Occurrence of Campylobacter spp. in Swedish calves, common sequence types and antibiotic resistance patterns

Abstract Aims Cattle are the second most important cause of human campylobacteriosis, after poultry, but there are knowledge gaps regarding Campylobacter in cattle. This study examined the occurrence of Campylobacter, the species present, sequence types and antibiotic resistance in Swedish cattle. Methods and Results Faeces samples collected from 154 calves on seven Swedish farms, and 69 follow‐up samples from a second collection occasion, were analysed. Campylobacter were isolated from 77% of calves at the first sampling, with Campylobacter jejuni as the most frequently isolated species. Animals kept on deep straw bedding were less likely to be colonized with Campylobacter. Whole‐genome sequencing of 90 C. jejuni samples resulted in 11 sequence types, among which ST‐19 and ST‐21 were most frequent. Antimicrobial resistance analyses showed that 46% of 142 isolates analysed were resistant to quinolones, while all isolates belonging to ST‐19, ST‐22 and ST‐441 were resistant to ciprofloxacin and nalidixic acid. Conclusions Campylobacter jejuni was the species most frequently isolated in calves and a strong association was found between sequence type and antimicrobial resistance pattern. Significance and Impact of the Study The high proportion of calves with quinolone‐resistant Campylobacter jejuni should be considered in a One Health perspective.


Introduction
Campylobacteriosis is the most frequently reported zoonosis in many countries. There were almost 250 000 confirmed cases in Europe in 2018, representing more than 50% of all human cases of zoonotic infections reported in Europe (EFSA and ECDC 2019). However, these figures are likely to be underestimates and the true incidence is probably higher (Boqvist et al. 2018). Most cases of campylobacteriosis are sporadic, but outbreaks can occur. Poultry is an important reservoir of Campylobacter jejuni, and consumption and handling of broilers or broiler meat pose high risks of human campylobacteriosis (Rosner et al. 2017;Berthenet et al. 2019;Cody et al. 2019;EFSA and ECDC 2019). However, there are other sources of human campylobacteriosis (Sheppard et al. 2009;Mughini Gras et al. 2012;EFSA and ECDC 2019). For example, 54% of strong-evidence outbreaks in the EU in 2017 were reported to be caused by milk, while in the US the most commonly identified sources of campylobacteriosis outbreaks 2010-2017 were milk-associated (EFSA and ECDC 2018;CDC 2020). Outbreaks of campylobacteriosis have also been reported among cattle farm workers and visitors (Gilpin et al. 2008;Heuvelink et al. 2009;Lahti et al. 2017b). Cattle were identified as a Campylobacter reservoir for 21-55% of human cases in the Netherlands and France (Mughini Gras et al. 2012;Th epault et al. 2018b), and as the second most important cause of human campylobacteriosis in Denmark (Boysen et al. 2014).
Cattle are asymptomatic carriers of thermotolerant Campylobacter and may shed the bacteria intermittently in the faeces (Hakkinen and H€ anninen 2009;Ramonait_ e et al. 2013;Tang et al. 2017). This means that Campylobacter spp. can easily contaminate the udder and milk (Bianchini et al. 2014;Arthursson et al. 2018;Hansson et al. 2020). However, there are still knowledge gaps regarding the epidemiology of Campylobacter spp. in cattle, for example, only eight EU member states reported monitoring data for Campylobacter in cattle in 2018 (EFSA and ECDC 2019). To our knowledge, there is no coordinated monitoring of Campylobacter spp. in cattle in any country.
Multi-locus sequence typing (MLST) has become the standard for genetic analyses, e.g., to study transmission and risk factors (Lahti et al. 2017a(Lahti et al. , 2017b. There is a strong host-genotype association of multi-locus clonal complex (CC), sequence type (ST), and allele level, particularly within C. jejuni, which has a greater host signal than geographical signal (Sheppard et al. 2010). Most MLST analyses of C. jejuni in Sweden have been performed on isolates from chickens and humans. There is thus insufficient knowledge of common sequence types in isolates from cattle, and of the importance of cattle for campylobacteriosis in humans .
There are also challenges with antimicrobial resistance (AMR) in Campylobacter spp., as reflected, for example, in increased resistance to quinolones such as ciprofloxacin and nalidixic acid (Riley et al. 2015;EFSA and ECDC 2017;Tang et al. 2017;CDC 2020). This problem has been highlighted by the WHO (2020). Ciprofloxacin resistance tended to increase over time among international travellers tested between 2007 and 2014 (Post et al. 2017). In Sweden, resistance to ciprofloxacin in Campylobacter isolated from humans increased from 14% in 2014 to 61% in 2019 (Swedres-Svarm 2019). In Swedish broilers, annual resistance of C. jejuni to fluoroquinolone varied between 4 and 24% during 2010-2018, although records of antibiotic sales show that fluoroquinolones were not used in commercial chicken production during this period (Swedres-Svarm 2019). The low fluoroquinolone use is likely due to the ban on administering the antibiotic via feed or water, which makes distribution to poultry difficult. However, individual animals can still be treated under certain circumstances, for example, in dairy cattle there were 0Á14 fluoroquinolone treatments per 100 completed/interrupted lactations in 2018 (Swedres-Svarm 2019). However, there is limited information on AMR in Campylobacter strains isolated from cattle in Sweden.
To fill some of the knowledge gaps highlighted above, it is imperative to determine the occurrence and AMR patterns of Campylobacter spp. isolated from cattle, since this species can play an important role in the epidemiology of human campylobacteriosis. The aim of this study was thus to provide bacteriological and epidemiological knowledge on the occurrence, species, and sequence types of Campylobacter spp. in cattle, and to increase understanding of the AMR pattern of Campylobacter spp. isolates from Swedish cattle.

Study design and study population
This study was part of a larger study analysing Shiga toxinproducing Escherichia coli (STEC) in Swedish calves, which was performed on seven dairy farms where presence of STEC had been confirmed through environmental sampling. Details of farm selection can be found in Tamminen et al. (2020). Five farms (A, B, C, D, G) were located on the island of € Oland, Farm E in southern Sweden (Sk ane) and Farm F in the south-eastern county of Sm aland (Fig. 1). Farm size (total number of cattle) varied between 130 and 600 animals (Table 1). Faeces samples from 154 calves aged between 8 and 306 days (mean 113 days) were collected between April and November 2016. A detailed description of selection of animals for individual sampling can be found in Tamminen et al. (2020). In short, up to 26 animals per farm were selected by systematic random sampling in pens where STEC had been detected. If the pens on the farm contained fewer than 20 animals, all were sampled.
On six farms, a second sampling was performed 4-5 weeks after the first sampling. On this occasion, animals from which STEC had been isolated at the first sampling were included, together with 2-3 previously negative controls. Farm E was only visited once, since no animals tested positive for STEC in the first sampling (Fig. 2). Sampling of animals was performed in accordance with ethical approval granted by the regional ethics committee (Uppsala Djurf€ ors€ oksetiska N€ amnd, Dnr: C 85/15).

Collection of faecal samples
Faecal samples from each animal were obtained from the rectum and transferred to 100 ml plastic jars. A new and clean pair of gloves were used for each sample and the jars were filled to a maximum of two-thirds, in order to decrease the risk of the lid opening during transport. The samples were transported chilled to the National Veterinary Institute (SVA, Uppsala, Sweden). All packages reached the laboratory within 48 h. The samples were analysed by direct culture according to ISO 10272: Part 1C (2017). In brief, faecal contents were spread on modified charcoal-cefoperazone-deoxycholate agar (mCCDA) (Oxoid, Basingstoke, UK) and the plates were incubated at

Species identification
Suspected Campylobacter colonies were re-cultured on horse blood agar (SVA, Uppsala, Sweden) and incubated in microaerobic atmosphere at 37 AE 1°C for 44 AE 4 h. If suspected colonies had different macro-morphological appearance, 2-3 isolates were re-cultured for identification. Genus and species identification were performed from purified colonies on blood agar by matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS), using a Microflex LT MALDI-TOF mass spectrometer (Bruker Daltonics, Billerica, MA). Identification of Campylobacter fetus to subspecies were performed by sequencing, due to difficulties to distinguish between Campylobacter fetus subsp. fetus and Campylobacter fetus subsp. veneralis by MALDI-TOF MS. At least one colony from each positive sample was stored in glycerol broth (15% glycerol and 85% serum broth) at À70°C.

Antimicrobial susceptibility testing
Susceptibility to selected antimicrobial substances was assessed with VetMIC TM panel analysis systems: Camp EU, version 2013-10 (SVA, Sweden), determining the antimicrobial minimum inhibitory concentration (MIC) against ciprofloxacin, erythromycin, gentamycin, nalidixic acid, streptomycin and tetracycline. Multi-drug resistance was defined as resistance to three or more antibiotic classes. For example, resistance to ciprofloxacin and nalidixic acid was considered resistance to one antibiotic class (quinolones). Susceptibility testing was performed on 142 strains of C. jejuni. Not all strains isolated could be tested, because not all survived at À70°C. Reference strains of C. jejuni (CCUG 33560) were used as controls. Epidemiological cut-off (ECOFF) values for determining susceptibility were obtained from the European Committee on Antimicrobial Susceptibility Testing (EUCAST) website https://www.eucast.org/mic_distributions_and_ ecoffs/. The ECOFF values classify isolates with acquired reduced susceptibility as 'non-wild type'. In this paper, non-wild type isolates are called 'resistant', in agreement with the Swedish Veterinary Antibiotic Resistance Monitoring report (Swedres-Svarm 2018). This classification is relevant for monitoring purposes, but it should be understood that resistance defined in this manner does not always refer to clinical resistance.

Whole-genome sequencing
Whole-genome sequencing (WGS) was performed on 90 C. jejuni isolates with at least seven isolates from each farm, including calves found to be colonized with C. jejuni on the two different sampling occasions. In order  to increase the chances of getting different types of sequences from each farm, strains with at least one titre level difference of at least one antibiotic were chosen. All C. jejuni isolates were subcultured twice from single colonies on horse blood agar plates (SVA B341180; National Veterinary Institute) for 48 h at 41Á5°C in microaerobic atmosphere, prior to DNA extraction. DNA was extracted using the EZ1 DNA Tissue Kit and the bacterial protocol for an EZ1 Advanced XL instrument (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The DNA was eluted in 100 µl elution buffer from the kit and quantified using the Qubit ds DNA High Sensitivity Assay Kit on a Qubitâ 2.0 Fluorometer (Invitrogen, Carlsbad, CA). Sequencing libraries were prepared using the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA). The libraries were then sequenced on an Illumina NextSeq 500 system with 2 9 150-bp pairedend reads, using the NextSeq 500/550 Mid Output kit V2.5. The sequence reads generated were analysed using the Ridom SeqSphere + v6.0.9 software (Ridom GmbH, M€ unster, Germany). The genomes were assembled de novo using SKESA (Souvorov et al. 2018), through a pipeline script in Ridom SeqSphere+, with an average mean genome coverage of 1839. The MLST profiles were assigned using the scheme available at https://pubmlst. org/campylobacter/ (Jolley et al. 2018) through a C. jejuni/coli MLST task template in Ridom SeqSphere+. Core genome MLST (cgMLST) analysis was performed using the C. jejuni/coli cgMLST task template v1.3 in Ridom SeqSphere+, which contains 637 loci. A minimum spanning tree (MST) based on these 637 loci was generated in Ridom SeqSphere + using default parameter

Epidemiological determinants
At the time of the first sampling, information about the age of the sampled animals was collected from farm records. In addition, pen size was measured using a laser telemeter and the number of animals in each pen was noted. Each pen was also categorized according to type (group pen with deep straw bedding, pen with straw/sawdust bedding, or other, for example, slatted floors or free-stall with cubicles). Hygiene (faecal contamination and wetness) was scored between 1 and 3 (1 = limited faeces visible, dry bedding, 2 = fecal contamination of bedding material clearly visible and/or bedding wet in part of the pen, 3 = faecal contamination visible and/or bedding wet in the whole pen).

Statistical analysis
Differences in resistance to nalidixic acid and ciprofloxacin between farms were analysed by Fisher's exact test, performed using a statistical program on the internet website 'Social Science Statistics' (https://www.socscistatis tics.com). A probability level of P < 0Á05 was considered statistically significant. Statistical analysis of the association between determinants and the dependent variable (calf testing positive or negative for Campylobacter spp.) was performed in R Statistical Software (R Core Team 2018). Univariable analysis of age and pen-level risk factors was performed using Fisher's exact test (categorical variables) and the Kruskal-Wallis rank sum test (continuous variables), using the package 'tableone' (Yoshida and Bohn 2019). This was followed by multivariable analysis by generalized logistic regression using lme4 (Bates et al. 2015). Numerical variables were scaled and centred before inclusion, and pen ID was included as a random effect. Variance inflation was investigated using the package 'car' (Weisberg and Fox 2011), with variance inflation >2Á5 considered to indicate variance inflation. The model was reduced using likelihood ratio test and confounding was evaluated by re-introducing each variable to the final model. Possible nonlinear associations of numerical variables were investigated using a generalized additive model before model reduction (Wood and Scheipl 2017). The outcome from STEC O157:H7 sampling (calf positive or negative) was forced into the final model to investigate the risk of bias due to a selection process based on presence of STEC O157:H7 in the environment.

Results
Occurrence of Campylobacter spp.

Associations between Campylobacter spp. and epidemiological determinants
The univariable analysis showed significant differences in age, number of animals in a pen, pen size, and pen type between calves testing positive for Campylobacter spp. and those testing negative (Table 3). Pen hygiene and number of animals in a pen introduced variance inflation in the multivariable model, most likely due to high correlation with pen type and pen size, and were excluded from the final model. After model reduction, pen type was the only variable left in the model. The results indicated that calves in pens with straw/sawdust bedding or concrete/rubber surfaces were more likely to test positive for Campylobacter spp. (odds ratio (OR) = 8 and 11  respectively; P = 0Á018) than calves in pens with deep straw bedding. Re-introduction of removed variables indicated that some of this effect may have been due to pen size and number of animals in a pen, since these variables influenced the estimates of pen type. On accounting for pen type, young age was not associated with presence of Campylobacter spp., but it should be noted that there may be systematic differences in age between pen type (e.g. calves kept on straw were younger than calves on deep straw bedding), which may have biased this result. The generalized additive model did not indicate nonlinear associations between presence of Campylobacter spp. and the determinants. Univariable and multivariable analysis showed no association between Campylobacter spp. and STEC O157:H7.

Antimicrobial susceptibility
Antimicrobial resistance to ciprofloxacin was detected in 66 (46%) of the 142 C. jejuni isolates that could be tested. No resistance to either gentamycin or erythromycin was recorded (Table 4), and none of the 142 isolates showed multi-drug resistance. There was variation between the farms regarding AMR of the C. jejuni strains detected. The MIC distributions for nalidixic acid and ciprofloxacin differed significantly between herds (P < 0Á001) ( Table 4). All 20 strains isolated from the calves on farm E were resistant to ciprofloxacin and nalidixic acid. In contrast, only one of the 19 strains of C. jejuni isolated from the calves on farm F showed resistance to ciprofloxacin and nalidixic acid, while the remaining 18 strains were sensitive to all antimicrobials tested (Table 5). There was no association between quinolone resistance (resistance to ciprofloxacin and nalidixic acid) and pen type or animal age.

Whole-genome sequencing
The two most frequently isolated sequence types were ST-21 (29%) and , both belonging to clonal complex CC-21 (Table 6). For one of the farms (E), all seven sequenced isolates belonged to ST-19, whereas 2-4 STs were identified on the other six farms (Fig. 4).  Among the 69 calves that were sampled twice, C. jejuni were isolated from 31 calves on both occasions. Fourteen of these C. jejuni had the same sequence type and resistance pattern on both sampling occasions, five of the calves had C. jejuni with the same sequence type but different resistance pattern, and the isolated C. jejuni were of different sequence types in 12 of the calves.
A distinct relationship was observed between sequence type and AMR pattern, since the eight different sequence types, with one exception, showed exactly the same pattern (Fig. 4). Of the 20 strains belonging to 18 were resistant to nalidixic acid and ciprofloxacin, one to nalidixic acid, ciprofloxacin and streptomycin, and one to nalidixic acid, ciprofloxacin and tetracycline. . The experimental set-up was designed to analyse occurrence and shedding of STEC, which might potentially introduce sampling bias in analysis of Campylobacter spp. However, there was no association between STEC status and presence of Campylobacter spp., and the results on occurrence of Campylobacter spp. correspond to those presented in other studies. In this study, C. jejuni was identified in 67% of the samples, C. hyointestinalis in 5%, and C. fetus subsp. fetus in 4%. These proportions are in agreement with results presented in other studies, in which C. jejuni was the dominant species (20-68%), followed by C. coli (0-24%), C. hyointestinalis (0-11%), C. lari (0-1%), and C. fetus subsp. fetus (0-1%) (Hakkinen et al. 2007;Ramonait_ e et al. 2013;Th epault et al. 2018a;Hansson et al. 2020). Calves kept on deep straw bedding were less likely to be colonized with Campylobacter spp., despite these pens often being associated with poor hygiene. However, pen type was also associated with calf age, pen size and number of animals in the pen, and the effects of these variables cannot be separated due to confounding. Thus, more studies on how environmental and management factors influence presence and survival of Campylobacter spp. are needed to reveal the underlying causal relationship, and to estimate the potential for using management-and environmentrelated measures to reduce presence of the pathogen.
In this study, only one colony from each sample was subcultured for species identification. However, in faeces samples from three calves, more than one Campylobacter species was isolated. It is likely that more calves were colonized with at least two different Campylobacter species   (Harvala et al. 2016). In a study by Jaakkonen et al.(2019), some C. jejuni strains from cattle, such as ST-883 and ST-1080, persisted for at least 11 months, whereas other C. jejuni types were found sporadically. The longest time interval between repeated sampling occasions in the present study was 34 days, and the same STs were found on both occasions in 14 calves.
There was high similarity in terms of both ST and AMR between the farms. This was partly expected, considering that five of the farms included in the study were situated on the island of € Oland and were only 4Á8-36 km apart (Fig. 1). Between-farm contacts (such as transporting animals together, animal contact on pasture, sharing manure spreader, and farmers visiting each other) are also common on the island . Additionally, € Oland hosts many migratory wild birds that could be colonized with Campylobacter and contribute to bacterial spread to other individuals (Broman et al. 2004;S€ oderlund et al. 2019). Migratory wild birds can also transfer antimicrobial resistance through horizontal gene transfer (Sj€ olund et al. 2008). € Oland is also a popular recreation area with a large number of visitors that may contribute to indirect spread of Campylobacter by transferring cattle or bird faecal material on, for example, boots and vehicles. However, the strains isolated from the two farms located in other regions of Sweden (E, F) were closely related to the strains isolated from the € Oland farms, indicating that geographical distance is not associated with genetic distance of Campylobacter. Thus, transmission routes other than between-farm spread may be important to consider in future studies.
The occurrence of ciprofloxacin and nalidixic acid resistance in this study was remarkably high (46% and ). The high level of resistance to quinolones is difficult to interpret, but could be due to the hyper-mutable nature of Campylobacter. Point mutations in the quinolone resistance-determining region of gyrA gene are most often responsible for resistance to fluoroquinolones (Payot et al.2006). The strong association that was found between specific genotypes and resistance to antimicrobials have also been observed in studies of chicken isolates of C. jejuni (Habib et al. 2009;Wirz et al. 2010). Previous studies in Sweden of C. jejuni isolates from faeces from healthy cattle, sampled at slaughter during four different years between 2001 and 2015, showed annual resistance of 2-21% to ciprofloxacin and 2-23% to nalidixic acid (Swedres-Svarm, 2018). The difference between the studies is difficult to explain, but could be due to different sample sizes and sampling frames. For instance, the cattle in the present study were younger and only from dairy farms, whereas those studied by Swedres-Svarm (2018) included older animals and cattle from both dairy and beef farms. The use of quinolones in livestock production is restricted under Swedish Board of Agriculture regulations on medicine and drug use (SJVFS 2013:42). The main clinical indication for treatment of cattle with quinolones is mastitis (SVS 2015), and obviously dairy cows are more likely to suffer from mastitis than beef cows. This could be another explanation for the different results obtained in this and previous studies. Additionally, different age groups are present more often in dairy herds compared with beef herds, and transmission of resistant bacteria to younger animals through the environment could be facilitated, unless an all in-all out system is used. None of the isolates was resistant to macrolides (erythromycin), the drug of choice for treatment of human campylobacteriosis. Low resistance to macrolides has also been found in the Swedish human population, with <1% of C. jejuni isolates from humans being resistant to erythromycin (Swedres-Svarm, 2019). Antibiotic resistance, particularly to quinolones and macrolides, in thermotolerant Campylobacter spp. is considered a serious threat to public health, as clinical treatment of campylobacteriosis may require use of those antibiotics. A high proportion of cattle colonized with quinolone-resistant C. jejuni could result in continuous contamination of the environment and food products, which should be considered in a One Health perspective.