Identification of Poly-N-acetylglucosamine as a Major Polysaccharide Component of the Bacillus subtilis Biofilm Matrix*

Background: Bacillus subtilis is a model organism for analyzing bacterial biofilms, but the carbohydrate components are undescribed. Results: Genes in the epsHIJK locus needed for biofilm formation encode proteins synthesizing the conserved bacterial polysaccharide poly-N-acetylglucosamine (PNAG). Conclusion: PNAG is a major carbohydrate component of B. subtilis biofilms. Significance: PNAG production is essential for formation of B. subtilis biofilms. Bacillus subtilis is intensively studied as a model organism for the development of bacterial biofilms or pellicles. A key component is currently undefined exopolysaccharides produced from proteins encoded by genes within the eps locus. Within this locus are four genes, epsHIJK, known to be essential for pellicle formation. We show they encode proteins synthesizing the broadly expressed microbial carbohydrate poly-N-acetylglucosamine (PNAG). PNAG was present in both pellicle and planktonic wild-type B. subtilis cells and in strains with deletions in the epsA–G and -L–O genes but not in strains deleted for epsH–K. Cloning of the B. subtilis epsH–K genes into Escherichia coli with in-frame deletions in the PNAG biosynthetic genes pgaA–D, respectively, restored PNAG production in E. coli. Cloning the entire B. subtilis epsHIJK locus into pga-deleted E. coli, Klebsiella pneumoniae, or alginate-negative Pseudomonas aeruginosa restored or conferred PNAG production. Bioinformatic and structural predictions of the EpsHIJK proteins suggest EpsH and EpsJ are glycosyltransferases (GT) with a GT-A fold; EpsI is a GT with a GT-B fold, and EpsK is an α-helical membrane transporter. B. subtilis, E. coli, and pga-deleted E. coli carrying the epsHIJK genes on a plasmid were all susceptible to opsonic killing by antibodies to PNAG. The immunochemical and genetic data identify the genes and proteins used by B. subtilis to produce PNAG as a significant carbohydrate factor essential for pellicle formation.

depending on the microbial species, strain, growth conditions, and overall environment. Interestingly, numerous species have developed diverse metabolic pathways for production of EPS, alluding to the possibility that these systems have evolved independently. A common EPS associated with microbial biofilm formation is a polymeric ␤-1,6-linked N-acetylglucosamine (PNAG) structure that is highly conserved and expressed by a range of bacterial, fungal, and protozoan microorganisms (6,7). PNAG was first isolated and characterized from Staphylococcus epidermidis (8), where it was referred to as the polysaccharide intercellular adhesin, and then later shown to be produced by Staphylococcus aureus (9,10). In these two species, PNAG is synthesized by proteins encoded by four genes in the ica operon (9 -11), and ica-deleted PNAG-deficient strains were unable to produce biofilms (9,11). However, PNAG-independent biofilm formation in some staphylococcal strains has also been described (12). In general, depending on the strain and culture conditions, PNAG is often necessary, but not sufficient or essential, for biofilm formation.
PNAG synthesis also occurs in various Gram-negative organisms, including Escherichia coli (13), Acinetobacter baumannii (14) and Burkholderia spp. (15). In these organisms PNAG is synthesized by four proteins encoded by genes in the pga operon. More recently, Cywes-Bentley et al. (7) showed that a much wider spectrum of microbes can synthesize PNAG, including many human bacterial pathogens such as Streptococcus pneumoniae, Streptococcus pyogenes, Streptococcus agalactiae, Neisseria gonorrhoeae, and Neisseria meningitidis, the nontypable Haemophilus influenzae and Mycobacterium tuberculosis, and eukaryotic organisms such as fungal pathogens, and protozoan parasites such as Trichomonas vaginalis and murine and human Plasmodia spp. that cause malaria. Among many of these pathogens there are no readily identifiable genes homologous to those in the ica or pga loci. Nonetheless, the broad conservation of PNAG synthesis implies an important role in microbial biology, perhaps related to growth phases involving aggregation of microbes living in different environments or protection from anti-microbial factors.
Bacillus subtilis is a spore-forming Gram-positive bacillus wherein certain strains grown under specific conditions develop complex biofilms (16 -18). A large amount of knowledge has been accumulated that explains how the production and composition of the B. subtilis biofilm are regulated (17), although many aspects of this process are not fully elucidated. Full biofilm formation by some strains of B. subtilis such as NCIB 3610 (3610) is dependent on 15 proteins encoded by genes in the eps locus (epsA-epsO) that are associated with the carbohydrate content and complexity of a surface pellicle (16) involved in the overall biofilm structure. The composition and structure of polysaccharides synthesized by the proteins within this complex are not well described, but mutations in most of the genes within the B. subtilis eps cluster lead to loss of biofilm formation (19). Some genes, such as epsE, have dual functions, affecting both polysaccharide synthesis and flagella-based motility (20,21).
Given the ubiquity of PNAG synthesis among a range of microbial organisms, we evaluated B. subtilis biofilms for PNAG production and further examined the B. subtilis eps locus for genes potentially encoding PNAG biosynthetic proteins. Both biofilm and planktonic cells produced PNAG, and within the available annotated genome, we identified four genes, epsH-K, as potentially being responsible for PNAG synthesis. These genes are predicted to encode two glycosyltransferases (GT) (epsH and epsJ), separated by another GT with potential EPS modifying enzymatic activity (epsI), and a transporter/facilitator of synthesis (epsK). Cloning of the epsH-K genes into PNAG-deficient E. coli or Klebsiella pneumoniae (⌬pga), or alginate-negative Pseudomonas aeruginosa, leads to synthesis of a polymer immunochemically equivalent to PNAG. Each of the four B. subtilis epsH-K genes could individually trans-complement E. coli strains deleted for the pgaA-D genes, respectively. Also the epsH and epsJ genes could partially complement E. coli strains deleted for the pgaC or pgaA genes, respectively. Extracts of both WT B. subtilis and E. coli (⌬pga) complemented with B. subtilis epsHIJK contained PNAG-immunoreactive, hexosamine-containing material that was destroyed by treatment with both the PNAG-degrading enzyme dispersin B (22) and by sodium periodate, which can only hydrolyze polymeric hexosamines in a 1-6-linkage. Synthesis of PNAG in E. coli from the epsH-K genes resulted in susceptibility of cells to killing in an opsonophagocytic assay using antibodies specific to PNAG, indicative of functional conservation of PNAG properties when B. subtilis gene products direct synthesis of PNAG in E. coli. Overall, we identify the presence of PNAG in B. subtilis biofilms and the genes within the eps locus that encoded proteins that synthesize PNAG. B. subtilis lacking epsH-K genes are unable to form biofilms, indicating PNAG is essential for biofilm/pellicle formation by this organism.

Experimental Procedures
Bacterial Strains and Plasmids-Bacterial strains (B. subtilis, E. coli, K. pneumoniae, and P. aeruginosa) and plasmids used in this study are listed in Table 1. B. subtilis was grown on minimal medium glucose (MMG) agar (23) or lysogeny broth (LB) agar (24) for 3-5 days at room temperature, whereas the other bacterial strains were grown overnight at 37°C in LB or on LB agar.
BLAST (Basic Local Alignment Search Tool) and Structural Analysis-Sequences of eps genes in B. subtilis strain 168 were obtained from the NCBI website (reference sequence, NZ_ CM000487.1). Amino acid sequences from four eps genes (epsH, epsI, epsJ, and epsK) were mapped on E. coli and S. aureus sequences using BLAST from the NCBI website. Amino acid sequences from the four pga genes of E. coli IHE3034 were similarly mapped on S. aureus sequences. The amino acid sequences of B. subtilis epsH-K were analyzed by the Protein Homology/analogY Recognition Engine Version 2 (PHYRE 2 ) (25) for structural predictions as to their function.
Strain Construction in Gram-negative Bacteria-Deletions of individual pga genes in a clinical isolate of an E. coli K1 strain from a child with meningitis, designated E11 and kindly provided by Kwan Sik-Kim of Johns Hopkins University School of Medicine, and deletion of the entire 4-gene pga locus in both E. coli E11 and K. pneumoniae NTUH-K2044 (26) were constructed as described previously (27). Briefly, a kanamycin resistance cassette flanked by FLP recombinase recognition target sites and homology arms to replace the DNA segments of interest in-frame were generated by PCR with deletion primers (supplemental Table 1). Recombination within the targeted chromosomal sequences was mediated by the red recombinase encoded on pRdET (28), resulting in the replacement of the targeted sequence with a kanamycin-resistant cassette; all allelic replacements were confirmed by PCR. Subsequently, the kanamycin marker was removed, using the FLP expression plasmid pCP20 (29).
Strain Construction in B. subtilis-All constructs were either directly integrated and resolved in the competent strain DK1042 (comI Q12L ) or integrated in the competent strain DS2569 (⌬pBS32), transferred to the 3610 background using SPP1-mediated generalized phage transduction, and resolved (30). All strains and plasmids used in this study are listed in Table 1. Primer sequences are delineated in supplemental Table 1.
In-frame Deletions-To generate the in frame marker-less deletion constructs, each plasmid was introduced by single crossover integration at the restrictive temperature for plasmid replication (37°C) using mls resistance as a selection. To evict the plasmid, the strain was incubated in 3 ml of LB broth at a permissive temperature for plasmid replication (22°C) for 14 h. Cells were then serially diluted and plated on LB agar at 37°C. Individual colonies were patched onto LB agar plates and LB agar plates containing mls to identify mls-sensitive colonies that had evicted the plasmid. Chromosomal DNA from colonies that had excised the plasmid was purified and screened by PCR to determine which isolate had retained the deletion allele.
To generate the ⌬epsA in-frame marker-less deletion construct, the region upstream of epsA was amplified by PCR using the primer pair 3971/3972 and digested with EcoRI and XhoI, and the region downstream of epsA was amplified by PCR using the primer pair 3973/3974 and digested with XhoI and BamHI. The two fragments were then simultaneously ligated into the EcoRI and BamHI sites of pMiniMAD that carries a temperature-sensitive origin of replication and an erythromycin resistance cassette to generate pMP201.
The same method was used to generate the other eps gene in-frame marker-less deletion constructs. Plasmids are detailed in Table 1 and primers in supplemental Table 1.
Genetic Complementation-Wild-type alginate overproducing mucoid P. aeruginosa FRD1 and an alginate-deficient strain (P. aeruginosa (Tn501::algF)) due to a polar effect of the Tn501 insert in the algF gene on alginate synthesis (31) were used to express the B. subtilis epsHIJK locus (29).
For complementation of the various mutations, pga or eps genes from E. coli K1 strain E11 or B. subtilis DS991, respectively, were amplified by PCR with complementation primers (see Table 2) using chromosomal DNA as a template. PCR products were cloned into the broad host range vector pUCP18Tc as XbaI/SbfI or SacI/XbaI fragments and transformed into E. coli Sm10 with selection on LB agar containing tetracycline (10 mg/liter). All constructs were confirmed by sequencing. E. coli and K. pneumoniae strains were complemented with individual plasmids by electroporation and selection on LB agar supplemented with tetracycline (10 mg/liter). For P. aeruginosa, complementation plasmids were conjugated from E. coli Sm10 as described previously (32). The complemented P. aeruginosa strains were selected on LB agar containing irgasan (25 mg/liter) and tetracycline (75 mg/liter).
Immunochemical Detection of PNAG on Microbial Cells-To detect PNAG in B. subtilis biofilms, strain DS991 was inoculated into 1 ml of MMG broth in 12-well tissue culture plates and left for 1 week at room temperature in a humidified environment. Bacterial biofilms were removed with as little remaining media as possible and placed onto glass slides in a demarcated well. Samples were allowed to air dry before fixing for 5 min at room temperature with ice-cold methanol, which was removed by gently tipping the slide onto its side and absorbing residual methanol with a tissue.
Planktonic B. subtilis strains were grown in MMG media at 25°C for 3-10 days at room temperature, whereas the other bacterial species were grown on LB agar overnight at 37°C and then held at room temperature for 48 h to promote PNAG expression. Strains harboring one of the pUC vectors were grown on LB agar plates with tetracycline (10 mg/liter) under the same conditions with protection from light. Microbial samples were suspended in PBS, then spotted onto microscope slides, air-dried, and fixed for 1 min at room temperature with ice-cold methanol. After washing, slides were reacted with control or PNAG-specific mAbs directly conjugated to Alexa Fluor 488 at 5.2 g/ml along with nucleic acid stain, Syto 83 (Molecular Probes) (7). After 2 h at room temperature or overnight at 4°C, slides were washed and evaluated by confocal microscopy. For enzymatic and periodate treatments, samples fixed to slides were incubated in Tris-buffered saline (pH 6.4) containing either 50 g of dispersin B/ml (digests PNAG) or 50 g of chitinase/ml (no effect on PNAG) overnight at 37°C or in 0.4 M periodate (destroys PNAG) for 2 h at 37°C in a humidified environment. After washing, cells were treated with the Alexa Fluor 488 directly conjugated mAbs.
Extraction and Detection of PNAG-B. subtilis DS991 and E. coli ⌬pga (pUCP18Tc-epsHIJK) were grown as described above on MMG agar or LB agar plates, respectively, and cells were used for extraction and immunologic detection of PNAG by a slot blot, as described previously (14), and chemical detection of hexosamines, as described previously (33).
B. subtilis DS991, B. subtilis DK2055 ⌬epsK, E. coli WT, and E. coli ⌬pga as a negative control were tested for intracellular PNAG by a direct-binding ELISA using lysates of these cells prepared by sonication. Briefly, cells from blood agar plates grown overnight at 37°C were suspended in normal saline, washed, and then suspended in Tris-buffered saline containing 12.5 g/ml dispersin B to remove surface PNAG during a 24-h 37°C incubation step. After washing, bacterial cells were suspended in 0.04 M phosphate buffer (pH 7.2) at an absorbance at 650 nm (A 650 nm ) of 0.8, sonicated with 10-s bursts, 10 times, and insoluble debris was removed by centrifugation, and dilutions of the lysate were used to directly sensitize Immulon 4 ELISA plates. A standard curve was generated using purified PNAG (0.015-0.1 g/ml) to sensitize ELISA wells in duplicate, and samples were probed with 10 g/ml mAb F598 followed by an alkaline phosphatase-conjugated goat antibody to human IgG. After assay development with the p-nitrophenyl phosphate substrate, A 405 nm readings were obtained; a standard curve of the PNAG in B. subtilis Biofilms JULY 31, 2015 • VOLUME 290 • NUMBER 31 purified PNAG concentration versus A 405 nm was calculated by linear regression, and the amount of PNAG in bacterial cell lysates was calculated after subtracting out the background from the negative control. Samples of cells taken before and after dispersin B treatment were probed as described above for immunochemical detection of PNAG n microbial cells.
Opsonophagocytic Assay-Opsonophagocytic killing (OPK) of B. subtilis DS991, E. coli WT, E. coli (⌬pga), and E. coli (⌬pga) (pUCP18Tc-epsHIJK) was carried out, as described previously (34), except differentiated human promyelocytic HL60 cells were used as the phagocyte source (35). The percent killing mediated by antibodies in immune sera raised to a conjugate of nine residues of ␤-1-6-linked glucosamine and tetanus toxoid (TT; 9GlcNH 2 -TT vaccine) (36) was calculated by dividing the colony-forming units (cfu) in the test sera by those in the corresponding dilution of the nonimmune control serum.

Results
Detection of PNAG in B. subtilis Biofilms/Pellicles-B. subtilis pellicles formed over 10 days of growth at the air-liquid interface in MMG medium were reacted with control mAb F429 or mAb F598 to PNAG, both directly conjugated to Alexa Fluor 488 and visualized for immunofluorescence by confocal microscopy. Bacilli embedded in a strongly immunoreactive matrix of PNAG were readily observed, and binding to mAb F598 was lost after treating the B. subtilis biofilms with the PNAG-degrading enzyme dispersin B or with PNAG-hydrolyzing sodium periodate (Fig. 1).
Expression of PNAG by WT and eps-mutant B. subtilis-Planktonic cells of B. subtilis that overproduced EPS (due to mutation of the master repressor SinR) and also defective for the biofilm-organizing protein TasA (DS991) were also positive for PNAG expression by immunofluorescence microscopy ( Fig.  2A), and immunoreactivity was lost following treatment with FIGURE 1. PNAG production in 10-day-old pellicles of B. subtilis. Pellicles/ biofilms formed at the air-liquid interface of MMG medium over 10 days at room temperature were analyzed via immunofluorescence for the presence of PNAG within these structures. Cells were identified using Syto 83 DNA stain and PNAG by reactivity with mAb F598 directly conjugated to Alexa Fluor 488. Loss or immunologic reactivity following digestion with dispersin B or exposure to 0.04 M periodate are both treatments that degrade PNAG due to their enzymatic and chemical specificity to attack the ␤-1-6-linked N-acetylglucosamine polymer. Bars, 10 m. FIGURE 2. Demonstration of PNAG production by B. subtilis WT and deletion strains by immunofluorescence microscopy. A, parental strain B. subtilis DS991 failed to react with mAb F429 to P. aeruginosa alginate after chitinase treatment but did bind mAb F598 to PNAG. Binding was lost after treatment of cells with dispersin B or sodium periodate. B, binding of mAb F429 (control) or F598 to PNAG to indicated B. subtilis strain with a deletion of in the indicated gene in the eps locus (⌬eps). DNA was stained with Syto 83 to visualize cells, and mAbs were conjugated directly to Alexa Fluor 488 (green). White bars, 10 m. dispersin B or periodate ( Fig. 2A). When PNAG production was evaluated in B. subtilis strains with in-frame deletions in all 15 identified eps genes ( Table 1), most of which are unable to form biofilms, only disruption of the epsH-K genes led to loss of PNAG production (Fig. 2B). The epsH and epsJ genes are annotated as putative GTs and show similarity to proteins encoded by S. aureus icaA and E. coli pgaC ( Table 2). The epsI gene is currently annotated as a putative polysaccharide pyruvyltransferase ( Table 2), but as shown below, its structural predictions suggest other functions. The epsK gene is currently annotated as a putative extracellular matrix component exporter similar to the wzx proteins in E. coli (Table 2). Two other eps genes (epsE and epsF) are annotated as putative GTs, but deletions in these genes did not lead to loss of PNAG production (Fig. 2B). Similarly, an epsG deletion mutant, known to be deficient in biofilm production like the epsH mutant (17), was still able to produce PNAG (Fig.  2B). As with other microbial species, PNAG production appears to be necessary, but not sufficient, for full biofilm formation in B. subtilis. Bioinformatic Analysis of B. subtilis PNAG Biosynthetic eps Genes and Proteins-The above results, along with bioinformatic analysis of the B. subtilis eps locus, indicated the genes required for PNAG production, epsH-K, could comprise a four-gene locus somewhat similar to the S. aureus icaADBC and E. coli pgaABCD loci encoding the PNAG biosynthetic proteins in these two organisms (Table 2) (11,13). To obtain a more precise idea of the functions of the EpsH-K proteins, as well as other proteins within the eps locus, the predicted protein structures were analyzed using the PHYRE 2 server (Fig. 3) (25). Overall, the entire eps locus contains genes encoding proteins known to be involved in polysaccharide synthesis using either the Wzx/Wzy or ABC transporter pathways for lipopolysaccharides, capsules, colonic acid, and S-layer components (37)(38)(39). Both EpsH and EpsJ have nearly identical predicted structures ( Fig. 3) with homology to the GT domains of IcaA and PgaC proteins used in S. aureus and E. coli for PNAG synthesis, respectively (13,40), as well as the BcsA protein used for cellulose (␤-1-4-linked glucose) synthesis (41). With 60 -70% sequence coverage and 15-22% sequence identity, EpsH and EpsJ are predicted with 100% confidence to be UDP-N-acetyl-glucosamine transferases with a GT-A fold. They are not predicted to have any transmembrane domains and therefore would be unable to translocate the synthesized polymer across the membrane.
BLAST analysis showed that EpsK belongs to the polysaccharide biosynthesis protein family, pfam01943. Members of this family are integral membrane proteins that encode for multiple antimicrobial extrusion protein (MATE)-like transporters, such as the O-antigen flippase Wzx. EpsK was predicted to be an ␣-helical membrane transporter with 14 transmembrane helices using PHYRE 2 and various transmembrane domain prediction servers, respectively (Fig. 3) (42). PHYRE 2 predicts EpsK to be structurally similar to DinF, a member of the MATE family, covering 84% of the sequence with 10 -12% amino acid identity and 100% confidence of the fold prediction.
The analysis of the EpsI protein (Fig. 3) indicated it was not related structurally to the PNAG deacetylases IcaB or PgaB (43,44) or any members of the family 4 carbohydrate esterases (deacetylases), but rather it is predicted to be a cytoplasmic GT with homology to genes in lipopolysaccharide biosynthesis. There were no predicted signal sequences or transmembrane helices but almost complete coverage (80 -90%) of the EpsI sequence with 95% confidence in fold prediction to known lipopolysaccharide transferases.
Functional Equivalence of B. subtilis EpsH-K Proteins to E. coli PNAG Biosynthetic Proteins-To determine whether each of these four B. subtilis eps genes encoded proteins that could replace those encoded by the pga operon, we constructed individual in-frame mutations in each of the four E. coli pga genes, pgaA-D, and complemented each of these different mutants with an eps-related gene judged to most likely encode a protein of potential similar function. As controls, we also complemented some of the E. coli mutants with a mismatched gene from B. subtilis. The WT E. coli strain produced PNAG, and deletion of any of the four individual pga genes abolished PNAG production (Table 3). Complementation with an empty vector (pUCP18Tc) did not restore the phenotype in any strain with an in-frame pga mutation. Complementation of the E. coli-pga mutant strains with individual clones of B. subtilis epsH, epsI, epsJ, or epsK resulted in restoration of PNAG production in each E. coli mutant strain with the gene from the B. subtilis locus judged most likely to be a functional equivalent (Table 3). Interestingly, we found that PNAG production in the Ϫ ϩϩϩ E. coli ⌬pgaC strain could be restored by complementation with either the B. subtilis epsJ gene and, to a lesser extent, the epsH gene (Table 3). Similarly, we could achieve phenotypic complementation of PNAG production in the E. coli ⌬pgaA mutant strain most strongly with the B. subtilis epsH gene and, to a lesser extent, the epsJ gene. The E. coli pgaB mutant was efficiently complemented with the B. subtilis epsI gene, and the E. coli pgaD mutant was complemented with the B. subtilis epsK gene ( Table 3). Cloning of the B. subtilis epsI and epsK genes into the E. coli ⌬pgaA mutant did not result in PNAG production. Finally, when the entire pga locus was deleted from either E. coli or K. pneumoniae, the loss of PNAG production could be restored in both strains by complementation with the entire B subtilis epsHIJK locus in pUCP18Tc-epsHIJK (Table 3). We also analyzed the B. subtilis DS991 WT and ⌬epsK strains as well as WT E. coli for the presence of intracellular PNAG. WT E. coli had a low, but detectable, level of intracellular PNAG (Table 4). Both WT and ⌬epsK B. subtilis had detectable intracellular PNAG (Table 4), with the B. subtilis ⌬epsK strain having about 60% more intracellular PNAG, indicating that in the absence of the predicted EpsK transporter, there was accumulation of the PNAG polysaccharide inside the B. subtilis ⌬epsK cells.
Finally, we found that the predicted MATE function of the B. subtilis EpsK protein had similarity to PelG and PslK in P. aeruginosa used for the synthesis of the PEL and PSL polysaccharides, respectively. P. aeruginosa, unlike E. coli and K. pneumoniae, does not normally synthesize PNAG, so we introduced the pUCP18Tc-epsHIJK plasmid into WT and algF-interrupted (Tn::algF) P. aeruginosa strain FRD1 to ascertain whether PNAG could be synthesized. We did not detect PNAG synthesis in either the WT P. aeruginosa strain or the strain carrying pUCP18Tc-epsHIJK, but we did show alginate production by virtue of binding of mAb F429 to these bacterial cells (Fig. 4). We hypothesized that it might be problematic to produce or detect low level PNAG production in WT FRD1 P. aeruginosa if this organism is expressing both the positively charged PNAG and the negatively charged alginate in the same cell. We thus cloned pUCP18Tc-epsHIJK into P. aeruginosa FRD1 (Tn501:: algF), wherein the Tn (transposon) insertion has a polar effect on the alginate biosynthetic locus, leading to loss of alginate production, as verified by loss of binding of mAb F429 to cells of this strain (Fig. 4). Addition of the pUCP18Tc-epsHIJK plasmid resulted in expression of PNAG on the recombinant P. aeruginosa FRD1 (Tn501::algF) cells as detected by immunofluorescence analysis of binding of mAb F598 to PNAG. This binding was lost after treatment of the recombinant P. aeruginosa strains with dispersin B and periodate, indicative of PNAG synthesis in alginate-negative P. aeruginosa by proteins encoded by the B. subtilis epsHIJK genes.
Detection of PNAG by WT and Recombinant B. subtilis and E. coli Strains-To confirm that PNAG was produced by WT B. subtilis but not B. subtilis ⌬epsH or ⌬epsJ, and by WT E. coli E11  4LZ9) shows a MATE transporter fold. The structure models were generated using PHYRE 2 (25) and are shown in schematic representations with ␣-helices, ␤-strands, and loops colored orange, blue, and gray, respectively. Predicted active site residues for the EpsH-J proteins are shown as magenta sticks. and E. coli E11 (⌬pga) (pUCP18Tc-epsHIJK) but not E. coli (⌬pga), we extracted surface material from cells as described (14) and used slot blots for immunologic detection of extracted antigen. As shown in Fig. 5, WT B. subtilis as well as WT E. coli E11 produced immunoreactive PNAG, whereas the B. subtilis ⌬epsH and ⌬epsJ strains did not, nor did E. coli (⌬pga). Complementation of the PNAG-deficient E. coli (⌬pga) strain with the B. subtilis epsHIJK genes restored detectable PNAG production. Confirmation that the immunoreactive material was likely PNAG was obtained by treating extracts with dispersin B or periodate, both of which destroy PNAG. Such treated materials had no reactivity with the mAb to PNAG in a slot-blot assay (Fig. 5). Finally, chemical analysis of the extracts indicated hexosamine was only detectable in the PNAG-positive extracts. Attempts to further purify PNAG to a degree sufficient for analysis by NMR were unsuccessful. This is consistent with almost all prior publications indicating that methods have not yet been developed to purify PNAG sufficiently for NMR analysis in the absence of hyper-expression of the biosynthetic proteins in organisms like S. aureus, E. coli, or A. baumannii (13,14,45,46). In organisms such as S. epidermidis (8) and Vibrio parahaemolyticus (47), isolation of PNAG-related small molecular weight fragments has been achieved only by use of natural hyper-producers of biofilms as sources of the initial extracts.
Opsonic Killing Mediated by Antibody to PNAG-The susceptibility of the B. subtilis and E. coli strains to OPK was tested in an assay using antibodies raised in either a rabbit or goat to the 9GlcNH 2 -TT vaccine (36). Antibodies in both of these antisera readily mediated OPK of WT B. subtilis and WT E. coli (Fig. 6). Deletion of the pga locus in E. coli resulted in no effect of antibody to PNAG on cell survival in an OPK assay, whereas introduction of the pUCP18Tc-epsHIJK plasmid into pga-deleted E. coli restored the susceptibility to OPK. Thus, B. subtilis EpsH-K proteins produced an antigen in E. coli functionally equivalent to native PNAG from this organism.

Discussion
B. subtilis has served as one of the major model organisms for scrutinizing biofilm formation and multicellular activity in bacterial communities (18,48). Many of the factors, genes, and conditions needed for producing and modulating formation of biofilms by this organism are well studied (49). Components of the B. subtilis biofilm include EPS and polymeric substances such as poly-DL-glutamic acid and proteins, including TapA, TasA, and BslA (48). However, the chemistry of the biofilm EPS constituents are not defined, and no definitive polysaccharide structures have previously been identified. Here, we found that within the 15-gene eps cluster of B. subtilis there was a 4-gene locus encompassing epsHIJK that encodes proteins that can synthesize either the conserved bacterial surface polysaccharide PNAG, a common component of microbial biofilms, or an antigenically cross-reactive material. In the B. subtilis biofilm, PNAG, or a related structure, likely serves as a scaffold as well as an anchoring substrate for the other components in the biofilm, which requires gene products from the eps locus other than epsHIJK for full matrix formation, as many of the eps gene products are needed to establish the biofilm phenotype (48).
To synthesize PNAG using the B. subtilis EpsH-K proteins, we speculate that EpsH is an undecaprenyl priming transferase that makes undecaprenyl-3-O-acyl N-acetylglucosamine. EpsI could either be modifying this first GlcNAc or possibly adding on another sugar monomer and also providing a deacetylase function. EpsJ is potentially the poly-GlcNAc transferase that is needed for long chain extension to the lipid linker of the UDP-N-acetylglucosamine precursor to synthesize the PNAG-like molecule, and EpsK either  Extracts of the indicated E. coli or B. subtilis strain were immobilized on membranes and probed with mAb F598 to PNAG followed by anti-human IgG conjugated to HRP. Both WT E. coli and E. coli ⌬pga carrying the cloned B. subtilis epsHIJK genes expressed PNAG, while the ⌬pga strain did not contain detectable PNAG. Similarly, extracts of cells of B. subtilis DS991 had detectable PNAG, but none was present in the strains lacking either the epsH or epsJ genes. Extracts with detectable PNAG lost reactivity with mAb F598 following dispersin B or periodate treatment and contained detectable hexosamine. N/A, not applicable; Pos, positive; Neg, negative.
facilitates the activity of EpsJ and/or transports the polysaccharide out of the cell.
In regard to EpsK, it is unclear how an ␣-helical membrane transporter from a Gram-positive organism would function in PNAG synthesis and transport in E. coli. Using BLAST analysis we noted that EpsK is a member of the Wzx family of proteins (Table 2), which show overall little conservation in their primary amino acid sequences, but Wzx proteins can interchangeably export different polysaccharides containing N-acetylglucosamine or N-acetylgalactosamine as the initiating sugar (50). Therefore, as a member of the Wzx family of proteins, it is entirely plausible that EpsK is transporting the PNAG polymer out of the cell in both B. subtilis and E. coli.
These structural inferences of the EpsHIJK proteins are consistent with the predictions that synthesis of PNAG, alginate, and cellulose, all components of microbial biofilms, is mediated by proteins with similar hallmarks and functions but not necessarily with conserved architectures (41,51). Moreover, it appears that some of the proteins, despite little overall amino acid sequence identity, nonetheless maintain sufficiently conserved and interchangeable functions, as documented here for the B. subtilis and E. coli PNAG biosynthetic proteins.
We found that deletion of these genes in B. subtilis disrupted PNAG production as detected by loss of immunoreactivity with mAb F598. Cloning either the epsHIJK genes individually into E. coli strains with in-frame deletions in the endogenous pgaA-D genes, or as a 4-gene cluster into pgaABCD-deleted E. coli or K. pneumoniae, resulted in production of immunochemically detectable PNAG, as did cloning the epsHIJK genes into alginate-deficient P. aeruginosa. Expression of epsHIJK in E. coli mutants deficient in pga genes could result in either production of authentic PNAG or a material that is a component of another E. coli factor, such as an LPS or a glycoprotein. We also found that the epsH and epsJ genes can both complement the E. coli pgaA and the pgaC genes, yet in both of these organisms loss of any one of these genes results in a PNAG-negative phenotype. It appears there is no cross-complementation within the host strain. Although the structural similarity of EpsH and EpsJ explains, in part, the ability of each to provide some crosscomplementation, it seems that they likely have some additional nonshared specific functions within B. subtilis that are provided by other proteins when they are expressed in E. coli.
Although PNAG itself, or a structure containing a PNAG component, appears to be an EPS element involved in B subtilis biofilm formation, it is likely that other EPS molecules are also needed for full pellicle formation. Studies of the eps locus indicate that deletion of numerous other genes also disrupts biofilm formation, and some of these appear to be classic EPS biosynthetic genes (17,18,48). The EpsM-O proteins are predicted to be an acetyltransferase (EpsM), an aminotransferase/sugar dehydratase (EpsN), and a GT with a GT-B fold (EpsO). The EpsE protein is involved in both motility and biofilm formation (21) acting as a clutch of the flagellar cellular motility apparatus to inhibit movement, while also having a predicted GT structure.
Recently, Elsholz et al. (52) reported that the B. subtilis EPS serves as a positive regulator of its own synthesis by binding to the extracellular portion of a receptor encoded by the epsA gene that interacts with a tyrosine kinase encoded by epsB and inhibits EpsB autophosphorylation. Dephosphorylated EpsB is associated with enhanced EPS production. In an epsH mutant unable to produce biofilm and, as shown here, PNAG, a His-tagged recombinant EpsB protein was highly phosphorylated due to lack of EPS production. We found neither EpsA nor EpsB was needed for synthesis of immunoreactive PNAG in B. subtilis. This is consistent with the finding from Elsholz et al. (52) that PNAG-related material prepared using a modified method from Mack et al. (8) that yields small molecular weight polysaccharide intercellular adhesin had no effect on EpsB autophosphorylation. Given that the epsH mutant was not producing a factor binding to the EpsA extracellular domain, it appears that other EPS components either depend on EpsH for synthesis or PNAG is needed to facilitate the inhibition of EpsB autophosphorylation. However, the actual factors regulating B. subtilis EPS production via EpsA and EpsB are currently uncharacterized in regard to specific chemical properties.
One final aspect to consider is that the polysaccharide synthesized by the B. subtilis EpsH-K proteins may not be PNAG but a molecular entity with sufficient ␤-1-6-linked N-acetylglucosamine in it to make it immunoreactive with antibodies to PNAG but also containing other components. mAb F598 binding to PNAG requires a minimum of seven ␤-1-6-linked N-acetylglucosamine residues (36,53), indicating a minimal PNAG-related constituent present in the material synthesized by the EpsH-K proteins. We do note that every microbial strain wherein we or others have identified a dispersin B and perio- date-sensitive, hexosamine-containing antigen that reacts with antibodies to native PNAG, and wherein the reactive material has been isolated, turns out to be chemically verified PNAG (14, 45, 54 -56). Furthermore, only 1-6-linked hexosamines, and no other possible amino-sugar linkages, are sensitive to periodate, and only ␤-1-6-linked N-acetylglucosamine can be digested by dispersin B (22). Also, for every microbe for which genes encoding PNAG biosynthetic proteins have been identified and deleted, there is loss of antibody reactivity with material on the cell surface due to gene loss. Thus, we consider it highly unlikely that the material produced by the EpsH-K proteins is not PNAG, but at a minimum, it is at least a PNAGcontaining molecular structure.
Overall, our results show PNAG, or a closely related entity, is a component of the B. subtilis biofilm matrix synthesized by the EpsH-K proteins. These proteins can also be expressed and are functional in Gram-negative organisms, including E. coli, K. pneumoniae, and P. aeruginosa. However, it is possible that the EpsH-K proteins are also required for producing other polysaccharide components of the biofilm matrix, inasmuch as PNAG does not appear to be the only carbohydrate entity in this organism's biofilm (48). However, no carbohydrate component other than PNAG identified here has been fully characterized as being synthesized by proteins encoded by genes within the B. subtilis eps locus, leaving open the question as to the chemical composition of additional eps-dependent polymeric carbohydrates present in this structure. Analysis of EPS composition would undoubtedly advance the understanding of the B. subtilis biofilm formation process, but it should be noted that published methods used to isolate EPS (52) would contain very little native PNAG, as this molecule is poorly soluble at neutral pH, particularly after alcohol precipitation (14,45). Thus, in the absence of an appropriate method, researchers have not been able to purify PNAG to obtain an appreciable yield for detailed chemical structural determination. We did find, however, that extracts from the WT B. subtilis cells and recombinant E. coli cells carrying the pUCP18Tc-epsHIJK plasmid contained immunoreactive dispersin B and the periodate-sensitive hexosamine-containing material present. Thus, the analysis of the products of the B. subtilis epsHIJK locus described here are fully consistent with the production by WT B. subtilis of PNAG, which is likely a necessary component of this organism's biofilm based on the genetic data showing the essentiality of epsH-K gene products for production of this structure (19).