The influence of bat ecology on viral diversity and reservoir status

Abstract Repeated emergence of zoonotic viruses from bat reservoirs into human populations demands predictive approaches to preemptively identify virus‐carrying bat species. Here, we use machine learning to examine drivers of viral diversity in bats, determine whether those drivers depend on viral genome type, and predict undetected viral carriers. Our results indicate that bat species with longer life spans, broad geographic distributions in the eastern hemisphere, and large group sizes carry more viruses overall. Life span was a stronger predictor of deoxyribonucleic acid viral diversity, while group size and family were more important for predicting ribonucleic acid viruses, potentially reflecting broad differences in infection duration. Importantly, our models predict 54 bat species as likely carriers of zoonotic viruses, despite not currently being considered reservoirs. Mapping these predictions as a proportion of local bat diversity, we identify global regions where efforts to reduce disease spillover into humans by identifying viral carriers may be most productive.

1 Table A1. Bat ecological traits used in boosted regression tree models. Numbers and percentages of bat species -total and those carrying virusesfor which each variable could be collected are included. These values are for the full dataset (n=812 species in the Shi and Rabosky (2015) phylogeny). Residuals taken from regression of square root transformed range area on species sympatry. Species sympatry is a measure of the number of species whose geographic range overlaps with the range of a focal bat species.  Table A2. Predicted relationships between bat ecological traits and viral richness. Bat traits were either predicted to have a positive (+) or (-) relationship with viral diversity.

Relationship with Viral Richness Justification
Forearm Length + In bats, forearm length is used to assess body size. Larger hosts may provide parasites with more resources, or encounter parasites at greater rates through ingestion and contact (Lindenfors et al. 2007, Kamiya et al. 2014).

Relative Wing Loading & Aspect Ratio
-RWL and AR serve as rough proxies for bat ecological niches (Norberg and Rayner 1987) and capture variation in daily activity patterns. The high body temperatures and metabolic rates associated with bat flight could play a role in pathogen control (O'Shea et al. 2014), given that immune responses are thought to be potentiated at high temperatures (i.e. fever; Hasday et al. 2000, Blatteis 2003. Species with wings designed for long periods of sustained flight (aerial insectivores; high AR, high RWL), leading to sustained increases in body temperature, may host a lower diversity of pathogens, owing to faster clearance and lower probability of detection.

+
Long lifespans provide more time for exposure to parasites (Poulin and Morand 2004).

Phytophagy > Predatory
Shared food resources (i.e. partially consumed pieces of fruit) may become contaminated with viruses (Dobson 2005), increasing opportunities for viral transmission between individuals. Frugivorous and nectivorous species (i.e. phytophagus) may therefore harbor greater pathogen diversity than predatory species.
Torpor Expression -Torpor use is predicted to correlate negatively with viral diversity, since viral replication and host contact rates will be reduced in hibernation (Luis et al. 2013, Guy et al. 2019, potentially limiting opportunities for viral transmission and leading to lower pathogen diversity.

Median Aggregation Size
-While conflicting evidence exists (Willig et al. 2003, Kamiya et al. 2014, we expect viral richness to increase towards the equator, similar to the latitudinal patterns of species richness observed in other groups (Willig et al. 2003) Longitude Range Midpoint + No species of bat is transoceanic. To control for this natural geographic barrier, we include longitude to separate old from new world species. Further, zoonotic emergence has occurred disproportionally in the eastern hemisphere (Jones et al. 2008) and so there may be longitudinal variation in viral transmission between species.
Geographic Range Area + Wide-ranging species may encounter a greater diversity of habitat types or a greater number of species, increasing opportunities for parasite exposure and transmission (Poulin and Morand 2004, Lindenfors et al. 2007, Luis et al. 2013).

+
For a given range size, species experiencing range overlap with a greater number of other bat species may have greater opportunities for contacts that facilitate pathogen transmission (Luis et al. 2013, Olival et al. 2017, Guy et al. 2019

+
Included to help control for sampling bias. Previous literature indicates parasite diversity scales positively with study effort (e.g., Lindenfors et al. 2007, Luis et al. 2013 Fig. A2). Note that we could not control for the effects of citations in the categorical variables of bat family and torpor, possibly explaining why torpor increases in importance in the residuals model.   Fig. A2). Note that we could not control for the effects of citations in the categorical variables of bat family and torpor, possibly explaining why torpor increases in importance in the residuals model.            Log Group Size Residuals*