The Campylobacter strains used for this study were the chicken isolates C. coli OR12 (El-Shibiny et al., 2005), C. coli RM2228 (Fouts et al., 2005), C. jejuni HPC5 (Loc Carrillo et al., 2005), and the human clinical isolate C. jejuni NCTC 11168 (Parkhill et al., 2000).

Aerotolerant C. coli OR12 (Aer) was serially passaged by streaking on blood agar plates (BA; Oxoid) containing 5% v/v defibrinated horse blood (TCS; Buckingham; UK) and incubating aerobically at 37°C for 2–7 days. Intermediate stocks were archived at regular intervals by collecting confluent aerobic growth from a BA plate into freezer stock solution (15% Glycerol, 85% Nutrient Broth No.2; Oxoid) and freezing at −80°C. Gram stains, using standard techniques, were performed at regular intervals to ensure contamination had not occurred. Four colonies were initially selected from the original growth and passaged independently. With the exception of pulsed-field gel electrophoresis (PFGE), all experiments were conducted using isolate B, as it had qualitatively demonstrated the best growth in the early passages. The original wild type C. coli OR12 (WT) was maintained and stored microaerobically on BA.

Blood agar 48 well plates were prepared by dispensing 1 ml of BA to each well (Nunclon Delta; Nunc; Denmark). Inoculum suspensions of approximately 10⁷ CFU/ml of microaerobically grown C. coli RM2228, C. coli OR12 WT, C. coli OR12 Aer and aerobically grown C. coli OR12 Aer were prepared by suspending confluent growth from BA plates into PBS. Triplicate wells, representing biological replicates, were inoculated with 10 μl of suspension (approximately 10⁵ CFU) per strain. Once dried, plates were incubated aerobically at 37°C. After 0, 6, 24, 30, and 48 h of incubation, well contents were suspended by gentle repeated pipetting in maximum recovery diluent (MRD; Oxoid). Suspensions were enumerated on CCDA using a modification of the Miles Misra method (Miles et al., 1938). Briefly, decimal dilutions were performed in MRD, 5 replicate 10 μl droplets from up to six dilutions were plated on modified Campylobacter blood-free selective agar (CCDA; Oxoid) with 2% agar. Once dry, plates were inverted, incubated microaerobically (5% CO₂, 5% O₂, 2% H₂, 88% N₂) in a modular atmosphere controlled cabinet at 42°C (Don Whitley Scientific modified atmospheric cabinet, Shipley, UK) and examined after 24 and 48 h. Dilutions giving rise to between 3 and 30 colonies were selected, the sum of the 5 spots was multiplied by 20, then multiplied by the inverse of the dilution factor to give the original CFU/ml.

Suspensions of approximately 10⁸ CFU/ml of the following strains were prepared: microaerobically grown C. jejuni NCTC 11168; C. jejuni HPC5; C. coli RM2228; C. coli OR12 WT; C. coli OR12 Aer and aerobically grown C. coli OR12 Aer. Each inoculum was decimally diluted five times and one 10 μl droplet of each dilution of each strain was dispensed onto two BA plates and two CCDA plates. Once dry, one of each was incubated microaerobically at 42°C and the other at 37°C aerobically. The microaerobic plates were examined after 48 h and the aerobic plates checked daily for 6 days.

Sterile 250 ml flasks containing 50 ml Mueller-Hinton broth (MHB; Oxoid) were inoculated in triplicate to an approximate cell density of 10⁶ CFU/ml of microaerobically grown C. jejuni HPC5 and C. coli OR12 WT and aerobically grown C. coli OR12 Aer P25. These were incubated aerobically at 37°C with orbital shaking at 100 rpm. Samples were collected at 0, 2, 4, 6, 24, and 30 h and Campylobacter enumerated on CCDA as above.

Sterile 100 ml flasks containing 30 ml nutrient broth no.2 (NB2; Oxoid) were inoculated in triplicate with approximately 10⁴ CFU/ml of microaerobically grown C. coli RM2228 and C. coli OR12 WT and aerobically grown C. coli OR12 Aer P38. The flasks were incubated microaerobically in anaerobic gas jars using gas replacement (7.3% CO₂, 5.6% O₂, 3.6% H₂, 83% N₂) at 42°C with orbital shaking at 100 rpm (John et al., 2011). Samples were collected and Campylobacter enumerated on CCDA at 0, 2, 5, and 8 h.

Motility was assessed using a motility agar assay and via membrane filtration. For the agar assay, inocula were prepared from freshly grown aerobic cultures of C. coli OR12 Aer or microaerobic cultures of C. coli OR12 WT. A pipette tip was used to collect a small plug of colony growth, which was then stab-inoculated into the center of a motility agar (Mueller Hinton broth with 0.4% v/v agar) plate, without touching the bottom. Plates were incubated microaerobically at 42°C and examined after 48 h.

For the membrane filtration assays, sterile nitrocellulose membrane filters with 0.45 μm pore size and 47 mm diameter (Millipore) were aseptically placed in the center of BA plates. The centers of the filters were inoculated with 50 μl of C. coli suspension taking care not to contaminate the surrounding agar. Once the liquid had passed through the filters, they were aseptically removed with sterile forceps. Plates were inoculated in duplicate, with one incubated microaerobically at 42°C and one aerobically at 37°C.

Resistance to peroxide stress was performed using a method similar to that of Rodrigues et al. (2015). Glass universal containers, filled with 10 ml of MRD containing 0, 0.5 or 1 mM H₂O₂ (Sigma; UK), were inoculated with identical optical densities of approximately 10⁸ CFU/ml microaerobically grown C. coli RM2228, C. coli OR12 WT, C. coli OR12 Aer P40 and aerobically grown C. coli OR12 Aer P40. Containers were incubated microaerobically at 42°C for 1 h with shaking at 100 rpm followed by immediate enumeration of Campylobacter on CCDA as above. Three universals were inoculated per strain, representing independent biological replicates.

Confluent BA cultures of microaerobically grown C. coli OR12 WT and aerobically grown C. coli OR12 Aer P37 were fixed in electron microscopy fixative (3% glutaraldehyde in 0.1 M cacodylate buffer) prior to centrifugation and resuspension in sterile water. Transmission electron microscopy (TEM) was performed at the Advanced Microscopy Unit, University of Nottingham. Samples were adsorbed onto copper Formvar/Carbon grids (AGS162-3; Agar Scientific) and negatively stained with 3% uranyl acetate. Grids were imaged using a Tecnai G12 biotwin TEM, run at 100Kv, with a SIS megaview camera system and Gatan Microscopy Suite software (Gatan Inc.).

Cell suspensions of four C. coli OR12 and aerotolerant derivatives were incorporated into agarose plugs, lysed and washed as described by Ribot et al. (2001). DNA plugs were cut into 2-mm slices and placed into 100 μl of 1x SuRE/Cut Buffer A (Roche) and equilibrated at 25°C for 15 min. This was replaced with 100 μl of 0.2 U/μl Sma I (Roche) in the same buffer and incubated for 2 h at 25°C. The plugs were incorporated in a gel consisting of 1% w/v Agarose (Bio-Rad) in TAE buffer. A 50–1000 kb DNA ladder (Lambda PFGE Ladder; New England BioLabs) was added as a marker. Electrophoresis was performed using a Bio-Rad CHEF-DRII. The gel was stained in 50 μg/ml ethidium bromide in TAE buffer and visualized under UV light using the Gel Doc XR system with the Quantity One basic software, version 4.6.5 (Bio-Rad).

Ross 308 broiler chickens (n = 30) were obtained as day old chicks from a commercial hatchery (PD Hook; UK) and reared in biosecure conditions. Chicks were group reared in pens with a bedding of wood shavings and transferred to individual cages with an astroturf floor and environmental enrichment at 18 days of age. A 12-h light-dark cycle was followed for the duration of rearing. Temperatures were as outlined in the Code of Practice for the Housing and Care of Animals Bred, Supplied or Used for Scientific Purposes, with minor adjustments made as guided by bird thermoregulatory behavior. They received commercial broiler diets (starter, grower, and finisher) for the duration. Cloacal swabs were taken at day 18 and tested for Campylobacter by direct plating on CCDA and Salmonella by enrichment. At 21 days of age, 15 birds (Group 1) were gavaged with an estimated 10⁸ CFU of C. coli OR12 WT and 15 birds (Group 2) with C. coli OR12 Aer P34. After 3 days, seven birds from each group were culled by parenteral barbiturate overdose, with the remaining eight birds culled at 7 days post challenge. Caecal contents were collected and enumerated for Campylobacter in triplicate as described above, with CCDA selective supplement (Oxoid) added to the medium. The membrane filtration assay was also performed on the 10⁻¹ dilution of each caecal content sample, as described above. A secondary aerotolerance assay was performed on 24 or 48 h growth on the microaerobically incubated BA filtration plates. Growth was suspended in PBS and normalized to an OD₆₀₀ of 0.05, decimally diluted to 10⁻³ and a 10 μl spot of each dilution inoculated to a BA plate which was incubated aerobically at 37°C.

Power calculations were performed using the Spreadsheet for Sample Size Calculation (Brown, 2016). This indicated that seven birds were required per group to detect a 1 log₁₀ CFU/g difference between groups with 95% confidence and a standard deviation of 0.5 log₁₀ CFU/g. One additional bird was added per group to allow for any unexpected mortality.

All experimental animal work was performed in accordance with UK and EU law. This study was approved by the Local Ethics Committee of the University of Nottingham and performed under Home Office license.

Viable counts were log₁₀-transformed for analysis. Statistically significant differences (p < 0.05) were assessed using parametric (ANOVA) and non-parametric (Mann–Whitney U) tests through the statistical packages available within Minitab statistical software (Minitab Inc.).

Genomic DNA was extracted from half of a confluent aerobically incubated BA plate of C. coli OR12 Aer P43. DNA was isolated using the GenElute Bacterial Genomic DNA Kit (Sigma–Aldrich) according to manufacturer’s instructions. Sequencing was performed using the MiSeq platform at Northumbria University. The sequence reads were aligned to the assembled sequence of C. coli OR12 (NCBI RefSeq NZ_CP013733.1) and appear in the database as CcOR12aero (NCBI RefSeq NZ_CP019977). DNA sequence variants were detected using the variant detector within CLC Genomics workbench 9.01 (Qiagen Bioinformatics). Candidate high probability nucleotide sequence variants detected were confirmed by careful inspection of the sequence reads, and any predicted changes to the protein encoding sequences confirmed using the Artemis genome browser (Rutherford et al., 2000). The predicted protein sequences were further compared with wild-type C. coli OR12 sequences, and with other similar protein sequences using the protein basic local alignment tool, BLASTP mounted on the NCBI website¹. Nucleotide BLAST¹ was used to identify orthologs of genes with known roles in Campylobacter oxidative stress resistance. Protein sequences of gene products in C. jejuni NCTC11168 and C. coli OR12 were compared using the protein BLAST function.

For the initial early passages, growth of aerotolerant C. coli OR12 was limited to the heavily inoculated region of the BA plate. After approximately 15 aerobic passages, growth of single colonies became visible within 2–5 days of aerobic incubation at 37°C. Gram staining confirmed typical Campylobacter morphologies with variable populations of Gram-negative spirals, filamentous, and coccoid forms. Approximately 10⁵ CFU of C. coli RM2228, wild type C. coli O12 (WT), aerotolerant C. coli OR12 (Aer) passage (P) 32, and a microaerobic preparation of C. coli OR12 Aer P32 were inoculated onto BA microplates and aerobically incubated at 37°C. No changes in viable count were apparent after 6 h; however, by 24 h the C. coli RM2228 and C. coli OR12 had declined to below the limit of detection (1.3 log₁₀ CFU). In contrast, both the C. coli OR12 Aer P32 and the microaerobically prepared C. coli OR12 Aer P32 had grown by more than 1 log₁₀ after 24 h and continued to grow, with viable counts of 7.4 and 8.3 log₁₀ CFU, respectively, by 48 h (Figure 1).

Differences between aerobic and microaerobic growth were examined on BA and CCDA plates that had been inoculated with six decimal dilutions of a panel of six Campylobacter strains. These included C. coli OR12 WT along with aerobically and microaerobically prepared C. coli OR12 Aer P38. After 48 h of microaerobic incubation, single colonies were present at the highest dilution (10⁻⁵) of all strains on BA and CCDA, correlating with viable count ranges of between 5.7 × 10⁷ and 2.5 × 10⁸ CFU/ml for the initial inocula. Viable counts of all strains were within 1 log₁₀ of each other (Figure 2). After 6 days of aerobic incubation, single colonies were present down to the 10⁻⁴ dilution of both aerobically and microaerobically prepared C. coli Aer P38 on BA and CCDA. No aerobic growth was present from any of the other strains examined (Figure 2). The growth rates of C. coli RM2228, C. coli OR12 WT and C. coli O12 Aer P38 were assessed by microaerobic incubation at 42°C in nutrient broth no. 2. Viable counts are shown in Supplementary Figure 1. Exponential growth rates for each strain were calculated in the logarithmic growth phase; rates were not significantly different between strains.

Aerobic survival in liquid medium was assessed by inoculating Mueller Hinton broth with an estimated 6 log₁₀ CFU/ml and incubating at 37°C. Viable counts are presented in Supplementary Figure 2. Following 6 h of aerobic incubation, the C. jejuni HPC5, C. coli OR12 WT and C. coli OR12 Aer had declined by 0.17, 0.23, and 0.42 log₁₀ CFU/ml, respectively. By 24 h, both C. coli OR12 strains and two of the C. jejuni HPC5 flasks had declined to below the limit of detection (1.3 log₁₀ CFU/ml) and one C. jejuni HPC5 flask had a viable count of 2.4 log₁₀ CFU/ml. Despite differences in the starting inocula, aerotolerant C. coli OR12 did not demonstrate superior survival compared to the wild type and C. jejuni HPC5.

Motility assays were performed to determine whether serial aerobic passage resulted in reduced motility. The C. coli OR12 WT strain was tested together with C. coli OR12 Aer P22 and P50. All isolates demonstrated motility, with diameters of growth of 35, 34 and 33 mm, respectively, therefore aerotolerance was not associated with any loss of motility.

Owing to their highly energetic corkscrew-like motility campylobacters can traverse membrane filters that would exclude other bacteria, a property that has been used as a basis for their isolation from several animal sources (Moreno et al., 1993). Filtration was therefore used to examine the motility of aerotolerant C. coli OR12. Duplicate BA plates with centered nitrocellulose membrane filters were inoculated with 50 μl of suspensions containing 7.6 log₁₀ CFU/ml C. coli OR12 WT or 7.6 log₁₀ CFU/ml C. coli OR12 Aer P38. One of each was incubated microaerobically at 42°C and one of each aerobically at 37°C. After 24 h, heavy growth was present in the center of the filtered regions of both plates incubated microaerobically. After 4 days, growth was evident on the aerotolerant C. coli OR12 incubated aerobically, with confluent colonies by 5 days. No growth was noted on the wild type C. coli OR12 incubated aerobically. Under aerobic conditions aerotolerant C. coli OR12 remained motile and metabolically active to be able to translocate the membrane filter.

To measure oxidative stress resistance, suspensions containing between 7.3 log₁₀ and 7.8 log₁₀ CFU/ml C. coli RM2228, C. coli OR12 WT, aerobically prepared C. coli OR12 Aer P40 and microaerobically prepared C. coli OR12 Aer P40 were exposed to different concentrations of H₂O₂. Marked differences in survival were apparent between aerobically prepared aerotolerant C. coli OR12 and the other microaerobically prepared strains (Figure 3).

Transmission electron micrograph (TEM) images are shown in Figure 4. Both wild type and aerotolerant C. coli OR12 contained populations of typical spiral cells as well as filamentous and coccoid forms.

Colonization gavage doses were titrated at 1.38 × 10⁸ CFU for group 1 (C. coli OR12 WT) and 3.4 × 10⁷ CFU for group 2 (C. coli OR12 Aer P34). Caecal Campylobacter counts at 3 and 7 days post gavage are shown in Figure 5. All the birds that received C. coli OR12 WT became colonized, with mean caecal Campylobacter counts of 8.71 (±0.2) and 8.57 (±0.3) log₁₀ CFU/g at 3 and 7 days post gavage, respectively. Of the birds challenged with C. coli OR12 Aer, only 2/7 had detectably colonized by 3 days and 6/8 by 8 days. Birds with no Campylobacter growth were considered ≤2.3 log₁₀ CFU/g, which was the limit of detection. The difference in caecal Campylobacter counts between the wild type and aerotolerant strain was significant at 3 days post challenge (p = 0.006), but not significant by day 7 (p = 0.0659) as determined by the two-tailed Mann–Whitney U test. Standard deviations for individual bird Campylobacter counts generated by triplicate technical replicates were all below 0.19 log₁₀ CFU/g.

All birds from which Campylobacter was detected on CCDA also yielded positive growth on microaerobically incubated membrane filtered BA plates. Colony morphology was consistent with Campylobacter (Supplementary Figure 3) and Gram stains provided confirmation. Of the aerobically incubated membrane filtration BA plates, growth was detected from one of the C. coli OR12 Aer challenged birds at 3 days post challenge and from 3 birds at 7 days post challenge. No growth was detected on any of the aerobically incubated membrane filtered BA plates from wild type challenged birds.

Secondary aerotolerance assays were performed on the campylobacters isolated by microaerobic membrane filtration. All C. coli isolated from birds challenged with C. coli OR12 Aer P34 retained their aerotolerance. Supplementary Figure 3A shows the serially diluted and aerobically incubated C. coli isolated from the two birds 3 days post challenge with C. coli OR12 Aer P34. No such growth was noted from isolates recovered from the seven birds challenged with the wild-type strain.

Macrorestriction patterns of wild type C. coli OR12 and 4 aerobically passaged isolates determined by PFGE were identical to each other (Supplementary Figure 4) and to the original isolate described by El-Shibiny (2006). The development of aerotolerance is not associated with any significant DNA insertion, deletion or genome rearrangement. Therefore whole genome sequencing was performed to examine the occurrence of any point mutations.

A total of 23 mutations were noted between the wild type and aerotolerant C. coli OR12 sequences. There were 11 single nucleotide substitutions, 2 nucleotide insertions and 10 insertions or deletions within homopolymeric GC tracts. The DNA changes within homopolymeric GC tracts are summarized in Table 1 and all others in Table 2. Three of the mutations were silent and one was within a non-coding region. All the predicted protein sequences were analyzed with BLASTP to confirm the protein identity and to determine if the changes induced were present in other similar sequences.

Gene pgk, which encodes the 400 amino acid protein phosphyglycerate kinase is located between nucleotides 380840–382042. BLASTP analysis of the amino acid sequence identified many C. coli and C. jejuni sequences with 99–100% coverage and identity. Aerotolerant C. coli OR12 has an A insertion at nucleotide position 380,879, which results in a frame shift at residue 19 and early protein termination at 21 amino acids. The entire amino acid sequence of the mutant protein is: MSDIISIKDIDLSKKKSFYKM. BLASTP analysis of this sequence did not find any sequences of similar length.

Gene cfiA encodes a 597 amino acid product, which was initially annotated as 2-oxoglutarate carboxylase. Subsequent BLASTP searches identified 99% amino acid sequence identity with many C. coli biotin attachment proteins. The sequence also shares identity with several conserved domains, including phosphate binding superfamily, carboxylase, carboxylase interaction sites and a biotinylation site. The sequence of aerotolerant C. coli OR12 has a T insertion at position 837,990. This results in a frame shift and premature termination at residue 419. BLASTP analysis of the mutant sequence identified no similar sequences of identical length, and lacked part of the carboxylase domain as well as the entirety of the biotinylation and carboxylase interaction sites.

Nucleotide position 455,044 lies within a poly-C tract. In wild type C. coli OR12 this tract is composed of 10 Cs in 83% of the sequence reads, and 9 Cs in 7% of the sequences. The aerotolerant C. coli OR12 sequence has become fixed with 100% 9 Cs at this location. This position is within genes ATE51_00982 and ATE51_00984. In the 10 C configuration, the reading frame ATE51_00982 encodes a 593 amino acid protein that shares high sequence identity with PseD proteins (97 to 100% identity) from many other C. coli and C. jejuni data base entries. The sequence also has 90% coverage and 71% identity with the sequence of PseD from C. jejuni NCTC 11168, which is involved in flagellar modification (McNally et al., 2006). The 9 C configuration, present in the minority of the wild type and all of the aerotolerant C. coli OR12 sequences, results in a frame shift at the fifth amino acid residue and protein termination at 21 amino acids. The 10 C configuration of gene ATE51_00984 encodes a 60 amino acid protein. However, the 9 C configuration results in a frame shift after 60 amino acid residues and fusion with ATE51_00982 resulting in a 648 amino acid protein, which has 99% coverage and 72% identity with the PseD from C. jejuni NCTC 11168.

Gene hxuA encodes a 1063 amino acid protein that was annotated as a heme/hemopexin binding precursor. However, BLASTP analysis of the protein sequence identified many C. coli filamentous haemagglutinin proteins with over 90% coverage and identity. The amino acid sequence also contains a conserved N-terminal haemagglutinin domain, conserved in many proteins with haemagglutination or haemolysin activity. Nucleotide 565,851 lies within a poly-G tract. Of the wild-type C. coli OR12 sequences, 96% of contained 10 Gs and 4% contained 9 Gs. Aerotolerant C. coli OR12 sequences were composed of 75% containing 9Gs and 25% containing 10 Gs. The 10 G configuration results in encoding of the full 1063 amino acid protein; however, the 9 G configuration results in a frame shift at residue 19 and immediate termination.

Gene pepD is a disrupted pseudogene in the majority of C. coli OR12 sequences. Nucleotide 828,036 lies within a poly-C tract. The wild-type C. coli OR12 sequences are composed of: 91% containing 8 Cs and 9% containing 9 Gs. Aerotolerant C. coli OR12 sequences are composed of: 77% containing 10 Gs and 23% containing 9 Gs. The 8 G configuration, which represents the majority of wild-type sequences, encodes a 221 amino acid protein that shares high sequence identity (83% coverage and 100% identity) with many C. coli aminoacyl-histidine dipeptidase proteins that are significantly longer. The 9 C configuration, present in the minority of wild-type and aerotolerant sequences, encodes a 423 amino acid protein, which has high sequence identity (99%) with other C. coli aminoacyl-histidine dipeptidase sequences of the same length. The sequence also shares similarity with a conserved zinc peptidase like superfamily domain. The 10 C configuration, present in the minority of aerotolerant sequences, encodes a 194 amino acid protein, which is identical to the 8 G product apart from the final 9 amino acid residues. Of the protein sequences that show high similarity with the 8 G and 10 G products, none ended in the same location or with similar C-terminal amino acid residues.

The first 192 amino acids of this predicted 205 amino acid protein product of gene ATE51_00954 share 67% identity with the first portion of the adjacent ATE51_00952 and nearby ATE51_00958, both of which are annotated as putative carbonic anhydrases. BLASTP analysis indicates 100% identity with several 196 amino acid proteins in C. coli and C. jejuni also similarly annotated as carbonic anhydrases. Aerotolerant C. coli OR12 has a single G homopolymer deletion at position 483,767 which causes a frame shift at the G196 residue, replacing residues GGHMEDTYLYS with GVIWRIPIFTHSKDSSLVCGWRD at the carboxyl terminus. BLASTP analysis of the 217 aa mutant configuration identified 88–90% identity with one C. jejuni and one C. coli hypothetical protein of 218 aa length, however, the majority of entries demonstrating sequence similarity were for 403–404 aa carbonic anhydrases.

Gene ATE51_00958 encodes a 405 aa protein, annotated as carbonic anhydrase with 97–100% identity with C. jejuni and C. coli carbonic anhydrase proteins of identical length. Aerotolerant C. coli OR12 has a G deletion at nucleotide 484,990 which induces a frame shift at residue 197, altering the aa sequence with a stop at 201 aa. BLASTP analysis of the mutant sequence revealed 97% identity with two C. coli hypothetical proteins of identical length, one of which was in C. coli RM2228.

Nucleotide position 1,158,035 lies within a homopolymeric tract, with wild-type C. coli OR12 sequences consisting of: 90% with 10 Cs; 7% with 9 Cs and 3% with 11 Cs. Aerotolerant C. coli OR12 sequences consist of: 85% with 8 Cs and 15% with 9 Cs. The original annotation of ATE51_02354 indicated a hypothetical protein; however, a more recent annotation of the C. coli OR12 genome within the Genbank database classified the region as a disrupted carbonic anhydrase pseudogene. The 10 C configuration, representing 90% of wild-type C. coli OR12 sequences, results in a predicted 190 aa hypothetical protein, with many highly similar hypothetical proteins in the database. One entry suggests high sequence identity with a partial carbonic anhydrase in C. coli (entry KQH53282.1). The 8 C configuration, representing 85% of aerotolerant C. coli OR12 sequence reads, results in a frame shift and exchange of the terminal GY residue for VLERFCLEIHMADI. This arrangement is also present in several similar sequences in the database, again mostly annotated as hypothetical proteins. The 9 C configuration, present in 7% of wild type and 15% of aerotolerant C. coli OR12 sequence reads results in a frame shift and fusion with the next open reading frame, ATE51_02352, resulting in a 408 amino acid protein. This sequence is also present in the database, with all similar entries annotated as hypothetical proteins. Residues 194–408 have a 99% similarity to a putative carbonic anhydrase domain in C. jejuni DFVF1099 (Takamiya et al., 2011). The 11 C configuration (3% of wild-type C. coli OR12 sequence reads) results in a frame shift similar to the 8 C configuration, but an additional G residue is retained; only the terminal Y amino acid residue is exchanged for the residues VLERFCLEIHMADI. This configuration is less frequently reported in the database than that of the 8C variant. All variants result in high similarity with the conserved domain DUF2920, a bacterial domain of unknown function. The three putative carbonic anhydrase protein sequences above share significant identity at their N-termini, as demonstrated by the alignment in Supplementary Figure 5.

Genes involved in Campylobacter oxidative stress resistance, as cataloged in a recent review by Flint et al. (2016b), were identified in C. coli OR12 and the corresponding protein sequences compared with orthologs from C. jejuni NCTC11168. These data are presented in Supplementary Table 1. All the functional genes cataloged by Flint et al. (2016b) could be identified with the exception of those encoding the regulators of response to peroxide A and B (rrpA and rrpB). No significant sequence similarity to the latter two genes could be identified in the sequences of C. coli OR12 or C. coli RM1875 or C. coli 15-537360.

Several authors have reported aerotolerance of C. jejuni and C. coli, however, the methods adopted to assess this are variable. For example, Jones et al. (1993) report aerobic growth of C. jejuni strains on blood agar after 2–3 days of aerobic incubation, but used an initial acclimation period of 18 h under microaerobic conditions. Similarly, Chynoweth et al. (1998) found 29 of 40 C. jejuni isolates from various sources could be adapted to grow aerobically on nutrient agar. All isolates were grown microaerobically before streaking on nutrient agar, and though some were initially slow to grow, they became adapted following aerobic subculture. Gaynor et al. (2004) reported the aerobic passage of C. jejuni NCTC 11168 on blood agar with antibiotic supplement at 37°C, which resulted in significant attenuation of colonization in a day-old chick model after thirteen passages.

In contrast, Oh et al. (2015) examined aerotolerance of 70 C. jejuni isolates obtained from chicken meat samples with respect to survival of approximately 10⁹ CFU/ml in aerobic conditions at 42°C in MHB. Of these, 50 were considered aerotolerant based on survival for 12 h and 25 classified as hyper-aerotolerant, with survival for 24 h. The importance of alkyl hydroperoxide reductase was demonstrated by mutation of ahpC leading to the loss of hyper aerotolerance (Oh et al., 2015). Purdy et al. (1999) studied the effect of sodB mutation on aerotolerance and in vivo colonization fitness of C. coli. Inactivation of sodB activity did not affect survival of cultures aerobically incubated at 25°C in MHB, however, the colonization of day old chicks was reduced (Purdy et al., 1999). The aerotolerance test used by Purdy may not be equivalent to that described by Oh et al. (2015), as oxidative stress has been demonstrated to have a significantly greater effect at 42°C that at 25°C (Garénaux et al., 2008).

Kaakoush et al. (2007) examined the growth at several cell densities of four C. jejuni strains from aerobically and microaerobically incubated brain heart infusion broth. High initial bacterial density cultures grew when incubated either aerobically or microaerobically; however, cultures with less than 10⁴ CFU/ml only grew under microaerobic conditions. A combination of lower oxygen saturation in liquid media at 42°C and dense cultures that have a greater oxygen demand will lower the oxygen tension and allow growth. No growth occurred when C. coli OR12 Aer or WT were aerobically incubated at 37°C in MHB; indeed similar decline rates were observed for the C. coli OR12 derivatives and C. coli RM2228.

Rodrigues et al. (2015) recently reported the aerobic growth of C. jejuni strain Bf on Karmali agar. For this strain visible growth was reported within 24 h of aerobic incubation at 42°C. Similar to C. coli OR12 Aer, growth in liquid medium could not be elicited, and a quantitative assay was used to measure growth on solid medium. Superior resistance to peroxide stress was also noted compared to C. jejuni NCTC 11168, which had failed to grow aerobically (Rodrigues et al., 2015). Growth on solid medium creates oxygen gradients, such that facultative anaerobes have been found to be more active on the bottom layer and aerobes more rapidly dividing on the top layer of colonies (Reyrolle and Letellier, 1979). Oxygen gradients in bacterial colonies and solid media have also been measured analytically, with the top 30 μm of a gelatine medium being considered aerobic (Jeanson et al., 2015). Therefore, once initial growth of Campylobacter has commenced on a solid medium, bacteria that are not directly exposed to the outside atmosphere may reside in an atmosphere more similar to microaerobic conditions.

Chicken colonization studies confirmed that wild type C. coli OR12 is a robust colonizing strain, with reproducible caecal counts between 10⁸ and 10⁹ CFU/g. However, the aerotolerant strain colonized only 2/7 birds at 3 days and 6/8 birds by 7 days. Birds that were colonized had relatively high counts, with all >7 log₁₀ CFU/g. The binary nature of the outcomes suggests that bottlenecking events and stochastic processes are significant factors in the colonization process. They are also suggestive of adaptation occurring between days 3 and 7, where campylobacters, presumably present at levels below the limit of detection, have become acclimated to the intestinal or caecal environment and proliferate to achieve higher colonization levels. These adaptations may not be identical between birds, as suggested by the wider range of caecal counts compared with the original isolate. This experiment further validates the importance of individual caging of chickens during Campylobacter colonization assays as advocated by Conlan et al. (2011). Had the birds been group housed, any chickens in which the aerotolerant C. coli OR12 had colonized would likely have shed the modified strain in large numbers in feces, potentially rapidly colonizing the other birds in the group. The reduction in initial colonization fitness could have been missed under these circumstances. A reduction in the challenge dose could potentially allow further resolution between groups and may be of value in future studies. Aerotolerant C. coli OR12 recovered from chickens were still capable of aerobic growth demonstrating the trait is stable and compatible with chicken colonization, albeit potentially at a lower initial level. However, it is difficult to distinguish the lower levels of colonization from the overall effect of in vitro passage, which is known to reduce the colonization fitness (Stern et al., 1988). Nevertheless these findings correlate with the observations of Jones et al. (1993) in that aerotolerant C. jejuni recovered from mice retained aerotolerance. The evident adaptation between days 3 and 7 did not lead to any loss of aerotolerance. The ability of C. coli OR12 to adapt to aerobic conditions and re-adapt to chicken colonization may be a driver of diversity within the Campylobacter population, where the plastic genome allows rapid adaptation to new or stressful conditions. Given the capacity of campylobacters for inter-strain recombination these changes may then be disseminated amongst cohabiting Campylobacter populations. The ability to withstand, and grow, in aerobic conditions could be of major importance not only in environmental persistence, allowing fomite transfer between farms, but could also allow prolonged survival on processed chicken carcasses and therefore be of public health importance. Survival of C. coli OR12 on chicken skin following refrigeration and freezing was examined by El-Shibiny et al. (2009). Survival rates were comparable with the other C. jejuni and C. coli strains tested. Similar experiments with aerotolerant C. coli OR12 could provide valuable insight as to whether improved aerotolerance in vitro translates to superior survival on product, which would be of public health significance. This may be further considered in the context of the recently described aerobic C. jejuni strain Bf, which was found to be able to adhere to human epithelial cells and form biofilms under aerobic conditions (Bronnec et al., 2016).

This report describes true aerobic growth, without initial support from microaerobic incubation. With the exception of C. coli OR12, no growth was noted from any aerobically incubated Campylobacter spp. in this study. However, the equality of growth following aerobic or microaerobic incubation, which had been reported by Jones et al. (1993) was not achieved with C. coli OR12. Nevertheless the ability to undergo direct growth in atmospheric oxygen, a trait that was retained post either microaerobic growth or in vivo passage through chickens, encouraged us to examine the genetic basis for C. coli OR12 aerotolerance.

Oxidative stress resistance has been proposed as a key feature of aerotolerance. Almost all the genes whose products have been implicated in oxidative stress resistance could be identified in C. coli OR12, none contained any mutations in the aerotolerant C. coli OR12. Orthologs of rrpA and rrpB, encoding MarR-like transcription factors implicated in oxidative stress resistance, could not be identified in the genome sequences of C. coli OR12 or RM1875 or 15-537360 (Gundogdu et al., 2015). Bronnec et al. (2016) identified a complete type VI secretion system in C. jejuni Bf and suggested that it may contribute to the aerotolerant phenotype. The C. coli OR12 genome contains a complete T6SS, although C. coli RM2228, which did not demonstrate aerotolerance, also possesses a full T6SS (Bleumink-Pluym et al., 2013). This would suggest that possession of T6SS alone does not confer aerotolerance, although it could be involved when in combination with other features. Changes in 23 coding sequences were observed between the wild type and aerotolerant C. coli OR12 genomes. The functional significances of these are discussed in the following paragraphs.

Phosphoglycerate kinase (PGK) is a magnesium dependent kinase involved in glycolysis and gluconeogenesis. In many organisms it performs the first ATP- generating step in glycolysis, however, C. jejuni lacks a homolog of 6-phosphofructokinase and therefore cannot metabolize hexose sugars (Velayudhan and Kelly, 2002). Under these circumstances C. jejuni PGK is likely to be limited to the gluconeogenic pathway, converting 3-phosphoglycerate to 1,3-diphosphoglycerate (Figure 6).

The aerotolerant C. coli OR12 sequence has an insertion early in the pgk gene that causes a frame shift and termination at 21 amino acids. No other copies of pgk are present, so it is likely that no functional PGK protein is produced. This may prevent the gluconeogenic pathway from occurring in aerotolerant C. coli OR12, so sugar anabolism must be occurring via different pathways. The C coli OR12 strain does possess orthologs to the serABC genes present in C. jejuni NCTC11168 (ATE51_01764; ATE51_03664; ATE51_03558, BLASTP identities of 94, 79, and 87%, respectively). Therefore, serine biosynthesis from 3-phosphoglycerate should be unaffected (Velayudhan et al., 2004). Circumvention of PGK and reinstatement of gluconeogenesis is theoretically possible via the formation of 2,3-bisphospho-D-glycerate, which would require the enzyme bisphosphoglycerate mutase. A putative phosphoglycerate/bisphosphoglycerate mutase has been annotated for C. jejuni NCTC11168, for which an ortholog is present in C. coli OR12 (ATE51_01624). The accumulation of 3-phosphoglycerate in the absence of PGK could drive this alternative reaction series. Aerobic growth of C. coli OR12 on BA plates prompted this study, and it is worth noting that erythrocytes are estimated to contain 4 to 5 mM 2,3-bisphospho-D-glycerate formed by the Luebering-Rapoport glycolytic shunt, which is used to promote the release of oxygen from deoxygenated hemoglobin at times of need by binding the T-state as an allosteric effector (Benesch and Benesch, 1967). The availability of 2,3-bisphospho-D-glycerate, albeit at declining concentrations, could be utilized by campylobacters to reduce or remove the requirement of PGK in gluconeogenesis and increase the affinity of any remaining hemoglobin to bind oxygen. However, the adaptation present in the aerotolerant C. coli OR 12 also enables aerobic growth on CCDA. Modifications to central metabolism are emerging amongst campylobacters, for example it was recently reported that 1.7% of thermophilic campylobacters have the potential to metabolize glucose via the Entner-Doudoroff pathway (Vegge et al., 2016). This pathway also permits the metabolism of L-fucose, which has been reported to provide a competitive advantage to C. jejuni in a piglet colonization model (Stahl et al., 2011).

The cfiA gene in the wild-type C. coli OR12 sequence encodes a putative biotin attachment protein. The protein sequence shares 85% identity with 100% coverage of the pyruvate carboxylase B subunit of C. jejuni NCTC 11168. Therefore, this protein is likely to be the ortholog of PycB, encoded by the gene cj0933c in C. jejuni NCTC 11168. Pyruvate carboxylase in C. jejuni is composed of two subunits, PycA and PycB (Velayudhan and Kelly, 2002). These correspond with the biotin carboxylase and biotin carboxyl carrier subunits of a two-subunit type of carboxylase (Goss et al., 1981). The complete enzyme contains biotin and catalyzes the ATP-dependent carboxylation of pyruvate, yielding oxaloacetate. Velayudhan and Kelly (2002) performed mutation studies of pycB and pycA, and they noted that mutation of pycB resulted in significantly reduced growth in BHI broth, and the inability to use lactate, pyruvate or malate as carbon sources.

The C. coli OR12 Aer sequence contains a T insertion at nucleotide position 837,990 compared to the wild-type sequence. This would result in a frame shift and premature termination at 419 amino acids. This is likely to affect the activity of the enzyme as it is missing almost one third of its amino acids, including part of the conserved carboxylase domain as well as the entirety of the putative biotinylation and carboxylase interaction sites. This is surprising as the aerotolerant C. coli OR12 has a growth rate very similar to the wild type when grown microaerobically in NB2. Colonies on solid medium from microaerobically incubated aerotolerant C. coli OR12 were slightly smaller than the wild type, but this is not similar to the pycB mutant which had a threefold reduction in growth compared to the wild type (Velayudhan and Kelly, 2002). Comparisons of carbon metabolism in C. jejuni and C. coli indicate these species share a core set of carbon sources including amino acids, citric acid cycle intermediates and carboxylic acids (Wagley et al., 2014). However, exceptions were noted such as the inability of C. coli strains to metabolize D-malic acid and the inability of C. jejuni strains to metabolize propionic acid.

Post-translational glycosylation of flagellar proteins is important for Campylobacter virulence. For example, mutation of pseA in C. jejuni 81-176 resulted in deficiency of flagellar glycosylation with pseudaminic acid and reduced virulence in a ferret model (Guerry et al., 2006). The function of the gene pseD, cj01333 in C. jejuni NCTC 11168, which is a member of the motility accessory family, was examined in the context of flagellar glycosylation in C. jejuni 81-167 by McNally et al. (2006). Mutation of pseD did not affect cell motility or flagellin glycosylation with pseduaminic acid. However, mutation did prevent glycosylation with the pseudaminic acid acetamidino derivative PseAm, though it did not prevent cellular production of PseAm. Changes within a homopoylmeric C tract of pseD in aerotolerant C. coli OR12 effectively reinstates a longer reading frame to fuse ATE51_00982 and ATE51_00984 to create a 648 amino acid predicted product that exhibits 99% coverage with the C. jejuni NCTC 11168 PseD.

The gene hxuA encodes a protein which shares high sequence identity with other C. coli proteins annotated as filamentous haemagglutinin. The sequence also contains a region with high similarity to several conserved haemagglutination activity domains. Filamentous haemagglutinin proteins are involved in virulence of several bacterial pathogens, including Bordetella pertussis and Avibacterium (formerly Haemophilus) paragallinarum (Hobb et al., 2002; Inatsuka et al., 2005). The genome of Arcobacter butzleri ATCC 49616 contains the gene hcpA which encodes an iron activated filamentous hemagglutinin (Douidah et al., 2012). BLASTP comparison of this protein with C. coli OR12 HxuA revealed no significant similarity, nor did comparison with the C. concisus HcpA, the HagA protein of A. paragallinarum, or the Bordetella proteins FhaA and FhaB. C. fetus subsp. fetus 82-40 encodes a putative filamentous haemagglutinin of the HcpA family. It may be coincidental that this species is associated with bacteraemic infections, however, it is unknown whether this potential virulence factor is functional in vivo (Miller, 2008). The protein may have a possible role in iron acquisition or metabolism, which is known to be involved in the oxidative stress response (Atack and Kelly, 2009). The function of the putative filamentous haemagglutinin encoded by hxuA in C. coli OR12 remains obscure but the reading frame is notable in that it is phase variable, being intact in 96% of the wild-type sequence reads but appears disrupted in aerotolerant C. coli OR12.

Carbonic anhydrase enzymes catalyze the reversible hydration of CO₂ to bicarbonate and can be divided into at least six different phylogenetic groups, with varying expressions across kingdoms (Bury-Moné et al., 2008; Al-Haideri et al., 2015). These groups have similar function but differing structures, which are believed to have arisen by convergent evolution (Smith and Ferry, 2000). Strains of C. jejuni are reported to encode a β and a γ carbonic anhydrase, the former class are oligomeric, containing between 2 and 8 monomers and the latter are homotrimers (Smith and Ferry, 2000). However, the putative carbonic anhydrase annotated genes affected in the aerotolerant C. coli OR12 are not similar to those that have been investigated in campylobacters thus far. Whether these represent a different class of carbonic anhydrases not yet characterized in Campylobacter, or encode a hitherto unknown function requires further investigation.

Eleven other genes contained coding mutations, all of which involved single nucleotide substitutions or changes of up to three amino acids at the C-terminus. None of these involved genes which could be expected to influence aerotolerance, such as those encoding oxidative stress resistance proteins or their regulatory genes. Further investigation of these changes would include protein modeling software to determine whether the changes induce significant structural alterations or cause relocation of the protein.