The impact of within-host coinfection interactions on between-host parasite transmission dynamics varies with spatial scale

Abstract

Within-host interactions among coinfecting parasites can have major consequences for individual infection risk and disease severity. However, the impact of these within-host interactions on between-host parasite transmission, and the spatial scales over which they occur, remain unknown. We developed and apply a novel spatially explicit analysis to parasite infection data from a wild wood mouse (Apodemus sylvaticus) population. We previously demonstrated a strong within-host negative interaction between two wood mouse gastrointestinal parasites, the nematode Heligmosomoides polygyrus and the coccidian Eimeria hungaryensis, using drug-treatment experiments. Here, we show this negative within-host interaction can significantly alter the between-host transmission dynamics of E. hungaryensis, but only within spatially restricted neighbourhoods around each host. However, for the closely related species E. apionodes, which experiments show does not interact strongly with H. polygyrus, we did not find any effect on transmission over any spatial scale. Our results demonstrate that the effects of within-host coinfection interactions can ripple out beyond each host to alter the transmission dynamics of the parasites, but only over local scales that likely reflect the spatial dimension of transmission. Hence there may be knock-on consequences of drug treatments impacting the transmission of non-target parasites, altering infection risks even for non-treated individuals in the wider neighbourhood.

Keywords: coinfection; disease ecology; disease spread; parasite ecology; parasite interactions; spatial ecology.

Figure 1.

Example trapping grid with wood mouse neighbourhoods. This trapping grid shows the relative locations of traps, with the current focal animal shown in black. Neighbouring traps are colour-coded either red for an animal infected with the nematode H. polygyrus, blue for an uninfected animal, or blank for no capture within the specified time window of the focal's capture. Five example neighbourhoods of increasing size around the focal are shown in dotted outline. The prevalence of nematode infection within each neighbourhood size of the focal is then calculated as the proportion of infected animals (red points) out of the total number of animals caught (red + blue points) within that neighbourhood, for the specified time window.

Figure 2.

Neighbourhood analysis of parasite data from wood mice (Apodemus sylvaticus) showing the associations between the effect size from Bayesian GLMMs of the neighbourhood-level prevalence of the GI nematode Heligmosomoides polygyrus on focal individual infection intensity of (left column) Eimeria hungaryensis (known from previous experimental perturbations [4], to undergo strong within-host coinfection interactions with the H. polygyrus), and (right column) E. apionodes (known not to undergo strong within-host interactions with H. polygyrus) for increasing neighbourhood sizes (x-axes). Each row shows results for different time windows between captures of animals in the neighbourhood and subsequent capture of the focal individual. Figures show median model estimates (points) and 95% credible intervals (bars) for each neighbourhood size and time window; see Material and methods for full specification of model structure. Red points denote neighbourhood sizes for which the 95% credible intervals do not cross 0; black and grey points denote neighbourhood sizes for which 95% credible intervals do cross 0.

Figure 3.

Results from neighbourhood analysis showing the associations between the effect size (from GLMMs) of the neighbourhood-level prevalence of the GI nematode H. polygyrus on focal individual E. hungaryensis infection intensity for increasing neighbourhood sizes, applied to a dataset that randomly assigned observed pairs of H. polygyrus and E. hungaryensis infection data from each animal to the spatial and temporal capture records of other animals in the dataset. Each panel shows results for different time windows between captures of animals in the neighbourhood and subsequent capture of the focal individual. Figures show median model estimates (points) and 95% credible intervals (bars) from Bayesian GLMMs of the effect of neighbourhood prevalence of H. polygyrus on individual-level intensity of E. hungaryensis, at each neighbourhood size, for each time window.