Novel parasite introductions are threatening wildlife communities globally. The degree of threat depends on the ecological context, namely host community characteristics and the presence of other parasites. How these two factors interact to shape the outcomes of parasite introductions is not known. We experimentally investigated how infection following the introduction of the amphibian parasite, Ranavirus (Rv), is influenced by host species identity, host species richness, and previous exposure to the fungal parasite, Batrachochytrium dendrobatidis (Bd). We varied Bd exposure history prior to introducing Rv in host communities of either larval common toads (Bufo bufo), larval common frogs (Rana temporaria), or both host species together. Interactive effects of host species identity and host species richness were responsible for driving Rv infection patterns, more so than Bd exposure history. Notably, the host species at highest risk of Rv infection changed between one-species and two-species host communities. In one-species contexts, frog tadpoles were more likely to exhibit Rv infections than toad tadpoles. When the two species were co-housed, Rv infection probability in toads increased to levels comparable to frogs. Prior exposure to Bd did not alter these findings. Overall, host community context was the key predictor of Rv infection risk following its introduction into amphibian communities. These results show that host roles in pathogen maintenance may be dynamic and change depending on host community characteristics. Such changes likely have important implications for biodiversity-disease relationships that have so far remain overlooked.

Approval for the animal experiment was granted by the Zoological Society of London’s Animal Welfare and Ethics Review Board (AWERB). Legal permission for procedural pathogen exposure and non-schedule 1 methods of euthanasia were provided by the UK Home Office (project licence P8897246A).

Animal collection and rearing

With landowner permission, we collected toad and frog egg masses from ponds near London and transported them by car to Home Office facilities at the Institute of Zoology (IoZ), Zoological Society of London (London, UK). Clutches were kept in outdoor tanks separated by species and left to hatch into tadpoles. Rearing animals from egg masses eliminated the risk of pre-existing infections with either pathogen (Garner et al. 2009, Duffus et al. 2014). We co-housed tadpoles in 19L plastic containers with air-driven sponge filters and natural plant material (Brazilian waterweed, Elodea densa). The containers contained aged tap water that was tested weekly for nitrates, nitrites, and ammonia levels. Water changes of ten percent total volume were carried out every other day, and animals were fed 2 Tetra TabiMin sinking tablets every other day. We reared tadpoles to Gosner stage 25 (Gosner 1960), consistent with previous infection experiments (Garner et al. 2009), at which point we transferred them to indoor animal facilities where the experiment was carried out. Rooms were kept at 18°C, temperatures suitable for all hosts and parasites (Garner et al. 2009, Duffus et al. 2014, Voyles et al. 2017, Price et al. 2019), and with a 14-hr light cycle (6am-8pm). We held tadpoles in 19L plastic containers containing 15L aged tap water. We co-housed each species separately and regardless of future Bd exposure treatment, at stocking densities of 100 individuals, until experimental host communities were formed (see below).

Experimental procedures

We carried out a mesocosm experiment following a factorial design to test the effects of sequential Bd-Rv exposure on the probability (i.e. risk) and intensity (mean infection loads) of Rv infection in both single-host and two-host communities (Fig. 1). We only considered scenarios in which Bd exposure occurred before Rv exposure to reflect the sequential histories of emergence of the two parasites in certain European localities (Rosa et al. 2017), to mimic the likely sequence of within-season proliferation of the two parasites (Bosch et al. 2020), and to keep sample sizes feasible. Although our focus was on infection outcomes of the secondary parasite, Rv, the experimental design also permitted assessments of how subsequent introduction by Rv may have altered existing infections of the background parasite, Bd. The experiment followed three procedures: (1) Bd exposure, (2) assembly of host communities, and (3) Rv exposure (Fig. 1). For Bd exposures, we randomly assigned toad and frog tadpoles to one of two treatments: (1) sham exposures to sterile media (control) or (2) exposure to live Bd zoospores in media. Bd transmits via free-swimming aquatic zoospores (Longcore et al. 1999). We administered six Bd  doses (GPL, Spanish isolate) into containers over two weeks of the experiment by pipetting 150,000 live zoospores, counted with a haemocytometer (Daversa et al. 2018), in liquid media into the 19L containers at a stocking density of 100 tadpoles. For sham exposures, we pipetted zoospore-free liquid media into tanks at the same volume. Because Bd can establish infection and begin shedding zoospores within one week post exposure (Maguire et al. 2016, DeMarchi et al. 2026), Bd-exposed tadpoles could have been exhibiting robust infections at the time of assembling communities, creating potential for subsequent interaction with Rv. Immune responses to Bd may have also created opportunities for (indirect) parasite-parasite interactions, but direct evidence of the timing and consequence of immune responses to Bd are lacking.

One week following the last Bd exposure, we assigned tadpoles to one of three host communities: one species toad communities (10 toads, 0 frogs), one species frog communities (0 toads, 10 frogs), or two species mixed communities (5 toads, 5 frogs), split by Bd exposure treatment. We housed communities in 5L plastic containers containing 3L of aged tap water for the duration of the experiment. This design therefore assumed substitutive assembly of host communities, which implicitly mimics a scenario of strong interspecific competition, whereby host species richness changed but the overall density of hosts was constant (Rudolf and Antonovics 2005). For logistical reasons we could not conduct both substitutive and additive (increasing richness and density simultaneously) designs, and we so chose a substitutive design to focus attention on the effects of host richness alone, rather than richness and density combined.

Two days after assembling host communities, we assigned them to one of two Rv exposure scenarios: exposure to Rv or a sham exposure. Rv has multiple transmission routes, including direct (host-to-host) transmission, environmental transmission and indirect transmission from necrophagy (feeding on carcasses or dead tissue) (Brunner et al. 2007). For this reason, we exposed host communities to Rv by providing Rv-exposed ‘seed’ carcasses as a source of Rv (Pearman et al. 2004, Brunner et al. 2007). We generated carcasses by flash-freezing common frog tadpoles that were previously bath exposed in separate containers (1.7L containing 425ml aged tap water) to either one dose of an FV3-like Rv (RUK13 isolate) in 14mL (10⁷ titre) of media (Rv present) or one dose of Rv-free media as a control (Rv absent). To confirm that exposures led to Rv infection, five reference frog tadpoles from the Rv-exposed group were assayed for Rv infection, all of which tested positive (see Rv assay details below). The carcasses were kept frozen until addition to experimental mesocosms, when they were thawed to room temperature. We removed carcasses from mesocosms after one week if they were not consumed or degraded.

The above procedures resulted in four parasite exposure histories (sham-sham, Bd-sham, sham-Rv, Bd-Rv), three host community compositions varying in species identity and richness (toad-only, frog-only, mixed) and twelve total treatments (2 Bd exposure treatments x 3 host species compositions x 2 Rv exposure treatments) in a fully factorial design. We replicated treatments six times, for a total of 72 experimental communities and 720 tadpoles (360 toad tadpoles, 360 frog tadpoles) (Table S1). We conducted daily health and welfare checks of tadpoles for the duration of the experiment. We fed tadpoles ground Tetra TabiMin sinking tablets every other day and did ten percent water changes with cleaning on non-feeding days for the duration of the experiment. If humane endpoints were observed (e.g., pathological symptoms of chytridiomycosis, Ranavirosis, including an inability to self-right, feed, maintain buoyancy) euthanasia was actioned via a non-schedule 1 method (MS-222 followed by submergence in 90% ethanol).  All tadpoles living at experiment day 14 were euthanized by an overdose of MS-222 and preserved in 90% ethanol for subsequent parasite detection assays.

Parasite detection assays

We detected Bd infection in tadpoles using standard molecular diagnostics involving DNA extraction and qPCR processing (Boyle et al. 2004). We extracted DNA samples from the mouthparts of tadpoles, which is the only keratinized part of larval bodies that supports Bd infection. Extraction of DNA was achieved using a bead-beating protocol (Bates et al. 2022), and extracts were prepared in a 1/10 dilution for qPCR. We used qPCR targeting the ITS1 gene region to measure the amount of Bd DNA in samples, and we ran samples in duplicate. We added negative controls and four concentration standards (0.1, 1, 10, 100) to each PCR plate, which served as positive controls and standards for generating estimation curves (Garner et al. 2009, Luquet et al. 2012, Bielby et al. 2015). We re-ran duplicate samples if only one sample amplified. Bd infection load is reported in genomic equivalents (GE), where one GE equals one Bd zoospore. A sample tested Bd positive when they amplified in duplicate and when the average GE of the two samples was greater than 0.01. We only considered the sample positive if both replicates tested positive.

DNA was extracted from the remaining frog tadpole carcasses for Rv detection. Again, we used a qPCR assay, this one targeting the viral major capsid protein (MCP) gene following Leung et al. (Leung et al. 2017). We included 10 standards ranging from 3 - 3 x 10⁹ viral copies to serve as positive controls. Again, we processed samples in duplicate and reran samples when only one replicate tested Rv-positive.

Data Analysis

We used generalized linear mixed models (GLMMs) in R (R Core Team 2023) to test the main and interactive effects of host species identity, host species richness, and prior exposure to Bd on Rv infection probability and intensity.  Although our primary focus was on Rv outcomes, we also analysed how the above factors influenced Bd infection probability and intensity (substituting Rv exposure for Bd exposure as an explanatory factor). Because host communities were assembled after Bd exposure in our experiments, no inferences about effects of host species richness on Bd infection probability and intensity were sought. For both parasites, models of individual infection status (infected or not) were run using the glmer() function with the lme4 package in R and used a binomial error structure (Bates et al. 2015). Models of individual infection intensity were run using the lmer() function with the lme4 package and used Bd and Rv loads as responses, expressed as genomic equivalents (GE) and viral copies respectively, with Gaussian error structures. We log-transformed infection intensity values to achieve normalization and only used data from hosts testing positive for infection. All models included the following fixed effects and their second-order interactions: host species identity (toads/frogs), host species richness (one species/two species), and history of exposure to the alternative parasite (exposed/unexposed). All models included tank ID as a random effect. We performed likelihood ratio tests based on a ‘Chi-squared’ distribution to determine the significance of the fixed effects and their interactions on model performance (using the dropterm function in R). In cases where statistically significant effects of a given factor were detected, we performed Tukey’s tests to evaluate specific differences between factor levels, using the emmeans package in R (Lenth 2023). Infection data included individuals that died before the end of the experiment.

We also evaluated host mortality using the survival package in R (Therneau 2020). Cox proportional hazard models included host species richness and parasite exposure treatment as fixed effects, as we did not have the power to explore interactive effects. An ‘event’ denoted death, which we tracked daily throughout the 14-day experimental period. Animals surviving on the 15^th day had an event status of 0, all others had an event status of 1 on the day we observed the mortality. We generated Kaplan-Meier plots in R to visualize the survival curves.