Heart rate reveals torpor at high body temperatures in lowland tropical free-tailed bats

Reduction in metabolic rate and body temperature is a common strategy for small endotherms to save energy. The daily reduction in metabolic rate and heterothermy, or torpor, is particularly pronounced in regions with a large variation in daily ambient temperature. This applies most strongly in temperate bat species (order Chiroptera), but it is less clear how tropical bats save energy if ambient temperatures remain high. However, many subtropical and tropical species use some daily heterothermy on cool days. We recorded the heart rate and the body temperature of free-ranging Pallas' mastiff bats (Molossus molossus) in Gamboa, Panamá, and showed that these individuals have low field metabolic rates across a wide range of body temperatures that conform to high ambient temperature. Importantly, low metabolic rates in controlled respirometry trials were best predicted by heart rate, and not body temperature. Molossus molossus enter torpor-like states characterized by low metabolic rate and heart rates at body temperatures of 32°C, and thermoconform across a range of temperatures. Flexible metabolic strategies may be far more common in tropical endotherms than currently known.

Reduction in metabolic rate and body temperature is a common strategy for small endotherms to save energy. The daily reduction in metabolic rate and heterothermy, or torpor, is particularly pronounced in regions with a large variation in daily ambient temperature. This applies most strongly in temperate bat species (order Chiroptera), but it is less clear how tropical bats save energy if ambient temperatures remain high. However, many subtropical and tropical species use some daily heterothermy on cool days. We recorded the heart rate and the body temperature of free-ranging Pallas' mastiff bats (Molossus molossus) in Gamboa, Panamá, and showed that these individuals have low field metabolic rates across a wide range of body temperatures that conform to high ambient temperature. Importantly, low metabolic rates in controlled respirometry trials were best predicted by heart rate, and not body temperature. Molossus molossus enter torpor-like states characterized by low metabolic rate and heart rates at body temperatures of 32°C, and thermoconform across a range of shortages and the use of low energy states (torpor) may be particularly advantageous in these scenarios. Molossus molossus spend 23 h in their roost with f H as low as 40-50 bpm and resting rates of 156 ± 71 bpm at night [43]. These resting rates are 50% lower than expected and indicate that the bats may employ torpor at high T a , and presumed high T b .
We hypothesized that M. molossus would thermoconform at high T a , with similar low MRs across a wide range of T a . Subsequently, low f H will be used by M. molossus across a wide range of T a . We established relationships among f H , T b , T a and energy consumption in captive individuals, and then applied these relationships to free-ranging bats roosting in their natural social groups.

Animal capture and marking
We captured 11 M. molossus (10.5 ± 0.7 g) with mist nets as they emerged from their roosts in holes and crevices underneath houses in Gamboa, Panamá. We marked each bat with a subcutaneous temperaturesensitive PIT-tag (BioThermo13, Biomark Inc., Boise ID, USA) [26,41] and fitted it with an external heart rate transmitter (ca 0.8 g; SP2000 HR Sparrow Systems, Fisher, IL USA) that emits a continuous long-wave signal, interrupted by cardiac muscle potentials [34,45]. To attach the heart rate transmitter, we trimmed the fur in the middle of the back below the shoulder blades, applied a topical analgesic (Xylocaine gel, Astra Zeneca, Wedel Germany), and disinfected the skin and electrodes with 70% EtOH [43]. We inserted the transmitter's two disinfected gold electrodes ca 3 mm dorsally through a puncture made with a 23 ga sterile needle. The transmitters were mounted on thin, flexible cloth and then glued over the electrode insertion points using a silicone-based skin adhesive (Sauer Hautkleber, Manfred Sauer, Germany). The electrodes were flexible and did not appear to disturb the animals. Transmitters represented 7.0 ± 0.7% (s.d.) of body mass [46]. We removed transmitters immediately after respirometry experiments or after 2-3 days of deployment on free-ranging bats. We saw no signs of infection at the lead insertion sites, and f H and T b records did not indicate elevation in the MR consistent with an immune challenge. Bats either maintained body mass or gained up to 1 g of mass (0.6 ± 0.3 g), revealing no measurable negative impact of the short-term deployment of the additional mass of the transmitter.

Laboratory measurements of heart rate, body temperature and metabolic rate
We used an open-flow, push-through respirometry system [47] to measure rates of oxygen consumption (VO 2 ), carbon dioxide production (VCO 2 ) and T b of six bats for 10-20 h continuously [48]. Owing to logistical constraints, these were not the same bats that were tracked in the field. We dried incurrent air (greater than 75% relative humidity, approx. 26°C) with Drierite (WH Hammond Drierite Co, Ltd, Xenia, OH, USA) and pumped it through a mass flow controller (FB8, Sable Systems International, Las Vegas, NV, USA) into a 1.97 l respirometry chamber fitted with a thermocouple within a 20 l insulated cooler that was dark and temperature-controlled (PELT5, Sable Systems). An additional empty chamber served as a reference (baseline) to the animal chamber. The flow rate was 300 ml min −1 and relative humidity and vapour production were measured with an RH-300 (Sable Systems). After drying the excurrent air again with Drierite, we measured CO 2 concentration (FOXBOX, Sable Systems), and after scrubbing the air of CO 2 with Ascarite (Thomas Scientific, Swedesboro NJ, USA), we determined O 2 concentrations (FOXBOX, Sable Systems). Chamber temperature, CO 2 , O 2 and relative humidity were recorded at 1 Hz directly with Expedata via the UI-2 data acquisition interface (Sable Systems). Bats had the option to roost on vertical or horizontal mesh platforms above a layer of mineral oil used to trap excrement. We kept the bats at 28 and 32°C for at least two hours at a time, which is equal to or lower than the lower critical threshold of the thermoneutral zone (TNZ) for this species [49]. We measured bat T b via PIT tag (BioThermo13, Biomark, Inc.) every minute using an antenna (HPR Plus, Biomark) in the insulated chamber [41], and we recorded f H as a sound file (see below). PIT tag calibrations against a thermometer traceable to the U.S. National Bureau of Standards showed a mean measurement error of 0.21 ± 0.2°C. f H was averaged over the 1 min preceding each T b measurement. This gave five T b and f H measures for each measurement ofVO 2 andVCO 2 . We used carbon dioxide (VCO 2 ) production to estimate metabolic rates using Equation 10.5 from Lighton [50]:VCO 2 = (FeCO 2 -FiCO 2 ) * FR/(1-FeCO 2 * (1-(1-RER))), where FiCO 2 is the incurrent CO 2 content, FeCO 2 is the excurrent CO 2 content, FR is the flow rate and RER is the respiratory exchange ratio (VCO 2 :VO 2 ), which we calculated to be 0.8 from empirical measurements of CO 2 and O 2 . We convertedVCO 2 (ml min −1 ) to metabolic rates [51,52] that would be comparable to field rates using the standard conversion of 25.0 J ml −1 CO 2 . After conclusion of the experiment, the heart on December 28, 2017 http://rsos.royalsocietypublishing.org/ Downloaded from rate transmitters were removed, bats were offered water via a transfer pipette and placed in the entrance to their roost.

Field heart rate and body temperature telemetry
We recorded f H and T b of five bats that were not part of the respirometry experiments during 1-3 days and nights in their natural roosts. We used telemetry receivers (AR8000, AOR Ltd) connected to 3-element Yagi antennae (Sparrow Systems) to detect the signal of the heart rate transmitter. This was then recorded via the mini-dv output to a wave file on a digital recorder (Tascam DR-05). The maximum recording distance for T b was 10 cm, and 100 m for f H transmitters; therefore, no flying data were recorded. Three bats (1646, 1721, 1732) roosted in the walls of a wooden house and had both f H and T b sampled every 10 min. The heart rate was recorded continuously for two bats 2253 and 2289, and T b was recorded once per minute. Bat 2253 roosted in the roof of a wooden structure under a metal roof and bat 2289 in the brick ground floor of a house. To temporally synchronize T b and f H , a smoothed f H of the previous 60 s was matched to each T b . Ambient temperature data were recorded at 15 min intervals by the Autoridad del Canal de Panamá (ACP) for Gamboa and provided by the Smithsonian Tropical Research Institute's Physical Monitoring Program. The daytime mean ambient temperature was 25.87 ± 1.21°C (mean daytime minimum to mean maximum: 23.38-28.24), and the mean nightly ambient temperature was 23.74 ± 0.50°C (mean nightly minimum to mean maximum: 22.74-24.78°C). The ACP monitoring station was located along the banks of the Panama Canal 400-800 m from the observation roosts. While these temperatures do not directly measure the more insulated and stable microhabitats the bats experience in their roosts, they show the potential T a that bats experience across the day.

Heart rate scoring
The heart rate from radio transmitters has been visually scored at sampling intervals of 5-10 min [33,34,45,53]. However, complete sampling can show novel energy-saving strategies like those in tentmaking bats that depress f H several times per hour [26], a pattern that would not have been detected by sampling every five to ten minutes. We therefore fully sampled the recorded data using an automated approach in R 3.3.2 [54] to identify and count heartbeats [26]. We applied a finite impulse response filter in seewave [55] with a window length of 1500-2000 samples to select the carrier frequency of the transmitter. We counted individual heartbeats by applying a timer function in seewave that ran over nonoverlapping windows of 500 samples. This created a resolution of 88-96 sampling windows per second. We then applied a kernel density filter in KernSmooth [56] to further eliminate noise that was outside of the 90% quantile. This approach is conservative and may have eliminated some heart rate outliers, but the autocorrelated nature of heart rate allowed us to filter out errors probably induced by static or other interference in the recordings. Automated samples were visually inspected periodically to validate the filtering method, particularly when there was high variation.
We could not apply all the automated methods to the respirometry data due to large amounts of interference from the PIT tag reader within the respirometry chamber. Here, we hand-scored f H by counting all heartbeats within the first 10 s of every minute. Sampling at regular intervals may underestimate short-term changes in f H [26], but the consistent low f H of the bats throughout our captive and field experiments did not warrant finer-scale sampling. This was the same time resolution as T b measures and provided five f H measures per unit of MR sampling. We then averaged these measures to create a single value for each 5 min respirometry sample.

Analysis
We tested the fit of heart rate (f H ), body temperature (T b ), and the difference between body temperature and ambient temperature (T diff = T b − T a ) on the metabolic rate (MR) using generalized linear mixedeffects models (GLMMs) with individual as a random intercept in lme4 after inspecting the data for normality and equal variance. We used both the field metabolic rate (FMR, kJ h −1 ) and massspecific metabolic rates (W g −1 ) to facilitate broader comparisons with published data, particularly those collected in field experiments. We compared the minimum metabolic rates of M. molossus, identified as the lowest 10% quantile of the MR for each individual, to minimum torpor metabolic rates for species found across a variety of temperature regimes (temperate, subtropical and tropical) that undergo daily torpor from three speciose mammalian orders (Chiroptera, Dasyuromorphia and Rodentia; data from Ruf and Geiser [13]). A nested model approach revealed that all three of our predictors (f H ,  contributed significantly to explaining the variation in both measures of energy consumption (FMR and mass-specific MR). We therefore took a model selection approach to evaluate which factors were the most efficient at predicting energy consumption in respirometry. We calculated the Akaike information criterion corrected for small sample sizes (AICc) for each model, as well as the confidence intervals for each model parameter in lme4, and both the marginal (R 2 m , fixed effects alone) and conditional (R 2 c , full model) R 2 values in MuMIn [57] using the approach outlined by Nakagawa & Schielzeth [58]. All analyses were performed in R 3.3.2 [54], and we present means ± s.d. for all variables unless otherwise noted.

Results
3.1. Respirometry, captive heart rate, body temperature and energy consumption All bats reduced their MR and T b in the respirometry chamber. We recorded a large range of MR (0.075-1.245 kJ h −1 or 0.0021-0.325 W g −1 ), f H (59-999 bpm) and T b (27.9-37.6°C) across the 10-20 h of continuous sampling. In general, T b followed T a when the MR was reduced across the small range of T a that we measured, and f H was not dependent on T b (figure 1 and electronic supplementary material, figure S1). After physiological arousal generated by observers tapping on chamber walls, f H dropped rapidly into a low metabolic state, and T b eventually followed at a slower rate. Bats showed a low MR of 0.1218 ± 0.021 (mean ± s.d.) kJ h −1 (0.00314 ± 0.00051 W g −1 ). When we further examined the stable MR at our two T a , we found that metabolic rates were lower at 28°C (0.131 ± 0.014 s.e. kJ h −1 ) than at 32°C (0.160 ± 0.001 s.e. kJ h −1 ; χ 2 1 = 445.77 , p < 0.001; figure 2). These lower MR at 28°C are consistent with lower f H at 28°C (82.4 ± 6.3 bpm) than at 32°C (115.8 ± 1.0 bpm; χ 2 1 = 1059.2, p < 0.001; figure 2). Molossus molossus generally used low mass-specific MR in the respirometry chamber. Bats showed a low minimum MR of 0.1138 ± 0.0167 kJ h −1 (0.00294 ± 0.00034 W g −1 ) and this did not differ from the minimum metabolic rates for other mammalian orders that use daily heterothermy (χ 2 3 = 5.525, p = 0.154; figure 3). Molossus molossus enter low metabolic states at substantially higher T b than these other mammals (χ 2 3 = 26.68, p < 0.001; figure 3). All three predictor variables (f H , T b , T diff ) explained substantial variation in the MR, but f H alone was the best model regardless of any temperature interaction (figure 4; electronic supplementary material, table S1). All models that included f H had low AICc values and high R 2 . Models including either T b or T diff resulted in better fits, but the increase in model fit provided by adding in a temperature measure came with a high penalty of AICc. The wide range of T b at any given MR and between the two temperature regimes illustrates the low predictive ability of both T b and T diff (figures 2 and 4).

Free-ranging heart rate and body-temperature telemetry
The mean f H of free-ranging M. molossus in their roosts were generally low across the 24 h period but were in the range of 58-1068 bpm. Heart rate varied most in the early evening period when bats returned from foraging or were returned to their roost after instrumentation (figure 5). The heart rate was elevated to over 1000 bpm during those times. Mean roosting f H were below 200 bpm for all periods of the day and night (147 ± 57 bpm) and remained in stable low-level states (electronic supplementary material, figure S3).
T b of roosting bats were in the range of 25.3-37.2°C and varied both within an hour and across the day (figure 6). T b increased with daily T a (χ 2 1 = 39.7, p < 0.001; slope = 0.481, R 2 m = 0.121, R 2 c = 0.461). f H in the lowest 10% of observed values (i.e. lower than 90 bpm) were observed in all hours of the day except 2, 5, 19 and 20 h. These low rates were observed at T b up to 33°C, and f H of 100 bpm was observed at up to 36°C (electronic supplementary material, figure S3).
Our in-roost T b measures indicated that M. molossus used heterothermy, with minimum T b occurring during 02.00-04.00 in the coolest portion of the night. As daily T a increased during the diurnal period, T b became less variable, with individuals maintaining a constant T b that was close to T a .   T diff explained 34% of the variation in f H when inter-individual variation was accounted for via random effects in our models. This is best illustrated by individual daily variation in these variables (figure 5). In the example we illustrate, the bat roosted between a wooden ceiling and a metal roof approximately 2 m above the ground. We could not measure the interior temperature of the roost, but T b measured through the PIT tag reader ranged from 25.3 to 37.2°C and heart rate was generally low.

Discussion
We found that Molossus molossus entered low energy states independent of T a , which results in substantial energy savings across much of the T a they experienced. In respirometry chambers, f H was the best predictor of metabolic rate. These results extended to free-ranging M. molossus that used low f H across nearly the full range of T b measured. Minimizing energy expenditure in tropical settings with low variation in high T a may then be possible through modifications of heart rate and oxygen consumption independent of T b . Such physiological adaptations are perhaps vital for tropical lowland endotherms because they free them from some of the energetic constraints of high T a , but these effects remain understudied like many aspects of tropical ecosystems [59].
A range of mammals achieve low metabolic rates at relatively high T b , described as potential 'hyperthermic daily torpor' by Lovegrove et al. [8]. For example, in gerbils (32-35°C) [40], fat-tailed dwarf lemurs (30°C) [25] and some desert dwelling bats (31-33°C, [18,19]), T b thermoconforms at high T a while the animals remain in low metabolic states, and the MR then increases with rising T a . Bats appear to be particularly flexible in their thermal profiles, and some are even able to thermoconform to T a > 45°C for extended periods of time, which would be lethal in many other mammals [19,60]. This   may be related to the high T b generated during flight [14], or to the high TNZ upper critical temperature (e.g. 38°C) found in many species [49], although the thermoneutral dynamics for most bat species are underexplored. In previous work on tropical and subtropical bats, the finding of a reduction in T b associated with heterothermy is variable, and they only show heterothermy when T a is below 20-25°C   Figure 6. Hourly means of f H (a) and T b (b) simultaneously measured from bats in their natural roosts. Scotophase is indicated by the filled bar along the bottom of each panel. [29,[61][62][63][64]. However, many of these species, particularly those studied in free-ranging field conditions, have only had T b measured. It is possible that like M. molossus, they enter low energy states at relatively high T b and maintain low heart rates across a wide range of T a and T b . Entering a low energy state at high T b , or a T b within the thermoneutral zone, would allow animals to lower their FMR without the costs of rewarming that are incurred as they leave torpid states [65], while taking advantage of the benefits of heterothermy with minimal costs [7,8]. Passive warming would be particularly advantageous for nocturnal animals that enter their active periods as T a lowers and is less likely to provide much passive thermal support in their transition towards active states. Our respirometry measures did not capture the full metabolic potential of M. molossus observed in the wild roosts. The T a that we used in our respirometry measures did not extend far beyond the lower critical temperature (ca 30°C) of the thermoneutral zone reported for M. molossus [49], but energy consumption at these temperatures was far lower than the basal metabolic rate (BMR) reported for this species, and steady-state MRs were lower at 28°C than at 32°C, with a mean difference of 0.029 kJ h −1 . The mean of the 30% quantile of respirometry MR (0.00314 ± 0.0005 W g −1 ) was only 39% of the previously measured BMR for M. molossus (0.008044 W g −1 [49]). We did not observe this value until the 90% quantile of our data, placing it among the highest MRs we recorded for our bats. The high temporal variability in metabolic rate indicates that bats did not employ a stable BMR in our respirometry measurements, despite extended measurement periods and T a that should be within a neutral range for this species. Accurately measuring the BMR and the thermoneutral zone is difficult when species thermoconform across a wide range of T a [7,66]. Our data show that measuring this type of energetic and thermal flexibility is important, and that we may find that this energetic flexibility is common in tropical bats.
Heart rate is an accurate measure of the metabolic rate during steady-state conditions and when transitioning between resting states. However, torpor with strong heterothermy is not just an extrapolation of a resting state as regressions of torpid bats would underestimate restingVO 2 by up to 75% [67]. The low metabolic states in M. molossus do not show such a curvilinear relationship, with linear models providing the best fit to our respirometry data. We are careful not to infer a continuous linear relationship once animals are exercising because this probably underestimates the metabolic rate during exercise [30]. Through careful calibration and the use of nonlinear exponential estimates of energy consumption from f H based on body mass and heart mass, it may be possible to even more accurately estimate the energetic expenditure of exercising free-ranging animals.
The regular low metabolic states used by M. molossus in both respirometry and natural roosts may be an overall energetic conservation strategy to compensate for their ephemeral food resources. These small bats typically forage over water bodies for less than an hour per night on unpredictable, but rich patches of insects at dusk and dawn [42]. When nights are particularly windy or rainy, these bats tend to forgo foraging (DKN Dechmann, unpublished data), and there is a positive relationship between the duration of foraging and f H in the roost [43]. Furthermore, we observed T b reductions across a wide range of T a , with free-ranging bats using T b of 25.3-37.2°C. Multi-day torpor bouts have not been observed in M. molossus and we did not attempt to measure this in our experiment. While M. molossus are capable of maintaining stable blood glucose levels for up to 48 h of fasting [68], total reduction of the MR through extended torpor-like states may allow them to cope with multiple nights of inclement weather. In some cases, T b was lower than the T a recorded at the Panama Canal (electronic supplementary material, figure S2). Bats then select cool, stable roost microhabitats regardless of their ability to maintain a low MR at high T b . The use of torpor by tropical and subtropical mammals [13,16,22,25,69,70] illustrates the utility of reductions in metabolic rates and T b during periods of low food availability. However, starvation risk is not the only driver of torpor, as well-fed bats will also use torpor to minimize time outside the roost [17]. These torpor and torpor-like states at high T b may be particularly important to reduce the evaporative water loss incurred by bats via the large naked membrane of their wings [14], which can be reduced by 90% during torpid states [71,72]. The possibility of low energy states at relatively high T b allows M. molossus to remain active and alert and move away from observers. This means that unlike the lethargic torpid bats with low T b in the temperate zone, bats at higher T b can escape from predators at any given time [71,73]. Daily torpor-like states in M. molossus allow them to minimize exposure to risks outside of their roost (such as predators and water loss), while maximizing their energy savings.

Conclusion
We suggest that the low energy state that we measure in M. molossus is torpor with shallow heterothermy. In other small mammals, minimum f H during daily torpor at low T b is near 70 bpm [36,37,[74][75][76]. This is similar to the minimum stable rates (58-75 bpm) of M. molossus both during respirometry and in their natural roosts, but much higher than the f H of 8 bpm for hibernating bats [67]. The metabolic rates used by our bats were well within the ranges reported for other mammalian orders that use daily heterothermy (figure 3), but occur at higher T b . Bats evolved in the tropics and our findings in M. molossus are probably an example of the basal form of heterothermy that evolved near the root of the mammalian radiation [77]. The ability to maintain low metabolic rates and subsequently low T b in deep torpor or hibernation would build upon this set of regulatory networks driven by cellular requirements for oxygen diffusion via f H , which reduced total metabolic rates while in thermoneutral conditions or above. Tropical heterothermy or torpor may allow individuals to flexibly adjust energetic expenditure to rapid environmental changes at fine timescales, as well as minimize energetic expenditure during pregnancy and lactation when T b reductions may be constrained [7,17,64,78].
Low f H in response to lowered oxygen demands is an effective, flexible strategy. Work integrating energetic expenditure via f H in birds repeatedly shows lower energetic expenditures than would be predicted [33,34,45,79,80], indicating that a diversity of adaptations may be possible by manipulating one of the primary drivers of energy delivery to metabolism. The widespread nature of heterothermy or torpor-like states in tropical species leads to challenges when using traditional cut-off T b or frequency distributions of T b . Heart rate studies in tropical bats, including this study, have shown surprising metabolic strategies to cope with life in warm ambient temperatures [21,26]. This direct information on the flow of energy through an individual allows us further insight into the variability of energetic strategies in tropical systems, particularly as ambient conditions become more unpredictable.