Bacterial strains used in this study are listed in Table 1. Both C. jejuni NCTC12662 and the phage resistant strains 12662X3 and 12662R were routinely cultured on Blood Agar Base II (Oxoid) supplemented with aseptically obtained 5% calf blood (BA) and incubated at 37°C under microaerobic conditions (6% CO₂, 6% O₂, and 88% N₂H₂). Standard Luria-Bertani broth (LB) and LB agar (LA) (Difco) were used for growing E. coli cells. Brain hearth infusion (BHI) supplemented with 5% calf blood (BHIb) was used for phenotypic expression of C. jejuni mutants. Where necessary, 100 μg/ml of kanamycin and 20 μg/ml of chloramphenicol were used for selection of transformants.

Overnight cultures of C. jejuni on BA were flooded with 1 ml of cation-adjusted (1 mM CaCl₂ and 10 mM MgSO₄) Brain Heart Infusion broth (Oxoid) (cBHI) and harvested with a sterile Pasteur pipette in a sterile eppendorf tube. Following adjustment to an optical density at 600 nm (OD₆₀₀) of 0.35 (10⁹ CFU/ml) in cBHI and incubation at 37°C for 4 h microaerobically, 500 μl of logarithmic phase culture was added to 5 ml of molten NZCYM overlay agar (NZCYM broth [Sigma] with 0.6% agar [Sigma] and 10 μg/ml vancomycin [Sigma]) that was adjusted to 45°C. After a brief vortex, this mix was poured on previously prepared NZCYM agar (with 1.2% agar and 10 μg/ml vancomycin) plates and allowed to dry for 45 min to 1 h in the flow hood. Phage F207 harboring lawns were prepared the same way, using 500 μl of 10⁶ PFU/ml F207 in SM buffer (0.1 M NaCl₂ [Merck], 8 mM MgSO₄.7H₂O [Merck], 0.01% gelatin [Sigma], 50 mM Tris-HCl [Sigma] in 1 l of ultrapure water with pH 7.5) instead.

Bacteriophages and the propagation strains used in this study are listed in Tables 1, 2, respectively. Propagation of bacteriophages was performed using plate lysis as previously described (Sørensen et al., 2011; Gencay et al., 2017a). Briefly, 15 μl of relevant phage stock was diluted in 500 μl SM buffer and 200 μl of the diluted phage stock was mixed with 400 μl of logarithmic phase cultures prepared as described above. Following an aerobic incubation for 15 min at 37°C for adsorption of the phages, 5 ml of molten NZCYM overlay agar was added and the phage-host-overlay agar mix was poured on NZCYM agar plates. Following 18–24 h of incubation at 37°C under microaerobic conditions, confluent lysed plates were flooded with 5 ml of SM buffer and left at 4°C with gentle agitation (60–80 rpm) overnight. The next day, excessive SM buffer in the plate lysates were pooled, syringe filtered (0.2 μm [Sartorious]) and stored under refrigeration as phage stocks.

The titers of the phages were determined by spotting 3 × 10 μl of the original phage stock and 10-fold serial dilutions in sterile SM buffer on the lawns of either C. jejuni NCTC12662 or 12662R as described above. For two independent biological replicates, mean plaque numbers were counted and calculated to give PFU/ml.

Campylobacter jejuni NCTC12662 was grown overnight on BA and harvested as described above. A cBHI broth culture of OD₆₀₀ of 0.35 was prepared and spiked with phage F207 at a multiplicity of infection (MoI) of 0.02 and incubated overnight microaerobically, at 37°C with gentle shaking (40–60 rpm) in parallel with a F207 negative culture. On the following day, OD₆₀₀ values of the cultures were determined and 10-fold dilutions were prepared to determine colony and plaque counts by spotting the dilutions of both F207 negative and positive cultures on BA and phage F207 lawns, respectively. The number of phages in the F207 positive culture was determined by spotting the dilutions of the supernatant, obtained after centrifugation at 6000 × g for 15 min, onto NCTC12662 lawn. Following incubation, a number of colonies from the phage exposed culture were isolated and re-streaked on BA plates for purification. A total of seven rounds of re-streaking were done; the single colony from the third round was named 12662X3, whereas the final one was named 12662R.

Colony counts were determined by spotting 3 × 10 μl of the corresponding 10-fold dilutions of C. jejuni in sterile physiological saline on BA in duplicate. Following microaerobic incubation at 37°C for 48 h, CFU/ml counts were calculated. For determination of F207 resistant C. jejuni counts, spotting was performed on a lawn of F207 and incubated microaerobically at 37°C for 24 h.

Campylobacter jejuni NCTC12662 and 12662R grown on BA were subjected to phage adsorption as described elsewhere with slight modifications (Scott et al., 2007). Overnight cultures grown at 37°C under microaerobic conditions on BA plates were flooded with 1 ml of cBHI, cells were harvested as described above and centrifuged at 6,000 × g for 5 min. Cells were washed for a total of three times with cBHI, adjusted to an OD₆₀₀ of 0.40–0.45 (~10⁹ CFU/ml), and infected with F207 at different MoIs (0.05, 0.0002, and 0.00003) in three different experiments with single replica and incubated aerobically at 37°C with agitation at 100 rpm. Prior to infection of bacterial cultures, 1 ml aliquots of samples were taken for determination of initial number of CFU/ml. Through the course of a 90 min incubation period, 1 ml aliquots of bacteria-phage mixtures were taken at 0, 30, 60, and 90 min, syringe filtered (0.20 μm) and F207 titers were determined on C. jejuni NCTC12662 lawns as described above.

Motility assay of both C. jejuni NCTC12662 and phage resistant isolate 12662R was performed as previously described (Baldvinsson et al., 2014). Briefly, bacterial cultures were grown, harvested in BHI as described above and adjusted to OD₆₀₀ of 0.1. One microliter of cell suspensions were inoculated in the middle of pre-dried (45 min) Heart Infusion Broth (Difco) plates (containing 0.25% agar), and plates were incubated at 37°C microaerobically. Average motility was determined from a total of five plates by measuring the growth zone in diameters from three different radiuses on each plate at 23 and 46.5 h.

High-resolution magic angle spinning nuclear magnetic resonance (HR-MAS NMR) analysis of intact bacterial cells was performed as described by McNally et al. (2005), and the spectra were acquired in D₂O on a Varian VNMRS 600 MHz spectrometer using VNMRJ and equipped with a ¹H{¹⁵N–³¹P} 4 mm pulsed field gradient (PFG) indirect-detection nanoprobe. The 1D ¹H spectra were acquired with 256 transients, referenced to an internal 3-(trimethylsilyl)propionic-2,2,3,3-d₄ acid sodium salt standard (δH 0.00 ppm), and the intensity of the residual H₂O peak was reduced with a standard presaturation pulse at 4.76 ppm. The 1D ¹H–³¹P HSQC spectra were acquired with 512 transients. Line broadening of 2.0 Hz was applied to all spectra before Fourier transformation.

Pure cultures of 12662X3 and NCTC12662R streaked on BA plates were harvested in physiological saline, and DNA was extracted using the DNAeasy kit (Qiagen) and eluted in 10 mM Tris-HCl. DNA libraries were prepared using Nextera XT v.3 kit (Illumina) and sequencing was executed in a MiSeq (Illumina) platform in paired-end (2 × 250-bp) operating mode. Sequencing runs resulted in 971,110 and 886,084 reads, respectively, with an average read length of 229.7-bp for both strains. Using the CLC Genomics Workbench 9.5.3. (Qiagen) with the settings: match score = 1; mismatch cost = 10; insertion and deletion costs = 3, all of the reads were individually mapped to the wild type C. jejuni NCTC12662 genome (GenBank accession no: CP019965). Each of the 19 polyG tracts found in NCTC12662 were inspected manually and found to have coverages between 28 and 131 ×. The number of reads with respective polyG numbers were counted manually and the percentage of different lengths of polyG harboring reads were calculated and compared to the percentages of reads obtained previously (Gencay et al., 2017b). Consensus sequences were extracted, aligned using Mauve (Darling et al., 2004) and inspected for gaps, rearrangements or single nucleotide polymorphisms (SNPs). Searches for gene similarities were primarily done using the BLAST algorithm available at NCBI (https://blast.ncbi.nlm.nih.gov) and the CLC Workbench. Further identification of protein orthologous/homologs was executed by protein family-associated hidden Markov models (HMMs) based HMMER3 (http://www.ebi.ac.uk/Tools/hmmer/). Global alignment of proteins were done using EMBOSS Needle (http://www.ebi.ac.uk/Tools/psa/emboss_needle/) with gap and extension penalties of 10 and 0.5, respectively, and transmembrane predictions were done using PHOBIUS (http://phobius.sbc.su.se/).

BLAST similarities with known MeOPN-transferases (cj1421 and cj1422) from NCTC11168 indicated that gene 06810 might be the sole MeOPN-transferase found in NCTC12662. Indeed, 06810 shares 83.2% identity with the N-termini (1–244 residues) of Cj1421 and Cj1422 which is followed by a transmembrane domain and a domain of unknown function (DUF2972, Pfam: PF11186) situated at the C-terminus. Disruption of gene 06810 encoding the putative MeOPN-transferase was achieved by allelic exchange using plasmid pYEG102 carrying the chloramphenicol acetyltransferase gene (cat) from pRY109 (Yao et al., 1993) flanked by 732-bp upstream and a 760-bp downstream fragments of 06810, resulting in a 1704-bp replacement of the coding region of gene 06810. Plasmid pYEG102 was constructed by using the In-Fusion HD cloning kit, combining three different fragments situated on an NdeI-XhoI digested pET28a (+) (Novagen) backbone. The 732-bp upstream and 760-bp downstream fragments with overlaps (underlined) both on digested pET28a (+) and cat gene were amplified from NCTC12662 using primer pairs B-F: 5′-TTTGTCGCACTGATAGTGCAAGGAGTTGGATGGCGATA-3′ and B2-R: 5′-GGTGGTGGTGCTCGAGCCAAGCTCCAAGATAACGCTCT-3′ as well as A2-F: 5′-CGCGCGGCAGCCATATAATACTTTGTGGCGGCTTAGGAACAAGGCT-3′ and A-R: 5′-TATATCATAAATCTATTTTGTAAAGTCAATCATAGCTTGACCTAGTTTATAAGCT-3′, respectively. The cat gene with overlaps on both upstream and downstream fragments was amplified from pRY109 using primers 23gcat-F: 5′-GATTGACTTTACAAAATAGATTTATGATATAGTGGATAGATTTATGATATAATGAGTTA-3′ and 23gcat-R: 5′-TCCAACTCCTTGCACTATCAGTGCGACAAACTGGGATTTTATTTATTC-3′. The ligation was transformed into E. coli Stellar™ competent cells according to manufacturer's recommendations (Clontech Laboratories) and transformations were selected by kanamycin and chloramphenicol resistance on LA plates at 37°C. Competent NCTC12662 cells were prepared and transformed with pYEG102 by electroporation. Cells were transferred to 1.5 ml of BHIb in a petri dish and following phenotypic expression at 37°C under microaerobic conditions for 18 h, cells were spread on BA plates containing chloramphenicol and incubated at 37°C under microaerobic conditions for up to 7 days. Correct homologous recombination and disruption of gene 06810 was confirmed by sequencing PCR products spanning the region.

To investigate phage resistance development in C. jejuni NCTC12662, we exposed a liquid culture of 1.5 × 10⁸ CFU/ml NCTC12662 to 3.0 × 10⁶ PFU/ml phage F207, an uncharacterized phage in our collection isolated from a duck abattoir previously shown to have a very distinct host range (Hansen et al., 2007). When incubated overnight, we found that phage exposure did not decrease the number of cells compared to the control culture neither did the culture show indications of lysis (Table 3). Furthermore, phage F207 could be isolated from the supernatant and had propagated approximately 10-fold without any particular influence on the growth of the culture (Table 3). When the phage-exposed culture was plated on a lawn of phage F207, 25% of the cells were able to grow (data not shown), suggesting selection of a phage resistant subpopulation of NCTC12662. To study the molecular mechanism behind resistance development, a phage resistant colony was isolated from the phage F207 lawn and re-streaked 7 times for purification. The resulting strain named C. jejuni 12662R was highly resistant to phage F207 as neither lysis nor plaques were formed on a lawn of 12662R when the 10⁷ PFU/ml stock and the serial dilutions were spotted. To identify the resistance mechanism, we determined the ability of phage F207 to bind to C. jejuni 12662R and wild type cells in an adsorption assay, enumerating free phages over time after co-incubation with C. jejuni cells. While the number of the free phages declined approximately 100-fold over 90 min in the presence of wild type cells, the number of free phages remained unchanged in the presence of 12662R, demonstrating that phage F207 is not able to adsorb to the resistant strain 12662R (Figure 1). Thus, exposure of C. jejuni NCTC12662 to phage F207 selected for a phage resistant adsorption mutant.

By using an acapsular kpsM mutant and a motA mutant carrying non-motile flagella filaments in a NCTC12662 background, we previously determined that the phages in our collection isolated in Denmark in 2011 are either dependent on capsule or a motile flagellum for successful infection (Table 2; Sørensen et al., 2015). To determine the target receptor of the remaining phages in our collection isolated in 2004 (Hansen et al., 2007), the same approach was used. We found that phage F207 (used to select the phage resistant 12662R), as well as F267 and F268 are dependent on CPS for infection of NCTC12662, whereas phage F325 depends on a motile flagellum (Table 4). Thus, all phages in our collection are either dependent on CPS or motile flagella for infection and as previously seen, the receptor type dependency correlated with the phage genera Cp8virus or Cp220virus, respectively (Hansen et al., 2007; Sørensen et al., 2015).

Following this, we performed plaque assays using all phages in our collection and found that none of the phages that are dependent on motility for infection could form plaques on a lawn of 12662R (Table 5). But in contrast to the flagellotropic phages, a mixed pattern of infectivity of 12662R was observed for the CPS-dependent phages (Table 5). CPS-dependent phages F198, F207, F303, F370, F371, F372, and F373 did not form any lysis or plaques on 12662R and thus, were unable to infect this phage resistant variant of NCTC12662. On the other hand, the remaining CPS-dependent phages were still able to infect C. jejuni 12662R, although with a variety of lower efficiencies compared to the wild type (Table 5). Interestingly, the different levels of infectivity of the CPS-dependent phages neither correlate with the time of isolation (2004 vs. 2011), isolation strain, or the origin of phages (Tables 2, 5). Thus, C. jejuni 12662R showed complete resistance to the flagellotropic phages, whereas the CPS-dependent phages were differently affected by the changes in 12662R, indicating a resistance mechanism that influences Campylobacter phages broadly.

The phage sensitivity profile of 12662R suggested that the motility was affected, since none of the flagellotropic phages in our collection were able to form plaques on lawns of 12662R. Using a soft-agar assay we confirmed that C. jejuni 12662R indeed was non-motile, whereas the wild type NCTC12662 was highly motile (Figure 2). Since C. jejuni 12662R was isolated after re-streaking several times, we tested the motility of the initial phage resistant isolate and could confirm the non-motile phenotype of this isolate also (data not shown). These findings suggested that C. jejuni 12662R became non-motile during the growth of the culture while being exposed to phage F207. However, as phage F207 is dependent on the CPS for infection, we hypothesize that the phage exposure was most likely not the selective cause for loss of motility. Our own experience, as well as the literature confirms that C. jejuni easily loses motility under in vitro conditions (Jerome and Mansfield, 2014) and it may thus be an event that occurred independently of the F207 exposure.

The MeOPN modification of the CPS has previously been identified as the receptor for phage F336 in C. jejuni NCTC11168 (Sørensen et al., 2011). We previously detected MeOPN in the CPS of the strain NCTC12662 (Sørensen, 2011) by HR-MAS NMR and subsequent genome sequencing has confirmed the presence of the MeOPN biosynthesis gene orthologous (06780-06795) in the CPS locus of this strain (Gencay et al., 2017b). Hence, we compared the CPS of the wild type NCTC12662 and 12662R by HR-MAS NMR and particularly looked for the presence of MeOPN. Interestingly, we found that the signal for the MeOPN modification was diminished in the decoupled 1D ¹H-³¹P HSQC spectra of 12662R compared to NCTC12662 (Figure 3), indicating that the MeOPN modification of the CPS was significantly reduced in the phage resistant mutant 12662R.

To determine the underlying genetic changes causing both the non-motile phenotype and the observed phage resistance, we performed whole genome sequencing of the phage resistant mutant 12662R as well as the mutant originating from the third re-streak of the original phage resistant colony (12662X3), to rule out the influence of the consecutive re-streaking during isolation of 12662R. Subsequent comparative genomic analyses only revealed a few genetic changes of the phage resistant variants that are explained below. Interestingly, the changes observed except for one, were associated with polyG tract length differences.

Two flagella associated changes were identified in 12662X3 and 12662R compared to NCTC12662; a SNP that causes Ala202Val in an integral membrane domain of an FlhA homolog (Gilbreath et al., 2011) encoded by 04195, and a change in the polyG tract length of the gene 06235, leading to an OFF state of expression of this gene (Table 6). Gene 06235 is a homolog of cj1295 of C. jejuni NCTC11168 (Szymanski and Wren, 2005), and both ON and OFF expression variants, as well as a NCTC11168Δcj1295 mutant were previously shown to be fully motile (Hitchen et al., 2010; Baldvinsson et al., 2014), indicating that loss of motility in 12662X3 and 12662R is unlikely to be associated with the OFF state of 06235. In contrast, FlhA is an important part of the flagellar export machinery and mutations in flhA renders NCTC11168 non-motile (Carrillo et al., 2004; Gilbreath et al., 2011), thus providing a plausible explanation for the observed non-motile phenotypes of 12662X3 and 12662R.

The genomic analyses also revealed a change in another phase variable gene. While a mixed population of ON and OFF expression states of gene 00195 was observed in the parental strain NCTC12662 (Gencay et al., 2017b), 00195 was dominantly in the ON expression state (9G) in the reads obtained from the phage resistant mutants 12662X3 and 12662R (Table 6). Gene 00195 encodes a protein homologous to Cj0031 of C. jejuni NCTC11168, encoding endonuclease and methyl-transferase activities combined, recently shown to reduce the efficiency of plaque formation of certain C. jejuni phages when they were propagated on a Δcj0031 background (Anjum et al., 2016). Since our phages were propagated on a mixed population of NCTC12662, it is not possible to interpret if 00195 had any influence on plaque formation of our phages. Most likely this altered phenotype is a random result of picking a single colony for the isolation of 12662R and not associated with phage F207 exposure (Bayliss and Palmer, 2012; Aidley et al., 2017).

The CPS locus of NCTC12662 encodes four putative phase variable genes carrying polyG tracts in coding regions (Gencay et al., 2017b). The sequence analysis confirmed that the polyG tract length within the putative MeOPN-transferase (06810) of 12662R was altered leading to a switched OFF state of this gene due to an early translational stop downstream of polyG tract. Moreover, 06810 remained in an OFF state throughout the consecutive re-streaks of the phage resistant variant (Table 6), thus explaining the observed reduction of the MeOPN presence in the CPS of 12662R (Figure 3). No other SNPs were detected in any of the other CPS genes. In summary, the genetic analysis of 12662R showed that a single SNP in FlhA as well as the phase variable switching of gene 06810 to an OFF expression state, may have caused the phage resistance phenotype of C. jejuni 12662R.

So far, the MeOPN modification of Galf NAc in strain NCTC11168 is the only identified phage receptor of C. jejuni CPS-dependent phages (Sørensen et al., 2011). Here we have shown that phage F207 does not bind to 12662R, which is not expressing the functional MeOPN-transferase gene (06810). Hence, we speculated if MeOPN is a common constituent of the receptors for CPS-dependent phages infecting C. jejuni. We therefore constructed a defined MeOPN-transferase knockout mutant in NCTC12662 by inserting a chloramphenicol resistance marker to disrupt gene 06810 and confirmed the absence of MeOPN by HR-MAS NMR (Figure 3). Subsequently, we performed plaque assays using the CPS-dependent phages in our collection and demonstrated a similar phage resistance phenotype of the MeOPN deletion mutant compared to 12662R (Table 5). However, three phages (F358, F359, and F366) showed a reduced efficiency of plaquing on NCTC12662Δ06810 as compared to 12662R, whereas phage F369 showed the opposite pattern. Since NCTC12662Δ06810 was completely resistant to infection by CPS-dependent phages F198, F207, F303, F370, F371, F372, and F373, we could conclude that these phages absolutely require the MeOPN modification for infection. On the other hand, the remaining CPS phages (except F358, F359, F366, and F369) were able to infect NCTC12662Δ06810 with differing reduced efficiencies similarly to what was observed for 12662R, suggesting that MeOPN is not essential for these phages to infect C. jejuni. In conclusion, some CPS phages show a complete dependency on MeOPN presence for infecting C. jejuni, while others do not require the MeOPN modification to perform a successful infection.

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.