Report

Human Symbionts Use a Host-Like Pathway for Surface Fucosylation

See allHide authors and affiliations

Science  18 Mar 2005:
Vol. 307, Issue 5716, pp. 1778-1781
DOI: 10.1126/science.1106469

Abstract

The mammalian intestine harbors a beneficial microbiota numbering approximately 1012 organisms per gram of colonic content. The host tolerates this tremendous bacterial load while maintaining the ability to efficiently respond to pathogenic organisms. In this study, we show that the Bacteroides use a mammalian-like pathway to decorate numerous surface capsular polysaccharides and glycoproteins with l-fucose, an abundant surface molecule of intestinal epithelial cells, resulting in the coordinated expression of this surface molecule by host and symbiont. A Bacteroides mutant deficient in the ability to cover its surface with l-fucose is defective in colonizing the mammalian intestine under competitive conditions.

The ability of humans to tolerate a complex gut microbiota despite their exquisite ability to distinguish self from nonself has been called an “immunological paradox” (1). One mechanism that may contribute to the tolerance of these resident microorganisms is molecular mimicry, whereby the bacteria display surface molecules resembling those of the host's surface to render them immunologically inert. Immunologic similarities between the abundant colonic microorganisms Bacteroides and tissues of the host are known (2, 3).

The surfaces of intestinal epithelial cells are covered with an abundance of terminally fucosylated glycoproteins and glycolipids (4, 5), which are induced by the intestinal microbiota and specifically by Bacteroides (6), which in turn cleave l-fucose moieties from the host's surface and internalize them for use as an energy source (7). Here we show that Bacteroides convert exogenously acquired l-fucose to guanosine diphosphate (GDP)–l-fucose to incorporate it into multiple surface capsular polysaccharides and glycoproteins. This study suggests a basis for molecular mimicry by these indigenous microorganisms and demonstrates that the synthesis of fucosylated surface molecules gives these symbionts a competitive colonization advantage.

Bacteroides fragilis 9343 is covered with multiple capsular polysaccharides whose expression is regulated by DNA inversions (8, 9). Of the eight capsular polysaccharides (termed PSA to PSH) known to be synthesized by this organism, the structures of only two, PSA and PSB, have been elucidated (10). The repeating unit of PSB contains a terminal α1,2-linked l-fucose moiety, and its biosynthesis locus encodes an α1,2 fucoslytransferase (11), which incorporates l-fucose into the PSB repeating unit. We found that three other capsular polysaccharide biosynthesis loci of B. fragilis 9343 (PSC, PSD, and PSH) also contain fucosyltransferase homologs.

The donor for the incorporation of fucose into eukaryotic glycoproteins, glycolipids, and bacterial polysaccharides is GDP-l-fucose [recently reviewed in (12)]. In bacteria, GDP-l-fucose is formed from GDP-d-mannose by GDP-mannose dehydratase (Gmd) and fucose synthetase (Fcl) (Fig. 1A) (13). Mammalian organisms also convert l-fucose to GDP-l-fucose through an l-fucose-1-phosphate intermediate (Fig. 2A). In B. fragilis 9343, gmd and fcl reside approximately 4 kb upstream of the oppositely transcribed PSB locus (Fig. 1A). Polymerase chain reaction (PCR) analysis (Fig. 1B) of 50 B. fragilis strains demonstrated that these genes are always present in tandem upstream of the PSB locus (14). Analysis of the completed genome sequence of B. fragilis 9343 confirmed that these are the only gmd-fcl genes in the genome.

Fig. 1.

Gmd and Fcl are essential for the synthesis of four of the capsular polysaccharides of B. fragilis 9343 when grown in media lacking fucose. (A) Biosynthesis pathway leading to the formation of GDP-l-fucose from GDP-d-mannose. Gmd represents GDP-mannose dehydratase and Fcl represents fucose synthetase. (B) Open reading frame (ORF) map of the chromosomal region containing gmd-fcl. A product was amplified by PCR1 and PCR2 for all 50 B. fragilis strains analyzed. The region deleted in Δgmd-fcl is shown. (C) Western blot analysis demonstrating the eight capsular phenotypes from organisms grown in rich medium or in defined media with glucose, mannose, or fucose. α-, antiserum. (D) Western blot analysis demonstrating that the capsular phenotypes of the Δgmd-fcl mutant grown in minimal glucose medium are restored when gmd-fcl from B. fragilis (pMJC20) or E. coli (pMJC22) is supplied in trans.

Fig. 2.

Contribution of Fkp to capsule bioynthesis. (A) Salvage pathway for the formation of GDP-l-fucose from l-fucose. (B) Fkp protein showing the N-terminal portion that is similar to mammalian l-fucose-1-P guanylyltransferase (light shading) and the C-terminal portion that is similar to l-fucose kinases (dark shading). (C) ORF map showing the location of fkp in the B. fragilis 9343 chromosome. (D) Western blot analysis of each of the eight polysaccharides from the wild type, mutants, and complemented mutants grown in rich medium.

Deletion of gmd and fcl from B. fragilis was expected to render the organism unable to synthesize the four polysaccharides predicted to contain l-fucose. However, when Δgmd-fcl was grown in standard medium, all eight polysaccharides were synthesized (Fig. 1C, lanes 1 and 2). When this mutant was grown in minimal glucose or minimal mannose medium, the synthesis of PSB, PSC, PSD, and PSE was abrogated (Fig. 1C, lanes 3 to 6). The PSE-null and PSH-positive phenotypes were not predicted on the basis of genetics; however, because B. fragilis synthesizes a large number of polysaccharides, it is likely that various gene products are not exclusive to the synthesis of their respective polysaccharide. When gmd-fcl genes were restored in trans (plasmid pMJC20), Δgmd-fcl recovered full polysaccharide synthesis (Fig. 1D, lane 3). The Escherichia coli K-12 gmd-fcl genes, involved in colanic acid biosynthesis (13), also restored capsule expression to near wild-type levels in minimal glucose medium (Fig. 1D, lane 4).

By contrast, when Δgmd-fcl was grown in minimal medium supplemented with l-fucose, the polysaccharide phenotypes resembled the wild type (Fig. 1C, lanes 7 and 8), suggesting that the organism directly uses l-fucose from the medium for incorporation into its capsular polysaccharides, a previously undescribed phenomenon in the prokaryotic superkingdom. Surface expression of a distinct monosaccharide, sialic acid, after acquisition from host molecules has been described for other microbes (15, 16). The use of exogenous fucose in this manner would require a mammalian-like pathway containing l-fucose-1-P guanylyltransferase, which converts l-fucose-1-phosphate to GDP-l-fucose (Fig. 2A). We detected a protein in the B. fragilis proteome with N-terminal similarity to the human l-fucose-1-P guanylyltransferase (Fig. 2B). Further analysis of this B. fragilis protein revealed C-terminal similarity to mammalian l-fucokinases, which convert l-fucose to l-fucose-1-phosphate. These findings suggested that this protein catalyzes both steps in the conversion of l-fucose to GDP-l-fucose. The gene encoding this hybrid protein, which we designate Fkp, is in a conserved genetic region upstream of the PSE capsule locus (Fig. 2C).

We deleted fkp from both wild-type and Δgmd-fcl backgrounds. Mutational analysis revealed that Δfkp expressed all capsular polysaccharides (Fig. 2D, lane 3), whereas the Δgmd-fclΔfkp mutant was unable to synthesize PSB, PSC, PSD, or PSE, even when grown in rich medium (Fig. 2D, lane 4). The four polysaccharides synthesized by this mutant are more homogeneous in size in the higher molecular-weight range as compared to the wild type, which is possibly attributable to less competition for nucleotide-activated monosaccharide precursors. When the Δgmd-fclΔfkp mutant is complemented with either gmd-fcl or fkp, polysaccharide synthesis is restored (Fig. 2D, lanes 5 and 6). These results demonstrate the involvement of Fkp in capsular polysaccharide biosynthesis and strongly suggest that this enzyme converts l-fucose to GDP-l-fucose.

To confirm the activity of Fkp, a Histagged fusion protein was used in enzymatic assays (14). Tritiated l-fucose (Fig. 3B, lane 1) was fully converted to GDP-l-fucose by His-Fkp (lane 6) but not by a control lysate purified in the same way (lane 2). When adenosine triphosphate (ATP) was omitted, no conversion occurred, because ATP is required to convert l-fucose to l-fucose-1-phosphate (lane 3). When guanosine triphosphate (GTP) was omitted, the reaction produced only l-fucose-1-phosphate (lane 4), because GTP is necessary to convert this intermediate to GDP-l-fucose. High-performance anion-exchange chromatography analysis of the enzymatic reaction confirmed the identity of all reaction products (fig. S1).

Fig. 3.

Fkp is able to convert l-fucose into GDP-l-fucose via an l-fucose-1-phosphate intermediate. (A) Coomassie-stained SDS–polyacrylamide gel showing the His-tagged Fkp and vector control used for enzymatic assays. The arrow indicates His-Fkp. (B) Phosphoimager scan of thin-layer chromatographic analysis demonstrating that His-Fkp converts l-fucose to GDP-l-fucose. Lane 1, l-fucose standard; lane 2, pET16b vector control; lane 3, no ATP added; lane 4, no GTP added; lane 5, reaction terminated at 30 min; lane 6, reaction terminated at 2 hours. (C) Colorimetric analysis demonstrating the migration of the standards compared with the substrate, intermediate, and product.

When grown in rich medium supplemented with 3H-l-fucose, B. fragilis incorporates this radiolabled sugar into its capsular polysaccharides in an Fkp-dependent manner (Fig. 4A). Moreover, Fkp also enables the 3H-l-fucose to associate with numerous proteins (Fig. 4B), which are sensitive to digestion by proteinase K (Fig. 4C). Antiserum raised to wild-type organisms and adsorbed with Δgmd-fclΔfkp detected numerous noncapsular polysaccharide molecules synthesized by the wild type but not by Δgmd-fclΔfkp (Fig. 4D), which were destroyed by proteinase K treatment (Fig. 4E). Thepresenceof l-fucose in these proteinacious molecules was confirmed by their binding of the Aleuria aurantia lectin, which is specific to fucose in α-1,3 or α-1,6 linkages (Fig. 4F). Therefore, B. fragilis synthesizes multiple fucosylated capsular polysaccharides and numerous fucosylated glycoproteins and is able to use an external source of l-fucose in an Fkp-dependent manner for their synthesis. Because dietary forms of l-fucose are absorbed with other simple monosaccharides in the small intestine, the Bacteroides must acquire exogenous l-fucose from the host (7). This is the first report of glycoprotein synthesis by an intestinal Bacteroides species and, to our knowledge, of bacterial glycoproteins containing l-fucose.

Fig. 4.

Fkp enables B. fragilis to incorporate exogenous l-fucose into capsular polysaccharides and glycoproteins. (A) Phosphoimager scan of a 4 to 12% SDS–polyacrylamide gel of the capsular polysaccharide complex isolated from the wild type, mutants, and complemented mutants grown in medium containing 3H-l-fucose. (B) Phosphoimager scan of a 12% SDS–polyacrylamide gel of whole-cell lysates grown in medium containing 3H-l-fucose. (C) Phosphoimager scan of a 12% SDS–polyacrylamide gel of wild-type bacteria grown in medium containing 3H-l-fucose and either untreated or subject to digestion with proteinase K. (D) Immunoblot analysis of wild-type and Δgmd-fclΔfkp strains probed with antiserum to wild-type organisms adsorbed with Δgmd-fclΔfkp. (E) Immunoblot analysis of wild-type strain and proteinase K–treated wild-type strain probed with antiserum to wild-type organisms adsorbed with Δgmd-fclΔfkp. (F) Western blot analysis showing that the A. aurantia lectin recognizes glycoproteins of the wild type but not of Δgmd-fclΔfkp. (G) Phosphoimager scan of a 12% SDS–polyacrylamide gel of Bacteroides species grown in 3H-l-fucose–supplemented medium and lysed or treated with proteinase K before lysis.

B. thetaiotaomicron is the only other sequenced bacterial genome containing an fkp homolog. Southern blot analysis using the l-fucose-1-P guanylyltransferase portion of fkp as a probe revealed homologs in all 20 B. fragilis strains tested and in all Bacteroides species analyzed (fig. S2). Porphyromonas gingivalis and Tannerella forsythensis, both members of the Bacteroidales order that colonize the oral cavity, do not contain fkp, suggesting acquisition after the Bacteroidales diverged into separate genera but before Bacteroides speciation. When representative Bacteroides species were grown in medium containing 3H-l-fucose, they also incorporated l-fucose into multiple glycoproteins, whereas Enterococcus faecalis and E. coli, used as controls, did not (Fig. 4G and fig. S3).

Full-length homologs of fkp were also detected in the genomes of Arabidopsis thaliana and Oryza sativa. This may explain why the A. thaliana mur1 mutant (17), which is deficient in the synthesis of GMD2 (GDP-mannose 4,6-dehydratase), synthesizes fucosylated glycoconjugates when grown on media containing fucose (17, 18).

We performed colonization experiments to determine whether the synthesis of fucosylated molecules confers an advantage on B. fragilis in the mammalian intestine by inoculating Swiss Webster germ-free mice with either the wild-type strain or Δgmd-fclΔfkp. Two days after inoculation, mice colonized with either strain showed similar numbers of bacteria in their feces (1.7 × 1010 to 3.1 1010 bacteria per gram of feces, table S1) × and maintained this level for the duration of the experiment (6 weeks). Thus, Δgmd-fclΔfkp is as fit as the wild-type strain to colonize the mouse intestine in the absence of competition. Next, a competitive colonization assay was performed, in which a bacterial mixture containing 50% wild-type and 50% Δgmd-fclΔfkp was used for inoculation. PCR analysis of the bacterial colonies from feces collected 24 hours after inoculation demonstrated that only 5% of the fecal bacteria were Δgmd-fclΔfkp. Three days after inoculation, no mutant bacteria were detected in the feces of any mice. This experiment was repeated, with the percentage of Δgmd-fclΔfkp in the inoculum increased to 60%. Twenty-four hours after bacterial inoculation, 3.2% of the fecal bacteria from mice in cage 1 and 8.6% of the fecal bacteria from mice in cage 2 were Δgmd-fclΔfkp. By day 3, the percentages of Δgmd-fclΔfkp in the feces of mice were reduced to 0% and 1.2% in cages 1 and 2, respectively.

Therefore, the synthesis of fucosylated surface molecules by B. fragilis gives these organisms a survival advantage in the competitive mammalian intestinal ecosystem; an ecosystem in which the synthesis of fucosylated host surface molecules is induced by the Bacteroides themselves (6). Because l-fucose is involved in Bacteroides' signaling of the synthesis of host fucosylated glycan production (7), it will be interesting to determine whether Δgmd-fclΔfkp is able to induce this symbiotic signal.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5716/1778/DC1

Materials and Methods

Figs. S1 to S3

Table S1

References

References and Notes

View Abstract

Navigate This Article