C. jejuni strains were grown on NZCYM (Difco) agar, supplemented with 50 μg/mL kanamycin where needed, at 37°C under microaerobic conditions (85% N₂, 10% CO₂, 5% O₂). E. coli was grown on LB agar supplemented with 50 μg/mL kanamycin where needed. The list and sources of bacterial strains used in this study are given in Table 1.

FlaGrab was expressed as an N-terminal glutathione-S-transferase (GST) fusion protein as described previously (Kropinski et al., 2011). CC-FlaGrab [the C-terminal quarter of FlaGrab, previously shown to harbor the growth inhibitory activity (Javed et al., 2015b)] and NC-FlaGrab [the N-terminal quarter of FlaGrab, previously shown to retain no growth inhibitory activity (Javed et al., 2015b)] were expressed in E. coli BL21 as GST-fused proteins and purified as described previously (Javed et al., 2013), with the exception that proteins were eluted in 10 mM reduced L-glutathione, 50 mM HEPES, and 1 mM DTT at pH 9.0. The list of plasmids used in this study are listed in Table 1.

Cells were harvested from overnight NZCYM plate cultures, pelleted and washed once in NZCYM broth and set to an OD₆₀₀ of 0.05 (2 × 10⁸ CFU/mL) in 20 mL NZCYM broth in 125-mL Erlenmeyer flasks. Cells were grown under microaerobic conditions with agitation at 200 rpm. After 4.5 h incubation, CC-FlaGrab was added to a final concentration of 25 μg/mL. As negative controls, 6 mL HEPES buffer or 6 mL NC-FlaGrab were used. Flasks were incubated 30 min before the entire contents of each was harvested and the RNA stabilized using 0.1 volumes of ice cold 10% buffered phenol in 100% ethanol (Palyada et al., 2004). Total RNA was extracted from each sample using a hot phenol method as previously described (Palyada et al., 2004). Genomic DNA was removed from the samples using RNAse-free DNAse I (Epicenter) (37°C for 30 min) and cleaned using the Zymo RNA Clean & Concentrator. PCR was used to confirm the absence of residual DNA and the DNase treatment was repeated until the absence of genomic DNA was confirmed. Total RNA quality was assessed using an Agilent Bioanalyzer and RNA was stored at −80°C until further use. This experiment was done in biological triplicate, with the exception of the NC-FlaGrab control, which was done only once.

Samples were subsequently depleted of rRNA using the RiboZero Bacterial kit according to the manufacturer's instructions and rRNA depletion was confirmed using the Agilent Bioanalyzer RNA 6000 Pico Kit. The Ion Total RNA-seq kit was used to construct strand-specific barcoded sequencing libraries according to the manufacturer's instructions. Following library construction, each library was quality-checked and quantified using the Bioanalyzer High Sensitivity DNA kit. Once all libraries were completed, they were pooled together in equimolar amounts and were templated using the Ion PI Hi-Q kit. The templated beads were sequenced on an Ion Torrent Proton using the Ion PI Hi-Q sequencing 200 kit on a single Proton V2 chip.

The raw sequencing reads were demultiplexed by the Ion Torrent suite software and mapped to the 11168 genome using STAR (Love et al., 2014). The raw demultiplexed sequencing reads have been deposited at the NCBI SRA archive under accession number PRJNA580017. Reads aligning to coding regions were counted using HT-seq using the default settings (Dobin et al., 2013). DESeq2 was used to identify differentially expressed transcripts between the control and CC-FlaGrab treated cells. Genes with a false discovery rate (FDR)-corrected p < 0.05 were considered differentially expressed. We also conducted gene set enrichment analysis (GSEA) on C. jejuni Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways using the generally applicable gene set enrichment (GAGE) method (Luo et al., 2009) with a minimum FDR cutoff of <0.1 (Love et al., 2014).

To generate motA and motB mutants, C. jejuni 11168 was transformed by natural transformation with the pDRH3330 (motA) or pDRH3331 (motB) mutation constructs published by Beeby et al. (2016). To generate a cj1295 mutant, C. jejuni 11168 was naturally transformed with the cj1295 mutation construct (p1295) described by Hitchen et al. (2010).

For natural transformation of mutant constructs, an overnight culture of C. jejuni cells was streaked onto BHI containing 2% yeast extract (BHIY), grown for 6 h under the conditions described above, and harvested into PBS. Cells were washed once in PBS, resuspended in 250 μL PBS, spotted (20–30 × 10-μL spots) onto pre-warmed (37°C) BHIY plates and allowed to absorb into agar ~5 min. p1295 DNA (20 μg/mL in water) was spotted (~10-μL spots) atop each spot of cell suspension and allowed to dry ~5 min. Plates were incubated for 12–16 h, then the contents of the plates were harvested and spread or streaked across 4–6 BHI plates containing 50 μg/mL kanamycin. Plates were incubated until colonies appeared (3–5 days). Isolated colonies were patched onto kanamycin-containing BHI plates and grown overnight prior to confirmation of successful mutagenesis by colony PCR (OneTaq, NEB) using gene-specific primers (Hitchen et al., 2010; Beeby et al., 2016). This protocol has been deposited into Protocols.io (dx.doi.org/10.17504/protocols.io.magc2bw).

Immunogold labeling and transmission electron microscopy was done as described previously (Javed et al., 2015a), with some exceptions. Briefly, C. jejuni cells (OD₆₀₀ = 3.0) were harvested from a plate of overnight growth in NZCYM broth, incubated with Formvar-coated copper grids for 45 min, washed 1 × 3 min with PBS, then fixed by incubating in 2.5% paraformaldehyde in PBS for 20 min. Grids were then washed 3 × 3 min with PBS, incubated in blocking solution (PBS containing 5% bovine serum albumin and 0.05% Tween) for 35 min, then in freshly purified GST-tagged FlaGrab (0.2 mg/mL diluted 1:25 in blocking solution) for 45 min. Grids were then washed 3 × 3 min in blocking solution, incubated in a rabbit anti-FlaGrab antibody solution (diluted 1:50 in blocking solution) for 45 min, washed 3 × 3 min as above, and incubated in a solution of goat anti-rabbit IgG (whole molecule) conjugated to 10-nm gold particles (Sigma, diluted 1:50 in blocking solution) for 45 min. Finally, grids were sequentially washed (3 × 3 min each) with blocking solution, PBS, and distilled water, then air-dried on Whatman filter paper overnight. All steps were done at room temperature. Grids were examined using a transmission electron microscope (JEOL JEM1011; JEOL, Inc., Peabody, MA, USA). Images were captured using a high contrast 2k × 2k mid-mount digital camera (Advanced Microscopy Techniques, Corp., Woburn, MA, USA). This protocol has been deposited into Protocols.io (dx.doi.org/10.17504/protocols.io.mv5c686).

Bacterial growth clearance was tested by spotting phage protein onto a freshly inoculated bacterial suspension using the overlay agar method described previously, with some modifications (Javed et al., 2015b). Briefly, overnight bacterial cultures were harvested in NZCYM broth and set to an OD₆₀₀ of 0.35. A 5-mL aliquot of this suspension was transferred to a standard sized empty Petri dish and incubated at 37°C without shaking for 4 h under microaerobic conditions. The suspension was then set to an OD₆₀₀ of 0.5, and 200 μL of this was mixed with 5 mL sterile 0.6% molten NZCYM agar at 45°C. This suspension was poured onto the surface of a pre-warmed (37°C) NZCYM plate containing 1.5% agar. Plates were allowed to solidify for 15 min and then 10 μL UV-sterilized FlaGrab or CC-FlaGrab solution (0.91–1.3 mg/mL in 10 mM reduced L-glutathione, 50 mM Tris, 1 mM DTT, pH 9.0) was spotted onto the agar surface and allowed to completely soak into the agar before inverting the plate and incubating at 37°C under microaerobic conditions. Zones of growth clearance were observed after 18–24 h.

Nucleic acid sequence information for cj1293-cj1342 (the flagellar glycosylation locus in 11168) for strains 11168, 12567, 12660, 12661, and 12664 was obtained from NCBI (Gundogdu et al., 2007; Sacher et al., 2018a,b). Nucleotide and amino acid sequence alignments and phylogenetic trees were generated using Geneious version 8 [http://www.geneious.com (Kearse et al., 2012)]. Alignment figures were generated using T-Coffee (Di Tommaso et al., 2011) and Boxshade (http://sourceforge.net/projects/boxshade/). Black shading indicates the consensus sequence, gray shading indicates similar residues.

Cell-free flagella were prepared for MS as described previously with some modifications (Javed et al., 2015a). Briefly, bacterial growth was harvested from 20 to 40 NZCYM agar plates, washed with PBS and resuspended in 100–150 mL PBS and stored up to 7 days at −80°C. Cells were thawed by vortexing and flagella were sheared from cells using a Polytron homogenizer for 6 × 30 s (allowing 30 s incubation on ice between rounds) at max rpm on ice. Cells were then removed by centrifugation at 8,700 × g for 20 min at 4°C. To collect flagellar filaments, the resulting supernatant was ultracentrifuged at 207,870 × g, 4°C for 1.5 h using 6 × 30-mL tubes (Beckman). Pellets were washed 3 times by resuspending each pellet in 20 mL distilled water and ultracentrifuging again as above, giving a total of 4 × 1.5 h ultracentrifuge runs. After washing, pellets were resuspended and pooled into a total of 1 ml distilled water and incubated at −20°C until use.

First, 100 μg of flagella protein sample was dissolved in 25 μL of digestion buffer [50 mM aqueous ammonium bicarbonate (NH₄HCO₃) buffer]. Then, 25 μL of dithiothreitol (DTT) solution (25 mM) was added to the samples and incubated at 45°C for 45 min. Subsequently, carbamidomethylation was performed by adding 25 μL of iodoacetamide (IAA) solution (90 mM) and incubating at room temperature for 45 min in the dark. The sample was then dialyzed against double distilled water and lyophilized. Flagellin was digested by adding 12.5 μL sequencing-grade trypsin (Promega, 0.4 μg/μL) and incubating at 37°C for 12 h. The digest was desalted by C18 solid phase extraction cartridges and then dried under a speed vacuum. The peptides and glycopeptides were subsequently re-dissolved in 0.1% formic acid in water and stored at −30°C until analysis by nano-LC-MS/MS.

Desalted flagellin digests were analyzed on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Scientific) equipped with a nanospray ion source and connected to a Dionex binary solvent system. Pre-packed nano-LC columns (15 cm in length, 75 μm in internal diameter) filled with 3 μm C18 material were used for chromatographic separation of glycoprotein digests. Precursor ion scanning was performed at 120,000 resolution in an Orbitrap analyzer, and precursors at a time frame of 3 s were selected for subsequent fragmentation using higher energy collision dissociation (HCD) at a normalized collision energy of 28 and collision-induced dissociation (CID) at an energy of 35. The threshold for triggering an MS/MS event was set to 500 counts. The fragment ions were analyzed on an Orbitrap after HCD and CID fragmentation at a resolution of 30,000. Charge state screening was enabled, and precursors with unknown charge state or a charge state of +1 were excluded (positive ion mode). Dynamic exclusion was enabled with an exclusion duration of 10 sec.

The LC-MS/MS spectra of the tryptic digest of proteins were searched against the FlaA and FlaB protein sequence of the respective strains in FASTA format using Byonic 2.3 software with trypsin as the digestion enzyme. Carbamidomethylation of cysteine and oxidation of methionine were selected as variable modifications. Glycan modifications such as Pse5Ac7Ac (m/z = 316.12705), Pse5Ac7Am (m/z = 315.14304), DMGA-Pse5Ac7Ac (m/z = 390.16383), and DMGA-Pse5Ac7Am (m/z = 389.17981) were used as variable modifications on serines/threonines. The LC-MS/MS spectra were analyzed manually for the glycopeptide fragmentation on HCD and CID with the support of Xcalibur software. To identify glycans, the spectra of the glycopeptides were evaluated for glycan neutral loss patterns, oxonium ions, and glycopeptide fragmentations. Unknown derivatives of pseudaminic acid were identified based on the presence of oxonium ions and neutral losses on the spectra of glycopeptides bearing glycosylation sites.

To determine the mechanism of FlaGrab growth inhibition of C. jejuni 11168 cells, we sequenced total mRNA (RNA-seq) from cells incubated with CC-FlaGrab [the C-terminal quarter of FlaGrab, previously shown to possess the growth inhibitory activity (Javed et al., 2015b)] for 30 min and compared their transcriptomic profiles with cells incubated with buffer. As an additional negative control, we extracted RNA from buffer-exposed cells and from cells exposed to NC-FlaGrab (the N-terminal quarter of FlaGrab, previously shown to retain no growth inhibitory activity). PCA plot analysis of all samples showed that the NC-FlaGrab clustered with the buffer controls, suggesting that FlaGrab only impacts gene expression when capable of binding to host targets (Figure S1).

Incubating C. jejuni 11168 with CC-FlaGrab altered the expression of genes encoding energy metabolism enzymes, membrane proteins, and periplasmic proteins (Figure 1 and Table 2). Upregulated genes include omp50, which encodes a major C. jejuni porin and phosphotyrosine kinase also known as Cjtk (Corcionivoschi et al., 2012). Other upregulated genes include cj1169c (a periplasmic protein-encoding gene upstream of omp50); cj0037c (cccC), which encodes cytochrome C (Guccione et al., 2017); dsbI, which encodes a disulfide bond forming protein; tlp4, which encodes a chemotaxis receptor; metB, which encodes a methionine biosynthesis enzyme; and cj0761, which encodes a hypothetical protein. Downregulated genes include sdhAB (mrfAB), which are misannotated as succinate dehydrogenases but are likely to encode methylmenaquinol:fumarate reductases, which function in the tricarboxylic acid (TCA) cycle (Weingarten et al., 2009; Kassem et al., 2014; Guccione et al., 2017). Other downregulated genes include cj0426, which encodes an ABC transporter ATP binding protein, the downstream gene cj0427, rpmF, which encodes a 50s ribosomal protein, and cj0911 and cj1485c, which encode hypothetical periplasmic proteins. KEGG pathway analysis showed that several categories were enriched in the downregulated genes upon CC-FlaGrab exposure, including the TCA cycle, carbon metabolism, and oxidative phosphorylation pathways (Table 3).

We next sought to understand how FlaGrab binding to bacterial flagella might lead to the observed effects on cell physiology. The proton motive force is hypothesized to power C. jejuni flagellar rotation through the MotA and MotB stator proteins, as in many other motile bacteria (Blair and Berg, 1990; Hosking et al., 2006; Morimoto et al., 2010; Kojima et al., 2018). Since flagellar motility is reduced upon FlaGrab binding (Javed et al., 2015a), we hypothesized that the cell might sense increased FlaGrab-induced flagellar stiffness and increase proton flow through the flagellar motor channel to compensate for this inhibition of flagellar rotation. This increased proton flow might in turn disrupt the electron transport chain (as proposed by Flint et al., 2014), which could prompt cells to respond by altering transcription of energy metabolism pathways. We predicted that if this were true, flagellar motility mutants, such as ΔmotA and ΔmotB, which do not express functional flagellar motors, would resist FlaGrab-induced clearing (Beeby et al., 2016; Ribardo et al., 2019). To test this, we generated insertional mutants in each of ΔmotA and ΔmotB in C. jejuni 11168 and confirmed that these mutants were non-motile (Figure S2). We then tested their binding and susceptibility to FlaGrab. We used immunogold labeling and transmission electron microscopy to show that binding to these mutants occurred to wild type levels (Figure 2), but when we tested growth inhibition of the ΔmotA and ΔmotB mutants using a soft agar spot assay, we found that FlaGrab was unable to clear the growth of these mutants (Figure 2, inset). This result supported our hypothesis that flagellar function is required for FlaGrab-mediated growth inhibition.

To better understand the mechanisms by which C. jejuni may resist FlaGrab growth inhibition, we sought to identify strains resistant to this activity. We identified four C. jejuni strains, 12567, 12660, 12661, and 12664, which were naturally reduced in their ability to be cleared by FlaGrab and CC-FlaGrab compared to 11168 (Javed et al., 2013) (Table 4). Strain 12567 was cleared to near-11168 levels, while strain 12664 was not cleared. Strain 12660 displayed reduced clearing, and strain 12661 was generally not cleared, although one biological replicate of this strain was cleared well. Motility of these strains did not correlate with susceptibility to FlaGrab-induced clearance; most strains displayed similar motility to 11168, with the exception of 12661, which was significantly less motile (49%, p = 0.008) than 11168 (Figure S3).

We hypothesized that reduced FlaGrab clearing would correlate with reduced binding. To test this, we probed cells with immunogold-labeled FlaGrab and examined them using transmission electron microscopy. We found that 12567 displayed robust FlaGrab binding (Figure 3), while 12660 displayed intermediate levels of binding. Interestingly, we found that biological replicates of 12661 showed either no binding or robust binding to FlaGrab (n = 2 each). Finally, we observed a complete lack of FlaGrab binding to 12664. Although neither the growth inhibition assay nor the immunogold-labeling assay provides quantitative results, the level of growth inhibition observed for these strains appeared to correlate with the levels of FlaGrab binding to the strains. These results suggested that reduced FlaGrab binding may explain the observed resistance to FlaGrab achieved by these strains.

We next sought to understand how cells achieved altered levels of FlaGrab binding. As we previously showed that FlaGrab requires Pse5Ac7Am for binding to C. jejuni 11168 flagella (Javed et al., 2015a), we hypothesized that strains displaying reduced levels or differing patterns of FlaGrab binding would express different levels of Pse5Ac7Am on their flagella. Pse5Ac7Ac is synthesized by the sequential action of PseB, PseC, PseH, PseG, PseI, and PseF (Logan, 2006). Pse5Ac7Am is formed by adding the Am group to Pse5Ac7Ac by PseA (Guerry et al., 2006). Pse5Ac7Ac and Pse5Ac7Am are then transferred to flagellin in the cytoplasm by the action of separate transferases: PseE transfers Pse5Ac7Ac, while PseD transfers Pse5Ac7Am (Karlyshev et al., 2002; Karlyshev and Wren, 2005; Guerry et al., 2006). C. jejuni and C. coli require Pse5Ac7Ac biosynthesis for flagellar biogenesis (Goon et al., 2003), and all strains examined here are flagellate, and thus, we predicted that all strains would encode the genes required for Pse5Ac7Ac synthesis and transfer onto Fla protein subunits. However, we hypothesized that 12664, which is unable to bind FlaGrab, would encode a non-functional version of a gene involved in Pse5Ac7Am biosynthesis and/or transfer (i.e., pseA and/or pseD). We hypothesized that 12661, which we found to variably bind FlaGrab, might express a phase-variable version of one or both of these genes. To test these hypotheses, we analyzed the sequenced genomes for these strains (Sacher et al., 2018a,b) and examined the region spanning cj1293-cj1343, which was previously established as the flagellar glycosylation locus in C. jejuni 11168 (Parkhill et al., 2000).

In terms of gene content, we found a high degree of similarity at the flagellar glycosylation loci of all five strains (Figure 4). Compared to 11168, which encodes 47 genes in this locus, we found that 12567, 12660, and 12664 had between 46–48 genes, while 12661 encodes only 40. This is in contrast to C. jejuni 81–176, another well-characterized strain, which encodes only 24 of the 47 flagellar glycosylation genes encoded by 11168 (Guerry et al., 2006). Notably, we found that similarly to 11168, all four strains (12567, 12660, 12661, and 12664) encoded copies of all genes required for Pse5Ac7Ac and Pse5Ac7Am synthesis and transfer.

We next examined whether amino acid sequence differences in PseA or PseD encoded by 12661 or 12664 could explain the observed differences in FlaGrab binding to these strains, since these proteins are the only two genes known to be required for Pse5Ac7Am display on flagella (the known FlaGrab receptor) in 11168 that are not required for flagellar filament biogenesis (Guerry et al., 2006). At the protein level, we found that PseA was upwards of 98.7% identical among strains, but that PseD diverged among strains (Figure S4A). PseD sequences from 11168, 12567, and 12660 were upwards of 96.9% identical to one another, but sequences from 12661 and 12664 were only 79.6–83.4% identical to PseD from 11168, 12567, and 12660. PseD sequences from 12661 and 12664 were 93.6% identical to one another. Most of the sequence divergence occurred within the third quarter of the PseD amino acid sequence. The fact that the two strains observed to express the most substantial reduction in FlaGrab binding grouped apart from the other strains based on their PseD sequence supported the hypothesis that changes in PseD sequence might explain the reduced FlaGrab binding of strains 12661 and 12664.

The fact that FlaGrab binding to 12661 flagella occurred variably, while 12664 displayed a consistent lack of FlaGrab binding, pointed to the existence of differences in Pse5Ac7Am expression between these strains. This prompted us to more closely examine the nucleotide sequence identity of pseD from these two strains, which led us to discover a 25-nt insertion near the 5′ end of pseD in 12661 that is not present in any of the other strains (Figure 5). Interestingly, this insertion included a 9-G homopolymeric (poly-G) tract. Homopolymeric tracts are commonly signs of phase-variability in bacterial genes, and have been implicated in C. jejuni phage-host interactions (Coward et al., 2006; Sørensen et al., 2012). We analyzed the raw Illumina MiSeq reads from our whole genome sequence analysis of 12661 (Sacher et al., 2018b), and indeed found variability in the number of Gs at this locus, supporting the notion that this strain could encode a phase-variable version of pseD (data not shown).

To test the hypothesis that slipped-strand mispairing in pseD drives variation in 12661 binding to FlaGrab, we first amplified the pseD region from five single colony isolates, then from a mixed population of cells, and then from purified genomic DNA, all in the absence of FlaGrab, and assessed poly-G tract lengths using Sanger sequencing. To ensure the region we were amplifying was pseD and not a homologous sequence, since C. jejuni strains tend to encode 6–7 highly similar maf (motility associated factor, including pseD) genes at the flagellar glycosylation locus, we designed our primers to amplify a region anchored within ptmA, the gene directly upstream of pseD. We found that in all cases, the poly-G tract in pseD_12661 was in the “on” (9 Gs) state, suggesting in-frame expression of PseD (data not shown). We next sought to determine whether FlaGrab binding would provide a pressure for the strain to phase-vary its pseD. We grew 12661 in the presence of FlaGrab and amplified pseD from 12661 within and outside the spot of FlaGrab-induced clearing using colony PCR followed by Sanger sequencing. Sequence analysis of amplicons derived from FlaGrab-exposed cells (n = 2 sections) and FlaGrab-unexposed cells (n = 3 sections) showed that the poly-G tract remained phased-on in each case, with no apparent secondary peaks underneath the primary nucleotide identified at each location (data not shown). From this, we concluded that under the conditions tested, FlaGrab pressure does not select for off-switching of pseD in C. jejuni 12661.

Along with the differences in pseD sequence observed in 12661 and 12664, we wanted to determine whether these strains, which displayed reduced FlaGrab binding, displayed flagellar glycans distinct from 11168, which displays robust FlaGrab binding. To test this, we used high-resolution liquid chromatography with tandem mass spectrometry (LC-MS/MS) to identify and compare the glycans present on trypsin-digested flagella isolated from 12661 (in its non-FlaGrab-binding state), 12664 and 11168. We hypothesized that we would see reduced levels (or a complete absence) of Pse5Ac7Am on 12661 and 12664 flagella compared with 11168, and we surmised that other glycans would instead be present on the flagella of 12661 and 12664.

To verify our methodology, we first examined 11168 flagellar glycans. In agreement with Ulasi et al. (2015), we detected four glycopeptides: ¹⁸⁰ISTSGEVQFTLK¹⁹¹, ²⁰⁴VVISTSVGTGLGALADEINK²²³, ³³⁹DILISGSNLSSAGFGATQF³⁵⁷, and ⁴⁶⁴TTAFGVK⁴⁷⁰, and identified several glycoforms for each (Figures 6, 7) (Ulasi et al., 2015). Also in agreement with Ulasi et al. and others, we detected oxonium ions corresponding to Pse5Ac7Ac (m/z = 317.13), Pse5Ac7Am (m/z = 316.15), dimethylglyceric acid (DMGA)-Pse5Ac7Ac (m/z = 391.16), and DMGA-Pse5Ac7Am (m/z = 390.18) (Logan et al., 2009; Zampronio et al., 2011; Ulasi et al., 2015). We also compared the relative glycan abundances at each location by comparing the relative peak intensities between the analyzed glycopeptides. We observed that for all four peptides, 3.81–16.44% of the glycan content corresponded to Pse5Ac7Ac, while 56.39–69.89% corresponded to Pse5Ac7Am or its Leg derivative, 2.64–10.41% corresponded to DMGA-Pse5Ac7Ac, and 12.01–23.65% corresponded to DMGA-Pse5Ac7Am (Figure 7).

We next sought to analyze 12661 flagellar peptides. To confirm the non-FlaGrab-binding state of the 12661 population used for MS, we performed FlaGrab immunogold labeling/TEM on the same cell preparation as used to prepare flagella for MS, and confirmed that the protein did not bind to flagella in that preparation (data not shown). Our MS analysis of trypsin-digested 12661 flagellin resulted in the detection of the four glycopeptides described for 11168, ¹⁸⁰ISTSGEVQFTLK¹⁹¹, ²⁰⁴VVISTSVGTGLGALADEINK²²³, ³³⁹DILISGSNLSSAGFGATQF³⁵⁷, and ⁴⁶⁴TTAFGVK⁴⁷⁰ (Figure 8). On all four peptides, we detected oxonium ions of m/z = 317.13 (Figure 8A), which corresponds to Pse5Ac7Ac. Also on all four peptides, we detected an oxonium ion of m/z = 332.15 (Figure 8B), which does not correspond to any known Campylobacter flagellar glycan, but based on the addition of 15 Da to the mass of Pse5Ac7Ac, may represent the addition of an amine group to this glycan (the proposed structure is depicted in Figure 8B, inset). Notably, we did not detect oxonium ions corresponding to Pse5Ac7Am, DMGA-Pse5Ac7Ac or DMGA-Pse5Ac7Am. In terms of relative glycan content, we observed that for the first three peptides, glycans corresponding to Pse5Ac7Ac made up 9.78–30.31% of the glycan content, while the remaining 69.69–90.22% of the glycan content corresponded to the unknown glycan of m/z = 332.15 (Figure 7). However, the ⁴⁶⁴TTTFGVK⁴⁷⁰ peptide displayed the opposite trend, with 83.00% of the glycan content corresponding to Pse5Ac7Ac and 17.00% corresponding to the unknown glycan of m/z = 332.15.

We next analyzed 12664 flagellin. We detected the first three glycopeptides described above, but not the ⁴⁶⁴TTAFGVK⁴⁷⁰ peptide, so its glycan profile could not be determined (Figures 7, 9). Interestingly, upon analysis of the three glycopeptides detected, we detected oxonium ions of m/z = 316.15, which corresponds to Pse5Ac7Am. Also on all three glycopeptides, we detected an unknown glycan of m/z = 331.16. Based on the addition of 15 Da to the mass of Pse5Ac7Am, this may represent the addition of an amine group to this glycan (the proposed structure is depicted in Figure 9B, inset). In addition, in this strain, as for 12661 but not for 11168, we did not detect the DMGA derivatives DMGA-Pse5Ac7Ac or DMGA-Pse5Ac7Am. In terms of glycan content, the glycan corresponding to Pse5Ac7Am was 43.86–58.60% abundant across the three detected glycopeptides, while the unknown glycan of m/z = 331 was 41.40–56.14% abundant. Also, the glycan corresponding to Pse5Ac7Ac was 3.17% abundant on the ²⁰⁴VVISTSVGTGLGALVEEINK²²³ peptide only, but was not detected on the others.

To rule out the possibility that the absence of DMGA-Pse5Ac7Am was the reason for the lack of FlaGrab binding in 12664 and 12661, we sought to determine whether FlaGrab binding to Pse5Ac7Am requires DMGA in 11168. Hitchen et al. (2010) have shown that a cj1295 mutant in C. jejuni 11168 is unable to synthesize DMGA-Pse5Ac7Ac or DMGA-Pse5Ac7Am (Hitchen et al., 2010). We therefore generated a cj1295 mutant in 11168 and used immunogold labeling/TEM to test whether FlaGrab binding still occurred in this strain. We found no difference in FlaGrab binding between the cj1295 mutant and wild type cells (Figure 10), suggesting that DMGA is not required for FlaGrab binding to Pse5Ac7Am.

To gain insight into the genetic basis for the observed differences in flagellar glycans detected on 11168, 12661, and 12664 flagella, we examined the differences in gene content at the flagellar glycosylation locus of each strain. In particular, we hypothesized that 12661 and 12664 would encode genes in this locus not encoded by 11168 that might explain their display of unknown glycans of m/z = 332 and 331 glycans, respectively. Notably, the m/z value for each unknown glycan represents a gain of ~15 Da relative to the other glycan detected for that strain (m/z = 317 and 316, respectively), suggesting that an amine group (NH₂, 16 Da) might substitute for a hydrogen atom on the glycan detected in each case. We therefore examined the genes within the flagellar glycosylation locus that were only present in 12661 and 12664 to see if any had homology to an aminotransferase, but did not identify any aminotransferase homologs.

To identify candidate genes that might be involved in the synthesis of the unknown glycans of m/z = 332 and 331, we examined other differences in gene content at the flagellar glycosylation locus between 12664, 12661, and 11168. We noted strain-specific differences in maf gene content and in distribution of poly-G tracts among these genes. Differences in maf genes were particularly noteworthy, as paralogues of these (pseD, pseE) are known to be involved in pseudaminic acid transfer to flagellin in Campylobacter. For instance, strain 12664 lacks the motility associated factor maf5 (cj1341c), while maf6 (cj1342c) is missing from both 12664 and 12661 (Figure 4, Region I). As well, the poly-G tract in maf2 is uniquely absent in 12661. We also noted differences in the number, location, and presence/absence of poly-G tracts in genes containing a DUF2920 domain (DUF = domain of unknown function) (Figure 4, Regions C, D, and I). In 12664, we noted the unique presence of the DUF2920-containing gene CJ12664_1248 downstream of the identically-annotated cj1310c (Figure 4, Region D). In addition, we found that 12664 encodes another DUF2920-containing gene, CJ12664_1277, just upstream of maf4 (cj1340c) (Figure 4, Region I), and we found that 12661 also encoded this second DUF2920-containing gene (CJ12661_1263). However, strain 12661 is also missing the DUF2920-containing gene cj1306 (Figure 4, Region C). While strains 11168 and 12664 encode a poly-G tract in cj1305c, another DUF2920-containing gene, this poly-G tract is disrupted in strain 12661. Outside of maf and DUF2920-containing genes, strain 12664 also uniquely lacks the fkbH domain-containing gene cj1302 (Figure 4, Region B) (it is present as a pseudogene), and 12661 uniquely lacks cj1301, which is annotated as an epimerase (Figure 4, Region B).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.