Amphibian larvae benefit from a warm environment under simultaneous threat from chytridiomycosis and ranavirosis

Rising temperatures can facilitate epizootic outbreaks, but disease outbreaks may be suppressed if temperatures increase beyond the optimum of the pathogens while still within the temperature range that allows for effective immune function in hosts. The two most devastating pathogens of wild amphibians, Batrachochytrium dendrobatidis (Bd) and ranaviruses (Rv), co-occur in large areas, yet little is known about the consequences of their co-infection and how these consequences depend on temperature. Here we tested how co-infection and elevated temperatures (28 and 30°C vs. 22°C) affected Bd and Rv prevalence, infection intensities, and resulting mortalities in larval agile frogs and common toads. We found multiple pieces of evidence that the presence of one pathogen influenced the prevalence and/or infection intensity of the other pathogen in both species, depending on temperature and initial Rv concentration. Generally, the 30°C treatment lowered the prevalence and infection intensity of both pathogens, and, in agile frogs, this was mirrored by higher survival. These results suggest that if temperatures naturally increase or are artificially elevated beyond what is ideal for both Bd and Rv, amphibians may be able to control infections and survive even the simultaneous presence of their most dangerous pathogenic enemies.


Introduction
During recent decades amphibians have experienced dramatic declines [1] and have become one of the most threatened vertebrate taxa, pushing them into the forefront of conservation efforts [2]. The causes behind population declines and extinctions are complex [3], with multiple stressors acting in synergy, but emerging infectious diseases likely play a decisive role [4]. The chytrid fungus Batrachochytrium dendrobatidis (Bd) was linked directly to the extinction of dozens of amphibian species and caused the worst pandemic of wildlife ever recorded in history [1,4]. At the same time, epidemics caused by Ranaviruses (family Iridoviridae, hereafter Rv) have also led to mass mortality events, resulting in local extinctions of amphibians [5][6][7].
Typically, several pathogens are present in natural populations [8], and these are likely to interactively determine consequences for hosts. Accordingly, co-infections are increasingly recognised as important drivers of disease dynamics [9, 10]. Co-infections can be disadvantageous, insignificant or even beneficial to hosts, and multiple levels of interactions can influence the outcomes [11]. Direct and indirect interactions among infectious agents within hosts include interference competition for physical space [12], exploitation competition for resources [13,14], as well as indirect interactions mediated by cross-reaction immunity and the immunosuppression of the host [15,16]. The outcome of co-infections is also shaped by the species-specific infectivity and virulence of pathogens [17], arrival order (i.e., 'priority effect' [18,19]), species and body condition of hosts [20,21], and abiotic factors [11].
The highly pathogenic and widespread Bd and Rv frequently co-infect amphibians [8,[22][23][24], and this can cause mass die-offs [25]. Naturally co-infected amphibians tend to experience higher Bd infection loads during simultaneous infection with Rv than individuals exclusively infected with Bd [22], while Rv infection intensity tends to be negatively associated Similarly, an increase in temperature from 10 to 25°C resulted in higher mortality and Rv copy numbers in tadpoles of four amphibian species exposed to FV3 [47]. In contrast, Rv infection probability and mortality were lower at 22°C than at 14°C in two species of Lithobates frogs infected with three different FV3 strains [48]. While these studies delivered clear evidence for the importance of temperature in the case of both chytridiomycosis and ranavirosis, contradictions in patterns may be due to interspecific differences in the temperature dependence of amphibian immune functions or to some other factor remaining to be explored. Importantly, manipulative studies testing how elevated temperature affects disease progression in amphibians co-infected with these two pathogens are lacking entirely.
Hence, to clarify the effects of high temperatures on disease progression in amphibians during single and co-infections with Bd and Rv, we experimentally infected tadpoles of agile frogs (Rana dalmatina) and common toads (Bufo bufo) and subsequently exposed them to elevated temperatures for six days, and finally assessed infection patterns and survival. We thereby aimed to deliver information about the effects of high temperatures that can occur under natural conditions on the severity of consequences of Bd and Rv infection and, especially, on the outcomes of co-infection. We also wanted to provide information for the parameterisation of predictive models on the consequences of climate change on future dynamics of wildlife diseases. Finally, we intended to test the potential of localised heating, an in situ mitigation method relying on the thermal treatment of Bd-infected amphibians [49], by assessing whether elevated temperatures decrease Bd prevalence and intensity without resulting in elevated Rv infection loads and excessive mortality in the presence of Rv.

Experimental design and procedures
We applied a full-factorial design with three thermal treatments: 22, 28 and 30°C, combined with six infection treatments: uninfected control ('control'); exposed to Bd ('Bd'); exposed to Rv in a low concentration ('Rv-low'); exposed to Rv in a high concentration ('Rv-high'); coexposed to Bd and Rv in a low concentration ('Bd + Rv-low'); co-exposed to Bd and Rv in a high concentration ('Bd + Rv-high'). We replicated each treatment combination 20 times (two individuals from each of ten families in each treatment combination) for a total of 360 animals per species. The experimental procedure started with a 19-days Bd treatment, followed by a 24hours Rv treatment and, finally, a 6-days thermal treatment (for a schematic representation, see On day 19, we haphazardly selected eight tadpoles from each rearing container and randomly assigned them to Rv × temperature combinations. Then we exposed tadpoles to FV3 by applying one of two concentrations, i.e., Rv-low (6.12×10 3 plaque forming unit (pfu)×ml -1 ) and Rv-high (6.25×10 5 pfu×ml -1 ) during 24-hour exposures where tadpoles were challenged individually in plastic cups containing RSW and the corresponding concentration of Rv, while control tadpoles received sham extract (for more details on experimental infection with Rv, see the Supplementary material). Subsequently, to assess initial infection prevalence and intensity for both pathogens, we haphazardly selected 20 tadpoles from each of the six infection treatment groups (12 tadpoles from each family, 120 individuals in total) and preserved them in 96% ethanol.
On day 20, we started thermal treatments as described in Ujszegi et al. [

Statistical analyses
We analysed data on the two species separately. In the main text, we focus on infection status after the thermal treatment; see the Supplementary material for analyses before thermal treatment. Because of low Bd prevalence in agile frogs and low variance in infection intensities of both pathogens in common toads (see Table S2 and Table S3), we did not perform statistical modelling in these three cases. To analyse the effects of co-infection, thermal treatment, and their interaction on pathogen prevalence after temperature treatment, we used generalised linear mixed-effects models (GLMM) with binomial error, including only those treatment groups that had been exposed to the given pathogen. To analyse the effect of treatments and their interactions on Rv infection intensity after temperature treatment within Rv-positive agile frogs, we used a linear mixed-effects model (LMM). We analysed the survival of agile frog tadpoles during thermal treatment using a mixed-effects Cox's proportional hazards model (COXME); mortality of common toads was negligible (1.1%; see Table S3), so we did not analyse it. In all models, we used family as a random factor. After stepwise model simplification, we performed pairwise comparisons by calculating linear contrasts and applying false discovery rate correction. For further details regarding the statistical analyses, see the Supplementary material.

Agile frogs
Initial Bd-prevalence, as assessed before thermal treatments, was extremely low: two out of 60 Bd-exposed individuals carried the fungus. Initial Bd infection intensities were also very low (Table S2). In contrast, the initial prevalence of Rv in Rv-exposed tadpoles was between 73.7 (Table S2) and 100% and Rv infection intensities were low to moderate ( Fig. S1; for details, see the Supplementary material).
By the end of the 6-days thermal treatments, Bd prevalence in Bd-exposed agile frog tadpoles remained low, with only three tadpoles testing positive, all in the Bd + Rv-low treatment (Table S3). Infection intensity of Bd was also very low after thermal treatments (<12.2 Genomic Equivalents (GE); Table S3).
The prevalence of Rv in Rv-exposed tadpoles after thermal treatments varied between 21.4 and 100% (Table S3; Fig. 2A) and was higher in the Rv-high treatment (GLMM; χ 2 1 = 14.49, P < 0.001) and in tadpoles not co-exposed to Bd (χ 2 1 = 13.34, P < 0.001). There was a tendency for thermal treatment to affect Rv prevalence (χ 2 2 = 4.62, P = 0.099), where the lowest infection probability was at 28 and the highest at 22°C (Table S4). The two-way interactions were non-significant (all P > 0.37).
The two-way interactions were all non-significant as well (all P > 0.17), but the three-way interaction between thermal treatment, previous exposure to Bd and Rv concentration were significant (χ 2 2 = 7.85, P = 0.02; Fig. 2A). To scrutinise the pattern behind this interaction we separately analysed the treatment groups exposed to the low and the high Rv concentration. In the treatment groups exposed to the low Rv concentration, the interaction between previous exposure to Bd and thermal treatment was significant (χ 2 2 = 13.12, P = 0.001), where Rv infection intensity tended to be higher in the previously Bd-exposed treatment groups, but this effect was abolished by the 30°C thermal treatment ( Fig. 2A; Table S5). In treatment groups exposed to the high Rv concentration, the two-way interaction between previous exposure to Bd and thermal treatment did not reach significance (χ 2 2 = 3.48, P = 0.18), and previous exposure to Bd did not have an effect (χ 2 1 = 1.49, P = 0.22), but thermal treatment did (χ 2 2 = 14.29, P < 0.001). Tadpoles in the 30°C treatment exhibited lower Rv copy numbers than those maintained at 22°C ( Fig. 2A; Table S6).

Common toads
The initial prevalence of Bd varied between 5 and 42%, and Bd infection intensities remained low until the start of thermal treatments (Table S2). At the same time, the initial prevalence of Rv varied between 20 and 100% and initial Rv infection intensities were low to moderate (Table   S2; for details, see the Supplementary material).
After thermal treatments, the prevalence of Bd in Bd-exposed common toad tadpoles varied between 0 and 45% (Table S3). The thermal treatment (GLMM; χ 2 2 = 10.943, P = 0.004) significantly influenced Bd prevalence, but the presence and concentration of Rv did not (χ 2 2 = 2.379, P = 0.304; Fig. 2C). The prevalence of Bd was higher in tadpoles treated at 22°C compared to those kept at 30°C (Table S7). The interaction between thermal treatment and the presence and concentration of Rv was non-significant (χ 2 4 = 3.681, P = 0.45).
The intensity of infection with Bd was highest after 22 °C treatments in all Bd-exposed groups, with a dramatic drop in GE values in groups receiving 28 and 30°C treatments (Table   S3; Fig. 2C). The effect of co-exposure to Rv could not be assessed across all thermal treatments due to the very low prevalence observed at higher temperatures, but within the 22 °C thermal treatment it was non-significant (χ 2 2 = 0.065, P = 0.968).
The prevalence of Rv in Rv-exposed tadpoles varied between 0 and 40% (  2B) and was significantly influenced by thermal treatment (χ 2 2 = 11.92, P = 0.002), Bd coexposure (χ 2 1 = 5.43, P = 0.02) and Rv concentration (χ 2 1 = 8.84, P = 0.003). The prevalence of Rv was higher when tadpoles were treated at 22°C compared to 28 or 30°C, while it did not differ between animals treated at 28 and 30°C (Table S7; Fig. 2B). Also, Rv prevalence was higher in the absence of Bd co-exposure and after receiving a high Rv concentration (Table S7; Fig. S3). Copy number of Rv was highest after the 22°C treatment, especially when tadpoles had been exposed to the high Rv concentration in the absence of Bd (Fig. 2B). Our study also revealed inter-specific differences in pathogen resistance and tolerance.

Previous reports showed that interactions between
We found that agile frogs were highly resistant to the chytrid fungus but susceptible to ranaviral infection. At the same time, common toads were moderately resistant to both pathogens. These differences in prevalence were mirrored by patterns in mortality. Agile frogs exposed to Rv In summary, our results suggest that high temperatures may be beneficial to amphibians exposed to both Bd and Rv. Also, while previous exposure to Bd affected Rv prevalence and in some treatment combinations Rv replication as well, superinfection with Rv did not influence the replication of Bd. Finally, temperature and co-infection appeared to also interact in their effects on pathogen replication and disease progression. Nonetheless, in our study, both species exhibited relatively low prevalence and infection intensities except for Rv in agile frogs, so we urge further experimental studies on more susceptible species to scrutinise the effects of coinfection and external factors modulating its outcomes in amphibians.  Widespread co-occurrence of virulent pathogens within California amphibian communities.

Animal collection and husbandry
In March and April 2019, we collected ca. 100 eggs from each of ten freshly laid clutches of agile frogs and common toads from natural populations located in and around Budapest, We reared the embryos separated according to sib-groups (families) in plastic boxes (25 × 17 × 13 cm) holding 1-L reconstituted soft water (RSW) and maintained a temperature of 16.3 ± 0.3°C (mean ± SD) and a 12:12 h light:dark cycle. Five days after hatching, the laboratory air temperature was raised to 18.45 ± 0.5°C (mean ± SD), and we maintained a 13:11 h light:dark cycle. At this time, each sib-group was haphazardly divided into groups of 10 larvae and placed into plastic rearing containers (37 × 27 × 16.5 cm), each holding 10 litres of reconstituted soft water (RSW). We changed the RSW twice a week and fed tadpoles with chopped and slightly boiled spinach ad libitum. Infection with Batrachochytrium dendrobatidis (Bd) was applied at each of the first five water changes as described below ("Experimental

infections").
Four days after the 5 th Bd inoculation, on day 19, we haphazardly selected eight tadpoles from each rearing container, placed them individually into 2-L plastic rearing boxes filled with 1.8 litres of RSW and randomly assigned them to Ranavirus (Rv) × temperature treatment combinations. We placed rearing boxes on a shelf system in randomised spatial blocks containing one replicate from each treatment combination. The tadpoles were housed in these individual boxes for a short time before the 24-hours Rv treatment (as described below; see "Experimental infections") and during the 6-days thermal treatment (see "Thermal treatments" below).

Experimental infections
We used a liquid culture of the Bd isolate IA042 (obtained from a dead Alytes obstetricans in 2004 in Spain), which belongs to the global pandemic lineage (Bd-GPL). We maintained the stock culture in mTGhL broth (8 g tryptone, 2 g gelatine hydrolysate, and 4 g lactose in 1000 ml distilled water) in 25 cm 2 cell culture flasks at 4°C and passaged every three months. One week before use in the experiment, we inoculated 100 ml mTGhL with 2 ml stock culture in 175 cm 2 flasks and incubated these cultures at 20°C for seven days. We estimated Bd-zoospore concentrations using a Bürker chamber at ×400 magnification and inoculated each rearing container assigned to the Bd and the Bd + Rv treatments with 10 ml to obtain final concentrations of approximately 2000 zoospores×ml -1 .
The FV3 strain was propagated in T75, and T175 cell-culture flasks on Epithelioma Papulosum Cyprini (EPC) cells (ATCC CRL-2872) in DMEM nutrient medium supplemented with 2% fetal bovine serum (FBS), without the use of antibiotics at 25°C with 5% CO2. Flasks were freeze-thaw harvested, and the medium was clarified from cell-debris by centrifugation for 10 min at 2500 × g to collect a sufficient amount of viral supernatant. Viral titre (TCID50) was determined on 96-well plates of EPC cells using the Spaerman-Karber formula and converted to plaque-forming units (pfu) using the equation: 1 TCID50 = 0.69 pfu. We exposed tadpoles to FV3 by applying one of two concentrations (i.e., Rv-low and Rv-high) during 24hour exposures immediately preceding the start of the thermal treatment. We placed tadpoles allocated to Rv treatments individually in 500 mL plastic cups containing 200 mL RSW and 6.12 × 10 3 pfu×ml -1 FV3 in the Rv-low group and 6.25 × 10 5 pfu×ml -1 FV3 in the Rv-high group.
We added the same quantity of sham extract (only the DMEM nutrient medium with 2% FBS without the virus) to the rest of the tadpoles.

Thermal treatments
We placed the 2-L rearing boxes in 80 × 60 × 12 cm trays filled with tap water to a depth of 8 cm (water level was lower than in rearing boxes by ca. 2 cm to avoid floating of the latter) and subsequently turning on submersible aquarium heaters (Tetra HT 200 in 28°C treatments and Tetra HT 300 in 30°C treatments) and water pumps (Tetra WP 300) placed opposite to each other on the longitudinal axis of trays. Thereby, water temperature increased gradually to the desired level in ca. two hours, allowing tadpoles to adjust to increasing temperatures. After heating up, the temperature did not change over time and varied only a little among/within trays (Table S1), as documented by automated temperature loggers (Onset HOBO Pendant Temperature/Light 8K) placed into one-third of the trays (i.e., 12 out of 36). Actual water temperatures in the tadpole rearing boxes were overall 21.4 ± 0.72, 28.16 ± 0.24 and 30.13 ± 0.35°C (mean ± SD) in the three temperature treatments, respectively.
During the six days of thermal treatment, we changed water twice with RSW pre-heated to the temperature of the respective thermal treatment group. We fed tadpoles with a lowered amount of spinach (one-third of the amount provided during the rearing period) to avoid water fouling and anoxia at high temperatures.

Molecular analyses
We assessed Bd infection intensity from dissected mouthparts in the case of Bd and from liver tissue in the case of Rv. The small body size of common toad tadpoles did not allow for precise separation of the liver, so in their case, we used all internal organs as a whole. We homogenised

Statistical analyses
We calculated the prevalence data with 95% confidence intervals presented in Table S2 and To analyse prevalence and pathogen load after thermal treatment, we included only those treatment groups that had been exposed to the given pathogen. We used generalised linear mixed-effects models (GLMM) to test the effects of treatments on pathogen prevalence. The model for Rv prevalence in agile frogs contained thermal treatment (22, 28 or 30°C), Rv concentration (low or high), co-exposure with Bd (yes or no) as fixed factors and all two-way interactions (there was not enough variance in the data to allow model fit for testing the threeway interaction). The model for Bd prevalence in common toads contained thermal treatment and a three-category factor that combined the information on the presence and concentration of Rv (no Rv, low Rv, high Rv) and the interaction of the two fixed factors. In both models, we entered family as a random factor. We assumed a binomial error distribution and used a logit link function. We fitted the models applying maximum likelihood estimation using the 'glmmTMB' function of the package 'glmmTMB' [75] and checked model-fit diagnostics using the 'DHARMa' package [76]. We did not run such analyses for Bd prevalence in agile frogs because zero prevalence in the majority of treatment groups would have led to very high estimation uncertainty (separation) and inability to test interactions.
We analysed the effect of treatments on infection intensity after temperature treatment Rv infection load in common toads, we used the same modelling approach, but we tested only the main effects because there was not enough variation in the data for testing interactions (i.e. prevalence was zero in 6 out of 12 treatment combinations, causing separation in binomial models). Low prevalence prohibited the analyses of infection intensity for Bd in both species.
To analyse the survival of agile frog tadpoles during thermal treatment, we ran a mixedeffects Cox's proportional hazards model (COXME; 'coxme' function of the 'coxme' package), entering family as a random effect [77]. We entered survival as an ordinal categorical dependent variable ranging 1-6 (each category representing the day of death during heat treatment, 1 being the first 24 hours); individuals that survived to the end of thermal treatment were treated as censored observations. We included thermal treatment, Rv concentration, Bd co-exposure and their two-and three-way interactions as predictors. Because mortality of common toads was negligible (1.1%; see Table S3), we did not analyse their survival.
We applied a backward stepwise model selection procedure to reduce noise in parameter estimates due to the inclusion of non-significant terms [78,79]. We obtained statistics for excluded terms by re-entering them to the final model. For these steps, we used type-3 analysisof-deviance tables ('Anova' function of the 'car' package). To perform pairwise comparisons, we calculated linear contrasts from the final models using the 'emmeans' function of the 'emmeans' package while applying the false discovery rate (FDR) correction method to adjust P values for multiple comparisons [80,81].

Prevalence and infection intensity of pathogens before the thermal treatment
Cross-contamination was not detected in either infection group except for one individual in the control group after 30°C thermal treatment, but this tadpole also exhibited very low Rv infection intensity (28 pfu×ml -1 ). Seven agile frog samples collected after thermal treatments were accidentally lost during sample procession, so that the sample sizes in the infection groups were Ncontrol = 59; NBd = 58; NRv-low = 58; NRv-high = 59; NBd+Rv-low = 60; NBd+Rv-high = 59; resulting in a sample size of 353 in agile frogs. We lost one common toad sample after the thermal treatment from the uninfected control group (Ncontrol = 59), resulting in a total sample size of 359 in common toads. Four samples preserved after experimental infections but before thermal treatments exhibited low DNA quality (one in the low concentration Rv treatment and one in the high concentration Rv treatment in the case of agile frogs; and one in the low concentration Rv treatment, and one in the Bd + high concentration Rv treatment in case of common toads), so we excluded these from the analysis of initial infection, resulting in a sample size of 118 individuals in both species.
All but two of the 120 agile frog tadpoles survived until the start of thermal treatment.
At that time, the initial prevalence of Bd in Bd-exposed individuals was extremely low: two out of 60 (Table S2). Initial Bd infection intensities were also very low, the Genomic Equivalent (GE) values being 0.36 and 2.43, respectively (Table S2). In contrast, the initial prevalence of Rv in Rv-exposed tadpoles was high, varying between 73.7 and 100% (Table S2), and was significantly higher after exposure to high Rv concentration compared to low Rv concentration (Fisher's Exact Test; P = 0.005), while the presence of Bd had no effect on Rv prevalence (Fisher's Exact Test; P = 0.45). Initial Rv infection intensities within the infected agile frog tadpoles were low to moderate (Table S2) and significantly higher in the Rv-high group than in the Rv-low group (LMM; χ 2 1 = 127.612, P < 0.001), while the presence of Bd had no effect (χ 2 1 = 0.003, P = 0.951). The interaction between the presence of Bd and the concentration of Rv was non-significant (χ 2 1 = 0.03, P = 0.861).
All but two of the 120 common toad tadpoles survived until the thermal treatment. The initial prevalence of Bd in Bd-exposed individuals varied between 5 and 42% (Table S2). The initial prevalence of Rv in Rv-exposed tadpoles varied between 20 and 100% (Table S2), and it was significantly higher after exposure to high Rv concentration compared to low Rv concentration (Fisher's Exact Test; P < 0.005), while the presence of Bd had no effect on Rv prevalence (Fisher's Exact Test; P = 0.45). Also, the initial infection intensity of Rv in Rv exposed toad tadpoles was higher in individuals that received high Rv concentrations (LMM; χ 2 1 = 130.721, P < 0.001) without an effect of the presence of Bd (χ 2 1 = 1.505, P = 0.219). The interaction between the presence of Bd and the concentration of Rv was non-significant (χ 2 1 = 0.643, P = 0.422).      S2. Survival probability of agile frog tadpoles (A) exposed to low Rv concentrations or (B) exposed to high Rv concentrations during the 6-days of thermal treatment visualised by Kaplan-Meier curves.