Species interactions in a parasite community drive infection risk in a wildlife population

Abstract

Most hosts, including humans, are simultaneously or sequentially infected with several parasites. A key question is whether patterns of coinfection arise because infection by one parasite species affects susceptibility to others or because of inherent differences between hosts. We used time-series data from individual hosts in natural populations to analyze patterns of infection risk for a microparasite community, detecting large positive and negative effects of other infections. Patterns remain once variations in host susceptibility and exposure are accounted for. Indeed, effects are typically of greater magnitude, and explain more variation in infection risk, than the effects associated with host and environmental factors more commonly considered in disease studies. We highlight the danger of mistaken inference when considering parasite species in isolation rather than parasite communities.

Fig. 1

The web of interactions between microparasite species within this community, showing the magnitude of effects. All associations shown obtained overwhelming support (accumulative weight in models >0.9) (see table S3). Positive associations are shown in red [odds ratio (OR) >1] and negative associations (OR < 1) in blue, with the strength of the line color reflecting the magnitude of the effect. ORs (exp β) (see table S3) relative to uninfected individuals, with 95% confidence intervals in brackets. Thus an individual with a new B. microti infection is ~5 times as likely to become infected with A. phagocytophilum than an uninfected individual (OR = 5.43). Infection history associated with each effect is also indicated: N, negative, P, positive. Thus, NP indicates no infection at t₋₁, infection at t₀. −P is used to signify that NP and PP show similar effects. Because B. microti induces a chronic infection, there are three infection status histories (uninfected, chronic, and new infections). For cowpox virus, a probability of infection was used in analyses, and results are for individuals that sero-converted at t₀ (i.e., NP infection history, probability of infection at t₀ = 0.5) (12).

Fig. 2

Predicted probabilities of acquiring an infection depending on infection status by other parasites for (A) A. phagocytophilum (AP), (B) Bartonella spp. (BT), (C) B. microti (BM), and (D) cowpox virus (CP). Predictions are based on the models in table S3. Gray bars represent individuals without other infections (uninf); red and blue bars highlight positive and negative effects on susceptibility, respectively. Each bar is labeled with the infection (e.g., CP) and the infection history (see Fig. 1 for explanations) associated with each effect. Error bars represent 95% confidence intervals, averaged over random effects. Predicted probabilities are defined with reference to an 18-g male in July at one specific site. In cases where NP and PP show similar effects, predicted probabilities are based on an NP infection history.