Bat species assemblage predicts coronavirus prevalence

The study addresses how changes in bat community composition and species diversity influence the prevalence and infection dynamics of coronaviruses (CoVs) in cave-roosting bats in Ghana. Against a backdrop of anthropogenic disturbance, habitat loss, and documented sensitivity of many bat species to environmental change, the authors hypothesize that shifts in bat assemblages—particularly increases in competent reservoir hosts and changes in age structure—alter CoV transmission and prevalence. Motivated by evidence that bats are important reservoirs for alpha- and beta-coronaviruses, frequent high-density roosting, and high mobility, the study explores whether lower species diversity is associated with higher CoV prevalence (consistent with a dilution-effect mechanism) and which community features (e.g., relative abundance of specific Hipposideros species and subadults) best predict infection risk.

The paper situates its research within a broad literature linking biodiversity changes to infectious disease risk. Prior work indicates bats are reservoirs for diverse viruses, including SARS-related beta-CoVs and alpha-CoVs linked to human HCoV-229E and NL63. The perceived high viral richness in bats may reflect sampling focus and their species richness. Bats' unique immune adaptations related to flight and their social, high-density roosting behavior can facilitate pathogen transmission. Habitat alteration and human encroachment can increase contact and crowding, potentially elevating spillover risk. The dilution effect hypothesis posits that higher host diversity can reduce disease transmission by increasing the proportion of non-competent hosts and altering transmission pathways; however, outcomes are context-dependent and influenced by host community composition, abundances, age structure, and ecological interactions. Empirical studies across taxa show mixed evidence, highlighting the need to disentangle effects of diversity per se from changes in host abundance and competence. This study contributes correlational evidence in a multi-host, multi-pathogen bat system while acknowledging these complexities.

Study system and sampling: Over two years, approximately 2,300 cave-dwelling bats were captured at five roosting caves in central Ghana, sampled bimonthly. Most species were identified morphologically; for cryptic taxa within the Hipposideros caffer complex, mitochondrial cytochrome b sequencing was used to resolve species identity for 1,172 individuals. Community composition and age structure (subadult vs adult) were recorded per site and period.

Virus screening: Fecal material preserved in RNAlater was processed for RNA extraction using the MagNA Pure 96 System (Roche). Real-time reverse transcriptase PCR assays targeted four CoV clades: alpha-CoV 229E-like; beta-CoV 2b; beta-CoV 2b basal; and MERS-like beta-CoV 2c, using clade-specific primers/probes (Supplementary Methods/Table S6). Positive cases were defined as Ct ≤ 30.6 (equivalent to >15 CoV RNA copies/µL). CoV prevalence was calculated as the proportion of bats with detectable CoV RNA per site and sampling period. Viral load comparisons used Ct values among positive individuals.

Statistical analysis: Community metrics (species richness, Shannon diversity, Simpson’s reciprocal index, evenness) and relative abundance of subadults were computed. Differences in species community composition across sites and periods were analyzed with mixed models (lme4). Infection probability (binary) was modeled using generalized linear mixed-effects models (glmmTMB), with fixed effects including one diversity index (species richness, Shannon, or Simpson), the relative abundance of common species (Hipposideros abae, H. caffer B/C/D, Coleura, Nycteris macrotis), and relative abundance of subadults; random effects were sampling period nested within site. Model selection used AICc (MuMIn::dredge), with models having ΔAICc ≤ 2 considered competitive; full-model results are reported for comparability. Collinearity was assessed (VIF), and highly correlated predictors were modeled separately. Multiple testing was controlled via FDR correction. Correlations between CoV prevalence and diversity were assessed using Spearman’s rank correlation. Viral load differences (Ct) by species and age were tested via ANOVA/t-tests.

• Community composition varied significantly across sites (F3,96 = 6.685; R2 = 0.31; p < 0.001) and across sampling periods (F1,19 = 1.919; R2 = 0.25; p = 0.019), with strong spatial heterogeneity and seasonal age-structure dynamics.
• Of 326 bats screened across five families, 113 CoV infections were detected among 15 species. Infections encompassed four clades: alpha-CoV 229E-like, beta-CoV 2b, beta-CoV 2b basal, and beta-CoV 2c. CoV presence was uneven among closely related species, with Hipposideros species frequently infected. Some CoVs (alpha 229E-like, beta 2b) behaved as multi-host pathogens, whereas beta 2b basal and 2c were detected in fewer host species.
• The most abundant species (e.g., Hipposideros caffer D) hosted multiple CoV clades.
• Diversity–prevalence correlations: alpha-CoV 229E-like prevalence showed a weak negative correlation with Shannon diversity (r(55) = -0.33, p = 0.015); beta-CoV 2b showed a moderate negative correlation (r(55) = -0.51, p < 0.001). Diversity indices emphasizing abundance/evenness exhibited stronger negative associations with prevalence.
• GLMMs (odds ratios, 95% CI, FDR-adjusted p): • alpha-CoV 229E-like: Higher Shannon diversity reduced infection odds (OR 0.32, 0.13–0.80, padj = 0.019). Greater relative abundance of H. caffer C increased odds (OR 1.02, 1.01–1.03, padj = 0.042). Higher relative abundance of subadults increased odds (OR 1.01, 1.00–1.03, padj = 0.034). • beta-CoV 2b: Higher Shannon diversity strongly reduced infection odds (OR 0.05, 0.02–0.14, padj < 0.001). Higher relative abundance of H. caffer D increased odds (OR 1.04, 1.03–1.06, padj < 0.001). Higher relative abundance of H. abae decreased odds (OR 0.97, 0.96–0.99, padj = 0.003). Higher subadult abundance increased odds (OR 1.01, 1.00–1.02, padj = 0.011). In an alternative competitive model, greater N. macrotis abundance reduced odds (OR 0.98, 0.96–1.00, padj = 0.042), while subadults remained positively associated (OR 1.02, 1.01–1.03, padj < 0.001), and Shannon diversity remained protective (OR 0.13, 0.05–0.34, padj < 0.001).
• Age and viral load: Viral load (Ct) differed by species for both CoVs (alpha-CoV 229E-like: T3,10 = 3.23, p = 0.023; beta-CoV 2b: T4,6 = 7.45, p < 0.001). Subadults had lower Ct values (higher viral loads) than adults for both viruses, indicating more acute infections and potentially higher transmissibility in younger bats.
• Overall pattern: Less diverse bat communities had higher prevalence of multi-host CoVs, coinciding with higher relative abundance of competent hosts and more subadults.

Findings support a diversity–disease relationship consistent with the dilution effect in this multi-host bat system: sites with lower species diversity exhibited higher prevalence and infection probability for alpha-CoV 229E-like and beta-CoV 2b. This pattern appears driven by community reassembly favoring competent host species (e.g., H. caffer D) in less diverse, likely disturbed assemblages, alongside a greater proportion of subadults who showed higher viral loads and increased infection likelihood. Thus, both the taxonomic composition (competent vs non-competent hosts) and demographic structure (age) of bat communities shape CoV dynamics. These results underscore that community evenness and the relative abundance of key species, rather than richness alone, strongly influence infection risk. However, the study emphasizes context dependence: diversity–disease outcomes vary with specific host–pathogen ecologies and changes in host abundances, making generalization beyond the studied system cautious. The work aligns with One Health principles, suggesting biodiversity conservation and maintaining more even, species-rich bat communities may reduce the circulation of CoVs with zoonotic potential.

The study demonstrates that bat species assemblage—particularly species diversity, evenness, and the relative abundance of competent hosts and subadults—predicts CoV prevalence and infection risk in cave-roosting bats in Ghana. Multi-host coronaviruses (alpha-CoV 229E-like and beta-CoV 2b) were more prevalent where communities were less diverse and dominated by competent species, with subadults contributing to higher infection likelihood through higher viral loads. These findings provide correlational support for a dilution-effect mechanism in a multi-host, multi-pathogen context and highlight the importance of community composition and demography in disease ecology. Future research should experimentally and longitudinally disentangle biodiversity effects from host abundance and competence, clarify mechanisms underlying age-related infection dynamics, integrate immunogenetic and microbiome data, and assess how habitat disturbance and conservation actions modify host community structure and pathogen transmission over time.

• Correlational design limits causal inference; diversity metrics covary with host abundances and evenness, making it difficult to disentangle biodiversity effects from changes in competent host dominance.
• Community and infection data were analyzed at site-period levels; although recapture rates were low and missing sampling events were not corrected for, residual temporal/spatial confounding may remain.
• Some predictor sets exhibited multicollinearity, necessitating separate models and potentially limiting joint inference across highly correlated variables.
• Virological screening numbers (subset of total captured bats) may affect precision of prevalence estimates across all species.
• Mechanistic drivers of higher infection in subadults (behavior, immunity, microbiome) remain unresolved; immunogenetic associations are suggestive but not definitive.
• Findings are context-dependent to Ghanaian cave bat communities and specific CoV clades; generalizability to other regions, host taxa, and pathogens may be limited.