Functional analysis of the Campylobacter jejuni N-linked protein glycosylation pathway

Authors

  • Dennis Linton,

    1. Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK.
    Search for more papers by this author
    • Present address: School of Biological Sciences, University of Manchester, Manchester M13 9PT, UK.

  • Nick Dorrell,

    1. Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK.
    Search for more papers by this author
  • Paul G. Hitchen,

    1. Department of Biological Sciences, Imperial College, London SW7 2AY, UK.
    Search for more papers by this author
  • Saba Amber,

    1. Institute of Microbiology, Department of Biology, Swiss Federal Institute of Technology, Zürich, CH-8092 Zürich, Switzerland.
    Search for more papers by this author
  • Andrey V. Karlyshev,

    1. Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK.
    Search for more papers by this author
  • Howard R. Morris,

    1. Department of Biological Sciences, Imperial College, London SW7 2AY, UK.
    2. M-SCAN Mass Spectrometry Research and Training Centre, Silwood Park, Ascot SL5 7PZ, UK.
    Search for more papers by this author
  • Anne Dell,

    1. Department of Biological Sciences, Imperial College, London SW7 2AY, UK.
    Search for more papers by this author
  • Miguel A. Valvano,

    1. Department of Microbiology and Immunology, University of Western Ontario, London, Ontario N6A 5C1, Canada.
    Search for more papers by this author
  • Markus Aebi,

    1. Institute of Microbiology, Department of Biology, Swiss Federal Institute of Technology, Zürich, CH-8092 Zürich, Switzerland.
    Search for more papers by this author
  • Brendan W. Wren

    Corresponding author
    1. Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK.
      E-mail brendan.wren@lshtm.ac.uk; Tel. +44 (0)207 927 2288; Fax +44 (0)207 637 4314.
    Search for more papers by this author

E-mail brendan.wren@lshtm.ac.uk; Tel. +44 (0)207 927 2288; Fax +44 (0)207 637 4314.

Summary

We describe in this report the characterization of the recently discovered N-linked glycosylation locus of the human bacterial pathogen Campylobacter jejuni, the first such system found in a species from the domain Bacteria. We exploited the ability of this locus to function in Escherichia coli to demonstrate through mutational and structural analyses that variant glycan structures can be transferred onto protein indicating the relaxed specificity of the putative oligosaccharyltransferase PglB. Structural data derived from these variant glycans allowed us to infer the role of five individual glycosyltransferases in the biosynthesis of the N-linked heptasaccharide. Furthermore, we show that C. jejuni- and E. coli-derived pathways can interact in the biosynthesis of N-linked glycoproteins. In particular, the E. coli encoded WecA protein, a UDP-GlcNAc: undecaprenylphosphate GlcNAc-1-phosphate transferase involved in glycolipid biosynthesis, provides for an alternative N-linked heptasaccharide biosynthetic pathway bypassing the requirement for the C. jejuni-derived glycosyltransferase PglC. This is the first experimental evidence that biosynthesis of the N-linked glycan occurs on a lipid-linked precursor prior to transfer onto protein. These findings provide a framework for understanding the process of N-linked protein glycosylation in Bacteria and for devising strategies to exploit this system for glycoengineering.

Introduction

Recently an N-linked protein glycosylation system was identified in the human enteropathogen Campylobacter jejuni (Szymanski et al., 1999; Wacker et al., 2002; Young et al., 2002), the first example of this modification in a species from the domain Bacteria. Glycosylation results in transfer of a heptasaccharide glycan with the structure GalNAc-α1,4-GalNAc-α1,4-[Glcβ1,3-]GalNAc-α1,4-GalNAc-α1,4-GalNAc-α1,3-Bac-β1 where Bac is bacillosamine or 2,4-diacetimido 2,4,6 trideoxyglucopyranose, to asparagine residues within the eukaryotic-like consensus sequon Asn-Xaa-Ser/Thr (Young et al., 2002; Larsen et al., 2004; Nita-Lazar et al., 2005). All identified glycosylated proteins are thought to have a periplasmic or surface location (Linton et al., 2002; Young et al., 2002). C. jejuni mutants with disrupted glycosylation have been constructed and shown to have reduced capacity for attachment/invasion of human epithelial tissue-derived cell lines (Szymanski et al., 2002; Karlyshev et al., 2004), impaired colonization of both mice and chicken intestinal tracts (Szymanski et al., 2002; Karlyshev et al., 2004), and reduced competence for DNA uptake (Larsen et al., 2004).

A C. jejuni genetic locus consisting of 12 putative genes (Fig. 1) can be transferred into E. coli so that this organism is able to produce N-linked glycoproteins (Wacker et al., 2002). One putative gene from this locus, termed  pglB,  encodes  a  protein  with  significant  levels of similarity to the STT3 protein, a component of the eukaryotic oligosaccharyltransferase (OST) complex (Dempski and Imperiali, 2002; Kelleher et al., 2003). Both knockout mutation of pglB and site-directed mutagenesis of a highly conserved region of the protein abrogate N-glycosylation, indicating its essential role in this process (Szymanski et al., 1999; Wacker et al., 2002; Young et al., 2002). N-linked glycosylation in eukaryotes is initiated by sequential addition of monosaccharides to the lipid carrier dolichylpyrophosphate (Dol-PP) on the cytoplasmic face of the endoplasmic reticulum (ER) (Burda and Aebi, 1999). The lipid-linked intermediate Man5GlcNAc2-PP-Dol is then translocated across the ER membrane into the lumen (Helenius et al., 2002). After further addition of monosaccharides to generate the glycolipid Glc3Man9GlcNAc2-PP-Dol, the glycan moiety is transferred from lipid carrier to nascent secreted polypeptides by the OST of which STT3 is a key component.

Figure 1.

Schematic representation of the C. jejuni locus that produces N-linked glycoproteins in E. coli. Predicted open reading frames are indicated by horizontal arrows with their respective gene designations indicated above the arrow. Knockout mutants were constructed in all putative genes represented by coloured arrows with colours coded according to the predicted function of the encoded putative protein: glycosyltransferases, green; enzymes involved in sugar biosynthesis, blue; oligosaccharyltransferase, red; predicted ABC transporter, yellow. The approximate location of insertion of the mutagenic cassette is indicated by the vertical black arrows. The exact location is indicated by the two numbers below the vertical arrow so that the lower of these numbers indicates the length in base pairs of the corresponding gene and the upper number indicates the distance in base pairs of the insertion site from the 5′ end of the gene. The waaC gene consists of 264 base pairs of the 1029 base pair gene, and the pglG gene consists of 708 base pairs of the 894 base pair gene. Sequence data were assembled from the data in accession numbers Y11648 and AJ131360.

The current model of N-linked glycosylation in C. jejuni (Szymanski et al., 2003; Wacker et al., 2002) is based on (i) the defining of a C. jejuni glycosylation locus through transfer of a gene cluster into E. coli cells that conferred the ability to glycosylate proteins with a C. jejuni-like glycan structure; (ii) the similarity of putative gene products from the glycosylation locus to functionally characterized proteins and (iii) by analogy to the eukaryotic N-linked glycosylation process inferred from the presence of a STT3 homologue, PglB, in the glycosylation locus. Thus, it has been proposed that the Campylobacter N-linked glycan is assembled on a lipid carrier at the cytoplasmic face of the inner membrane through the action of glycosyltransferases encoded by putative genes of the glycosylation locus, flipped across the inner membrane and into the periplasm and finally transferred through the action of the PglB protein to (glyco)proteins in the periplasm. However, this model lacks supporting experimental evidence. Herein we report mutational and structural analyses providing experimental evidence for the role in the biosynthesis of the C. jejuni N-linked glycan of gene products encoded by the glycosylation locus, including the specific individual function of five glycosyltransferases. We also demonstrate that the C. jejuni-derived glycosylation apparatus interacts with E. coli-encoded proteins involved in biosynthesis of a lipid-linked glycan, thus providing the first experimental evidence for involvement of a lipid-linked intermediate in biosynthesis of N-linked glycoproteins.

Results

Production of C. jejuni glycoprotein PEB3 in pgl gene mutant backgrounds

To understand the function of the individual proteins involved in the N-linked glycosylation system of C. jejuni, we exploited the ability of this system to function in E. coli. Thus, plasmid pACYCpgl, encoding the C. jejuni 81116 glycosylation locus (Fry et al., 1998), can direct glycosylation of a C. jejuni-derived glycoprotein in E. coli (Wacker et al., 2002). The principal advantage of this approach is that by carrying out our analysis in E. coli rather than C. jejuni we were able to readily engineer a C-terminal polyhistidine tag onto the glycoprotein so that it can be easily purified, irrespective of glycosylation status, for downstream structural analysis. As part of this strategy we constructed plasmid pPeb3his that produces a polyhistidine-tagged version of the C. jejuni glycoprotein PEB3 suitable for purification via nickel affinity chromatography. E. coli BL21-AI cells containing pACYCpgl and pPeb3his produced PEB3 that could be detected with lectin SBA, which binds to the terminal α-linked GalNAc residue of the heptasaccharide glycan (Linton et al., 2002), indicating a fully glycosylated protein (Fig. 2, lane 2), in contrast to cells containing only pPeb3his (Fig. 2, lane 1).

Figure 2.

Effect of mutations of genes from the C. jejuni glycosylation locus on glycosylation of the PEB3 reporter glycoprotein produced in E. coli assessed by lectin SBA binding. Partially purified His-tagged PEB3 produced in various E. coli genetic backgrounds was analysed by SDS-PAGE followed by blotting and probing with lectin SBA. The host strains contained plasmids as indicated.

To investigate the role of individual genes from the C. jejuni glycosylation locus we constructed insertional knockout mutants in 11 genes from the C. jejuni 81116 glycosylation locus on pACYCpgl and transformed this panel of plasmids into E. coli BL21-AI cells containing pPeb3his. All strains produced His-tagged PEB3 protein as detected by both an anti-PEB3 antibody and a monoclonal antibody against the polyhistidine tag (data not shown). However, PEB3 produced in a strain containing a glycosylation locus with an insertion in the pglB gene (pACYCpglB::kan) was no longer detectable with lectin SBA (Fig. 2, lane 4). This is consistent with previous data indicating an essential role for the putative oligosaccharyltransferase PglB in the glycosylation process (Wacker et al., 2002; Young et al., 2002). There was only very slight presumably non-specific SBA reactivity of the PEB3 protein from strains containing pACYCpgl derivatives with insertional mutations in pglA, pglJ and galE genes (Fig. 2, lanes 3, 11 and 12 respectively) and the SBA-reactive band was of increased mobility. These findings suggest that mutation of pglA, pglJ and galE genes had a significant effect on PEB3 glycosylation. In contrast, knockout mutants in pglC, pglD, pglE, pglF, pglI and wlaB genes resulted in production of SBA-reactive protein of similar mobility to that produced from cells containing the wild-type locus (Fig. 2, lanes 5, 6, 7, 8, 10 and 13 respectively). The lower Mr weakly SBA-reactive band in the lane corresponding to the pglE mutant is, we believe, due to non-specific interaction of SBA with non-glycosylated Peb3. With a knockout mutation of the pglH gene strongly SBA-reactive PEB3 was produced but of increased mobility when compared with SBA-reactive PEB3 from cells containing the wild-type locus (Fig. 2, lane 9 compared with lane 2), indicating presence of a smaller glycan that retains a terminal GalNAc residue. A second well characterized C. jejuni glycoprotein, AcrA, when expressed in E. coli in the presence of the pgl locus and mutants described above and probed with a glycosylation-specific antiserum, gave essentially identical results (data not shown).

Structural characterization of PEB3 glycan produced in pgl mutant backgrounds

We employed mass spectrometry to precisely define the oligosaccharide structures on the PEB3 proteins produced in E. coli, as described above. As previously shown for the C. jejuni glycoprotein AcrA (Wacker et al., 2002), PEB3 produced in E. coli containing pACYCpgl was glycosylated with a heptasaccharide that gave a fragmentation pattern (Fig. 3A) similar to that derived from PEB3 purified from C. jejuni (Wacker et al., 2002). Furthermore, a second fragmentation pattern with all major fragment ions shifted down by 25 Da was also produced from the E. coli-derived PEB3 glycopeptide (data not shown). As we have previously suggested (Wacker et al., 2002), this can be accounted for by the presence of an N-Acetylhexosamine (HexNAc) residue replacing a bacillosamine residue as the sugar linked to the peptide. Data presented below provide an explanation for the origins of these two structural variants.

Figure 3.

Tandem mass spectrometry (MS–MS) of glycosylated peptides derived from tryptic cleavage of PEB3 protein produced in E. coli strains containing mutant versions of the C. jejuni glycosylation locus. Spectra produced from fragmentation of the tryptic peptide DFNVSK containing the N-glycosylation site of the glycoprotein PEB3 are presented along with the inferred structures of the corresponding attached glycans. Spectrum A is derived from PEB3 produced in an E. coli strain containing plasmid PACYCpgl, spectrum B from a strain containing plasmid pACYCpglD::kan and spectrum C from a strain containing plasmid pACYCwlaB::kan. Spectrum D is derived from PEB3 produced in a strain lacking the glycosylation locus and similar spectra were produced from strains containing plasmids pACYCpglB::kan (spectrum E) and pACYCgalE::kan (spectrum F). The spectra labelled G, H, I, J, K, L and M are derived from PEB3 produced in strains containing the plasmids pACYCpglI::kan, pACYCpglA::kan, pACYCpglJ::kan, pACYCpglH::kan, pACYCpglC::kan, pACYCpglE::kan and pACYCpglF::kan respectively.

In two cases where genes from the glycosylation locus were inactivated, the resultant tandem mass spectrometry (MS–MS) fragmentation patterns (Fig. 3B and C) indicated the presence of a heptasaccharide structurally identical to that produced with a wild-type glycosylation locus. These were the wlaB and the pglD genes that encode a putative inner membrane-located ABC transporter and a putative acetyltransferase respectively (Parkhill et al., 1999). These data could be explained by functional complementation mediated by E. coli-derived proteins. In contrast, inactivation of pglB and galE genes, encoding the putative oligosaccharyltransferase and a sugar epimerase, respectively (Parkhill et al., 1999), resulted in a peptide fragmentation pattern indicating a complete lack of glycan (Fig. 3E and F), as obtained for PEB3 derived from E. coli cells lacking the glycosylation locus (Fig. 3D). Thus, pglB and galE gene products are essential for glycosylation. Inactivation of the remaining genes from the glycosylation locus resulted in production of PEB3 glycopeptide MS–MS fragmentation patterns, indicating attachment of structurally variant glycans, as described below.

Inactivation of pglI resulted in production of a PEB3-derived glycopeptide with an MS–MS fragmentation pattern (Fig. 3G) consistent with a hexasaccharide structure that when compared with the heptasaccharide produced in the background of a wild-type glycosylation locus, lacks a single hexose residue. These data are in complete agreement with similar analysis of a pglI knockout mutant of C. jejuni (data not shown). Given that the structure of the N-linked heptasaccharide contains only one hexose, a β1,-3 linked glucose residue, we propose that pglI encodes a β1,-3 glucosyltransferase. There are four other genes from the glycosylation locus (pglA, pglJ, pglH and pglC) encoding proteins with significant levels of sequence similarity to glycosyltransferases (Fig. 1). Inactivation of pglA resulted in production of a glycopeptide with an MS–MS fragmentation pattern consistent with a glycan consisting of a single residue of bacillosamine attached to the peptide (Fig. 3H). This strongly suggests that pglA encodes the transferase responsible for addition of the subsequent sugar residue an α1,-3 linked GalNAc residue (Young et al., 2002). Thus, we propose that the pglA gene encodes an α1,3 N-acetylgalactosaminyltransferase. In a similar manner, inactivation of pglJ results in production of a PEB3-derived glycopeptide with an MS–MS fragmentation pattern consistent with an attached disaccharide (GalNAc-α1,3-Bac) (Fig. 3I). We therefore infer that the pglJ gene product is an α1,4 N-acetylgalactosaminyltransferase responsible for the addition of the subsequent GalNAc residue. Inactivation of pglH results in production of a PEB3-derived glycopeptide with an MS–MS fragmentation pattern consistent with a GalNAc-α1,4-GalNAc-α1,3-Bac glycan (Fig. 3J). Likewise, we infer that the pglH gene product is also an α1,4 N-acetylgalactosaminyltransferase that is responsible for addition of the fourth residue of the heptasaccharide. However, the glycan derived from a strain with inactivated pglH lacks two further α1,4-linked GalNAc residues. The transferase(s) responsible for the addition of these two residues remain unidentified.

The final putative transferase of the pgl locus is encoded by pglC. In the presence of pACYCpglC::kan the PEB3-derived glycopeptide generated an MS–MS fragmentation pattern consistent with a heptasaccharide structure  containing  HexNAc  rather  than  bacillosamine as the first sugar residue (Fig. 3K). Similar PEB3 glycopeptide-derived MS–MS fragmentation patterns were observed with strains containing pACYCpglE::kan and pACYCpglF::kan (Fig. 3L and M respectively). These data implicate all three genes (pglC, pglE and pglF) in biosynthesis/transfer of the first sugar residue bacillosamine. It has been proposed that the pglC-encoded putative transferase is responsible for transfer of the first sugar residue onto a lipid carrier (Fry et al., 1998; Wacker et al., 2002; Young et al., 2002). To account for the presence of a heptasaccharide in the pglC mutant strain we hypothesized that the function of the pglC gene product was complemented by an E. coli derived enzyme able to transfer a HexNAc sugar residue onto a lipid carrier. The principal candidate was the wecA gene product, a UDP-GlcNAc: undecaprenylphosphate GlcNAc-1-phosphate transferase involved in biosynthesis of enterobacterial common antigen and many E. coli O antigens (Meier-Dieter et al., 1992; Klena and Schnaitman, 1993; Alexander and Valvano, 1994).

Complementation of the pglC gene product by WecA

To test our hypothesis that the wecA gene product complements PglC function we employed E. coli strain CLM37 with a deletion of the wecA gene. When the C. jejuni glycoprotein AcrA was expressed in E. coli W3110 strain along with an intact pgl glycosylation locus, mono- and diglycosylated AcrA protein was produced as visualized by the glycoprotein-specific R12 antiserum and AcrA-specific antiserum (Fig. 4, lane 1). Also, in E. coli W3110 containing the glycosylation locus with an inactivated pglC gene, an antiserum R12-reactive glycosylated form of AcrA was observed (Fig. 4, lane 2). In strain CLM37 (ΔwecA), AcrA was still glycosylated because of the action of the glycosylation locus on pACYCpgl (compare Fig. 4, lanes 3 and 6). However, in this background, knockout of PglC function resulted in production of non-glycosylated AcrA and loss of reactivity with R12 antiserum (Fig. 4, lane 4). Restoration of WecA function in trans restored glycosylation of AcrA (Fig. 4, lane 5). Thus, E. coli WecA can complement PglC function. This indicates that PglC is responsible for transfer of the first sugar residue of the heptasaccharide onto the lipid carrier.

Figure 4.

Complementation of PglC function by the E. coli-encoded wecA gene product. Whole cell extracts from E. coli strains producing the C. jejuni glycoprotein AcrA were separated by SDS-PAGE, blotted and immunodetected using either the glycosylation-specific R12 antiserum or the non-glycosylation-specific anti-AcrA antiserum (Wacker et al., 2002). Lanes 1 and 2 contain proteins derived from the E. coli strain W3110 and lanes 3–6 contain extracts from a wecA gene knockout E. coli W3110 strain termed CLM37. Lanes 5 and 6 show WecA expression in trans in CLM37 cells.

Discussion

We have recently identified a C. jejuni glycosylation locus that when transferred into E. coli is capable of glycosylating target proteins with a heptasaccharide of identical structure to that produced in C. jejuni (Wacker et al., 2002). Although mutational studies in C. jejuni have demonstrated the involvement of individual genes from this locus in the glycosylation process (Szymanski et al., 1999; Linton et al., 2002; Young et al., 2002), until now there was no direct evidence for their exact roles. Production of His-tagged PEB3 in an E. coli strain containing a plasmid-borne N-linked glycosylation locus from C. jejuni, has allowed us to knockout individual genes from the locus, isolate the target glycoprotein irrespective of its glycosylation status and characterize attached variant glycan structures by mass spectrometry. In this way, we have provided clear evidence for the role of a number of gene products from the C. jejuni glycosylation locus in the glycosylation process.

We propose that in bacteria, as in eukaryotes, a lipid (most likely undecaprenyl pyrophosphate) serves as an anchor for the assembly of the heptasaccharide glycan prior to transfer onto protein (Wacker et al., 2002). The complementation of PglC function by WecA, as discussed later, is strong evidence for the involvement of a lipid-linked oligosaccharide (LLO) intermediate. Our analysis of protein-bound oligosaccharide, as opposed to LLO analysis, is a more indirect analysis of the phenotypes of specific mutants. However, the observation that glycans of different length were readily transferred onto the reporter glycoprotein PEB3 was an important finding demonstrating that the putative oligosaccharyltransferase PglB responsible for transfer of glycan from LLO to protein has, at least in terms of glycan structure, a relaxed specificity. This is similar to the yeast oligosaccharyltransferase where glycans as small as Man2GlcNAc2 and Man1GlcNAc2 are transferred from lipid to protein (Jackson et al., 1989). Also in bacteria, the WaaL protein, implicated in the transfer of sugars to the lipid A-core oligosaccharide of the lipopolysaccharide (LPS) molecule, has a relaxed specificity (Feldman et al., 1999; Raetz and Whitfield, 2002). These data indicate that there is considerable scope for engineering glycan structures through the action of combinations of functionally characterized glycosyltransferases without compromising oligosaccharyltransferase function, thus highlighting the potential of this system for generating defined glycans that can be transferred through an N-linkage onto recombinant proteins produced in E. coli.

Structural characterization of variant glycans in knockout mutants of putative glycosyltransferases from the glycosylation locus clearly demonstrated the individual roles of glycosyltransferases in the synthesis of the heptasaccharide. Thus, the pglA, pglH, pglI and pglJ genes were shown to encode specific glycosyltransferases responsible for sequential addition of monosaccharides to form the heptasaccharide glycan (Fig. 5). The identity of the transferase(s) responsible for addition of the two terminal α-1,4-linked GalNAc residues remain(s) unknown. The most likely scenarios are that either PglH is responsible for addition of all three terminal GalNAc residues or that PglH and PglJ act alternately, adding two GalNAc residues each to synthesize the four terminal α-1,4-linked GalNAc residues. The mechanism of termination of the heptasaccharide glycan extension is also currently not known. PglA, an α1,3 N-acetylgalactosaminyltransferase responsible for addition of the GalNAc residue to bacillosamine (Fig. 5), can also transfer GalNAc to a WecA-derived HexNAc residue. Similar functional redundancy has been proposed for the pglA gene product of N. meningitidis (Power et al., 2003) that shares significant levels of amino sequence identity with PglA from C. jejuni.

Figure 5.

Function of glycosyltransferase involved in biosynthesis of the C. jejuni N-linked heptasaccharide glycan. Arrows indicate the sugar residue added by the indicated glycosyltransferase.

Knockout of the pglB and galE genes resulted in complete abolition of glycan from the target glycoprotein (Fig. 3E and F). For the pglB gene this is in agreement with previous data (Wacker et al., 2002; Young et al., 2002) and confirms the essential role of this protein in the glycosylation reaction as the presumed oligosaccharyltransferase. The galE gene is thought to encode a UDP-glucose 4-epimerase involved in lipooligosaccharide biosynthesis (Fry et al., 2000). Our data clearly demonstrate a role for the galE gene product in the biosynthesis of the N-linked glycan. Given the structure of the heptasaccharide glycan it seems likely that GalE is involved in biosynthesis of the sugars GalNAc and perhaps bacillosamine.

Our results revealed that when expressed in E. coli the C. jejuni glycosylation process interacted with E. coli pathways. The predicted pglC gene product has a high level of similarity to the C terminal region of the Salmonella enterica serovar Typhimurium WbaP protein (Fry et al., 1998), that is responsible for transfer of galactosyl-1-phosphate from UDP-galactose to carrier lipid undecaprenylphosphate, an initial stage in O-antigen LPS biosynthesis (Wang et al., 1996). This suggests a role for PglC in transfer of the first sugar residue of the N-linked heptasaccharide, bacillosamine, onto the putative lipid carrier. However, analysis of PglC function was complicated by partial complementation by the E. coli-encoded enzyme WecA so that both wecA and pglC must be disrupted to abolish glycosylation (Fig. 4). Although WecA could mediate synthesis of glycan in strains lacking PglC, the glycan produced differed from that produced with PglC function in that the sugar linked to the peptide had a mass consistent with a HexNAc residue rather than bacillosamine (Fig. 3). This also explains the two forms of glycan produced in E. coli with a wild-type glycosylation locus (Wacker et al., 2002). Presumably one form, with a HexNAc as the first sugar, is derived from WecA function and the second form, containing bacillosamine as the first sugar, is derived from PglC function.

Our data from knockouts of the pglD and wlaB genes indicate further possible interactions of the C. jejuni-derived glycosylation apparatus with E. coli-encoded proteins. Despite that individual knockout mutation of pglD and wlaB in C. jejuni result in disruption of N-linked glycosylation as measured by SBA-binding (data not shown), mutants of pglD and wlaB in E. coli produce a PEB3 that is glycosylated with heptasaccharide identical in structure to that produced in C. jejuni and in E. coli cells with a wild-type glycosylation locus (Fig. 3). The pglD gene encodes a putative acetyltransferase (Parkhill et al., 1999) postulated to be involved in biosynthesis of the diacetamido sugar bacillosamine (Wacker et al., 2002; Young et al., 2002) and we propose that an as yet unidentified E. coli-encoded acetyltransferase might complement PglD function. The wlaB gene encodes a putative inner membrane-located ABC transporter (Parkhill et al., 1999) hypothesized to transport lipid-linked oligosaccharide across the inner membrane into the periplasm prior to transfer onto protein (Wacker et al., 2002; Young et al., 2002). We presume that in the absence of WlaB an E. coli-encoded protein is able to carry out this function. The relaxed specificity (at least in terms of the saccharide moiety) of such transporters has been described (Feldman et al., 1999; Marolda et al., 2004).

The pglC and pglD genes are part of a cluster of four genes (pglC, pglD, pglE and pglF) that encode putative proteins with high levels of amino acid similarity to the products of a cluster of Neisseria meningitidis genes (pglB, pglC and pglD) involved in pilin O-linked glycosylation (Power et al., 2000). Both C. jejuni N-linked glycan and O-linked N. meningitidis pilin glycan contain the sugar residue 2,4-diacetimido-2,4,6-trideoxyhexose (Stimson et al., 1995; Wacker et al., 2002) that, in C. jejuni we know to be 2,4-diacetimido 2,4,6 trideoxyglucopyranose or bacillosamine (Young et al., 2002). The C. jejuni pglD, pglE and pglF genes are annotated as encoding a putative acetyltransferase, an aminotransferase and a dehydratase respectively (Parkhill et al., 1999). Therefore, it seems likely that both sets of genes encode proteins involved in the biosynthesis of bacillosamine presumably from a HexNAc precursor as previously speculated (Stimson et al., 1995; Wacker et al., 2002; Young et al., 2002). The results presented are supportive of a role for PglE and PglF in bacillosamine biosynthesis in that knockout of these genes results in a glycan with a HexNac residue replacing bacillosamine as the first sugar of the heptasaccharide.

In summary, mutational and structural approaches have provided the first direct evidence for the function of individual proteins involved in the biosynthesis of the C. jejuni N-linked heptasaccharide glycan and demonstrated interaction of E. coli-encoded proteins with the glycosylation process. Especially noteworthy is the role of WecA in biosynthesis of the glycan clearly demonstrating that the glycan is assembled on a lipid at the cytoplasmic face of the inner membrane. Finally, our results demonstrate that in E. coli PglB, the putative oligosaccharyltransferase is able to transfer biosynthetic intermediate glycans onto protein. This finding suggests that this and related systems (a similar putative N-linked glycosylation locus was recently discovered in a second related bacterial species; Baar et al., 2003), offer great potential for the production in E. coli of recombinant N-linked glycoproteins glycosylated with diverse glycan structures.

Experimental procedures

Bacterial strains and growth conditions

Campylobacter jejuni strains were grown as previously described  (Linton  et al.,  2002).  E.  coli strains  were  grown in Luria–Bertani (LB) broth or on LB agar at 37°C supplemented as needed with ampicillin (100 µg ml−1), kanamycin (50 µg ml−1), tetracycline (10 µg ml−1), trimethoprim (50 µg ml−1) and chloramphenicol (20 µg ml−1).

In vitro mutagenesis of the C. jejuni 81116 pgl locus cloned in pACYC184

Mutagenesis of 11 genes in the C. jejuni 81116 glycosylation locus cloned in pACYC184 (pACYCpgl) was performed in vitro using a customised EZ::TN transposon system (Epicentre, Madison, WI, USA). Briefly, a kanamycin resistance cassette (Trieu-Cuot et al., 1985) lacking a transcriptional terminator and therefore unable to exert downstream polar effects was amplified by PCR and cloned into the multiple cloning site of the vector pMOD™<MCS> (Epicentre). This construct was linearized by ScaI digestion and the kanamycin resistance cassette along with flanking mosaic ends was amplified by PCR using primers FP-1 and RP-1 (Epicentre). The PCR product was combined with plasmid pACYCpgl (Wacker et al., 2002) in an in vitro transposition reaction performed according to manufacturer's instructions (Epicentre). The resultant pool of mutated pACYCpgl plasmids was electroporated into E. coli XL1-Blue MRF′ (Stratagene) and putative mutants were screened by PCR to identify the location and orientation of the kanamycin cassette. We only used those mutants having the kanamycin resistance cassette inserted with the same transcriptional orientation as the genes of the glycosylation locus, which were also confirmed by sequence analysis.

Production of glycosylated C. jejuni PEB3 in E. coli

The peb3 gene of C. jejuni NCTC 11168 was amplified by PCR and cloned directly into the plasmid pETBlue (Novagen) according to manufacturer's instructions. Primers (forward: 5′-ATG AAA AAA ATT ATT ACT TTA TTG G-3′ and reverse: 5′-TTA GTG ATG GTG ATG GTG ATG TTC TCT CCA GCC GTA TTT TTT AAA AAT T-3′) were designed to incorporate a 6-histidine residue tag onto the C-terminal of the PEB3 protein. Expression of glycosylated PEB3 was performed in E. coli strain BL21-AI™ (Invitrogen). An overnight culture was used to seed 200 ml of LB broth supplemented with the appropriate antibiotics and containing 0.1% glucose. Following incubation at 37°C with shaking for 2 h, arabinose was added to 0.1% and cultures were incubated for a further 2 h. Following centrifugation the bacterial pellet was frozen at −20°C and PEB3 partially purified using Ni-NTA agarose beads according to manufacturer's instructions (Qiagen). Soybean agglutinin (SBA) reactivity of the PEB3 protein was determined as described previously (Linton et al., 2002). The E. coli W3310 [rph-1 IN(rrnD-rrnE)1] derivative CLM37 was used in some experiments for the production of glycosylated proteins. This strain contains a deletion of the wecA chromosomal gene that was constructed by the procedure described by Datsenko and Wanner (Datsenko and Wanner, 2000). Briefly we generated primers composed of 40–45 nucleotides corresponding to regions adjacent to the wecA coding sequence. The primers also contained 20 additional nucleotides that annealed to the template DNA from plasmid pKD4, which carries a kanamycin-resistance gene flanked by FRT (FLP recognition target) sites. Competent cells were prepared by growing E. coli W3110 carrying pKD46 in LB containing 0.5% (w/v) arabinose and the PCR products were introduced by electroporation. The plasmid pKD46 encodes the Red recombinase of the λ phage, which was placed under the control of the arabinose-inducible promoter PBAD. Kanamycin-resistant colonies were screened by PCR using primers annealing to regions outside of the mutated gene. Next, the antibiotic gene was excised by introducing the plasmid pCP20 encoding the FLP recombinase. Plasmids pKD46 and pCP20 are both thermosensitive for replication and they were cured at 42°C.

Identification of glycan structures by mass spectrometry

Glycan structures present on the C. jejuni glycoprotein PEB3 were determined by mass spectrometry. Partially purified PEB3 was run on 10% Novex™ precast gels (Invitrogen) and stained with Colloidal blue stain (Invitrogen). Relevant bands were excised, destained and digested overnight with trypsin (Promega). Tryptic peptides were extracted from gel pieces and purified using a C-18 microtrap peptide cartridge (Presearch) in preparation for sequencing by mass spectrometry (MS) and tandem mass spectrometry (MS–MS) using a hybrid quadrupole orthogonal acceleration time of flight (Q-TOF) mass spectrometer (Micromass, UK). MS and MS–MS spectra were collected in the positive ion mode as described previously (Wacker et al., 2002). Data were acquired and processed using Masslynx software (Micromass, UK). The instrument was precalibrated using a 1 pmol µl−1 solution of [Glu1]-fibrinopeptide B in acetonitrile/5% aqueous acetic acid (1:3, v:v).

Addendum

Following submission of our manuscript a report was published demonstrating that the gene previously, and within our report, termed galE encodes a bifunctional UDP-GlcNAc/Glc 4-epimerase and hence the gene was renamed gne (Bernatchez et al., 2004).

Acknowledgements

D.L. is supported by a Wellcome Trust Career Development Fellowship. B.W.W. is supported by funding from the Biotechnology and Biological Sciences Research Council. A.D. and H.R.M. are supported by funding from the Biotechnology and Biological Sciences Research Council and the Wellcome Trust. A.D. is a Biotechnology and Biological Sciences Research Council Professorial Fellow. M.A. is supported by a grant from the Swiss National Science Foundation. M.A.V. holds a Canada Research Chair in Infectious Diseases and Microbial Pathogenesis and is supported by a grant from the Canadian Institutes of Health Research. We thank Diane Newell for the gift of an anti-PEB3 antiserum and Ian S. Roberts for critical reading of the manuscript.

Ancillary