Identification of the Wzx flippase, Wzy polymerase and sugar‐modifying enzymes for spore coat polysaccharide biosynthesis in Myxococcus xanthus

The rod‐shaped cells of Myxococcus xanthus, a Gram‐negative deltaproteobacterium, differentiate to environmentally resistant spores upon starvation or chemical stress. The environmental resistance depends on a spore coat polysaccharide that is synthesised by the ExoA‐I proteins, some of which are part of a Wzx/Wzy‐dependent pathway for polysaccharide synthesis and export; however, key components of this pathway have remained unidentified. Here, we identify and characterise two additional loci encoding proteins with homology to enzymes involved in polysaccharide synthesis and export, as well as sugar modification and show that six of the proteins encoded by these loci are essential for the formation of environmentally resistant spores. Our data support that MXAN_3260, renamed ExoM and MXAN_3026, renamed ExoJ, are the Wzx flippase and Wzy polymerase, respectively, responsible for translocation and polymerisation of the repeat unit of the spore coat polysaccharide. Moreover, we provide evidence that three glycosyltransferases (MXAN_3027/ExoK, MXAN_3262/ExoO and MXAN_3263/ExoP) and a polysaccharide deacetylase (MXAN_3259/ExoL) are important for formation of the intact spore coat, while ExoE is the polyisoprenyl‐phosphate hexose‐1‐phosphate transferase responsible for initiating repeat unit synthesis, likely by transferring N‐acetylgalactosamine‐1‐P to undecaprenyl‐phosphate. Together, our data generate a more complete model of the Exo pathway for spore coat polysaccharide biosynthesis and export.

morphology and novel characteristics. The best-studied examples of environmentally induced bacterial differentiation include spore formation in three phylogenetically widely separated species Bacillus subtilis (Tan & Ramamurthi, 2014), Streptomyces coelicolor (Flärdh & Buttner, 2009) and Myxococcus xanthus (Konovalova, Petters, & Søgaard-Andersen, 2010). While the spore formation process varies among these three species, the resulting spores have in common that they have a spore coat that confers resistance to environmental stress.
In B. subtilis, sporulation is initiated in response to starvation and depends on an unusual cell division event in which the division septum is placed asymmetrically close to a cell pole, resulting in the formation of a large mother cell and a smaller forespore. Next, the mother cell engulfs the forespore and lysis of the mother cell finally releases the mature spore (Tan & Ramamurthi, 2014). The spore envelope, partly generated by the mother cell and partly by the forespore, consists of a multilayered structure comprising from inside to outside: the cytoplasmic membrane, peptidoglycan (PG), an outer membrane, which is originally the cytoplasmic membrane of the mother cell and a proteinaceous coat (Driks & Eichenberger, 2016;McKenney, Driks, & Eichenberger, 2013). In response to nutrient depletion, S. coelicolor generate aerial hyphae, and here, multiple synchronous cell divisions give rise to the spores (Flärdh & Buttner, 2009;Sigle, Ladwig, Wohlleben, & Muth, 2015). The spore envelope of S. coelicolor is less well-studied but contains PG, a proteinaceous sheath made of chaplins and rodlins and spore wall teichoic acids (Flärdh & Buttner, 2009;Sigle et al., 2015). In the Gram-negative deltaproteobacterium M. xanthus, sporulation is also typically induced by starvation (Konovalova et al., 2010). However, in this bacterium, spores are formed independently of a cell division event and during the sporulation process, PG is replaced by a spore coat consisting mainly of polysaccharide. Here, we focus on the identification of proteins important for formation of the spore coat polysaccharide in M. xanthus.
In Wzx/Wzy pathways, the individual repeat units of the polysaccharide chain are synthesised in the cytoplasm on the lipid carrier undecaprenyl-phosphate (Und-P) in a process that is primed by a polyisoprenyl-phosphate hexose-1-phosphate transferase (PHPT) or a polyisoprenyl-phosphate N-acetylhexosamine-1phosphate transferase (PNPT). Next, specific glycosyltransferases (GTs) transfer the additional sugar building blocks from nucleotide-sugar donors to the Und-PP-sugar primer molecule to generate the Und-PP-repeat unit, which can be further modified by additional enzymes. Individual repeat units are transported across the inner membrane (IM) by the Wzx flippase, assembled into the polysaccharide by the Wzy polymerase together with a polysaccharide co-polymerase (PCP) protein and transported across the OM by a Wza OM polysaccharide export (OPX) protein . In the Exo pathway (Figure 1a), ExoE is a predicted PHPT responsible for priming synthesis of individual repeat units (Holkenbrink et al., 2014). The integral membrane protein ExoC together with the cytoplasmic ExoD tyrosine kinase form part of a bipartite Wzc protein of the PCP-2 family, in which ExoD (formerly BtkA (Kimura, Yamashita, Mori, Kitajima, & Takegawa, 2011)) is thought to participate in regulating ExoC activity (Holkenbrink et al., 2014;Kimura et al., 2011). ExoA (formerly FdgA [Ueki & Inouye, 2005]) is a homolog of Wza OPX proteins (Holkenbrink et al., 2014). ExoG and ExoI are N-acetyltransferase homologs that could be involved in modifying sugars before or after incorporation into the Und-PP-repeat units before export; ExoH is homologous to aminotransferases, ExoF is a putative gluconeogenesis factor and ExoB is an OM β-barrel protein of unknown function (Holkenbrink et al., 2014). All Exo proteins except for ExoF are essential for sporulation and synthesis of an intact spore coat polysaccharide (Holkenbrink et al., 2014;Licking et al., 2000;Ueki & Inouye, 2005). Generally, Wzc proteins of the PCP-2 family are components of Wzx/Wzy-dependent pathways for polysaccharide synthesis and export (Morona, Purins, Tocilj, Matte, & Cygler, 2009) supporting the notion that the ExoA-I proteins are part of a Wzx/Wzy pathway. Notably, such an Exo pathway is incomplete and lacks several key enzymes including the GTs that add sugars from nucleotide-sugar donors to the Und-PP-sugar primer molecule, the Wzx flippase and the Wzy polymerase ( Figure 1a).
Here, we report the identification of two additional gene clusters encoding seven proteins that have homology to enzymes involved in polysaccharide synthesis and/or modification, and show that they are essential for sporulation and by implication for synthesis of the spore coat polysaccharide. We identify MXAN_3260 as the Wzx flippase (renamed ExoM) and MXAN_3026 (renamed to ExoJ) as the Wzy polymerase. We also identify five additional proteins important for spore coat polysaccharide synthesis including three GTs and determine the nucleotide sugar specificity of the ExoE priming enzyme, thus, generating a more complete model of the Exo pathway for spore coat polysaccharide biosynthesis.

| Identification of two loci encoding a Wzx flippase, a Wzy polymerase and other proteins involved in polysaccharide synthesis
To identify missing components for spore coat polysaccharide biosynthesis, we used a two-pronged strategy. First, as polysaccharide biosynthesis genes are often clustered (Rehm, 2010) (Giglio, Zhu, Klunder, Kummer, & Garza, 2015;Kimura et al., 2011;Licking et al., 2000;Müller, Treuner-Lange, Heider, Huntley, & Higgs, 2010;Ueki & Inouye, 2005;Wartel et al., 2013), we identified those candidate genes whose transcription pattern was similar to that of the exoA-I and nfsA-H genes during chemically induced sporulation using published data (Müller et al., 2010). Among the seven candidate genes, only the genes for the Wzx homolog MXAN_3260 and the Wzy_C domain protein MXAN_3026 were upregulated (Figure 1b) suggesting these two proteins could be the missing Wzx flippase and Wzy polymerase, respectively, for spore coat polysaccharide synthesis.
Further, mutation of MXAN_1035 was previously reported to only slightly affect spore formation (Holkenbrink et al., 2014), while MXAN_1052 is in the same polysaccharide biosynthesis gene cluster as MXAN_1035, and therefore, likely also not involved in spore coat synthesis. MXAN_7416 and MXAN_7442 are part of the eps locus, which is important for EPS synthesis (Lu et al., 2005). Finally,  (Valvano, 2011). The Wzy polymerase candidate MXAN_3026 F I G U R E 1 Model and expression analysis of gene clusters involved in spore coat polysaccharide synthesis. (a) Current model for spore coat polysaccharide synthesis in M. xanthus (see text). Proteins indicated in stippled lines have not been identified. Note that the number of GTs is unknown. Right panel, the colour code indicates predicted functions and is used throughout all figures. (b,e) Transcription pattern of indicated genes in response to chemical induction of sporulation with 0.5 M glycerol (+gly) based on data from (Müller et al., 2010) and shown as log2fold change compared to cells in the absence of glycerol (-gly). Note that the expression pattern of the genes in the exo I gene cluster were previously published (Müller et al., , 2010 and are included for comparison. (c) Domain and TMH prediction of ExoM (MXAN_3260) and ExoJ (MXAN_3026), where domains are indicated in red and green, respectively. Grey rectangles indicate TMHs, red and black lines indicate periplasmic and cytoplasmic regions, respectively. Numbers indicate domain borders. (d) exo gene cluster I, II and III. Genes are drawn to scale and MXAN number and gene name are indicated (see also Table S1). Gene orientation is indicated by the direction of arrows. Note that the exoJ-K, exoL-M and exoN-P genes are likely in operons. The exoA-I genes form an operon . Gene coordinates are relative to the first nucleotide of the first gene in an operon and are indicated by the same gene colour, except for exoE, for which coordinates are indicated relative to its first nucleotide. DNA fragments comprising promoter and structural gene used in complementation experiments are indicated by a line above the corresponding region/s contains a PF04932 (Wzy_C) domain (Figure 1c), which is also found in O-antigen ligases, Wzy polymerases and O-linked oligosaccharyltransferases (Hug & Feldman, 2011;Schild, Lamprecht, & Reidl, 2005) and multiple TMHs whose topology depended on the prediction programme used (Figure 1c).
MXAN_3260 and MXAN_3026 are encoded by genes in two distinct gene clusters that we renamed to exo gene cluster III and II, while we renamed the exoA-I gene cluster to Exo gene cluster I  (Müller et al., 2010). As discussed in details below, exo gene cluster I and III make up one cluster in Vulgatibacter incomptus and all three clusters are present as one cluster in Anaeromyxobacteraceae supporting that the gene products of the three clusters may function together.
As shown below, the genes of exo gene cluster II and III are important for sporulation and our data support that the encoded proteins form part of the same machinery. For simplicity and to facilitate identification of the genes throughout this study, we renamed MXAN_3026, MXAN_3027 and MXAN_3259-MXAN_3263 to ExoJ-P following the Exo nomenclature (Holkenbrink et al., 2014;Müller et al., 2012) (Figure 1d).

| ExoJ-ExoP are important for chemically induced sporulation
To determine the importance of the seven exoJ-P genes in sporulation, we generated in-frame deletion mutations in each of the genes separately and determined their importance for sporulation using chemical induction (Figure 2a Cells of the ΔexoE mutant, which cannot produce spore coat polysaccharide (Holkenbrink et al., 2014), served as a negative control.
As previously described (Holkenbrink et al., 2014), ΔexoE cells initially shortened becoming ovoid by 4 hr; by 24 hr, most ΔexoE cells had reverted to a non-phase-bright rod-shape while a few remained non-phase-bright ovoid-shaped or were branched and non-phasebright. ΔexoE cells were not resistant to heat and sonic treatment.
The ΔexoM and ΔexoJ mutants formed large round cells by 4 hr; by 24 hr, many cells had reverted to rod-shape, however, a significant fraction were ovoid, branched or formed large spheres; cells of these two mutants were neither phase-bright nor resistant to heat and sonic treatment. The ΔexoK, ΔexoO and ΔexoP mutants had cell morphologies and sporulation defects similar to those of the ΔexoM and ΔexoJ mutants. By 4 hr, ΔexoL cells were ovoid; by 24 hr, many of these cells had reverted to rod-shape, however, a significant fraction were ovoid and a few were branched or had turned into large spheres. None of these cells were phase-bright or resistant to heat and sonic treatment. Finally, the ΔexoN mutant formed phase-bright spores that were resistant to heat and sonic treatment but at a twofold reduced level compared to WT; moreover, a significant fraction of cells at 24 hr were non-phase-bright rod-shaped or ovoid while a small fraction were branched or formed large spheres. Sporulation of all eight in-frame deletion mutants was restored by ectopic expression of the corresponding full-length gene under the control of the native promoter (P nat ) on a plasmid integrated in a single copy at the Mx8 attB site (Figures 1d and 2).
We conclude that all seven ExoJ-P proteins, except ExoN, are essential for chemically induced sporulation. These data agree with the idea that ExoM is the Wzx flippase, ExoJ the Wzy polymerase, ExoK/-O/-P GTs and ExoL a polysaccharide deacetylase essential for synthesis of an intact spore coat polysaccharide and are also consistent with a previous report that an insertional mutation in exoJ caused a sporulation defect . Of note, cells lacking ExoE, which are blocked in the initiation of repeat unit synthesis because they lack the PHPT for spore coat polysaccharide synthesis, mostly reverted from ovoid at 4 hr to rod-shaped at 24 hr while the remaining mutants, which would be impaired in spore coat polysaccharide synthesis at later steps, formed morphologically highly abnormal cells (ovoid-shaped, branched or large spheres) by 24 hr (Figure 2b) (see Section 3).

| Loss of ExoE and ExoJ-ExoP neither affects EPS and LPS synthesis, cell morphology nor motility
In addition to the spore coat polysaccharide, M. xanthus synthesises two additional polysaccharides, that is, EPS and LPS, both of which are important for fruiting body formation. Because, blocking synthesis of one glycan polymer can affect synthesis of other polymers including PG by sequestration of Und-P through accumulation of Und-PP intermediates (Burrows & Lam, 1999;Jorgenson, Kannan, Laubacher, & Young, 2016;Ranjit & Young, 2016;Valvano, 2008), we determined whether lack of Exo proteins interferes with EPS, LPS or PG synthesis during growth.
EPS synthesis was tested using nutrient-rich agar supplemented with Congo red. As a result of binding of the dye to EPS, WT colonies acquired a red colour while the negative control, a ΩdifE mutant (Yang, Geng, Xu, Kaplan, & Shi, 1998), did not ( Figure 3a). All exo mutants tested accumulated EPS at WT level. LPS was extracted from cell extracts from growing cells and detected by Emerald green staining. For WT as well as all tested exo mutants, a fast running Moreover, these observations support that the Exo machinery is dedicated to spore coat synthesis and that the EPS, LPS and PG machineries function independently of the Exo proteins during growth.
M. xanthus possesses two distinct motility systems that are important for fruiting body formation; one of them depends on type IV pili (T4P) and the other one depends on the Agl/Glt gliding motility complexes (Schumacher & Søgaard-Andersen, 2017;Zhang, Ducret, Shaevitz, & Mignot, 2012). EPS and LPS are important for motility (Lu et al., 2005;Pérez-Burgos et al., 2019). On 0.5% agar, which favours T4P-dependent motility, WT displayed the flares characteristic of this type of motility; by contrast, the negative control ΔpilA strain, which lacks the major pilin of T4P (Wu & Kaiser, 1996), did not. On 1.5% agar, which favours gliding motility, single cells were F I G U R E 2 Chemically induced sporulation in Δexo mutants. (a) Sporulation was induced by addition of glycerol to a final concentration of 0.5 M. At 0, 4 and 24 hr, cell morphology was observed by phase contrast microscopy. In images labelled resistant spores, cells were exposed to sonic and heat treatment before microscopy. Sporulation frequency after sonic and heat treatment is indicated as the mean ± SD from three biological experiments relative to WT. P nat is short for the native promoter and is used throughout the study. Scale bars, 5 µm. (b) Quantification of cell morphology at 24 hr before sonic and heat treatment relative to WT (100%); n = 300 combined from three biological replicates (a) (b) observed at the colony edge of WT, while the ΔaglQ mutant, which lacks an essential component of the gliding machinery motor (Nan et al., 2013;Sun, Wartel, Cascales, Shaevitz, & Mignot, 2011), had a smooth edge. By contrast, all tested exo mutants were indistinguishable from WT ( Figure S1; Table 1). Together, these results indicate that loss of the Exo machinery does not interfere with motility during growth.

| ExoJ-M and ExoO-P are important for starvation-induced sporulation
Having shown that the Exo proteins are neither important for EPS, LPS or PG biosynthesis nor for motility during growth, we asked whether the exo mutants are able to generate starvation-induced spores during fruiting body formation. Previous analyses showed F I G U R E 3 EPS and LPS synthesis, and cell length determination in the Δexo mutants. (a) Determination of EPS accumulation. 20 μl of cell suspensions at 7 × 10 9 cells/ml were spotted on 0.5% agar supplemented with 0.5% CTT and Congo red and incubated at 32°C for 24 hr. The ΩdifE mutant served as a negative control. (b) Detection of LPS O-antigen. Extracted LPS samples from the same number of cells were separated by SDS-PAGE and visualised using Pro-Q Emerald 300. The ΔwbaP mutant served as a negative control. (c) Cell length determination. Cell length distribution is shown in a violin plot. Each violin indicates the probability density of the data at different cell lengths. Mean and median values are represented by a continuous and dashed line, respectively. For each strain, mean cell length ± standard deviation is indicated; n = 800 combined from four biological replicates. Samples were compared using a Mann-Whitney test, * indicates p value < .05 and ns, not significant that ΔexoA, ΩexoC, ΔexoC, ΩexoD and ΩexoJ mutants are generally able to form fruiting bodies but have a reduced sporulation efficiency (Kimura et al., 2011;Licking et al., 2000;Müller et al., 2012;Ueki & Inouye, 2005). Because, these experiments were performed under different conditions, we tested the developmental proficiency of eight exo mutants on TPM 1.5% agar and under submerged conditions.
Similar to WT, all eight exo mutants had aggregated to form fruiting bodies by 24 hr under both conditions ( Figure S2). However, all exo mutants with the exception of the ∆exoN mutant had a strong sporulation defect ( Figure S2; Table 1). Importantly, the sporulation defects were partially or completely complemented by ectopic expression of the relevant full-length gene from the native promoter (Figures 1d and S2; Table 1). Because, the sporulation defects of the exo mutants during chemically induced sporulation were fully complemented, we speculate that the partial complementation observed for some of the exo mutants is caused by insufficient expression of the relevant gene during starvation. As in the case of chemically induced sporulation, we conclude that ExoJ-P proteins, with the exception of ExoN, are essential for starvation-induced sporulation and by implication in formation of an intact spore coat.

| ExoE has GalNAc-1-P transferase activity
ExoE was suggested to be the PHPT homolog that initiates repeat unit biosynthesis for spore coat polysaccharide biosynthesis (Holkenbrink et al., 2014); however, it is unknown which sugar ExoE transfers to Und-P. Similarly to the Escherichia coli WcaJ Ec , which transfers Glc-1-P to Und-P and the Salmonella enterica  (Lehrman, 1994). In WcaJ Ec , the fifth TMH forms a helixbreak-helix structure and does not fully span the IM resulting in the cytoplasmic localisation of the C-terminal catalytic domain. This depends on the residue P291 that forms part of a DX 12 P motif highly conserved among PHPTs (Furlong et al., 2015). ExoE also carries the DX 12 P motif and contains all the other conserved essential residues important for catalytic activity that have been identified in the C-terminal catalytic domain of WbaP Se (Patel, Furlong, & Valvano, 2010) (Figures 4b and S3). Thus, ExoE is a PHPT with a predicted topology similar to that described for WcaJ Ec and WbaP Se .
Compositional analysis of the spore coat polysaccharide showed that it is composed of 1-3-, 1-4-linked GalNAc, 1-4-linked Glc (GalNAc:Glc ratio 17:1) and glycine (Holkenbrink et al., 2014 which lacks the ability to produce colanic acid as previously reported (Patel et al., 2012;Pérez-Burgos et al., 2019). For this experiment, native ExoE was synthesised in the ΔwcaJ Ec strain also containing pWQ499, which encodes the RcsA regulator that increases the production of colanic acid (Furlong et al., 2015). Cells growing in the absence and presence of arabinose were examined for a glossy and mucoid colony phenotype characteristic of colanic acid capsule production ( Figure 4c). Only cells containing the FLAG WcaJ Ec -encoding plasmid pLA3 exhibited the distinct mucoid phenotype representing colanic acid production, whereas the strain synthesising native ExoE or containing the pBAD24 vector control did not display this phenotype. As shown in Figure S4a, FLAG ExoE accumulated although at a slightly lower level than FLAG WcaJ Ec . Together, these results indicate ExoE is not a Glc-1-P transferase.
PHPT proteins were initially described as hexose-  Merino et al., 2011;Power et al., 2000). Because, ExoE did not transfer Glc-1-P, we next investigated whether ExoE could transfer GalNAc-1-P. To this end, we performed heterologous expression experiments in E. coli in which we first tested for transfer of GlcNAc-1-P and subsequently for transfer of GalNAc-1-P. GlcNAc-1-P transferase activity can be tested using the E. coli strain MV501, which has a transposon insertion in wecA Ec . As described, WecA Ec is a PNPT that uses UDP-GlcNAc for initiating synthesis of O7 polysaccharide antigen (Alexander & Valvano, 1994). Native ExoE in   Figure S3. (c-g) Heterologous complementation experiments to characterise ExoE specificity.
(c) E.coli XBF1 (ΔwcaJ Ec ) containing pWQ499 (rcsA + ) was transformed with the indicated plasmids and plated on LB agar in the absence or presence of 0.2% arabinose (ara) to induce gene expression. Cells were incubated for 24 hr at 37°C, and then, 24 hr at room temperature before scoring the mucoid phenotype. (d,e) Silver-stained polyacrylamide gels of LPS extracted from the E. coli wecA::Tn10 mutant strain MV501 carrying the indicated plasmids. In (e), MV501 also contained the plasmid pGEMT-Gne, which encodes the UDP-GlcNAc/UDP-GalNAc epimerase Gne Ah . VW187 is the parental wecA + strain. Arabinose was added as indicated. protein accumulating (Figures 4f,g and S4a). By contrast, the control plasmids pJD132 and pSM13, which encode WbaP Ec O9:K30 of E. coli and WbaP Se of S. enterica, respectively, both restored O-antigen synthesis (Figure 4f,g). Collectively, the heterologous expression experiments support that ExoE has specificity for UDP-GalNAc but lacks specificity for UDP-Glc, UDP-GlcNAc and UDP-Gal. These data, together with the observation that the spore coat polysaccharide contains Glc and GalNAc, suggest that ExoE is a GalNAc-1-P transferase forming Und-PP-GalNAc and that GalNAc is likely the first sugar added to Und-P during the biosynthesis of the spore coat polysaccharide repeat.

| The exo and nfs gene clusters co-occur only in a subset of sporulating Myxococcales
Because, the majority of the members of the order Myxococcales can sporulate (Reichenbach, 1999), we hypothesised that the Exo and Nfs machineries for formation of the rigid spore coat would be conserved in Myxoccocales. We, therefore, searched for orthologs of each Exo and Nfs protein in Myxococcales with fully sequenced genomes using a reciprocal best BlastP hit method (Section 4).

| D ISCUSS I ON
Cells of M. xanthus generate at least three different polysaccharidic cell surface structures, namely LPS, EPS and the spore coat polysaccharide. Here, we focused on identifying the proteins that would function together with the ExoA-I proteins in spore coat polysaccharide biosynthesis and export.
Using bioinformatics and gene co-expression analyses, we identified two loci that encode proteins important for sporulation.
One of them, named the exo gene cluster II, encodes a homolog of is only partially required, are essential for sporulation, and therefore, predicted to function in formation of the intact spore coat polysaccharide. Based on these findings, we propose a revised model for spore coat polysaccharide biosynthesis ( Figure 6).
The M. xanthus spore coat polysaccharide is composed of 1-3-, 1-4-linked GalNAc, 1-4-linked Glc and glycine (Holkenbrink et al., 2014) and with the latter proposed to form glycine bridges between polysaccharide chains (Holkenbrink et al., 2014). The spore coat polysaccharide is also acetylated (Filer, White, Kindler, & Rosenberg, 1977b;Holkenbrink et al., 2014). However, the precise structure of the spore coat polysaccharide is unknown. The data of Holkenbrink et al. (Holkenbrink et al., 2014), together with our results, suggest a model in which ExoE is the PHPT homolog responsible for the first step in repeat unit synthesis by catalysing the transfer of a sugar-1-P donor to Und-P (Holkenbrink et al., 2014). Here, we demonstrate that ExoE is functionally similar to WecP Ah , a GalNAc-1-P transferase from A. hydrophila (Merino et al., 2011) Table S2. S, species that form spores; (S), tested for sporulation but with ambiguous results; NT, sporulation not tested; N, sporulation tested and not observed (dos Santos et al., 2014;Fudou, Jojima, Iizuka, & Yamanaka, 2002;Garcia et al., 2014;Garcia & Müller, 2014a, 2014bMohr, Garcia, Gerth, Irschik, & Müller, 2012, Sanford, Cole, & Tiedje, 2002Yamamoto et al., 2014). For the exo, nfs and glt gene clusters, a reciprocal best BlastP hit method was used to identify orthologs. Generally, the exo gene clusters are marked in green (cluster II), blue (cluster I) and red (cluster III) and the nfs and glt gene clusters in orange (nfs), light brown (glt cluster I) and dark brown (glt cluster II). To evaluate gene proximity and cluster conservation, 10 genes were considered as the maximum distance for a gene to be in a cluster. Genes found in the same cluster are marked with the same colour. If two or three gene clusters are within a distance of <10 genes, all genes are marked in the same colour (e.g., two of the exo clusters in V. incomptus, all three exo clusters in Anaeromyxobacteraceae and the glt clusters in Anaeromyxobacteraceae, Nannocystineae and Sorangineae). Light grey indicates a conserved gene that is found somewhere else on the genome (>10 genes away from a cluster); dark grey and black indicate conservation of the marked genes but in a separate cluster; a cross indicates no homolog found. Note that the nfsB gene in L. luteola DSM 27648 is found in close proximity to the gltC, glG and gltI genes and is most likely a gltB homolog. (b) nfs and glt gene clusters in M. xanthus. Genes are not drawn to scale, MXAN number or gene name are indicated and gene orientation is indicated by arrows. The tree in A was prepared in MEGA7 (Kumar, Stecher, & Tamura, 2016) using the Neighbour-Joining method (Saitou & Nei, 1987). Bootstrap values (500 replicates) are shown next to the branches (Felsenstein, 1985) (a)

(b)
predict that the GTs ExoK, ExoO and ExoP transfer sugar building blocks to the repeat unit, which is likely a tetrasaccharide. The N-acetyltransferase homologs ExoG and ExoI, the aminotransferase homolog ExoH and the polysaccharide deacetylase homolog ExoL presumably modify sugars before or after incorporation into the repeat unit.
Based on the composition of the spore coat polysaccharide (Holkenbrink et al., 2014), we suggest that the GTs ExoK, ExoO and ExoP incorporate GalNAc and Glc into the repeat unit. Acetylation of the spore coat polysaccharide may involve the ExoG and ExoI N-acetyltransferases (but see also below). ExoL is the first identified potential polysaccharide deacetylase implicated in M. xanthus spore coat synthesis. Interestingly, phase-bright spores were not detected in the exoG and exoI mutants (Holkenbrink et al., 2014); similarly, the exoL mutant did not form phase-bright spores (here) suggesting that proper acetylation of the spore coat polysaccharide is important for its synthesis, stability and/or function. However, it is unknown which residue is modified by ExoG, ExoI and ExoL, and whether these proteins function on the same or independent targets. In Caulobacter crescentus, the polysaccharide deacetylase HfsH and the N-acetyltransferase HfsK affect acetylation of the holdfast polysaccharide; in the absence of any of these two proteins there is a defect in adhesive and cohesive properties of the holdfast polysaccharide without affecting its synthesis (Sprecher et al., 2017;Wan, Brown, Elliott, & Brun, 2013).
ExoH is predicted to be a pyridoxal phosphate-dependent (PLP) aminotransferase with a DegT/DnrJ/EryC1/StrS family domain (PF01041), which generally catalyses the transfer of an amino group from an amino acid to an amino acceptor (John, 1995). Similarly to the aminotransferase ArnB that transfers an amino group to arabinose in S. enterica (Noland et al., 2002) or the PLP aminotransferase PseC from Helicobacter pylori, which transfers an amino group to a sugar moiety prior to acetylation by PseH (Ud-Din, Liu, & Roujeinikova, 2015), we suggest that ExoH may add an amino group to monosaccharides before their incorporation into the repeat unit or modify sugar(s) in the repeat unit.
The glycine in the spore coat polysaccharide was proposed to form glycine bridges between polysaccharide chains (Holkenbrink et al., 2014). Holkenbrink et al. also suggested that glycine is added to the spore coat polysaccharide in the cytoplasm. Interestingly, a structure-based search with HHPred revealed that the closest ho-

molog of the N-acetyltransferases ExoG and ExoI is FemX from
Staphylococcus aureus, that is, for ExoG and ExoI, the probabilities of homology to FemX is 100% with an E-value of 1.9e-31 and 100% with an E-value of 4.1e-30. The Fem proteins belong to GCN5related N-acetyltransferases (GNAT) that generally transfer acetylated molecules to an amino acceptor of different target molecules including sugars (Favrot, Blanchard, & Vergnolle, 2016;Reith & Mayer, 2011;Ud-Din et al., 2015). In S. aureus, the FemA/B/X proteins add five glycine residues to the lysine in the stem peptide of the lipid II PG precursor using glycyl-charged tRNA molecules as substrates (Favrot et al., 2016). The pentaglycine modification crosslinks PG glycan chains (Favrot et al., 2016). Therefore, it is tempting to speculate that one or both of ExoG and ExoI rather than being involved in acetylation of the spore coat repeat unit could be involved in glycine addition to amino group(s) in the repeat unit. In this context, we speculate that the amino group added by ExoH could serve as an acceptor for glycine transfer. This is also consistent with the absence of glycine modified sugars after acid hydrolysis of the spore coat polysaccharide (Holkenbrink et al., 2014). Amino acid modified sugars, in this case with serine, have also been identified in the K40 capsular polysaccharide of E. coli O8 and the modification demonstrated to be essential for the polymerisation of the capsular repeat unit (Amor, Yethon, Monteiro, & Whitfield, 1999).
Two Exo proteins are important, but not essential, for formation of phase-bright spores and by implication spore coat synthesis: ExoF (Holkenbrink et al., 2014) and ExoN (here). ExoN is a putative serine O-acetyltransferase, which are commonly involved in the first step F I G U R E 6 Model of spore coat polysaccharide biosynthesis in M. xanthus. Colour code indicates predicted functions. Stippled lines indicate that the site of action is hypothetical and remains to be determined experimentally. See Section 3 for details. ExoE adds the first sugar of the repeat unit (blue), which is likely GalNAc, followed by addition of monosaccharides (Glc and GalNAc) by the GTs (yellow) of cysteine synthesis from serine. M. xanthus utilises amino acids and lipids as carbon and energy sources and does not grow on carbohydrates because it lacks required catabolic enzymes (Dworkin, 1962;Hemphill & Zahler, 1968;Watson & Dworkin, 1968). During glycerol-induced sporulation, genes for large portions of the tricarboxylic acid cycle are downregulated, whereas genes for the glyoxylate shunt and gluconeogenesis are upregulated (Müller et al., 2010), for example, the activity of at least six enzymes putatively involved in synthesis of the major spore coat component UDP-GalNAc increases in response to glycerol addition prior to shortening of cells (Filer, Kindler, & Rosenberg, 1977a). Given these metabolic changes, we speculate that ExoN may contribute to synthesis of monosaccharides or other metabolites important for spore coat polysaccharide synthesis, and therefore, without ExoN, cells may lack this precursor(s). The M. xanthus genome encodes two additional serine O-acetyltransferase homologs (MXAN_1572 and MXAN_7449), which may function redundantly with ExoN, and therefore, the ΔexoN mutant is still able to form some phase-bright spores. Similarly, the partially dispensable ExoF, which is a putative gluconeogenesis factor, has been suggested to be important for biosynthesis of activated sugar precursors (Holkenbrink et al., 2014).
After the repeat unit has been synthesised on the cytoplasmic side of the IM, translocation occurs via the Wzx flippase homolog ExoM. In the periplasm, the Wzy polymerase ExoJ elongates the chain with the help/control of the Wzc homolog formed by the integral membrane protein ExoC and the cytoplasmic ExoD tyrosine kinase, which could regulate ExoC activity (Kimura et al., 2011). A fascinating aspect of the sporulation process in M. xanthus is that the PG is degraded during spore morphogenesis (Bui et al., 2009). It has been suggested that the spore coat protects cells from bursting due to intracellular turgor in the absence of PG (Bui et al., 2009;Müller et al., 2012). Therefore, we predict that the removal of PG must be closely coordinated with synthesis of the spore coat polysaccharide. Previous research on chemical induction of sporulation in the exoA-I mutants showed that mutant cells initiate the sporulation process with cell shortening and widening. However, at a certain point the sporulation process is aborted and cells regain rod-shape even in the continued presence of glycerol. Of note, after abortion of the sporulation process, many cells display severe morphological defects including branching, formation of spiral-shaped cells and formation of large spherical cells (Holkenbrink et al., 2014;Müller et al., 2012). Here, we observed similar morphological defects in mutants impaired in spore coat polysaccharide synthesis after chemical induction of sporulation. Interestingly, cells lacking the PHPT ExoE have less severe shape defects after 4 and 24 hr of glycerol-induction than mutants lacking enzymes suggested to act downstream of the priming step. These observations have two implications. First, we speculate that the abortion of the sporulation process in the exo mutants (as opposed to cell lysis due to lack of PG as well as spore coat) is caused by a coupling between spore coat polysaccharide synthesis and the PG removal process. Therefore, in the absence of proper spore coat polysaccharide synthesis, PG would not be completely removed and cells regain rod-shape through de novo synthesis of PG. Because, the Exo proteins are not important for PG synthesis during growth, we speculate that the coupling between spore coat polysaccharide synthesis and the PG removal process is regulatory rather than involving shared proteins.
Second, we speculate that in the absence of ExoE PHPT activity, Und-P is not sequestered in intermediates for spore coat polysaccharide biosynthesis, and therefore, PG can be resynthesised. By contrast, in the ΔexoJ-M, O-P mutants (here) and the previously described ΔexoA-D, G-I (Holkenbrink et al., 2014) mutants, Und-P would be sequestered in intermediates for spore coat polysaccharide, thus, titrating Und-P away from PG metabolism resulting in more cells with an abnormal shape. A future goal will be to understand how spore coat polysaccharide synthesis and PG removal are coordinated.
Our analysis of the taxonomic distribution of the exo and nfs gene clusters lend support to the notion that the spore coat could be synthesised by a different mechanism in sporulating Cystobacterineae compared to sporulating Nannocystineae and Sorangineae.

| Motility assays
Exponentially growing cultures of M. xanthus were harvested (6,000 g, room temperature (RT)) and resuspended in 1% CTT to a calculated density of 7 × 10 9 cells/ml. About 5 µl aliquots of cell suspensions were spotted on 0.5% and 1.5% agar supplemented with 0.5% CTT and incubated at 32°C. Cells were visualised after 24 hr using a M205FA Stereomicroscope (Leica) and imaged using a Hamamatsu ORCA-flash V2 Digital CMOS camera (Hamamatsu Photonics). Pictures were analysed using Metamorph® v 7.5 (Molecular Devices).

TA B L E 3 Plasmids used in this work
7.0, 1 mM of CaCl2) to a calculated density of 7 × 10 9 cells/ml. About 10 μl of aliquots of cells were placed on TPM agar (10 mM of Tris-HCl pH 7.6, 1 mM of K 2 HPO 4 /KH 2 PO 4 pH 7.6, 8 mM of MgSO 4 ), while for development in submerged culture, 50 μl of aliquots were mixed with 350 μl of MC7 buffer and placed in a 24-well polystyrene plate (Falcon).
Cells were visualised at the indicated time points using a M205FA Stereomicroscope (Leica) and imaged using a Hamamatsu ORCA-flash V2 Digital CMOS camera (Hamamatsu Photonics) and a DMi8 Inverted microscope and DFC9000 GT camera (Leica). After 120 hr, cells were collected and incubated at 50°C for 2 hr, and then, sonicated as described for chemically induced spores. Sporulation levels were determined as the number of sonication-and heat-resistant spores relative to WT using a Helber bacterial counting chamber (Hawksley, UK).
Briefly, exponentially growing cells were harvested, and resuspended in 1% CTT to a calculated density of 7 × 10 9 cells/ml. About 20 µl aliquots of cell suspensions were placed on 0.5% CTT 0.5% agar supplemented with 40 μg/ml Congo red. The plates were incubated at 32°C and documented at 24 hr.

| Cell length determination
About 5 µl aliquots of exponentially growing cultures were spotted on 1.5% agar supplemented with 0.2% CTT, immediately covered with a cover slide, imaged using a DMi8 Inverted microscope and DFC9000 GT camera (Leica) and cell length determined and visual-

| Detection of colanic acid biosynthesis
E. coli strains were grown at 37°C overnight on LB plates with antibiotics plus 0.2% (w/v) arabinose, when needed, to induce protein synthesis. Incubation was prolonged to 24-48 hr at RT to visualise the mucoid phenotype (Furlong et al., 2015).

| Statistics
Statistical analyses were performed using SigmaPlot v14. All data sets were tested for a normal distribution using a Shapiro-Wilk test.
For all data sets without a normal distribution, the Mann-Whitney test was applied to test for significant differences.

ACK N OWLED G EM ENTS
The authors thank Jana Jung for constructing SA7455 and Signal Response' as well as by the Max Planck Society.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon request.