All bacterial strains and C. jejuni phages are listed in the Supplementary Tables 1, 2, respectively. Thirty-one C. jejuni strains were isolated from Danish broiler chickens in 2007–9 and were selected to represent the most frequent sequence types (ST) in this reservoir (Boysen et al., 2014). C. jejuni strains were standardly grown on base II agar plates containing 5% calf blood at 41.6°C under microaerobic conditions (6% O₂, 6% CO₂, 88% H₂N₂). When appropriate chloramphenicol was added to plates at a final concentration of 20 μg/ml. For phage-related work C. jejuni strains were routinely cultivated in brain heart infusion (BHI) broth (Oxoid) supplemented with 10 mM MgSO₄ and 1 mM CaCl₂ (CBHI).

Phages were propagated and titrated as previously described (Gencay et al., 2017; Sørensen et al., 2017). Briefly, for the phage propagation, the plate lysis method was performed using corresponding C. jejuni propagation strains and original phage stocks. Propagated phage stocks were stored at 4°C. Phage titration (enumeration) and host range analyses were performed using spot assays (plaque assays) on bacterial lawns. Briefly, C. jejuni strains were standardly grown overnight, harvested into CBHI and adjusted to an optical density at 600 nm (OD₆₀₀) of 0.35. Bacterial cultures were then incubated for 4 h at 41.6°C under microaerobic conditions before 500 μl was mixed with 5 ml NZCYM overlay agar (NZCYM broth [Sigma], 0.6% agar) tempered to 45°C and subsequently poured onto premade NZCYM basal agar plates (1.2% agar, 10 μg/ml vancomycin). Plates were then dried for 45 min in a flow hood. Phage stocks were tenfold serial diluted in SM buffer (100 mM NaCl, 8 mM MgSO₄, 50 mM Tris-HCl, pH 7.5) up to 10^–7, and three aliquots of 10 μl of the undiluted stock (10⁰) and each serial dilution were spotted on the bacterial lawns made with relevant C. jejuni strains. Plates were incubated for 18–24 h at 41.6°C under microaerobic conditions. Following incubation plaques were counted and the mean plaque forming units per ml (pfu/ml) was calculated. For the host range analyses, all experiments were performed in duplicate and the data represent the mean of two or three independent experiments.

Selected C. jejuni strains isolated from broiler chickens were genome sequenced using the Illumina platform (Illumina HiSeq) and assembled using CLC genomics workbench version 9.0.1. Average coverage of the genomes ranged from 80-fold to 269-fold. The genomes were annotated using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP). CPS loci were fully assembled using CLC genomics workbench version 9.0.1 by mapping reads to complete C. jejuni CPS reference sequences downloaded from GenBank. A CPS reference sequence was identified by blasting (BLASTN) annotated contigs containing CPS genes against all C. jejuni genomes available in the NCBI database and choosing the closest matching complete CPS sequence. Fully assembled CPS loci were subsequently confirmed by mapping sequencing reads and annotated using RASTtk (Brettin et al., 2015). The complete annotated CPS loci can be found as an appendix at Mendeley Data: https://data.mendeley.com/datasets/2n9r622g8d/1. Comparison of complete CPS loci was illustrated using Easyfig (Sullivan et al., 2011).

CAMSA2002 and CAMSA2038 were additionally sequenced on a PacBio RS II device (Pacific Biosciences, Menlo Park, CA, United States) using P6/C4 chemistry. One SMRT cell was used for each C. jejuni genome. Average coverage for CAMSA2002 was 535-fold while the average coverage for CAMSA2038 was 820-fold. Complete de novo genomes of these two strains were assembled using SMRT Analysis version 2.3 and the HGAP3 algorithm. Annotation was performed using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP).

Comparative genomics and McrBC protein sequence alignment was performed using the CLC genomics main workbench 20 and the Whole Genome Alignment plugin. Conserved domains in selected hypothetical protein sequences were identified using InterPro (Mitchell et al., 2019) and HHpred (Zimmermann et al., 2018). Pairwise comparisons between divergent genes were further conducted using BLASTN available at NCBI.

PolyG regions in genes homologous to cj1421, cj1422, and cj1426 from C. jejuni NCTC11168 were amplified and sequenced (Sanger sequencing) using the primers cj1421F and cj1421R (cj1421), cj1422F and cj11422R (cj1422) and cj1426F and cj1426R (cj1426). Primer sequences are listed in Supplementary Table 3.

Phage adsorption assays were performed as previously described with minor modifications (Baldvinsson et al., 2014). Briefly, C. jejuni were grown overnight on BA plates and harvested in CBHI. Cells were pelleted by centrifugation at 6,000 × g for 5 min and re-suspended in CBHI. This washing procedure was repeated twice (total number of washes, three), and the final cell suspension was adjusted to an OD₆₀₀ of 0.4. Bacteriophages were added to the bacterial suspension at a final concentration of 10⁶ PFU/ml (multiplicity of infection [MOI] of approximately 0.0025) and incubated at 37°C with gentle shaking (50–80 rpm) for a total of 90 min. Samples containing free-phages were collected at 0, 15, 30, 60, and 90 min, filtered through an 0.22-μm sterile syringe filter (Millipore), and stored at 4°C until enumeration (plaque assay). Experiments were repeated twice and for each experiment, the calculated phage pfu/ml at time zero was designated as 100% free phages. The percentage of free phages was calculated for the remaining time points according to time zero. All experiments were performed in duplicate and the data represent the mean percentages of free phages and standard deviations thereof from the two independent experiments.

Motility assays were performed as previously described (Sørensen et al., 2011). Briefly, C. jejuni strains were grown under standard conditions for 18–24 h, harvested into BHI and adjusted to an OD₆₀₀ of 0.1. One microliter of the bacterial suspension was then placed in the center of five to six Heart infusion broth (HIB) (Difco) 0.25% agar plates that had been pre-dried for 45 min in a flow hood. Plates were then incubated under standard growth conditions and growth zones demonstrating movement in the soft agar were measured after 24 h. Measurements and standard deviations represent the mean counts from two independent experiments.

A deletion mutant of mcrB in CAMSA2038 was constructed by replacing most of mcrB (203 bp to 2009 bp) with a cat cassette by homologous recombination. The vector construct used for homologous recombination was created by fusing three PCR amplicons as described below using In-fusion cloning (Takara). A PCR fragment (778 bp) of the flanking region upstream the mcrB deletion was amplified from C. jejuni CAMSA2038 using the primers FLOR06_OH2_F2 and FLOR06_OH2_R2. A PCR fragment (878 bp) of the flanking region downstream the mcrB deletion was amplified using the primers FLOR06_OH1_F4 and FLOR06_OH1_R3. The cat gene (723 bp) was amplified from pRY109 (Yao et al., 1993) using the primers FLOR06_Cat1_fwd and FLOR06_Cat2_rev. The cat gene was then inserted between the upstream and downstream mcrB deletion fragments and simultaneously inserted into the pET28a+ vector (Novagen) using the In-Fusion kit according to the manufacturer’s instructions. The resulting plasmid pFLOR06 was transformed into E. coli Stellar™ Competent Cells (Takara) according to manufacturer’s instructions and verified by sequencing. Purified pFLOR06 plasmid was transformed into C. jejuni CAMSA2038 by electroporation to allow homologous recombination of the flanking mcrB regions to create the CAMSA2038ΔmcrB. The mutant was verified by PCR and sequencing using the primers FLOR06_UP1_fwd, FLOR06_UP2_fwd, and FLOR06_DOWN1_rev. Plasmids and primers are listed in Supplementary Table 3.

Data in the graphs show the mean values of two independent experiments with standard deviations depicted as error bars. When appropriate one-way ANOVA (GraphPad Prism 9.2.0) was used to assess if differences were statistical significant.

To identify putative novel phage resistance mechanisms in C. jejuni, we obtained a collection of 31 strains from Statens Serum Institute, which were isolated from broilers in Denmark between 2007 and 2009 and had previously been characterized by MLST typing. These strains represent the ST-types commonly found in broilers in Denmark and included several strains of the two dominant ST-types, ST-21 and ST-45. We then determined phage susceptibility by screening these strains for plaque formation using our collection of 51 C. jejuni phages belonging to either Fletchervirus or Firehammervirus. This host range analysis showed that 21 strains belonging a variety of STs were infected by at least one phage and most of the strains were sensitive to several phages. In contrast, ten strains belonging to different ST-types were completely resistant to all 51 phages (Figure 1). Most ST’s showed a mixed pattern of phage sensitivity, however interestingly, the phage sensitive ST-21 strains were only infected by Fletchervirus phages dependent on the CPS for infection. On the contrary, strains belonging to ST-45 were only infected by Firehammervirus phages. While our Fletchervirus phages are all dependent on the CPS for infection, the Firehammervirus are flagellotropic phages dependent on motile flagella (Sørensen et al., 2015). Thus, our results suggest a correlation between sensitivity toward a specific phage genus and the two major C. jejuni MLST types, ST-21 and ST-45, often associated with human disease.

To further investigate the genetic relationship between ST-21 and ST-45, we genome sequenced all strains belonging to these MLST types. As Fletchervirus phages are dependent on CPS for infection, we fully assembled and performed comparative genomics of the CPS loci encoded by the ST-21 and ST-45 strains (Figure 2). Our analysis demonstrated that all ST-21 strains encoded identical CPS loci typical of the Penner serotype HS2 group, similar to the CPS locus of the well-characterized C. jejuni NCTC11168 strain. The HS2 CPS locus encodes all biosynthesis genes needed to produce the MeOPN modification and two MeOPN transferases cj1421 and cj1422 responsible for attaching MeOPN to GalfNAc and heptose residues in the CPS, respectively. We previously demonstrated that MeOPN forms a receptor recognized by several Fletchervirus phages both in C. jejuni strains NCTC11168 and NCTC12662 (Sørensen et al., 2011; Holst Sørensen et al., 2012; Gencay et al., 2018). Thus, all ST-21 strains encode the common MeOPN phage receptor of the Fletchervirus phages. On the contrary, the ST-45 strains encode diverse CPS loci, but interestingly no MeOPN biosynthesis or transferase genes were found. These results show that the common MeOPN phage receptor is not present in any of the ST-45 strains, explaining their resistance toward the CPS-dependent Fletchervirus phages.

Although all ST-21 strains encode the common MeOPN phage receptor, three of these strains (CAMSA2021, CAMSA2038, and CAMSA2147) are resistant to phage infection (Figure 1). We previously showed that phase variable expression of the MeOPN transferases cj1421 and cj1422 and the 6-O-Me transferase cj1426 lead to rapid phage resistance development in strain NCTC11168 (Sørensen et al., 2011; Holst Sørensen et al., 2012; Aidley et al., 2017). This was caused by variations in the N-terminal polyG tracts turning the expression of these genes either off (cj1421) or on (cj1422 and cj1426) promoting the loss or masking of the MeOPN receptor, respectively. To investigate if phase variation could be involved in generating phage resistance in these three ST-21 strains, we compared the polyG tract length in the cj1421, cj1422, and cj1426 homologs with sequencing data obtained following PCR amplification of polyG tract regions in these genes (Table 1). As a control, we also included the highly phage sensitive CAMSA2002 strain in our analysis. For all strains, varying polyG tracts length were detected, suggesting mixed populations with some cells expressing the transferases while others did not. Overall, expression of cj1421 was mostly turned on in all four strains leading to expression of the MeOPN receptor attached to GalfNAc. In contrast, variable tract lengths were detected in cj1422 and cj1426 turning expression of the genes both on and off, hence the presence of MeOPN attached to Hep and the 6-O-Me group varied. Thus, these results indicate that phase variable expression of surface structures leading to loss or masking of the MeOPN receptor could not be solely responsible for the observed phage resistant phenotypes. To further investigate if phage resistance was associated with lack of phage binding to the surface of the phage resistant ST-21 strains, we used phage F367 as a representative Fletchervirus and determined adsorption to CAMSA2002 (phage sensitive), CAMSA2038 (phage resistant) and CAMSA2147 (phage resistant). Our data showed that phage F367 indeed bind to both resistant strains (CAMSA2038 and CAMSA2147) at levels comparable to the phage sensitive CAMSA2002 (Figure 3). Thus, our data demonstrate that phage resistance toward Fletchervirus in CAMSA2038 and CAMSA2147 is not associated with lack of phage binding due to phase variable expression of surface structures and involves potentially novel internal phage resistance mechanisms.

While the receptor recognized by Firehammervirus phages has not been identified, it was previously demonstrated that lack or reduced motility can lead to phage resistance in C. jejuni toward this group of phages (Coward et al., 2006; Sørensen et al., 2015). Since all ST-21 strains are resistant to all Firehammervirus phages, we compared the motility of the ST-21 strains with motility patterns observed for the ST-45 that are only infected by this group of phages. However, even though motility varied between strains, the motility of ST-21 strains was comparable to that observed for strains belonging to ST-45 (Supplementary Figure 1). Thus, absence of or reduced motility is not responsible for Firehammervirus phage resistance observed in the ST-21 strains. To further investigate if other differences related to flagella biosynthesis could play a role in Firehammervirus phage resistance, we compared the flagellar O-linked glycosylation loci of the ST-21 and ST-45 strains. Similar to our results obtained from the comparison of the CPS loci, the flagellar O-linked glycosylation loci of all the ST-21 strains was similar to the O-linked glycosylation encoded by C. jejuni NCTC11168, whereas this locus differs among the ST-45 strains (data not shown). Since not much is known about the receptor of Firehammervirus, we could not conclude if specific components related to flagella synthesis and glycosylation was responsible for the differences in Firehammervirus sensitivity of the ST-21 and ST-45 strains. However, our results show that motility per se was not impaired in any of the strains.

To identify possible genetic changes responsible for phage resistance in the ST-21 strain CAMSA2038, we performed comparative genomics of CAMSA2038 with the phage sensitive ST-21 CAMSA2002. Overall, we found that the genomes were highly conserved over the entire length on the chromosome including the CPS and O-linked glycosylation locus (99.57 average nucleotide identity, 99.11 alignment percentage) (Figure 4A). In addition, the phage sensitive CAMSA2002 contains a plasmid with high sequence similarity to the pTet plasmid of C. jejuni strain 81–176 (Supplementary Figure 2), a plasmid encoding tetracycline resistance as it contains a tetO gene (Bacon et al., 2000). The pTet plasmid found in CAMSA2002 (38.0 kb) is, however, substantially smaller compared to pTet in 81–176 (45.2 kb) and it does not encode the tetO gene. Interestingly, only few genes differed between the CAMSA2002 and CAMSA2038 when not considering SNP mutations (Table 2). Also, sequence divergence was detected in the clustered regularly interspaced short palindromic repeats (CRISPR) region of these two strains. In the following, we will present the major genetic differences detected in the phage sensitive CAMSA2002 and phage resistant CAMSA2038.

Sequence analysis of the CRISPR region identified more spacers in the phage sensitive CAMSA2002 strain (six spacers) compared to the phage resistant CAMSA2038 (three spacers) (Table 3). Only one spacer sequence was present in both CAMSA2002 and CAMSA2038 whereas the sequences of the remaining spacers differed between the strains (Table 3). The identical spacer shows 100% similarity to phage DA10, an unclassified phage belonging to the Myoviridae family unrelated to the Fletchervirus and Firehammervirus (Hooton et al., 2020). Furthermore, this spacer sequence also showed a match (28 out of 30 bp) to gene PC5_00034 (hypothetical protein) encoded by Campylobacter phage PC5 belonging to the Fletchervirus genus (Janež et al., 2014). However, the PC5_00034 gene is only found in the PC5 and DA10 genomes, and not present in any other currently available Fletchervirus phage genomes. Another spacer sequence found in the phage sensitive CAMSA2002 strain showed some sequence similarity to a conserved putative phosphatidyl-serine decarboxylase encoded by all Fletchervirus phages (Table 3). But the similarity was rather weak (23–25 bp out of 30 bp) explaining why this spacer does not protect against Fletchervirus phage infection, as CAMSA2002 is sensitive to 34 phages belonging to this genus. The remaining spacers found in CAMSA2038 and two spacers in CAMSA2002 show 100% sequence similarity to the genome of phage DA10. These findings show that CAMSA2002 and CAMSA2038 mainly contain sequences originating from phage DA10 in their CRISPR array. In conclusion, in silico analysis and host range data support that CRISPR is not responsible for the phage resistant nature of CAMSA2038.

Phage sensitive CAMSA2002 and phage resistant CAMSA2038 both encode a DUF262-containing protein (DDR89_00040 in CAMSA2002, DDV75_00040 in CAMSA2038) in the same location of the genomes, but the protein sequences show no similarity (Table 2). We therefore conducted a detailed in silico analysis to predict putative functions of these two proteins (Supplementary Table 4). In CAMSA2038, the DUF262 is present in the N-terminal of DDV75_00040, but the protein also contains a DUF1524 located in the C-terminal and is predicted by HHpred as a SspE homolog across the entire protein length. In contrast, the corresponding protein DDR89_00040 in CAMSA2002 contains a smaller N-terminal DUF262, but no domains in the C-terminal and thus only the N-terminal region was predicted as a SspE homolog (Supplementary Table 4). Recently, SspE coupled with SspABCD was identified as a phage defense system, where SspE senses and nicks phage DNA not containing phosphorothioate (PT) modifications of the sugar-phosphate DNA backbone produced by SspABCD (Xiong et al., 2020). While the DUF262 encode NTPase and PT sensing activity, the DUF1524 domain is responsible for the nicking endonuclease activity. Thus, both domains are needed to ensure the overall structure, enzymatic and phage resistance activities of SspE (Xiong et al., 2020). It is therefore likely that only CAMSA2038 encodes a true SspE homolog (Supplementary Table 4). Furthermore, SspE only exerts its phage resistance properties when SspABCD is present. It is currently not known if C. jejuni DNA contains PT modifications needed for SspE to distinguish and recognize foreign DNA and thus to function. We therefore searched the CAMSA2038 genome for hypothetical proteins containing DUF499 (SspD) and DUF4007 (SspB) domains, however, none such proteins were found. Thus, future experimental verification is needed to show if SspE indeed is associated with phage resistance in CAMSA2038.

Several divergent genes of CAMSA2002 and CAMSA2038 encode putative components of restriction-modification (RM) systems (Table 2). Both CAMSA2002 and CAMSA2038 encode homologs of the RM type IIG system encoded by cj0031 in C. jejuni NCTC11168 (DDR89_00180 in CAMSA2002, DDV75_00180 in CAMSA2038). However, significant sequence divergence is observed in the C-terminal region of these genes. cj0031 encodes both the endonuclease and methylase in a single gene and variations in the C-terminal region has been associated with differences in site-specific methylation (Anjum et al., 2016). Thus, DDR89_00180 and DDV75_00180 may methylate different and potential novel sites in CAMSA2002 and CAMSA2038. In addition, the phage resistant CAMSA2038 contains a polyG tract in DDV75_00180 promoting phase variable expression of the gene.

Other differences include a Type I restriction enzyme HsdR N-terminal domain-containing protein (DDV75_03095) encoded by CAMSA2038 that is not found in CAMSA2002 (Table 2). Type I restriction-modification systems are multi-functional complexes encode by three hsd genes, hsdR (restriction of unmethylated DNA), hsdM (modification) and hsdS (recognize target sequence) where the gene products form a pentameric R₂M₂S complex (Cooper et al., 2017). Both CAMSA2002 and CAMSA2038 encode other hdsR, hsdM and hsdS genes, thus the role of this hypothetical HdsR protein remains unclear.

Finally, CAMSA2038 encodes a putative McrB homolog (DDV75_00695) of the McrBC type IV restriction endonuclease system, while CAMSA2002 in the corresponding position on the chromosome encodes a pseudogene (DDR89_00700) annotated as a putative AAA family ATPase (Table 2). Comparison of the two genes demonstrated a query coverage of 54% with 96% similarity, suggesting that CAMSA2002 also encodes a putative McrB homolog, yet with a significant different sequence (Figure 4B). Further In silico analysis demonstrated that the putative mcrB gene in CAMSA2002 contains frameshift mutations thereby resulting in the annotation as a pseudogene. Thus, while both CAMSA2002 and CAMSA2038 encode mcrC homologs forming the restriction endonuclease, only CAMSA2038 seems to encode a complete mcrB homolog responsible for the DNA binding (Figure 4B). Aligning the mcrB genes from CAMSA2002 and CAMSA2038 showed that the far C-terminals were highly conserved, but the N-terminals showed significant sequence divergence (Figure 4B). The N-terminal of the McrB subunit comprises the actual DNA binding region while the C-terminal region contains an AAA + GTPase domain responsible for GTP hydrolysis (Gast et al., 1997; Pieper and Pingoud, 2002; Loenen et al., 2014). Thus, the McrB encoded by CAMSA2038 may recognize a different methylation site or different type of DNA modification as compared to McrB in CAMSA2002. An additional small conserved region was observed near the middle of the two mcrB genes (position 965–1235 bp in DDV75_00695), suggesting that conservation of this region could be necessary for McrB to establish its function. Further analysis demonstrated that this region covers the beginning of the GTPase domain (data not shown).

In summary, in silico analysis indicate that different types of RM systems could be associated with phage resistance in CAMSA2038.

In E. coli, McrBC forms a methyl-specific endonuclease specifically cleaving modified DNA (Zagorskaitė et al., 2018). While McrB homologous are found in several C. jejuni strains, not much is known about their function and role in this species (Gardner and Olson, 2012). To investigate the role of McrBC in relation to phage resistance in C. jejuni we created a mcrB (DDV75_00695) deletion mutant in CAMSA2038 thereby rendering the McrBC complex non-functional. We then tested the mutant for phage sensitivity by determining plaque formation using 43 Fletchervirus and Firehammervirus phages from our collection. Our data showed that CAMSA2038ΔmcrB was sensitive to 17 out of the 43 phages tested (Table 4). Fourteen Fletchervirus phages were able to infect CAMSA2038 in the absence of mcrB. However, in all cases we observed reduced efficiencies of plating as compared to the number of plaques formed on the phage propagation hosts. In addition, plaque formation varied between experiments. Also, three Firehammervirus phages (F377, F378, and F379) were able to infect and produce plaques on the mcrB mutant. None of the ST-21 strains including CAMSA2038 were otherwise infected by Firehammervirus phages. While plaque formation of phage F377 and F379 also demonstrated reduced and variable efficiencies, F378 was able to consistently form plaques at level almost equal to what is observed on its propagation host. Thus, our data demonstrate that McrBC cause phage resistance towards both Fletchervirus and Firehammervirus phages in C. jejuni strain CAMSA2038.