Host life traits can have important influence on the way a particular host species interfaces with conspecifics, animal communities, humans, and the environment, creating or eliminating opportunities for pathogen transmission. For example, a study carried out in Northern Botswana found that antibiotic resistance is higher in carnivores and water-associated species, demonstrating that the way an organism interacts with its environment can alter pathogen transmission and carriage (Alexander et al., 2010; Jobbins and Alexander, 2015).

Multi-host pathogens, such as Campylobacter, can have altered transmission dynamics in urban landscapes increasing pathogen prevalence (Woolhouse et al., 2001; Bradley and Altizer, 2007) in relation to changes in interspecific contact rates, a key determinant of multi-host pathogen transmission (Woolhouse et al., 2001; Renteria-Solis et al., 2014). Indeed, urbanization is increasingly identified as a potential driver of infectious disease emergence as it can change host spatial dynamics and pathogen transmission patterns in ways that are unpredictable (Neiderud, 2015; Hassell et al., 2017). For example, urban landscapes may alter host-pathogen dynamics directly by changing the way pathogens move through a system, or indirectly through changes to host immunocompetence or behavior (Brearley et al., 2013; Brunton et al., 2020). Urbanization has been shown to increase individual dispersal between troops in the banded mongoose (Mungos mungo), which may heighten the transmission risk of pathogens that spread through close contact (Verble et al., 2021). Evidence of increased stress (typically measured by fecal cortisol) and immunosuppression in urban-associated wildlife, however, is mixed. Some species associated with urban landscapes have in fact demonstrated lower stress levels and more robust immune responses, likely associated with consistent resource availability (Audet et al., 2015; Lyons et al., 2017; Iglesias-Carrasco et al., 2020). This was demonstrated in a population of banded mongoose in northern Botswana, where access to anthropogenic food sources was associated with lower levels of circulating glucocorticoids (Laver et al., 2012).

While Campylobacter occurs in many wildlife species, clinical signs of gastrointestinal illness or adverse health effects have rarely been observed. Two noctule bats (Nyctalus noctula; Hazeleger et al., 2018), a dozen rhesus monkeys (Macaca mulatta; Kalashnikova et al., 2002), and several vervet monkeys (Chlorocebus pygerythrus; Ngotho et al., 2006) infected with Campylobacter spp. presented with signs of acute gastrointestinal disease. These infections have the potential to cause mortality, with one wild vervet monkey dying from suspected antibiotic-resistant Campylobacter infection during laboratory quarantine, while several other monkeys in poor condition were euthanized (Ngotho et al., 2006). This outcome suggests that some free-ranging wildlife may be susceptible to acute Campylobacter infections, potentially affecting the quantity of pathogen shed into the environment and subsequent transmission to other animals and humans (Chaban et al., 2010). Even in the absence of acute signs of disease, however, evidence suggests that Campylobacter spp. infection in wildlife can affect body mass and overall health (Taff and Townsend, 2017).

Campylobacter transmission at the human-wildlife interface is also likely influenced by the intrinsic properties of the pathogen itself, including its persistence within infected animals across time (Inglis et al., 2004). In a study of clinical Campylobacter isolates, it was found that there is extensive evolutionary potential due to frequent recombination events, and that genetic exchange occurs regularly between C. coli and C. jejuni, important human pathogens (Wilson et al., 2009). This exchange may be accelerated by human activity and could increase the host range of these organisms (Sheppard et al., 2008; Waldenstrom et al., 2010; Lawton et al., 2018; Wei et al., 2019). For example, C. jejuni sequence type (ST)-137 was found to be widespread across wild bird taxa and has been implicated more recently in sporadic cases of campylobacteriosis (Wei et al., 2019). Hypervariable regions have also been discovered in C. jejuni that likely contribute to rapid evolutionary change, host range expansion, and antibiotic resistance (Parkhill et al., 2000).

A worrying trend is the emergence of multi-drug resistant (MDR) Campylobacter. In a study that took place in the United Kingdom, MDR C. jejuni was found to be widespread in poultry (Lopes et al., 2019). As livestock waste is considered an important reservoir of antibiotic resistance genes (ARGs; He et al., 2020), the risk of ARGs spreading to wildlife through this route is substantial. However, even in the absence of antibiotic use, high levels of resistance have been found in Campylobacter isolated from wild birds (Molina-Lopez et al., 2011; Du et al., 2019). Examples of this phenomenon are also observed in domestic species. For instance, 15% of Campylobacter isolates from a population of Australian chickens were fluoroquinolone-resistant without any history of exposure to this antibiotic class (Abraham et al., 2020). It has been demonstrated that gyrA mutations conferring fluroquinolone resistance are particularly stable and unlikely to incur fitness costs in Campylobacter (Luo et al., 2005; Luangtongkum et al., 2009). There are additional mechanisms that may allow for the persistence of antibiotic resistance genes in the absence of selective pressure, such as fitness-enhancing compensation from other regions of the genome and co-selection with genes that confer an evolutionary advantage (Melnyk et al., 2015). Monitoring antibiotic resistance in Campylobacter will remain key in reducing its spread and preserving the therapeutic agents available to treat disease.

While wildlife is considered an important reservoir of Campylobacter spp., our understanding of infection and transmission dynamics is mostly limited to avian species. The current body of evidence of Campylobacter among mammalian wildlife hosts is inadequate to fully utilize a One Health approach to disease control strategies. Here, we review and summarize the literature concerning Campylobacter infection in wildlife hosts, identifying and expanding our current understanding of Campylobacter in mammalian wildlife species. Using infection detection data, we assess the potential influence of host life history traits such as trophic level, urban association, and sociality on the probability of Campylobacter carriage in wildlife species. We discuss these results and study limitations as well as implications to Campylobacter transmission dynamics and surveillance and management approaches.

This study followed guidelines provided under the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), a framework for the standardization of reviews and meta-analyses (Page et al., 2021). Briefly, published studies of Campylobacter in wildlife were retrieved from online scientific search engines (Web of Science, PubMed, Google Scholar) using the key words: ‘wildlife + Campylobacter’ and ‘wild mammals + Campylobacter.’ Only the first 300 records from each query in Google Scholar were screened, as this resource was considered a secondary database to capture articles not located on Web of Science or PubMed (Haddaway et al., 2015; Gusenbauer and Haddaway, 2020). Duplicate records were removed automatically using EndNote 20, a reference manager (The EndNote Team, 2013). Reference lists of pertinent publications were searched for additional related articles. Information was retrieved on the location of sample collection, order, genus, and species studied, animal status (captive, wild feces or wild carcass), detection of Campylobacter spp. and particular Campylobacter species identified, and method of testing used (culture and PCR or other molecular methods).

Mammals that had Campylobacter detection data were categorized by trophic level (herbivore, omnivore, and carnivore) and sociality using the PanTHERIA database (Jones et al., 2009) and the Animal Diversity Web (ADW; Myers et al., 2019). In cases where species are occasionally omnivores but are primarily herbivores or carnivores, the predominant trophic level was chosen for analysis. Sociality of the species was defined as either social or solitary, depending on whether a species lived alone or within a social group (Myers et al., 2019). Mammals were also classified based on their association with urban areas. The classification system from Santini et al. (Santini et al., 2019) was used to classify species into urban ‘dwellers’ and urban ‘visitors’, with species meeting more than one criterion labeled as both (i.e., ‘dwellers\visitors’). Per this classification system, species that are classified as urban ‘dwellers’ have been demonstrated to utilize urban areas for all resources, while ‘visitors’ may exist in urban areas for some parts of their life cycle, but still obtain most resources from natural environments. Urban association for species that were not included in the Santini et al. (Santini et al., 2019) review were searched for using ‘taxon name (genus and species) + urban, urbanization or humans’ (Supplementary Table S1). An additional category was added for purposes of this study to classify species that had no urban interaction, which were labeled as urban ‘avoiders’ based on an absence of urban interaction in the literature. Several marine and alpine species were assumed to be avoiders due to habitat needs that did not overlap with urban areas. Information on consumption of species by humans was collected from the IUCN Red List (IUCN, 2019) and Mildenstein et al. (2016) for some bat species.

Once data had been compiled, duplicate records were removed from the original database through an automated process using EndNote 20 (The EndNote Team, 2013). If the same species was studied in two different regions of the world, and no subspecies are apparent, only one entry for this species was kept in our database. Due to the significant differences in sample sizes from different manuscripts (sample sizes varied from one to 1,168; Sus scrofa), we used a dummy categorical variable for Campylobacter presence in wildlife species, with 1 indicating that Campylobacter spp. had been detected (culture and/or polymerase chain reaction (PCR) based methods) or 0 indicating no Campylobacter spp. detection. This dummy variable was used to create graphs and carry out statistical analyses of the data. If at least one study of a species reported the detection of Campylobacter spp., it was recorded as present for the purposes of analysis. For studies that only identified Campylobacter to the genus level or where the authors did not attempt to detect C. jejuni or C. coli, no data was recorded at the species level and this field was left blank. All statistical analyses were conducted in R 4.1.2, an open source integrated programming environment (R Core Development Team, 2022). To investigate the effect of the categorical input variables (trophic level, urban association, and sociality) on the binary presence/absence of Campylobacter, C. jejuni, and C. coli, a series of logistic regression (LR) models were created in order to determine the ability of a model to predict carriage of either Campylobacter genus or the two species of Campylobacter analyzed in this study (C. jejuni and C. coli). An important assumption of LR models is that multicollinearity does not exist among the predictors. The generalized variance inflation factor (GVIF) was used to assess for multicollinearity in the predictor variables of all models used for analysis. GVIF is generally corrected by ½ degrees of freedom (df) when applied to categorical variables, and a cutoff of 1.73 following (Usio, 2020) was used, with independent variables below this threshold having little-to-no correlation. To obtain the GVIF^1/2df value, we used the vif function of the car package (Fox et al., 2012). No independent variables in the models had GVIF^1/2df values significantly above 1, indicating that multicollinearity was not present in this dataset.

We use the Akaike information criterion (AIC) to score and select the model with the best fit for the data balanced against the number of model parameters included (Aho et al., 2014). For this analysis we use the AICc, which is a sample size adjusted formula to calculate AIC (Bedrick and Tsai, 1994), due to the relatively low sample size (n = 128). AICc values were generated using AICcmodavg (Mazerolle and Mazerolle, 2017) and the model with the lowest AICc value was chosen as the best fit and used in additional analyses. The predict function in R 4.1.2 was used to estimate the probability of pathogen carriage for each level of the factors included in the in our top model. Data were visualized and graphed using ggplot2 (Wickham et al., 2016). Summary statistics were generated using the janitor package (Firke, 2021).

This study provides a broad overview of Campylobacter spp. infection in free-ranging mammalian wildlife. Statistical analyses in this study were based on presence/absence and did not include Campylobacter prevalence data, removing some of the complexity of the original datasets. Prevalence often varied greatly between species and studies, ranging between 1 to 87.5%. These differences in prevalence likely create differential transmission risks, a potential area of future research. Culture and detection methods also varied greatly between the studies which can have a significant impact on the prevalence, species and genotypes isolated (Williams et al., 2012; Hetman et al., 2020). Detection methods often are biased toward C. jejuni and C. coli isolation which can skew the representation of Campylobacter species present (Lastovica and le Roux, 2000). Importantly, there have been relatively few studies about Campylobacter upsaliensis, especially in free-ranging wild mammals. This species is suspected to be an emerging human pathogen and has been isolated from domestic animals such as dogs with some frequency (Bourke et al., 1998; Man, 2011; Couturier et al., 2012). The age and storage of feces also impacts Campylobacter isolation, with fresh fecal samples significantly more likely to likely to be positive for Campylobacter spp. than older feces using culture methods (Abdollahpour et al., 2015). In a study performed using sheep feces sampled and placed back into the environment (Jackson et al., 2009), the authors found that Campylobacter died rapidly. Another phenomenon in Campylobacter spp. that complicates isolation is the progression to a viable but non-culturable (VBNC) state under harsh conditions (Moriarty et al., 2011). This may be particularly relevant for Campylobacter isolation from wildlife samples, especially if samples are taken from remote sites. Culture methods have also been found to be inferior to immunoassay and molecular methods for sensitivity and specificity of Campylobacter detection (Buss et al., 2019).

Pathogen prevalence and transmission have been well studied in European and North American wildlife and bird species, but relatively few studies on Campylobacter prevalence and transmission in wildlife have occurred in developing countries where Campylobacter burden is highest (Troeger et al., 2017), transmission dynamics may differ (Platts-Mills and Kosek, 2014), and limited resources may increase the difficulty of investigations. There is also a lack of phylogenetic information on Campylobacter spp. in mammalian wildlife. Campylobacter spp. possess the ability to evolve rapidly (Wilson et al., 2009) which suggests that the relationship between wildlife-associated Campylobacter and human-associated Campylobacter may be obscured. Most papers that incorporate genetic or phylogenetic information rely on MLST techniques which have been the gold-standard for defining clonal groups but have been shown to be inferior at establishing epidemiological dynamics (Llarena et al., 2017). Only two recent papers incorporate WGS on free-ranging wildlife and both papers found human-associated Campylobacter genotypes in their respective wildlife species (Baily et al., 2015; Hazeleger et al., 2018). Future research should incorporate WGS methods into targeted studies of species at the human-wildlife interface that are most likely to contribute to Campylobacter transmission and human disease.