The majority of emerging infectious diseases of pandemic potential are zoonotic diseases[1, 2]. Examples include the H1N1 influenza pandemic (1918), severe acute respiratory syndrome (SARS) coronavirus (2002–2004), Ebola virus (2013–2016), and the recent severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that started in 2019[3], whose origin is still being debated upon[4]. Bats, non-human primates, and duikers have been reported as spillover sources of the Ebola virus [5, 6]. In turn, bat species have been identified as the most probable wildlife origin of SARS-CoV-2[7] with pangolins acting as intermediate hosts[8]. These zoonotic diseases have severe health, economic, and social impacts[9, 10]. The SARS-CoV-2 outbreak has resulted in more than 5.4 million deaths[11] and a loss of 2.4 trillion US dollars in global GDP[12]. Due to the huge financial and public health cost of zoonotic diseases, it is imperative to monitor and prevent zoonotic disease outbreaks[10]. A potential pathway for prevention is the development of predictive models that can identify locations and wildlife species of zoonotic risk potential.
Tropical species with high pathogen richness may be important zoonotic reservoirs[2, 13, 14]. Future zoonotic infectious disease events have been predicted to emerge mostly in tropical regions due to their high richness of host species[2]. Additionally, species that harbor more pathogens are more likely to lead to zoonotic spillover when there is close human-wildlife contact. Hence, the monitoring and prediction of pathogen richness in tropical host species is crucial to prevent future zoonotic disease outbreaks[13, 14].
Previous studies have identified the main predictors of pathogen richness in wildlife species, including evolutionary history, species geographic distribution, species traits, and anthropogenic factors. There is contrasted evidence that phylogenetically similar species have higher spillover risk[15, 16] and that certain taxonomic groups (i.e. Rodentia and Carnivora) harbor higher zoonotic pathogen richness[17]. In turn, species geographic distribution affects pathogen exposure[13, 15, 18, 19]. Because species with greater geographic range area and altitude breadth are more likely to occupy different habitats, the chances of coming into contact with a variety of host species and fomites are higher, thus increasing their pathogen exposure[13, 20]. Indeed, geographic range overlap with a high number of species has been reported to be positively correlated with greater virus richness[18, 19].
Fast-lived species are hypothesized to have higher pathogen richness. Fast-lived species are more resilient to human disturbance and thrive in anthropogenic environments[21]. This allows them to make use of new available resources (e.g. anthropogenic food sources and declining competing species population due to hunting) and live in close proximity to humans. This leads to higher population density of fast-lived species, promoting the spread of pathogens. Another hypothesis is based on the trade-offs between reproduction and immunity[22, 23]. Fast-lived species invest more energy into reproduction, resulting in nonspecific and weaker immune responses to diseases[23]. For instance, fast-lived birds with shorter incubation periods have lower natural antibody levels[24] and lymphocytes[25]. However, there is limited evidence that support this claim in mammals[22]. A study on three wild rodent species found that fast-lived species have a smaller antibody response than slow-lived species[26]. However, Cooper et al. (2012)[14] found that lower longevity (fast-living) in ungulates, carnivores, and primates was not correlated with lower white blood cell counts. Hence, the pathway in which life history traits influence pathogen richness may not be simply because of the reproduction and immunity trade-offs, it may also include how different traits affect pathogen exposure and persistence.
An alternative hypothesis suggests that slow-lived species have higher pathogen richness[22]. Species with longer life spans are hypothesized to encounter more pathogens and provide more time for pathogen multiplication and colonization[27]. Further, larger-bodied species provide more niches for pathogen establishment (e.g. greater surface area for ectoparasites)[27] and have greater energy requirements, food intake, and thus, greater risk of foodborne pathogens[13, 27].
Anthropogenic factors such as hunting pressure and livestock density can also affect pathogen transmission and exposure[28–31]. Hunting results in direct human-animal contact which not only increase the risk of zoonotic spillover[29], but also promotes the transmission of pathogens from humans to wildlife, increasing pathogen richness in wildlife hosts[31]. High livestock density also facilitates pathogen transmission as it increases the risk of pathogen spillback from livestock to wildlife[32]. Furthermore, wildlife hunting and livestock expansion reduces biodiversity[33, 34] which in turn, increases pathogen exposure and richness for unhunted host species due to the dilution effect[30]. At low species diversity there is a reduction in interspecific competition which can result in higher host species population density and susceptible species abundance. This increases the chance of interaction between infected vectors, infected hosts, and susceptible individuals[30], increasing pathogen exposure and richness. On the other hand, hunting reduces species population density which can reduce pathogen transmission and persistence within the population[28] by decreasing the number of potentially interacting host individuals.
Other driver to high pathogen richness is the alteration in environmental conditions which can result in changes in species pathogen exposure and energy allocation to immunity[35]. Land-cover changes such as urbanization and conversion of habitats for agriculture result in increased human-wildlife contact and changes in resource availability[35, 36]. For instance, urbanization can result in the loss of habitats and resources, leading to poorer health conditions of wild animals[37]. The reduction in resources can subsequently lead to decrease energy allocation to immunity and increase pathogen richness[35]. Conversely, some species are able to take advantage of anthropogenic food sources (e.g. rubbish bins, plants from backyards and streetscapes, and domestic animals), allowing them to adapt to urban environments[36, 38]. Urban and agricultural areas may also release contaminants into the environment which can directly affect wildlife health and immunity[35, 37].
Anthropogenic and intrinsic (species traits) predictors of pathogen richness have been examined in recent literature. For example, Gibb et al. (2020)[21] found that there was higher wildlife host species abundance in agricultural and urban land-use whereas the relationship between pathogen richness and life history traits was assessed by Cooper et al. (2012)[14], Kamiya et al. (2013)[13], and Plourde et al. (2017)[16]. These studies show that pathogen transmission and persistence is greatly affected by both intrinsic (e.g. species life history traits) and extrinsic (e.g. land-cover changes) factors[13, 21, 39]. Yet, the examination of the interactions between both extrinsic and intrinsic factors and the relative importance of each predictor for explaining pathogen richness has not been properly studied, hence limiting our understanding of the determinants of pathogen richness. To address this, here we analyzed the extrinsic and intrinsic predictors of pathogen richness in tropical mammal species collectively. Furthermore, we added a new hunting pressure predictor that have not been modelled with pathogen richness previously. We focused on the tropics as it is the area with the highest likelihood of zoonotic risk events and used a model ensemble approach 1) to identify the extrinsic factors and species traits that are associated with pathogen richness, 2) to assess their relative importance for predictive purposes, and 3) to project pathogen richness across the tropics.
We analyzed the proposed predictors of pathogen richness in tropical mammals using a model ensemble composed of machine learning and zero-inflated regression-based models. Using multiple models allows the comparison of results and ensure that the findings are robust and not overly dependent on a single model. Additionally, ensemble model prediction combines predictions of multiple models, reducing uncertainty and can result in a more reliable and accurate prediction[40]. Based on previous research results, we hypothesize that species with higher reproductive rate have greater pathogen richness. Furthermore, higher urban and agricultural land-cover change within a species geographic range increases pathogen richness due to increase in exposure to contaminants and domesticated animals. Knowledge gained from this study will help inform pathogen surveillance in wildlife to prevent future zoonotic emerging infectious diseases and help shed light on the pathways in which these predictors affect pathogen richness.