Introduction

N-glycosylation is a prevalent protein co/post-translational modification which has been found in all three domains of life (Aebi 2013; Jarrell et al. 2014; Nothaft and Szymanski 2010). In the Archaea, bioinformatics analysis detected at least one copy of the predicted oligosaccharyltransferase (OST) gene (aglB), encoding the most conserved enzyme in the N-glycosylation system, in 166 of 168 sequenced archaeal genomes, indicating the almost universal presence of this protein modification throughout the third domain (Kaminski et al. 2013). This is in contrast to the, so far, limited observation of N-glycosylation in Bacteria where the pathway seems restricted to most members of the Epsilon-subdivision of Proteobacteria (Campylobacter, Wolinella, Helicobacter) and a subset of the Deltaproteobacteria, including Desulfovibrio species (Nothaft and Szymanski 2010, Nothaft and Szymanski 2013). Archaea also have the most diverse N-glycan structures, with a great variety of sugar components (including unique sugars), modifications and glycosidic linkages, as well as a range of linking sugars, including simple hexoses (Jarrell et al. 2014).

Although the N-glycosylation system is widespread in Archaea, genetic studies of the archaeal N-glycosylation pathway, in combination with biochemical and structural approaches, are mainly restricted to three model microorganisms: Methanococcus maripaludis, Sulfolobus acidocaldarius and Haloferax volcanii (Jarrell et al. 2014). Notably, however, structures for AglB are only available from other species including Archaeoglobus fulgidus and Pyrococcus furiosus (Igura et al. 2007, Matsumoto et al. 2013). Based on the studies in these three microorganisms, an archaeal N-glycosylation model has been proposed (Jarrell et al. 2014). On the inner leaflet of the cytoplasmic membrane, various glycosyltransferases (GTs) sequentially add sugar monomers onto a dolichol lipid carrier (either mono- or diphosphorylated; Dol-P or Dol-PP) embedded in the cytoplasmic membrane. This lipid-linked N-glycan precursor is then translocated across the cytoplasmic membrane via a flippase, and at the exterior face of the cytoplasmic membrane, the N-glycan is transferred en bloc by AglB from the lipid carrier onto select asparagine residues in an Asn-Xaa-Ser/Thr (Xaa ≠ Pro) sequon in the acceptor glycoprotein. Additional sugars may still be added to the glycan after its transfer to the acceptor protein (Cohen-Rosenzweig et al. 2012). The OST, as well as a large number of archaeal GTs and enzymes involved in the biosynthesis of the individual sugars comprising the glycan have been identified in the three model microorganisms (Jarrell et al. 2014).

To date, archaeal N-glycoproteins are typically external proteins or membrane-bound proteins and include archaellins (formerly archaeal flagellins (Albers and Jarrell 2015, Jarrell and Albers 2012)), S-layer proteins, pilins, cytochrome b558/556 in S. acidocaldarius, and even an archaeal viral protein (Jarrell et al. 2014). Most studies of the archaeal N-glycosylation pathway have employed either archaellins or S-layer proteins as the reporter protein (Jarrell et al. 2010).

In M. maripaludis, the swimming organelle archaellum is composed of three structural proteins: the major archaellins FlaB1 and FlaB2 and the minor archaellin FlaB3, the latter forming the curved, hook-like region (Chaban et al. 2007). All three archaellins are N-glycosylated at multiple sites with the tetrasaccharide Sug-4-β-ManNAc3NAmA6Thr-4-β-GlcNAc3NAcA-3-β-GalNAc, where Sug is (5S)-2-acetamido-2,4-dideoxy-5-O-methyl-L-erythro-hexos-5-ulo-1,5-pyranose, which to date has been found exclusively in this species (Kelly et al. 2009). The Epd pili of M. maripaludis (Nair et al. 2014) are also composed of glycoproteins, however the major pilin, EpdE, has an N-linked pentasaccharide consisting of the archaellin tetrasaccharide with an additional hexose added as a branch to the linking sugar GalNAc (Ng et al. 2011).

Using both genetic and biochemical methods, a series of enzymes have been demonstrated in M. maripaludis to be involved in the N-glycosylation or sugar biosynthesis pathways. The genes for these components have been localised mainly to two regions. The first region extends from mmp1079 to mmp1088, including aglO (mmp1079), aglA (mmp1080) and aglL (mmp1088) encoding the second, third and fourth GTs, respectively (Vandyke et al. 2009); aglXYZ (mmp1081–1083) encoding enzymes involved in the acetamidino modification of the third sugar (Jones et al. 2012); aglU (mmp1084) encoding the enzyme that transfers the threonine onto the third sugar; and aglV (mmp1085) encoding a methyltransferase for the methylation of the unique fourth sugar (Ding et al. 2013). The second region is located from mmp0350 to mmp0359. Several of the genes located in this area are involved in the biosynthesis of the second and third sugars of the glycan, including agl17 (mmp0350), agl18 (mmp0351), agl19 (mmp0352), agl20 (mmp0353) and agl21 (mmp0357) (Namboori and Graham 2008a; Siu et al. 2015; VanDyke et al. 2008). The gene for the OST, aglB (mmp1424) is found at a separate locus (Vandyke et al. 2009). Besides the work mentioned above, in vitro studies employing heterologously expressed and purified enzymes have identified other genes, including mmp0705, mmp0706, mmp1077 and mmp1680, that are involved in acetamido sugar biosynthesis (Namboori and Graham 2008b). However, genetic studies are still needed to confirm the involvement of these genes in archaellin N-glycan sugar biosynthesis. Key enzymes in the N-glycosylation pathway, i.e. the GT responsible for the transfer of the first sugar to the dolichol carrier and the putative flippase, have not yet been identified in M. maripaludis, although MMP1423 is a strong candidate for the first glycosyltransferase. MMP1423 is a putative family 4 glycosyltransferase with a GT_GPT_archaea domain, found in UDP-GlcNAc:dolichol-P GlcNAc-1-P transferase (GPT)-like proteins in archaea. In Eukaryotes, GPT catalyse the initial step in the N-glycosylation pathway, i.e., the transfer of GlcNAc-1-P from UDP-GlcNAc to dolichol-P resulting in the formation of GlcNAc-P-P-dolichol. Previous attempts to delete mmp1423, however, have been unsuccessful (Vandyke et al. 2009).

In M. maripaludis, N-glycosylation plays an important role in archaella formation. Archaella are only assembled using archaellins that are modified with disaccharide or longer N-glycans, i.e. in ΔaglA and ΔaglL mutants, but not with monosaccharide-modified or non-glycosylated archaellins, i.e. ΔaglO or ΔaglB mutants (Siu et al. 2015; Vandyke et al. 2009). While archaella are assembled with a disaccharide or trisaccharide attached to the archaellins, motility is reduced in these mutants compared to that of wild type (WT) cells as measured in swarm plate assays. The swarming diameters in these assays are directly related to the size of the N-linked glycan, with the cells producing a disaccharide glycan swarming the least (Vandyke et al. 2009). Interestingly, this minimum disaccharide N-glycan rule does not apply to pili assembly, since pili were observed on the cell surface of a ΔaglB mutant where the constituent pilins would be non-glycosylated (Vandyke et al. 2009). On the other hand, not all of the N-glycosylation sites normally occupied in the archaellins in WT cells are necessary for archaella formation (Ding et al. 2015). Archaella could be assembled using FlaB2 in which 3 out of 4 N-glycosylation sites were removed by site-directed mutagenesis (SDM) (leaving only the 1st N-glycosylation site intact), but not with non-glycosylated FlaB2 (Ding et al. 2015).

In this work, we studied a six-gene operon mmp1089–1094, neighbouring the first identified N-glycosylation genetic region, and demonstrated that one gene, mmp1090, is involved in the unique fourth sugar biosynthesis pathway. No significant change in the archaellin N-glycan was observed in either Δmmp1091 or Δmmp1092 mutant cells. Several attempts to delete the other genes in the operon, namely mmp1093 (annotated as coaD, (Sarmiento et al. 2013)), mmp1094 (annotated as ppsA (Sarmiento et al. 2013)) or mmp1089 (Vandyke et al. 2009) were unsuccessful, suggesting that these genes might be essential for M. maripaludis.

Materials and methods

Strains and growth conditions

M. maripaludis S2 Δhpt (Mm900) (Moore and Leigh 2005) and the mutants derived from Mm900 (Table 1) were cultured anaerobically in sealed serum bottles containing 10 mL medium under an atmosphere of CO2–H2 (20:80) at 35 °C with shaking. Unless otherwise specified, cells were cultured in Balch medium III (Balch et al. 1979). McCas medium was used at various steps during the creation of in-frame deletions (Moore and Leigh 2005). A final concentration of 1 mg/mL neomycin was added to McCas medium as needed for selection following transformation. To obtain M. maripaludis single colonies, cells were plated onto McCas medium-Noble agar (1.5 %, w/v) plates containing 240 µg/mL 8-aza-hypoxanthine (Acros Organics, NJ) for selection and incubated inside of an anaerobic canister at 37 °C for 5 days (Moore and Leigh 2005). M. maripaludis strains carrying a complementation plasmid were cultured in nitrogen-free medium supplemented with either 10 mM l-alanine or 10 mM NH4Cl (Lie et al. 2005) as sole nitrogen source in the presence of puromycin (2.5 µg/mL) for plasmid selection. Escherichia coli TOP10 cells used for molecular cloning steps were cultured at 37 °C with shaking in Luria–Bertani broth (LB) or on LB plates with 100 µg/mL ampicillin for selection.

Table 1 Strains and plasmids used in this study
Full size table

Reverse-transcription (RT)-PCR

To determine if the mmp1089 to mmp1094 genes were co-transcribed, RT-PCR was performed as per the manufacturer’s protocol using a One-Step RT-PCR kit (Qiagen Inc.) with primers (Table 2) which amplify across each of the intergenic regions between neighbouring genes from mmp1089 to mmp1094. The RNA template was extracted from Mm900 cells using a High Pure RNA Isolation Kit (Roche Inc.), followed by an additional DNase treatment (Turbo DNA-free Kit, Ambion) at 37 °C for 30 min. PCR amplifications with the same primer pairs were also performed using either purified RNA not subjected to reverse transcription as template to exclude the possibility of genomic DNA contamination of the RNA, or genomic DNA from Mm900 as template to confirm the amplicon size and primer specificity.

Table 2 Primers used in this study
Full size table

Plasmid constructions for in-frame gene deletions

Plasmids used for in-frame deletions of the targeted genes in M. maripaludis (Table 1) were constructed as previously described using the primers listed in Table 2 (Chaban et al. 2007, Moore and Leigh 2005). Briefly, ~1 kb of the DNA upstream (gene specific P1 and P2 primers) and downstream (gene specific P3 and P4 primers) of the targeted gene was amplified by PCR and then ligated via the AscI restriction sites incorporated into the two interior primers P2 and P3. After ligation, the majority of the target gene is deleted, leaving a short residual 5′ and 3′ piece which is still in frame. This ligation product was used as template to perform a second PCR with the exterior primers P1 and P4 and the resulting amplification product was cloned into the BamHI or XbaI restriction site in the vector pCRprtNeo, which carries a neomycin resistant gene but lacks an origin of replication for M. maripaludis (Moore and Leigh 2005). DNA sequencing confirmed that all gene deletions in the pCRprtNeo vectors were in frame. These recombinant plasmids were then used for the creation of the specific in-frame gene deletions in M. maripaludis.

Generation of an in-frame deletion of targeted genes in M. maripaludis

Markerless in-frame deletions of mmp1090, mmp1091 and mmp1092 were successfully generated using the procedure described previously (Moore and Leigh 2005). Several attempts to create in-frame deletions of mmp1089, mmp1093 and mmp1094 were also conducted using the same method, but all were unsuccessful. Single transformant colonies picked from the McCas-Noble agar plates containing 8-aza-hypoxanthine were inoculated into Balch medium III and then screened by using washed whole cells in PCR reactions with primers (listed in Table 2) that would amplify across the target gene to identify deletion mutants. The PCR products were examined by agarose gel electrophoresis and the sizes were compared to that predicted for the WT and deletion versions of the gene in order to identify the specific gene deletion mutants. Deletion mutants were re-streaked for purity on Balch medium III agar plates and the presence of the desired gene deletion reconfirmed by PCR.

Complementation of the Δmmp1090 deletion strain

mmp1090 was amplified by PCR using Mm900 genomic DNA as template and primers listed in Table 2. The forward primer had an NsiI site incorporated while the reverse primer had a MluI site introduced. The PCR product was digested with NsiI and MluI, and ligated into NsiI/MluI digested vector pHW40, putting transcription of the cloned gene under the control of the regulatable nif promoter (Lie et al. 2005). The fidelity of the gene in the complementation plasmid was confirmed by DNA sequencing. This complementation plasmid, designated pKJ1189, was then transformed into Δmmp1090 mutant cells via the PEG-based methodology (Tumbula et al. 1994). Following recovery overnight, the cells were then sub-cultured in the presence of 2.5 µg/mL puromycin for plasmid selection. Transformant cells were subsequently grown in nitrogen-free medium supplemented with either 10 mM l-alanine (nif promoter is induced) or 10 mM NH4Cl (nif promoter is repressed) in the presence of 2.5 µg/mL puromycin. Complementation was also performed in Δmmp1090 mutant cells with the same vector backbone containing mmp1090 carrying a mutation at either Y151 (Y155A) or K155 (K155A). These mutations were created using site-directed mutagenesis (SDM) and primers listed in Table 2, essentially as reported previously (Bardy and Jarrell 2003).

Western blot analysis

Whole cell lysates from the various M. maripaludis strains were separated by 15 % SDS-PAGE and transferred onto an Immobilon-P membrane (Millipore Inc.) (Towbin et al. 1979). Chicken anti-FlaB2 specific antibodies were used as primary antibody to recognise the major archaellin FlaB2, which has been routinely used as the reporter protein for N-glycosylation in our previous studies (Ding et al. 2013; Jones et al. 2012). Horseradish peroxidase-conjugated rabbit anti-chicken immunoglobulin Y (Jackson Immuno Research Laboratories, West Grove PA) was used as secondary antibody, and the blot was developed using Immobilon Western Chemiluminescent HRP Substrate (Millipore Canada Inc., Etobicoke ON).

Archaella purification

Archaella from the Mm900 (WT) strain as well as the Δmmp1090, Δmmp1091 and Δmmp1092 deletion strains were isolated as previously described (Bardy et al. 2002).

Mass spectrometry analysis of archaellin N-glycan in deletion strains

Each archaella sample (40 µg) was incubated overnight at 37 °C with trypsin (Promega, Madison, WI) at an approximate ratio of 20:1 (protein:enzyme, wt/wt) in 50 mM ammonium bicarbonate. The digests were then analysed by nano-liquid chromatography-tandem mass spectrometry (Nano-LC–MS/MS) using a NanoAquity UPLC system (Waters, Milford, MA) coupled to an Ultima hybrid quadrupole time-of-flight (QTOF) mass spectrometer (Waters). The digests were injected onto an Acclaim PepMax100 C18 µ-precolumn (5 mm by 300 µm i.d.; Dionex/Thermo Scientific, Sunnyvale CA) and resolved on a 1.7- µm BEH130 C18 column (100 µm by 100 mm i.d.; Waters, Milford, CA) using the following gradient conditions: 1–45 % ACN, 0.1 % formic acid in 36 min and 45–95 % ACN, 0.1 % formic acid in 2 min. The flow rate was 400 nL/min. MS/MS spectra were acquired on doubly, triply and quadruply charged ions and searched against the NCBInr database using the Mascot search engine (Matrix Science, Ltd., London, United Kingdom). The spectral datasets were searched for glycopeptide MS/MS spectra which were then interpreted by hand.

Electron microscopy

M. maripaludis cells from an overnight culture were briefly washed with 2 % NaCl and resuspended in phosphate-buffered saline. Cells were then loaded onto carbon-Formvar-coated copper grids and stained with 2 % phosphotungstic acid. Grids were examined in a Hitachi 7000 electron microscope operating at an accelerating voltage of 75 kV.

Results

Numerous genes immediately adjacent to the gene cluster encompassed by mmp1089 to mmp1094 have been previously shown to be involved in the N-glycosylation or N-glycan sugar biosynthesis pathways in M. maripaludis (Ding et al. 2013; Jones et al. 2012, Vandyke et al. 2009). This led us to examine the possible involvement of mmp1089 to mmp1094 in the N-glycosylation pathway, first by examination of their annotations. The annotation of MMP1089 is as a polysaccharide biosynthesis protein containing RfbX or MATE (multidrug and toxic compound extrusion) Wzx-like domains often found in polysaccharide export proteins and flippases, MMP1090 as an UDP-glucose 4-epimerase (also named as UDP-galactose 4-epimerase or GalE), MMP1091 as an ADP-glucose pyrophosphorylase, MMP1092 as an auxin efflux carrier, MMP1093 (CoaD) as a 4′-phosphopantetheine adenylyltransferase, and MMP1094 (PpsA) as a phosphoenolpyruvate synthase (Table 3). In addition, most of the proteins encoded by mmp1089–mmp1094 have high sequence identity/similarity over almost their entire length to well-studied proteins, as listed in Table 3. While the annotations and bioinformatic analyses of MMP1093 and MMP1094 indicate that these proteins are likely involved in central metabolic pathways, the annotations of MMP1089–MMP1091 suggest that they all may be involved either directly in N-glycosylation or in sugar biosynthesis pathways that may be related to N-glycan synthesis.

Table 3 Annotations of MMP1089–1094
Full size table

RT-PCR analysis of the co-transcription of mmp1094 to mmp1089

mmp1094 to mmp1089 are orientated in the same direction on the complementary strand in the M. maripaludis genome, and opposite from the direction of the neighbouring genes aglL (mmp1088) and mmp1095, as shown in Fig. 1a. This six-gene cluster starts with mmp1094 and relatively short intergenic regions separate the neighbouring genes: 71 bp between mmp1094 and mmp1093, 73 bp between mmp1092 and mmp1091, 104 bp between mmp1091 and mmp1090, and 25 bp between mmp1090 and mmp1089. Interestingly, mmp1093 and mmp1092 are predicted to share a 4 bp overlap. These observations suggest that the six genes might be transcriptionally linked and this was examined by RT-PCR experiments. Using RNA extracted from Mm900 cells as template and subjected to a reverse transcriptase step (RT lanes), amplification products were obtained using primers that could amplify across the intergenic regions linking mmp1094 to mmp1093 (predicted size of amplicon is 442 bp), mmp1093 to mmp1092 (541 bp), mmp1092 to mmp1091 (628 bp), mmp1091 to mmp1090 (529 bp) and mmp1090 to mmp1089 (533 bp) (Fig. 1b). No amplification products were obtained from PCR reactions using the same purified RNA as template but omitting the reverse transcriptase step (RNA lanes), indicating that the purified RNA used for RT-PCR was not contaminated with genomic DNA. Using the same primer pairs and genomic DNA as template, PCR products (DNA lanes) were obtained to confirm the amplicon size and primer specificity. All amplicons were of the size predicted from the location of the primers. These results indicate that mmp1094 to mmp1089 form an operon.

Fig. 1
figure 1

mmp1094 to mmp1089 are an operon. a. Genomic regions between aglL and mmp1095 that were targeted for RT-PCR. Black lines below show the anticipated amplicons obtained from RT-PCR. b. RT-PCR confirmation of the co-transcription of mmp1089–1094. Amplicons were obtained from RT-PCR using primers amplifying intergenic regions between mmp1089–1090, mmp1090–1091, mmp1091–1092, mmp1092–1093 and mmp1093–1094. No products were obtained from reactions using the same primer pairs and RNA that had not been subjected to the reverse transcription step as template (RNA lanes), indicating that the RNA template used for RT-PCR was not contaminated with genomic DNA. PCR using the same primer pairs and Mm900 genomic DNA as template (DNA lanes) was conducted to show the amplicon size and primer specificity

Full size image

Generation of Δmmp1090, Δmmp1091 and Δmmp1092 mutants

In order to determine the possible involvement of mmp1089–1094 in the N-glycosylation pathway, each of the six genes was targeted for in-frame deletion. Mutants carrying deletions in each of mmp1090, mmp1091 and mmp1092 mutants were obtained (Fig. 2). In each case, PCR products, obtained using primers that would amplify across the targeted gene and with mutant cells as template, were smaller than those obtained when Mm900 genomic DNA was used as template, indicating the successful deletion of the corresponding gene. PCR products from each mutant were also sequenced to confirm that each deletion was in-frame. Several attempts to delete mmp1093 and mmp1094 were also performed but screening of over 100 transformants in different experiments by PCR failed to identify any potential deletion mutants. Attempts to delete mmp1089, examined as a possible flippase in previous studies, were also unsuccessful (Vandyke et al. 2009). The failure to obtain mutants in any of these three genes suggests that they might be essential for the survival of M. maripaludis under the laboratory culture conditions used.

Fig. 2
figure 2

PCR confirmation of the in-frame deletion of mmp1090, mmp1091 and mmp1092. Sequencing primer pairs amplifying across the deletion area of mmp1090, mmp1091 or mmp1092 were used in PCR with either Mm900 genomic DNA (WT lanes) or corresponding washed deletion mutant cells as template. In all cases, the sizes of the amplicons were as predicted

Full size image

Western blot analysis of archaellin FlaB2 from Δmmp1090, Δmmp1091 and Δmmp1092 mutants

To test if the in-frame deletion of mmp1090, mmp1091 or mmp1092 resulted in a detectable truncation in the archaellin N-glycan, whole cell lysates of each mutant were first subjected to Western blot analysis, using anti-FlaB2 antibodies, as it has been shown previously that even small truncations of the N-glycan results in a faster migration of FlaB2 that is detectable on Western blots (Ding et al. 2013; Jones et al. 2012; Siu et al. 2015; Vandyke et al. 2009). As shown in Fig. 3, FlaB2 from the Δmmp1090 migrated faster than that from WT cells, indicating a possible truncation in its N-glycan. Using whole cell lysates from mutants deleted for ΔaglO, ΔaglA and ΔaglL as an indication of FlaB2 electrophoretic mobility corresponding to attached N-glycans missing three, two or one sugar(s), respectively, it was predicted that FlaB2 from mmp1090 deleted cells would have a glycan lacking the terminal sugar. No reduction of FlaB2 apparent molecular weight was observed in strains deleted for either mmp1091 or mmp1092 in Western blots.

Fig. 3
figure 3

Western blot analysis of FlaB2 from Δmmp1090, Δmmp1091 and Δmmp1092 mutants. Whole cell lysates of Mm900 (WT), as well as the mutants ΔaglO, ΔaglA and ΔaglL (missing the 2nd, 3rd, or 4th GT, respectively) were included for comparison to the Δmmp1090, Δmmp1091 and Δmmp1092 mutants. The WT, ΔaglO, ΔaglA and ΔaglL strains synthesize FlaB2 with a tetra-, mono-, di-, or tri-saccharide N-glycan, respectively. The electrophoretic mobility of FlaB2 from Δmmp1090 was the same as that from ΔaglL, indicating that FlaB2 from the Δmmp1090 mutant was likely modified with the truncated trisaccharide reported for the ΔaglL mutant. The electrophoretic mobilities of FlaB2 from the Δmmp1091 and Δmmp1092 mutants could not be distinguished from that of the WT cells

Full size image

In-frame deletion of mmp1090, mmp1091 or mmp1092 does not interfere with archaella assembly

In M. maripaludis, a minimum length disaccharide glycan attached to archaellins is required for archaella formation (Vandyke et al. 2009). When Δmmp1090, Δmmp1091 and Δmmp1092 mutants were examined for archaella formation by electron microscopy, all mutants were found to be archaellated (Fig. 4). This is in agreement with the Western blot results suggesting a WT-sized glycan in the Δmmp1091 and Δmmp1092 mutants and a three-sugar glycan for the Δmmp1090 mutant.

Fig. 4
figure 4

Electron micrographs of WT cells and the Δmmp1090, Δmmp1091 and Δmmp1092 mutants. Archaella were observed on the cell surface of the WT (WT) and all three mutants. Bars, 500 nm

Full size image

Mass spectrometry analysis of N-glycan structures from Δmmp1090, Δmmp1091 and Δmmp1092 mutants

To specifically identify the N-glycan structure in each mutant, archaella were isolated from the ∆mmp1090, ∆mmp1091 and ∆mmp1092 cells and the attached glycan structure was determined by mass spectrometry (Fig. 5). The structure of the tetrameric WT glycan (Fig. 5a) has been described previously (Kelly et al. 2009). The glycopeptides from the Δmmp1090 mutant (Fig. 5b) are modified with a trisaccharide composed of the linking GalNAc, the di-N-acetyl glucuronic acid (GlcNAc3NAcA) and a third sugar (ManNAc3NAmA) that lacks the threonine modification as well as the fourth sugar residue observed on WT glycan. The archaellin glycopeptides from the Δmmp1091 mutant are modified almost exclusively (>95 %) with WT glycan (Fig. 5c). The small differences observed (<5 %) could be attributed to side peaks/adducts and these were observed in both parent and mutant samples analysed. This was also true of the glycan structure in the archaellins of the Δmmp1092 mutant (data not shown).

Fig. 5
figure 5

NanoLC-MS/MS analysis of the FlaB2 tryptic glycopeptide, T53−81. The tryptic digests of archaellin isolated from a WT as well as b Δmmp1090 and c Δmmp1091 mutant strains of M. maripaludis were analyzed by nanoLC-MS/MS on a Nanoaquity UPLC system (Waters) coupled to a Q-TOF Ultima mass spectrometer (Waters). The triply protonated glycopeptide ion (MH3 3+) was selected for MS/MS analysis in each case. The FlaB2 tryptic peptide T53−81 contains one site of N-glycosylation. The sequence of this glycopeptide is provided in both panels b and c in order to illustrate the difference in the nature of the glycan modification observed in the two mutants. The major carbohydrate oxonium ions are identified in the MS/MS spectra using symbols to indicate the sugar residues present. The symbols are identified in the inset in panel a. The major b and y ions arising from fragmentation of the peptide bonds are also shown

Full size image

Complementation of the Δmmp1090 strain with WT and mutant versions of mmp1090

To examine if the in-frame deletion of mmp1090 was the sole contributor to the defect in N-glycan length observed by mass spectrometry, the Δmmp1090 mutant was complemented with a plasmid bearing a WT copy of mmp1090 under the control of the nif promoter. Complemented cells were cultured in nitrogen-free medium supplemented with 10 mM of either l-alanine (nif promoter is induced) or NH4Cl (nif promoter is repressed) as the sole nitrogen source. FlaB2 from complemented cells cultured in both media was analysed by Western blotting (Fig. 6). Compared to FlaB2 from the Δmmp1090 mutant cells, the apparent molecular weight of FlaB2 from Δmmp1090-complemented cells under alanine growth conditions was restored to that of the WT cells. Under NH4Cl growth conditions, the electrophoretic mobility of FlaB2 in the Δmmp1090-complemented cells was also restored to that from WT cells. Similar results have been observed in several previous studies using this vector system (Ding et al. 2013, Jones et al. 2012), a result which we attribute to a small amount of transcription that can occur from the nif promoter even in the presence of NH4Cl (Lie et al. 2005). This basal expression under NH4Cl growth conditions may lead to the synthesis of enough MMP1090 to complement the defect in the deletion strain.

Fig. 6
figure 6

Western blot analysis of FlaB2 in lysates of the ∆mmp1090 mutant complemented in trans with WT or mutant versions of mmp1090. Both WT mmp1090 (1090) and two mmp1090 mutants, Y151A and K155A, were cloned in the shuttle vector pHW40 under an inducible nif promoter and transformed into Δmmp1090 mutant cells. Complementation cells were cultured in nitrogen-free medium supplemented with either l-alanine (Ala) (promoter on) or NH4Cl (NH4 +) conditions (promoter off). The left triplet of lanes shows that the complementation of the Δmmp1090 mutant cells with the WT copy of mmp1090 supplied in trans (Δmmp1090 comp Ala) restored the FlaB2 apparent molecular weight to the WT size. The right group of lanes shows the apparent molecular weight of FlaB2 in the Δmmp1090 cells complemented with the mutant versions of mmp1090 was not returned to the size of FlaB2 found in WT cells

Full size image

UDP-glucose 4-epimerase belongs to the Short-chain Dehydrogenases/Reductases (SDR) superfamily whose members typically contain a conserved YxxxK motif important for catalysis (Jörnvall et al. 1995). Two mutant versions of MMP1090, Y151A and K155A, where the Y151 and K155 of the conserved YxxxK motif were changed to alanine, were also used to complement the mmp1090 deletion strain but neither could return the FlaB2 to WT size as determined by Western blots (Fig. 6), suggesting that MMP1090 lost its function due to the individual point mutation.

Discussion

In this study, we have investigated the possible roles of a six-gene operon (mmp1089–1094) on the formation of the tetrasaccharide N-linked to archaellins. While deletions of mmp1090, mmp1091 and mmp1092 were obtained and studied, attempts to delete the remaining three genes were unsuccessful, suggesting that these three genes are essential under our normal growth conditions. This is consistent with the results from a recent genome-wide transposon mutagenesis study which also indicated that mmp1089, mmp1093 and mmp1094 are likely essential (Sarmiento et al. 2013).

We have examined close relatives of M. maripaludis S2 for the presence and order of genes homologous to mmp1089–1094. The complete genome sequences of several strains of M. maripaludis (C5,C6, C7, X1) are available for comparison. In all four of these other strains, genes homologous to mmp1090–1094 are found adjacent to each other and in the same order as in the S2 strain. However, only in strains C7 and X1 is the gene homologous to mmp1089 found adjacent to the mmp1090 homologue. No N-linked glycan structures have been reported in these additional M. maripaludis strains for comparison to that from the S2 strain. Other species of Methanococcus were also examined and a variety of formats were observed with regards to the mmp1089–1094 homologues. Methanococcus vannielii has a complete set of mmp1089–1094 homologues in the same order as in M. maripaludis S2. Methanococcus aeolicus has the mmp1090–1091 homologues adjacent but the other genes are located around the genome. Methanococcus voltae is unusual in not having the mmp1090–1091 homologues adjacent to each other but the mmp1092–1094 homologs are clustered.

The annotations of mmp1093 and mmp1094 suggest that both gene products are likely to be involved in intermediary metabolism and not N-glycosylation (Table 3). MMP1093 is annotated as phosphopantetheine adenylyltransferase (PPAT, encoded by coaD), which catalyses the penultimate step of the CoA biosynthesis pathway, i.e. the reversible adenylation of 4′-phosphopantetheine to generate 3′-dephospho-CoA (Geerlof et al. 1999). CoA is an essential cofactor for many enzymatic reactions and genes encoding the last 4 steps of the CoA biosynthesis pathway, sequentially coaB, coaC, coaD and coaE (sometimes coaB and coaC are fused into one gene coaBC encoding a bifunctional protein) are found throughout archaeal genomes, indicating the last 4 steps of the CoA biosynthesis pathway are conserved among all the three domains of life (Genschel 2004; Kupke and Schwarz 2006). coaBC (mmp1606), coaD (mmp1093) and coaE (mmp1282) are not grouped in the M. maripaludis S2 genome and all are essential according to a recent transposon mutagenesis study (Sarmiento et al. 2013). MMP1094 is annotated as a phosphoenolpyruvate synthase (PPS) or pyruvate, water dikinase (EC number 2.7.9.2), encoded by ppsA, which catalyses the conversion of pyruvate to phosphoenolpyruvate. PPS activity was previously detected in crude cell extracts of M. maripaludis S2, although experiments that formally confirm that this activity is due to the product of mmp1094 are lacking.

MMP1089 is annotated as a polysaccharide synthesis protein. According to BLAST (Basic Local Alignment Search Tool), it belongs to the MATE-like superfamily with a RfbX (Wzx) domain (Table 3). Wzx is considered to be the translocase (flippase) in the Wzy (polymerase)-dependent O-antigen biosynthesis pathway, flipping the isoprenoid lipid-linked O-antigenic unit across the cytoplasmic membrane in Gram-negative bacteria (Islam and Lam 2013; Liu et al. 1996). MMP1089 is an integral membrane protein and predicted to contain 14 transmembrane helices (TMHMM, http://www.cbs.dtu.dk/services/TMHMM/; Sonnhammer et al. 1998), sharing similar protein topology with various Wzx proteins from Salmonella enterica, E. coli and Pseudomonas aeruginosa (Cunneen and Reeves 2008; Islam et al. 2010; Marolda et al. 2010). MMP1089 also shares 26 % identity (Table 3) with AglR from Hfx. volcanii, the only putative flippase so far identified in archaeal N-glycosylation pathways. AglR is thought to be the enzyme that flips Dol-P-mannose across the cytoplasmic membrane; the mannose is subsequently transferred onto the protein-bound N-glycan as the terminal sugar of the pentasaccharide that is N-linked to the S-layer protein (Eichler 2013; Kaminski et al. 2012). Both its annotation and its genomic location immediately adjacent to sugar biosynthetic and GT genes known to be involved in N-glycosylation suggest a role for MMP1089 in N-glycosylation. In our previous studies, mmp1089 had been targeted for deletion in the belief that it may encode the flippase involved in the N-linked glycosylation pathway, but none of these attempts were successful (Vandyke et al. 2009). The subsequent transposon mutagenesis study from the Whitman group also suggests the gene is essential (Sarmiento et al. 2013). This is somewhat puzzling since the N-glycosylation pathway is not essential in M. maripaludis as evidenced by the deletion of aglB, encoding the OST which catalyses the critical terminal step in the pathway (Vandyke et al. 2009). It may be that deletion of the flippase leads to a sequestering of the dolichol carrier that prevents its turnover and use in an essential pathway, as suggested for the putative Wzx-like flippase involved in capsule formation in Streptococcus pneumoniae (Xayarath and Yother 2007). Otherwise, MMP1089 may be involved in other, presumably essential, processes in addition to its predicted role as a flippase in N-glycosylation.

Successful deletion of the remaining three genes of the operon (mmp1090, mmp1091 and mmp1092) was accomplished, although only one had a demonstrable involvement in N-glycosylation. Inactivation of mmp1090 resulted in a truncated archaellin N-glycan missing the terminal sugar residue and the threonine attached to the third sugar residue, the same archaellin glycan structure found in a mutant deleted for the 4th GT, AglL (Vandyke et al. 2009). As it is known that the threonine residue is transferred onto the third sugar by the threonine transferase AglU only after AglL transfers the terminal sugar to the glycan being assembled on the Dol-P carrier (Ding et al. 2013; Jarrell et al. 2014), MMP1090 is most likely to be involved in the biosynthesis of the unique terminal sugar.

MMP1090 is annotated as a UDP-glucose 4-epimerase (Table 3). BLAST searches reveal numerous proteins, described as UDP-glucose-4-epimerases, in both Archaea and Bacteria that have very high sequence identity to MMP1090 over essentially the entire length of the protein. Among these homologues, MMP1090 shares 33 and 32 % amino acid identity with two studied archaeal UDP-glucose 4-epimerases, from the hyperthermophiles Pyrobaculum calidifontis (Sakuraba et al. 2011) and Pyrococcus horikoshii (Chung et al. 2012). In support of the annotation, several signature motifs or amino acid residues possessed by UDP-glucose 4-epimerases are also found in MMP1090, including an YxxxK motif and a glycine-rich motif (Kallberg et al. 2002; Persson and Kallberg 2013). The tyrosine and lysine in the YxxxK motif, together with a conserved upstream serine residue, establish a catalytic S-Y-K triad (Jörnvall et al. 1995; Oppermann et al. 2003). The glycine rich motif, which is located in the N-terminus of the protein, is important for the binding of NAD+ as cofactor (Jörnvall et al. 1995). UDP-glucose 4-epimerases are members of the extended family of SDRs and MMP1090 possesses a perfect match (TGGAGFIGSHIVDMLIENGHDV) to the conserved glycine rich motif [ST]Gx2G[FMQY][DILV]Gx6[FILMV][ILMV]x2Gx2[ILV] of this family (Kallberg et al. 2002).

Based on their substrate specificity, the UDP-glucose 4-epimerase family has been subdivided into three groups (Ishiyama et al. 2004). In the first group, UDP-glucose 4-epimerase catalyses the interconversion of UDP-glucose and UDP-galactose, e.g. GalE from E. coli (eGalE hereafter). A second group of UDP-glucose 4-epimerase, e.g. human UDP-glucose 4-epimerase (hGalE hereafter), catalyses not only the interconversion of UDP-glucose and UDP-galactose, but also the interconversion of UDP-N-acetyl-glucosamine (UDP-GlcNAc) and UDP-N-acetyl-galactosamine (UDP-GalNAc). The third group, e.g. WbpP from P. aeruginosa, preferably catalyses the interconversion of acetylated UDP-hexoses (UDP-GlcNAc and UDP-GalNAc). After comparing protein structures from the three groups of UDP-glucose 4-epimerase, Ishiyama et al. (Ishiyama et al. 2004) proposed that the differences in the catalytic pockets lead to the different substrate specificities. The catalytic pocket from hGalE is ~15 % larger than that from eGalE due to the smaller side chain from N207 and C307 in hGalE compared with that from the corresponding N198 and Y299 in eGalE; thus the steric hindrance prevents the latter enzyme from catalysing the acetylated UDP-hexoses. The two corresponding amino acids from WbpP, A208 and S306, also lead to a larger catalytic pocket. Furthermore, the existence of ordered solvent molecules in the WbpP catalytic pocket results in the preference for acetylated UDP-hexoses as substrates. In MMP1090, BLAST analysis with eGalE, hGalE and WbpP shows the corresponding amino acids are G193 and I281. The short side chains from these two amino acids presumably result in a relatively larger catalytic pocket, indicating that MMP1090 might belong to either group 2 or group 3 UDP-glucose 4-epimerase.

We initially hypothesised that M. maripaludis would need a UDP-GlcNAc 4-epimerase to produce UDP-GalNAc from UDP-GlcNAc. UDP-GalNAc is likely the substrate for the first GT to begin the assembly of the tetrasaccharide on the dolichol phosphate carrier, since the linking sugar in the archaellin N-glycan is GalNAc. There is no gene annotated as a UDP-GlcNAc 4-epimerase in the M. maripaludis genome sequence but there are examples of UDP-glucose 4-epimerases also having UDP-GlcNAc 4-epimerase activity, such as the bifunctional GalE involved in the synthesis of the N-linked heptasaccharide of Campylobacter jejuni, which contains 5 GalNAc residues (Bernatchez et al. 2005). If MMP1090 has this activity and it is necessary for the production of UDP-GalNAc, then its deletion should result in completely non-glycosylated archaellins, migrating with the same apparent molecular weight as those from the aglB (oligosaccharyltransferase) mutant. Clearly this was not the case and the MMP1090 role appears to be in the synthesis of the terminal sugar, even though our examination of the pocket size of MMP1090 suggested it might be able to utilise acetylated UDP-hexoses as substrates.

Complementation studies demonstrated that the electrophoretic mobility of FlaB2 in Δmmp1090 mutant cells expressing the WT version of mmp1090 in trans was indistinguishable from that of FlaB2 from WT cells in Western blots, indicating that MMP1090 is the sole contributor to the loss of the terminal sugar residue of the archaellin N-glycan (the missing threonine is transferred onto the N-glycan precursor afterwards (Ding et al. 2013; Vandyke et al. 2009)). In addition, two mutant versions of mmp1090 were created which resulted in a change of one amino acid, either Y151 or K155, located in the conserved YxxxK motif, to alanine. The corresponding amino acids in hGalE, Y157 and K161, are two of the key amino acids anchoring the cofactor NAD+ within the enzyme. In addition, Y157 also serves as the active site base by directly interacting with the C4 hydroxyl group in the glucosyl moiety (Thoden et al. 2000). In agreement with our hypothesis that Y151 and K155 are key amino acids in MMP1090, neither of the two mutant versions of MMP1090 could restore the FlaB2 of the mmp1090 deletion strain to WT size in complementation studies (Fig. 6). These findings are consistent with MMP1090 being a UDP-glucose 4-epimerase involved in the terminal sugar biosynthesis pathway of the archaellin N-glycan.

No detectable phenotype related to archaellin glycosylation was observed in either the Δmmp1091 or Δmmp1092 deletion mutants, as the electrophoretic mobility of FlaB2 in Western blots, the archaellin N-glycan structure determined by mass spectrometry and the archaellation state determined by electron microscopy all appeared identical to those observed in WT cells. The annotation of MMP1091 indicates that it is an ADP-glucose pyrophosphorylase (ADPG-PPase), while BLAST results show that it is a putative UDP-glucose pyrophosphorylase (UDPG-PPase). ADPG-PPase and UDPG-PPase catalyse similar reactions, i.e. the synthesis of ADP-glucose or UDP-glucose from glucose-1-phosphate and either ATP (ADPG-PPase) or UTP (UDPG-PPase), respectively. ADP-glucose and UDP-glucose can be used as glucosyl donors by glycogen or starch synthases in the biosynthesis of glycogen and starch, which are common energy storage forms in the three domains of life (Henrissat et al. 2002). Archaeal glycogen synthases can use both ADP-glucose and UDP-glucose as donor substrates (Gruyer et al. 2002; Horcajada et al. 2006). In M. maripaludis, glycogen was reported to comprise 0.34 % of the cell dry weight in the early stationary phase (Yu et al. 1994). A BLAST search of the M. maripaludis S2 genome using MMP1091 as query revealed a second protein, MMP1076, as 37 % identical with a query cover of 91 % and an E value of 1e-26. This protein is annotated as a glucosamine-1-phosphate N-acetyltransferase. It is a homologue of the E. coli enzyme GlmU, which is a bifunctional enzyme that catalyses the acetylation of glucosamine 1-phosphate and the subsequent transfer of the sugar to UTP. The purified Methanocaldococcus jannaschii homologue, MJ1101, also shows both of these activities (Namboori and Graham 2008b) while other archaeal homologues, such as from Sulfolobus tokodaii, lack the acetyltransferase activity (Zhang et al. 2005). Interestingly, MJ1101 as well the homologues from S. tokodaii and P. furiosus, could all utilise glucose-1-phosphate as an additional substrate to generate UDP-glucose (Mizanur et al. 2004; Namboori and Graham 2008b, Zhang et al. 2005). Given that MJ1101 is 68 % identical to MMP1076, it seems likely that the M. maripaludis enzyme may also be able to activate glucose-1-phosphate and thus compensate for the loss of MMP1091, although this has not been tested (Namboori and Graham 2008b). Although no archaellin N-glycan deficiency was observed in the Δmmp1091 mutant, it is premature to conclude that MMP1091 is not involved in the N-glycosylation pathway considering its annotation and its genomic position in an operon adjacent to mmp1090, as well as the possibility of the lost activity being compensated by MMP1076. In the case of H. volcanii, formation of the hexuronic acid found at position 3 of the pentasaccharide N-linked to the S-layer protein does depend on the activity of a glucose-1-phosphate uridylyltransferase (AglF), which acts in concert with the dehydrogenase AglM (Yurist-Doutsch et al. 2008).

Deletion of mmp1092 also did not have a detectable effect on the archaellin N-glycan. MMP1092 is predicted to be an integral membrane protein containing 10 transmembrane helices, according to bioinformatics tools such as TMHMM and PSORTb (http://www.psort.org/psortb/). While MMP1092 is annotated as an auxin efflux protein, members of this family have also been found in both bacteria and archaea. In bacteria, there are reports of these homologues being malate transporters (TCDB: Transport Classification DataBase; http://www.tcdb.org/). Unlike the case with MMP1091, there are no other proteins encoded by M. maripaludis S2 with high sequence identity to MMP1092 over a large portion of the protein. Since no detectable phenotype was observed in the Δmmp1092 mutant, further work is needed to identify the function of MMP1092 but at present no role in N-glycosylation can be assigned to it.

The mmp1089–mmp1094 operon appears to be the end of the large number of genes involved in N-glycosylation that are found in this region. Examination of the annotations of genes mmp1095 and downstream display a number of genes predicted to be involved in ion transport and other functions seemingly unrelated to N-glycosylation.

In this study, mmp1090, likely encoding a UDP-glucose 4-epimerase, was identified by genetic and mass spectrometry techniques to be involved in the biosynthesis pathway of the terminal sugar residue in the archaellin N-glycan. The terminal sugar in the archaellin N-linked tetrasaccharide, with the structure of (5S)-2-acetamido-2,4-dideoxy-5-O-methyl-α-L-erythro-hexos-5-ulo-1,5-pyranose, is a unique sugar so far found exclusively in M. maripaludis (Kelly et al. 2009). The knowledge of the biosynthesis of this dialdose is very limited and only one protein, AglV, the enzyme responsible for the transfer of the methyl group onto C5 via O-linkage, has been genetically identified in its biosynthesis pathway (Ding et al. 2013). This is in contrast to the second and third sugars of the glycan where the biosynthesis pathways are now well known (Ding et al. 2013; Jones et al. 2012; Siu et al. 2015, Vandyke et al. 2009). The involvement of MMP1090 as a UDP-glucose 4-epimerase in the biosynthesis pathway of the terminal sugar provides evidence that this unique sugar is probably synthesised from UDP-glucose or UDP-GlcNAc. We are currently attempting heterologous overexpression and purification of the active enzyme in E. coli to identify the substrate specificity of this enzyme to help in the elucidation of its role in the biosynthetic pathway of this unusual sugar. Based on this demonstrated involvement in the N-glycosylation pathway, mmp1090 is here designated as aglW, in keeping with the nomenclature scheme for genes involved in archaeal N-glycosylation (Chaban et al. 2006; Eichler et al. 2013).