An Alternative and Conserved Cell Wall Enzyme That Can Substitute for the Lipid II Synthase MurG

Almost all bacteria are surrounded by a cell wall, which protects cells from environmental harm. Formation of the cell wall requires the precursor molecule lipid II, which in bacteria is universally synthesized by the conserved and essential lipid II synthase MurG.

strong deformations of the cell membrane, followed by the release of progeny cells by blebbing, tubulation, and vesiculation (32,38). Given that lipid vesicles without any content can proliferate in a manner similar to that observed for L-forms led to the hypothesis that this mode of proliferation may be comparable to that used by early life forms that existed before the cell wall had evolved (39,40).
Here, we exploited the unique properties of a K. viridifaciens L-form strain that readily switches between a wall-deficient and filamentous mode of growth to discover an alternative MurG-like enzyme that is important for building the PG-based cell wall. Our data surprisingly show that K. viridifaciens produces wild-type peptidoglycan in the absence of murG, which has so far been considered essential for lipid II biosynthesis in all bacteria. The MurG activity is taken over by a homologue called MglA, which is widespread in filamentous actinobacteria and able to substitute for the absence of MurG across different genera.

RESULTS
Morphological transitions of the shape-shifting strain alpha. We recently generated a K. viridifaciens L-form lineage by exposing the parental wild-type strain to high levels of penicillin and lysozyme. This strain, designated alpha, proliferates indefinitely in the cell wall-deficient state in media containing high levels of osmolytes (32). On solid L-phase medium (LPMA), containing high levels of sucrose, alpha forms greenish viscous colonies, which exclusively contain L-form cells (Fig. 1A). In contrast, the parental strain forms compact, yellowish colonies composed of mycelia and S-cells on LPMA medium (Fig. 1B). Likewise, in liquid Baird-Parker liquid (LPB) medium, which like LPMA medium contains high levels of sucrose, alpha proliferates exclusively in the wall-deficient state in a manner that is morphologically similar to that described for other Lforms (35,41,42) (see Movie S1A in the supplemental material; Fig. 1C). Following strong deformations of the mother cell membrane (see panels for 56, 150, and 200 min in Fig. 1C), small progeny cells are released after approximately 300 min. The mother cell, from which the progeny were released (indicated with an asterisk in Fig. 1C), lysed after 580 min. Characterization using transmission electron microscopy (TEM) confirmed that alpha possessed no PG-based cell wall when grown on media containing high levels of osmolytes ( Fig. 1D; Fig. S1). Notably, when alpha is plated on maltoseyeast extract medium (MYM; lacking high levels of osmolytes) the strain can switch to the mycelial mode of growth (Fig. 1E). However, unlike the wild-type strain (Fig. 1F), the mycelial colonies of alpha fail to develop aerial hyphae and spores. Subsequent transfer of mycelia to LPMA medium plates stopped filamentous growth and reinitiated wall-deficient growth, during which L-form cells were extruded from stalled hyphal tips (Movie S1B; Fig. 1G). Given the ability of these wall-deficient cells to proliferate, they eventually dominated the culture (not shown). Taken together, these results demonstrate that alpha can switch between a walled and wall-deficient state.
Deletion of divIVA abolishes switching of alpha from the wall-deficient to the filamentous mode of growth. The ability of alpha to efficiently switch between the walled and wall-deficient state provides an ideal platform to delete genes essential for cell wall biosynthesis. As a proof of concept, we focused on divIVA, which is essential for polar growth in filamentous actinomycetes (24). In actinobacteria, divIVA is located adjacent to the conserved dcw gene cluster (Fig. S2). divIVA is present in Gram-positive rod-shaped (Mycobacterium, Corynebacterium, Bacillus), filamentous (Streptomyces and Kitasatospora), and coccoid (Staphylococcus and Streptococcus) bacteria but absent in Gram-negative bacteria, such as Escherichia coli. In B. subtilis and Staphylococcus aureus, the DivIVA proteins share only 29% (BSU15420) and 26% (SAOUHSC_01158) amino acid identity to the Streptomyces coelicolor orthologue. To localize DivIVA, plasmid pKR2 was created, allowing constitutive expression of DivIVA-enhanced green fluorescent protein (eGFP) (see Materials and Methods). Fluorescence microscopy revealed that the fusion protein localized to hyphal tips (Fig. S3A), as in streptomycetes (24). When alpha was grown in the wall-deficient state in LPB medium, typically one or two foci of DivIVA-eGFP were detected per cell, which invariably were localized to the membrane. In contrast, no foci were detected in L-form cells containing the empty plasmid (pKR1) or those expressing cytosolic eGFP (pGreen [43]). We then constructed the plasmids pKR3 to delete divIVA and pKR4 to delete a large part of the dcw gene cluster, including divIVA (see Materials and Methods). Introduction of these plasmids into alpha by polyethylene glycol (PEG)-mediated transformation and a subsequent screening yielded the desired divIVA and dcw mutants (Fig. S4). Analysis of growth in LPB medium or on solid LPMA plates indicated that the L-form cells proliferated normally in the absence of divIVA or part of the dcw gene cluster ( Fig. 2A). However, when L-form cells were plated on MYM (lacking osmoprotectants), only the alpha strain was able to switch to the mycelial mode of growth (Fig. 2B). Introduction of plasmid pKR6, which expresses divIVA from the constitutive gap1 promoter, complemented the growth of the divIVA mutant on MYM (Fig. 2B). In agreement, Western blot analysis using antibodies against DivIVA of Corynebacterium glutamicum confirmed the absence of DivIVA in both the divIVA and the dcw mutant and showed that expression was restored in the divIVA mutant complemented with pKR6 (Fig. 2C).
To analyze if the switch from the wall-deficient to the walled state in the absence of DivIVA was blocked due to the failure to produce the cytosolic precursors required for peptidoglycan synthesis in the L-form state, we performed a comparative liquid chromatography-mass spectrometry (LC-MS) analysis (Fig. 2D). We noticed that the LC-MS profiles of the divIVA and dcw mutant strains were similar to that of alpha with respect viridifaciens alpha strain on LPMA medium yields greenish, mucoid colonies exclusively consisting of L-form cells, unlike the wild-type strain that forms yellowish colonies consisting of mycelia and S-cells (B). (C) Time-lapse microscopy stills of alpha proliferating in the wall-deficient state in liquid LPB medium. The arrowhead shows the mother cell, which generates progeny and lyses after 580 min (marked with an asterisk). Stills were taken from Movie S1A in the supplemental material. (D) Transmission electron microscopy of a wall-deficient cell of alpha. (E) Growth of alpha on solid MYM yields compact, nonsporulating colonies unlike those of the wild-type strain, which forms gray-pigmented sporulating colonies (F). (G) Time-lapse microscopy stills of the mycelium of alpha transferred to LPMA medium, showing the extrusion of L-forms by filaments (arrowheads). Stills were taken from Movie S1B. Scale bars represents 20 mm (A, B), 10 mm (C, E, F, G), and 500 nm (D).
to the cytosolic PG building blocks (Fig. 2D). Importantly, MS-MS analysis identified the last cytosolic precursor in the PG biosynthesis pathway, UDP-MurNAc-pentapeptide (M w = 1,194.35) in all strains (Fig. 2E). Taken together, these results demonstrate that DivIVA is essential for filamentous growth but not required for synthesis of the cytosolic PG precursors.
Identification of a distant MurG homologue as an alternative lipid II synthase. Having a mutant lacking many genes of the dcw cluster offers many opportunities for the study of individual genes. The constructed dcw mutant lacks ftsW, murG, ftsQ, ftsZ, ylmD, ylmE, selF, sepG, and divIVA. Surprisingly, introduction of only divIVA (expressed from the constitutive gap1 promoter) (Fig. S5) restored the ability of the dcw mutant to switch to the walled mode of growth on solid media lacking osmoprotectants (Fig. 3A). Discovery of an Alternative Lipid II Synthase ® The colonies that were formed were small and heterogeneous compared to the mycelial colonies formed by alpha (Fig. 3A). Furthermore, expression of divIVA in the dcw mutant was not able to restore filamentous growth in liquid cultures (data not shown). To verify that the dcw mutant expressing divIVA produced normal PG on solid medium, we performed a peptidoglycan architecture analysis using LC-MS (Fig. 3B). This surprisingly revealed that all expected muropeptides were formed at levels comparable to those formed by alpha and the wild-type strain, despite the absence of a functional murG gene ( Fig. 3B; Table 1).
The ability of the dcw mutant expressing divIVA to become filamentous inevitably means that another protein had functionally replaced the activity of MurG. BLAST analysis of the amino acid sequence of MurG from Streptomyces coelicolor (MurG Sco , SCO2084) against the genome sequence of K. viridifaciens revealed that this actinomycete contains two putative, but distant, MurG homologs ( Table 2). The two additional homologs (BOQ63_RS12640 and BOQ63_RS05415) showed 31.2% and 16.5% sequence  identity, respectively, to MurG (Fig. S7). Further investigation revealed that MurG proteins possess two characteristic domains: an N-terminal domain that contains the lipid I binding site (PF03033) (44) and a C-terminal domain that contains the UDP-GlcNAc binding site (PF04101) (Fig. S6), both of which are required for the UDP-N-acetylglucosamine transferase activity. Of the two distant MurG homologs, only BOQ63_RS12640 contained both domains (Fig. S6). A broader search of MurG-like proteins in other Streptomyces and Kitasatospora spp. revealed that 38% of the strains possess one, two, and sometimes even three genes for MurG-like proteins containing both the necessary N-terminal (PF03033) and C-terminal (PF04101) domains (Fig. 4A), in addition to canonical MurG, which is present in all strains and encoded in the dcw gene cluster. A sequence similarity network was produced by pairwise comparison of the 1,553 MurG and MurG-like proteins extracted from all translated Streptomyces and Kitasatospora genomes, which showed that nearly all MurG proteins encoded by the orthologue of murG in the dcw gene cluster grouped together. However, the MurG-like proteins clustered in many different groups (Fig. 4B).
To corroborate that murG is not required for filamentous growth, we decided to delete murG in alpha using knockout construct pKR8 (see Materials and Methods). The genotype of the mutant was verified by PCR ( Fig. S8A) and showed that the absence of murG had no effect on L-form or filamentous growth (Fig. 5A). Likewise, inactivation of mglA in alpha using construct pKR9 had no effect on L-form growth and did not prevent switching to mycelial growth (Fig. 5A). We then attempted to create a double mu-  Discovery of an Alternative Lipid II Synthase ® tant by deleting mglA in the murG mutant. PCR analysis of a putative double mutant strain with the highly sensitive Q5 DNA polymerase indicated, however, that a small proportion of the multinucleated L-forms had retained a copy of mglA (Fig. S8A). Also, further subculturing of this merodiploid strain in the presence of antibiotics that counterselected for maintenance of mglA did not lead to a complete loss of this gene, suggesting that the ability to produce lipid II is essential in these L-forms (see Discussion). Nevertheless, plating this merodiploid strain on MYM essentially blocked mycelial growth, and only at very high cell densities were infrequent shifters found (see encircled colony in Fig. 5A).
Having demonstrated that murG is not required for the filamentous growth of alpha, we then wondered whether murG would also be dispensable for the filamentous growth of the wild-type strain. Notably, murG deletion mutants could not be obtained if transformants were selected on MYM, unlike with a mglA deletion mutant that was readily found. However, when transformants were selected on LPMA medium containing high levels of sucrose, a murG mutant could be created in K. viridifaciens (Fig. S8B). As shown in Fig. 5B, the generated murG and mglA mutants were able to develop and sporulate normally on MYM, compared to the parental wild type. However, exposing the strains to low levels of penicillin and ampicillin revealed that the murG mutant was more susceptible to these cell wall-targeting antibiotics than the wild type and its mglA mutant. In contrast, no difference effect was observed when tetracycline was added to the plates (Fig. 5C). All together, these results demonstrate that MurG and MglA have overlapping activities, whereby MglA is able to functionally replace the canonical lipid II synthase MurG.
MglA from K. viridifaciens can functionally replace MurG in S. coelicolor. The observations that mglA can functionally replace murG in K. viridifaciens and that strains expressing only MglA produce wild-type peptidoglycan strongly suggest that the mglA gene product synthesizes lipid II. To further substantiate this, we investigated whether mglA could also functionally complement murG in another actinobacterium, namely, the model organism S. coelicolor M145, which itself does not harbor an orthologue of mglA. For this, we created the construct pGWS1379, expressing mglA from the constitutive modified ermE* promoter (78) in the integrative vector pMS82, and introduced it into S. coelicolor. As a control, we used the empty vector pMS82. We then applied CRISPR interference (CRISPRi) (45) to knock down the native murG Sco gene to assess viability. CRISPRi works only when the spacer of the endonuclease-deficient Cas9 (dCas9)/ single guide RNA (sgRNA) complex targets the nontemplate strand of murG Sco , and not the template strand, or when the spacer is absent (45,46). The functionality of the CRISPRi constructs was evident in control cells without mglA; colonies expressing the dCas9/sgRNA complex targeting the nontemplate strand of murG Sco in M145 formed small colonies, likely due to leaky expression of the essential murG Sco gene. Conversely, control transformants harboring CRISPRi constructs targeting the template strand or without the spacer (empty plasmid) grew normally (Fig. 6A). Excitingly, S. coelicolor transformants expressing mglA formed normal-size colonies under all conditions, even when murG Sco expression was knocked down by the CRISPRi system. Restoration of normal growth was also observed when these transformant colonies were transferred to fresh agar plates, while colonies of transformants lacking mglA remained small (Fig. 6B). This validates the concept that mglA of K. viridifaciens can functionally replace canonical murG in S. coelicolor. Taken together, our experiments show that the MglA enzyme can functionally replace the lipid II biosynthetic enzyme MurG, both in Kitasatospora and in Streptomyces.

DISCUSSION
The cell wall is a hallmark feature of bacterial cells, and the steps involved in its biosynthesis are widely conserved across the bacterial domain. In all bacteria, the final Discovery of an Alternative Lipid II Synthase ® cytosolic step in precursor biosynthesis is the conversion of lipid I to lipid II by MurG, encoded in the dcw gene cluster. We here show for the first time that the enzyme MglA can replace the activity of MurG and demonstrate that murG is dispensable in the filamentous actinomycete K. viridifaciens in the presence of mglA. MglA alone is sufficient to produce wild-type peptidoglycan. MglA is in fact widespread among the Streptomycetaceae and was identified in the genomes of 38% of all Streptomyces and Kitasatospora strains. Furthermore, introduction of K. viridifaciens mglA into S. coelicolor M145, which itself lacks an orthologue of mglA, allowed the knockdown of the canonical murG gene using CRISPRi, showing that the gene is a bona fide cell wall biosynthetic gene that is functional in different actinobacteria.
Filamentous actinomycetes are multicellular bacteria that form networks of interconnected hyphae, whereby sporulating aerial hyphae are established after a period of vegetative growth. Streptomyces is a wonderful model system for the study of cell division because, among other reasons, canonical cell division is not required for the normal growth of this bacterium (21,25,47). Most of the cell division proteins are encoded by genes located in the conserved dcw gene cluster. In streptomycetes, many cell division genes, such as ftsI, ftsL, ftsW, and divIC, are required only for sporulation and do not affect normal growth (48)(49)(50). Our data, surprisingly, show that many genes within the dcw cluster can be deleted simultaneously in K. viridifaciens, including divIVA, which is essential for polar growth in actinobacteria, by using a strain (alpha) with the ability to readily switch between a wall-deficient and filamentous mode of growth. The alpha strain thus provides a unique system for the identification of proteins that are required for polar growth. As a proof of concept for this principle, divIVA, which is required for polar growth, was successfully deleted. The absence of divIVA arrested growth in the CRISPRi constructs were introduced into S. coelicolor M145 or with the control plasmid pMS82 and into a recombinant strain with pGWS1379 integrated in its genome, thus expressing K. viridifaciens MglA. Expectedly, no effect was seen when we introduced CRISPRi constructs that either had no spacer or contained a spacer targeting the template strand (T) of murG Sco . However, constructs targeting the nontemplate strand (NT) resulted in severe phenotypic defects and sick colonies of S. coelicolor that lacked mglA, but not in pGWS1379 transformants that expressed mglA. (B) Morphology of colonies of the strains carrying the CRISPRi constructs after their transfer to fresh mannitol soya flour (MS) agar plates. Images were taken after 5 days of incubation at 30°C. Bar, 2 mm. wall-deficient state but had no effect on the synthesis of the PG building blocks, consistent with its role in driving apical growth. This indicates that the block in PG formation occurred in a later step of the PG biosynthesis pathway. Introduction of only divIVA in the dcw mutant restored polar growth, which was a rather surprising discovery given the absence of a whole string of genes involved in cell division and cell wall synthesis and, in particular, murG. MurG catalyzes the coupling of GlcNAc to lipid I, yielding the PG precursor lipid II, and this enzymatic activity is therefore essential for cell wall synthesis. The ability of alpha to produce a cell wall with an apparently normal architecture, as shown by the analysis of the peptidoglycan, indicated that K. viridifaciens possesses other enzymes capable of synthesizing lipid II in the absence of murG. An in silico search in the genome of K. viridifaciens identified mglA (BOQ63_RS12640), which is a distant relative of MurG with the likely ability to replace the activity of canonical MurG. This is based on, among other things, the presence of the two domains that are known to be required for the transfer of GlcNAc to lipid I. Many actinobacteria possess proteins carrying these two domains, suggesting that MglA proteins are common in these bacteria. In fact, some species even contain three genes for MurG-like proteins, in addition to the canonical MurG encoded in the dcw gene cluster. Interestingly, both murG and mglA could be individually deleted in the wild-type strain, whereby the resulting mutants showed normal growth and development when strains were grown in nonstressed environments. However, the murG mutant was more susceptible to cell wall-targeting antibiotics than the wild-type strain or its mglA mutant. Considering that MglA alone suffices to produce normal peptidoglycan, this suggests that MurG is required to build a more robust cell wall. Deletion of murG was possible only after exposing transformants to hyperosmotic growth conditions. We hypothesize that the hyperosmotic conditions activated the transcription of mglA, thus allowing deletion of murG specifically under these growth conditions. This implies that the function of mglA is to synthesize lipid II under specific growth conditions, for instance during hyperosmotic stress.
In further support of the function of MglA as an alternative lipid II synthase, we tested if it could also take over the function of murG in another bacterium. For this, we chose the model streptomycete S. coelicolor M145, which is a distinct genus within the Streptomycetaceae (31, 51) but lacks a copy of mglA. Importantly, murG could be readily depleted using CRISPRi in strains expressing mglA from a constitutive promoter, while knockdown of murG in colonies of S. coelicolor harboring control plasmids led to very severe growth defects. This not only validates our data that mglA encodes a lipid II synthase but also indicates that this is a more universal phenomenon that does not occur only in specific strains of Kitasatospora or connect to strains that have the capacity to produce natural wall-less cells. Furthermore, it shows that no additional Kitasatospora genes are required to allow mglA to functionally complement murG in Streptomyces.
We also attempted to delete murG and mglA simultaneously in alpha. While the single mutants were readily obtained, we never obtained strains that were completely devoid of both murG and mglA, despite many attempts. Like mycelia, L-forms are multinucleated cells, and some cells of the population retained mglA, most likely to ensure minimal levels of lipid II. Consistent with this idea is the finding that antibiotics that target lipid II, such as vancomycin, are lethal to alpha (our unpublished data). We hypothesize that this lethality is caused by depletion of the lipid carrier undecaprenyl diphosphate, which is also used in other pathways and which may be essential for these L-forms. Removing mglA in strains lacking murG virtually blocked the ability to switch to the filamentous mode of growth, whereas each of the single mutants switched as efficiently as the parental alpha strain. Thus, we show that MglA is an enzyme involved in cell wall metabolism, which appears to facilitate switching between a wall-deficient and a walled lifestyle.

MATERIALS AND METHODS
Strains and media. Bacterial strains used in this study are shown in Table 3. To obtain sporulating cultures of K. viridifaciens and S. coelicolor, strains were grown at 30°C for 4 days on MYM (52). For general cloning purposes, E. coli strains DH5a and JM109 were used, while E. coli ET12567 and SCS110 were Discovery of an Alternative Lipid II Synthase ® used to obtain unmethylated DNA. E. coli strains were grown at 37°C in LB medium supplemented with chloramphenicol (25 mg ml 21 ), ampicillin (100 mg ml 21 ), apramycin (50 mg ml 21 ), kanamycin (50 mg ml 21 ), or viomycin (30 mg ml 21 ), where necessary.
To support the growth of wall-deficient cells, strains were grown in liquid LPB medium while being shaken at 100 rpm or grown on solid LPMA medium at 30°C (32). To switch from the wall-deficient to the filamentous mode of growth, L-form colonies grown on LPMA for 7 days were streaked on MYM. If needed, mycelial colonies of switched strains were transferred after 4 days to liquid tryptic soy broth sucrose (TSBS) medium and grown for 2 days at 30°C, while being shaken at 200 rpm.
Construction of plasmids. All plasmids and primers used in this work are shown in Tables 4 and 5, respectively.
(i) Construction of the DivIVA localization construct pKR2. To localize DivIVA, we first created plasmid pKR1 containing a viomycin resistance cassette cloned into the unique NheI site of pIJ8630 (53). To this end, the viomycin resistance cassette was amplified from pIJ780 (54) with the primers vph-FW-NheI and vph-RV-NheI. Next, we amplified the constitutive gap1 promoter as a 450-bp fragment from the genome of S. coelicolor with the primers Pgap1-FW-BglII and Pgap1-RV-XbaI. We also amplified the divIVA coding sequence (the bp 11 to 11335 region relative to the start codon of divIVA [BOQ63_RS32500]) from the chromosome of K. viridifaciens using primers divIVA-FW-XbaI and divIVA-Nostop-RV-NdeI (55). Finally, the promoter and divIVA coding sequences were cloned into pKR1 as BglII/XbaI and XbaI/NdeI fragments, respectively, yielding plasmid pKR2.
(ii) Construction of the deletion constructs pKR3, pKR4, pKR8, pKR9, and pKR10. The divIVA mutant was created in K. viridifaciens using pKR3, which is a derivative of the unstable plasmid pWHM3 (56). In the divIVA mutant, nucleotides 1205 to 1349 relative to the start codon of divIVA were replaced with the loxP-apra resistance cassette as described previously (57). A similar strategy was used for the deletion of the partial dcw cluster (plasmid pKR4) and for the deletion of murG (plasmid pKR8) and mglA (plasmid pKR9). For the deletion of the partial dcw cluster, the chromosomal region from bp 1487 relative to the start of the ftsW gene (BOQ63_RS32460) until bp 1349 relative to the start of the divIVA gene was replaced with the apramycin resistance marker. For the deletion of murG (BOQ63_RS32465, located in the dcw cluster), bp 110 to 11077 relative to the start codon of murG were replaced with the loxP-apra resistance cassette, while for the mglA (BOQ63_RS12640) deletion, the chromosomal region from bp 118 to 11105 relative to the start of mglA was replaced DmurG strain in which mglA is replaced by the vph viomycin resistance cassette This work with the apramycin resistance marker. To construct the murG mglA double mutant, pKR10 was created, replacing the apramycin resistance cassette in pKR8 by a viomycin resistance cassette. To this end, the viomycin resistance cassette was amplified from pIJ780 (54) with the primers vph-Fw-EcoRI-HindIII-XbaI and vph-Rv-EcoRI-HindIII-XbaI. The viomycin resistance cassette contained on the PCR fragment was then cloned into pKR8 using XbaI, thereby replacing the apramycin cassette and yielding pKR10.
(iii) Construction of the complementation constructs pKR6 and pKR7. For complementation of divIVA under the control of the strong gap1 promoter (43), construct pKR6 was made. First, we created plasmid pKR5 with the strong gap1 promoter. The promoter region of gap1 (SCO1947) was amplified with the primers Pgap1-FW-BglII and Pgap1-RV-XbaI using S. coelicolor genomic DNA as the template. Next, the gap1 promoter was cloned as a BglII/XbaI fragment into the integrative vector pIJ8600 (53) to generate plasmid pKR5. Afterwards, the divIVA coding sequence was amplified from the genome of K. viridifaciens with the primers divIVA-XbaI-FW and divIVA-NdeI-RV. Finally, to create the plasmid pKR6, the XbaI/NdeI fragment containing the divIVA coding sequence was cloned in pKR5.
(iv) Construction of the mglA expression construct pGWS1379. A DNA fragment containing the modified ermE* promoter was obtained as an EcoRI/NdeI fragment from pHM10a (58), while mglA was amplified by PCR from K. viridifaciens chromosomal DNA using the primer pair mglA_F14_ENdeI and mglA_R11146_HX. The ermEp* promoter fragment and NdeI/XbaI-digested mglA were simultaneously cloned into EcoRI/XbaI-digested pSET152 to generate construct pGWS1378. The insert of pGWS1378 was then introduced as a PvuII fragment into EcoRV-digested pMS82 (59) to generate construct pGWS1379. This construct was then introduced into S. coelicolor M145 via protoplast transformation as described previously (60).
Transformation of L-forms. Transformation of alpha essentially followed the protocol for the rapid small-scale transformation of Streptomyces protoplasts (60), with the difference that cells (50 ml) from a mid-exponential growing L-form culture were used instead of protoplasts. Typically, 1 mg DNA was used for each transformation. Transformants were selected by applying an overlay containing the required antibiotics in P buffer after 20 h. Further selection of transformants was done on LPMA medium Microscopy. Strains grown in LPB or LPMA were imaged using a Zeiss Axio Lab A1 upright microscope equipped with an AxioCam MRc digital camera. A thin layer of LPMA (without horse serum) was applied to the glass slides to immobilize the cells prior to the microscopic analysis.
(i) Fluorescence microscopy. Fluorescence microscopy pictures were obtained with a Zeiss Axioscope A1 upright fluorescence microscope equipped with an AxioCam MRc5 camera. Aliquots of 10ml of live cells were immobilized on top of a thin layer of LPMA (without horse serum) prior to analysis. Fluorescent images were obtained using a 470/40-nm-band-pass excitation and a 505/560-nm-band-pass detection, using an 100Â-numerical-aperture 1.3 objective. To obtain a sufficiently dark background, the background of the images was set to black. These corrections were made using Adobe Photoshop CS5.
(ii) Time-lapse microscopy. To visualize the proliferation of alpha, cells were collected and resuspended in 300 ml LPB (containing 4 to 22% sucrose) and placed in the wells of a chambered 8-well m-slide (ibidi). Cells were imaged on a Nikon Eclipse Ti-E inverted microscope equipped with a confocal spinning disk unit (CSU-X1) operated at 10,000 rpm (Yokogawa), using a 40Â Plan Fluor lens (Nikon), and illuminated in bright field. Images were captured every 2 min for 10 to 15 h by an Andor iXon Ultra 897 high-speed electron microscope charge-coupled device (EM-CCD) camera (Andor Technology). Zstacks were acquired at 0.2-to 0.5-mm intervals using an NI-DAQ-controlled Piezo element. During imaging, wall-less cells were kept at 30°C using an INUG2E-TIZ stage top incubator (Tokai Hit).
(iii) Electron microscopy. For transmission electron microscopy, L-forms obtained from a 7-day-old liquid-grown alpha culture were trapped in agarose blocks prior to fixation with 1.5% glutaraldehyde and a postfixation step with 1% OsO 4 . Samples were embedded in Epon and sectioned into 70-nm slices. Samples were stained using uranyl acetate (2%) and lead citrate (0.4%), if necessary, before being imaged using a JEOL 1010 or an FEI Tecnai 12 BioTWIN transmission electron microscope.
DivIVA detection using Western analysis. To detect DivIVA using Western analysis, the biomass of L-form strains was harvested after 7 days of growth in LPB medium, while the biomass of mycelial strains was obtained from liquid-grown TSBS cultures after 17 h. Cell pellets were washed twice with 10% phosphate-buffered saline (PBS), after which they were resuspended in 50 mM HEPES, pH 7.4, 50 mM NaCl, 0.5% Triton X-100, 1 mM PFMS (phenylmethylsulfonyl fluoride), and P8465 protease inhibitor cocktail (Sigma). The cells and mycelia were disrupted with a Bioruptor Plus sonication device (Diagenode). Complete lysis was verified by microscopy, after which the soluble cell lysate was separated from the insoluble debris by centrifugation at 13,000 rpm for 10 min at 4°C. The total protein concentration in the cell lysates was quantified by a bicinchoninic acid (BCA) assay (Sigma-Aldrich). Equal amounts of total proteins were separated with SDS-PAGE using 12.5% gels. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes (GE Healthcare) with the Mini Trans-Blot Cell (Bio-Rad Laboratories) according to the manufacturer's instructions. DivIVA was detected using a 1:5,000 dilution of polyclonal antibodies raised against Corynebacterium glutamicum DivIVA (kindly provided by Marc Bramkamp). The secondary antibody, anti-rabbit IgG conjugated to alkaline phosphatase (Sigma), was visualized with the BCIP (5-bromo-4-chloro-3-indolylphosphate)-NBT (nitroblue tetrazolium) color development substrate (Promega).
Isolation of cytoplasmic peptidoglycan precursors. For the cytoplasmic PG precursor isolation and identification, we used a modification of the method previously described (61). The alpha strain and the divIVA and dcw mutants were grown in LPB for 7 days, while the wild-type K. viridifaciens strain was grown for 3 days in a modified version of LPB lacking sucrose. The cells were harvested by centrifugation at 4°C and washed in 0.9% NaCl. Cells were extracted with 5% cold trichloric acid (TCA) for 30 min at 4°C. The extracts were centrifuged at 13,000 rpm for 5 min at 4°C, after which the supernatants were desalted on a Sephadex G-25 column (Illustra NAP-10 columns; GE Healthcare, Pittsburgh, PA) and concentrated by rotary evaporation. The concentrated precursors were dissolved in 200 ml high-performance liquid chromatography (HPLC)-grade water.
Peptidoglycan extraction. The peptidoglycan architecture was analyzed as described previously (62). Mycelia of the wild-type strain, alpha, and the dcw mutant complemented with divIVA were grown on top of cellophane discs on modified LPMA medium lacking sucrose and horse serum. Following growth, the mycelial mass was removed from the cellophane, washed in 0.1 M Tris-HCl, pH 7.5, and lyophilized. Ten milligrams of the lyophilized biomass was used for PG isolation. Therefore, the biomass was boiled in 0.25% SDS in 0.1 M Tris-HCl, pH 6.8, thoroughly washed, sonicated, and treated with DNase, RNase, and trypsin. Inactivation of these enzymes was performed by boiling the samples, followed by washing them with water. Wall teichoic acids were removed with 1 M HCl (63). PG was digested with mutanolysin and lysozyme. Muropeptides were reduced with sodium borohydride, and the pH was adjusted to 3.5 to 4.5 with phosphoric acid.
LC-MS analysis of PG precursors and muropeptides. The LC-MS setup consisted of a Waters Acquity UPLC system (Waters, Milford, MA, USA) and an LTQ Orbitrap XL hybrid ion Trap-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with an Ion Max electrospray source. Chromatographic separation of muropeptides and precursors was performed on an Acquity ultraperformance LC (UPLC) HSS T3 C 18 column (1.8 mm, 100 Å, 2.1 by 100 mm). Mobile phase A consisted of 99.9% H 2 O and 0.1% formic acid, while mobile phase B consisted of 95% acetonitrile, 4.9% H 2 O, and 0.1% formic acid. All solvents used were of LC-MS grade or better. The flow rate was set to 0.5 ml min 21 . The binary gradient program consisted of 1 min of 98% phase A, 12 min of from 98% A to 85% A, and 2 min of from 85% A to 0% A. The column was then flushed for 3 min with 100% phase B, after which the gradient was set to 98% and the column was equilibrated for 8 min. The column temperature was set to 30°C, and the injection volume used was 5 ml. The temperature of the autosampler tray was set to 8°C. Data were collected in the positive electrospray ionization (ESI) mode, with a scan range of m/z 500 to 2,500 in high-range mode. The resolution was set to 15,000 (at m/z 400).
Discovery of an Alternative Lipid II Synthase ® Sequence homology analysis of dcw gene clusters. The homology search of the different dcw clusters was done using MultiGeneBlast (64). The query used for the search was the dcw cluster from Streptomyces coelicolor A3(2), for which the required sequences were obtained from the Streptomyces Annotation Sever (StrepDB). The homology search included the loci from SCO2077 (divIVA) to SCO2091 (ftsL). A database was constructed with genome assemblies obtained from the NCBI. The analyzed species have the following NCBI accession numbers: AL645882.2 [S. coelicolor A3(2)], NZ_MPLE00000000.1 (Kitasatospora viridifaciens DSM40239), CP000480 (Mycobacterium smegmatis MC2 155), AL123456 (Mycobacterium tuberculosis H37Rv), CP014279 (Corynebacterium stationis ATCC 6872), BX927147 (Corynebacterium glutamicum ATCC 13032), AL009126 (Bacillus subtilis subsp.168), U00096 (Escherichia coli K-12), CP000253.1 (Staphylococcus aureus NTC8325), and AE007317 (Streptococcus pneumoniae R6). In the homology search, the BLAST parameters were set to a minimal sequence coverage of 25% and a minimal identity of 30%. The first 11 hits of the MultiGeneBlast output are shown in Fig. S1, where homologue genes are represented by arrows with the same colors.
Phylogeny analysis of Streptomyces and Kitasatospora species. A set of 1,050 Streptomyces and Kitasatospora genomes was downloaded from the NCBI database by querying the fasta files in combination with the taxonomic identifier. To this set, 116 unpublished draft genome sequences of an in-house collection of actinomycetes were added (65). Complete protein sets encoded within the genomes of Streptomyces and Kitasatospora spp. were extracted. The Pfam domains of four housekeeping proteins, AtpD (ATP synthase subunit beta), RecA (recombinase A), TrpB (tryptophan synthase beta chain), and GyrB (DNA gyrase subunit B), were retrieved from https://pfam.xfam.org/ and are annotated as PF00213, PF00154, PF06233, and PF00204, respectively. Using the selected Pfam domains, the Hmmsearch program of the HMMER v3.0 package (66) was employed to identify analogous proteins within the chosen species. MAFFT was used to perform a multiple-sequence alignment (67). Aligned sequences were concatenated using SeqKit (68), and maximum-likelihood phylogenetic trees were calculated with RAxML (69). iTOL (70) was used for the visualization of the phylogenetic tree.
Detection of murG genes in Streptomyces and Kitasatospora species. MurG domains were predicted using the Pfam database (44). Proteins with the predicted MurG domains were used to search in the complete protein sets encoded within the extracted genomes using Hmmsearch. Instead of a multiple-sequence alignment, each protein domain sequence was aligned to its profile hidden Markov model from Pfam using the hmmalign tool (71). For each protein, a pairwise distance was calculated for all detected MurG proteins, and the threshold was set at 0.9. Network visualizations were constructed using Cytoscape (v. 3.7.1) (72).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.