INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Many studies and reviews have discussed the important roles of polysaccharides in bacterial physiology and pathogenesis (2, 15, 46). As components of bacterial cell walls, capsules, lipopolysaccharides of gram-negative organisms, and accessory polymers of gram-positive cell walls, they not only offer structural rigidity for maintaining the shape of cells but also affect surface permeability, charge, and hydrophobicity and consequently affect the way bacteria interact with the environment. One important aspect of polysaccharides relates to their immunological and biological properties in tissues of the host, such as inhibition of phagocytosis and stimulation of cytokine production (17). These properties are critical in determining the outcome of the intricate interactions between bacteria and the host immune system. Many polysaccharides are also the immunodeterminants of serotype-specific antigens, including the O-antigen side chains of lipopolysaccharide and group and type antigens of some streptococci. Thus, it is clear that knowledge of cell surface architecture is important in understanding bacterial behavior inside the host, as well as in designing new therapeutics and serodiagnostic tools.

Biosynthesis of polysaccharides generally involves synthesis of sugar precursors in the cytoplasm, formation and polymerization of repeating units, and export to the cell surface. The formation of repeating units usually is initiated by the transfer of a sugar precursor onto a lipid carrier, such as undecaprenol-phosphate, to which subsequent sugar residues can be added. The sugar residue attached to the lipid carrier is often the first sugar in the repeating units (1, 44, 56), although there are exceptions as in Escherichia coli O8 and O9 serotype-specific O antigens in which the sugar residues attached to the lipid carrier do not appear in the final polysaccharides (41). Genes for the biosynthesis of polysaccharides are generally arranged in clusters of one or several transcriptional units (13, 32, 44, 56).

Enterococci (formerly group D streptococci) are among the leading causes of nosocomial infections in the United States, with the majority of clinical isolates being classified as Enterococcus faecalis (34). Besides their low-level intrinsic resistance to many antibiotics, enterococci have also developed high-level resistance to other antibiotics, including vancomycin and -lactams (34). Very little information is available on cell surface-associated polysaccharides of enterococci. Early studies on the type-specific antigens in enterococci indicated that they were polysaccharides in the cell walls (6, 19, 45, 47). Type 1 antigen seems to be the most frequently identified type among E. faecalis isolates, at least in some studies (45, 47). Crude cell wall extracts from E. faecalis type 1 strains were found to contain rhamnose, glucose, galactose, N-acetylglucosamine (GlcNAc), and N-acetylgalactosamine (6). Pazur et al. demonstrated that enzymatic activities to synthesize thymidine diphosphate glucose (TDP-Glc) and to convert TDP-Glc to thymidine diphosphate rhamnose (TDP-Rha) were present in the cell extracts of E. faecalis N (40). TDP-Glc and TDP-Rha are the precursors from which glucose and rhamnose can be transferred to synthesize polysaccharides. An immunogenic glycan was also found in the cell walls of E. faecalis N (38, 39). The glycan was shown to be a diheteroglycan of glucose and galactose, but no rhamnose was present. No genetic information on polysaccharide biosynthesis in enterococci is available.

In a previous study, we reported a cosmid clone carrying a DNA fragment from E. faecalis OG1RF that produced a nonprotein antigen in E. coli (57). We hypothesized that the antigen could be a polysaccharide of OG1RF. To test this hypothesis, we have carried out further studies including DNA sequence analysis and characterization of transposon mutants. The results showed that the DNA fragment contains a gene cluster with homology to genes important in the assembly and export of polysaccharide from various organisms. Transposon insertions in three open reading frames (ORFs) within the cluster caused the loss of antigen expression in E. coli.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains, plasmids, and culture conditions. Strains and plasmids used in this study are listed in Table 1. E. coli cells were grown in Luria-Bertani (LB) broth or on LB agar with appropriate antibiotics overnight at 37°C. Enterococci were grown in brain heart infusion (BHI) broth or on BHI agar (Difco) overnight at 37°C for routine purposes. Antibiotics were used at the following concentrations: chloramphenicol at 25 µg/ml, kanamycin at 50 µg/ml, and ampicillin at 50 µg/ml. Isopropylthio--D-galactoside (IPTG) and 5-bromo-4-chloro-3-indolyl--galactoside (X-Gal) were used at 0.5 mM and 80 µg/ml, respectively.

In vitro transposition. Cosmid DNA from clone TX5159 (formerly BO4B6I) (57) was prepared by the alkaline-sodium dodecyl sulfate (SDS) method and purified in a CsCl-ethidium bromide gradient as described by Sambrook et al. (43). In vitro tranposition using a Tn7-based system (3, 4, 48) was performed by Matthew C. Biery and Nancy L. Craig, Department of Molecular Biology and Genetics, Howard Hughes Medical Institute, Johns Hopkins University.

DNA manipulations and transformation of E. coli. DNA preparation, purification, restriction digestion, agarose gel electrophoresis, and ligation were performed by using standard methods (43) unless otherwise stated. Routine preparation of E. coli competent cells and transformation of DNA into E. coli were performed by the one-step procedure (9). Transformation of cosmid DNA that had been subjected to in vitro transposition was carried out by using Subcloning Efficiency DH5 competent cells (GIBCO BRL, Life Technologies, Gaithersburg, Md.); transformants were randomly picked and stored in 96-well microtiter dishes in LB broth containing 15% glycerol.

Proteinase K, periodate treatment, SDS-polyacrylamide gel electrophoresis (PAGE), and Western blot analysis. Proteinase K treatment of the cell lysates of E. coli clones was based on a method described previously (57), with slight modifications. Briefly, 20 ml of an overnight culture of E. coli clones was centrifuged and resuspended in 700 µl of phosphate-buffer saline. The suspension was split into two aliquots; 50 µl of proteinase K (50 mg/ml); Fisher Scientific, Fair Lawn, N.J.) was added to the first aliquot, and 4 µl of 100 mM phenylmethylsulfonyl fluoride and 50 µl of H₂O were added to the second. The aliquots were incubated at 50°C overnight and stored at 20°C. Treatment by periodate followed part of the procedure for the ECL Glycoprotein Detection Module (Amersham Life Science, Little Chalfont, Engalnd). Twenty-five microliters of each of the above aliquots (proteinase K treated and untreated) were mixed with 25 µl of 200 mM acetate buffer (pH 5.5.), then 25 µl of 30 mM sodium metaperiodate (freshly prepared) was added, and the mixture was incubated at room temperature for 1 h in the dark. The reaction was stopped by addition of 25 µl of 20 mM sodium metabisulfite and stored at 20°C in the dark. For controls, 25 µl of 200 mM acetate buffer instead of 30 mM sodium metaperiodate was used.

Immunoblot analysis of bacterial colonies. Colonies were inoculated onto NitroBind nitrocellulose transfer membranes (Micron Separations, Inc., Westborough, Mass.) placed on LB agar plates containing proper antibiotics and were then incubated at 37°C overnight. The membranes were lifted, placed in a chloroform chamber for 15 min, and incubated in lysis buffer. Subsequent steps were as described previously (57).

PCR amplification from the ends of the transposon and purification of PCR products. NLC94 and NLC272 are primers that allow DNA synthesis to proceed outward from the ends of the Tn7 transposon. Primers GW248, GW250, GW251, GW254, GW273, GW280, and GW283 are primers hybridizing to regions in the insert of the cosmid. The sequence and location of each primer are summarized in Table 2 (see also Fig. 3).

Sublconing of restriction fragments into pBluescript SK (). Subcloning was based on the method described previously (57), with sight modifications. Cosmid DNA was digested with EcoRI, BamHI, or DraI and purified by phenol-chloroform extraction followed by ethanol precipitation. Vector pBluescript SK () DNA was digested with EcoRI, BamHI, or EcoRV (to ligate with DraI) and dephosphorylated with shrimp alkaline phosphatase (United States Biochemical, Cleveland, Ohio). After ligation, the mixture was transformed into E. coli DH5 and plated on LB plates with ampicillin, IPTG, and X-Gal. White colonies were picked and analyzed by restriction digestion of miniprep DNA.

DNA sequencing and sequence analysis. Three types of templates were prepared for sequencing: DNA from pBluescript SK () clones prepared by using a Qiagen Plasmid Mini kit (Qiagen Inc., Chatsworth, Calif.); purified PCR products; and TX5159 cosmid DNA purified in a CsCl-ethidium bromide gradient. Primers hybridizing to the T3 and T7 regions in pBluescript SK () and to the ends of the Tn7 transposon, as well as primers hybridizing to the ends of the sequences from the previous rounds, were used as sequencing primers. Sequencing reactions were performed by the Taq Dye-deoxy Terminator method and a model 377 DNA sequencing system (Applied Biosystems, Foster City, Calif.). The GelAssemble program in the Genetics Computer Group (GCG; Madison, Wis.) software package was used to assemble the nucleotide sequences. Gapped BLASTx at the National Center for Biotechnology Information (NCBI) was used to search for homologous sequences in the database. ORF Finder at NCBI was used to identify potential ORFs in the sequences. The terminator program in the GCG package was used to search for terminator-like sequences. The ProtParam program at ExPASy was used to calculate the theoretical parameters including molecular weight, isoelectric point, and hydrophobicity of each deduced amino acid sequence. Hydropathy plots were obtained by using the ProtScale program at ExPASy according to the method of Kyte and Doolittle (24) with a window size of 9. The DAS (dense alignment surface method) (10) transmembrane segment prediction program was also used to predict transmembrane domains. Potential signal peptide cleavage sites were analyzed by using SignalP (37). Overall amino acid sequence comparison was done with the Bestfit program in the GCG package, using the default parameters. Potential ribosome-binding sites (RBSs) and promoter sequences were examined manually.

Elution of antibodies specific to the polysaccharide. Cell lysates of TX5159 were treated with proteinase K, subjected to SDS-PAGE, transferred to nitrocellulose membranes, and incubated with one of the patient sera as described above. The membranes were then incubated with 10 ml of 100 mM glycine (pH 2.5) for 30 min at room temperature. After neutralization with 1 ml of 1 M Tris (pH 8.0), the solution was transferred to a clean tube and stored at 20°C until ready to use (16).

Nucleotide sequence accession number. The nucleotide sequence reported here was submitted to GenBank under accession no. AF071085.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Periodate treatment of E. coli(pTEX5159) lysates. In our previous study, TX5159 (formerly BO4B6I) was found to react with four patient sera but not with a rabbit serum prepared against surface protein extracts from an E. faecalis strain isolated from a patient with endocarditis (57). The supernatant from an overnight culture of TX5159 also reacted with one of the patient sera in a dot blot (data not shown). The antigenic material in the TX5159 cell lysates was resistant to extensive proteinase K degradation, suggesting the material was not a protein. To further examine the nature of this antigen, the lysate was subjected to treatment with sodium metaperiodate, which preferentially oxidizes carbohydrates, and then subjected to Western blot analysis using one of the patient sera (Fig. 1). The immunostaining disappeared completely after periodate treatment, suggesting that the antigenic material is a carbohydrate. The material was produced in E. coli in sufficient amounts so that after a 1:32 dilution of the sample (equivalent to 3 to 4 µl of the overnight culture), there was still a strong reaction on the Western blot. The appearance of the diluted material in the Western blot resembled the ladder pattern seen with repeating units of polysaccharides; 10 bands could be clearly observed. A cosmid clone previously found to encode a protein antigen, the autolysin of E. faecalis (57), was used as a control. This antigen band disappeared after proteinase K treatment but remained intact after incubation with periodate.

View larger version (56K): [in this window] [in a new window]   FIG. 1.   Effects of proteinase K and periodate on cell lysates of TX5159. Cell lysates of an overnight culture were treated with either proteinase K or periodate. The samples were loaded for SDS-PAGE and transferred to a membrane after electrophoresis. The membrane was blotted with patient serum at 1:500 dilution in 1% skim milk. Lane 1, TX5159; lane 2, TX5159 plus proteinase K; lane 3, TX5159 plus periodate; lane 4, TX5159, diluted 32-fold; lane 5, diluted TX5159 lysates plus periodate; LO6E6II was (used as a control for the treatments); lane 6, LO6E6II lane 7, LO6E6II plus proteinase K; lane 8, LO6E6II plus periodate.

DNA sequencing of pTEX5159 and sequence analysis of contig C. DNA sequencing resulted in four cotigs (A to D) along the insert of pTEX5159 (Fig. 2). The total number of bases sequenced is 94,434, and the total unique sequence is 34,170 bases long. The average coverage of the four contigs is 2.76×, and the coverage of contig C, is 3×, slightly higher than for the other contigs. About 17.5% of the total unique sequence are single pass, most at the ends of each contig. A restriction map of the 43-kb insert of TX5159 was generated by using the information from DNA sequencing and restriction enzyme digestion (RED) analysis of transposon insertion clones made by in vitro tranposition (described below) (Fig. 2). Based on the restriction map, gaps between contigs B and C and between contigs C and D should be less than 500 bp, while a gap of about 4.5 kb is present between contigs A and B. Database comparison of the sequences from the four contigs showed that contig C contains ORFs with similarity to genes involved in polysaccharide biosynthesis.

View larger version (10K): [in this window] [in a new window]   FIG. 2.   Restriction map of the insert in pTEX5159. The top line shows the scale. The shaded boxes represent contigs A to D. Below the boxes, the horizontal line represents the insert, the short vertical lines above the horizontal line represent the EcoRI sites, and the vertical lines below show the HindIII sites. The EcoRI fragments were designated r1 to r10, with r1 being the largest in size and r10 the smallest, while the HindIII fragments were designated h1 to h15. Fragment h2A had the same mobility in agarose gel electrophoresis as h2B; the same was true of h3A and h3B.

View larger version (21K): [in this window] [in a new window]   FIG. 3.   The polysaccharide biosynthesis gene cluster and flanking regions. The ORFs are shown as open boxes, the direction of the arrows underneath the boxes indicates the direction of transcription, and the best-gapped BLASTx hits of some of the ORFs are indicated above the boxes. The six ORFs that showed some similarities to glycosyltransferases are labeled G. T. Tn7 transposon insertions are shown as lollipops. The name of each cosmid carrying the corresponding insertion is given underneath followed by pluses and/or minuses in parentheses to indicate the immunoreactivity. The sign before the slash indicates the immunoreactivity in DH5, and the sign after the slash indicates the immunoreactivity in S874. Open-circle lollipops represent transposon insertions that rendered the clones immunonegative, filled-circle lollipops represent those that did not affect the immunoreactivity of the clones, and shaded-circle lollipops in contig D are those that had undergone DNA rearrangements in S874. The black bar below orfde4 and orfde5 indicates the region where the initial 10 immunonegative clones were mapped. Three of these 10 clones, pTEX5159D to pTEX5159F, are shown as open-circle lollipops below the black bar. Primers used in characterizing transposon insertions are shown at the bottom, with the direction of arrows indicating the direction of DNA synthesis. Primers are not drawn in scale. Bm, BamHI; HIII, HindIII; Kn, KpnI; RI, EcoRI; Xb, XbaI.

Database similarities of the genes in contig C and characteristics of the deduced gene products. BLASTx was performed to search for sequence similarities in the databases (Table 4). Orfde1 showed strong similarity with a putative methionine aminopeptidase A (map) from Synechocystis sp. (21) and various other organisms. Orde2 showed some similarity to the serum resistance locus BrkB protein of Bordetella pertussis (11), which is required for the resistance of B. pertussis to complement-dependent killing by normal human serum and contains domains with homology to some transporters. Hydrophobicity analysis indicated that the orfde2 gene product is a hydrophobic protein with four to six potential transmembrane regions. A putative signal peptide cleavage site was found between amino acids 50 and 51 (LTA-VG).

View larger version (23K): [in this window] [in a new window]   FIG. 4.   Comparison of amino acid sequences of dTDP-rhamnose biosynthesis genes from various bacteria. Percentage identity was obtained by using Bestfit; each number in parentheses indicates the length of the segment that gave the percentage identity in amino acids. Default parameters were used for comparisons. Orfde6 to Orfde9 are from E. faecalis OG1RF. Cps19fL to -O, Streptococcus pneumoniae type 19F; smu, Streptococcus mutans; eco, Escherichia coli; sen, Salmonella enterica; sf1, Shigella flexneri. Cps19fL, Rm1Asmu, RfbAeco, Rm1Asen, and RfbAsf1, glucose-1-phosphate thymidyltransferases; Cps19fM, Rm1Csmu, RfbCeco, and RfbCsf1, dTDP-4-dehydrorhamnose 3,5-epimerases; Cps19fN, Rm1Bsmu, RfbBeco, RmlBsen, and RfbBsf1, dTDP-glucose-4,6-dehydratases; Cps19fO, Rm1Dsmu, RfbDeco, RmlDsen, and RfbDsf1, dTDP-4-keto-L-rhamnose reductases.

Sequence analysis of contig D. Contig D has 2,943 bases. One ORF and 1,029 bp was identified in the middle of contig D and did not reveal any database similarities. The end of contig D was from vector pBeloBAC11. The rest of the sequence in the contig did not show any similarities to sequences in the database.

Immunoscreening of TX5159 with transposon insertions. We screened 282 transposon insertion clones, generated by in vitro transposition of a Tn7 derivative, by using a patient serum and subjected immunonegative clones to a second screen. Ten clones, DH5(pTEX5159D) to DH5(pTEX5159M), were found to be immunonegative. Five of the 10 cones had single insertions in EcoRI fragment r4, while the other five had double insertions, with one in r4 and the other in either r1 or r5. Clones with single insertions in only r1 or r5 were still immunopositive, suggesting that the insertions in r4 were responsible for the loss of the immunoreactivity of these clones. Cosmid DNA from three clones, DH5(pTEX5159D), DH5(pTEX5159E), and DH5(pTEX5159F), were transformed into S874, and the transformants were tested by colony immunoblotting. S874(pTEX5159D) was negative, while transformants with pTEX5159E and pTEX5159F were not able to regrow after the initial appearance on the primary transformation plates.

Characterization of transposon insertion mutants. To determine the sites of the insertions, cosmid DNA from the 10 clones that were immunonegative was prepared and subjected to PCR amplification (Table 5). All insertions in EcoRI fragment r4 were localized to a small region, shown as a black bar in Fig. 3. PCR products amplified by using primer pairs GW248-NLC94 and GW250-NLC272 from three immunonegative clones, DH5(pTEX5159D), DH5(pTEX5159E), and DH5(pTEX5159F), were gel purified and subjected to DNA sequencing. The sequences showed that the insertions were in orfde4 and orfde5, which encode potential glycosyltransferases (Fig. 3). The insertion in DH5(pTEX5159D) was at nucleotide position 4191 in contig C, 361 bp downstream of the start site of orfde4 (equivalent to amino acid 119 from the N terminus of Orfde4); the insertion in DH5(pTEX5159E) was at 5238, in the middle of orfde5 (equivalent to amino acid 214 from the N terminus of Orfde5); and the insertion in DH5(pTEX5159F) was at 4701, 105 bp downstream of the start of orfde5 (equivalent to amino acid 35 from the N terminus of Orfde5). The sites of insertion in the other seven clones were estimated based on their PCR product sizes in relation to those from the three clones that had been sequenced. It appeared that all were in orfde4 or orfde5. The insertion in DH5(pTEX5159L) could be either at the end of orfde4 or at the beginning of orfde5.

View larger version (16K): [in this window] [in a new window]   FIG. 5.   Complementation of DH5(pTEX5159D) with pTEX5175. Samples were treated with proteinase K and analyzed as described in the text. Lane 1, DH5(pTEX5175); lane 2, DH5(pTEX5159D); lane 3, DH5(pTEX5159D + pTEX5175).

Complementation by E. coli O-antigen biosynthesis genes. After RED analysis of random clones, three clones, DH5(pTEX5159A), DH5(pTEX5159B), and DH5(pTEX5159C) were found to have insertions in r4 but still reacted in colony immunoblots. PCR products amplified by using primer pairs GW283-NLC94 and GW280-NC272 from these three clones were subjected to DNA sequencing (Table 5). The insertion in DH5(pTEX5159C) was at nucleotide position 3728 in contig C, which is in the region between orfde3 and orfde4 (Fig. 3). This finding suggested that orfde4 was transcribed independently.

Detection of the expression of this polysaccharide in E. faecalis. Antibody was eluted from nitrocellulose membranes following Western blotting of proteinase K-treated extracts from TX5159 with a patient serum. The eluted antibody reacted with TX5159 but not with the cloning host DH5. The eluted antibody was used to examine E. faecalis OG1RF and TX52 (57) under different growth conditions, including exposure to horse and rabbit sera, at different temperatures, and in minimal growth media. Several clinical from patients with E. faecalis endocarditis were also examined after growth in BHI broth. These clinical strains include isolates from patients from whom the tested sera were collected. No positive reaction was detected (data not shown).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In a previous study, we described a cosmid clone of E. faecalis (TX5159) that produced proteinase K-resistant antigenic material in E. coli (7). In this report, we provide evidence that the antigenic material was carbohydrate in nature and that its production did not involve the E. coli rfb genes. Furthermore, the antigen appears as a ladder after PAGE, suggesting a structure of repeating units, such as those of polysaccharides. Since the clone contains a number of genes with sequence similarity to polysaccharide biosynthesis functions, we conclude that genes for the biosynthesis of an enterococcal antigenic polysaccharide had been cloned and expressed in E. coli.

Early studies indicated that antigenic polysaccharides could be detected in the cell walls of enterococci and, moreover, that many of these were rhamnose-containing polysaccharides (6, 19, 45, 47). However, no studies to our knowledge suggested the presence of capsules in enterococci. In our previous report, TX5159 was found to react with four patient sera but not with a rabbit serum prepared against surface protein extracts from an E. faecalis strain grown in BHI (57). The rabbit serum was made from rabbits immunized with zwittergent surface extracts of E. faecalis, and it is possible that the polysaccharide was not present. In this study, we failed to detect the antigen in E. faecalis OG1RF and TX52 (57) under different growth conditions or in several E. faecalis clinical isolates. This raised the possibility that the antigen is due to E. coli and not enterococcus. PCR amplification of OG1RF genomic DNA by using intragenic primers to orfde2 to orfde16 resulted in PCR products of the expected sizes (data not shown). Southern blot analysis using pTEX5159 DNA as a probe against genomic DNA of OG1RF, TX52, and DH5 indicated that the probe hybridized to OG1RF and TX52 but not to DH5 (data not shown), suggesting that they are enterococcal genes. Therefore, in this model, the patients would also be making antibody against the E. coli polysaccharide, and the enterococcal genes would be complementing E. coli genes that were defective in the laboratory strains. We note that in our screens for immunonegative transposon mutants in DH5, where all the mutants carried insertions in orfde4 and orfde5, orfde4 shows some sequence similarity to orf264 in the gnd-rfc intergenic region of E. coli (27, 59). In most of the E. coli K-12 strains, there is an IS5 insertion in orfde264, which is responsible for the loss of O-antigen production in these K-12 strains. This finding raised the possibility that orfde4 complemented the defective orf264 of DH5 and thus restored its ability to make an O antigen. This possibility can be ruled out because of the following observations. Firstly, unlike DH5, which still maintains most of the rfb genes, S874 lacks the entire rfb region for O-antigen biosynthesis. However, TX5169 had the same reaction pattern with the patient serum as TX5159. Second, plasmid pTEX5175, which contains the complete sequence of orfde4, orfde5, and their flanking regions, alone did not make DH5 immunoreactive. Thus, the production of the antigenic polysaccharide required the presence of pTEX5159 but not the E. coli rfb genes.

The four patient sera used were collected from endocarditis patients infected with E. faecalis in different regions in the United States. The fact that TX5159 reacted strongly with all four sera suggested that the polysaccharide was produced in all of these patients and was not due to one particular strain or one particular patient. It has long been known that the biosynthesis of many polysaccharides (e.g., E. coli group IA and group II K antigens, and colanic acid [56]) is regulated, and many regulatory strategies have been used to achieve control of expression in response to different changes in the environment. Thus, expression of the polysaccharide in E. faecalis may be controlled so that it is only expressed under certain conditions, such as those found in infection. The use of more sensitive detection methods with a wider range of growth conditions is under way. Preliminary results from reverse transcription-PCR using primers to orfde4 and orfde6 suggest that the mRNA ttanscripts to those two genes could be detected in RNA extracts from OG1RF in exponential phase but not in stationary phase (data not shown). The other ORFs in contigs C and D are now being tested.

Comparison of the ORFs in contig C to database sequences revealed similarities to genes involved in polysaccharide biosynthesis and export from both gram-negative and gram-positive organisms. Orfde3 showed similarities to the undecaprenyl-phosphate -N-acetylglucosaminyltransferase encoded by the rfe genes from various organisms and Llm of S. aureus (29). The activity of the lipophilic protein encoded by llm is not known, although it has been proposed to be involved in the metabolism of cell surface components such as peptidoglycan (29). Rfe is required for the synthesis of the first lipid-linked intermediate in the biosynthesis of enterobacterial common antigen and several O antigens (1, 41, 42). Hydropathy plots of the enterococcal Orfde3 and Rfe of E. coli showed remarkable similarities between the two, indicating similar overall distribution of hydrophobic residues. However, a Tn7 insertion near the 3'end of orfde3 (pTEX5159A) did not affect the expression of the polysaccharide in E. coli 21548 (rfe), S874 (rfb), AB1133 (rfe⁺), or DH5. Several possibilities may explain this result: (i) the C-terminal 60 or so amino acids may not be required for enzymatic activity, (ii) the gene is not required for the synthesis of the polysaccharide in E. coli, and (iii) orfde3 codes for some other function(s) instead of the undecaprenyl-phosphate -N-acetylglucosaminyltransferase and is complemented by similar functions from E. coli. Further mutational studies of orfde3 may help answer some of these questions. One conclusion that can be drawn from results with pTEX5159A is that the insertion did not affect the expression of downstream genes, consistent with the result for DH5(pTEX5159C), which had an insertion between orfde3 and orfde4 and was immunopositive, as well as the observation that orfde4 appeared to have its own promoter, based on DNA sequence.

orfde6 to orfde9 showed high sequence similarities to genes in the dTDP-rhamnose biosynthesis pathway of various organisms. As noted above, rhamnose has been found as a component in enterococcal polysaccharides; the transposon insertion in orfde8 (pTEX5159B) resulted in loss of immunoreactivity in S874 (rfb) but not in DH5. This finding indicated that the E. coli rfbBDAC genes could provide this function and provides further support for the conclusion the orfde6 to orfde9 encode a dTDP-rhamnose biosynthesis pathway.

The G+C contents of orfde6 to orfde9 are very close to each other, from 38.2 to 38.5%, while those of the other ORFs in the cluster vary from 32.6 to 40.6%, suggesting that orfde6 to orfde9 may have originated from one source while the other ORFs in the cluster may have been assembled from several sources during evolution. This is consistent with observations in other bacteria such as E. coli and Salmonella spp. (20, 30, 49, 55, 59).

Several reports have shown the involvement of ABC transport systems in the translocation of sugar polymers across the cytoplasmic membrane in both gram-negative and gram-positive bacteria )14, 25, 56, 61) or the presence of ABC transporter genes within polysaccharide biosynthesis clusters (5, 61). It seems that the use of ABC transporters could be one of the common mechanisms for exporting sugar polymers. The strong sequence similarities of orfde11 and orfde12 to components in the ABC transport system, in particular to the ABC tranporters for teichoic acids of B. subtilis (25) and the O antigen of M. xanthus (14), suggest that these two ORFs may play similar roles (14) in the export of the E. faecalis polysaccharide to the cell surface.

The products of six ORFs, orfde4, orfde5, orfde5_6, orfde10, orfde15, and orfde16 (3' partial), showed various degrees of similarities to glycosyltransferases. One feature of these ORFs is that, except for orfde10, they encode hydrophilic proteins, consistent with the findings for other glycosyltransferases and the notion that the assembly of repeat units occurs on the cytoplasmic side of the membrane. The sequence homology of these ORF products to glycosyltransferases was not very strong, as has been recognized in general among glycosyltransferases. Thus, it would be difficult to propose, based on sequence comparison, the substrate specificity encoded by these ORFs. Further analysis, such as enzyme activity assays, is needed to elucidate their functions. Transposon insertions in orfde4 and orfde5 abolished the immunoreactivity in both DH5 and S874, suggesting that these two genes are required for the production of the polysaccharide and that there are not equivalent functions in the E. coli host strains. The introduction of pTEX5175 into DH5(pTEX5159D) restored its immunoreactivity, suggesting that its transposon insertion did not affect the expression of orfde6 and downstream genes. This is consistent with the results from sequence analysis that orfde4 and orfde6 may be transcribed from their own promoters and are in two individual transcriptional units.

Analysis of Tn7 transposon insertion mutants also provided evidence that orfde1 and orfde2 are not essential for antigen production. The involvement of orfde3 is not certain and requires further investigation. No Tn7 insertions in the other ORFs in contig C were identified. However, based on sequence analysis, it is likely that the other ORFs are also involved in the synthesis and export pathway. Tn7 insertions in contig D resulted in large deletions or rearrangement in the cosmid, suggesting that disruption in this region may be toxic to the host. Whether contig D is required in the biosynthesis pathway remains unclear.

In conclusion, a gene cluster of OG1RF was cloned and produced an antigenic polysaccharide in E. coli. Analysis of DNA sequences and transposon mutants indicated that it is a multicistronic region containing genes for dTDP-rhamnose synthesis, assembly of repeat units, and export of the polysaccharide. This is the first report of an enterococcal gene cluster involved in the biosynthesis of an antigenic polysaccharide. Further study of this cluster and the effects of such a polysaccharide on bacterial cell surface permeability, response to antimicrobial agents, and the host immune system may have important implications for research in enterococci as well as for polysaccharides in other organisms.