Invasive Burmese pythons alter host use and virus infection in the vector of a zoonotic virus

The study addresses how an invasive apex predator, the Burmese python, indirectly alters the transmission dynamics of a zoonotic arbovirus by restructuring vertebrate host communities in southern Florida. Vector-borne pathogens often exhibit a dilution effect, wherein higher diversity of poor reservoir (dilution) hosts suppresses pathogen prevalence in vectors. The establishment and spread of Burmese pythons have caused severe declines in medium- and large-sized mammals in the Everglades, potentially reshaping mosquito host use. Prior observations suggest mosquito blood meals shifted from diverse medium/large mammals toward rodents, notably the hispid cotton rat (Sigmodon hispidus), a confirmed amplifying host of Everglades virus (EVEV/EEVV). The research tests the hypothesis that python-driven reductions in non-rodent mammal diversity increase the proportion of vector feedings on cotton rats, thereby elevating EVEV infection prevalence in vector mosquitoes, with implications for human risk.

The paper situates the work within broader evidence that biodiversity can dilute transmission of vector-borne zoonoses and that anthropogenic or invasive-species-mediated perturbations restructure communities and disease risk. Previous studies documented 85–100% declines of common native mammals in southern Florida following python establishment and a marked shift in mosquito blood meal sources from medium/large mammals to rodents. Historical Everglades studies identified cotton rats as the key amplifying hosts of Everglades virus; recent data indicate rodent community changes with cotton rats dominant. Analogous systems (e.g., Lyme disease) show pathogen prevalence in vectors depends on host community composition and vector host use, supporting the perturbation hypothesis that ecosystem disruptions drive spillover via altered interspecific interactions.

Study design and sites: Fieldwork was conducted at 12 sites across southern Florida spanning a gradient of modeled Burmese python presence: Big Cypress National Preserve, Fakahatchee Strand Preserve State Park, and Florida Panther National Wildlife Refuge. Sites (>300 m apart) were in hardwood swamp habitat with surrounding freshwater wetlands. Python presence was estimated via a species distribution model (ensemble approach using biomod2 in R) fit to 1,165 georeferenced python observations (2014–2017) from EDDMapS; sampling bias was mitigated by bias-corrected background point generation. Mammal community sampling: Non-rodent (medium/large) mammal activity and diversity were quantified via stratified camera trapping along movement corridors (Bushnell Trophy Cam). Activity was measured as independent visits in 20-min increments; diversity metrics and evenness were calculated with Hill numbers using the vegan R package. Rodent activity was quantified primarily via live-trapping (including Sherman-style and Tauber folding traps), with trapping conducted four consecutive nights per month for three months; unique individuals informed minimum activity estimates. Indices included cotton rat activity, non-rodent activity, and non-rodent diversity. Mosquito collections and blood meal analysis: Female Culex (Melanoconion) mosquitoes were sampled using CO2-baited miniature light traps, pop-up/resting shelters, and daytime aspiration, over three monthly weeks of sampling per site. Blood-engorged females were processed individually. Genomic DNA was extracted (e.g., InstaGene Matrix) and vertebrate blood meal sources identified via PCR assays targeting mitochondrial cytochrome b and ribosomal 16S genes, with Sanger sequencing and GenBank comparisons. Virus detection and infection rate estimation: Mosquitoes were pooled by site and screened for Everglades virus (EVEV/EEVV) by PCR. Infection prevalence was estimated from pooled data (e.g., MLE per 1,000), with model weights accounting for the number of pools screened per site. Statistical analysis: Binomial GLMMs (glmmTMB in R 3.6.1) related (1) relative cotton rat host use (ratio of cotton rat blood meals to all other mammal blood meals) to mammal community metrics (cotton rat activity, non-rodent activity, non-rodent diversity) and (2) EVEV infection rates in vectors to relative cotton rat host use, cotton rat activity, and non-rodent diversity. Site-level random effects accounted for overdispersion and site heterogeneity; covariates were weighted by sample sizes (blood meals or pools screened). Model selection used AICc weights; effect sizes and confidence intervals were visualized using the effects package. Residual diagnostics included checks for spatial autocorrelation and residual clustering.

- Relative cotton rat host use varied widely among sites (0–63.4% of total blood meals at sites with ≥10 blood meals). - The most parsimonious GLMM for predicting relative cotton rat host use included cotton rat activity and non-rodent diversity (AICc weight ≈ 0.592). Both predictors were relevant; non-rodent diversity showed a strong negative association with cotton rat host use. - Model predictions suggested relative cotton rat host use increased substantially as non-rodent diversity decreased (approximately fivefold change across the observed diversity gradient), indicating loss of dilution hosts concentrates vector feeding on cotton rats. - For EVEV infection rates, models consistently identified relative cotton rat host use as the key positive predictor; it appeared in the top models and was the only variable with consistent relevance across the most parsimonious models. Increases in relative cotton rat host use from 0 to 50% of blood meals were associated with an approximately threefold increase in EVEV infection rates in mosquitoes. - Non-rodent activity and species richness declined sharply with increasing modeled python presence. At sites with the highest python presence (Big Cypress), non-rodent activity was very low (≈0.008–0.013), while sites with lower python presence had 2.8–14 times higher non-rodent activity (mean ≈9.1-fold). - The highest observed infection rate (example: 6 positives among 1,545 females; MLE ≈ 32.34) occurred at a site with the lowest mammal diversity and exclusively rodent blood meals, supporting the mechanism that absence of dilution hosts increases feeding on amplifying hosts and elevates infection prevalence. - Candidate dilution hosts included marsh rabbit and white-tailed deer; these species were preferred by the vector at sites with low to moderate python presence but were absent from high python presence sites.

Findings support an indirect pathway whereby an invasive predator (Burmese python) reduces non-rodent mammal diversity and activity, shifting mosquito feeding toward a competent amplifying host (cotton rat) and increasing Everglades virus infection prevalence in vectors. The strong negative relationship between non-rodent diversity and cotton rat blood meal ratio indicates community composition primarily affects infection indirectly by altering vector host use rather than directly impacting infection rates. Identification of potential dilution hosts (marsh rabbit, white-tailed deer) suggests that maintaining medium and large mammal populations can buffer pathogen transmission. Patterns mirror other vector-borne systems (e.g., Lyme disease) where host community diversity and vector host selection modulate pathogen prevalence. These results emphasize that conserving mammal diversity can mitigate zoonotic risk and that invasive predator management may have public health benefits by maintaining dilution host communities.

The study demonstrates that invasive Burmese pythons restructure mammal communities in southern Florida, decreasing non-rodent diversity and indirectly elevating Everglades virus infection prevalence in mosquito vectors by increasing feeding on cotton rats. GLMM analyses highlight non-rodent diversity and cotton rat activity as key drivers of vector host use, and relative cotton rat host use as the primary predictor of vector infection rates. Conservation of diverse mammal communities and management of invasive predators may reduce zoonotic transmission risk. Future work should improve inference by expanding spatial replication, integrating remote sensing, refining python abundance estimation, and incorporating additional ecological drivers (e.g., flooding, fire, invasive plants) to better predict human risk.

- Python presence was inferred from citizen science (EDDMAPs) observations, potentially biased by uneven effort and road-based detections; background sampling bias correction was applied, but uncertainties remain. - Mosquito infection prevalence estimates were based on variable sample sizes across sites; smaller samples yield wider confidence intervals. Models weighted estimates by the number of pools screened. - Higher infection prevalence in vectors does not directly translate to human disease risk, which also depends on human–vector contact patterns. - Some site-level extremes (e.g., high cotton rat host ratio and high infection rate) may disproportionately influence model outcomes. - No rigorous, standardized methods for python population sampling are yet available, limiting precision of predator pressure estimates.