Prevalence and risk factors associated with Campylobacter spp. occurrence in healthy dogs visiting four rural community veterinary clinics in South Africa

Reports on the occurrence of Campylobacter spp. in dogs in South Africa are non-existent. This study investigated the prevalence of Campylobacter spp. in 481 dogs visiting four rural community veterinary clinics in South Africa. Dogs were screened for Campylobacter spp. by culture and polymerase chain reaction (PCR), and logistic regression analysis was performed to assess the association between sex, clinic, breed and age and the occurrence of Campylobacter spp. in dogs. The prevalence of Campylobacter spp. was 41.50% (95% confidence interval [CI], 37.39% – 46.04%). Campylobacter jejuni, C. upsaliensis and C. coli were detected in 29.31% (95% CI, 25.42% – 33.54%), 13.10% (95% CI, 10.37% – 16.42%) and 5.41% (95% CI, 3.71% – 7.82%) of dogs, respectively. Dogs carrying more than one species of Campylobacter spp. accounted for 6.23% (95% CI, 4.40% – 8.78%). Campylobacter upsaliensis and C. jejuni were detected in 3.74% (95% CI, 2.37% – 5.86%), whereas C. coli and C. jejuni were found in 2.49% (95% CI, 1.42% – 4.34%) of dogs. Age and clinic were the risk factors significantly associated with Campylobacter spp. occurrence, while age, breed and clinic were predictors of C. jejuni carriage. Furthermore, age was the only risk factor associated with a higher likelihood of carrying C. upsaliensis. The prevalence of Campylobacter spp. C. jejuni and C. upsaliensis increased significantly as dogs grew older. In addition, the odds of carrying Campylobacter spp. were higher in the Staffordshire bull terrier breed compared to crossbreed dogs. In conclusion, this study shows that dogs visiting rural community veterinary clinics in South Africa are reservoirs of Campylobacter spp. and may be potential sources of Campylobacter spp. for humans living in close proximity of the dog populations under study.

Although dogs are considered an important reservoir of Campylobacter spp. (Hald & Madsen 1997;Hald et al. 2004), current data on the prevalence of Campylobacter spp. in dogs in South Africa and on the African continent are lacking. Therefore, the aim of this study was to investigate the prevalence and risk factors associated with Campylobacter spp. occurrence in healthy dogs visiting four rural community veterinary clinics in South Africa.

Study design, area and population
This cross-sectional study was conducted at four rural community veterinary clinics located in Gauteng Province, South Africa (see Figure 1)

DNA extraction
Briefly, a sterile inoculating loop was used to harvest colony sweeps from all Campy CVA plates that showed growth after 48 h -72 h. A loop-full of colony sweeps was suspended in a 1.5 mL Eppendorf tube containing 1 mL of FA buffer (Becton, Dickinson and Company). Bacterial suspensions were mixed and washed by vortexing, followed by centrifugation (15 000 g) for 5 minutes. After the first wash and centrifugation cycle, the supernatant was discarded and the bacterial pellet was resuspended in FA buffer (Becton, Dickinson and Company). Two additional washes and centrifugation cycles were carried out, after which the pellet was suspended in 500 µL of sterile water, vortexed and the homogeneous cell suspension was boiled to 100 °C for 15 min, then stored at -20 °C for further processing.

Campylobacter spp. screening
A Campylobacter spp. specific multiplex polymerase chain reaction (PCR) protocol (Forbes & Horne 2009) was used to screen DNA for Campylobacter spp. http://www.ojvr.org Open Access

Campylobacter speciation
Colony sweeps were obtained from all Campy CVA plates that were positive for Campylobacter spp. on PCR screening, streaked on horse blood agar and incubated at 37 °C for 48 h -72 h to obtain single colonies. Three suspect Campylobacter spp. single colonies were taken from each horse blood agar plate with a sterile plastic inoculating loop or swab, spreadplated separately on horse blood agar plates and incubated at 37 °C for 48 h -72 h to multiply and purify the single colonies. After incubation, pure single colony bacterial sweeps were harvested using a sterile plastic inoculating loop or swab, and the bacterial cells were suspended in 1.5 mL FA buffer in an Eppendorf tube. DNA was extracted from the single colony sweeps by the boiling method and screened for C. jejuni, C. coli and C. upsaliensis using the aforementioned primers and multiplex PCR protocol (Forbes & Horne 2009;Klena et al. 2004). Single colony isolates that were confirmed as C. jejuni, C. coli or C. upsaliensis on PCR were stored at -80 °C in cryovials containing a sterile freezing mixture (70% Brucella broth and 30% glycerol).

Statistical analysis
All statistical analyses were performed using the 'base', 'epiDisplay' and 'aod' packages of the R software version 3.3.3 (R Foundation for Statistical Computing 2017, Vienna, Austria, http://www.R-project.org/). The prevalence of Campylobacter spp., C. jejuni, C. coli and C. upsaliensis, were computed on 481 dogs using a general linear model considering the error distribution as binomial and the link function 'logit' (logistic regression). The following risk factors were tested: sex, breed, clinic, age and number of vaccinations. As a categorical variable, 'breed' included crossbreed, molosser, Staffordshire bull terrier, toy and other breeds.
To prepare for the logistic regression model, the potential relationship or association between the number of vaccinations and the age of the dog was tested. Because of non-normality in the distribution of ages in relation to time and number of vaccinations and heterogeneity of variance, an analysis of variance (ANOVA) could not be applied. Instead, a general linear model using count data (family = poisson) was used to align the ages of dogs with the number of vaccinations. Because dog ages were statistically linked with the number of vaccinations, the number of vaccinations as a risk factor was removed from the logistic regression model and only sex, breed, clinic and age were kept for risk factor analysis.
All the records for which sex and breed were not determined or missing (n = 100) were removed from the database, and only 381 dog samples were used in the model. A general linear model (family = binomial) with a full model encompassing all risk factors and all possible interactions was initially performed. Using a backward stepwise model selection based on the Akaike information criterion (AIC), the best model was kept (with the smallest AIC). A likelihood ratio test was performed to test the overall significance of each risk factor (multilevel comparison). Because age was analysed as a continuous variable, odds ratio (OR) and prevalence predictions were not calculated for this risk factor. In terms of levels within each categorical risk factor, for comparison purposes, the risk factor associated with the lowest Campylobacter spp. prevalence was used as a reference. Adjusted OR that took into account all cofounder variables were calculated, and confidence intervals (CIs). Odds ratio significance was computed using the Wald's test. To predict the prevalence of Campylobacter spp. within different categories for each risk factor, the 'predict' function was applied on logistic regression results. Finally, risk factors for C. jejuni, C. upsaliensis and C. coli were tested using the same approach that was applied for Campylobacter spp. risk factor analysis. For all analyses, p-values < 0.05 were considered significant.

Risk factors
A total of 381 dogs were included in the final model. Logistic regression showed that clinic, age and breed were significant risk factors for carrying Campylobacter spp. (see Table 1). No interactions between any of these risk factors were statistically significant in the model. Overall, visiting a particular clinic and the age of the dog were significant risk factors for carrying C. jejuni while age was the only risk factor associated with carrying C. upsaliensis. Dogs visiting clinics B, C and D were respectively 9.61, 5.68 and 12.3 times more likely to carry Campylobacter spp. in comparison to dogs visiting clinic A (reference clinic) (see Table 1). In addition, breed was a predictor of Campylobacter spp., with the odds of carrying Campylobacter spp. significantly higher in the Staffordshire bull terrier breed in comparison to crossbreed dogs, which had the lowest prevalence of Campylobacter spp. Furthermore, the prevalence of Campylobacter spp., C. jejuni and C. upsaliensis increased significantly as dogs grew older.  (Chaban et al. 2010;Engvall et al. 2003;Leahy et al. 2017;Parsons et al. 2011).

Discussion
Variations in Campylobacter spp. occurrence rates in dogs have been ascribed to a number of factors, including dog living conditions -whether a dog is confined in a house, a shelter or kennel, or is a stray. Higher Campylobacter spp. occurrence rates have been reported in dogs living in shelters or kennels and stray dogs (Baker, Barton & Lanser 1999;Parsons et al. 2011;Procter et al. 2014;Tsai et al. 2007;Workman et al. 2005) in comparison to in-house dogs. Additional factors such as Campylobacter spp. culture conditions, including the incubation temperature and atmosphere, as well as antimicrobial supplements that are used for selection of Campylobacter spp. in various recovery media or enrichment broths, may also influence Campylobacter detection rates (Allos & Lastovica 2008;Aspinall et al. 1996;Lastovica & Le Roux 2001).
Campylobacter jejuni was the most frequent Campylobacter species in dogs, followed by C. upsaliensis and C. coli to a lesser extent. This finding is in agreement with previous studies, which have reported C. jejuni as the most frequent species in dogs compared to other Campylobacter species (Carbonero et al. 2012;Giacomelli et al. 2015;Tsai et al. 2007). However, a number of reports have also found C. upsaliensis to be the most frequent species in dogs (Acke et al. 2009;Chaban et al. 2009;Holmberg et al. 2015;Parsons et al. 2011;Rossi et al. 2008). A number of studies have found that C. upsaliensis was more frequent in dogs confined in household compounds while C. jejuni was more common in stray dogs and shelter or kennel dogs (Carbonero et al. 2012;Leonard et al. 2011;Parsons et al. 2011;Procter et al. 2014).
The majority of dog owners in this study indicated that their dogs were not housed in fenced yards and dogs were allowed to leave their living premises and freely roam in the neighbourhood, thereby living a 'semi-stray' life. The stray nature of dogs sampled in this study may have played a role in the predominance of C. jejuni over C. upsaliensis. Roaming, scavenging and hunting behaviours of dogs living a semistray life exposes dogs to environments, food and water sources that may favour higher environmental contamination levels with C. jejuni compared to C. upsaliensis, which has been found to be more frequent in dogs living in-house that are fed home-cooked food (Leonard et al. 2011).
While the role played by C. jejuni in human disease is well recognised globally, of particular interest in this study was the presence of dogs infected with C. upsaliensis. Campylobacter upsaliensis has emerged in the last 20 years (Bourke et al. 1998) as an important species in dogs worldwide (Chaban et al. 2010;Engvall et al. 2003;Parsons et al. 2010) and a cause of campylobacteriosis in humans (Allos & Lastovica 2008;Couturier et al. 2012;Labarca et al. 2002;Nakamura et al. 2015). This is the first time C. upsaliensis has been reported in dogs in South Africa. This finding is of public health significance as C. upsaliensis has been previously reported as the third most frequent Campylobacter species in South Africa over a period of 10 years in paediatric patients, accounting for 23% of Campylobacter spp. cases (Lastovica & Engel 2001).
Carriage of more than one Campylobacter species was observed in 6.2% of dogs. Dogs with mixed infections carried C. jejuni and C. upsaliensis or C. jejuni and C. coli. Dogs carrying multiple Campylobacter species may have been exposed to environments and sources that allow these Campylobacter species to thrive favourably. Similar findings have been reported elsewhere (Bojanić et al. 2017;Chaban et al. 2010;Engvall et al. 2003;Hald et al. 2004;Koene et al. 2004). A number of studies have recommended the use of more than one Campylobacter spp. culture medium to facilitate the isolation and increase the chance of recovering multiple Campylobacter species of public health importance from faecal samples (Endtz et al. 1991;Baker et al. 1999;Koene et al. 2004). In the aforementioned studies in which multiple Campylobacter species were detected in individual dogs, at least two media were used to isolate Campylobacter spp. While evaluation of the sensitivity of the medium used in this study to recover Campylobacter spp. is beyond the scope of this investigation, detection of dogs carrying multiple Campylobacter species indicates that Campy CVA agar was a reliable single medium for direct and simultaneous recovery of more than one Campylobacter species from dog faeces.
Concerning the different risk factors that were investigated in this study, our findings showed that the overall prevalence of Campylobacter spp. and particularly the prevalence of C. jejuni and C. upsaliensis increased as dogs grew older, with predominance of C. jejuni in dogs younger than 1 year in comparison to dogs older than 1 year. Similar studies have reported that dogs less than 1 year old were more likely to be colonised by Campylobacter spp. (Acke et al. 2009;Guest, Stephen & Price 2007;Hald et al. 2004;Leahy et al. 2017;Parsons et al. 2010;Procter et al. 2014;Sandberg et al. 2002). The high prevalence of Campylobacter spp. in younger dogs may be most probably ascribed to an immature immune system and an underdeveloped enteric microbiota that is unable to outcompete and displace Campylobacter spp. in the intestine. This finding was not surprising as the dog population under study was skewed towards a higher number of dogs younger than 1 year (88%) compared to dogs that were older than 1 year.
Consistent with previous studies, the prevalence of Campylobacter spp. was not significantly different in male and female dogs (Nair et al. 1985;Olson & Sandstedt 1987;Sandberg et al. 2002;Torre & Tello 1993). However, breed was overall a predictor for Campylobacter carriage, with the Staffordshire bull terrier breed more likely to carry Campylobacter spp. in comparison to crossbreed dogs. Although breed has never been reported as a risk factor for Campylobacter spp. occurrence in dogs, this finding may indicate that dogs belonging to the Staffy breed, which is a pure breed, may be more susceptible to disease in comparison to crossbreeds, which are generally considered more resistant to disease.
Visiting a particular clinic was identified as a risk factor for carrying C. jejuni, with dogs visiting clinics B, C and D presenting a higher risk of carrying Campylobacter spp. and particularly C. jejuni. While the reasons behind this finding are not clear, the authors postulate that there are yetunidentified factors such as dog living conditions (in-house vs. stray dogs) that may be favouring a higher occurrence rate of C. jejuni in dogs living in the communities serviced by clinics B, C and D in comparison to clinic A, which had the lowest Campylobacter spp. prevalence.

Conclusion
This study provides useful information on the prevalence and risk factors of C. jejuni, C. upsaliensis and C. coli in dogs visiting rural community veterinary clinics in South Africa.
Our results indicate that dogs visiting the veterinary rural community clinics under study are reservoirs and may be an important source of Campylobacter spp. for humans. However, a limitation of this study is that the dogs studied were not recruited randomly and the prevalence of Campylobacter presented in this study may not be a reflection of the larger dog population of South Africa. Future epidemiological and characterisation studies comparing dog and human Campylobacter spp. isolates are needed to establish the zoonotic potential of Campylobacter spp. carried by dogs in South Africa.