the mutant exhibited a 5-hour lag phase compared to its parent strain (Fig. 6A). This phenotype was complemented by introducing the positive allele of pyrB, cloned in a plasmid under control of the nisin-inducible promoter. Note that in the medium without nisin, partial complementation was observed, in keeping with reports about the modest leakiness of this promoter (37). Interestingly, the growth delay was also restored by the addition of uracil to the medium, indicating that even in rich medium the ∆pyrB mutant is starved for nucleotides (Fig. 6). We further examined whether the growth impediment of the ∆pyrB mutant was due only to starvation for uracil, or also to changes in the CW, possibly influencing its thickening and/or rigidity and consequently cell division and separation. To this end, we introduced into the ∆pyrB mutant a plasmid that carried a nisin-inducible gene encoding PGH Lc-P40 of Lactobacillus casei (K. Regulski and M.-P. Chapot-Chartier, personal communication). Lc-P40 (Lcabl_00230) is a PG-specific peptidase that hydrolyzes PG at the interpeptide bridges, which are formed by D-Asp in both Lb. casei and L. lactis (2,38). We expected that Lc-P40 would "relax" the overly thick and rigid CW of the ∆pyrB mutant by creating breaks in PG. Indeed, the expression of Lc-P40 in the ∆pyrB mutant markedly increased growth, especially when the nisin inducer was added. It is important to note that the improvement in growth was observed in M17G medium without uracil, indicating that the growth delay of the ∆pyrB mutant was not only due to starvation for uracil.
Following the same reasoning, we mutated the ponA gene to create another "relaxation" factor in the ∆pyrB mutant. The ponA gene encodes PBP1A, an enzyme needed for the formation of PG cross-links, and mutations in it have been reported to favor the appearance of breaks in PG (15). As expected, the ponA mutation reduced the growth lag of the ∆pyrB mutant in M17G medium without uracil. Surprisingly, both "relaxing" genetic backgrounds (PGH-positive and ponAnegative) not only shortened the lag time but also increased the final OD of bacteria in the medium without uracil (Fig. 7).
We further examined how the growth of the ∆pyrB mutant was affected by certain factors that could reduce the number of PG breaks. For this, we introduced a multicopy plasmid that encoded the PG deacetylase gene pgdA in the ∆pyrB mutant. It was previously shown in L. lactis that this plasmid causes increased resistance to PG hydrolysis (6,7), and for this reason we expected that overexpression of the pgdA + allele would increase the growth delay of the ∆pyrB mutant. Indeed, when the ∆pyrB mutant was grown in the presence of uracil, the introduction of multiple copies of pgdA + on a plasmid caused growth retardation (Fig. 8). In the WT strain, introduction of the same plasmid also caused growth delay, but to a lesser extent.
We also examined acmA, the main lactococcal autolysin (14), as another factor that affects the number of breaks in PG (15). Deletion of acmA in the ∆pyrB mutant had a similar effect on growth delay, albeit less pronounced, as introduction of the plasmid carrying pgdA did. Only a slight effect of the ∆acmA mutation was observable in the WT strain (Fig. 8 B). Note that growth retardation caused by pgdA overexpression or by acmA deletion in the ∆pyrB mutant was seen in a medium with uracil, indicating that starvation for uracil is not responsible for the observed effects in the ∆pyrB mutant. In conclusion, the observed results are consistent with the increased CW rigidity of the ∆pyrB mutant compared to WT.
Inactivation of pyrB increases D-Asp/Asn-containing muropeptides in PG and cross-linking -We determined the PG structure of exponentially growing L. lactis strains VES4883 (∆guaA) and VES6497 (∆pyrB) by separation of the constituent muropeptides with RP-HPLC, and compared their muropeptide profiles with that of their parent, MG1363, which was grown to the exponential and early stationary phases. In keeping with the role of pyrB in the regulation of L-Asp availability for PG synthesis, more D-Asp/D-Asn was present in PG of both mutants (Table 3).
In particular, the RP-HPLC profiles of the ∆guaA and ∆pyrB mutants had fewer muropeptides that lacked D-Asp/Asn linked to their stem peptide (2.1% and 1.9%, respectively) relative to the control strain MG1363 in the same growth phase (4.0%). Furthermore, these two strains displayed elevated PG cross-linking (34.1 and 34.7%, respectively, versus 31.7% in WT). The MG1363 cells in early stationary phase (OD 600 1.2) behaved similarly to ∆pyrB or ∆guaA mutants in the exponential growth phase (both at OD 0.2). Among the muropeptides that contained D-Asp/Asn, a relative increase in D-Asn was seen in the CW of ∆pyrB or ∆guaA mutants and MG1363 in the early stationary phase: the Asn/Asp ratio in these strains was 3.1, 4.1, and 3.8, respectively, versus 2.2 for exponentially grown MG1363 (Table 3). Overall, the PG muropeptide analysis confirmed that there was more D-Asp/Asn in the PG of ∆pyrB and ∆guaA mutants, supporting the hypothesis that pyrB plays a pivotal role in the switch between L-Asp utilization for PG and its use in pyrimidine biosynthesis.

DISCUSSION
The bacterial PG sacculus needs to be strong in order for cells to be able to withstand high turgor pressure. Rigid and thick PG is advantageous in the stationary phase, when bacterial cells are exposed to different stress conditions, e.g., exhaustion and acidification of the medium (39). However, during exponential growth, PG must be flexible to allow fast cell division and enlargement. A balance between CW rigidity and flexibility may be maintained through the concerted activities of various enzymes involved in CW synthesis (PBPs) and hydrolysis (PGHs) (22,40). Despite recent achievements in this field, the elucidation of the mechanisms that govern the maintenance of this balance remains a challenge. Notwithstanding extensive studies on bacterial PG modifications in response to PGhydrolysis-induced CW stress (41), little is known about the regulation of these processes under non-stress conditions.
Through the isolation of lysozymeresistant mutants of L. lactis and the identification of the genes responsible for this phenotype, we were able to shed light on the mechanisms that govern the equilibrium between hydrolysis and synthesis of PG in this organism. Of the independent lysozymeresistant isolates we obtained, 14% were guanine auxotrophs, which indicated that guaA was involved in the resistance phenotype. Whole-genome expression studies of the ∆guaA mutant showed decreased expression of the gene that encodes the regulator PyrR. This may be the reason why these mutants exhibited repression of the whole pyr operon (pyrR, pyrP, pyrB and carA) and other genes involved in pyrimidine metabolism (32,42). More studies are required to understand the precise regulatory mechanism that links the downregulation of genes involved in pyrimidine metabolism with the inactivation of guaA.
Since guaA is involved in the synthesis of guanosine monophosphate, and consequently of the global regulatory alarmones (p)ppGpp (43), it might be that the latter participates in the regulatory scheme.
Most importantly, studies of the ∆guaA mutant led us to discover the contributions of pyrB to CW structure and properties. The aspartate carbamoyltransferase PyrB is responsible for the utilization of L-Asp for pyrimidine synthesis. L-Asp is also a precursor for CW synthesis (see Fig. 9): it is converted to D-Asp by racemase RacD, then attached to the stem peptide of the PG precursor by AslA ligase (10) and converted to D-Asn by AsnH (9). Simultaneous utilization of L-Asp for PG and for pyrimidine biosynthesis could provide the means of coordinating CW structure in a manner dependent on pyrimidine biosynthesis, in which PyrB could play a pivotal role. We hypothesize that down-regulation of pyrB expression, as in the ∆guaA mutant, or its absence, as in the case of pyrB deletion, results in more L-Asp, which is transformed into D-Asp and used for the formation of PG crossbridges. This would lead to the observed increase in PG rigidity and possibly to lysozyme resistance. To our knowledge, the data presented here indicate for the first time that a link exists between pyrimidine metabolism and CW plasticity in bacteria.
This hypothesis is supported by the exceptional CW-related features of the ∆pyrB mutant: i) the observed growth defects, which were restored by the introduction of additional mutations designed to weaken the CW (Fig. 6,  Fig. 7, Fig. 8), ii) increased PG cross-linking and D-Asp/D-Asn content (Table 3), iii) elevated lysozyme resistance (Fig. 3), and iv) increased CW thickness (Fig. 4) and rigidity (Fig. 5).
In particular, we observed a longer lagphase and lower final optical density in ∆pyrB mutant cultures than in the control strain MG1363 (Fig. 6). This growth defect was suppressed by introducing CW "relaxing" determinants in the ∆pyrB mutant, such as PGH of L. casei Lc-P40, which introduces breaks in PG interpeptide bridges, or a mutation in the ponA gene (Fig. 7). In addition, we exploited the fact that the growth defect was corrected by uracil to show that the opposite was also true, namely that the introduction of genetic factors that potentially lead to a decrease in PG breaks resulted in growth retardation. Thus, overexpression of the PG deacetylase PgdA, which leads to resistance to autolysis and lysozyme through increased N-deacetylation of PG (6,7) resulted in growth retardation in the ∆pyrB mutant.
Removal of the main lactococcal autolysin AcmA (14) produced similar albeit less pronounced effect.
RP-HPLC analysis revealed that PG of the pyrB-negative strain as well as of the ∆guaA mutant (in which expression of pyrB was drastically decreased), contained fewer muropeptides that lacked D-Asp or D-Asn (1.9% and 2.1%, respectively) than the control strain (4.0%) in the exponential growth phase (Table 3). This was true regardless of whether the medium was supplemented or not with uracil (data not shown). The increased amount of D-Asp and D-Asn corresponds to an increased cross-linking index value (31.7% for WT versus 34.7% and 34.1% for ∆pyrB and ∆guaA mutants, respectively). Remarkably, the D-Asn/D-Asp ratios in PG of the ∆pyrB and the ∆guaA mutants were higher (3.1 and 4.1, respectively) than in the WT strain (2.2). Preferential accumulation of D-Asn in PG could be explained by the simultaneous increase in D-Asp supply in ∆guaA and ∆pyrB mutants, by the activity of AsnH, and/or increased affinity of flippase for disaccharide pentapeptides carrying D-Asn. It has been reported that amidation of D-Asp renders L. lactis more resistant to lysozyme due to a decrease in the net negative charge of the CW (9). Therefore a higher amount of D-Asn in PG could result in a less negative CW charge, leading to elevated resistance to this cationic antimicrobial.
Interestingly, a relatively small increase in PG cross-linking and D-Asn/D-Asp content and ratio corresponded to considerable differences in other phenotypes, such as CW rigidity (Fig. 5), thickness (Fig. 4), growth defects (Fig.7), and lysozyme resistance (Fig.3). This might be explained by assuming that more intensively cross-linked PG is specifically localized, which, if true, would not have been observable in our experiments because of the sample-averaging nature of HPLC analyses (44). Similarly, it has been reported that a relatively moderate shift in PG O-acetylation (from 2.6% to 4.2%) caused complete growth arrest in L. lactis (7).
The pyrB mutant phenotype could in part explain the high lysozyme resistance of the guaA mutant. The results presented here imply that the ∆guaA mutant possesses several regulatory mechanisms that enable it to prevent lysozyme activity, all potentially affecting cell envelope charge and/or PG cross-linking. The first is induction of the dlt operon, which is involved in alanylation of anionic TA (29,45). The consequent decrease in negative charge of the CW increases resistance to the positively charged lysozyme (7). Second, an increase in the D-Asp/Asn supply due to pyrB downregulation could result in elevated PG crosslinking and thus counteract the muramidase activity of lysozyme. Third, the presence of a higher amount of D-Asn in PG also results in a decrease in the net negative charge of the CW, which would increase lysozyme resistance. Moreover, the ponA gene, which encodes PBP 1A (30), was four-fold over-expressed in the lysozyme-resistant ∆guaA mutant. It has been reported that up-regulation of PBP expression is a common strategy of Gram-positive bacteria responding to cell wall stress (41). Despite this, the ponA mutation did not increase the sensitivity of MG1363 to lysozyme activity (Fig. 3), suggesting that ponA over-expression only marginally affects this phenotype. Strikingly, the deletion of only one gene, namely guaA, protected against both types of lysozyme activities, the muramidase enzyme and the cationic antibacterial peptide.
The link described here between pyrB expression and CW rigidity could have an important role in maintaining the balance between CW flexibility and rigidity during the growth of a bacterial community. This function is possible because the expression of pyrimidine metabolism genes is tightly linked to pyrimidine availability in the medium, which in turn depends on the growth phase of the culture (25,32). Differential pyrB expression during culture growth was documented in chrono-transcriptomics studies of L. lactis MG1363 grown in milk (46) or in M17G (47,48). Notably, it was observed that the pyr operon is considerably more expressed during the exponential growth than it is in the stationary phase. If the D-Asp/D-Asn content in PG is dependent on pyrB expression, this would enable the availability of flexible PG during exponential growth and more rigid and thick PG in the stationary phase. In keeping with this reasoning, we observed a higher amount of Asp/Asn-less muropeptides in PG of MG1363 in the exponential phase than we did in the stationary phase. Correspondingly, the cross-linking index of MG1363 PG was lower in the exponential than in the stationary phase, where it was comparable to that of the ∆pyrB mutant (Table 3).
Such "metabolic programming" would enable the maintenance of the needed CW plasticity without the need to transcriptionally control CW-synthesis genes. It could be considered advantageous because the majority of CW-synthesis genes are essential, and it has been suggested that essential genes in bacteria generally are not regulated at the transcriptional level (49). To avoid the transcriptional regulation of genes, encoding autolytic PGHs could also be advantageous, since these are secreted extracellular enzymes and thus their regulation via intracellular transcription could be challenging. In keeping with this interpretation, the lactococcal response to CW hydrolysis proceeds by modulating the target (O-acetylation of PG) rather than by decreasing transcription of the gene that codes for the secreted autolysin AcmA (7). Indeed, no decrease in expression of lactococcal-PGH-encoding genes was observed in response to lysozyme treatment in our transcriptomics study (Table 2).
PG hydrolases are necessary for CW synthesis in growing cells (21,50). The decrease of L-Asp availability in exponential phase, regulated by PyrB, leads to greater PG sensitivity to hydrolysis, which would be beneficial in terms of making PG more apt for incorporation of newly synthesized.
The described hijacking of the regulatory circuit of pyrimidine metabolism, which is tightly regulated by growth requirements, in order to maintain an optimal balance between CW rigidity and flexibility, may also take place in various other Grampositive bacteria with D-Asp/D-Asn in their PG cross-bridges, such as Enterococcus faecium, L. casei, Lactobacillus delbruckei, Lactobacillus brevis, and others (51). Interestingly, this mechanism is not used by L. lactis to respond to CW stress, since pyrB is not down-regulated in cells treated with lysozyme ( Table 2). It is probable that the tight regulation of genes involved in nucleotide metabolism precludes the possibility of changing pyrB expression in response to CW stress.
This work shows that, in order to maintain optimal CW plasticity, bacteria regulate substrate availability (in this case L-Asp) for CW synthesis. The availability of L-Asp for CW synthesis might also be affected by expression of the aspartyl-tRNA synthetase gene aspS. This possibility is supported by the down-regulation of aspS in the stationary growth phase, as observed in chronotranscriptomics assays of L. lactis MG1363 grown in M17G (47,48); this activity could also favor the availability of L-Asp for CW synthesis.
An analogous mechanism to the one described here might be at play in bacteria that have L-amino acids in their PG cross-bridges, such as, e.g., Streptococcus pneumoniae (L-Ala-L-Ser or L-Ala-L-Ala in the cross-bridge). Like the competitive "switch" between PyrB and RacD for L-Asp and the resulting change in CW of L. lactis, in this case the competition between the protein synthesis machinery and tRNA-dependent aminoacyl ligases for aminoacylated-tRNA (52) could be a means of regulating PG plasticity. Ribosomal content is strictly regulated according to nutrient availability and is higher in the exponential phase than in the stationary phase (53). This would raise the probability of L-amino acids being incorporated in the PG cross-bridges of bacteria in stationary phase. The incorporation of amino acids in PG cross-bridges could therefore be viewed as an evolutionary adaptation enabling bacteria to tune PG plasticity to growth requirements.

Bacterial
strains and growth conditions -The bacterial strains and plasmids used in this study are listed in Table 1. All L. lactis strains used here are derivatives of strain MG1363 (54). L. lactis was grown at 30°C in M17 medium (BD Biosciences) that contained 0.5% glucose (M17G). Erythromycin (2.5 μg/ml) and chloramphenicol (5 μg/ml) (Sigma) were added when needed. E. coli was grown in LB medium (BD Biosciences) at 37°C, unless otherwise indicated, in the presence of ampicillin (100 μg/ml), erythromycin (100 μg/ml), or chloramphenicol (10 μg/ml), when needed. Nisin was prepared in Me 2 SO (Sigma) and added at a final concentration of 0.1 ng/ml. Growth was monitored by optical density measurement at 600 nm (OD 600 ) with a spectrophotometer (Spectronic 20 Genesys).
Lysozyme resistance tests -A concentrated hen egg white lysozyme (Fluka, Buchs, Switzerland) solution was freshly prepared in GM17 medium and then diluted in molten GM17 agar (1.5%) at 45°C. Overnight bacterial cultures were successively diluted 10fold and 5 µl of each dilution was spotted on GM17 agar plates supplemented with different concentrations of lysozyme.
Similarly, the guaA deletion was introduced in strain VEL1378 (dltD), creating strain VES5160; and the pyrB deletion in strain VES4883 (∆guaA), yielding VES6530. Also, the ΔacmA mutation was introduced in strain VES6497 by using the plasmid pINTAA and following the procedure described in (14), thus creating strain VES6831.
The ponA pyrB double mutant VES6949 was constructed by transforming strain VES6497 with plasmid pVE1837 [pRV300 carrying a 210-bp internal fragment of ponA (30)] and selecting for erythromycin resistance.
DNA sequence determination of this fragment revealed that it contained the IS905 transposon DNA sequence, flanked by sequences that are situated upstream and downstream of IS905 transposons, indicating that VES2824 is missing the 33893-bp chromosomal fragment situated between the two IS905s.
Peptidoglycan structure analysis -PG was extracted from cultures in the exponential [OD 600 0.4 for MG1363, 0.2 for VES6497 (∆pyrB) and VES4883 (∆guaA)] and early stationary growth phase (OD 600 1.2 for MG1363) as described previously (58). PG was then hydrolyzed with mutanolysin, and the resulting soluble muropeptides were reduced and separated by RP-HPLC with an Agilent UHPLC 1290 system using an ammonium phosphate buffer and linear methanol gradient, as described previously (59). The eluted muropeptides were detected by UV absorbance at 202 nm. Muropeptides were identified according to their retention time by comparison with a reference chromatogram (58). The different muropeptides were quantified by integration of the peaks on the chromatogram. The relative amount of each muropeptide was expressed as the ratio of its peak area over the sum of all of the peak areas.
The PG cross-linking index was calculated as described in (60). DNA microarray analysis -For DNA microarray experiments, L. lactis MG1363 and L. lactis VES4883 (∆guaA) cells were grown in M17G medium and harvested at the midexponential growth phase. In order to assess the effect of the lysozyme treatment on the transcriptome, 5 mg/ml of lysozyme (Sigma-Aldrich, St. Louis, MO) was added to the culture. Cells were collected after 20 min of incubation with lysozyme. Total RNA of lysozyme-treated MG1363 and of VES4883 was compared to RNA isolated from MG1363 (WT) cells. Slides were scanned with a Genepix 4200 laser scanner at 10-µm resolution. ArrayPro 4.5 (Media Cybernetics Inc., Silver Spring, MD) was used to analyze slide images; processing and normalization were performed using MicroPrep software (61,62) as described in (62). Gene expression was considered to be significantly altered when the Cyber T Baysian P-value was ≤0.001.
Transmission electron microscopy (TEM) -Pellets of bacteria (OD 600 0.5 for MG1363, MG1363 + uracil, and ∆pyrB mutant + uracil, and OD 600 0.2 for the ∆pyrB mutant) were fixed with 2% glutaraldehyde in 0.1 M Na cacodylate buffer pH 7.2, for 3 hours at RT. Samples were contrasted with Oolong Tea Extract 0.5% in cacodylate buffer, fixed with 1% osmium tetroxide that contained 1.5% potassium cyanoferrate, gradually dehydrated in ethanol (30% to 100%), substituted gradually in a mixture of propylene oxide-epon, and embedded in Epon (Delta microscopie, Labège, France). Thin (70-nm) sections were collected onto 200 mesh copper grids, and counterstained with lead citrate. Grids were examined using a Hitachi HT7700 electron microscope operated at 80kV (Elexience, France). Images were acquired with a charge-coupled device camera (AMT, Japan). Cell wall thickness was measured on TEM micrographs of at least 3 cells at magnification 70000x, taking at least 5 measurements on each cell.
Sample preparation and atomic force microscopy (AFM) experiments -L. lactis strains MG1363 and its isogenic ∆pyrB mutant were grown in M17G broth in the presence or absence of 100 µg/ml uracil at 30°C under static conditions, until the exponential phase was reached (OD 600 0.5 for MG1363, MG1363 + uracil, and ∆pyrB mutant + uracil, and OD 600 0.2 for the ∆pyrB mutant). Bacterial cells were concentrated by centrifugation, washed twice and resuspended in phosphate-buffered saline (PBS), and immobilized on PDMS stamps prepared as described previously (35). Briefly, microstructured PDMS stamps were covered by a total of 100 µl of the cell suspension. Cells were then deposited into the microstructures of the stamps by convective/capillary assembly.
Images were recorded in PBS in Quantitative Imaging TM (QI TM ) mode (33) using MSCT AUWH (Bruker, Billerica, USA) cantilevers (nominal spring constants of 0.1 and 0.01 N/m), and with an applied force of 0.5 nN. Force spectroscopy mode was used to perform local nanoindentation measurements on areas of 200 × 200 µm² on top of cells. The applied force was kept between 0.5 and 2 nN depending on the strains and growth conditions probed. The cantilevers' spring constant was determined using the thermal noise method (63). For imaging and force spectroscopy we used a Nanowizard III (JPK Instruments, Berlin, Germany). For rigidity measurements, the force distance curves obtained in QI TM mode and during nanoindentation experiments were transformed into force-indentation curves by subtracting the cantilever deflection on a solid surface. The indentation curves were then fitted to the Hertz model, which links the force (F) as a function of the Young's modulus value (E) with the square of the indentation (δ) for a conical indenter according to the following equation: F = [2E tanα/π(1-ν²)]δ², where α is the tip opening angle (17.5°) and ν the Poisson ratio, assumed to be 0.5. In each condition, the fitted indentation segment was kept constant at 50 nm. In each case, Young's modulus values were measured on 12 cells (n = 12288 curves per conditions), and Young's modulus medians were calculated from fits in a Gaussian model. All results were analyzed using the data processing software provided by JPK Instruments.  Comparison of lysozyme resistance of L. lactis WT MG1363 and its mutants VES2824, in which a large region that included the guaA gene was deleted, and VES 4883, in which only the guaA gene was deleted. Serially diluted cultures were grown on M17G agar plates supplemented with lysozyme and guanine.   Gaussian model. The three asterisks show significant differences between the rigidity of strain MG1363 and that of the ∆pyrB mutant at a p-value < 0.0001 (unpaired t-test).      Genes of VES4883 (∆guaA) and of lysozyme-treated MG1363 that were up-and downregulated compared with those in MG1363 (WT). Only genes that were related to CW, involved in purine and pyrimidine metabolism, or whose expression was affected by more than 1.5-fold are indicated.