Body mass and hibernation microclimate may predict bat susceptibility to white‐nose syndrome

Abstract In multihost disease systems, differences in mortality between species may reflect variation in host physiology, morphology, and behavior. In systems where the pathogen can persist in the environment, microclimate conditions, and the adaptation of the host to these conditions, may also impact mortality. White‐nose syndrome (WNS) is an emerging disease of hibernating bats caused by an environmentally persistent fungus, Pseudogymnoascus destructans. We assessed the effects of body mass, torpid metabolic rate, evaporative water loss, and hibernaculum temperature and water vapor deficit on predicted overwinter survival of bats infected by P. destructans. We used a hibernation energetics model in an individual‐based model framework to predict the probability of survival of nine bat species at eight sampling sites across North America. The model predicts time until fat exhaustion as a function of species‐specific host characteristics, hibernaculum microclimate, and fungal growth. We fit a linear model to determine relationships with each variable and predicted survival and semipartial correlation coefficients to determine the major drivers in variation in bat survival. We found host body mass and hibernaculum water vapor deficit explained over half of the variation in survival with WNS across species. As previous work on the interplay between host and pathogen physiology and the environment has focused on species with narrow microclimate preferences, our view on this relationship is limited. Our results highlight some key predictors of interspecific survival among western bat species and provide a framework to assess impacts of WNS as the fungus continues to spread into western North America.


| INTRODUC TI ON
Disease is energetically costly. In particular, the interaction between the energetic demands of an infection and the energetic demands of the environment experienced by the host can drive variation in mortality (Bonneaud et al., 2003). When a pathogen is also impacted by the environment and can persist outside the host for extended periods of time, understanding the interplay of all sides of the "disease triangle" (host, pathogen, and environment) becomes critical . Furthermore, in multihost disease systems, variation in mortality often reflects host physiological adaptations and the susceptibility of the host to disease (Dobson, 2004;Langwig et al., 2016).
Environments that require low energetic input by the host in the absence of disease may become unsuitable if conditions facilitate extensive pathogen growth or transmission (Nowakowski et al., 2016). For example, Pseudogymnoascus destructans, the pathogenic fungus that causes white-nose syndrome (WNS) in hibernating bat species, thrives in cool, moist conditions (Langwig et al., 2012;Verant et al., 2012). Hibernacula that align with these conditions are the least energetically demanding environments for healthy bats (Geiser, 2004). Throughout hibernation, bats alternate between periods of reduced metabolic rate and body temperature (i.e., torpor) and periods of euthermia. Although the concept of "optimal" environments for hibernation is debated, it is generally accepted that healthy bats can reduce their fat loss over winter by selecting microclimates that reduce arousal frequency (Boyles, Boyles et al., 2017;Boyles, Dunbar et al., 2007;Boyles et al., 2020;Humphries et al., 2003;Nowack et al., 2019). As bats drop body temperature to near-ambient temperature during torpor, minimum torpid metabolic rate is used at colder ambient temperatures until hibernacula reach a species' minimum defended temperature (Geiser, 1988(Geiser, , 2004. The need to arouse to euthermic body temperatures during hibernation has been attributed to many factors, including the need to replenish water lost through evaporation (Ben-Hamo et al., 2013). Humid environments decrease evaporative water loss, increase torpor bout duration, and decrease energy expenditure (McGuire et al., 2017;Thomas & Cloutier, 1992). However, the populations and species suffering the highest WNS mortality rates hibernate in humid environments (Langwig et al., 2012) that allow for extensive fungal growth.
In sum, the environments that once provided low energetic demands for the host could also lead to high fungal burdens and the demise of bat populations as the fungus moves across species' ranges.
Many hibernating bat species in North America are often found roosting in drier conditions than the populations that have been well-studied in eastern North America (Baerwald, 2017;Gillies et al., 2014;Grieneisen, 2011;Jagnow, 1998;Kuenzi et al., 1999;Neubaum et al., 2006). Inter-and intraspecific studies have shown that bats that roost in drier conditions have mechanisms to decrease rates of evaporative water loss when exposed to extreme conditions (Boratyński et al., 2015;Gearhart et al., 2019;. We surmise that these species may be less susceptible to the impacts of WNS, as drier hibernacula could limit fungal growth (Marroquin et al., 2017), and these species may have the mechanisms to reduce water loss associated with these conditions. Variation in hibernaculum microclimates facilitates a natural experiment to test the effects of environmental conditions, host physiology, and pathogen growth on mortality. Furthermore, we can assess if previously suitable conditions for minimum energy consumption for the host become unsuitable in the presence of a pathogen due to increased fungal growth and thus higher susceptibility to disease.
Multiple studies have investigated factors that are important to WNS survival, including body mass (Haase et al., 2019), prehibernation fat stores (Cheng et al., 2019), physiological mechanisms for resistance to infection (Auteri & Knowles, 2020;Hoyt et al., 2015), and hibernaculum microclimate (Langwig et al., 2012;Verant et al., 2014). However, these studies have related these covariates independently and often within a single species. When discussing WNS survival across species, studies have focused on species that hibernate in a narrow range of microclimate conditions, which can miss critical aspects of survival. For example, Langwig et al. (2012) presented variation in survival with WNS across species as a function of hibernaculum relative humidity and temperature. However, the minimum measured relative humidity was 90%, whereas hibernaculum environments experienced by North American bat species susceptible to WNS range from 20%-100%.
As WNS has the greatest impacts in eastern North America, most research on the interplay between host and pathogen physiology and hibernaculum microclimate has focused on eastern bat species. These species roost in stable microclimates, and thus, we have a limited view on how interspecific variation in bat physiology may influence disease dynamics of WNS. Here, we combine field data with mathematical modeling to test the relative importance of bat morphology, bat physiology, and hibernation microclimate in explaining variation in survival from WNS across nine species roosting in a range of hibernaculum conditions. We applied the Haase et al. (2019) hibernation energetics model in an individual-based model framework to assess survival as a function of western bat species and provide a framework to assess impacts of WNS as the fungus continues to spread into western North America.

K E Y W O R D S
bats, disease, evaporative water loss, hibernation energetics, microclimate, Pseudogymnoascus destructans, white-nose syndrome species-specific bat characteristics, hibernaculum microclimate, and fungal growth rate at eight sampling sites across North America. Hayman et al. (2016) and Haase et al. (2019) incorporated the effect of fungal growth on fat consumption and evaluated the sensitivity of the model to bat morphometric parameters. Our model incorporates the main physiological variables believed to drive variation in WNS survival across species and assess the relative importance of these physiological variables.

| ME THODS
We used a combination of field data and energetic modeling to We predicted the probability of survival with WNS for each species in an individual-based model framework. Microclimate data (temperature, water vapor deficit), bat morphometrics (body mass), and bat physiological characteristics (torpid metabolic rate, evaporative water loss) were collected from the field from eight sites (Table 1, Figure 1; Supporting Information). We predicted winter duration at each sampling site given estimates from Hranac et al. (unpublished data, CRH, CGH, NWF, JCM, CLL, LPM, SHO,) and assumed that all individuals at that site would enter into and emerge from hibernation associated with those predictions to remove the effect of intraspecific variation in hibernation duration. Finally, we estimated hourly fungal growth rate as a function of hibernaculum temperature and water vapor deficit given equations described by Hayman et al. (2016).
We ran the hibernation energetics model for 100 independent bats within 100 variations of hibernaculum microclimate for 10,000 bat-runs of fat expenditure per species. We characterized a hibernaculum microclimate environment by randomly selecting temperature and relative humidity values from normal distributions fitted to the mean and standard deviation from our measured microclimate data from the sampling location (where bats were found roosting) within the site associated with that species (Table 1). The randomly sampled temperature and relative humidity combination represented the "winter" condition for each individual bat-run. We used relative humidity, rather than water vapor deficit to characterize the hibernaculum microclimate in order to prevent impossible water vapor deficit and temperature combinations. We then converted to water vapor deficit (difference in water vapor pressure TA B L E 1 Information for each hibernaculum site, including predicted winter duration (with 95% confidence intervals; CRH unpublished data) and microclimate conditions Note: Temperature (T a ), relative humidity (RH), and water vapor deficit (dWVP) mean and standard deviation (in brackets) are reported for all data loggers within the hibernaculum and at the specific location where bats of any species were observed.
[kPa] between air and saturation at measured air temperature and relative humidity) once the microclimate was characterized using equations from Campbell and Norman (1998 We fit a linear model to the probability of survival with body mass, mass-specific torpid metabolic rate, mass-specific evaporative water loss, hibernaculum temperature, and hibernaculum water vapor deficit as predictors. Because mass-specific evaporative water loss and hibernaculum water vapor deficit are mechanistically linked, we fit two models, one with mass-specific evaporative water loss and one with hibernaculum water vapor deficit, and selected the model with the highest adjusted R 2 value. We then calculated the partial correlation coefficient (PCC; Baba et al., 2004)  covariates. Therefore, we determined which variables (hibernaculum temperature and water vapor deficit, mass-specific minimum torpid metabolic rate, mass-specific evaporative water loss, and body mass) explained the most variation in survival with WNS and fulfilled our predictions.
We were also interested in how these variables differed among species and whether there was any significant variation in survival during hibernation with WNS. We first performed a multiple comparisons Kruskal-Wallis rank-sum test of differences (Daniel, 1990) to test for pairwise differences between species. We also calculated Tukey's honest significant differences (Tukey, 1949) to determine pairwise differences in body mass, mass-specific torpid metabolic rate, and mass-specific evaporative water loss between species. We assume that if the species share the same predictor traits, we predict there will be no significant differences in survival.

| RE SULTS
There were species differences in survival during hibernation with WNS predicted at our sampling sites (X 2 = 529.6, df = 8, pvalue < .001, critical difference = 117.53; Table S1). There were no predicted to survive in some microclimate scenarios. P. subflavus had a wide range of survival across microclimate scenarios, but on average, did not have enough fat to survive hibernation. Between species, there were no patterns of pairwise differences in body mass, mass-specific torpid metabolic rate, mass-specific evaporative water loss, or hibernaculum microclimate, as we predicted (Table S1). In other words, similar pairwise trait patterns did not appear among the bats predicted to survive or succumb to WNS.
When determining which morphometric, physiological, or microclimate variables were strong predictors, we found hibernaculum temperature and water vapor deficit, mass-specific minimum F I G U R E 2 Difference between predicted days until fat exhaustion and predicted winter duration (with standard deviation error bars). There were significant species differences in survival (Table S1). All small Myotis species, including M. ciliolabrum, M. evotis, M. lucifugus, M. thysanodes, and M. volans, did not exhibit differences in predicted survival at our measured sampling sites. Additionally, M. velifer did not differ from C. townsendii nor E. fuscus in predicted survival, but these three did differ from the rest of the Myotis species torpid metabolic rate, mass-specific evaporative water loss, and body mass explained 58% of the variation in winter survival with WNS (Table 3). Hibernaculum water vapor deficit explained more variation (9.0%) than mass-specific evaporative water loss (1.5%) and thus the model with water vapor deficit explained more overall variation in survival (adjusted R 2 = 0.89) compared to mass-specific evaporative water loss (adjusted R 2 = 0.75). Body mass (p-value = 0.004) and hibernaculum water vapor deficit (pvalue = 0.043) were the only significant predictors of the probability of survival through hibernation with WNS (Table 3). When assessing each covariate alone, water vapor deficit (8.27%) and body mass (47.0%) explained over half of the variation in survival, while the other covariates explained less than 4% (Table 3).
According to the estimated parameter values, high humidity and smaller body mass resulted in decreased survival with WNS.

| D ISCUSS I ON
Unraveling the complex interactions between the host, pathogen, and environment (the "disease triangle") helps us understand the influence of external pathogens on hibernation physiology and winter survival. In hibernating bats of North America, WNS is a disease where the environment clearly influences both the host and the pathogen. We used the natural physiological and environmental host diversity of western bats in order to address the importance of each of these factors on bat overwintering mortality. Our application of hibernation modeling to empirical field data allows for the estimation of survival from WNS across multiple species that use different hibernation conditions. We show that hibernaculum water vapor deficit and host body mass are likely strong predictors of susceptibility to WNS. Our results indicate that body mass is the strongest predictor for survival with WNS and we note pairwise differences between species with differing body mass. We found that species roosting in drier environments are predicted to have higher survival than those that roost in hibernacula at or near saturation.
Saturated conditions, though less energetically costly for hibernation by healthy individuals, lead to greater fungal growth rates and thus higher mortality (Haase et al., 2019;Langwig et al., 2012), which suggests a trade-off between water conservation and fat conservation in bats infected with WNS.
Taken together, our results suggest that the importance of body mass for survival from previous work may be related to the amount of surface area available for evaporative water loss. For example, though E. fuscus is a much larger bat than M. fuscus, it does not necessarily mean they have proportionally more fat than Myotis species.
Instead, Myotis have greater rates of evaporative water loss per unit of body mass than larger bats due to the greater surface area per unit volume, which is an important component to arousal rates in hibernating bats (Ben-Hamo et al., 2013;McGuire et al., 2017). The fact that small species which roost in dry hibernacula (i.e., Myotis ciliolabrum) can have low survival (Figure 3) indicates that greater body mass, and therefore less surface area relative to body size, is still imperative for survival.
The relationship between predicted survival and water vapor deficit within the hibernaculum demonstrates how microclimate can influence WNS susceptibility. Our results are consistent with results from empirical studies that find an association between population declines and humid environments (Langwig et al., 2012). Suboptimal environments for fungal growth are potentially unsuitable environments for bat hibernation given each species physiology and behavior. However, we found that if bats have the predisposition, and thus adaptations, for successfully hibernating in unsaturated environments, then these individuals may survive hibernation with WNS.
Therein lies a trade-off between fat conservation and water conservation-if bats are adapted to roosting in saturated conditions, then drier environments would still result in high mortality due to increased evaporative water loss, with or without the impact of the fungus. Therefore, we cannot explicitly state that suboptimal environments for the fungus would result in higher rates of survival, but rather need to take a holistic approach to understanding the specific hibernation strategies of each bat species, recognizing that intraspecific differences can exist (e.g., . Additionally, our models only predict survival based on sufficient fat stores, but do not consider fat stores posthibernation, when remaining stores are necessary for flight and reproduction, nor the energy required for the costly inflammatory response to fight off infection (Fuller et al., 2020). E. fuscus populations from more arid regions had lower rates of evaporative water loss than those from more mesic regions when measured in dry (0% relative humidity) conditions (Klüg- . Additionally, E. fuscus maintains more variable torpid metabolic rates in northern latitudes compared to southern latitude conspecifics at temperature below the minimum defended temperature, suggesting a continuum in thermoregulatory responses and minimum defended temperatures to hibernaculum environments (Dunbar & Brigham, 2010). However, these species vary in microclimate selection, while other species, such as M. lucifugus, are noticeably more selective in their hibernacula (Brack, 2007). M. lucifugus tends to roost in hibernacula that are at or very close to saturation with stable temperatures (Grieneisen, 2011).
When determining the conditions that may favor survival, it is important to consider how species and populations are adapted. It may be generally true that low humidity sites are detrimental to the fungus, and therefore, some species may experience high survival in low humidity, but that is only true for those species that are adapted to those conditions. For instance, we predicted survival for five species that were found in dry conditions (Figure 3b), but only three of them are predicted to have high survival (> 75% survival, torpid metabolic rate also had no impact. Further work to understand variation in physiology and behavior is warranted, specifically in the adaptations that may or may not influence bat survival through hibernation with WNS (e.g., low evaporative water loss).
We modeled nine bat species at eight overwintering sites and the model framework could be extended to different species and hibernacula at different latitudes and longitudes as data become available. Until then, extrapolation to sites with parameters that fall outside those of our study sites should be done with caution.
Future improvements to our model could include interspecific differences in hibernation behavior across species; for example, one could incorporate clustering behavior and its effect on heat loss and evaporative water loss (Boratyński et al., 2015;Boyles et al., 2008). However, there is evidence that clustering can cause additional or partial arousals in individuals when a neighbor arouses (Hayman et al., 2017), suggesting that clustering may increase fat expenditure in colonies with WNS (Langwig et al., 2012). Solitary roosting has also been observed in many western bat species, where individuals roost alone in crevices and talus slopes (Neubaum et al., 2006