PelX is a UDP-N-acetylglucosamine C4-epimerase involved in Pel polysaccharide-dependent biofilm formation

Pel is an N-acetylgalactosamine rich polysaccharide that contributes to the structure and function of Pseudomonas aeruginosa biofilms. The pelABCDEFG operon is highly conserved among diverse bacterial species, and thus Pel may be a widespread biofilm determinant. Previous annotation of pel gene clusters led us to identify an additional gene, pelX, that is found adjacent to pelABCDEFG in over 100 different bacterial species. The pelX gene is predicted to encode a member of the short-chain dehydrogenase/reductase (SDR) superfamily of enzymes, but its potential role in Pel-dependent biofilm formation is unknown. Herein, we have used Pseudomonas protegens Pf-5 as a model to understand PelX function as P. aeruginosa lacks a pelX homologue in its pel gene cluster. We find that P. protegens forms Pel-dependent biofilms, however, despite expression of pelX under these conditions, biofilm formation was unaffected in a ΔpelX strain. This observation led to our identification of the pelX paralogue, PFL_5533, which we designate pgnE, that appears to be functionally redundant to pelX. In line with this, a ΔpelX ΔpgnE double mutant was substantially impaired in its ability to form Pel-dependent biofilms. To understand the molecular basis for this observation, we determined the structure of PelX to 2.1Å resolution. The structure revealed that PelX resembles UDP-N-acetylglucosamine (UDP-GlcNAc) C4-epimerases and, using 1H NMR analysis, we show that PelX catalyzes the epimerization between UDP-GlcNAc and UDP-GalNAc. Taken together, our results demonstrate that Pel-dependent biofilm formation requires a UDP-GlcNAc C4-epimerase that generates the UDP-GalNAc precursors required by the Pel synthase machinery for polymer production.


INTRODUCTION 52
Exopolysaccharides are a critical component of bacterial biofilms. The opportunistic pathogen 53 Pseudomonas aeruginosa is a model bacterium for studying the contribution of exopolysaccharides to 54 biofilm architecture because biofilms formed by this organism use exopolysaccharides as a structural 55 scaffold (1). P. aeruginosa synthesizes the exopolysaccharides alginate, Psl, and Pel, and each have been PelX, using BLAST and Phyre 2 suggest that it likely encodes an SDR family enzyme (18,19). In total, we 109 identified 136 pel loci containing a pelX gene (Fig. 1, Data Set S1). 110 In order to determine whether pelX plays a role in Pel polysaccharide dependent biofilm formation, 111 we set out to characterize pelX in a species of bacteria for which the regulation of pel gene expression has 112 been studied. In P. protegens, which contains a pelX gene upstream of pelA, the pel gene cluster is under 113 the control of the same Gac/Rsm global regulatory cascade as in P. aeruginosa (17). In addition, two 114 putative recognition sequences for the enhancer binding protein FleQ are found upstream of pelX 115 (PFL_2971), not pelA, suggesting that in contrast to P. aeruginosa, pelX may be the first gene of the pel 116 operon in this species (20). Given that this operon is likely regulated in a similar manner to the pel locus of 117 P. aeruginosa and that these two species are closely related, we used P. protegens to characterize the role 118 of PelX in biofilm formation. 119 120

P. protegens forms Pel-dependent biofilms that are enhanced by elevated levels of c-di-GMP 121
In addition to the pel genes, psl gene expression has been shown to be regulated by the Gac/Rsm pathway 122 in P. protegens and this regulatory cascade is required for P. protegens biofilm formation (17). 123 Interestingly, some strains of P. aeruginosa, including PAO1, use Psl as their predominant biofilm matrix 124 exopolysaccharide whereas others, such as PA14,use Pel (21). Therefore, in order to determine whether P. 125 protegens biofilms are dependent on Pel and/or Psl, we generated strains lacking pelF or pslA, genes 126 previously shown to be required for Pel-and Psl-dependent biofilm formation, respectively, and examined 127 whether these strains could form biofilms (8,22). After five days of static growth in liquid culture, we found 128 that wild-type and DpslA strains of P. protegens adhered similarly to a polystyrene surface, whereas a strain 129 lacking pelF displayed a marked reduction in surface attachment ( Fig. 2A). The level of surface adherence 130 of a DpelF DpslA double mutant was comparable to that of the DpelF strain. Based on these data, we 131 conclude that the Pel polysaccharide is a critical component of P. protegens  Previous analysis of the region upstream of P. protegens pelX identified a FleQ consensus binding 133 sequence (20). FleQ is a bis-(3′,5′)-cyclic dimeric guanosine mono-phosphate (c-di-GMP) responsive 134 transcription factor that binds to specific sequences upstream of the pel operon in P. aeruginosa, blocking 135 their transcription (23). When the intracellular concentration of c-di-GMP is high, FleQ switches to an 136 activator and upregulates transcription of the pel genes (24). Based on these observations, we reasoned that 137 expression of the P. protegens pel operon is likely upregulated in the presence of elevated levels of c-di-138 GMP (23). To test this hypothesis, we expressed the well-characterized diguanylate cyclase WspR of P. 139 aeruginosa from an IPTG-inducible plasmid in P. protegens (25). Because WspR activity can be inhibited 140 by c-di-GMP binding to an allosteric site of the enzyme, we inactivated this autoinhibitory site by 141 introducing a previously characterized R242A point mutation into the sequence of the protein (WspR R242A ; 142 (26)). Upon induction of WspR R242A expression, approximately 2.3-fold more P. protegens adhered to 143 polystyrene surfaces compared to a vector control strain (Fig. 2B). Taken together, our data suggests that 144 P. protegens Pel-dependent biofilm formation is enhanced in response to elevated intracellular c-di-GMP 145 levels. 146 147 pelX is expressed under biofilm promoting conditions but is functionally redundant with PFL_5533 148 Since Pel-dependent biofilm formation is enhanced in the presence of c-di-GMP, and FleQ is predicted to 149 bind upstream of the pelX gene, we reasoned that pelX is most likely expressed in a c-di-GMP dependent 150 manner along with the rest of the pel genes. To test this, we probed for the expression of PelX by fusing a 151 vesicular stomatitis virus glycoprotein (VSV-G) tag to its C-terminus at the native pelX locus on the P. 152 protegens chromosome. To examine expression of the pel operon, a VSV-G tag was similarly added to the 153 C-terminus of the putative Pel synthase subunit, PelF (13). Strains expressing either WspR R242A or a vector 154 control were grown under biofilm-conducive conditions and analyzed by Western blot. In strains lacking 155 WspR R242A , neither PelX nor PelF could be detected; however, in the WspR R242A expressing strains, both 156 PelX and PelF were detected at their expected molecular weights of 34 and 58 kDa, respectively (Fig. 3A). 157 These data suggest that pelX and pelF expression are positively regulated by c-di-GMP in P. protegens, and that PelX is expressed under conditions where the Pel polysaccharide is produced. However, when we 159 deleted pelX, we found that P. protegens biofilm biomass was unaffected, indicating that PelX is not 160 essential for Pel-dependent biofilm formation (Fig. 3B). These findings led us to hypothesize that the P. 161 protegens genome might encode a second SDR enzyme that renders PelX functionally redundant. We 162 queried the PelX amino acid sequence against the P. protegens Pf-5 proteome using BLASTP to identify 163 similar proteins (18). This search identified several proteins from the SDR superfamily (Table 1), however, 164 one protein in particular, PFL_5533, stood out because it shares 68% sequence identity with PelX. To 165 determine whether PFL_5533 is expressed during P. protegens biofilm formation, we fused a C-terminal 166 VSV-G tag to PFL_5533 at its native chromosomal locus and examined its expression in the presence and 167 absence of WspR R242A . We detected similar levels of VSV-G tagged PFL_5533 in both vector control and 168 WspR R242A -expressing strains suggesting that in contrast to pelX, the expression of this gene does not 169 change in response to c-di-GMP (Fig. 3A). The observation that PFL_5533 is expressed during biofilm 170 growth conditions and that it possesses high sequence homology to pelX led us to probe its potential role in 171 Pel polysaccharide production. 172 To determine whether PFL_5533 contributes to biofilm formation by P. protegens, we generated a 173 strain lacking this gene and examined biofilm formation in our WspR R242A overexpression background. 174 Similar to our DpelX strain, we detected no significant difference in biofilm formation between DPFL_5533 175 and wild-type strains (Fig. 3B). In contrast, a DpelX DPFL_5533 double mutant exhibited a defect in biofilm 176 formation comparable to that of a DpelF strain, which is incapable of producing Pel (Fig. 3B). To confirm 177 that this reduction in biofilm formation was due to decreased Pel polysaccharide secretion, P. protegens 178 culture supernatants were analyzed using a lectin from Wisteria floribunda (WFL) that specifically 179 recognizes terminal GalNAc moieties, and Pel-specific antisera generated using P. aeruginosa Pel 180 polysaccharide (10,27). Culture supernatants from wild-type P. protegens displayed a strong signal when 181 analyzed by both of these detection methods, while a DpelF strain exhibited no signal, indicating that these 182 tools can be used to monitor Pel polysaccharide produced by this bacterium (Fig. 3C; (9)). In line with our biofilm data, Pel was detected in culture supernatants from DpelX and DPFL_5533 strains at levels 184 comparable to wild-type whereas no Pel polysaccharide was detected in the DpelX DPFL_5533 double 185 mutant. Taken together, these data indicate that pelX and PFL_5533 have genetically redundant functions 186 in biofilm formation under our experimental conditions, and that the activity of a predicted SDR family 187 enzyme is essential for Pel polysaccharide biosynthesis and Pel-dependent biofilm formation by P. 188

protegens. 189 190
PelX is a UDP-GlcNAC C4-epimerase that preferentially epimerizes N-acetylated UDP-hexoses 191 To gain further insight into PelX function, we initiated structural and functional studies on recombinant 192 PelX protein. Initial efforts to purify His6-tagged PelX overexpressed in E. coli yielded two species 193 consistent with a monomer and dimer of PelX when analyzed by SDS-PAGE. Addition of reducing agent 194 significantly lowered the abundance of the putative PelX dimer, suggesting that this higher molecular 195 weight species likely arose from the formation of an intermolecular disulfide bond. This intermolecular 196 disulfide bond is likely not biologically relevant given that the bacterial cytoplasm is a reducing 197 environment. As sample heterogeneity can be problematic for both the interpretation of biochemical data 198 and protein crystallization, we generated a PelX variant in which the cysteine residue presumed to be 199 involved in disulfide bond formation (C232) was mutated to serine (PelX C232S ). This PelX C232S variant 200 appeared as a monomer on SDS-PAGE and its purification to homogeneity was straightforward. When 201 examined by size exclusion chromatography, PelX C232S had an apparent molecular weight of 64 kDa 202 compared to its expected monomeric molecular weight of 35 kDa, suggesting that like other characterized 203 SDR enzymes, PelX forms non-covalent, SDS-sensitive dimers in solution ( Fig. S1; (28)). 204 The SDR superfamily of enzymes are known to catalyze numerous chemical reactions including 205 dehydration, reduction, isomerization, epimerization, dehalogenation, and decarboxylation (14). We 206 hypothesized that PelX likely functions as an epimerase because UDP-GalNAc, the putative precursor for 207 C4 position of the hexose ring. Characterized SDR C4-epimerases are classified into three groups based on 209 their substrate preference (29). Group 1 epimerases preferentially interconvert non-acetylated UDP-210 hexoses, group 2 epimerases are equally able to interconvert non-acetylated and N-acetylated UDP-hexoses, 211 while group 3 epimerases preferentially interconvert N-acetylated-UDP-hexoses. Given that the Pel 212 polysaccharide is GalNAc rich, we hypothesized that PelX likely functions as either a group 2 or group 3 213 epimerase. To examine the potential epimerase activity of PelX, we used 1 H NMR to monitor the 214 stereochemistry of UDP-GlcNAc, UDP-GalNAc, UDP-Glc, or UDP-galactose (UDP-Gal) in the presence 215 or absence of purified PelX C232S . Two 1 H NMR resonances with characteristic multiplicities in the 5.4-5.7 216 ppm H-1 " region allow for the differentiation of UDP-GalNAc/UDP-Gal from UDP-GlcNAc/UDP-Glc, 217 respectively ( Fig. 4A and 4B). Using these resonances, we found that PelX C232S readily converts UDP-218 GalNAc to UDP-GlcNAc and vice versa ( Fig. 4A and 4C). PelX C232S also converted a minor amount of 219 UDP-Gal to UDP-Glc, however, we did not observe significant conversion of UDP-Glc to UDP-Gal (Fig.  220   4B). Collectively, these data define PelX as a group 3 UDP-hexose C4-epimerase. 221 To corroborate our biochemical data, we next performed absolute quantification of cellular GalNAc 222 and GlcNAc levels in our WspR R242A -expressing P. protegens wild-type, DpelX, DPFL_5533, and DpelX 223 DPFL_5533 strains. While GalNAc levels were below the limit of our detection methods, we found that 224 GlcNAc levels were significantly elevated in the epimerase deficient background compared to both wild-225 type and the individual epimerase mutant strains (Fig. 4D). Taken together with our 1 H NMR results, these 226 data suggest that PelX and its homologue PFL_5533 function to generate pools of UDP-GalNAc precursors 227 for polymerization into Pel polysaccharide. 228 229

PelX resembles members of the SDR enzyme superfamily 230
Having established that PelX is a UDP-GlcNAc C4-epimerase, we next sought to determine its structure to 231 obtain further insight into substrate recognition by this enzyme. Despite its straightforward purification and 232 homogenous oligomeric state, we found PelX C232S to be recalcitrant to crystallization. We next attempted to crystallize PelX C232S in complex with its confirmed substrate UDP-GlcNAc. Crystals of PelX C232S 234 incubated with UDP-GlcNAc appeared within three days and the structure of the complex was solved to 235 2.1 Å resolution using molecular replacement with the SDR family member WbpP (PDB ID: 1SB8) as the 236 search model (28). PelX crystallized in space group P21212 and contains a dimer in the asymmetric unit, an 237 arrangement observed for many other structurally characterized SDR family members ( Fig. 5A; (30)). The 238 dimer interface of PelX C232S is similar to that observed in the WbpP crystal structure where each protomer 239 contributes two a-helices to a four-helix bundle. 240 The overall structure of PelX C232S shows that it possesses the characteristic domains associated with 241 the SDR family, which includes an N-terminal NAD + -binding Rossmann-fold (residues 1-172 and 218-242 243) and a C-terminal a/b-domain involved in substrate-binding (residues 173-217 and 244-310; Fig. 5A). 243 PelX C232S contains the GxxGxxG motif required for binding NAD + that is found in all SDR family members 244 as well as the active site catalytic triad Sx24Yx3K (31). Although NAD + was not exogenously supplied in 245 the purification or crystallization buffers, electron density for this cofactor was clearly observed, suggesting 246 it was acquired during PelX C232S overexpression in E. coli. While the addition of UDP-GlcNAc was 247 essential for the formation of crystals, we were unable to model the GlcNAc moiety of this molecule due 248 to the poor quality of the electron density (Fig. S3). We speculate that the sugar moiety may be disordered 249 because PelX C232S is catalytically active and converting a portion of the UDP-GlcNAc to UDP-GalNAc. 250 Modeling UDP alone rather than UDP-GlcNAc improved the refinement statistics of the overall model and 251 resulted in ligand B-factors comparable to the surrounding protein atoms (Table 2). 252 Previous studies on a catalytically inactive variant of the UDP-Gal 4-epimerase GalE from E. coli 253 allowed for the co-crystallization and modeling of UDP-Glc and UDP-Gal in the active site of this enzyme 254 (32). In their study, these authors targeted the serine and tyrosine residues of the consensus Sx24Yx3K active 255 site motif. Guided by this approach, we generated a variant of PelX C232S with S121A and Y146F mutations 256 and confirmed that this variant is catalytically inactive (Fig. S2). PelX C232S/S121A/Y146F crystallized readily 257 with either UDP-GlcNAc or UDP-GalNAc, and both structures were solved to a resolution of 2.1 Å using molecular replacement (Table 2). The final models of PelX C232S/S121A/Y146F in complex with UDP-GlcNAc 259 or UDP-GalNAc were both refined to an Rwork/Rfree of 15.6%/19.5% (Table 2). In these structures, the 260 electron density for the sugar moieties was well defined compared to the PelX C232S -UDP-GlcNAc co-261 crystal structure and allowed for the unambiguous modelling of the expected sugar-nucleotides (Fig. S3). 262 Given that both structures showed improved ligand density for their respective substrates, these structures 263 substantiate our biochemical data showing that UDP-GlcNAc and UDP-GalNAc are substrates for PelX. 264 Examination of the active site of our PelX C232S/S121A/Y146F -substrate complexes did not show any significant 265 differences in the positions of active site residues suggesting that both sugar-nucleotides are recognized by 266 the enzyme in a similar manner (Fig. 5B). We next compared our substrate-bound PelX C232S/S121A/Y146F 267 structures to the UDP-GalNAc bound structure of the aforementioned UDP-hexose C4-epimerase WbpP 268 from P. aeruginosa. WbpP shares 32% sequence identity to PelX and also catalyzes the epimerization of 269 UDP-GlcNAc to UDP-GalNAc (28). The overall structure of WbpP is highly similar to PelX C232S/S121A/Y146F 270 (PDB code 1SB8, rms deviation 1.9 Å over 306 Cα) except that WbpP possesses an additional N-terminal 271 α-helix not found in PelX. The active site residues identified as being important for sugar-nucleotide 272 interaction in WbpP are invariant in PelX ( Fig. 5C) with the exception of A81, A122 and G189 in PelX, 273 which correspond to residues G102, S143 and A209 in WbpP, respectively (28). These differences are not 274 predicted to impair specificity towards the UDP-GlcNAc/GalNAc substrate. Rather, Demendi et al found 275 that bulkier residues (G102K, A209N), and mutation of S143A actually displayed enhanced specificity 276 towards acetylated substrates (33). However, while the positions of the PelX C232S/S121A/Y146F and WbpP active 277 site residues and NAD + cofactor are highly similar, comparison of the bound UDP-GalNAc substrate 278 between the two structures reveals distinct differences in the conformations of the GalNAc moiety ( Fig.  279 5C). We suspect that this difference in conformation may be a result of the co-crystallization of UDP-280 GalNAc with wild-type WbpP whereas to observe electron density for the GalNAc moiety of UDP-GalNAc 281 in complex with PelX we had to mutate two active site residues, S121A and Y146F. The residues equivalent 282 to S121 and Y146 in WbpP make contact with the C4 hydroxyl group of GalNAc and thus are likely 283 involved in substrate orientation. These observations suggest that the conformation of UDP-GalNAc in our mutant PelX co-crystal structure may not represent a state adopted during catalysis, but demonstrate a high 285 degree of conformational freedom of the sugar moiety within the relatively large substrate binding pocket. 286 The GlcNAc moiety of UDP-GlcNAc in our PelX C232S/S121A/Y146F -UDP-GlcNAc co-crystal structure was 287 also found in a similar orientation as in our UDP-GalNAc-containing structure. Taking these considerations 288 into account and given that WbpP and PelX share a high degree of sequence similarity and interconvert 289 identical substrates with similar preference, we speculate that the epimerization of N-acetylated UDP-290 hexoses by PelX most likely occurs via a similar catalytic mechanism as proposed for WbpP (28,33). In 291 sum, our structural data support our biochemical studies showing that PelX belongs to the group 3 family 292 of UDP-N-acetylated hexose C4-epimerases. 293

DISCUSSION 295
In this study, we report the characterization of the Pel polysaccharide precursor-generating enzyme PelX. 296 Using P. protegens Pf-5 as a model bacterium, we found that pelX is required for Pel polysaccharide-297 dependent biofilm formation in a strain that also lacks the pelX paralogue, PFL_5533. Guided by our 1 H 298 NMR analyses and multiple crystal structures, we have shown that PelX functions as a UDP-GlcNAc C4-299 epimerase and that it preferentially interconverts UDP-GlcNAc/UDP-GalNAc over UDP-Glc/UDP-Gal, 300 defining it as a group 3 UDP-N-acetylhexose C4-epimerase. Based on these observations and the data 301 presented herein we propose naming PFL_5533 polysaccharide UDP-GlcNAc epimerase (pgnE). 302 Functional redundancy of sugar-nucleotide synthesizing enzymes in biofilm producing bacteria is 303 not unprecedented. For example, in P. aeruginosa PAO1, PslB and WbpW both catalyze the synthesis of 304

GDP-mannose, a precursor molecule required for Psl polysaccharide and A-band lipopolysaccharide (LPS). 305
Like PelX and PgnE, PslB and WbpW have been shown to be genetically redundant as a defect in Psl 306 polysaccharide or A-band LPS is only observed when both pslB and wbpW are deleted (22). Although P. 307 aeruginosa PAO1 has another paralogue of PslB and WbpW, AlgA, the algD promoter responsible for 308 transcription of the algA gene is not significantly activated in non-mucoid strains such as PAO1 (34). Psl 309 biosynthesis, like Pel, is also regulated by c-di-GMP through FleQ (23) whereas being an integral component of the P. aeruginosa outer membrane, the genes responsible for A-band LPS synthesis are 311 constitutively expressed (35). Although, at present what additional glycans PgnE may be involved in 312 producing is unknown, it is clear that the existence of paralogous sugar-nucleotide synthesizing enzymes 313 may be a means of keeping up with metabolic demand during the synthesis of multiple cell surface 314

polysaccharides. 315
We previously reported the isolation of Pel polysaccharide from P. aeruginosa PAO1 and 316 carbohydrate composition analyses showed that it is rich in GalNAc (9). Therefore, the co-regulation of a 317 UDP-GlcNAc C4-epimerase with the pel genes likely ensures that adequate quantities of UDP-GalNAc are 318 available for Pel biosynthesis when a biofilm mode of growth is favoured. In contrast to P. protegens Pf-5, 319 P. aeruginosa PAO1 does not contain a pelX gene in its Pel biosynthetic gene cluster, yet this bacterium is 320 also capable of producing Pel polysaccharide (36). In the PAO1 genome, the poorly characterized PA4068 321 gene is found in the same genomic context as pgnE whereby both genes are part of a two-gene operon, with 322 the second gene predicted to encode a dTDP-4-dehydrorhamnose reductase (PA4069/PFL_5534; (37)). In 323 addition, the protein encoded by PA4068 shares 76% identity with PgnE, suggesting that this gene may 324 function analogously to pgnE and by extension pelX. A DPA4068 mutant was found to display a surface 325 attachment defect during secretin induced stress suggesting a role for this gene in surface glycan production 326 (37). However, it has been established that Psl is the primary polysaccharide required for P. aeruginosa 327 PAO1 biofilm formation even though this strain is genetically capable of synthesizing Pel (36). 328 Consequently, studies characterizing Pel polysaccharide production by PAO1 have relied on an engineered 329 strain that lacks the ability to produce Psl and expresses the pel genes from an arabinose-inducible promoter. 330 It may be that only low levels of UDP-GalNAc are required to sustain Pel polysaccharide production by 331 wild-type PAO1 and thus a second UDP-GlcNAc C4-epimerase that is dedicated to Pel production is not 332 required. In contrast, Pel polysaccharide appears to be a major biofilm matrix constituent in P. protegens 333 Pf-5 and thus the higher levels of Pel production in this organism may necessitate the need for increased 334 synthesis of UDP-GalNAc precursors.
The epimerization of UDP-Gal to UDP-Glc by PelX occurs much less efficiently than its N-336 acetylated counterpart. Creuzenet and colleagues noted a similar trend for WbpP, a UDP-GlcNAc C4-337 epimerase involved in P. aeruginosa PAK O-antigen biosynthesis, and hypothesized that the poor 338 efficiency displayed by this enzyme towards non-acetylated substrates means that this reaction is unlikely 339 to occur in vivo (38). The equilibrium of the PelX catalyzed epimerization between UDP-GalNAc and UDP-340 GlcNAc in vitro is skewed towards the more thermodynamically stable UDP-GlcNAc epimer. A similar 341 balance for this equilibrium has been documented for other epimerases (38,39). We speculate that the 342 continuous polymerization of UDP-GalNAc by the putative Pel polysaccharide polymerase, PelF, would 343 keep the cellular concentration of UDP-GalNAc low and thus drive the equilibrium towards its production. 344 In conclusion, this work demonstrates the involvement of a Pel polysaccharide precursor generating 345 enzyme required for biofilm formation in P. protegens. Our data linking the production of UDP-GalNAc 346 to Pel polysaccharide production lends genetic and biochemical support to the chemical analyses that 347 showed Pel is a GalNAc-rich carbohydrate polymer (9). Furthermore, the identification of a new Pel 348 polysaccharide-dependent biofilm forming bacterium provides an additional model system that can be used 349 for the characterization of this understudied polysaccharide secretion apparatus.

EXPERIMENTAL PROCEDURES 351
Bacterial strains, microbiological media and physiological buffers. All bacterial strains and plasmids used 352 in this study are listed in Table S1. Jensen's medium contained per liter of MilliQ water: 5 g NaCl, 2.51 g 353 K2HPO4, 13.46 g glutamic acid, 2.81 g L-valine, 1.32 g L-phenylalanine, 0.33 g/L MgSO4•7H2O, 21 mg 354 respectively. Prior to this study, the POGs were unnamed, therefore based on our observations we have 377 named these POGs as pelX and pgnE. PelX primary amino acid sequences were aligned using MUSCLE 378 (44) to identify highly conserved amino acid residues. Additionally, the P. protegens PelX sequence was 379 submitted to Phyre 2 to determine the predicted fold of the protein (19). The PelX and PgnE protein 380 sequences from P. protegens Pf-5 were obtained from the Pseudomonas Genome Database (41). 381 Comparison of the PelX structure to previously determined structures was performed using the DALI 382 pairwise comparison server (45). in Table S1). The pslA, pelF, and pelX, alleles were generated using forward upstream and downstream 389 reverse primers tailed with EcoRI and XbaI, restrictions sites, respectively ( Table S1). The PFL_5533 allele 390 was generated using forward upstream and downstream reverse primers tailed with EcoRI and HindIII 391 restriction sites, respectively (Table S1). This PCR product was gel purified, digested and ligated into 392 pEXG2, and the resulting constructs, pLSM33, pLSM34, and pLSM35, pLSM36 were identified and 393 sequenced as described above. 394 The VSV-G tagged pelF, pelX, and PFL_5533 constructs were generated by amplifying flanking 395 upstream and downstream regions surrounding the stop codon of the ORFs of each gene. The reverse 396 upstream and forward downstream primers (Table S1) were tailed with complementary sequences encoding 397 the VSV-G peptide immediately before the stop codon. Amplified upstream and downstream fragments 398 were joined by splicing-by-overlap extension PCR using forward upstream and reverse downstream primers 399 tailed with EcoRI and HindIII restriction sites, respectively (Table S1). These PCR products were gel purified, digested, and ligated into pEXG2, as described above. Clones with positive inserts were verified 401 by Sanger sequencing to generate pLSM37, pLSM38, and pLSM39. 402 The aforementioned pEXG2 based plasmids were introduced into P. protegens Pf-5 via biparental 403 mating with donor strain E. coli SM10 (47). Merodiploids were selected on LB containing 60 µg mL -1 404 Gentamicin (Gen) and 25 µg mL -1 Irgasan (Irg). SacB-mediated counter-selection was carried out by 405 selecting for double-crossover mutations on no salt lysogeny broth (NSLB) agar containing 5% (w/v) 406 sucrose. Unmarked gene deletions were identified by PCR with primers targeting the outside, flanking 407 regions of pslA, pelF, pelX, and PFL_5533 (Table S1, S2). These PCR products were Sanger sequenced 408 using the same primers to confirm the correct deletion. 409 410 Generation of WspR overexpression strains -The wspR nucleotide sequence from P. aeruginosa PAO1 411 was obtained from the Pseudomonas Genome Database and used to design primers specific to full-length 412 wspR (Table S1). The forward primer encodes an EcoRI restriction site and a ribosomal binding site, while 413 the reverse primer encodes a HindIII restriction site. The amplified PCR products were digested with EcoRI 414 and HindIII restriction endonucleases and subsequently cloned into the pPSV39 vector (Table S1). 415 Confirmation of the correct nucleotide sequence of wspR was achieved through DNA sequencing (The 416 Center for Applied Genomics, The Hospital for Sick Children). R242 was mutated to an alanine to prevent 417 allosteric inhibition of WspR using the QuickChange Lightning Site Directed Mutagenesis kit (Agilent 418 technologies), as described previously. The resulting expression vector (pLSM-wspR R242A ) encodes residues 419 1-347 of WspR. Introduction of the pPSV39 empty vector or pSLM-wspR R242A into P. protegens was carried 420 out by electroporation. Positive clones were selected for on LB agar containing 30 µg mL -1 Gen. h. Non-attached cells were removed and the wells were washed thoroughly with water, and stained with 425 1.5 mL 0.1% (w/v) crystal violet. After 10 minutes, the wells were washed again and the stain solubilized using 2 mL of 95% (v/v) ethanol for 10 minutes. 200 µL was transferred to a fresh 96-well polypropylene 427 plate (Nunc) and the absorbance measured at 550 nm. For strains containing empty pPSV39 or pLSM-428 wspR R242A , the above protocol was modified slightly. As c-di-GMP significantly upregulated biofilm 429 formation, crystal violet staining for these strains was performed as described previously using 96-well 430 polypropylene plates that were incubated statically for 6 h or 24 h at 30 °C. All strains were grown in KBM 431 containing 30 µg ml -1 Gen and 30 µM IPTG. 432 433 Dot blots -Pel antisera was obtained as described in Colvin et al. from P. aeruginosa PA14 PBADpel (10). 434 The adsorption reaction was conducted as described by Jennings et al (9). Culture supernatants containing 435 secreted Pel were harvested by centrifugation (16,000 × g for 2 min) from 1 mL aliquots of P. protegens 436 grown overnight at 30 °C in LB containing 30 µg ml -1 Gen and 30 µM IPTG, and treated with proteinase K 437 (final concentration, 0.5 mg mL -1 ) for 60 min at 60 °C, followed by 30 min at 80 °C to inactivate proteinase 438

K. 439
Pel immunoblots were performed as described by Colvin et al (10) and Jennings et al (9). 5 µL of 440 secreted Pel, prepared as described above, was pipetted onto a nitrocellulose membrane and left to air dry 441 for 10 min. The membrane was blocked with 5% (w/v) skim milk in Tris-buffered saline (10 mM Tris-HCl 442 pH 7.5, 150 mM NaCl) containing 0.1% (v/v) Tween-20 (TBS-T) for 1 h at room temperature and probed 443 with adsorbed a-Pel at a 1:60 dilution in 1% (w/v) skim milk in TBS-T overnight at 4 °C with shaking. 444 Blots were washed three times for 5 min each with TBS-T, probed with goat α-rabbit HRP-conjugated 445 secondary antibody (Bio-Rad) at a 1:2000 dilution in TBS-T for 45 min at room temperature with shaking, 446 and washed again. All immunoblots were developed using SuperSignal West Pico (Thermo Scientific) 447 following the manufacturer's recommendations. 448 For WFL-HRP immunoblots, 5 µL of secreted Pel, prepared as described above, was pipetted onto 449 a nitrocellulose membrane and left to air dry for 10 min. The membrane was blocked with 5% (w/v) bovine 450 serum albumin (BSA) in TBS-T for 1 h at room temperature and probed with 10 µg/mL of WFL-HRP (EY Laboratories) in 2% (w/v) BSA in TBS-T with 0.2 g/L CaCl2 overnight at room temperature with shaking. 452 Membranes were washed twice for 5 min and once for 10 min with TBS-T, then developed as described were imaged using a BioRad ChemiDoc imaging system. 473

474
Cloning and mutagenesis -The pelX nucleotide sequence from P. protegens Pf-5 (PFL_2971) was obtained 475 from the Pseudomonas Genome Database (41) and used to design primers specific to full-length pelX 476 (Table S1). The amplified PCR products were digested with NdeI and XhoI restriction endonucleases and subsequently cloned into the pET28a vector (Novagen). Confirmation of the correct nucleotide sequence 478 of pelX was achieved through DNA sequencing (ACGT DNA Technologies Corporation). The resulting 479 expression vector (pLSM-PelX) encodes residues 1-309 of PelX fused to a cleavable N-terminal His6 tag 480 (His6-PelX) for purification purposes (Table S2) analysis revealed that the resulting His6-PelX C232S was ~99% pure and appeared at its expected molecular 505 weight of 35 kDa. Fractions containing PelX C232S were pooled and concentrated to a volume of 2 mL by 506 centrifugation at 2 200 × g at a temperature of 4 °C using an Amicon ultra centrifugal filter device 507 (Millipore) with a 10 kDa molecular weight cut-off. PelX C232S was purified and buffer exchanged into 508 Buffer B [20 mM Tris-HCl pH 8.0, 150 mM NaCl, 5% (v/v) glycerol, 1 mM TCEP] by size-exclusion 509 chromatography using a HiLoad 16/60 Superdex 200 gel-filtration column (GE Healthcare). PelX C232S 510 eluted as a single Gaussian shaped peak, and all PelX C232S containing fractions were pooled and 511 concentrated by centrifugation, as above, to 8 mg mL -1 and stored at 4 °C. PelX C232S/Y146F/S121A was purified 512 similarly. 513 514

Determination of the PelX oligomerization state by gel filtration analysis -Oligomerization of PelX C232S 515
was determined using a Superdex 200 10/300 GL column (GE Life Sciences). The column was equilibrated 516 in Buffer B. Molecular weight standards (Sigma, 12-200 kDa) were applied to the column as directed. 517 PelX C232S was applied to the column at 7.5 mg ml -1 (100 µL) and protein elution was monitored at 280 nm. 518 519 NMR activity assay -The following method has been adapted from Wyszynski et al (39). Enzymatic 520 reactions were performed in 30 mM sodium phosphate pH 8.0, with 50 µg of freshly purified PelX C232S and 521 10 mM UDP-GlcNAc, UDP-Glc, UDP-Gal, or 5 mM UDP-GalNAc in a total reaction volume of 220 µL. 522 After incubation at 37 °C for 1 hour, the mixture was flash frozen and lyophilized. The resulting material 523 was dissolved in 220 µL of D2O and analyzed by 1 H NMR. As control experiments, the same procedures 524 were applied to samples lacking PelX or UDP-GlcNAc. Data were collected on a Varian 600 MHz NMR 525 spectrometer. 526 527 Intracellular metabolite extraction -P. protegens Pf-5 wild-type, ΔpelX, ΔPFL_5533 and ΔpelX 528 ΔPFL_5533 strains that had been transformed with a plasmid expressing WspR R242A (pLSM21) were streaked out twice in succession on Jensen's agar containing 30 µg/mL gentamicin, and these first and 530 second subcultures were grown for 48 h at 30 °C. For each biological replicate, cells from the second 531 subcultures were collected using a polyester swab and suspended in 30 mL of Jensen's medium to match 532 an optical density at 600 nm (OD600) of 0.6. Subsequently, two 10 mL aliquots of this standardized culture 533 were each passed through a syringe filter (0.45 µm, PVDF, Millipore) to collect the bacteria. These filters 534 were placed face-up on Jensen's agar using flame sterilized tweezers and were then incubated at 30 °C for 535 3 h. 536 Following this incubation, the first filter was placed in 2 mL of sterile PBS containing 1 mM 537 purified recombinant PelA, and this filter was incubated for 30 min at room temperature to break up 538 aggregates (11). An established microtiter dilution method for viable cell counting was used to determine 539 the number of bacteria on the filter (48). The second filter was put into a Petri dish (60 x 15 mm) containing 540 2 mL of cold 80% (v/v) LC-MS grade methanol, which was incubated for 15 min. Afterwards, 1 mL of 541 80% LC-MS grade methanol was used to wash the filter, and then the 3 mL of the methanol extract were 542 transferred to a 5 mL microcentrifuge tube. These tubes were placed in a centrifuge at 7000 × g for 30 min 543 at 4 °C, and then 2 mL of the supernatant were transferred to a 2 mL microcentrifuge tube. The methanol 544 was evaporated using a speed vac, and then the dried cell extracts were suspended in 200 µL of 50% (v/v) 545 LC-MS grade methanol. Cell extracts were then stored at -80 °C until LC-MS analysis. Mass spectrometry 546 measurements of GalNAc were normalized to viable cell counts. PelX was unable to form crystals in the absence of UDP-GlcNAc. 576 Crystals of PelX C232S were cryoprotected in well solution supplemented with 20% (v/v) ethylene 577 glycol by briefly soaking the crystal in a separate drop. Crystals were soaked for 2-3 s prior to vitrification 578 in liquid nitrogen, and subsequently stored until X-ray diffraction data were collected on beamline X29A 579 at the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory. A total of 360 580 images of 1° ∆φ oscillation were collected on an ADSC Q315 CCD detector with a 250 mm crystal-to-detector distance and an exposure time of 0.4 s per image. The data were processed using DENZO and 582 integrated intensities were scaled using SCALEPACK from the HKL-2000 program package (50). The data 583 collection statistics are summarized in Table 2. The structure was solved by molecular replacement using 584 WbpP as a model with PHENIX AutoMR wizard. The resulting map was of good quality and allowed 585 manual model building using COOT (51,52). The model was then refined using PHENIX.REFINE (52) to 586 a final Rwork/Rfree of 16.7% and 19.7%, respectively. 587 PelX C232S/S121A/Y146F in complex with UDP-GalNAc or UDP-GlcNAc was crystallized under the 588 same conditions as the wild-type protein, and data collection and refinement were performed as described 589 above. The corresponding statistics can be found in Table 2.     Figure S2: PelX C232S/Y146F/S121A is catalytically inactive towards UDP-GalNAc. 1 H NMR spectrum from 850 the reaction of PelX C232S S121A Y146F with UDP-GalNAc. 851 852 Figure S3: PelX C232S/Y146F/S121A density of the ligands. PelX is displayed with its N-terminal Rossmann-fold domain shown in green, and its C-terminal substrate-binding α/β-domain in purple as in Figure 5.

(A) 854
PelX C232S with density shown for UDP and NAD + (B) PelX C232S/Y146F/S121A in complex with UDP-GalNAc 855 and NAD + and (C) PelX C232S/Y146F/S121A in complex with UDP-GlcNAc and NAD + . All three structures were 856 modeled with NAD + and nucleotide or sugar-nucleotide shown in stick representation, with the 857 corresponding |2mFo-DFc| map displayed as black mesh contoured at 2.0 σ. 858