|Home | About | Journals | Submit | Contact Us | Français|
Bacterial N-linking oligosaccharyl transferases (OTase enzymes) transfer lipid-linked glycans to selected proteins in the periplasm and were first described in the intestinal pathogen Campylobacter jejuni, a member of the ε-proteobacteria-subdivision of bacteria. More recently, orthologues from other ε-proteobacterial Campylobacter and Helicobacter species and a δ-proteobacterium, Desulfovibrio desulfuricans, have been described, suggesting that these two subdivisions of bacteria may be a source of further N-linked protein glycosylation systems. Whole-genome sequencing of both ε- and δ-proteobacteria from deep-sea vent habitats, a rich source of species from these subdivisions, revealed putative ORFs encoding OTase enzymes and associated adjacent glycosyltransferases similar to the C. jejuni N-linked glycosylation locus. We expressed putative OTase ORFs from the deep-sea vent species Nitratiruptor tergarcus, Sulfurovum lithotrophicum and Deferribacter desulfuricans in Escherichia coli and showed that they were able to functionally complement the C. jejuni OTase, CjPglB. The enzymes were shown to possess relaxed glycan specificity, transferring diverse glycan structures and demonstrated different glycosylation sequon specificities. Additionally, a permissive D. desulfuricans acceptor protein was identified, and we provide evidence that the N-linked glycan synthesized by N. tergarcus and S. lithotrophicum contains an acetylated sugar at the reducing end. This work demonstrates that deep-sea vent bacteria encode functional N-glycosylation machineries and are a potential source of biotechnologically important OTase enzymes.
Asparagine-linked glycosylation is a common post-translational modification of eukaryotic proteins, and is involved in many cellular functions such as quality control, protein folding and secretion (Helenius and Aebi 2004). The eukaryotic N-glycosylation machinery is located within the endoplasmic reticulum, where the glycan is assembled on the lipid carrier dolichyl pyrophosphate in the membrane by the action of several glycosyl transferases, flipped across the membrane and finally transferred to an acceptor protein at the consensus sequon N-X-S/T, where X can be any amino acid except proline (Aebi 2013). This transfer is accomplished by the action of the oligosaccharyltransferase (OTase) enzyme complex. Initially believed to be limited to eukaryotes, N-linked glycosylation was also identified in Archaea, where a surface layer (S-layer) protein was shown to be N-glycosylated in Halobacterium salinarum (Mescher and Strominger 1976). Further investigation revealed that this type of protein modification was present in many Archaeal species, displaying a variety of N-glycan structures (Eichler 2013). The S-layer protein is the best characterized Archaeal glycoprotein, but other proteins have been identified such as the ABC transporter SSO1273 in Sulfolobus solfataricus P2 (Palmieri et al. 2013) and an unusual type IV pilus protein in Methanococcus maripaludis (Ng et al. 2011). It was not until 1999 that the first bacterial N-linked protein glycosylation system was identified in the intestinal pathogen Campylobacter jejuni (Szymanski et al. 1999). Intriguingly, while the transfer of the N-linked glycan to acceptor proteins in the eukaryotic model organism, Saccharomyces cerevisiae requires the action of the multi-enzyme OTase complex (Aebi 2013), a single enzyme, termed PglB, with significant levels of amino acid similarity to the STT3 subunit of the yeast OTase complex, was sufficient for glycan transfer (Wacker et al. 2002; Young et al. 2002). The functional transfer of the C. jejuni N-linked protein glycosylation machinery encoded by the protein glycosylation (pgl) locus into Escherichia coli allowed structural determination of the glycan and functional characterization of individual gene products involved (Wacker et al. 2002; Linton et al. 2005). Subsequent demonstration of relaxed glycan substrate specificity of the C. jejuni OTase PglB (CjPglB) paved the way for the development of recombinant glycoengineering, an approach by which a desired glycoconjugate can be generated entirely in E. coli by co-expression of a glycan-coding locus, an acceptor protein and the OTase enzyme (Feldman et al. 2005). The C. jejuni N-linked glycosylation acceptor sequon was shown to differ from the eukaryotic N-X-S/T with an extended motif containing a negatively charged amino acid at the −2 position D/E-Z-N-X-S/T (where Z, like X, can be any amino acid except proline) (Kowarik, Young, et al. 2006). The most recent survey of C. jejuni glycoproteins identified 154 glycopeptides corresponding to 53 glycoproteins, confirming the general nature of the glycosylation machinery (Scott et al. 2011).
Orthologs of the CjPglB enzyme have been found in all Campylobacter spp. genome sequences, a subset of Helicobacter species, some members of the genus Desulfovibrio, as well as some δ- and ε-proteobacterial species found in deep-sea vent habitats (Nothaft and Szymanski 2010). Three of these orthologues have been functionally characterized; those from Campylobacter lari (ClPglB) (Schwarz et al. 2011), Helicobacter pullorum (HpPglB) (Jervis et al. 2010) and the δ-proteobacterium, Desulfovibrio desulfuricans (DdPglB) (Ielmini and Feldman 2011).
Deep-sea hydrothermal vents are light-independent environments on the sea-floor that host a large community of chemolithoautotrophic proteobacterial symbionts, predominantly ε-proteobacteria (Huber et al. 2007). The genome sequences of several bacterial species isolated from deep-sea vents have been determined (Inagaki et al. 2004; Nakagawa et al. 2005, 2007; Takaki et al. 2010) and several of the species were found to possess orthologs of the C. jejuni pglB gene. In the ε-proteobacteria N. tergarcus and S. lithotrophicum, the region adjacent to the putative pglB genes contains genes encoding proteins predicted to be involved in the generation and transfer of nucleotide-activated sugars, indicating the presence of at least a partial pgl operon similar to the one encoded by C. jejuni (Figure 1A) (Nothaft and Szymanski 2010). In contrast, the pglB gene in the δ-proteobacterium D. desulfuricans appears to be an orphan gene and not part of a pgl operon. The presence of these putative pglB genes suggests that these uncharacterized organisms may possess functional protein N-glycosylation machineries. Investigation of the putative N-glycosylation systems directly in these species is complicated by their relatively complex and unusual growth requirements, such as growth in supplemented synthetic sea water and at higher temperatures (Takai et al. 2003; Nakagawa et al. 2005). We therefore characterized the putative OTase enzymes in E. coli by co-expression with an acceptor protein and a lipid-linked glycan substrate, as previously reported for the OTase enzymes of D. desulfuricans and H. pullorum (Jervis et al. 2010; Ielmini and Feldman 2011).
We present the functional expression and characterization of three novel OTase enzymes from deep-sea vent bacteria, identify a possible glycoprotein encoded by one of these species and gain an insight to the nature of the native N-linked glycan structures.
In order to identify putative orthologs of CjPglB, the amino acid sequence was used as the query against all databases of prokaryotic proteins using blastp. Numerous non-Campylobacter orthologs were identified, and three encoded by deep-sea vent bacteria were chosen for further analysis (Figure 1B; Supplementary data, Figure 1). These included Nitratiruptor tergarcus (54% amino acid similarity, 34% identity to the C. jejuni PglB enzyme) and Sulfurovum lithotrophicum (55% similarity, 37% identity), two representatives of the ε-proteobacteria and Deferribacter desulfuricans (40% similarity, 24% identity), a species classified within the phylum Deferribacteres from the δ-proteobacteria.
Several amino acid residues and structural features have been identified as important for OTase activity in ClPglB (Lizak et al. 2011; Ihssen et al. 2012, 2014; Gerber et al. 2013). An amino acid alignment of the three deep-sea vent OTase enzymes with CjPglB indicates that the deep-sea vent OTase enzymes possess all the important residues for the function of ClPglB (highlighted in Supplementary data, Figure S2 and Table I), aside from residues R331 and I572 that are absent in DfdPglB.
In most Campylobacter species, the OTase gene is present within the locus encoding for the assembly of the N-linked oligosaccharide (Nothaft and Szymanski 2010). The genes flanking the deep-sea vent putative pglB genes from N. tergarcus (NtPglB) and S. lithotrophicum (SlPglB) include genes predicted to encode an initiating undecaprenol-phosphate sugar phosphotransferase (PglC) and a number of glycosyltransferases. However, no gene encoding a “flippase” enzyme required for membrane translocation of the lipid-linked oligosaccharide (LLO) into the periplasm was identified. In contrast, there are no orthologs of C. jejuni N-linked glycosylation pathway genes adjacent to the D. desulfuricans (DfdPglB) ortholog (Figure 1A).
For functional analysis, the putative OTase enzymes were tested for their ability to complement CjPglB in E. coli. The predicted ORFs coding for NtPglB, SlPglB and DfdPglB were codon-optimized, synthesized and cloned into inducible expression vectors of the pEXT family (see Methods) (Dykxhoorn et al. 1996). Initial activity assays tested the ability of the OTase enzymes to transfer the C. jejuni N-linked heptasaccharide to the commonly used C. jejuni reporter glycoproteins, AcrAand Cj0114. The three deep sea vent OTase enzymes (and CjPglB as a positive control) were co-expressed with hexa-his-tagged AcrA and the C. jejuni pgl gene locus with an insertionally inactivated pglB gene in E. coli strain CLM24 (Feldman et al. 2005). Analysis of purified AcrA by western blotting, using anti-hexa-his and anti-heptasaccharide antibodies demonstrated CjPglB-mediated glycosylation of AcrA at two sites as previously reported (Wacker et al. 2002), but no modification was observed by the three putative deep-sea vent OTase enzymes (Figure 2A and B).
Analysis of purified Cj0114 confirmed CjPglB-dependent glycosylation of Cj0114 at four extended N-linked glycosylation sequons as previously reported (Jervis et al. 2010) (Figure 2C and D). Two of the four Cj0114 glycoforms were observed with NtPglB and SlPglB (Figure 2C and D) demonstrating OTase activity of these two enzymes. No extra bands were observed for DfdPglB, suggesting that this enzyme was not able to transfer the C. jejuni heptasaccharide to the Cj0114 protein. These data confirm protein N-linked glycosylation activity of two of three deep-sea vent OTase enzymes.
To further investigate NtPglB-mediated Cj0114 glycosylation, we employed Cj0114 mutants in which each of the asparagine residues within the four glycosylation sequons (N100, N154, N172 and N178) were replaced with glutamine (Jervis et al. 2010). Co-expression of these four variants with pACYCpglΔpglB and CjPglB generated three Cj0114 glycoforms as expected (Figure 3A, left panel). Co-expression of Cj0114 N100Q and N178Q with pACYCpglΔpglB and NtPglB resulted in two glycoforms as for wild-type Cj0114, however, Cj0114 N154Q and N172Q produced only a single glycoform (Figure 3A right panel). This demonstrates that asparagine residues N154 and N172, but not N100 and N178, are required for modification with the C. jejuni heptasaccharide by NtPglB. The sequons surrounding N154 and N172 contain an aspartic acid at the −2 position, while the sequons surrounding N100 and N178 contain glutamic acid, suggesting that NtPglB may display a preference towards the former. However, replacement of aspartic acid at position −2 of N172 with a glutamic acid (D170E) did not disrupt glycosylation with NtPglB (data not shown). To investigate the requirement for a negatively charged residue at the −2 position for activity of NtPglB and SlPglB, the aspartic acid at the −2 position of N172 was replaced with alanine (mutant D170A). This resulted in one less Cj0114 glycoform produced by CjPglB, NtPglB and SlPglB (Figure 3B), demonstrating that as for CjPglB both NtPglB and SlPglB require a negatively charged amino acid at the −2 position.
CjPglB possesses relaxed glycan specificity (Feldman et al. 2005). In order to assess the ability of the three deep-sea vent OTase enzymes to transfer a variety of glycan moieties, the enzymes were co-expressed with Cj0114 in E. coli strain E69 that synthesizes the O9 O-antigen with N-acetylglucosamine (GlcNAc) at the reducing end (McCallum et al. 1989), and in E. coli strain CLM24 producing the F. tularensis O-antigen with QuiNAc (2-acetamido-2,6-dideoxy-O-d-glucose) at the reducing end (Cuccui et al. 2013). All three OTase enzymes were able to transfer both structures to Cj0114 (Figure 4), demonstrating that similar to CjPglB, the deep-sea vent OTase enzymes possess relaxed glycan specificity. Interestingly, both NtPglB and SlPglB preferentially transferred shorter chains of the F. tularensis O-antigen compared with CjPglB, while the transfer efficiency of DfdPglB appeared very low.
As Cj0114 was glycosylated by the deep-sea vent OTase enzymes in the recombinant E. coli system, a BLAST search was performed to identify potential orthologs encoded by the three species. A Cj0114 ortholog was identified in each species, and designated Nt0114 (34% identity to Cj0114), Sl0114 (30% identity to Cj0114) and Dfd0114 (29% identity to Cj0114). Only Nt0114 contained an extended bacterial N-glycosylation sequon, whilst Sl0114 and Dfd0114 contained two and three eukaryotic N-X-S/T sequons, respectively (Figure 5, Supplementary data, Figures S3 and S4).
Both Nt0114 and Dfd0114 were tested for their capacity to be glycosylated through expression in N-linked glycosylation competent E. coli CLM24 producing either the C. jejuni N-linked heptasaccharide glycan or F. tularensis O-antigen. In this system, Nt0114 was glycosylated with the C. jejuni heptasaccharide and F. tularensis O-antigen by CjPglB but not by any of the three deep-sea OTase enzymes (Supplementary data, Figure S3). Interestingly, when Dfd0114 lacking extended bacterial N-linked glycosylation sequons was co-expressed in E. coli CLM24 with OTase enzymes and the C. jejuni heptasaccharide, a glycosylated form was detected in the presence of CjPglB, suggesting transfer of glycan to a eukaryotic sequon (Figure 5B). When Dfd0114 was co-expressed with the F. tularensis O-antigen and the OTase enzymes, glycoforms were observed in the presence of CjPglB and DfdPglB, but not NtPglB or SlPglB (Figure 5B). This suggested glycan transfer by both CjPglB and DfdPglB to a eukaryotic-type sequon. CjPglB was previously believed to strictly require a negatively charged residue at the −2 position of the sequon (Kowarik, Young, et al. 2006). However, recent work has demonstrated modification of asparagine residues not located within an extended sequon by CjPglB, both in the native host as well as in E. coli (Ollis et al. 2014; Scott et al. 2014). To identify the Dfd0114 sequon glycosylated by CjPglB, site-directed mutant constructs were generated by individually changing the asparagine residue in each of the three eukaryotic sequons to alanine. The mutant N107A was no longer glycosylated by CjPglB, while mutants N101A and N118A remained glycosylated, suggesting transfer of the glycan to asparagine N101 by CjPglB (Figure 5C).
Unlike C. jejuni, the genes involved in the synthesis of the N-linked LLO in N. tergarcus, S. lithotrophicum and Deferribacter desulfuricans are not encoded within a single locus (Nothaft and Szymanski 2010) (Figure 1A). However, a putative undecaprenol-phosphate UDP-glycosyl transferase (pglC) is located immediately downstream of pglB in both N. tergarcus and S. lithotrophicum (Figure 1A). Additionally, orthologs of two of the three enzymes involved in the synthesis of UDP-N,N′-diacetylbacillosamine (diNAcBac) from UDP-N-acetylglucosamine (UDP-GlcNAc), pglE and pglF, but not pglD were identified downstream of pglB and pglC (Morrison and Imperiali 2014). No orthologs of pglC, E or F were identified from D. desulfuricans, suggesting that the native glycan transferred in this organism is unlikely to contain diNAcBac at the reducing end. Recombinant expression of NtPglC or SlPglC together with Cj0114 and pACYCpglpglC::Km in E. coli strain CLM37, which lacks the initiating transferase of the O-antigen biosynthesis repeat unit synthesis machinery (WecA), resulted in Cj0114 glycosylation with a glycan recognized by the anti-C. jejuni heptasaccharide antiserum HR6 (Figure 6A). This demonstrates that both deep-sea vent PglC orthologs can functionally complement Cj PglC and are able to transfer a glycan moiety to undecaprenol-phosphate, on which the full C. jejuni heptasaccharide can be synthesized by enzymes encoded by the C. jejuni pgl locus. It has previously been shown that CjPglB can transfer heptasaccharides with either a diNAcBac or a GlcNAc residue at the reducing end (Linton et al. 2005). However, CjPglC has been shown to catalyze only the transfer of UDP-Bac to undecaprenol-phosphate (Glover et al. 2006). Elucidation of the nature of the reducing end glycan requires mass spectrometric analysis; however, the full-length Cj0114 protein was unsuitable for mass spectrometry analysis. A modified C-terminal truncation of Cj0114 termed NGRP, which displays improved performance of the N172 tryptic peptide during MS analysis has previously been generated (A. J. Jervis, unpublished). This protein was used in an analogous PglC complementation experiment, and the resultant glycoform subjected to tandem MS analysis to identify the nature of the glycan at the reducing end of the oligosaccharide (Figure 6B). This showed the presence of a 228 Da residue consistent with a diNAcBac at the reducing end of the glycan, demonstrating that this was the substrate for both NtPglC and SlPglC. Two further minor peaks were observed in the NtPglC sample. The peak at 2822 corresponds to a C. jejuni heptasaccharide with a HexNAc at the reducing end based on mass difference. The peak at 2832 either corresponds to a C. jejuni heptasaccharide with an unknown glycan at the reducing end or is the result of partial fragmentation. This result suggests that NtPglC may also be able to transfer a HexNAc residue to undecaprenol-phosphate with low efficiency. Unfortunately, the intensity of the peaks was too low to sequence the glycan and confirm this hypothesis.
The discovery and functional characterization of a bacterial N-linked general protein glycosylation system in C. jejuni (Szymanski et al. 1999; Wacker et al. 2002) challenged the dogma that this type of post-translational protein modification was limited to the eukaryotic and archaeal kingdoms. Since then, further bacterial N-linking OTase enzymes have been functionally characterized (Jervis et al. 2010; Ielmini and Feldman 2011; Schwarz et al. 2011). Intriguingly, genome sequencing of three bacterial species from a deep-sea vent habitat identified genes encoding putative orthologs of the C. jejuni OTase PglB (Nakagawa et al. 2007; Nothaft and Szymanski 2010; Takaki et al. 2010). We report the functional characterization of the OTase orthologs encoded by these three species (Figure 1) using a recombinant approach in E. coli.
We have demonstrated that the N. tergarcus, S. lithotrophicum and D. desulfuricans pglB orthologs encode functional OTase enzymes that are able to transfer lipid-linked oligo- and polysaccharides to an acceptor protein. However, while it has been shown that CjPglB is able to glycosylate any N-glycosylation sequon as long as the protein is targeted to the periplasm and the glycosylation sequon is present within a flexible, accessible loop (Kowarik, Numao, et al. 2006; Kowarik, Young, et al. 2006; Fisher et al. 2011), our data suggest a more stringent acceptor protein requirement for the three deep-sea vent OTase enzymes, as only one acceptor protein, Cj0114, was glycosylated and at only two of the four possible sites. Such an acceptor protein specificity has not been reported previously for N-linking OTase enzymes, but has been demonstrated for two bacterial O-linking OTase enzymes (Horzempa et al. 2006; Harding et al. 2015). It is therefore possible that the deep-sea vent enzymes present a new class of N-linking OTase enzymes with more stringent acceptor protein specificity.
Even within the one protein that was successfully glycosylated by the deep-sea vent OTase enzymes, a preference for particular sequons was observed. This preference was not a result of the primary amino acid sequence of the sequon, as altering the non-modified sequons did not result in glycosylation. The sequon preference was also not due to general unavailability of the sequon within the secondary structure of the protein, as CjPglB was able to glycosylate all four sequons. It has previously been shown that the H. pullorum PglB1 was only able to glycosylate two of the four sequons within the Cj0114 protein (Jervis et al. 2010), and alternative glycosylation of the C. jejuni protein AcrA by the enzymes from C. lari and D. desulfuricans has also been reported (Ielmini and Feldman 2011; Schwarz et al. 2011), suggesting that sequon usage may be variable among bacterial N-OTase enzymes.
As both NtPglB and SlPglB possess the R331 amino acid residue (Table I) which is implicated in the interaction of ClPglB with the negatively charged residue at the −2 position of the extended glycosylation sequon (Lizak et al. 2011), it was unsurprising that glycosylation by both enzymes required this negative charge. R331 is absent in DfdPglB, similar to the ortholog encoded by D. desulfuricans, which is able to glycosylate sequons lacking the negatively charged residue at the −2 position (Ielmini and Feldman 2011). However, it was not possible to identify the Cj0114 sequon preference of this OTase due to low levels of activity. Similar to CjPglB (Wacker et al. 2006; Cuccui et al. 2013), all three OTase enzymes displayed relaxed glycan specificity, and were able to transfer both the E. coli O9 O-antigen and the F. tularensis O-antigen.
In order to support the hypothesis of a fully functional N-glycosylation machinery encoded by deep-sea vent bacteria, we sought to identify putative native glycoproteins. We identified orthologs of Cj0114 in all three bacterial species. The S. lithotrophicum ortholog (Sl0114) does not contain any glycosylation sequons, while the N. tergarcus protein (Nt0114) contains one extended sequon and the D. desulfuricans ortholog (Dfd0114) contains three eukaryotic sequons lacking the −2 negatively charged residues. Recombinant expression of Dfd0114 with the C. jejuni N-linked heptasaccharide and CjPglB resulted in transfer of the glycan to Dfd0114 at an asparagine residue within the sequence PNNNIS. The C. lari PglB has been shown to glycosylate asparagine residues not located within an extended bacterial sequon both in vivo (Schwarz et al. 2011) and in vitro (Gerber et al. 2013) and recent evidence suggests that CjPglB has similar activity, both in the native host as well as in a recombinant E. coli system (Ollis et al. 2014; Scott et al. 2014). No glycosylation of Dfd0114 with the C. jejuni heptasaccharide was observed for DfdPglB. However, low levels of Ddf0114 glycosylation by DfdPglB was observed when co-expressed with the F. tularensis O-antigen. This indicates that DfdPglB, similar to the enzyme from Desulfovibrio desulfuricans is able to glycosylate short, eukaryotic sequons (Ielmini and Feldman 2011). Further studies in Deferribacter desulfuricans are required to confirm the glycosylation status of Dfd0114 in the native organism as well as to identify the native glycoproteome and investigate the presence of eukaryotic glycosylation sequons within these proteins. In contrast, Nt0114 was only glycosylated by CjPglB, and not by any of the deep-sea vent OTase enzymes, including the “native” NtPglB (Supplementary data, Figure S3). Glycosylation of Nt0114 by CjPglB demonstrated that the protein is present in the correct subcellular compartment and the sequon is in principle accessible to the OTase enzymes. This suggests that the lack of glycosylation of Nt0114 by the deep-sea vent OTase enzymes is likely due to the more stringent acceptor protein specificity of these enzymes. This does not, however, rule out the possibility that this protein may be glycosylated in N. tergarcus. Further studies in N. tergarcus are required to address this and to identify the complete native glycoproteome.
No data are available regarding N-linked glycan structure in N. tergarcus or S. lithotrophicum. To investigate the nature of the sugar residue present at the reducing end of the N-glycan produced by these species, a CjPglC complementation experiment using the two putative initiating glycosyl transferase enzymes encoded by N. tergarcus and S. lithotrophicum was performed. This demonstrated that both NtPglC and SlPglC are able to transfer a diNAcBac residue to undecaprenol-phosphate, on top of which the remaining heptasaccharide was assembled. In the case of NtPglC, a small amount of glycan likely containing a HexNAc at the reducing end as judged by mass difference was also observed, suggesting that NtPglC is able to transfer a HexNAc to undecaprenol-phosphate at a very low rate. Analysis of the structure of the glycans in the native organisms, or attempts to reconstitute the complete N-glycan biosynthesis pathway in E. coli are required to fully investigate the structure of the glycan synthesized by the two species.
The demonstration of functional N-linked glycosylation systems encoded by deep-sea vent bacteria raises interesting questions regarding the evolution of this post-translational modification system and the role of N-linked protein glycosylation in the biology of these species. While this type of post-translational modification is ubiquitous among higher organisms, and has also been found to be encoded by 166 of 168 archaeal genome sequences obtained to date (Kaminski, Lurie-Weinberger, et al. 2013), it has so far only been shown in a small number of bacterial species (Jervis et al. 2010; Nothaft and Szymanski 2010; Ielmini and Feldman 2011; Schwarz et al. 2011). In eukaryotes, the role of the N-glycan is multifunctional, ranging from protein quality control to secretion and interaction between proteins, and the modification is essential for the function of the cell (Aebi 2013). In the archaeal species H. volcanii and members of the genus Methanococcus, the N-glycosylation machinery is not essential for cell viability (Jarrell et al. 2010). However, disruption of N-glycosylation in H. volcanii led to decreased ability to grow in high salt concentrations (Kaminski, Naparstek, et al. 2013). Additionally, H. volcanii cells containing a disrupted OTase gene were unable to produce intact flagella and were non-motile (Tripepi et al. 2012). Glycosylation of the S-layer protein in H. volcanii was recently shown to be dependent on salinity levels, suggesting a role in survival in the relatively harsh environment (Kaminski, Guan, et al. 2013). It has been demonstrated that CjPglB can function as a hydrolase in addition to an OTase, resulting in the release of the N-glycan as a free oligosaccharide into the periplasm of the bacterium (Nothaft et al. 2009). This release has been shown to be influenced by altering the salt and osmolyte concentration of the environment, suggesting an adaptive function similar to that observed for archaea. It can be speculated that the OTase enzymes encoded by deep-sea species, living predominantly under harsh environmental conditions such as high temperature and high osmolarity, may contribute to the survival in those conditions by generation of free oligosaccharides to counteract the high osmotic levels of sea water. It has also been shown that a glycosylated form of the C. jejuni glycoprotein PEB3 is more thermostable than a non-glycosylated protein (Min et al. 2009). Therefore, protein glycosylation in the deep-sea vent bacteria may increase the overall thermostability of the glycosylated subsection of the proteome.
Studies in C. jejuni using cells deficient in either the OTase CjPglB or enzymes involved in synthesis of the lipid-linked heptasaccharide have suggested roles for N-linked glycosylation in chicken colonization and adhesion and invasion of epithelial cells (Szymanski et al. 2002; Karlyshev et al. 2004). However, as a total number of 53 C. jejuni proteins have been shown to be N-glycosylated (Scott et al. 2011), it has not been possible to identify the precise role of glycosylation in this pathogen. It has also been demonstrated that the C. jejuni N-glycan is recognized by the human galactose-type lectin MGL, suggesting a potential role for the glycan in modulation of the immune system (van Sorge et al. 2009). As many deep-sea vent bacterial species can exist as free-living biomass as well as symbionts on other deep-sea vent animals such as polychaetes and shrimps (Polz and Cavanaugh 1995), it can be speculated that the bacteria may employ an N-linked glycan to interact with and modulate the immune system of their symbiotic partners. However, more studies in the native organisms are needed to address the function and scope of protein glycosylation in these species.
From a glycoengineering point of view, it is interesting to note that two of the three organisms that encode the OTase enzymes characterized here require higher growth temperature than those used for the recombinant expression in the E. coli host. This has two potential implications. Firstly, these enzymes may possess different biophysical characteristics compared with the best studied CjPglB, such as better thermostability/“shelf life”. Secondly, the activity of the recombinant OTase enzymes in the in vivo E. coli expression host may be different from the activity in the native host. Further studies in the native organisms, as well as adaptation of published in vitro glycosylation experiments (Kowarik, Numao, et al. 2006; Jervis et al. 2010) for testing of these OTase enzymes under different conditions are required to investigate these possibilities.
In summary, this is the first functional characterization of bacterial OTase enzymes encoded by three bacterial species from a deep-sea vent habitat and paves the way for further studies of the role of protein N-glycosylation in these specialized bacteria and other bacterial species such as C. jejuni. Additionally, the study provides a deeper understanding of a biotechnologically important class of enzymes.
Escherichia coli strains were grown in Lysogeny Broth (LB) or on LB agar at 37°C. Where required, the medium was supplemented with antibiotics at the following concentrations: 100 µg mL−1 ampicillin, 34 µg mL−1 chloramphenicol, 50 µg mL−1 kanamycin and 100 µg mL−1 spectinomycin. Escherichia coli DH5α library efficiency cells (Invitrogen, Carlsbad) were routinely used as a host for cloning experiments. Escherichia coli strain E69 was kindly provided by Chris Whitfield (University of Guelph, Canada). All strains and plasmids are listed in Table II.
The ORFs encoding NtPglB, SlPglB and DfdPglB, as well as the putative acceptor proteins were codon-optimized for expression in E. coli, synthesized by Celtek Genes (Celtek-genes) and delivered in vector pGH flanked by recognition sites for restriction endonucleases SacI and XbaI for the pglB genes and EcoRI and XbaI for the acceptor protein and pglC genes. The ORFs were subcloned by restriction digestion into vectors pEXT20, pEXT21, pEXT22 or pMLBAD (Table II) and sequence verified.
Escherichia coli CLM24 cells were transformed with plasmids pACYCpglΔpglB, along with a plasmid encoding an acceptor protein (see Table II for details) and either a control plasmid or a plasmid encoding a PglB. One colony was grown in LB broth to an OD600 of 0.4–0.6 at 37°C, and pglB expression induced with 1 mM IPTG. Cultures were grown for an additional 16 h, the cells harvested by centrifugation, lysed (see below) and acceptor proteins purified by immobilized metal ion affinity chromatography (IMAC).
Plasmid pBRCj0114 was co-expressed with a plasmid encoding the desired PglB protein in E. coli strain E69 which synthesizes the O9 O-antigen, and induction and sample preparation were performed as above.
Escherichia coli CLM24 cells were transformed with plasmids pGAB2, pBRCj0114 and either a control plasmid or a plasmid encoding the desired PglB protein. Protein expression and sample preparation were performed as above.
Briefly, cell pellets obtained after overnight induction of the acceptor proteins were resuspended in lysis solution (500 mM NaCl, 25 mM NaH2PO4, 15 mM imidazole containing 1 mg/mL lysozyme, pH 7.5), and sonicated. Lysates were clarified by centrifugation, and Ni-NTA agarose added. After 1 h with mixing, the slurry was loaded onto a Pierce Spin cup, washed five times with wash solution (500 mM NaCl, 25 mM NaH2PO4, 25 mM imidazole, pH 7.5), and bound proteins eluted with elution buffer (500 mM NaCl, 25 mM NaH2PO4, 500 mM imidazole, pH 7.5).
Purified acceptor proteins were separated by SDS–PAGE using NuPAGE™ Novex™ 4–12% Bis-Tris protein gels (Invitrogen), transferred to nitrocellulose membrane and analyzed by two color immunoblot using anti-hexahistidine, anti-glycan and corresponding fluorescent-labelled secondary antibodies (Table III) using an Odyssey near-infrared imager (LI-COR Biosciences).
Site-directed mutagenesis of acceptor protein sequons was performed using the QuikChange XL site-directed mutagenesis kit (Agilent) according to the manufacturer's instructions.
Coomassie-stained bands in SDS–PAGE gels were excised and subjected to in-gel trypsin digestion followed by cleanup on a C18 Zip-Tip (Millipore). MALDI-TOF MS and MALDI-LIFT-TOF/TOF MS spectra were acquired by laser-induced dissociation using a Bruker Ultraflex II mass spectrometer in the positive-ion reflection mode with 2,5-dihydroxybenzoic acid (20 mg mL−1 in 0.1% formic acid, 30% acetonitrile) as the matrix. Data were analyzed with FlexAnalysis 3.0 software (Bruker Daltonics).
This work was supported by the Wellcome Trust (grant 102978/Z/13/Z) and the Biotechnology and Biological Sciences Research Council (grant BB/H017437/1).
GlcNAc, N-acetylglucosamine; IMAC, immobilized metal ion affinity chromatography; LB, Lysogeny Broth; LLO, lipid-linked oligosaccharide; diNAcBac, N,N'-diacetylbacillosamine, OTase, oligosaccharyltransferase; UDP-GlcNAc, UDP-N-acetylglucosamine.
Plasmid pGVXN114 encoding CjPglB was obtained from Glycovaxyn AG, Switzerland under MTA. We thank David Knight and the staff of the Biomolecular Analysis Core Research Facility, Faculty of Life Sciences, University of Manchester, for assistance with MS.