Behavioural fever reduces ranaviral infection in toads

Abstract

1. Host behaviour is known to influence disease dynamics. Additionally, hosts often change their behaviours in response to pathogen detection to resist and avoid disease. The capacity of wildlife populations to respond to pathogens using behavioural plasticity is critical for reducing the impacts of disease outbreaks. However, there is limited information regarding the ability of ectothermic vertebrates to resist diseases via behavioural plasticity.

2. Here, we experimentally examine the effect of host behaviour on ranaviral infections, which affect at least 175 species of ectothermic vertebrates. We placed metamorphic (temporal block 1) or adult (block 2) Southern toads (Anaxyrus terrestris) in thermal gradients, tested their temperature preferences before and after oral inoculation by measuring individual-level body temperature over time, and measured ranaviral loads of viral-exposed individuals.

3. We found significant individual-level variation in temperature preference and evidence for behavioural fever in both metamorph and adult A. terrestris during the first two days after exposure. Additionally, we found that individual-level change in temperature preference was negatively correlated with ranaviral load and a better predictor of load than average temperature preference or maximum temperature reached by an individual. In other words, an increase in baseline temperature preference was more important than simply reaching an absolute temperature.

4. These results suggest that behavioural fever is an effective mechanism for resisting ranaviral infections.

Graphical Abstract:

Introduction

Pathogens can impose strong selective pressures on their hosts, driving the evolution of host behaviours that reduce disease risk (Moore 2002, Han, Bradley et al. 2008). Additionally, many hosts cope with pathogens via phenotypic plasticity, modulation of trait values within the lifespan of the host (Agrawal 2001). For example, some ectothermic hosts can respond to pathogen exposure by exhibiting a behavioural fever, which is an acute increase in temperature preference (T_pref) (Kluger, Kozak et al. 1998). The slightly warmer environmental temperatures chosen by the host after pathogen exposure are presumably more favorable for the host than the pathogen (Reynolds, Casterlin et al. 1977, Burns, Ramos et al. 1996, Ouedraogo, Goettel et al. 2004). Fever is primarily considered a method for improving host immune function by stimulating immunological defenses (Evans, Repasky et al. 2015). However, fever has also been proposed as a mechanism for directly killing or slowing pathogen growth with heat (Richards-Zawacki 2009). Understanding how phenotypic plasticity, such as behavioural fever, can limit infection and lower disease risk is crucial for predicting how hosts might be affected by emerging diseases in a changing climate. This is especially true for ectothermic hosts, which cannot regulate body temperature independent of the environment and are especially sensitive to climatic abnormalities.

Studies demonstrating that ectothermic hosts use behavioural fever as an effective method of pathogen resistance, defined as a host strategy that limits or inhibits infection (Roy and Kirchner 2000), are limited (Adamo and Lovett 2011, de Roode and Lefèvre 2012). Here, we test for an effective behavioural fever response to a ranavirus in an amphibian. Ranavirus is an emerging, widespread viral disease caused by viruses in the genus Ranavirus. It has caused mass mortality events in amphibian hosts contributing to amphibian population declines in last few decades (Chinchar 2002, Brunner and Yarber 2018). Ranaviruses are also known to cause frequent infection in other ectothermic vertebrates, such as in fishes and reptiles (Chinchar 2002, Duffus, Waltzek et al. 2015). Some of these viruses are even capable of transmission across taxa, or interclass transmission (Brunner, Schock et al. 2004, Bandín and Dopazo 2011). Hosts likely encounter ranaviruses frequently because of their broad distributions and lack of host specificity (Miller, Gray et al. 2011), and thus ranaviruses might impose strong selective pressures on hosts.

In fact, mounting evidence suggests that many hosts are adapting immunological and behavioural strategies to combat ranaviral infections (Whittington, Philbey et al. 1994, Parris, Davis et al. 2004, Teacher, Garner et al. 2009, Duffus, Waltzek et al. 2015). For example, larval amphibians appear to prefer warmer temperatures while infected with ranaviruses (Parris, Davis et al. 2004). However, this study did not measure temperature preference before exposure and therefore could not differentiate between preexisting differences in T_pref and differences caused by exposure to ranavirus. Additionally, this study did not measure ranaviral load or prevalence, so there is no evidence that the apparent difference in T_pref is effective at resisting infection. Other studies have shown that experimentally warming ranavirus-infected hosts can reduce host mortality and viral loads (Rojas, Richards et al. 2005, Echaubard, Leduc et al. 2014) despite successful ranviral growth in culture at these higher temperatures (Chinchar 2002). In these studies, the warming was induced by the experimenters, not the hosts, and exceeded the magnitude and duration of most behavioural fever responses. Additionally, these studies only tested for temperature effects in larval amphibians. Hence, it is unclear if hosts more generally respond to ranaviral infections with behavioural fever and whether fever is effective at reducing ranaviral loads within hosts.

Here, we test whether metamoprhic and adult Southern toads, Anaxyrus terrestris, adjust their preferred temperature after infection with ranavirus and whether any change in temperature preference reduces ranaviral load within hosts. To accomplish these goals, we exposed A. terrestris to a ranaviral isolate in thermal gradients ranging in temperature from 12–33 °C (Sauer, Sperry et al. 2016) to assess individual T_pref before and after exposure to this virus. We also measured viral load within individual hosts to assess whether variation in T_pref affected virus replication.

Materials and Methods

Animal husbandry

Adult and metamorphic A. terrestris were collected from Hillsborough County, Florida from sites where ranavirus has not previously been detected. Individuals were maintained in individual containers (23.5 × 16.8 × 10 cm or 11.7 dia. X 13.5 cm, respectively) on top of folded paper towels soaked with artificial spring water (Cohen, Neimark et al. 1980). These toads were held in a laboratory maintained between 24–25°C with a 12h photoperiod for at least two weeks before the start of the experiment. The toads were fed mineral-dusted crickets ad libitum until the start of the experiment and their containers and paper towels were changed weekly. Only adult individuals were tested for ranavirus before exposure (see Ranaviral exposures and quantification for details) as metamorphic A. terrestris were too small to safely test using the oral swabbing methods described below.

Experimental design

Experimental design was based on a prior experiment testing for behavioural fever in amphibians to the chyrid fungus and thus we only briefly describe the methods here; see Sauer, Fuller et al. (2018) for additional details on the design. Experiments were conducted in Tampa, Florida with two temporal blocks, the first being conducted with metamorphs and the second with adults. In each experiment, we first measured baseline non-infected temperature preference (baseline T_pref) in thermal gradient apparatuses (see next paragraph for the frequency of these measurements). These apparatuses were previously shown to provide variation in temperature that is independent of humidity and which does not confound amphibian and prey temperature preferences (see Fig. S1 and supplemental methods, and Sauer et al. (2016) for thermal gradient construction and validation details). The apparatuses provided an ecologically relevant ambient temperature range of 12 to 33°C (Fritts, Grodsky et al. 2015, Sauer, Sperry et al. 2016, Sauer, Fuller et al. 2018). After measuring baseline T_pref, individuals were split into two groups of similar mean body masses and baseline T_pref (N = 53), 1) a sham-exposed control group (metamorph: n = 12; adult n = 14) and 2) a ranavirus-exposed treatment group (metamorph: n = 13; adult n = 14). While individual-level body mass was consistent across treatments within a block, body mass was much smaller in the metamorph (mean mass = 0.48 g ± 0.05 SE) than adult block (mean mass = 17.70 g ± 1.42 SE).

Throughout the experiment, temperature measurements were taken each day, every four hours, between 10:00h and 22:00h (two during the lighted part of the photoperiod and two during the dark) at the center of each animal’s dorsum (Rowley and Alford 2007) (Extech® High Temperature IR Thermometer; accuracy: ±2% < 932°F). The only exception is that these measurements were not taken during feeding periods, during which all individuals were fed 10 live crickets in containment bags to prevent crickets from moving freely within the thermal gradient (see Sauer, Sperry et al. (2016) and Figure S4 for more details). Temperature measurements were taken for four days before ranaviral or sham exposure. The mean T_pref of those four days for each individual toad is referred to hereafter as the baseline T_pref ($T_{pre{f}_{baseline}}$ in equations). Temperature measurements were taken daily for two weeks (metamorph block) or four days (adult block) after ranaviral or sham exposure. We shortened our time between exposure and sampling for the adult block after discovering that the metamorphs were uninfected two weeks after exposure to ranavirus.

Ranaviral exposures and quantification

We used a ranavirus strain isolated from infected wood frogs in Michigan with 99% similarity to Frog Virus 3 (GenBank Accession number: PRJNA504607). We cultured the virus on fathead minnow cells and Eagle’s minimum essential media containing 5% fetal bovine serum (MEM) to a titer of 3.6×10⁵ plaque-forming units (PFU) mL⁻¹. The virus was stored at −80°C until used in the experiments. Before exposure, virus stock was thawed and homogenized and then each individual was dosed orally with 77 µL of the virus (2.8×10⁴ PFU) or the same volume of a sham inoculum of MEM. The individuals from the metamorph block were euthanized two weeks after ranavirus exposure and dissected to remove spleen, kidney, and liver for quantification of ranaviral loads. The individuals in the adult block were sampled prior to exposure, to ensure they were not already infected, then sampled again four days after exposure by inserting and twirling a sterile swab in their mouth for thirty seconds and then freezing these swabs at −80°C (Allender, Mitchell et al. 2013). Non-destructive sampling via oral swab was preferred because it allowed us the option to sample individuals multiple times. However, we had to destructively sample metamorphic individuals because they were too small to swab orally.

Ranaviral DNA was extracted from each metamorph and adult sample using the Quiagen DNeasy Blood and Tissue protocol (Qiagen, Inc., Valencia, CA). To determine viral load, we used qPCR methods based on Forson and Storfer (2006), with a 250-bp fragment of the major capsid protein (MCP) gene used as a standard (gBlocks® plasmid-based standards; Integrated DNA Technologies, Skokie, IL). The qPCR reaction mixture contained 10 μL TaqMan® Universal Master Mix (Applied Biosystems, Foster City, CA), 0.6 μL of primer (Thermo Fisher Scientific, Waltham, MA), 0.625 μL of TaqMan® TAMRA florescent probe (Applied Biosystems, Foster City, CA), and 6.275 μL of nuclease-free water per well with 2.5 μL of either DNA sample, standard, or nuclease-free water added to each well. All samples and standards were run in duplicate and thus, an individual-level load represents the mean DNA copies of these two subsamples. The duplicates agreed in all but one instance where one of the two did not amplify. For that individual we ran a third subsample, which did amplify and used the mean of the two positive subsamples.

Data analysis

All statistics were conducted with R 3.4.0 (Team 2017). To test for repeatability in baseline T_pref within individuals and variation in baseline T_pref among individuals, we conducted a one-way repeated measures ANOVA (stats package, aov function). This analysis tested whether baseline T_prefs of individuals varied significantly across days (main effect of day) and whether they varied among individuals (among-individual variance). Using the ANOVA table from this analysis, we calculated repeatability or the variance explained by individual-level behaviour: ($r=\frac{M{S}_W}{M{S}_W+\left(\frac{M{S}_W-M{S}_A}{n}\right)}$; where MS_W is the within-group variance component and MS_A is the among-groups variance component (Lessells and Boag 1987).

To test for behavioural fever, we conducted multiple two-factor (treatment and time) repeated measures linear mixed effects models with individual treated as a random effect (lme4 package, lmer function) followed by log-likelihood ratio tests to determine significance (car package, Anova function). For each model, we paired baseline T_pref with each post-exposure day T_pref (time; one model for each post exposure day) and looked for an interaction between treatment (ranavirus-exposed or sham-exposed) and time on the z-score of ΔT_pref. Change in T_pref (ΔT_pref) is calculated as:

$$\mathrm{\Delta }{\text{T}}_{pref}=T_{pre{f}_{i,j}}-\overline{T_{pre{f}_{baseline,i,j}}}$$

where $T_{pre{f}_{i,j}}$ is the temperature preference for individual i at time point j, and $\overline{T_{pre{f}_{baseline,i,j}}}$ is the mean temperature preference of all baseline time points for individual i. Before statistical analysis, we transformed our data to standardized deviations away from the mean of the ΔT_pref of all individuals in a day. This ensured that temperature preferences were independent of any change in room temperature over time and that we were comparing T_prefs between treatments within rather than across days. A significant interaction between treatment and time would mean that the two treatments behaved differently after ranaviral or sham exposure.

To test whether individual-level T_pref affects ranaviral load (log-transformed ranaviral DNA copies divided by mass of the individual), we conducted multiple linear regressions (stats package, glm function; normal error distribution). Each analysis tested for an effect of one of three metrics for body temperature on ranaviral load during a specified time interval. Our three metrics for body temperature were: 1) ΔT_pref ($\mathrm{\Delta }{\text{T}}_{pref}=T_{pre{f}_{i,j}}-\overline{T_{pre{f}_{baseline,i,j}}}$), 2) mean T_pref (mean of the four body temperature measurements per day), and 3) maximum T_pref (maximum of the four body temperature measurements per day). We conducted an analysis for each of the first four days after exposure (one model per independent variable, per day; 12 total models) as well as pooled across all four days (one model per independent variable; 3 total models). Additionally, we tested for an effect of baseline T_pref on ranavial load to check for underlying differences in susceptibility that happened to be correlated with baseline T_pref.

Results

There was no mortality during either block of this experiment. Before ranaviral exposure, we were able to detect consistency in the baseline T_pref of individuals (repeatability: r > 0.98; Fig. S2) and variation in baseline temperature preferences among individuals (all blocks and treatments combined; main effect of individual on baseline T_pref: F_52,1537 = 7.60, P < 2.0×10⁻¹⁶). A. terestris metamorphs and adults did not have significantly different mean T_prefs (F_1,51 = 3.90, P = 0.54) and together their mean T_pref was 23.07 °C ± 0.66 SE. For both temporal blocks, we found evidence of behavioural fever after ranaviral exposure (effect of the interaction between treatment and time; metamorphs: day 1 χ² = 22.0, P < 0.001 and day 3 χ² = 5.70, P = 0.017; adults: day 2 χ² = 5.51, P = 0.019; Fig 1 & Table S1). One day after exposure, metamorphs exposed to ranavirus increased their preferred temperature by 3.52 °C ± 0.78 SE relative to controls (Fig. 1A). Two days after exposure, adults exposed to ranavirus significantly increased their preferred temperature by 1.43 °C ± 0.59 SE relative to controls (Fig. 1B). We did not test whether the magnitude of behavioural fever between metamorphs and adults were different, because the two life stages were not tested simultaneously.

Figure 1 |.

Difference between the daily mean for each treatment group and sham exposed (control) treatments ($\mathrm{\Delta }{\text{T}}_{pref,i,j}-\overline{\mathrm{\Delta }{\text{T}}_{pref,control,j}}$) in change in individual-level temperature preference ($\mathrm{\Delta }{\text{T}}_{pref}=T_{pre{f}_{i,j}}-\overline{T_{pre{f}_{baseline,i,j}}}$) through time for a) metamorphic and b) adult Anaxyrus terrestris toads. Where $T_{pre{f}_{i,j}}$ is the temperature preference for individual i at time point j and $\overline{\mathrm{\Delta }{\text{T}}_{pref,control,j}}$ is the mean temperature preference of all control animals from time point j. Asterisks denote time points where the exposed and sham-exposed groups differ significantly (P < 0.05) error bars represent ±1 SE.

For metamorphs, we were unable to detect ranavirus in sampled toads two weeks after exposure. For adults, we intentionally sampled after four days rather than 14 days in the hopes that we would detect ranavirus before it was cleared. We detected ranavirus in 92.9% of the adult toads. We found that ΔT_pref (t₁₁ = −4.89, P < 0.001; Fig. 2B), mean T_pref (t₁₁ = −3.11, P = 0.01; Fig. S3), and max T_pref (t₁₁ = −3.1, P = 0.004; Fig. S4) two days after exposure were all associated negatively with ranaviral load four days after exposure. Additionally, ΔT_pref on day one, three, and four post-exposure were also significant negative predictors of ranaviral load on day four (t₁₁ = −3.19, P = 0.009; t₁₁ = −2.42, P = 0.03; t₁₁ = −3.78, P = 0.003, respectively; Fig. 2). Finally, we found a positive effect of mean ΔT_pref (t₁₁ = −4.77, P < 0.001) and mean T_pref (t₁₁ = −2.42, P = 0.03) on ranaviral load but no effect of overall max T_pref (t₁₁ = −1.23, P = 0.25), and baseline T_pref (t₁₁ = 1.04, P = 0.32) on ranaviral load.

Figure 2|.

Relationship between change in individual-level temperature preference ($\mathrm{\Delta }{\text{T}}_{pref}=T_{pre{f}_{i,j}}-\overline{T_{pre{f}_{baseline,i}}}$) A) one, B) two, C) three, and D) four days after ranaviral exposure and ranaviral loads on adult Anaxyrus terrestris. Where $T_{pre{f}_{i,j}}$ is the temperature preference for individual i at time point j and $\overline{\mathrm{\Delta }{T}_{pre{f}_{baseline,i}}}$ is the mean temperature preference of individual i prior to ranaviral exposure. Frogs that exhibited the greatest increase in T_pref had the lowest ranaviral abundance (t₁₁ = −4.89, P < 0.001). The shaded gray area represents the 95% confidence band.

Discussion

We set out to determine whether A. terrestris responded to ranaviral exposure with behavioural fever and whether fever facilitated ranaviral resistance by limiting pathogen loads. By measuring the thermal preference of individuals in thermal gradients both before and after exposure, we found that A. terrestris individuals responded to ranaviral exposure with behavioural fever. We also demonstrated that individual-level change in T_pref during the first two days after exposure was the greatest predictor of ranaviral load in adult toads. Individuals that increased their T_pref the most had the lowest ranaviral loads. These results suggest that behavioural fever appears to be effective at resisting ranaviral infections.

We demonstrated that variation in baseline T_pref among individuals before ranaviral exposure was greater than the variation in baseline T_pref within individuals (Fig. S2). In other words, individuals showed consistency in their preferred temperature through time and individual toads exhibited different preferred temperatures. However, once exposed to ranavirus, we found that individuals moved from their baseline T_pref to warmer locations. For A. terrestris metamorphs, this behavioural fever response peaked one day after exposure while for adults, it peaked two days after exposure (Fig. 1). These differences might be partly because of large differences in body size. As both blocks were exposed to the same dose, the dose per g was higher in the metamorph than adult block. Additionally, given that there is a negative relationship between body size and metabolic rates and processes, immune and fever responses to ranavirus might have been triggered more quickly in the smaller-bodied metamorphs than adults (Garner, Rowcliffe et al. 2011, Rohr, Civitello et al. 2018).

We did not sample metamorphic A. terrestris for ranavirus until 14d after exposure, and by that time, no individuals were infected. Thus, we do not discuss the effects of thermoregulatory behavour on ranaviral loads in the metamorph block. For adult A. terrestris, which were sampled for ranavirus 4d after exposure, there was a significant negative correlation between ranaviral load four days post exposure and ΔT_pref, mean T_pref, and maximum T_pref during the second day post exposure (Fig. 2). We also found that ΔT_pref was negatively correlated with ranaviral loads across all four days post exposure. Thus, an increase in temperature, regardless of an individual’s baseline T_pref, helped to reduce ranaviral infection. We did not find the same overall effect of mean T_pref or maximum T_pref on ranaviral load; in fact, maximum T_pref was the worst predictor of ranaviral load of the three measurements (ΔT_pref, mean T_pref, and maximum T_pref). This result supports the hypothesis that the main purpose of fever is to increase the immune system’s efficiency by raising body temperature to promote both innate and adaptive immunity (Evans, Repasky et al. 2015, Rakus, Ronsmans et al. 2017, Boltana, Aguilar et al. 2018), not to maximize absolute preferred temperature within the range that a host can tolerate. Though fever or increased temperature can slow or stop pathogen growth directly for some host-pathogen systems (Anderson, Blanford et al. 2013, Sauer, Fuller et al. 2018), any increase in body temperature from baseline T_pref should benefit the host. Behavioural fever should be beneficial to the host even if the pathogen tends to grow better in warmer temperatures; as is the case for ranaviruses and many other viral and bacterial pathogens, to which fever is a common method of resistance (Evans, Repasky et al. 2015). Assuming the temperature increase falls within the bounds of the host’s thermal performance breadth (Evans, Repasky et al. 2015, Cohen, Venesky et al. 2017, Sauer, Fuller et al. 2018). For example, some ectothermic host species that are adapted to cooler climates cannot tolerate temperature increases when infected with ranavirus (Bayley, Hill et al. 2013, Brand, Hill et al. 2016) and other pathogens (Thomas and Blanford 2003, Cohen, Civitello et al. 2018).

While we do not have any load data for metamorphic A. terrestris, the ranaviral load results from the adults suggests that behavioural fever increased A. terrestris resistance to ranavirus. Interestingly, previous studies have shown that ranavirus grows in vitro and in vivo up to temperatures even higher than those reached here by feverish A. terrestris (Chinchar 2002, Ariel, Nicolajsen et al. 2009). Thus, the decrease in ranaviral load associated with behavioural fever might be the result of a host-mediated mechanism and not a direct effect of increased temperature. Behavioural fever might be one way that ectothermic hosts resist ranaviral infections, but it is unclear how widespread that response is given the paucity of studies examining behavioural fever as a mechanism for resistance to ranaviruses (Parris, Davis et al. 2004), especially if the fairly heat-tolerant A. terrestris exhibits thermal regulatory behaviour that is not representative of more cold-adapted species (Sauer, Fuller et al. 2018). Nevertheless, assuming that behavioural fever is in part used as a method of improving immune efficiency and not simply as a method of heat-killing the pathogen, this strategy could be employed more broadly by ectothermic hosts than a strategy of simply reaching an absolute temperature to heat-kill ranavirus. That is, ectothermic hosts with critical thermal maxima lower than that of the pathogen should still benefit from behavioural fever as method to improve immunological resistance, assuming the appropriate temperatures are available and the host does not have a very small thermal safety margin. However, there are likely costs of behavioural fever that might be balanced against its benefits (Sheldon and Verhulst 1996, Lochmiller and Deerenberg 2000). For example, behavioural fever can increase predation risk in ectothermic hosts if it requires hosts to increase activity or leave refugia (Bundey, Raymond et al. 2003, Parris, Davis et al. 2004, Han, Bradley et al. 2008, Otti, Gantenbein-Ritter et al. 2012, Todd, Jodrey et al. 2016).

In summary, both metamorphic and adult A. terrestris responded to ranaviral exposure with behavioural fever. Additionally, we found that adult A. terrestris were successful at reducing their viral loads by increasing their body temperature after exposure. Hence, for A. terrestris, behavioural fever is a successful method of ranaviral resistance, a host strategy for limiting or inhibiting infection (Roy and Kirchner 2000). Our results support the idea that behavioural fever is primarily used as a method of improving immunological resistance rather than simply damaging pathogens with heat (Evans, Repasky et al. 2015). Thus, behavioural fever could be an effective mechanism of resistance used by ectothermic hosts that are less heat tolerant than the infecting pathogen. More experimental work is needed to determine how widespread the use of behavioural fever as a method of pathogen resistance is in ectothermic vertebrates. Understanding how ectothermic hosts rapidly respond to ranaviruses and other emerging pathogens and how changes to environmental temperature affect these host-parasite interactions is crucial given that ectotherms are increasingly experiencing population declines and die-offs due, in part, to increases in emerging diseases that often appear to be exasperated by abnormal temperatures (Harvell, Altizer et al. 2009, Raffel, Romansic et al. 2013, Grayfer, Edholm et al. 2015).