Human activities such as agriculture can create abundant and predictable food subsidies that significantly alter infectious disease dynamics in wildlife. Theory and meta-analysis suggest that how food subsidies affect host immune defense can determine whether resource provisioning amplifies or dampens pathogen transmission. However, little work has demonstrated multi-scale links between food availability, immunity, and infection in natural systems. We tested these relationships through a longitudinal study of 300 vampire bats (Desmodus rotundus) across 10 sites in Peru and Belize that differ in the availability of livestock, an important food source for vampire bats in human-modified landscapes. We predicted greater feeding on livestock prey could reduce starvation stress and energy spent foraging, allowing bats to invest more in immune defenses that promote pathogen resistance or clearance. We quantified leukocyte counts from blood slides, microbicidal ability of plasma against Escherichia coli, and levels of immunoglobulin G from plasma and used PCA, PERMANOVA, and LMMs to test for differential immunity as a function of livestock density. We next used PCR to determine bat infection status with Bartonella spp. and hemotrophic Mycoplasma spp. and used GLMMs and causal mediation analysis to test how bacterial infection status varies with livestock density and if these relationships are mediated through associations between provisioning and immune function.
Our data suggest that livestock expansion into rainforest habitats is associated with significant shifts in bat immune profiles. Individuals from livestock-dense habitat showed greater numbers of neutrophils and microbicidal ability but lower numbers of lymphocytes and levels of immunoglobulin G. This relationship was most pronounced in reproductive bats, which were also more common in high-livestock habitats, suggesting negative feedback between fitness benefits of resource subsidies and investment in immune function. We found the odds of Bartonella infection to be lowest in high-livestock sites, and this association was significantly mediated through the relationship between immunity and provisioning. In contrast, odds of hemotrophic Mycoplasma infection weakly increased with livestock density and were not mediated through resource-altered immunity. These differences among bacterial pathogens may reflect distinct mechanisms of infection clearance or different transmission modes. Our results highlight the complexities in predicting how provisioning affects infectious disease and the importance of considering how within-host dynamics and different transmission modes respond to resource shifts.