|Home | About | Journals | Submit | Contact Us | Français|
The Bacteroides are a numerically dominant genus of the human intestinal microbiota. These organisms harbor a rare bacterial pathway for incorporation of exogenous fucose into capsular polysaccharides and glycoproteins. The infrequency of glycoprotein synthesis by bacteria prompted a more detailed analysis of this process. Here, we demonstrate that Bacteroides fragilis has a general O-glycosylation system. The proteins targeted for glycosylation include those predicted to be involved in protein folding, protein-protein interactions, peptide degradation, as well as surface lipoproteins. Protein glycosylation is central to the physiology of B. fragilis and is necessary for the organism to competitively colonize the mammalian intestine. We provide evidence that general O-glycosylation systems are conserved among intestinal Bacteroides species and likely contribute to the predominance of Bacteroides in the human intestine.
The human intestine is home to one of the densest microbial ecosystems in the world. In this intensely competitive environment, the best adapted microbial species colonize at high densities and persist for long periods (Ley et al., 2006; Whitman et al., 1998). Members of the Bacteroides genus are specifically adapted for survival in this ecosystem and collectively comprise one of the most abundant bacterial genera in the human colon, in some individuals accounting for 10-20% of the microbial population (Eckburg et al., 2005; Kurokawa et al., 2007).
Intestinal microorganisms provide many functions essential to the well-being of the host. The Bacteroides contribute to human metabolism, development and immunity. They provide energy to the host in the form of short-chain fatty acids (Hooper et al., 2002), and are involved in the recycling of bile acids (Midtvedt, 1974). During development, Bacteroides stimulate intestinal angiogenesis (Hooper et al., 2001) and induce local and systemic immune functions (Hooper et al., 2003; Mazmanian et al., 2005). The intestinal microbiota are also implicated in a number of major health issues, including inflammatory bowel diseases, asthma and other allergic disorders, colon cancer and the spread of antibiotic resistance (O’Keefe, 2008; Penders et al., 2007; Salyers et al., 2004; Strober et al., 2007).
To better understand how these organisms affect the host both beneficially and deleteriously, we need a more thorough analysis of the bacterial mechanisms that are instrumental for their survival in their ecosystem. Bacteroides spp. have several unique adaptations that contribute to their successful habitation of this niche. They can rapidly change their cell surface architecture through the production of an unusually large number of phase-variable capsular polysaccharides (Krinos et al., 2001), an adaptation that appears specific to the intestinal environment, as it is not found in closely related oral species (Coyne and Comstock, 2008). Bacteroides species also have an unprecedented repertoire of genetic systems devoted to acquiring and metabolizing carbohydrates — including mucosal glycans and plant polysaccharides that cannot be digested by the host — which allow a rapid response to food supplies that are shifting and often times scarce (Cerdeno-Tarraga et al., 2005; Xu et al., 2003; Xu et al., 2007). Also, at least one Bacteroides species has been shown to stimulate the expression of fucosylated glycoconjugates on the intestinal epithelia of colonized mice (Bry et al., 1996). Bacteroides species produce enzymes to harvest this fucose from host mucosal glycans and have a rare bacterial pathway for incorporation of this exogenous fucose directly into their capsular polysaccharides and glycoproteins (Coyne et al., 2005). Recent studies suggest that the ability of B. fragilis to synthesize fucosylated glycoproteins is essential for it to competitively colonize the mammalian intestine (Coyne et al., 2008; Coyne et al., 2005).
Eukaryotic organisms have general N- and O-glycosylation systems that modify as many as half of their cellular proteins, but protein glycosylation is rare in bacteria. In those bacteria that do glycosylate proteins, the glycosylated molecules are usually limited to one or two abundant polymeric surface proteins such as flagellins, pilins and S-layer proteins (Benz and Schmidt, 2002; Fletcher et al., 2007; Schaffer et al., 2001). In these specialized bacterial glycosylation systems, the glycans are O-linked to serine and theronine residues. The only well characterized general glycosylation systems of bacteria are those of Campylobacter jejuni and related species, where the glycans are N-linked to asparagine residues (Szymanski et al., 1999; Wacker et al., 2002), and a recently described O-glycosylation system of Neisseria gonorrhoeae (Vik et al., 2009).
Here, we investigated the production of fucosylated glycoproteins by B. fragilis. We demonstrate that this organism has a general O-glycosylation system that is central to the physiology of B. fragilis and its ability to colonize its ecological niche.
We previously showed that B. fragilis produces many glycoproteins that bind the fucose-specific Aleuria aurantia lectin (AAL) (Coyne et al., 2005), but no glycosylated proteins from this species had been molecularly identified. For this, we used AAL affinity chromatography to purify fucosylated molecules from a whole cell lysate and separated the AAL-purified fraction by 2D isoelectric focusing/PAGE (Fig 1A). Ninety-five spots were excised from the 2D gel and individually subjected to MS/MS analysis to identify proteins. We unequivocally identified at least one protein for 83 of the 95 spots, for a total of 48 different proteins (Fig 1A and Table S1).
Two of the most abundant proteins identified in this primary screen were BF2494 and BF3567. We cloned the genes encoding these proteins, expressed them in E. coli with His-tags, purified them, then used the purified proteins to immunize rabbits to obtain antisera specific to the protein components of these molecules. On western blots of B. fragilis whole cell lysates, each of these antisera recognized multiple bands with molecular sizes slightly higher than predicted by the polypeptide sequences, consistent withthese molecules being glycosylated (Fig 1B). The B. fragilis mutant Δgmd-fclΔfkp is defective in both pathways for biosynthesis of GDP-fucose and consequently cannot incorporate fucose into glycoproteins (Coyne et al., 2005). The BF2494 and BF3567 antisera each recognized proteins of lower molecular weight in whole cell lysates of this mutant compared to wild-type, consistent with less or no glycosylation of these proteins (Fig 1B).
Based on these observations, we developed a secondary screen to assay for glycosylation of candidate proteins. We modified the Bacteroides expression vector pFD340 (Smith et al., 1992) so that a C-terminal His-tag was added to the product of a cloned gene. This vector, pCMF6, allows us to express proteins in B. fragilis and detect them with an antibody to the His-tag or to purify them by nickel affinity chromatography. We selected 24 candidate glycoproteins that were abundant on the 2D gel and/or appeared in multiple spots, particularly diagonal “trains” that are typical of glycoforms. The genes of candidate proteins were cloned into pCMF6 and expressed in both wild-type and Δgmd-fclΔfkp. Whole cell lysates were analyzed by western blot using an antibody to the C-terminal His-tag. Of the 24 candidates, 5 proteins in addition to BF2494 and BF3567 were of smaller size in Δgmd-fclΔfkp than in wild-type, consistent with a glycosylation defect, 6 proteins showed no change, and 11 were not produced in sufficient quantity to be detected (Fig 1C and Table S1).
The seven proteins that had a reduced molecular size in Δgmd-fclΔfkp were subjected to additional analyses to confirm glycosylation. Each of these seven His-tagged molecules was purified from B. fragilis using nickel affinity chromatography and its glycosylation was assayed by periodate-Schiff staining and AAL reactivity. These tests were validated using His-tagged BF2494 (BF2494-His) purified from both B. fragilis (glycosylated) and E. coli (unglycosylated; Fig 1D). Glycosylation of six of these proteins was confirmed by these methods (Fig 1D and 1E). The seventh protein, BF0522, could not be purified in sufficient quantity for these analyses, but positive results in both the primary and secondary screens strongly suggest that this molecule is glycosylated. An eighth protein, BF0994, was not present in whole cell lysates in sufficient quantity to be detected in the secondary screen but its glycosylation was confirmed upon purification of the protein (Fig 1E). Thus, we identified eight glycoproteins produced by B. fragilis (Table 1).
Next, we sought to determine if the glycan component of each of these glycoproteins is similar. For this, we generated an antiserum specific to the glycan component of BF2494 by immunizing rabbits with glycosylated BF2494-His purified from B. fragilis, with subsequent removal of the antibodies to the protein component using unglycosylated BF2494-His from E. coli immobilized on a nickel column. The resulting adsorbed antiserum recognizes glycosylated BF2494-His purified from B. fragilis but not BF2494-His purified from E. coli (Fig 1D). Western blot analysis demonstrated that this antiserum recognized each of the other six purified glycoproteins, suggesting that their glycans are similar or identical to those of BF2494 (Fig 1E).
The eight identified glycoproteins represent a variety of biochemical activities (Table 1). BF0447 is similar to Skp of E. coli, a chaperone of outer membrane proteins, and BF0994 has a sequence motif identifying it as a proline cis-trans isomerase, an activity usually associated with chaperone function (Duguay and Silhavy, 2004). BF0935 and BF3918 have motifs characteristic of zinc endopeptidases, while BF2334 and BF2494 contain tetratricopeptide repeat (TPR) domains that mediate protein-protein interactions. Therefore, B. fragilis glycosylates proteins with a range of different functions, distinct from the polymeric surface molecules that are typical of bacterial glycoproteins.
The putative functions of the B. fragilis glycoproteins suggested that many of them are likely internal rather than located on the cell surface. All eight of the glycoproteins are predicted by LipoP (Juncker et al., 2003) to have signal peptides. Three of the glycoproteins (BF0522, BF3567 and BF3918) are predicted to have type II signal peptidase cleavage sites, indicating that they are lipoproteins. The absence of an aspartate residue at position +2 following the cleavage site indicates that these proteins are targeted to the outer rather than the inner membrane (Yamaguchi et al., 1988).
The remaining five glycoproteins (BF0447, BF0935, BF0994, BF2334 and BF2494) are predicted to have type I signal peptidase cleavage sites and lack membrane targeting regions suggesting that they are periplasmic. To confirm their localizations, we prepared cytoplasmic and periplasmic fractions from wild type B. fragilis for analysis of native BF2494, and from wild type expressing each of these five His-tagged glycoproteins. The purity of each fraction was determined by measuring the activity of cytoplasmic and periplasmic marker enzymes (Table S2). We analyzed the fractions by western blot with antibodies to the protein component of BF2494 or to the His-tag. The results show that native BF2494 is targeted to the periplasm (Fig 2A) as is BF2494-His (Fig 2B), validating the use of these engineered proteins for localization experiments. BF0447-His, BF0935-His and BF2334-His are also located in the periplasm (Fig 2B). BF0994-His was not detected by this method, perhaps due to low abundance or solubility. However, given its predicted function as a proline cis-trans isomerase, BF0994 is likely to be located in the periplasm (Duguay and Silhavy, 2004).
The orientations of the three glycosylated outer membrane lipoproteins (BF0522, BF3567 and BF3918) were next investigated. Outer membrane lipoproteins of E. coli are generally anchored in the inner leaflet of the membrane (Tokuda and Matsuyama, 2004), but in other Gram negative bacteria, lipoproteins are also present on the cell surface (Nsofor et al., 2006; Reilly et al., 1999). To elucidate their localizations, whole bacteria were treated with proteinase K to digest surface proteins. As controls, we used BF2494, which we have shown here to be periplasmic, and AapA, a cell surface protein (Weinacht et al., 2004). BF3567 and BF0522-His were completely digested by proteinase K treatment demonstrating that they are cell surface proteins (Fig 2C and 2D). BF3918-His was not digested, indicating that it is likely anchored to the inner leaflet (Fig 2D).
Thus, all eight of these glycoproteins are secreted from the cytoplasm and targeted either to the periplasm or the inner or outer leaflet of the outer membrane (Table 1). Glycosylation of periplasmic proteins is unusual: it has only been previously observed for glycosylated proteins of C. jejuni, Desulfovibrio gigas, and N. gonorrhoeae (Santos-Silva et al., 2007; Young et al., 2002, Vik et al., 2009).
BF2494 was selected as a prototype glycoprotein for more detailed investigation due to its moderate size and abundance. Glycans are most commonly attached to proteins by N-linkages to asparagine or O-linkages to serine (S) and threonine (T). To determine the nature of the protein-glycan linkage in BF2494, BF2494-His was purified from B. fragilis and subjected to three procedures to release glycans from the proteins: β-elimination, which releases glycans that are O-linked to S or T, and two different hydrazinolysis procedures optimized to release either O-linked glycans, or both N- and O-linked glycans. All three procedures released oligosaccharides, suggesting that this protein has glycans O-linked to S and T (data not shown).
Given this information, site-directed mutagenesis was used to identify the glycosylation sites of BF2494 and confirm the O-linkages to S and T. BF2494 contains 22 S and T residues. To identify residues likely to be glycosylated, we focused on tryptic peptides that were not frequently detected by MS/MS analysis of BF2494 due to a mass increase from glycosylation. Nine S and T residues were present in rarely observed peptides (Fig 3A). Each of these nine residues was individually mutated to alanine and the resulting proteins were purified from B. fragilis and analyzed by western blot to detect changes in molecular weight resulting from lack of glycosylation at the mutated site. Mutation of candidate residues T60, S123, S126, T188, T309 and S321 had no detectable effect (data not shown). However, mutation of T87, T178 or T231 to alanine resulted in loss of the largest form of the molecule (Fig 3B). Mutation of two of these three residues in any combination resulted in loss of the two largest forms, while mutation of all three residues resulted in a single form of lower molecular weight than any form observed for the wild-type glycoprotein (Fig 3B). Furthermore, all three double T→A mutants are still glycosylated, but the triple mutant is not (Fig 3C). Cell fractionation showed that the triple T→A mutant protein is present in the periplasm, so its lack of glycosylation is not due to incorrect localization (Fig 3E). These results demonstrate that T87, T178 and T231 are the only glycosylation sites of BF2494 and therefore all the glycans of this protein are O-linked.
Inspection of the protein sequence around the three glycosylation sites of BF2494 revealed that each has an aspartate (D) preceding the glycosylated T and is followed by an amino acid with one or more methyl groups (alanine, isoleucine or leucine; Fig 3F). None of the 17 unglycosylated S and T residues of BF2494 (excluding two in the signal peptide), have a preceding D, although seven are followed by A, I or L and one by V. Furthermore, the other seven identified glycoproteins all contain at least one site with this motif. Because O-glycosylation systems in other organisms do not require a motif other than S or T, we made site directed mutations of the putative glycosylation site around T231 of BF2494. The predicted three residue glycosylation site motif (D)(S/T)(A/I/L/V) was first tested by altering the first residue, D230, to either an A, or the most conservative substitution, glutamate (E). These alterations were not tolerated and resulted in lack of glycosylation at this site (Fig 3G). The second residue of the motif, T231, was changed to an S, which did not alter glycosylation and therefore could be used interchangeably as the glycosylated residue. The third residue of the motif was predicted to require a methyl group. Substitution of the A232 at the third position with either of the two remaining methyl containing residues, methionine (M) or T, were tolerated; however, substitution with the non-methyl containing residues D or glycine (G) abrogated glycosylation at this site (Fig 3G). Therefore, the glycoyslation motif of the O-glycosylation system of B. fragilis is (D)(S/T)(A/I/L/V/M/T). A similar O-glycosylation motif has been predicted in Elizabethkingia meningoseptica, which belongs to the same phylum as B. fragilis (Plummer Jr et al., 1995).
Most glycosylation systems modify proteins that have been transported out of the cytoplasm, into the bacterial periplasm or the ER and golgi of eukaryotic cells (Szymanski and Wren, 2005). To test if secretion into the periplasm is required for protein glycosylation in B. fragilis, we mutated BF2494-His so that all but the first and last residues of the signal peptide were removed (amino acids 2-18; Fig 3A) and expressed this mutant protein in B. fragilis. Fractionation confirmed that this signal peptide mutant is retained in the cytoplasm (Fig 3D). Moreover, the purified mutant protein is not glycosylated (Fig 3C) and has the same molecular size as the triple glycosylation site mutant described above (Fig 3B).
B. fragilis strain 9343 produces eight different capsular polysaccharides, each synthesized by the products of separate operons (Krinos et al., 2001). As in many bacteria, the repeat units of the capsular polysaccharides are assembled at the inner surface of the cytoplasmic membrane, transported across the membrane by a polysaccharide flippase (Wzx) and polymerized by a polysaccharide polymerase (Wzy). Each of the eight capsular polysaccharide operons encodes both a flippase and a polymerase. As secretion into the periplasm is necessary for protein glycosylation, a genomic region involved in protein glycosylation would likely encode a flippase, but would lack a polymerase. On inspecting the B. fragilis genome, we found a region spanning genes BF4298 to BF4306 that encodes a putative flippase, five putative glycosyltransferases and other genes likely to be involved in oligosaccharide synthesis, but no polymerase (Fig 4A and Table S3).
We deleted all nine of these genes and examined the effect of this deletion on protein glycosylation. We previously showed that wild-type B. fragilis incorporates 3H-fucose from its growth medium into glycoproteins and that Δgmd-fclΔfkp is deficient in this process (Coyne et al., 2005). Fig 4B shows that mutant Δ(BF4298-4306) also can not incorporate 3H-fucose into glycoproteins. As a further test, we compared the sizes of three His-tagged glycoproteins expressed in wild-type B. fragilis and Δ(BF4298-4306). All three glycoproteins were of lower molecular weight in the mutant, identical to their sizes in Δgmd-fclΔfkp, consistent with a glycosylation defect (Fig 4C and 4D). These results demonstrate that the BF4298-4306 region is involved in protein glycosylation. Furthermore, since this region is required for synthesis of all the fucosylated glycoproteins, it appears to be part of a general glycosylation system. Based on these results, we named the BF4298-4306 region lfg for locus of fragilis glycosylation.
Interestingly, the lfg region is adjacent to metG, which encodes the essential enzyme methionyl-tRNA synthetase (Fig 4A). metG and wzx (BF4298) are transcribed in the same direction and are separated by only 90 bp, with no recognizable B. fragilis promoter (Bayley et al., 2000) in the intervening region. We used RT-PCR to amplify a region including part of the coding regions of both metG and wzx and confirmed that these genes are co-transcribed (Fig 4E). The transcriptional linkage of the protein translation machinery (metG) with the protein glycosylation machinery (lfg) suggests a high level of importance for protein glycosylation in B. fragilis.
Two other important features of the lfg region distinguish it from the capsular polysaccharide biosynthesis loci. The capsular polysaccharide loci have a low G+C content (32.6-37.9%) compared to the G+C content of the genome (42.3%), indicating relatively recent acquisitions by horizontal transfer events. In contrast, the G+C content of the lfg region is 45.4%, similar to that of the genome and the adjacent housekeeping gene metG (47.0%). In addition, analysis of the two other sequenced B. fragilis genomes reveals that the lfg region is highly conserved (the 9 encoded proteins are 98.8-100% identical between strains), whereas the capsular polysaccharide loci are heterogeneous (Comstock et al., 2000; Coyne et al., 2001). Therefore, the lfg region is likely evolutionarily ancient and essential for survival of the organism in its niche.
Our analysis of five other intestinal Bacteroides species (B. caccae, B. ovatus, B. thetaiotaomicron, B. uniformis, B. vulgatus) demonstrated that they all have regions similar to lfg, with a flippase gene downstream of metG and several glycosyltransferase genes (Fig 5A). This finding, coupled with our previous observation that other Bacteroides species produce fucosylated glycoproteins (Coyne et al., 2005), prompted us to investigate if the general O-glycosylation system of B. fragilis is conserved in other intestinal Bacteroides species.
We used the glycan-specific antiserum described above in western blots of whole cell lysates of all five species. Figure 5B shows that this antiserum recognizes many molecules in all the Bacteroides species analyzed, indicating that these species modify their proteins with a glycan similar to that of B. fragilis (Fig 5B). Furthermore, this antiserum recognizes a similar pattern of molecules in these species, suggesting that orthologous proteins of each species are glycosylated (Fig 5B).
There are conserved orthologs of the B. fragilis glycoprotein BF2494 in four of the five other intestinal Bacteroides species analyzed (70-75% identity, 82-88% similarity), with the B. vulgatus BF2494 ortholog demonstrating only 39% identity and 56% similarity. When we probed whole cell lysates of these five other Bacteroides species with antiserum to the protein component of BF2494, orthologs of the expected molecular weight were detected in all species except B. vulgatus. In each species, the antiserum recognized molecules of multiple sizes, suggesting that the BF2494 orthologs have various glycoforms (Fig 5C).
Alignment of the sequence of BF2494 with its orthologs shows that the three glycosylation sites of the B. fragilis molecule align with sequences from the other five orthologs that match perfectly with the glycosylation motif (D)(S/T)(A/I/L/V/M/T) (Fig 5D). This finding suggests that the same three sites are glycosylated in these BF2494 orthologs. Therefore, B. fragilis BF2494 should be glycosylated at these same three sites if transferred into these species. To test this prediction, we expressed B. fragilis BF2494-His and the BF2494-His triple glycosylation site mutant T87A.T178A.T231A in four other Bacteroides species. When whole cell lysates from these bacteria were probed with antibody to the His-tag, the B. fragilis 2494-His protein displayed multiple bands in each of these species, indicative of multiple glycoforms, whereas the 2494-His triple glycosylation site mutant was of a single, lower molecular size and was much less abundant, consistent with our observations in B. fragilis (Fig 5E). Together with the preceding results, these data suggest that these intestinal Bacteroides species have similar general O-glycosylation systems.
As our results suggested that protein glycosylation is an important process in B. fragilis, we carried out experiments to determine whether the attenuation of protein glycosylation in Δlfg and Δgmd-fclΔfkp affects growth of the bacterium. We found significant in vitro growth defects in both of these mutants. Most dramatically, both mutants had fewer viable cells at all stages of growth compared to the wild-type (Fig S1). This result is consistent with a central role for protein glycosylation in B. fragilis physiology.
We tested the importance of protein glycosylation for intestinal colonization using Δlfg in a gnotobiotic mouse model. When mice were monoassociated with Δlfg, it achieved a density of 2-4 × 1010 CFU per gram of fecal material from the first measurement (3 days after inoculation) to the end of the experiment (12 days). This is the same level we have previously observed for wild-type B. fragilis (Coyne et al., 2005) and indicates that the mutant is equally able to colonize the mouse intestine in the absence of competition. However, Δlfg is not able to compete with wild-type B. fragilis for intestinal colonization. In three separate experiments, even with a 4:1 numerical advantage in the initial inocula, the mutant strain was reduced to a small minority of the bacteria in fecal samples 3 days after inoculation, and was almost undetectable after 7 days (Table 2). These results show that protein glycosylation is required for B. fragilis to colonize its ecological niche.
In this paper, we investigated the production of fucosylated glycoproteins by B. fragilis. We identified eight glycoproteins with similar or identical glycans, which are located in the periplasm or outer membrane and have a variety of predicted functions. Based on the number of glycoprotein candidates identified in our primary screen (48) and the success rate of the secondary screen and confirmation procedure (8 out of 14 proteins detected were proven to be glycosylated), we expect that at least 20 of these 48 proteins are glycosylated. We showed that a prototype glycoprotein is O-glycosylated and confirmed that glycosylation requires a three residue motif. Furthermore, deletion of a single genomic region affected all of the fucosylated glycoproteins. Taken together, these results indicate that B. fragilis has a general O-glycosylation system, one of the few described general glycosylation sustems in bacteria. The well characterized general N-glycosylation system of C. jejuni appears to be conserved in closely related species (Szymanski and Wren, 2005), much as the general O-glycosylation system of B. fragilis is conserved in other intestinal Bacteroides species.
The predicted functions of the glycoproteins BF0447 and BF0994 suggest that they are involved in a chaperone system, possibly serving as chaperones of outer membrane proteins like that described for E. coli (Duguay and Silhavy, 2004). Other glycoproteins may be involved in the response to oxidative stress suggested by a recent study in which BF0522 and BF2334 were upregulated 6.1 and 3.9 fold respectively on exposure of B. fragilis to air (Sund et al., 2008). BF4280, which was detected in our primary screen, has been implicated in binding to the host protein plasminogen (Sijbrandi et al., 2008). Therefore, glycoproteins may also contribute to the pathogenesis of B. fragilis in extra-intestinal infections.
The contributions of glycans to protein function can be placed in two broad categories: providing labels for binding and recognition by other proteins (Haltiwanger and Lowe, 2004; Helenius and Aebi, 2004), and stabilizing proteins, including thermodynamic stability, solubility, and protection from proteases (Langsford et al., 1987; Meldgaard and Svendsen, 1994; Nakatsukasa et al., 2004). Protein glycosylation in B. fragilis may provide a stabilizing function, as we do not observe unglycosylated forms in wild-type bacteria (Figs 1D, 1E and and2B),2B), and the triple T→A unglycosylated mutant of BF2494 is expressed at an extremely reduced level (Fig 5D). It was recently reported that periplasmic proteins of B. fragilis do not form intramolecular disulfide bonds (Dutton et al., 2008). Therefore, protein glycosylation may provide an alternative means of protein stabilization.
The mechanism of O-glycosylation in B. fragilis appears to have some similarities to protein glycosylation in other bacteria, which itself is similar to eukaryotic N-glycosylation (Szymanski and Wren, 2005). The likely common features are assembly of the glycan on a lipid carrier by cytoplasmic glycosyltransferases using nucleotide-activated sugar substrates, flipping of the glycan into the periplasm (the ER in eukaryotes), with subsequent block transfer of the oligosaccharide to the protein. B. fragilis O-glycosylation therefore appears to be similar to pilin O-glycosylation in bacteria (Faridmoayer et al., 2007) and to a recently described S-layer O-glycosylation system of Geobacillus stearothermophilus (Steiner et al., 2008), but different than O-glycosylation in eukaryotes, which generally involves sequential transfer of monosaccharides to proteins. A proposed model for the general O-glycosylation pathway of B. fragilis is shown in Figure S2.
Several lines of evidence suggest that protein glycosylation is central to B. fragilis physiology; the substantial number of proteins that are glycosylated, their various locations both inside and outside the cell, their variety of biochemical functions and their possible roles in fundamental processes such as protein folding. Secondly, we do not observe unglycosylated forms of the glycoproteins, suggesting that glycosylation is essential for their function. In addition, the transcriptional linkage of the lfg region to metG, a gene required for translation of essentially all proteins made by the organism, suggests that the protein glycosylation system is a vital component of the protein synthesis machinery. Lastly, deletion of the lfg region results in a substantial growth deficiency in vitro and a complete inability to compete with wild-type bacteria in the mouse intestine. The conservation of the glycosylation system in all of the intestinal Bacteroides species examined suggests that it contributes to the predominance of these organisms in the human intestine.
All oligonucleotides used in this study are listed in Table S4.
500 μg protein eluted from an AAL affinity column was resuspended in 7 M urea, 2 M thiourea, 50 mM DTT, 0.5% Pharmalyte 3-11, 0.4% CHAPS and focused on 24 cm pH 3-11 non-linear IPG strips (GE Healthcare). These were reduced, alkylated and loaded onto 25 × 25 cm 12-25%T polyacrylamide gels, which were run at 10°C, fixed and stained with Coomassie Blue. Spots were visualized with an Odyssey 700 nm laser (Li-Cor) and images analysed using Delta2D (Decodon, Germany). Background was defined as any non-Gaussian pixel array and subtracted from the image. Spots were excised for identification by MS as described in supplemental procedures.
The coding regions of BF2494 and BF3567 were amplified from chromosomal DNA and ligated into the NcoI site of pET16b to produce proteins with C-terminal His-tags. The resulting plasmids were transformed into E. coli BL21(DE3). His-tagged proteins were purified using ProBond resin (Invitrogen) under denaturing conditions.
pFD340 (Smith et al., 1992) was modified to produce proteins from cloned genes with C-terminal His-tags. A double stranded insert encoding 10 histidine residues with a stop codon and SstI/BamHI ends was created by annealing two oligonucleotides and cloned into SstI/BamHI-digested pFD340 creating pCMF6. The coding regions and ribosome binding sites of candidate glycoproteins were amplified from the B. fragilis chromosome, digested with BamHI or BglII and ligated into the BamHI site of pCMF6. The recombinant proteins are modified by addition of the amino acids GSH10 at the C-terminus.
Cultures (2 L) of B. fragilis harboring plasmids encoding His-tagged proteins were grown to stationary phase and harvested. His-tagged proteins were purified using 0.5 ml Ni-NTA agarose resin (Invitrogen). Cells were resuspended in buffers containing an EDTA-free Complete Protease Inhibitor Tablet (Roche) and lysed by sonication. Membrane-associated proteins were purified under denaturing conditions, washed with 16 ml buffer at each of pH 8.0, 6.0 and 5.3 and eluted at pH 4.0. Soluble proteins were purified under native conditions with 16 ml washes containing 0 and 20 mM imidazole, a final wash with 16 ml buffer containing 1 mM EDTA, and eluted with 250 mM imidazole.
To generate antisera specific to BF2494 and BF3567, His-tagged molecules were purified from E. coli and used to immunize rabbits using the EXPRESS-LINE protocol of Lampire Biological Laboratories. Antiserum was generated in the same way to glycosylated BF2494-His purified from B. fragilis. Antibodies to the protein component of BF2494 were adsorbed from this antiserum by incubation with BF2494-His purified from E. coli immobilized on ProBond resin. This adsorption was performed twice and the flow-through antiserum was confirmed to be devoid of antibodies to 2494-His of E. coli. A monoclonal antibody to C-terminal His-tag was purchased from Invitrogen.
Fractionation was achieved as previously described (Salyers and Kotarski, 1980) with a 4 minute incubation for spheroplast formation. The amount of cytoplasmic and periplasmic material in each fraction was determined by assaying the activity of the cytoplasmic enzyme phosphoglucose isomerase (Noltzmann, 1966) and the periplasmic enzyme alkaline phosphatase (Garen and Levinthal, 1960).
Proteolysis was carried out as previously described (Shipman et al., 1999) except the buffer was PBS containing 10 mM MgCl2. To determine the orientation of BF3567, we used B. fragilis Δtsr15M8, in which the invertible promoter of the aap operon is locked in the ON position, giving increased expression of AapA (Weinacht et al., 2004). As our proteinase K treatment was sufficient to degrade AapA (external control) but did not affect BF2494 (internal control), we used the same treatment to determine the orientations of BF0522-His and BF3918-His expressed in wild-type B. fragilis.
Site-directed mutagenesis of the BF2494 coding sequence in plasmid pCMF6 was carried out with the QuikChange XL kit (Stratagene) with an elongation time of 12 minutes. All mutations were confirmed by sequencing.
A total of 10,040 bp of DNA were deleted, including all but the first 23 bp of the coding region of BF4298 and the last 33 bp of BF4306. DNA segments flanking the region to be deleted were amplified from the chromosome, digested with BamHI and MluI and ligated together into the BamHI site of the suicide vector pJST55 (Thompson and Malamy, 1990). After conjugal transfer to B. fragilis, co-integrates resulting from homologous recombination were selected for erythromycin (Em) resistance, passaged, plated on non-selective medium and replica plated onto medium containing Em. EmS colonies were screened by PCR to identify mutants.
B. fragilis was grown to OD600 0.8, harvested, and total RNA was extracted with the RNeasy Mini Kit using RNAprotect with on-column digestion using the RNase-Free DNase Set (all products from Qiagen) and further DNase treatment was performed with the DNA-free kit (Ambion). RT-PCR was performed using 100 ng of RNA with the SuperScript III One-Step RT-PCR System with Platinum Taq DNA Polymerase (Invitrogen). A separate set of reactions was performed that included Taq but no RT. The primers encompassed a region of 1142 bp including the last 928 bp of the coding region of metG and the first 121 bp of BF4298.
Mouse studies were approved by the Harvard Medical Area standing committee on animals. Male Swiss Webster germ-free mice (3-5 weeks old) were purchased from Taconic. The mice were housed in gnotobiotic isolators and fed an autoclaved rodent chow diet (Zeigler Bros). Mice were innoculated with bacteria by spreading log-phase cultures on their fur. For monoassociation assays, we used one cage containing two mice. Fresh fecal samples were collected at various intervals, diluted in PBS and plated for colony counts. For competitive colonization assays, we used three cages of two mice each. Wild-type and Δlfg strains from fecal samples and the innoculum culture were identified by performing colony PCR with a mixture of primers: resulting in a 770 bp product from wild-type and a 460 bp product from the mutant. We screened 96 colonies from each cage at each time point.
We thank P. Azadi and S. Park of the Complex Carbohydrate Research Center (CCRC) for glycan release and Z. Waldon of Children’s Hospital Boston for MS. This work was supported in part by NCRR grant RR018502 to the CCRC and mostly by NIH/NIAID grant AI067711.