Dynamics of influenza transmission in vampire bats revealed by longitudinal monitoring and a large-scale anthropogenic perturbation

Abstract

Interrupting pathogen transmission between species is a priority strategy to mitigate zoonotic threats. However, avoiding counterproductive interventions requires knowing animal reservoirs of infection and the dynamics of transmission within them, neither of which are easily ascertained from the cross-sectional surveys that now dominate investigations into newly discovered viruses. We used biobanked sera and metagenomic data to reconstruct the transmission of recently discovered bat-associated influenza virus (BIV; H18N11) over 12 years in three zones of Peru. Mechanistic models fit under a Bayesian framework, which enabled joint inference from serological and molecular data, showed that common vampire bats maintain BIV independently of the now assumed fruit bat reservoir through immune waning and seasonal transmission pulses. A large-scale vampire bat cull targeting rabies incidentally halved BIV transmission, confirming vampire bats as maintenance hosts. Our results show how combining field studies, perturbation responses, and multi-data-type models can elucidate pathogen dynamics in nature and reveal pathogen-dependent effects of interventions.

Fig. 1.. Observed and model-predicted seroprevalence of influenza virus across three zones of Peru.

In (A) to (C), solid lines show the mean predicted trajectory for the recovered class in an SEIR model. Shaded regions show the 95% credible intervals (CIs) for the (A) North, (B) Central, and (C) South zones. Points show monthly population-level seroprevalence data with binomial confidence intervals, with observations that fell outside the 95% CI of the posterior cumulative distribution shown in black. The gray bars in (C) show monthly applications of vampiricide during a culling campaign in the South. Black points above each time series show the collection dates for samples subjected to metagenomic sequencing. (D) Colors show the administrative regions of Peru, which were grouped for this study, and the points show the locations of sampled bat colonies. (E) A summary of change in serostatus in the paired samples from longitudinally sampled bats where 0 = seronegative and 1 = seropositive. N, North; C, Central; S, South.

Fig. 2.. Inference from multiple data types refines the within host biology of BIV.

Prior distributions (dashed) and posterior distributions for (A) δ (incubation period; yellow) and α (infectious period; red), and (B) γ (immune period; blue) produced by model fitting to the full dataset including population level serology, individual-level serology from recaptures, and metagenomic sequencing. Gray lines show the posterior distribution for each parameter when fitted to the base dataset of only the population-level serology data. The mean of each posterior distribution is shown by a solid vertical line, and 95% CIs are shown by the shaded areas. Full comparison of data inclusion can be found in fig. S5. (C) Model fitting to simulated data comparing the full dataset and the base dataset. The x axis of each graph shows the input parameter value, and the y axis shows the mean of the posterior distribution when model fitting to test data. The shaded areas show 95% credible intervals. RMSE, root mean square error.

Fig. 3.. Regionally synchronized long-term dynamics with distinct seasonal peaks of BIV transmission and infection.

Trajectories for (A) the time varying rate of transmission β, and (B) the proportion of the population actively infected by BIV predicted by model fitting for each of the three zones of Peru. Solid lines show the mean trajectory, and shaded areas show the 95% CIs. (C) The predicted seasonal peak of transmission varied by geographic zone. Box plots show the median (horizontal black line), quartiles (colored area), and range (vertical black line) of the predicted seasonal peaks.

Fig. 4.. Culling vampire bats reduces BIV transmission.

(A) Posterior (solid blue line) and prior (dashed line) distributions of the culling scaling factor (csf) evaluating the impact of culling on BIV transmission. The mean of the posterior is shown by the vertical blue line, and the 95% credible interval by the shaded area. (B) The predicted trajectory for the total vampire bat population in the South zone, as a proportion of the starting population. The mean predicted trajectory is shown by the solid black line, and the 95% credible interval by the shaded area. The predicted trajectory for the infected individuals in the presence (orange line) or absence (gray line) of culling is shown as a proportion of the population at each point in time, i.e., of the culled population size for the orange line. Vertical dashed lines show the start and end dates of the culling campaign.