We gathered data on parasite-host associations, host traits, and the transmission mode of parasites from the Brazilian Mammal Parasite Occurrence Data (Cruz et al., 2023a), with further details provided in the Supplementary material under ‘Description of the Dataset’. The spatial distribution of wild mammals was obtained from SALVE (ICMBio, 2022).

The MCMC convergence of the HMSC models developed for all the zoonotic microparasite taxonomic groups indicates reliable and stable model performance, with an average potential scale reduction factor for ꞵ-parameters—which measure the parasite species occurrence related to host trait covariates—of 1.02, and a maximum of 1.05. The models exhibited good data fit, as quantified by AUC, Tjur’s R², and RMSE (Table 1, Fig. S1–S3).

Variance partitioning of the explanatory variables included in the models indicated that fixed effects (i.e., body mass, litter size, dietary guild, activity period, and foraging stratum) explained a substantial amount of model variation (Table 1, Fig. S4–S6). Among these variables, body mass (fixed effects) and geographical units (random effects) emerged as the most important predictors of parasite occurrence between hosts. Host body mass displayed a positive relationship with parasite richness for a broad spectrum of host taxa (Fig. S7). Overall parasite richness was relatively consistent across the main guild, activity period, and foraging stratum (Fig. S8–S10). Further, models’ performance, assessed through AUC, Tjur’s R2, and RMSE values, along with variable importance, was similar between analyses that incorporated E. coli and rabies virus into the models and those that excluded them (Table 1). While the model performance for protozoa, which include T. cruzi, remained unchanged (Table 1), the variance partitioning for body mass and spatial random effect was considerably reduced (Fig. S7, panel b).

Transmission mode of the parasites covariates with host traits for each group of zoonotic microparasites, except for Protozoa. We observed a positive association between transmitted bacteria through direct contact and terrestrial hosts (Fig. 1, panel a). Conversely, the model showed a positive correlation between direct transmission and host litter size for viruses, excluding rabies virus (Fig. 1, panel b). However, this correlation is not evident when the rabies virus is considered (Fig. 1, panel d), and a new pattern appears, with a negative correlation between direct transmission and body mass. Therefore, these findings should be approached with caution and warrant further in-depth investigation.

Fig. 1.

Heatmap of estimated parameters linking parasite transmission mode and parasite species responses to host trait covariates. Red and blue colours indicate parameters estimated as positive and negative, respectively, with at least 95% posterior probability, while responses lacking strong statistical support are shown in white. Panels (a) show the model excluding E. coli, panel (b) show the model excluding rabies virus, and panels (c) and (d) models include these species. Panel (e) depicts shared zoonotic microparasite responses to host traits based on parasite taxonomy. Species are ordered according to their taxonomy, as illustrated by the taxonomy tree displayed in the panels.

Finally, the response of parasite species occurrence to host trait covariates displayed a high taxonomic relationship in the presence-absence model for Bacteria (P(rho) >0.9, estimated rho >1) and for Protozoa when T. cruzi was included in the data of species occurrence (P(rho) >0.9, estimated rho >1). For instance, when considering a relationship with at least a 95% posterior probability, 7% (2 out of 29) of the investigated bacteria species demonstrated a negative response to body mass, while 24% (7 out of 29) showed a positive response (Fig. 1, panel e). These proportions correspond to the species visually identified in Fig. 1 as having positive or negative associations with the host trait. Likewise, 27.5% (N = 8 from 29) of the species showed negative evolutionary signals associated with litter size and invertebrate diet, as illustrated by the species responses in the figure. The nocturnal activity trait negatively influenced only 3.5% (1 out of 29) of the investigated species, whereas 27.5% (8 out of 29) of the investigated zoonotic microparasite displayed a positive occurrence probability in nocturnal host species, and 34.5% (10 out of 29) presented a positive response to the diurnal activity trait (Fig. 1, panel e).

We found that Primates comprised the mammal order with the widest predicted range of zoonotic microparasite (Fig. 2). Examples of parasite occurrence probability across mammal orders include the protozoan Toxoplasma gondii in Primates (e.g., Ateles spp.), Carnivora (e.g., Lontra longicaudis), Cingulata (e.g., Cabassous spp.), and Cetartiodactyla (Dicotyles tajacu), Leishmania amazonensis, L. infantum, and Rotavirus A in Chiroptera (e.g., Micronycteris spp.), and Andes orthohantavirus, Juquitiba virus, Rio Mamore hantavirus, and Rickettsia rickettsii (bacteria) in didelphid marsupials (e.g., Gracilinanus spp. and Monodelphis dimidiata). The Brazilian territories exhibiting the highest estimated parasite richness, with at least four zoonotic species, include all states in the Northern region, except Tocantins (with a richness of 3.6); Mato Grosso in the Central-Western region; Bahia, Rio Grande do Norte, and Ceará in the Northeastern region; Rio Grande do Sul and Santa Catarina in the Southern region; and all states in the Southeastern region, except Espírito Santo (with a maximum richness of 3.6) (Fig. 3).

Fig. 2.

HMSC models fitted for each group of zoonotic microparasites (bacteria, protozoa, and viruses) based on body mass (log-transformed), litter size, main trophic guild, activity period, foraging stratum (fixed factors), host species identity (at species and order levels), and study site (random factor). The heatmap displays the unified results, with colours indicating the probability of zoonotic microparasite occurrence (vertical axis) between 0 and 1 (yellow to purple), according to host species suitability (horizontal axis). The probability of the parasite’s occurrence in a host species was not calculated when the association between them is already known (grey areas). Microparasite species are coloured by group.

Fig. 3.

Potential distribution of zoonotic microparasites according to host trait composition across study sites. Study sites abbreviations: AC - Acre, AL - Alagoas, AP - Amapá, AM - Amazonas, BA - Bahia, CE - Ceará, DF - Federal District, ES - Espírito Santo, GO - Goiás, MA - Maranhão, MG - Minas Gerais, MT - Mato Grosso, MS - Mato Grosso do Sul, PA - Pará, PB - Paraíba, PR - Paraná, PE - Pernambuco, PI - Piauí, RJ - Rio de Janeiro, RN - Rio Grande do Norte, RS - Rio Grande do Sul, RO - Rondônia, RR - Roraima, SC - Santa Catarina, SP - São Paulo, SE - Sergipe, TO - Tocantins. Eight states presented the highest probabilities of parasite occurrence: Roraima, Amazonas, Pará, and Tocantins in the Northern region; Mato Grosso and Goiás in the Central-Western region; and Maranhão and Piauí in the Northeastern region, where we estimated that the maximum parasite richness exceeds five zoonotic species. Primates and Chiroptera comprised the mammal orders with the broadest predicted range of zoonotic microparasites.

A novel insight from our results is the differential influence of body mass depending on the microparasite group, which challenges the assumption of a uniformly positive or negative relationship between host size and parasite richness. The attenuated effect of body mass in the protozoa model upon inclusion of T. cruzi suggests that widely distributed generalist parasites can mask trait-based patterns in parasite-host associations. This adjustment may also arise from the HMSC model’s reliance on community-wide mean traits, which can lead to an underestimation of probabilities for other common species (for an in-depth discussion, see Ovaskainen et al., 2017; Erickson and Smith, 2023), particularly due to T. cruzi’s broader host range. These findings underscore the need for refined modelling approaches that account for the ecological prominence of generalist parasites.

The variance explained by the fixed effects, such as body mass, litter size, diet, and foraging stratum, is also a significant predictor of zoonotic microparasite richness. Still, it only partially explains the observed variation. The magnitude of a significant spatial random effect suggests that unmeasured ecological and biogeographical factors, like host community composition and landscape structure, also play a substantial role (Andreazzi et al., 2023). These results emphasise the importance of considering both intrinsic host traits and broader geographic and environmental contexts when assessing patterns of zoonotic parasite distribution.

Our analysis indicates a correlation between nocturnal species and bacteria transmitted through direct contact, highlighting how host activity shapes interactions and encounters with parasites (Gao et al., 2021). Additionally, some parasites may synchronise their activity patterns with those of their hosts and vectors (Rund et al., 2011). Consequently, both factors contribute to the increased potential for direct or vector-borne transmission. No association was found between transmission mode and host traits for protozoa, possibly because our dataset included primarily vector-dependent species, with little variation in transmission modes to reveal patterns in host traits. Although host group size and density were not incorporated into the models due to data limitations, these traits are recognised to influence parasite transmission dynamics (Altizer et al., 2003; Lindenfors et al., 2007; Han et al., 2015). Smaller-bodied species, like rodents and bats, often form larger social groups, increasing direct contact and potentially facilitating virus transmission (Clutton-Brock, 2009; Webber et al., 2017). However, correlation does not imply causation, and other unmeasured traits may also contribute to the observed patterns.

The bacteria used in our model encompass a highly diverse array of taxonomically distant species. Our results indicate that taxonomically closely related bacteria exhibit similar responses to host trait covariates. They suggest that some taxonomically conserved phenotypic traits in bacteria, potentially related to transmission dynamics, share an evolutionary signal associated with host traits, a positive, neutral, or negative signal, contingent upon the taxonomic lineage. However, research on the relationship between host and parasite ecological and life-history traits is still in its early stages (Martiny et al., 2015; Levin et al., 2021). Similarly, studies on the phylogenetic conservation of ecological traits among microparasites remain limited (Finlay, 2004; Philippot et al., 2010; Martiny et al., 2015). Future research should investigate which parasite traits, potentially influenced by phylogenetic relationships, are associated with the mammalian traits examined in this study.

We found that including T. cruzi or rabies virus into the models altered correlations between transmission mode and host traits, as well as the evolutionary signal of parasites. Such shifts reflect the sensitivity of hierarchical models like HMSC to dataset composition, since these models share information across species by implementing trait means and residual structures (Ovaskainen et al., 2017). Using a taxonomic tree instead of phylogenetic distances is also less accurate for evolutionary analyses. For instance, in our analysis, most protozoan species belong to the same genus or family, which makes them taxonomically close, potentially contributing to the lack of an evolutionary signal related to host trait covariates. Thus, the absence of an evolutionary signal may reflect either a true biological pattern that remains undetected or a methodological artefact resulting from limited taxonomic variability, such as in protozoa, which can reduce phylogenetic resolution (Vamosi et al., 2009). Additionally, using taxonomy as a proxy for phylogeny can introduce error, as taxonomically similar groups may be phylogenetically distant, and vice versa. For these reasons, we recommend interpreting the observed evolutionary signal with caution and emphasise the need for future analyses using robust genomic or multigene phylogenies.

Previous studies have shown that Chiroptera and Rodentia harbour the highest number of empirically detected zoonotic microparasites in Brazil (Cruz et al., 2023b). However, our findings reveal that Primates, Didelphimorphia, and Chiroptera exhibit the highest predicted parasite richness among terrestrial mammals. Notably, Primates display a greater proportion of zoonotic parasites than rodents or bats (Han et al., 2016; Cruz et al., 2023b).

Their close phylogenetic proximity to humans heightens the prominence of non-human primates as zoonotic potential sources (Cooper et al., 2012b; Olival et al., 2017), making them likely hosts for pathogens that can infect humans, while also being vulnerable to human-associated pathogens (Davies and Pedersen, 2008). However, phylogenetic proximity alone does not fully explain their role as reservoirs (Davies and Pedersen, 2008; Cooper et al., 2012b; Mollentze and Streicker, 2020), as viruses often infect distantly related hosts (Cooper et al., 2012b). Conversely, bats are considered ‘special reservoirs’ due to their long lifespans relative to body size (Munshi-South and Wilkinson, 2010), which increases their exposure to parasites (Luis et al., 2013). These results underscore the importance of integrating factors such as geographical distribution, host morphology, ecology, and life-history traits to improve our understanding of species that are more likely to be exposed to zoonotic microparasites due to shared ecological characteristics, for instance, overlapping periods of activity.

We identified territories in the Cerrado (a savannah-like biome situated in the Central-Western region), the Amazon, and the Atlantic Rainforest as particularly suitable for the occurrence of zoonotic microparasites, given the presence of potential host species. Previous studies highlight the risk of zoonotic disease emergence in Northern and Central-Western Brazil, mainly due to the high diversity of wild wild mammals and habitat loss (Han et al., 2016; Vale et al., 2021; Winck et al., 2022). Hence, our findings emphasise the importance of host traits and geography in predicting zoonotic parasite occurrences within a One Health approach, particularly as host distributions shift in response to human-driven changes in land use (Magioli et al., 2021) and climate (Carlson et al., 2022). These shifts can alter parasite-host dynamics and affect pathogen exposure in both human and wild species.

While our approach provides insights into parasite richness regarding parasite taxonomy, transmission mode, and their covariation with host traits, we acknowledge limitations that may affect model accuracy, including potential sampling bias, limited data on mammal life history, and unknown parasite transmission modes (Wille et al., 2021). Although the major host traits analysed are crucial for understanding parasite sharing, other factors that we were unable to test in our study are also expected to play a role in determining parasite occurrence, such as immunological factors, longevity, home range, and group size (Lee, 2006; Han et al., 2015; Olival et al., 2017). Data gaps hinder the incorporation of these factors into statistical models.

Our approach predicts areas with a high probability of zoonotic microparasite occurrence based on host traits. However, the apparent high probability of zoonotic occurrence in the Atlantic Rainforest, particularly in the Southern region, might be influenced by an overrepresentation of studies conducted there (Cruz et al., 2023b). To refine our predictions, further research should focus on targeted field studies in regions identified as favourable for zoonotic microparasite occurrence. Based on our findings, heightened surveillance of wild mammal species, particularly primates (Ateles spp.), bats (Micronycteris spp.), and marsupials (Gracilinanus spp. and Monodelphis dimidiata) is recommended, especially for key zoonotic pathogens such as Andes orthohantavirus, Juquitiba virus, Rio Mamore hantavirus, Vaccinia virus, Rotavirus A, Leishmania spp., and Toxoplasma gondii.

Surveillance efforts should be implemented through: (1) indirect monitoring via environmental faecal/urine sampling, and (2) direct, non-invasive monitoring of handled animals (e.g., during rescues, roadkill specimens) for faecal/urine/swab collection. These efforts are critical in regions where predicted zoonotic parasite richness is highest, notably in the Cerrado, Amazon, and Atlantic Rainforest biomes. Practical implementation could combine these sampling methods with community-based reporting of unusual wildlife mortality or behaviour. Integrating these approaches into existing biodiversity monitoring and public health initiatives may enhance early detection capacities and support a more proactive One Health surveillance system.

Moving forward, public health awareness programmes can address risks associated with human interactions with potential hosts, promoting responsible behaviours to reduce parasite transmission. Also, whenever a human zoonotic infection is documented in a new area, simultaneous investigations in non-human hosts should accompany public health initiatives. This proactive approach can help mitigate challenges posed by the absence of systematically organised data on parasite-host associations. Finally, integrating landscape and climate variables into the modelling framework can enhance our understanding of how environmental changes influence host traits and zoonotic dynamics, contributing to human and animal health, as well as biodiversity conservation efforts.