Seven-transmembrane receptor protein RgsP and cell wall-binding protein RgsM promote unipolar growth in Rhizobiales

Members of the Rhizobiales (class of α-proteobacteria) display zonal peptidoglycan cell wall growth at one cell pole, contrasting with the dispersed mode of cell wall growth along the sidewalls of many other rod-shaped bacteria. Here we show that the seven-transmembrane receptor (7TMR) protein RgsP (SMc00074), together with the putative membrane-anchored peptidoglycan metallopeptidase RgsM (SMc02432), have key roles in unipolar peptidoglycan formation during growth and at mid-cell during cell division in Sinorhizobium meliloti. RgsP is composed of a periplasmic globular 7TMR-DISMED2 domain, a membrane-spanning region, and cytoplasmic PAS, GGDEF and EAL domains. The EAL domain confers phosphodiesterase activity towards the second messenger cyclic di-GMP, a key regulatory player in the transition between bacterial lifestyles. RgsP and RgsM localize to sites of zonal cell wall synthesis at the new cell pole and cell divison site, suggesting a role in cell wall biogenesis. The two proteins are essential for cell wall biogenesis and cell growth. Cells depleted of RgsP or RgsM had an altered muropeptide composition and RgsM binds to peptidoglycan. RgsP and RgsM orthologs are functional when interchanged between α-rhizobial species pointing to a conserved mechanism for cell wall biogenesis/remodeling within the Rhizobiales. Overall, our findings suggest that RgsP and RgsM contribute to the regulation of unipolar cell wall biogenesis in α-rhizobia.


Introduction
Rod-shaped bacteria have evolved diverse modes of cell growth.In Bacillus subtilis, Escherichia coli and Caulobacter crescentus, cells grow by elongating along the lateral sidewall, incorporating peptidoglycan (PG) cell wall material in a dispersed pattern along the sidewall [1].Other bacteria elongate by zonal growth in which PG is incorporated at one or both cell poles [2].Unipolar growth is observed in α-proteobacterial Rhizobiales, such as Agrobacterium tumefaciens and Sinorhizobium meliloti [3], while Gram-positive Actinomycetales including Mycobacterium tuberculosis grow bipolarly [4].Later in the cell cycle, all species switch PG synthesis completely or partially to mid-cell for cell division [1].During PG synthesis nascent material is incorporated into the pre-existing PG structure (sacculus), by the activity of hydrolases which selectively cleave bonds in the stress-bearing sacculus [5].Because the integrity of the PG sacculus is essential for maintaining cell shape and resisting turgor [6], cell growth and PG synthesis require precise spatial and temporal regulation of the incorporation of new material into the PG sacculus [7].Rod-shaped bacteria with a dispersed mode of PG incorporation along the lateral cell wall utilize the cytoplasmic actin-like MreB protein to direct PG synthesis.Dynamic filaments or patches of MreB are believed to serve as platforms for the intracellular and extracellular PG synthesis machineries [8,9].By contrast, most rod-shaped bacteria with polar growth do not contain MreB homologs [10] and it is currently unknown how polar cell wall elongation is regulated in these bacteria.
Bis-(3 0 -5 0 )-cyclic dimeric guanosine monophosphate (c-di-GMP) has a central role in the regulation of motility, adherence, biofilm formation and virulence and in several bacterial species, also for promoting cell cycle progression, growth and development [11][12][13][14][15]. c-di-GMP is synthesized by diguanylate cyclases (DGCs) with a conserved GGDEF domain and degraded by phosphodiesterases (PDEs) with either an EAL domain or a HD-GYP domain [16].In the α-proteobacterium C. crescentus, the spatial organization of proteins involved in c-di-GMP metabolism contributes to cell polarity and cell cycle progression [14], and we showed previously that strong overproduction of this second messenger inhibited growth and resulted in cell filamentation in S. meliloti [17].
Seven-transmembrane receptors (7TMRs) form the largest, most ubiquitous and most versatile family of membrane receptors.In eukaryotes, they are involved in signaling via interaction with cytoplasmic G-proteins [18].A distinct class of bacterial 7TMRs consists of so-called 7TMR-DISMs, which stands for 7TMR with diverse intracellular signaling modules [19].7TMR-DISM proteins contain seven transmembrane α-helices (7TMR-DISM_7TM domain) fused to various cytoplasmic signaling and/or extracytoplasmic 7TMR-DISMED1 or 7TMR-DISMED2 domains [19].These extracytoplasmic domains have been predicted to bind ligands such as carbohydrates [19].Common cytoplasmic signaling modules in 7TMR-DISM proteins include histidine protein kinase domains, GGDEF and EAL domains involved in cdi-GMP homeostasis and sensory Per-Arnt-Sim (PAS) domains.PAS domains are known to sense small molecules, ions, gases, light or redox state [20].With up to 14 paralogs per genome, for example in the spirochete Leptospira interrogans, 7TMR-DISMs are widely distributed in both Gram-negative and Gram-positive bacteria [19].However, only a few of these proteins have been functionally characterized.These include the Pseudomonas aeruginosa 7TMR-DISM histidine kinases RetS and LadS, which are involved in regulation of biofilm formation, and the GGDEF domain containing protein NicD, which is involved in biofilm dispersal [21,22].
The 7TMR-DISM protein SMc00074 (renamed RgsP for rhizobial growth and septation cdi-GMP phosphodiesterase) was previously shown to be essential in S. meliloti [17,23,24].Here, we provide evidence that RgsP is important for PG synthesis in certain α-rhizobial species.We identify SMc02432 (a putative membrane-anchored periplasmic PG metallopeptidase, renamed RgsM for rhizobial growth and septation metallopeptidase) as a RgsP interaction partner and show that both are required for unipolar cell growth in S. meliloti and related Rhizobiales.RgsP was also found to be an active PDE, substantially contributing to c-di-GMP homeostasis and hence possibly connecting c-di-GMP signaling with spatiotemporal control of PG synthesis.

RgsP is required for normal growth and cell morphology
RgsP is composed of 7TMR-DISMED2, 7TMR-DISM_7TM, PAS, GGDEF and EAL domains.Previous systematic mutagenesis of c-di-GMP-related genes identified rgsP (SMc00074) as a potentially essential gene in S. meliloti Rm2011 [17].C-terminally 3×FLAG-tagged RgsP accumulated in growing cells and was only detected at very low levels in stationary phase cells (S1A Fig) , suggesting a role during cell growth.To further study the function of this protein, we constructed a RgsP depletion strain (Rm2011 rgsP dpl ) by placing the native rgsP gene under the control of an IPTG-inducible promoter.Using the same promoter, we confirmed depletion of a C-terminally 3×FLAG-tagged RgsP variant in the absence of IPTG (S1B Fig).
Growth of Rm2011 rgsP dpl was strongly dependent on IPTG (Fig 1A).Cells cultured in the presence of IPTG showed wild type-like growth and morphology, whereas RgsP-depleted cells lost the rod shape and had an irregular, uneven cell surface (Fig 1B and 1C; S2 and S3A Figs).4.7% of these cells were stained by the dead cell indicator propidium iodide (S4A Fig) .This suggests that the physiological effect of RgsP depletion probably is mostly bacteriostatic after 24 h of incubation in absence of IPTG.To estimate the capability of RgsP-depleted cells to resume growth, we analyzed colony formation and microscopically monitored growth on TY medium containing IPTG.In cultures of Rm2011 rgsP dpl grown without added IPTG for 12 or 24 h, the number of colony forming units (CFU ml -1 OD 600 -1 ) was reduced by 67.8% or 97.4%, respectively, compared to IPTG-supplemented cultures (S4B Fig) .Following the fate of RgsPdepleted cells (previously grown in the absence of IPTG for 24 h), 25.8% were able to recover and to give rise to microcolonies on TY agarose pads with IPTG.In contrast, only 3.0% of the cells were able to divide at least once during the 12 h observation period on pads lacking IPTG (S4C Fig) .Thus, RgsP depletion predominantly resulted in growth arrest, which was relieved upon IPTG addition in a minor fraction of the cells only.However, when interpreting these results, it has to be taken into account that the cells grown under depletion conditions may contain some residual amounts of RgsP.
Zones of PG synthesis were visualized by HADA (7-hydroxycoumarin-3-carboxylic acid-D-alanine)-labeling [25].In IPTG-supplemented cultures with induced rgsP expression, 98% of the cells were stained at one of the cell poles and/or septal region.By contrast, only 33% of the RgsP-depleted cells incorporated HADA (Fig 1D

RgsP localizes dynamically during the cell cycle
To determine the subcellular localization of RgsP, we replaced rgsP with rgsP-egfp at its native genomic location.In exponentially growing S. meliloti cells, RgsP-EGFP co-localized with HADA-labeled sites of PG synthesis at one of the cell poles and the division site (Fig 2A and  2B).Time-lapse microscopy showed that RgsP-EGFP remained at the pole during the entire phase of cell elongation until it relocated to the mid-cell region (Fig 2C).We next localized RgsP-EGFP simultaneously with ParB-mCherry.ParB binds to the parS sites close to the chromosomal origin of replication [26].Early in the cell cycle, a single fluorescent ParB-mCherry focus localized at the old cell pole and RgsP-EGFP localized at the opposite, new pole (Fig 2C).Later, a second ParB-mCherry focus moved to the new pole.Following a period of co-localization (~90 minutes) of RgsP-EGFP and ParB-mCherry at the new pole, the polar RgsP-EGFP signal disappeared and appeared at mid-cell.Thus, RgsP-EGFP localizes at the new cell pole early in the cell cycle and later at the cell division site.Importantly, these are the sites of PG synthesis.
To relate the temporal pattern of RgsP relocation to the mid-cell to known pre-divisional processes, we co-localized RgsP with the early divisome component FtsZ [7].mCherry-FtsZ showed diffuse fluorescence in growing cells, and localized to the mid-cell in pre-divisional cells (Fig 2A).Septal localization of RgsP-EGFP always coincided with the presence of a mCherry-FtsZ focus at mid-cell, but not vice versa (Fig 2A and 2B), suggesting that RgsP accumulates at the septal site later than FtsZ.Time-lapse microscopy showed localization of RgsP-EGFP at mid-cell about 24 minutes after occurrence of the mCherry-FtsZ focus.This RgsP-EGFP relocalization was immediately followed by the onset of septum constriction, which was completed by cell division about 24 minutes later (Fig 2D).This implies that RgsP may also have a function in septation.
The dynamics of RgsP relocation to the mid-cell region was analyzed at a higher time resolution relative to PleD.We found previously that the DGC PleD localizes to the new cell pole within 20 minutes before cell division [17].Simultaneous tracking of PleD-EGFP and RgsP-mCherry showed that the accumulation of PleD-EGFP at the growing cell pole temporally correlated with the RgsP-mCherry signal fading away at the pole and relocating to the division site (S6   3D).
Point mutations targeting the conserved inhibitory site (I site) and the DGC active site motifs in the GGDEF domain (RxxD to AxxA and GGDQF to GAAAF) and the PDE active site in the EAL domain (EAL to AAL), or removal of the EAL domain did not significantly affect protein localization or complementation of the cell growth and morphology defects (Fig 3B , 3C and 3D).RgsP and RgsP-EGFP variants lacking the GGDEF or both GGDEF and EAL domains were unable to fully complement growth and morphology defects unless they were expressed at higher levels from P syn (Fig 3B and 3C).Nevertheless, the EGFP-tagged versions showed normal cellular localization (Fig 3D).When gene expression was driven by P Ã rgsP on pABCS2-mob, levels of corresponding 3×FLAG-tagged wild type protein and variants were similar, except for RgsP ΔGGDEF , which was detected at 67% of the corresponding wild type protein level (S1C Fig) .This implies that RgsP lacking the EAL domain fully supports cell growth, whereas lack of the GGDEF domain impairs protein stability or function.
To investigate the possible role of c-di-GMP in RgsP function, we tested complementation of RgsP depletion by rgsP and rgsP-egfp variants in the c-di-GMP-deficient strain Rm2011 ΔXVI, which lacks all genes predicted to encode active DGCs and in which the level of c-di-GMP was below the detection limit [17].The complementation properties of each of these RgsP versions were indistinguishable in strains with or without c-di-GMP, and RgsP-EGFP showed normal localization in Rm2011 ΔXVI (S8 Fig) .This suggests that c-di-GMP is not required for RgsP localization and the essential function of this protein in cell growth.
Taken together, these data suggest that the N-terminal part of RgsP determines the essentiality of this protein, whereas the EAL domain is dispensable for the growth-promoting function of RgsP.

RgsP is an active c-di-GMP phosphodiesterase
Since the GGDEF and EAL domains may have a regulatory role, we analyzed the enzymatic activities and ability to bind c-di-GMP in vitro.A purified His 6 -tagged variant of RgsP containing the PAS, GGDEF and EAL domains (His 6 -RgsP PAS-GGDEF-EAL ) hydrolyzed [α-32 P]-c-di-GMP, whereas no cleavage product was detected with the PDE active site mutant variant His 6 -RgsP PAS-GGDEF-EAL -AAL (S9A Fig) .In a DGC activity assay with [α-32 P]-GTP as a substrate, His 6 -RgsP PAS-GGDEF-EAL -AAL did not synthesize c-di-GMP, in contrast to C. crescentus DgcA used as a positive control (S9B Fig).This is in agreement with the degenerate active site GGDQF in the RgsP GGDEF domain and our previous in vitro DGC activity assay with a RgsP fragment comprising the PAS, GGDEF and EAL (active site intact) domains [17].Since an intact I site RxxD is present in the GGDEF domain of RgsP, we assayed for the ability of RgsP to bind c-di-GMP in a differential radial capillary action of ligand assay (DRaCALA) using a His 6 -tagged RgsP variant containing only the PAS and GGDEF domains (His 6 -RgsP PAS-GGDEF ).In this assay, the positive control His 6 -DmxB from Myxococcus xanthus produced the characteristic DRaCALA pattern, whereas His 6 -RgsP PAS-GGDEF was not able to prevent the diffusion of [α-32 P]-c-di-GMP (S9C Fig) .Thus, RgsP PAS-GGDEF-EAL has c-di-GMP PDE activity, presumably catalysed by the EAL domain, but the GGDEF domain does not bind or synthesize c-di-GMP.
To evaluate c-di-GMP PDE activity of RgsP in vivo, we determined the c-di-GMP content of RgsP-depleted Rm2011 rgsP dpl complemented with rgsP wt , rgsP AAL and rgsP GAAAF expressed from P Ã rgsP .The c-di-GMP content of cells expressing rgsP wt and rgsP GAAAF was similar, whereas expression of the PDE active site mutant variant rgsP AAL resulted in a two-fold increase in c-di-GMP content (S9D Fig).This data is consistent with the in vitro data and provides evidence that RgsP substantially contributes to c-di-GMP degradation in vivo.

RgsP interacts with the putative periplasmic metallopeptidase RgsM
To identify protein interaction partners of RgsP, we performed co-immunoprecipitation (Co-IP) experiments with a C-terminally 3×FLAG-tagged variant of RgsP, encoded by rgsP-3×flag replacing the native rgsP at its chromosomal location.27 RgsP interaction partner candidates were identified (S1 Table ).Among these, the hypothetical transmembrane protein RgsM (SMc02432) was most abundant and identified with the highest number of unique peptides.Co-IP with RgsM-3×FLAG resulted in identification of RgsP, further supporting an interaction between the two proteins (S2 Table ).
Analysis of RgsM amino acid sequence with transmembrane topology and signal peptide prediction tool PHOBIUS [27] suggested cytoplasmic localization of the amino acids 1-31, a short hydrophobic transmembrane α-helical region (amino acids 32-57) and periplasmic localization of the remaining C-terminal portion (Fig 4A).Amino acids 508-606 of RgsM represent a conserved peptidase M23 domain (pfam01551) often referred to as a LytM domain.This domain is predicted to have PG endopeptidase activity and is characteristic for zincdependent metallopeptidases.The LytM domain of RgsM contains a conserved HxxxD motif (S3 Table ), which is required for zinc ion coordination and hydrolysis of glycine-glycine bonds in the peptides of staphylococcal PG by Staphylococcus aureus LytM [28][29][30].The remaining RgsM amino acid sequence did not provide any hint about its possible function.To verify the predicted membrane topology of RgsM, we fused this protein to a truncated E. coli alkaline phosphatase PhoA, which is only active in the periplasm, and is missing its own signal peptide.The phosphatase detection assay revealed that PhoA, following RgsM 1-66 , indeed localized to the periplasm in both S. meliloti and E. coli.This strongly suggests the periplasmic localization of RgsM 60-646 (S10 Fig).
Interaction between RgsP and RgsM was verified by a bacterial two-hybrid assay [31].In this assay, putative interaction partners, fused to Bordetella pertussis adenylate cyclase fragments T18 and T25, are produced in an E. coli adenylate cyclase-deficient strain.Interaction between the two fusion proteins results in reconstitution of a functional enzyme, which activates expression of the lacZ reporter gene.Simultaneous production of T18-RgsM and RgsP ΔGGDEFΔEAL -T25 fusion proteins, as well as of T18-RgsM and T25-RgsM, resulted in increased β-galactosidase activity, indicating protein-protein interactions (Fig 4B).Thus, RgsM was able to interact with RgsP and to homodimerize in a heterologous host.This is consistent with the Co-IP data that suggested RgsP-RgsM interaction in S. meliloti.

RgsM is essential and co-localizes with RgsP
Attempts to generate a rgsM knockout mutant failed, suggesting that rgsM is essential.Similar to the growth phase-dependent accumulation of RgsP, C-terminally 3×FLAG-tagged RgsM was only detected in growing cells but not in stationary phase cells (S1A

Overexpression of rgsM compromises cell wall integrity
To further characterize the role of RgsM, we analyzed effects of rgsM overexpression.When grown in TY medium, RgsM-overproducing cells were indistinguishable from the empty vector control (S11 Since the most prominent difference between the composition of TY and LB media is the content of NaCl (86 mM in LB, none in TY) and CaCl 2 (2.7 mM in TY, none in LB), we analyzed the effects of these salts on the growth of RgsM-overproducing cells.Increasing the NaCl concentration to 300 mM in LB alleviated the rgsM overexpression-associated morphology and growth defects, and addition of CaCl 2 to 2.7 mM resulted in wild type-like growth (S12 Fig) .Replacement of Zn 2+ with Ca 2+ in RgsM may inactivate the protein and thereby mitigate the effect of rgsM overexpression.However, this explanation for the effect of CaCl 2 is unlikely since the presence of Ca 2+ in the growth medium did not negatively affect growth of S. meliloti, as RgsM depeletion did.Thus, the growth defect resulting from rgsM overexpression in LB may be a combined effect of an artificial increase in RgsM abundance and outer membrane destabilization in the absence of calcium [32].This was probably counteracted by elevated medium osmolarity, which is supposed to reduce turgor.
To analyze the putative RgsM metallopeptidase active site H 510 xxxD, we constructed the RgsM H510A variant.The H510A mutation mitigated the strong cell morphology defect caused by RgsM overproduction in LB.Notably, both in TY and LB, growth of the rgsM H510A overexpressing cells was inhibited and morphology was altered similar to RgsM-depleted cells (S11 Overexpression of rgsM, but not rgsM H510A , compromised the cell envelope of E. coli and resulted in cell lysis in LB lacking NaCl.This was detected using the β-galactosidase substrate chlorophenol red-β-D-galactopyranoside (CPRG) as an indicator of cell lysis and increased membrane permeability [33]

RgsP and RgsM affect PG composition
To investigate the role of RgsP and RgsM in cell wall biogenesis, we determined the muropeptide composition of PG isolated from S. meliloti Rm2011 depleted and non-depleted of RgsP  ).
Rm2011 cells, overproducing RgsM in LB, accumulated more Penta and TetraPenta muropeptides compared to cells harboring the empty vector control, whereas overproduction of RgsM H510A resulted in similar but less pronounced changes in the muropeptide profiles (S16A Fig) .The accumulation of pentapeptides upon either depletion or overproduction of RgsP and RgsM points to altered incorporation and/or processing of new material into the growing PG sacculus, and the increase in 3-3 cross-links suggests increased LD-transpeptidase activity.Both might be indirect effects in cells with impaired PG growth, in response to stress.Muropeptides of E. coli S17-1 overproducing RgsM or RgsM H510A and grown in LB remained unchanged suggesting that the phenotypic changes of E. coli cells were independent of a putative PG hydrolase activity of RgsM (S16B Fig).
To gain further mechanistic insights into RgsP and RgsM functions in S. meliloti, purified His 6 -tagged variants of both proteins were assayed for PG binding in vitro.Whereas It cannot be excluded that the assay conditions were not optimal or that an additional factor is required to activate RgsM.Overall, the effects of RgsM and RgsP depletion or overproduction on the muropeptide profiles and strong PG binding by RgsM further support involvement of both proteins in PG biogenesis.

RgsP and RgsM functions are conserved in α-rhizobial species
Comparative protein sequence analysis using BLASTP revealed conservation of RgsP and RgsM in Rhizobiales, Rhodobacterales, and in a single γ-proteobacterial species, whereas no homologs were detected in the remaining eubacterial phyla (S3 Table ).Thus, we asked whether the homologous proteins from other species were also functionally conserved.Rhizobium etli and A. tumefaciens rgsP homologs RHE_CH00976 (rgsP Re ) and Atu0784 (rgsP At ) were translationally fused to egfp at their native genomic locations.These tagged proteins displayed polar and septal localization corresponding to the HADA-staining zones in R. etli and A. tumefaciens, similar to the labeled RgsP variant in S. meliloti (S18A Fig; Fig 2).We also tested the ability of A. tumefaciens and R. etli rgsP and rgsM homologs, ectopically expressed from a taurine-inducible promoter, to complement S. meliloti RgsP and RgsM depletion strains.Without promoter induction ('leaky expression'), rgsP Re , rgsP At and rgsM Re fully complemented the growth and morphology defects of the respective S. meliloti depletion strains, similar to the S. meliloti homologs, whereas induction with taurine was required for complementation of RgsM depletion with rgsM At (S18B and S18C Fig) .Similar subcellular localization of RgsP homologs in S. meliloti, R. etli and A. tumefaciens, and cross-complementation of rgsP and rgsM between these species provide evidence for functional conservation of both proteins in the Rhizobiales.

Discussion
Members of the Rhizobiales, including S. meliloti, show unipolar cell growth [3].However, the molecular mechanisms governing polar growth of the PG sacculus are largely unknown in these bacteria.In this study, we provide evidence suggesting important roles for RgsP (a member of the seven-transmembrane receptor family 7TMR-DISM) and its interaction partner RgsM (a putative PG metallopeptidase) in polar cell wall growth.
Proteins associated with the polar PG growth zones in A. tumefaciens include the division scaffold proteins FtsZ and FtsA, PG synthase PBP1a and LD-transpeptidase Atu0845 [34,35], which are directly or indirectly involved in PG biosynthesis.RgsP does not show homology to known cell division or PG biosynthesis proteins; however, it localizes to sites of zonal cell wall synthesis in S. meliloti, R. etli and A. tumefaciens.This feature is likely to be conserved in α-rhizobia containing a RgsP homolog.The phenotypes caused by RgsP depletion in S. meliloti − i.e. growth inhibition, altered cell shape, reduced incorporation of HADA and accumulation of penta-muropeptides in the PG sacculus − imply a regulatory role of RgsP in PG biosynthesis.Although known protein components of the cell elongation and division machineries have not been detected in the RgsP and RgsM pull-down assays, indirect, transient or low affinity interactions of RgsP or RgsM with such proteins may have escaped this analysis.
In E. coli, pentapeptides are specific for newly synthesized PG and are quickly processed as PG matures [36] by trans-, endo-and carboxypeptidase reactions [7].In agreement with a previous report [3], we did not detect pronounced pentapeptide peaks in the S. meliloti wild type muropeptide profile.Perhaps in the wild type these pentapeptides were quickly processed but accumulated in RgsP-depleted cells due to impaired PG maturation.In E. coli, lack of the PG carboxypeptidase PBP5 results in an increase in the PG pentapeptide content; an effect which is further augmented by eliminating endopeptidases PBP4 and PBP7 [37].Along with accumulation of monomeric pentapeptides, tetrapeptide dimer and trimer levels decreased upon RgsP depletion, indicating impaired PG cross-linking or enhanced hydrolysis of the peptide bridges.The cell shape can be influenced by the degree of PG cross-linking [38][39][40][41][42][43][44].Hence, the impaired growth and altered cell shape of RgsP-depleted cells might have resulted from dysregulation of PG remodeling.
Cell wall biogenesis by necessity involves incorporation of new PG material into the existing PG sacculus, and this requires local hydrolysis of peptide bridges by PG endopeptidases.These include proteins with a M23 metallopeptidase (or LytM) domain.In E. coli and Vibrio cholerae, single genetic knockdowns of LytM domain proteins were not lethal because of genetic redundancy [45,46].By contrast, the LytM domain protein RgsM is essential in S. meliloti, which is in agreement with a previous report [24].The A. tumefaciens RgsM and RgsP orthologues were also shown to be essential [47].Although we were not able to detect RgsM PG hydrolase activity in vitro, several findings are in agreement with an enzymatic activity in vivo.The RgsM LytM domain contains conserved histidine residues, which have been found to be essential for PG hydrolase activity [28,30,48].Moreover, rgsM overexpression in S. meliloti destabilized the cell envelope and caused perturbations in the muropeptide composition.However, expression of rgsM in E. coli resulted in cell lysis but did not cause changes in the muropeptide profile.Structural studies of S. aureus LytM and Neisseria meningitidis LytM domain protein NMB0 315 suggested the full-length proteins to adopt conformations interfering with enzymatic function, implying activation by proteolysis or protein-protein interactions in vivo [28,49].Thus, we speculate that in vivo, RgsM may hydrolyze PG once activated by a yet-unknown factor.Alternatively, RgsM may have a regulatory role similar to E. coli and C. crescentus LytM domain proteins, which activate amidases to cleave septal PG [50][51][52][53][54].Some LytM domain proteins also contain PG-binding domains [53,55].PG binding by RgsM, demonstrated in vitro, is in agreement with an enzymatic or regulatory role of this protein.
Based on the mutual pull-down of RgsP and RgsM, an interaction in a bacterial two-hydrid assay and similar phenotypes of RgsM-and RgsP-depleted cells, we suggest that both proteins are involved in the same regulatory pathway.We narrowed down the essential part of RgsP to the N-terminal section including the 7TMR-DISM and PAS domains, which both may have a sensory function.In bacteria, 7TMR-DISMs are best understood in P. aeruginosa.In this bacterium, ligand binding to 7TMR-DISMED2 domains and their homodimerization were described as regulatory cues [21,22,56,57].We speculate that RgsP either homodimerizes or, once triggered by an unknown cue, interacts with RgsM, to modulate RgsM dimerization and activity.This model is in agreement with the similar phenotypes of RgsP-and RgsM-depleted cells and the dominant effect of RgsM H510A overproduction in S. meliloti.This enzymatically inactive protein variant may compete with native RgsM for the interaction sites on RgsP or other interacting proteins.
Although abundance of RgsP lacking the PAS domain was only moderately reduced, the localization and growth-promoting function of this protein was dramatically affected.A role of PAS domains or PAS-like motifs in polar protein localization was previously reported for example in C. crescentus.These include the C. crescentus single PAS domain protein MopJ, which directly interacts with the polarly localized histidine protein kinases DivJ and CckA, which are both involved in cell cycle regulation [58,59].We speculate that the PAS domain may be important for recruitment of RgsP to PG biosynthesis sites.The nature of the signal, perceived by the RgsP PAS domain, and its regulatory output remain to be investigated.
RgsP PDE activity substantially contributes to c-di-GMP turnover in S. meliloti.To our knowledge, RgsP is the only c-di-GMP PDE which is polarly localized in S. meliloti.Polarly localized c-di-GMP PDEs in C. crescentus and P. aeruginosa contribute to heterogeneity in the cellular c-di-GMP content [60][61][62].Lack of the RgsP PDE activity did not cause apparent phenotypic defects.Yet, we cannot exclude that RgsP PDE activity has a regulatory function in S. meliloti, which may be linked to its localization.A regulatory function may have escaped our phenotypic analyses or may not have been detected because of compensation by any of the twelve additional c-di-GMP PDEs encoded by the S. meliloti genome [17].
Whereas the RgsP Sm GGDEF domain seems to be inactive because of a degenerate active site [17], RgsP At contains an intact GGDEF motif, suggesting enzymatic activity of this protein.Since RgsP At complemented depletion of RgsP Sm in S. meliloti, the DGC activity of RgsP is unlikely to be relevant for its growth-promoting function.However, the GGDEF domain may modulate PDE activity of the EAL domain, as for example in case of C. crescentus PdeA or P. aeruginosa BifA and RmcA [61,63,64].
Overall, our study suggests that compared to well-studied γ-proteobacterial models, α-rhizobia utilize different sets of proteins for PG metabolism during cell elongation and division, and for relocating the PG growth machinery from a pole to the cell division site.Identification of further enzymes involved in PG synthesis and remodeling, and understanding the regulatory roles of RgsP and RgsM will lead to the clarification of specific mechanisms for polar PG biogenesis in α-rhizobia.
Growth assays were performed using 100 μl cultures in flat-bottom 96-well plates (Greiner), grown at 30 o C with shaking at 1,200 rpm.Three to six culture replicates were analyzed per strain.Optical density (OD 600 ) was recorded using Infinite M200 PRO fluorescence reader (Tecan).For growth assays involving protein depletion, cultures with or without IPTG were inoculated with 0.15 μl of stationary TY preculture with IPTG and grown for 24 h.Relative growth was calculated as a ratio of OD 600 of cells grown without IPTG and OD 600 of cultures supplemented with 0.5 mM IPTG.For growth assays, involving taurine-or IPTG-induced gene overexpression, the cultures containing or not containing taurine or IPTG were inoculated with 0.15 μl of stationary TY preculture and the growth was recorded at the indicated time points.

Construction of strains and plasmids
Constructs used in this work were generated using standard cloning techniques and are listed in S5 Table .The primers used are listed in S6 Table .All constructs were verified by sequencing.Plasmids were transferred to S. meliloti by E. coli S17-1-mediated conjugation as previously described [68].Electroporation was used to introduce plasmids to R. etli and A. tumefaciens following the protocol previously described [69].
To generate chromosomally integrated constructs encoding RgsP or RgsM with C-terminally fused enhanced green fluorescent protein (EGFP) or triple FLAG-tag (3×FLAG), the 700 to 800 bp 3' portion of the gene excluding the stop codon was cloned into suicide plasmids pK18mob2-egfp or pG18mob-3×flag yielding translational fusions of the C-terminal portion of the protein coding sequence to the corresponding tag.Integration of these gene fusion constructs into the S. meliloti genome by homologous recombination resulted in a replacement of the native gene copy with egfp-or 3xflag-tagged gene copy at the corresponding native chromosomal location.
To construct markerless translational fusions of rgsP and rgsM to egfp or mCherry at the native chromosomal location, the 700-800 bp 3' portion of the gene fused to egfp or mCherry was cloned into suicide plasmid pK18mobsacB [70] together with 700-800 bp of adjacent downstream genomic region.The resulting constructs were introduced into S. meliloti and transconjugants were subjected to sucrose selection to obtain double recombinants that have lost the integrated vector [70].Correct positions of chromosomally encoded gene fusions were verified by PCR.
To obtain RgsP and RgsM depletion strains, plasmids designed to uncouple the native promoter from the coding sequence and to place the coding sequence under the control of IPTGinducible promoters were constructed.The rgsP gene is most likely co-transcribed with the preceding rimJ gene, encoding a probable ribosomal-protein-alanine acetyltransferase (S7A Fig) .To uncouple transcription of rimJ and rgsP, and to place rgsP under the control of an IPTG-inducible promoter, the lac-T5 tandem promoter sequence was inserted between rgsP and rimJ without altering the rimJ open reading frame.To this end, a DNA fragment containing the IPTG-inducible T5 promoter and a Shine-Dalgarno sequence followed by the partial 5' rgsP coding sequence starting from the start codon (586 bp) was PCR-amplified from rgsP expression plasmid pWBT-SMc00074 [17].This fragment was inserted into pK18mob2 [70] downstream of the IPTG-inducible lac promoter, thus generating a lac-T5 tandem promoter.In the case of RgsM depletion, the partial 5' rgsM coding sequence starting from the start codon (406 bp) was PCR-amplified from S. meliloti genomic DNA and inserted into pK18 mob2 downstream of the IPTG-inducible lac promoter.Integration of these constructs into the genome placed the full-length protein coding sequence under the control of the corresponding IPTG-inducible promoter.Adjacent open reading frames were not affected by integration of these constructs.Conditional RgsP and RgsM depletion was then achieved by constitutively expressing the lacI repressor gene from vector pSRKGm in the strains Rm2011 rgsP dpl and Rm2011 rgsM dpl , respectively.
Gene overexpression constructs were generated by insertion of the corresponding coding sequences downstream of either the IPTG-inducible lac promoter in medium-copy vectors pSRKKm and pSRKGm [71], the IPTG-inducible T5 promoter in medium-copy vector pWBT, the taurine-inducible tauA promoter in low-copy vector pR-P tau , or a constitutive synthetic promoter (P syn ) in low-copy vector pR_egfp [72].In the case of pR_egfp, a 114 bp region upstream of the rgsP coding region was included.
To generate single-copy ectopic rgsP or rgsP-egfp expression constructs, we first determined the native rgsP promoter and its activity levels.At its native chromosomal locus, rgsP is most likely in an operon with rimJ.Therefore, a 300 bp genomic region upstream of rgsP as well as a 944 bp region including the putative rimJ promoter and the whole rimJ coding sequence (S7A Fig) were tested for promoter activity using fusions to egfp (S7B Fig).Since higher levels of egfp expression were observed in the case of the 944 bp fragment, we used this DNA region as native rgsP promoter.To exclude possible non-desirable effects of an additional rimJ copy, a nonsense mutation was introduced 64 nucleotides downstream of the rimJ start codon yielding promoter construct P Ã rgsP (S7A Fig) .The rgsP, rgsP-egfp or rgsP-3×flag coding sequences were inserted downstream of the P Ã rgsP sequence in single-copy plasmid pABC2S-mob [73].For generation of amino acid substitutions or protein variants lacking specific domains splicing by overlap extension PCR was applied.
The rgsP promoter-egfp fusions were generated by insertion of the upstream non-coding region (either long (944 bp) or short (300 bp)) and the three first codons of rgsP into replicative medium-copy number plasmid pSRKKm-egfp [17].Constructs for purification of His 6 -tagged proteins were generated by insertion of the coding sequence excluding the start codon into expression vector pWH844 [74].Fusions of rgsM to phoA were assembled from the full-length or partial rgsM coding sequence and the E. coli phoA coding sequence missing the first 26 codons.

Fluorescence measurements
For promoter-egfp activity assays, TY overnight cultures were diluted 1:500 in 100 μl of TY medium or 30% MM and grown in 96-well plates at 30˚C with shaking at 1,200 rpm.EGFP fluorescence (excitation 488 ± 9 nm; emission 522 ± 20 nm, gain 82) and OD 600 were determined using the Infinite 200 Pro multimode reader (Tecan) and calculated as relative fluorescence units (RFU), which represent fluorescence values divided by OD 600 .Background EGFP fluorescence was determined using a control strain harboring pSRKKm-egfp.Fluorescence of three independent transconjugants was measured as biological replicates.
Treatment of S. meliloti, R. etli and A. tumefaciens cells with fluorescently-labeled D-amino acid HADA was performed as previously described [75].Briefly, cells were grown for 24 h in liquid medium in glass tubes to an OD 600 of 0.4-0.6.80 μl of the cultures were then mixed with 0.25 μl 100 mM HADA solution and incubated for 2.5 min (A.tumefaciens), 3 min (S.meliloti) or 3.5 min (R. etli) at 30 o C with shaking at 800 rpm.After addition of 186 μl 100% ethanol and 5-20 min incubation at room temperature (RT), cells were washed three times with 0.9% NaCl and subsequently placed onto 1% agarose pads.
Cell viability was assessed by DNA staining with the fluorescent intercalating agent propidium iodide.100 μl of S. meliloti liquid cultures (OD 600 0.4-0.8)were mixed with 1 μl of 2 mg/ ml propidium iodide stock solution and incubated for 5 min at room temperature.Cells were washed three times with 0.9% NaCl and subsequently placed onto 1% agarose pads.

Transmission electron microscopy
Concentrated S. meliloti cell suspensions were high pressure frozen (HPF Compact 02, Wohlwend, CH) and freeze-substituted (AFS2, Leica, Wetzlar, Germany) in a medium based on acetone, containing 0.25% osmium tetroxide, 0.2% uranyl acetate and 5% ddH 2 O according to the following protocol: -90˚C for 20 h, from -90˚C to -60˚C in 1 h, -60˚C for 8 h, -60˚C to -30˚C in 1 h, -30˚C for 8 h, -30˚C to 0˚C in 1 h, 0˚C for 3 h.Still at 0˚C, samples were washed three times with acetone before a 1:1 mixture of Epon 812 substitute resin (Fluka, Buchs, CH) and acetone was applied at room temperature for 2 h.The 1:1 mixture was substituted with pure resin to impregnate the samples overnight.After another substitution with fresh Epon, samples were polymerized at 60˚C for 2 days.The sample containing polymerized Epon blocks were then trimmed with razor blades and cut to 50 nm ultrathin sections using an ultramicrotome (UC7, Leica, Wetzlar, Germany) and a diamond knife (Diatome, Biel, Switzerland).Sections were applied onto 100 mesh copper grids coated with pioloform.For additional contrast, mounted sections were post-stained with 2% uranyl acetate for 20-30 min and subsequently with lead citrate for another 1-2 min.The sections were finally analyzed and imaged using a JEM-2100 transmission electron microscope (JEOL, Tokyo, Japan) equipped with a 2k x 2k F214 fast-scan CCD camera (TVIPS, Gauting, Germany).

Quantification of intracellular c-di-GMP
Strain Rm2011 rgsP dpl ectopically expressing rgsP variants from P Ã rgsP was grown in triplicates in liquid TY medium without IPTG and harvested in the exponential growth phase 24 h after inoculation.Quantification of intracellular c-di-GMP was performed as previously described [76].Briefly, cells were collected by centrifugation and nucleotides were extracted three times with acetonitrile/methanol/water (2:2:1), dried and subjected to liquid chromatography-tandem mass spectrometry.c-di-GMP was normalized to total protein, determined using Bradford reagent (Bio-Rad).

Protein purification
Heterologous protein expression and purification was performed as previously described [17].E. coli BL21(DE3) harboring expression plasmids were grown in LB medium in flasks to OD 600 of 0.5-0.6 and protein expression was induced with 0.4 mM IPTG overnight at RT. Cells were lysed using French press (pressure 1,000 lb/in 2 ) and the lysates were centrifuged for 60 min at 24,000 ×g and 4˚C.Cleared lysates were applied to His SpinTrap columns (GE Healthcare) following the manufacturer's instructions and eluted with 0.5 M imidazole.Purity of isolated proteins was assessed by SDS-PAGE and Coomassie brilliant blue staining.Protein concentration was determined using Bradford reagent (Bio-Rad).

Preparation of [α-32 P]-labeled c-di-GMP
[α-32 P]-labeled c-di-GMP was synthesized using purified Caulobacter crescentus His 6 -DgcA at 10 μM from GTP and [α-32 P]-GTP (0.1 μCi/μl) at 1 mM in the reaction buffer (50 mM Tris-HCl, 300 mM NaCl, 10 mM MgCl 2, pH 8.0), overnight at 30˚C.The reaction was then treated with 5 units of calf intestine alkaline phosphatase (Fermentas) for 1 h at 22˚C to hydrolyze unreacted GTP and stopped by incubation for 10 min at 95˚C.The precipitated proteins were removed by centrifugation (10 min, 20,000 ×g, 22˚C) and the supernatant was used for the PDE activity and the c-di-GMP binding assays.

In vitro c-di-GMP binding assay
c-di-GMP binding was determined using a differential radial capillary action of ligand assay (DRaCALA) with [α-32 P]-labeled c-di-GMP, as previously described [77].This assay is based on the ability of dry nitrocellulose to prevent diffusion of bound proteinligand complexes and thereby separate them from free ligand.Reaction mixtures (50 μl) containing [α-32 P]-labeled c-di-GMP and 20 μM of indicated protein in the binding buffer (10 mM Tris, 100 mM NaCl, 5 mM MgCl 2, pH 8.0) were incubated for 10 min at RT. 10 μl of this reaction mixture was spotted onto nitrocellulose membrane and allowed to dry prior to exposing a phosphor-imaging screen (Molecular Dynamics).Data were collected using a STORM 840 scanner.

In vitro DGC and PDE activity assay
DGC and PDE activities were determined as previously described [78,79].Reaction mixtures (40 μl) containing purified proteins at 10 μM, in reaction buffer (50 mM Tris-HCl, 300 mM NaCl, 10 mM MgCl 2 , pH 8.0) were first pre-incubated for 5 min at 30˚C.DGC reactions were initiated by adding GTP/[α-32 P]-GTP (0.1 μCi/μl) to 1 mM, incubated at 30˚C for the indicated periods of time and stopped by adding an equal volume of 0.5 M EDTA.PDE reactions were initiated by adding [α-32 P]-labeled c-di-GMP and stopped by adding an equal volume of 0.5 M EDTA after indicated time periods.2 μl of the PDE or DGC reaction mixtures were spotted on polyethyleneimine-cellulose TLC chromatography plates, developed in 2:3 (v/v) 4 M (NH 4 ) 2 SO 4 /1.5 M KH 2 PO 4 (pH 3.65).Plates were dried prior to exposing a phosphor-imaging screen (Molecular Dynamics).Data were collected and analyzed using a STORM 840 scanner (Amersham Biosciences).

Co-immunoprecipitation (Co-IP) and protein identification by mass spectrometry
Co-IP and protein identification by mass spectrometry was performed as previously described including small modifications [80].Cultures of Rm2011 rgsP-3×flag, Rm2011 rgsM-3×flag and control strain Rm2011 harboring the empty vector pWBT were grown in TY medium in flasks to an OD 600 of 0.6 and cross-linked with 0.1% formaldehyde for 15 min at RT. Reaction was quenched by adding glycine at a final concentration of 0.35 M. Cells were washed, resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 2 mM phenylmethylsulfonyl-fluorid [PMSF], pH 7.4) and lysed using a French press (pressure 1,000 lb/in 2 ).Cleared lysates obtained after ultracentrifugation (100,000 ×g, 1 h, 4˚C) were incubated with anti-FLAG M2 affinity gel (FLAG Immunoprecipitation Kit, Sigma) overnight at 4˚C on a rolling shaker.Bound proteins were eluted with 3×FLAG peptide solution.
For mass-spectrometry analysis, proteins were digested by Sequencing Grade Modified Trypsin (Promega) at 37˚C overnight.The mass spectrometric analysis was performed using an Orbitrap Velos Pro mass spectrometer (Thermo Fisher Scientific).An Ultimate nanoRSLC-HPLC system (Dionex), equipped with a custom 20 cm x 75 μm C18 RP column filled with 1.7 μm beads was connected online to the mass spectrometer through a Proxeon nanospray source.1-15 μl of the tryptic digest were injected onto a C18 pre-concentration column.Automated trapping and desalting of the sample was performed at a flow rate of 6 μl/ min using water/0.05%formic acid as solvent.Separation of the tryptic peptides was achieved with the following gradient of water/0.05%formic acid (solvent A) and 80% acetonitrile/ 0.045% formic acid (solvent B) at a flow rate of 300 nl/min: holding 4% B for 5 min, followed by a linear gradient to 45% B within 30 min and linear increase to 95% solvent B in additional 5 min.The column was connected to a stainless steel nanoemitter (Proxeon, Denmark) and the eluent was sprayed directly towards the heated capillary of the mass spectrometer using a potential of 2,300 V.A survey scan with a resolution of 60,000 within the Orbitrap mass analyzer was combined with at least three data-dependent MS/MS scans with dynamic exclusion for 30 s either using CID with the linear ion-trap or using HCD combined with orbitrap detection at a resolution of 7,500.Data analysis was performed using Proteome Discoverer (Thermo Fisher Scientific) with SEQUEST and MASCOT (version 2.2; Matrix science) search engines using either SwissProt or NCBI databases.The combined transmembrane topology and signal peptide prediction online tool PHOBIUS [27] was used to predict regions with high hydrophobicity within candidate protein interactions partners of RgsP and RgsM.

Bacterial two-hybrid analysis and β-galactosidase activity assay
Bacterial two-hybrid analysis was performed as previously described [31].The adenylate cyclase-deficient strain E. coli BTH101 was co-transformed with plasmids carrying rgsP ΔGGDE- FΔEAL or rgsM translationally fused to T25 and T18 fragments of Bordetella pertussis adenylate cyclase.Transformant colonies were grown in 100 μl LB supplemented with antibiotics at 30˚C for 6 hours with shaking 1,200 rpm. 10 μl of each culture was spotted onto LB agar plates containing kanamycin, ampicillin, X-Gal and IPTG.Plates were imaged after 30 h of incubation at 30˚C.β-galactosidase activity was determined as previously described with small modifications [52].Cells grown on LB agar containing IPTG were resuspended in 1 ml of Z-buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , pH 7.0), in triplicates and the OD 600 was recorded.Cell permeabilization was facilitated by addition of 50 μl of chlorophorm and 50 μl of 0.05% SDS.The aqueous phase was mixed with an equal volume of Z-buffer containing 50 mM β-mercaptoethanol, and ortho-nitrophenyl-β-D-galactopyranoside (ONPG) was added to the final concentration of 0.5 mg/ml.A 420 was recorded using the Infinite M200 PRO fluorescence reader (Tecan).Miller Units (MU) were calculated as follows: MU = 1000 Ã A 420 / (t Ã V Ã OD 600 ).t represents the time in min and V the volume in ml.

Western blot
Western blot analysis was performed as previously described [68].Briefly, S. meliloti and E. coli strains expressing 3×flag-tagged rgsP or rgsM variants were grown in glass tubes supplemented with corresponding antibiotics.Unless otherwise specified, cells were grown in TY medium without IPTG and collected at an OD 600 of 0.4-0.824 h after inoculation.Cells were adjusted to an OD 600 of 1, 10 μl of lysed cells were loaded to SDS-PAGE gel and separated proteins were transferred to PVDF membrane (Thermo Fisher Scientific).RgsP protein variants were detected using anti-FLAG M2-Peroxidase (HRP) antibody (Sigma-Aldrich).Membranes were incubated with Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) and imaged using the luminescence image analyzer LAS-4000 (Fujifilm).

Peptidoglycan purification and muropeptide analysis
Isolation of PG sacculi was achieved following a published protocol [81].Briefly, S. meliloti strains were grown in 400 ml TY and LB media in flasks for 24 h at 30˚C until OD 600 reached 0.4-0.8.E. coli strains were grown in 400 ml LB for 4 h at 37˚C until OD 600 reached 0.5-0.9.Cells were harvested by centrifugation, resuspended in PBS buffer, added dropwise to boiling 8% SDS solution and stirred for 30 min.PG sacculi were pelleted by ultracentrifugation in a Beckman Coulter Optima at 440,000 ×g, at ambient temperature for 1 h and resuspended in water.Centrifugation and resuspension steps were repeated until the sacculi were free of SDS as verified by a published assay [82].The sacculi were then incubated with 100 μg/ml of αamylase in reaction buffer (10 mM Tris-HCl, 10 mM NaCl, pH 7.0) for 2 h at 37˚C, followed by incubation with 200 μg/ml pronase E for 1 h at 60˚C, to remove high molecular weight glycogen and proteins, respectively.The enzymes were removed by incubation in 4% SDS solution for 15 min at 80˚C.Sacculi were washed free of SDS as described above and resuspended in 0.02% sodium azide.
Purified sacculi were digested with the muramidase cellosyl (Hoechst, Frankfurt, Germany) at 37˚C overnight.The reaction was terminated by boiling the sample at 100˚C for 10 min.The samples were centrifuged (15,000 ×g for 10 min) and the soluble muropeptides were recovered from the supernatant, reduced with NaBH 4 and separated by HPLC as described for E. coli [81].Muropeptides were then separated on a C18 reversed-phase column (ProntoSIL) using an Agilent 1220 infinity HPLC system.Separation was carried out over a 180 min linear gradient from buffer A (50 mM sodium phosphate, pH 4.31) to buffer B (75 mM sodium phosphate, pH 4.95, 15% methanol).Muropetides were detected by their absorbance at 205 nm.Muropeptides of interest corresponding to major peaks were manually collected and analyzed by tandem mass spectrometry (MS/MS) as previously described [83].

Peptidoglycan binding assay
Purified PG (~100 μg) from S. meliloti strain Rm2011 was centrifuged at 15,000 ×g, 4˚C for 14 min and resuspended in binding buffer (10 mM Tris-maleate, 10 mM MgCl 2 , 50 mM NaCl, pH 6.8). 10 μg of protein of interest was incubated with or without PG in a final volume of 100 μl, and incubated at 4˚C for 30 min.Samples were centrifuged as described above and the supernatant was collected (supernatant fraction) whilst the pellet was resuspended in 200 μl of binding buffer.Another centrifugation step as described above was carried out (wash fraction).Bound proteins were released from PG by incubation with 100 μl of 2% SDS, at 4˚C for 1 h before being collected by a final centrifugation step as done at earlier steps.The proteins present in the different fractions were analyzed by SDS-PAGE.

Peptidoglycan hydrolase activity assay
Proteins (2 μM final concentration) were incubated with 10 μg of sacculi from S. meliloti, overnight at 37˚C, in a volume of 100 μl (containing 10 mM HEPES-NaOH, 1 mM ZnCl 2, 150 mM NaCl, 0.05% Triton X-100, pH 7.5).Proteins were inactivated by boiling for 10 min.Cellosyl was added (1 μM final concentration) and the samples were incubated overnight again at 37˚C.Samples were boiled for 10 min and the soluble muropeptides were collected by centrifugation at 15,000 ×g for 10 min, and taking the supernatant fraction.Muropeptides were reduced with NaBH 4 and analyzed by HPLC as described above.
; S5A Fig).Moreover, upon RgsP depletion, the proportion of pre-divisional cell doublets with visible septum constriction increased to 42% compared to 13% in RgsP-replete cultures (Fig 1D; S5B Fig).These results suggest that polar cell wall synthesis and late stages of cell division were impaired in the absence of RgsP.

Fig 1 .
Fig 1. RgsP function is related to cell growth and division.(A) Growth of S. meliloti RgsP depletion strain Rm2011 rgsP dpl in TY medium in presence or absence of IPTG.OD 600 was recorded every 30 min.Error bars represent the standard deviation of six biological replicates.(B-D) Microscopy analysis of Rm2011 rgsP dpl , grown in TY medium in presence or absence of IPTG for 24 h.(B) DIC microscopy images.(C) Transmission electron microscopy images.(D) Phase contrast and fluorescence microscopy images of cells pulse-labeled with HADA for 3 min.(B,D) Bar, 5 μm.(C) Bar, 1 μm.https://doi.org/10.1371/journal.pgen.1007594.g001 Fig), indicating mutually exclusive localization of RgsP and PleD at the new pole.

Fig 2 .
Fig 2. RgsP localizes to sites of zonal cell wall synthesis.(A) Fluorescence microscopy images of exponentially growing Rm2011 rgsP-egfp (gene fusion at the native genomic location) harboring pSRKGm-mCherry-ftsZ.Cells were pulse-labeled with HADA for 3 min.Cells with a mCherry-FtsZ focus (yellow arrowheads) and additional RgsP-EGFP and HADA signals at mid-cell (white arrowheads) are indicated.Bar, 5 μm.(B) Quantitative analysis of localization of HADA-stained regions of zonal cell wall synthesis and mCherry-FtsZ relative to localization of RgsP-EGFP in exponentially growing cells.The population was first divided into three classes dependent on EGFP signal localization and each class was evaluated for location of mCherry-FtsZ and HADA signals.n, total number of cells considered for statistical analysis.(C,D) Time-lapse microscopy images of Rm2011 rgsP-egfp harboring either pSRKGm-parB-mCherry (C) or pSRKGm-mCherry-ftsZ (D).The doubling time of analyzed bacteria was approximately 135 min (C).Time is given in minutes.Bar, 1 μm.Signals: HADA, blue; EGFP, green; mCherry, red.https://doi.org/10.1371/journal.pgen.1007594.g002

Fig 3 .
Fig 3. RgsP essentiality resides in its N-terminal part.(A) RgsP domain architecture and generated mutant variants.(B) Complementation of the RgsP depletion strain Rm2011 rgsP dpl growth defect by rgsP variants expressed from P syn (elevated), rgsP variants expressed from P Ã rgsP (native) or egfp-tagged rgsP variants expressed from P Ã rgsP (native, EGFP).Relative growth was calculated as a ratio of OD 600 of cultures without IPTG (depleted of RgsP produced from the chromosome) and OD 600 of cultures with IPTG (non-depleted of RgsP produced from the chromosome), 24 h after inoculation.Error bars represent the standard deviation of three biological replicates.The absolute OD 600 values are shown in S8A Fig. (C,D) Microscopy images of Rm2011 rgsP dpl , ectopically expressing the indicated (C) nontagged and (D) egfp-tagged rgsP variants, acquired after 24 h of growth in TY medium without IPTG.Bars, 5 μm.https://doi.org/10.1371/journal.pgen.1007594.g003

Fig 4 .
Fig 4. RgsP interacts with RgsM in bacterial two-hybrid system.(A) Schematic representation of the putative periplasmic metallopeptidase RgsM containing a transmembrane segment at the N-terminus and a C-terminal LytM domain.(B) Quantitative analysis of the interaction between RgsP and RgsM.β-galactosidase activity of E. coli BTH101, co-transformed with plasmids carrying rgsP ΔGGDEFΔEAL and full-length rgsM translationally fused to T25 and T18 fragments of Bordetella pertussis adenylate cyclase.The combination of yeast GCN4 leucin-zipper region (Zip) served as positive control.Error bars represent the standard deviation of three biological replicates.Blue-staining of colonies grown on a single LB agar plate containing X-Gal and IPTG is shown above.https://doi.org/10.1371/journal.pgen.1007594.g004 Fig).Next, we constructed a RgsM depletion strain Rm2011 rgsM dpl by placing the chromosomal rgsM gene under the control of an IPTG-inducible promoter.Depletion of a C-terminally 3×FLAGtagged RgsM variant was confirmed using the same genetic setup (S1B Fig).The growth of Rm2011 rgsM dpl was significantly impaired in the absence of IPTG (Fig 5A).Strikingly similar to cells depleted of RgsP, RgsM-depleted cells lost the wild type rod shape (Figs 5B and 1B; S2 Fig).Electron microscopy revealed regions of low electron density in RgsM-depleted cells (Fig 5C; S3B Fig).7.1% of these cells were stained by propidium iodide, indicating both lethal and bacteriostatic effects of RgsM depletion, similar to RgsP depletion (S4A Fig).Following depletion of RgsM for 12 or 24 h, the number of colony forming units was reduced by 89.4% or 99.5%, respectively, relative to cultures pre-incubated under non-depletion conditions (S4B Fig).We microscopically monitored growth of cells, previously grown under depletion conditions for 24 h.10.9% of the previously RgsM-depleted cells gave rise to microcolonies on medium supplemented with IPTG, whereas 99.1% of these cells did not divide on medium lacking IPTG during the 12 h observation period (S4C Fig).

Fig 5 .
Fig 5. RgsM function is related to cell growth and division.(A) Growth of S. meliloti RgsM depletion strain Rm2011 rgsM dpl in TY medium with or without added IPTG.OD 600 was recorded every 30 min.Error bars represent the standard deviation of six biological replicates.(B-D) Microscopy analysis of Rm2011 rgsM dpl , grown in TY medium in presence or absence of IPTG for 24 h.(B) DIC microscopy images.(C) Transmission electron microscopy images.Bar, 1 μm.(D) Phase contrast and fluorescence microscopy images obtained after 3 min HADA labeling pulse.(B,D) Bar, 5 μm.https://doi.org/10.1371/journal.pgen.1007594.g005 Fig).Contrastingly, in LB medium they grew very poorly and appeared enlarged and spherical (Fig 7A and 7B), with dramatically enlarged periplasm and inner membrane invaginations (Fig 7C; S3C Fig).Overexpression of rgsM in LB impaired PG biosynthesis, as judged from a very weak dispersed HADA staining, only visible after adjusting HADA fluorescence signal intensity, in contrast to cells carrying an empty vector, which showed polar and septal HADA signals (Fig 7D; S5A and S11C Figs).Noteworthy, even in the absence of RgsM overexpression, the rod cell shape differed slightly in TY and LB, with broader and shorter cells grown in LB (S11B Fig).
Fig), suggesting a possible dominant negative effect of RgsM H510A overproduction on the native RgsM function.Accumulation of overproduced RgsM wt and RgsM H510A in Rm2011 was confirmed by detecting the corresponding 3×FLAG-tagged variants (S1E Fig).

Fig 6 .
Fig 6.Localization of RgsP and RgsM throughout the cell cycle is concurrent and mutually interdependent.(A) Fluorescence microscopy images of exponentially growing Rm2011 mVenus-rgsM rgsP-mCherry (both gene fusions at the native genomic location).Arrowheads indicate the mid-cell region of pre-divisional cells.Bar, 5 μm.(B) Time-lapse microscopy images of cells described in panel A. The doubling time of analyzed bacteria was approximately 150 min.Time is shown in minutes.Bar, 1 μm.(C,D) Fluorescence microscopy images of RgsP depletion strain carrying mVenus-rgsM (gene fusion at the native genomic location) (C) and RgsM depletion strain carrying rgsP-mCherry (gene fusion at the native genomic location) (D) acquired after 24 h of growth in TY medium with or without IPTG.Bars, 5 μm.https://doi.org/10.1371/journal.pgen.1007594.g006