Keeping your cool: thermoregulatory performance and plasticity in desert cricetid rodents

Physiological ecologists have shown a long-standing interest in the ability of small mammals to cope with the hot and arid conditions found in desert environments (Schmidt-Nielsen and Schmidt-Nielsen, 1950a). Many cricetid rodents, Neotoma (woodrats) and Peromyscus (mice), are very successful inhabitants of the extremely hot and arid Mojave, Colorado and Sonoran deserts. As a group, they show few physiological adaptations to desert life (Walsberg, 2000), and their successful occupation of hot deserts appears to be facilitated by avoidance of extreme heat via nocturnality and the use of burrows that minimize exposure to challenging conditions on the desert surface during the daylight hours (Brown, 1968; Bradley and Yousef, 1972; Lee, 1963; MacMillen, 1965, 1983). Recent work, however, suggests that these behavioral strategies may not fully buffer small mammals from high environmental temperatures (Tracy and Walsberg, 2002; Walsberg, 2000) and that even nocturnal activity may be constrained by T_a (Levy et al., 2016; Murray and Smith, 2012). Currently, we have little information on the thermoregulatory responses and limits of wild cricetid rodents to environmental temperatures approaching or exceeding normothermic body temperature (T_b), and, as a consequence, we lack physiological and ecological contexts for understanding future responses to a warming climate.

Deserts are defined by water scarcity, and desert rodents, because of their small size, must balance their water budgets over a daily cycle. Consequently, understanding the costs of thermoregulation at environmental temperatures that approach and exceed T_b is essential for understanding the persistence of desert cricetids. Avenues of water intake include water produced by metabolism, preformed dietary water, and water from drinking. Avenues of water loss include water lost via urine, feces, and evaporative losses from the skin and respiratory tract. Under benign thermal conditions (air temperature, T_a<25°C), some rodents such as kangaroo rats, e.g. Dipodomys merriami, and a few cricetids such as Peromyscus crinitus, can offset daily water losses and balance their water budgets solely via metabolic water production (MacMillen and Christopher, 1975; MacMillen and Hinds, 1983; Schmidt-Nielsen, 1964). During the spring and summer in the southwest deserts, benign environmental conditions largely disappear and burrow temperatures may exceed 30°C (Tracy and Walsberg, 2002). When environmental temperatures approach and exceed normothermic T_b, both metabolic heat and environmental heat loads must be dissipated through evaporative cooling to keep T_b from exceeding lethal levels. Under these conditions, rates of evaporative water loss (EWL) increase markedly and metabolic water production alone is not sufficient to offset large evaporative losses (Blackwell and Pivorun, 1979; Bradley and Yousef, 1975; Chew, 1951; Edwards and Haines, 1978; Hainsworth, 1968; MacMillen, 1983).

To date, most studies of the thermoregulatory performance of rodents at T_a above T_b have been carried out in laboratory rats and mice (Rattus and Mus). In laboratory animals and the few studies on wild rodents, evaporative cooling is accomplished by copious salivation, which is sometimes accompanied by spreading of saliva onto the pelt (Chew, 1965; Hainsworth, 1968; Hart, 1971). Research on thermoregulatory performance in wild cricetid rodents at T_a>37°C has primarily focused on metabolism (Hart, 1971; Yousef, 1980) and only a handful of studies have characterized rates of EWL and variation in T_b (Brower and Cade, 1966; Layne and Dolan, 1975; Lee, 1963; Musser and Shoemaker, 1965; Nelson and Yousef, 1979; Weiner and Heldmaier, 1987). Efforts to partition evaporative heat loss into various pathways in rodents indicate that evaporative cooling via panting appears to be poorly developed (Adolph, 1947; Chew, 1951; Hudson, 1965; Leon et al., 2005; Wiegert, 1961). Although cutaneous evaporation increases substantially in Peromyscus and house mice with increasing T_a, it appears insufficient to offset large exogenous heat loads (Edwards and Haines, 1978). In all laboratory species studied to date, exposure to T_a above T_b results in rapid increases in T_b and a very limited evaporative cooling. In wild cricetids, the capacity for evaporative cooling is poorly understood and the efficiency of the thermoregulatory response and its costs in terms of the animal's water budget are unknown.

Rapidly increasing global temperatures (IPCC, 2014) suggest that a more integrative understanding of thermoregulation in the heat would provide insight into the responses of cricetid populations to future climates. The desert rodent community has remained remarkably stable over the last century of climate change, likely as a result of the ability to avoid extreme heat in their burrows (Riddell et al., 2021). However, predicting future responses to climate change will require an understanding of thermoregulatory performance at both the intraspecific and interspecific levels. Few studies, however, have characterized thermoregulatory performance of rodents inhabiting hot desert environments (Brown, 1968; Nelson and Yousef, 1979; McNab and Morrison, 1963); among these, Brown (1968) found that low versus high desert populations of Neotoma albigula have lower body mass, higher conductance and superior vascular mechanisms for cooling, which lead to greater heat tolerance. There are no data for wild rodents that provide an integrative view of thermoregulatory performance that includes simultaneous measurements of T_b, resting metabolic rate (RMR) and EWL for animals exposed to T_a exceeding normothermic T_b. Data are also lacking that explore differences in seasonal and geographic thermoregulatory performance.

In this study, we examined both the spatial and temporal variation in thermoregulatory performance of rodents in the Mojave desert of California and characterized the costs associated with maintaining T_b within critical thermal limits. To this end, we obtained an integrated view of thermoregulatory performance by simultaneously measuring metabolic rate (MR), EWL and T_b over T_a ranging from 10 to 48°C in four sympatric rodents [Neotoma lepida Thomas 1893 (desert woodrat), Neotoma albigula Hartley 1894 (white-throated woodrat), Peromyscus eremicus (Baird 1858) (cactus mouse) and Peromyscus crinitus (Merriam 1891) (canyon mouse)] at different sites during both summer and winter. Our study focused on the following questions. (1) Does heat exposure produce a marked hyperthermic response and/or evoke a strong evaporative cooling response in cricetid rodents, and if so is the thermal response metabolically costly? We predict that heat exposure will produce rapid increases in T_b and that any up-regulation of evaporative cooling will be metabolically costly because of the costs of salivation and wiping behavior. (2) Do cricetid rodents show heat tolerance limits that are well above T_b, demonstrating highly effective evaporative cooling, and if so how does the effectiveness of this cooling strategy compare with that of other taxa such as birds? Because most birds are diurnal and confront desert environments directly, we predict that birds will show higher heat tolerance and cooling capacity than the nocturnally active rodents (3) Do the thermoregulatory responses of cricetids across sites and seasons show significant plasticity within and among populations? We predict that a limited capacity for responding to heat stress within species will confer limited flexibility across sites and seasons. (4) Finally, we use our measurements to ask what are the potential water balance costs of cooling that might allow continued activity in a warming environment?

The Cricetidae is a large family of ubiquitous rodents that includes the New World rats and mice. Representatives of this family are found on three continents in virtually all environments. In the deserts of the American Southwest, N. albigula ranges from Colorado to Central Mexico and from California to Texas (Macêdo and Mares, 1988), N. lepida and P. crinitus range from Oregon to the southern tip of the Baja Peninsula and from California to Utah (Johnson and Armstrong, 1987; Verts and Carraway, 2002), and P. eremicus occurs from Nevada to Central Mexico and California to Texas (Veal and Caire, 1979. Neotoma are nocturnal herbivorous rats that feed primarily on seeds, green vegetation and succulents. Peromyscus are nocturnal omnivorous mice that feed on insects, seeds and green vegetation (Orr et al., 2015).

We examined Cricetidae thermoregulation during the summers (June–August) and winters (January–February) of 2017 and 2018. We collected data at two sites in the summer and winter: Joshua Tree National Monument and Zzyzx, and only during the summer at the two Midhills sites (Fig. 1). The biome at the Joshua Tree National Park site (33.72898, −115.82505, elevation 1000 m) was desert dry wash and creosote brush scrub with an annual precipitation, and a mean and mean annual high temperature of 137 mm, and 17 and 25°C, respectively. Zzyzx (35.12731, −116.11793, 300 m) was the hottest and driest site with 108 mm of annual precipitation, and a mean and mean annual high temperature of 21 and 30°C. At Zzyzx we trapped in the salt brush scrub biome. The Midhills low site (35.1669, −115.45377, 1400 m), was predominantly Joshua tree woodland which received 175 mm annual precipitation, with a mean and mean annual high temperature of 15 and 22°C. Our coolest site, Midhills high (35.14148, −115.4083, 1600 m) was sagebrush scrub with 209 mm of precipitation, and mean and mean annual high temperature of 14 and 20°C. At the extreme, T_a at Zzyzx routinely exceeds 43°C each year for more than 30 days during the summer, while minimum temperatures at Midhills high commonly drops below −5°C during the winter. The climate data for each site above were estimated using the 30 year daily normals (1981–2010) from the PRISM Climate and Weather System (PRISM Climate Group, Oregon State University).

We used Sherman folding traps (H.B. Sherman Traps, Inc., Tallahassee, FL. XLF15) baited with mixed birdseed (Audubon Park Wild Bird Food) for animal collection. Trapping arrays were line transects approximately 120 m in length, with traps set every 3 m. We also set traps at woodrat middens and burrow entrances. We only studied adult males or adult females that did not show obvious signs of pregnancy or lactation. All trapping occurred more than 200 m away from any occupied human structure to avoid capturing animals acclimatized to air-conditioned environments. Following each experiment, we returned rodents to their traps and offered seeds and celery for rehydration. We held animals in Sherman traps in a cool, quiet area for no longer than 12 h and released animals at their capture location. We obtained research permits from the CA Department of Wildlife, Joshua Tree National Park (JOTR-2015-SCI-0032) and Mojave National Preserve (MOJA SC-13686, MOJA-2015-0036). The Animal Care and Use Committee of the University of New Mexico approved all protocols (16-200518-MC).

We measured CO₂ production and total EWL using positive pressure flow-through respirometry. A pump (DAA-V 15-EB, Gast, Benton Harbor, MI, USA) pushed atmospheric air through a column of Drierite™. On humid days, we also used a column of silica gel before the Drierite™ column. The air then passed through a mass flow controller (MC-5SLPM, MC-20SLPM, accuracy: ±0.8% of reading+0.2% of full scale, Alicat Scientific, Tucson, AZ, USA) at flow rates ranging from 0.5 to 15 l min⁻¹ STP. We adjusted the flow rate to maintain respirometry chambers at dew points <5°C, which allowed for uninhibited evaporative cooling (Lasiewski et al., 1966). Dry air then moved through clear, airtight plastic containers containing the animals (Peromyscus in 1.1 l volume, 11.4×13.3×7.6 cm; Neotoma in 3.6 l, 22.8×15.2×10.1 cm). A wire mesh platform in the bottom of the chambers suspended the animals above 2–3 cm of mineral oil. The oil captured urine and feces, preventing fecal and urinary water from entering the airstream. A thermocouple (model RET-4, Physitemp, Clifton, NJ, USA) connected to a National Instruments (Austin, TX, USA) thermocouple module recorded the chamber temperature. We subsampled excurrent air from the chambers (∼400 ml min⁻¹) and directed the air to the CO₂/H₂O gas analyzers (model 840A, accuracy ±1 ppm CO₂, ±0.01% reading for H₂O, Li-Cor, Lincoln, NE, USA). For data acquisition, we connected the gas analyzers, thermocouples and mass flow controllers to a National Instruments™ DAQ and visualized using LabVIEW (National Instruments™).

We zeroed and spanned the CO₂ analyzers using Ascarite (Thomas Scientific, Swedesboro, NJ, USA) and gas with a known concentration of 1840 ppm CO₂. We used Drierite™ and a dew point generator (model 610, Li-Cor) to zero and span the water channels, and the zero on these analyzers was checked daily. We verified the mass flow controller calibrations at the beginning of each field season against a precision calibrated reference mass flow meter (M-20SLPM-D, 0–20 l min⁻¹, accuracy: ±0.4% of reading+0.2% of full scale, Alicat Scientific) over its full range. The mass flow controllers varied from the reference mass flow meter by <±0.04 l min⁻¹.

Experimental trials began with a rest period where animals were held in the dark in an unconnected metabolic chamber at a temperature of 25–28°C and allowed to sit quietly for 50–75 min. After this period, we placed the chambers in the environmental chamber and connected airflow lines and thermocouples to begin the trial. Experimental trials lasted 4–5 h, and we held animals at each experimental T_a until we obtained stable CO₂ readings. Once CO₂ values stabilized, we recorded a 15 min sample period for each temperature treatment, followed by a 5 min baseline period. We ran metabolic trials during the rest phase of the circadian cycle, and we tracked animal activity continuously using infrared security cameras (UDS655, Uniden^® Guardian, Tokyo, Japan). Animals that showed moderate to high activity in the chambers were removed from the analysis. Experimental T_a in the chamber ranged from 10 to 44°C, and we ramped the chamber T_a at 2–8°C intervals over 30–60 min during trials. We typically obtained data at five T_a values for each animal, and we limited trials to lengths where mass loss did not exceed 5% of body mass (M_b) to minimize the effects of dehydration on cooling efficiency. We recorded and monitored T_b in real-time during trials via a temperature-sensing passive integrated transponder (PIT) tag (BioThermo13, accuracy ±0.02°C, BioMark^®, Boise, ID, USA) (Whitfield et al., 2015) using a Biomark^® HPR Plus reader. The PIT tag was implanted sub-dermally between the shoulders of each rodent using a tag injector (Biomark^® MK10) at least 1 h prior to placing the animal in a chamber. Immediately prior to and after each trial, we obtained rectal temperatures and compared these values with the PIT tag readings. We used a measured approach to minimize mortality by estimating heat tolerance limit (HTL) values using criteria based on behavior, continuous escape attempts, and physiological responses including a depression in CO₂ production, rapid spikes in T_b or T_b increasing to 43°C (McKechnie and Wolf, 2019). Animals exposed to high temperatures were cooled rapidly after removal from the experimental trial by holding the animals in front of a fan in the cool room.

We used R version 3.5.1 (http://www.R-project.org/) to calculate the MR and EWL rates from our raw values of EWL using equations 10.5 and 10.9 of Lighton (2008a). Using a running mean, we selected the lowest average stable minute from each 15 min sampling period. We assumed that animals fed in the traps before testing, and thus we assumed a respiratory quotient (RQ) of 0.95, a value derived from the proportion and nutritional content of the seeds used as bait (Walsberg and Wolf, 1995). We converted rates of CO₂ production into MR using a thermal equivalent of 21.5 J ml⁻¹ CO₂ (Withers, 1992). We converted rates of EWL to heat loss using 2.43 kJ g⁻¹ H₂O (Lighton, 2008b). We used the segmented package in R to identify the T_a associated with the upper and lower critical temperatures as well as inflection T_a for T_b and EWL (Muggeo, 2003). We determined individual HTLs as the highest T_a where T_b remained below lethal levels (43°C). We estimated maximum values for MR, EWL and evaporative heat loss (EHD=EHL/MHP, where EHD is evaporative heat dissipation, EHL is evaporative heat loss and MHP is metabolic heat production) at the HTL using the slope intercept from the regression lines drawn above the critical points and at the HTL. We defined metabolic and evaporative scope as the ratio of EWL or MR at the HTL divided by the respective value at thermal neutrality (EWL­_HTL/EWL_TNZ, MR_HTL/MR_TNZ, where TNZ is thermoneutral zone).

We used R version 3.5.1 (http://www.R-project.org/) to conduct the statistical analyses. We compared physiological responses across seasons and sites with a species-specific analysis using Type-II analysis of covariance (ANCOVA), with individual as a random effect. To generate P-values, we used a Type-II ANCOVA from the lmerTest() package that uses a Kenward–Rogers approximation to Satterthwaite's methods for denominator degrees of freedom (Kenward and Roger, 1997). We analyzed the effects of mass across sites and seasons, using linear regression analysis with a Type-II analysis of variance (ANOVA) from the car package to evaluate significance (Langsrud, 2003). We conducted separate analyses on MR below the lower critical temperature (T_lc), above the upper critical temperature (T_uc), and within the TNZ. Similarly, we analyzed changes in T_b at T_a above and below the inflection of the hyperthermic response. For EWL, we examined changes in rates of EWL above and below the inflection point at which animals showed a marked increase in evaporative cooling. For each analysis, we removed sites and species with only a single season of data to balance the statistical design.

Neotoma albigula M_b (124.1 g, 95% confidence interval CI: 118–130.2 g, N=13 animals) was the largest of all the species, followed by N. lepida (M_b=112.8 g, 95% CI: 111.2–114.3 g, N=155), P. eremicus (M_b=18.9 g, 95% CI: 18.7–19.2 g, N=88) and P. crinitus (M_b=12.7 g, 95% CI: 12.6–2.8 g, N=111). We found significant differences among sites in M_b of N. lepida (F_4,139=5.32, P<0.001) and P. crinitus (F_2,100=9.54, P<0.001), with M_b varying by as much as 23.9% and 31.9% among sites, respectively (Table 1). We recorded the lowest M_b in populations from Zzyzx, the warmest habitat.

During high temperature trials, animals were observed licking their inguinal area and forearms at T_a≈38–39°C. Animals could clearly be seen salivating and at higher temperatures (T_a>41°C), and the fur was often noticeably wet when they were removed from the chamber at the end of a trial.

In both Neotoma and Peromyscus, we found no systematic offset between the rectal T_b values and PIT tag values at any T_a (Figs S1 and S2, and Table S1). There was little variation in normothermic T_b across species, with values averaging ∼36°C, when T_a varied between 10 and 31°C (Fig. 2, Table 1). All species showed a hyperthermic response to increasing T_a starting in Peromyscus at T_a≈31°C and Neotoma at T_a≈33°C (Fig. 2, Table 1) Species-specific analyses indicated that the differences associated with the interaction between site and season were primarily driven by N. lepida (F_1,62.3=25.40, P<0.001) and P. crinitus (F_1,73.7=3.97, P=0.050). Neotoma lepida T_b rose 0.02°C and P. crinitus T_b increased 0.01°C above the inflection point for the onset of hyperthermia during the summer at Joshua Tree. Conversely, at Zzyzx, T_b above the inflection point during the summer decreased 0.01°C for N. lepida and 0.002°C for P. crinitus. In contrast, P. eremicus exhibited a trend for a slightly higher T_b above the inflection point in the winter compared with the summer at Zzyzx (F_1,37.2=3.17, P=0.083).

We selected an endpoint T_b of 43°C for all trials as its exceedance produced rapid death (see also Leon et al., 2005). Some animals that were removed from the chamber at T_b=43°C continued to experience increases in T_b from increased activity or thermal inertia. We observed death in individual animals at T_b=43.5°C in N. lepida; T_b=42.7, 43.2, 44, 45°C in P. eremicus; and T_b=45.8°C in P. crinitus. In general, N. lepida showed slightly higher HTL at T_a≈44.5°C, in comparison with P. crinitus at T_a≈43.5°C and P. eremicus at T_a≈44°C (Table 1). At the HTL, N. lepida and P. eremicus were able to defend a T_b–T_a gradient of approximately 1.5°C, while P. crinitus could only defend a gradient of 0.5°C (Table 1). We did not observe a consistent relationship between HTL and site or season.

The larger Neotoma showed a slightly broader TNZ (N. albigula ∼27–37°C, N. lepida ∼27.5–35°C) in comparison with the smaller Peromyscus (P. crinitus ∼29.5–35°C, P. eremicus ∼28–36°C) (Fig. 3, Table 2). The RMR of all species increased from summer to winter; N. lepida increased RMR during the winter by 39.3% (F_1,76.3=47.10, P<0.001), P. eremicus by 37.8% (F_1,44.04=9.74, P=0.003) and P. crinitus by 19.2% (F_1,89.9=13.41, P<0.001). At Zzyzx, N. lepida increased MR by 43.8% from summer to winter, which was much higher than the 13.3% increase observed at Joshua Tree (F_1,76.6=8.65, P=0.004). T_uc, the T_a marked by a steep increase in MR, varied seasonally and was higher in the larger Neotoma than in Peromyscus (Table 2). Above the upper critical point, N. lepida increased MR from summer to winter by 30.9% (F_1,67.9=37.17, P<0.001), P. eremicus by 32.5% (F_1,31.3=15.92, P<0.001) and P. crinitus by 17.5% (F_1,71.5=17.34, P<0.001). At Zzyzx, N. lepida increased MR from summer to winter by 64.4%, which was much greater than the 19.3% increase observed in Joshua Tree animals (F_1,71=5.76, P=0.019).

Thermoneutral EWL was highly associated with M_b and T_a at the species level for N. lepida (M_b: F_1,81.99=7.09, P=0.009; T_a: F_1,260.1=46.83, P<0.001), but not P. eremicus (M_b: F_1,39.42=0.093, P=0.340; T_a: F_1,159=1.98, P=0.161). For P. crinitus, thermoneutral EWL was associated with M_b (F_1,95.72=10.61, P=0.002) and T_a (F_1,347.7=6.71, P=0.01) (Fig. 4, Table 3). Seasonally, thermoneutral values for EWL were 20.0% higher during the winter than during the summer for N. lepida (F_1,81.65=5.72, P=0.02), 43.7% higher during the winter for P. eremicus (F_1,38.4=10.18, P=0.002) and 25.1% higher during the winter for P. crinitus (F_1,93.1=20.22, P<0.001). Average rates of EWL within the TNZ ranged from 1.6 to 4.4 mg g⁻¹ h ⁻¹ across all species, sites and seasons. Maximal rates of EWL at the HTL were 29.5 mg g⁻¹ h⁻¹ for N. albigula, and ranged from 20.7 to 34.7 mg g⁻¹ h⁻¹ for N. lepida, 42.4 to 104.1 mg g⁻¹ h⁻¹ for P. eremicus and 47.3–70 mg g⁻¹ h⁻¹ for P. crinitus (Table 3).

Rates of EWL above the inflection point for evaporative cooling increased from summer to winter and varied among species (Table 3). EWL values increased by 65.0% from summer to winter in N. lepida (F_1,72.6=209.26, P<0.001), and by 66.6% in P. eremicus (F_1,38.5=20.20, P<0.001) and 68.4% in P. crinitus (F_1,81.4=18.88, P<0.001). Seasonal responses within species also sometimes varied by site, with the population at Zzyzx showing a greater seasonal increase compared with the Joshua Tree population (Table 3). Neotoma lepida from the Zzyzx population showed a 128% increase from summer to winter, which was much greater than the 37.5% increase observed in the Joshua Tree population (F_1,55.5=5.23, P=0.026). The Zzyzx P. crinitus population also showed a large seasonal increase in EWL of 95.6% which was much greater than the 34.0% increase observed in this species in Joshua Tree (F_1,73.4=3.89, P=0.052). Seasonal increases in rates of EWL generally did not translate into increases in species’ heat tolerance limit.

The values for evaporative scope (EWL_HTL/EWL_TNZ) varied among species, season and site (Table 3) and was lowest in Neotoma, with values ranging from 12.2 to 17.4 compared to EWL scope values of 14.3 to 24.3 observed in the smaller P. eremicus and values of 13.9 to 18.8 observed in P. crinitus. The maximum EHD value (where EHD is defined as the percentage of metabolic heat dissipated via evaporation) at the HTL ranged from 1.9 to 2.5 for Neotoma and from 2.3 to 4.3 in Peromyscus (Table 3).

Our goal was to quantify the thermoregulatory responses used by wild cricetid rodents to cope with environmental temperatures exceeding T_b and examine how these responses varied among species, sites and seasons. To this end, we made simultaneous measurements of MR, EWL and T_b on wild-caught, individuals at T_a ranging from 10 to 48°C during summer and winter from three different sites in the Mojave Desert. The common thermoregulatory pattern observed with increasing T_a started with stable T_b (∼35–37°C) in the TNZ followed by marked increases in T_b as T_a increased to ∼31–34°C (Fig. 2), which then increased steeply until the animal approached its HTL at T_b≈43°C. Estimated HTL for all species ranged from a T_a of 42 to 46°C. Increases in evaporative cooling lagged behind increases in T_b, with evaporation rates increasing starting at T_a≈33–38°C accompanied by sharp increases in metabolism. Metabolic increases represent the mechanical costs of increasing evaporative cooling (increased breathing rate, salivation and wiping behaviors) and resulted in large metabolic scopes (2–3) at the HTL. Scopes for evaporative cooling at the HTL were also notably large (13–24) and represent the acute response needed to offset environmental and metabolic heat loads to avoid lethal hyperthermia (Table 3). Overall, we found that desert cricetids have a very limited ability to cool evaporatively and are unable to maintain a T_a–T_b gradient of more than a few degrees Celsius. We also found limited seasonal variation in evaporative cooling ability, with winter animals showing modestly higher evaporative capacities. The thermal limits within species also did not appear to vary among individuals from different elevations. Overall, evaporative cooling in cricetid rodents is a very ineffective process with high rates of evaporation and metabolism, yielding only limited cooling ability and heat tolerance.

Increases in T_b are common in endotherms as environmental temperatures approach normothermic T_b (McKechnie and Wolf, 2019). We observed hyperthermic responses in all species with T_b inflection points occurring at T_a of ∼31°C and ∼33°C in Peromyscus and Neotoma, respectively. Increasing T_b serves to maximize the gradient for insensible heat loss with increasing T_a, minimize the gradient for heat gain between the animal and its environment, and allow for heat storage in tissues; these phenomena reduce evaporative cooling costs and conserve water (Chew, 1951; Schmidt-Nielsen and Schmidt-Nielsen, 1952). Our values are similar to the inflection T_a measured by Nelson and Yousef (1979) (T_a=30°C) and Lee (1963) (T_a=35°C) for N. lepida, and by Brown (1968) (T_a=32°C) for N. albigula. We were unable to observe any general patterns in T_b values within and among species or across sites and seasons. Increasing T_a above the inflection point was accompanied by steep concomitant increases in T_b that rapidly drove rodent T_b towards lethal values.

We estimated the HTL for each species by site and season by quantifying the T_a at which heat loads became uncompensated and resulted in uncontrolled increases in T_b or mortality. Our HTL values provide a non-lethal means of estimating and comparing rodent thermal tolerance across sites and seasons and linking thermal limits to EWL and metabolism. These estimates of HTL represent a trade-off between the goal of establishing the maximum T_a tolerated by an individual species at differing sites and seasons and not killing all of the test subjects. Erskine and Hutchison (1982) determined the critical thermal maximum (CT_max) for five species of rodents using rapid heating (30–40 min to lethality), resulting in 100% mortality among study animals. We used a more measured approach to minimize mortality by estimating HTL values using criteria based on behavior, continuous escape attempts, physiological responses including depression in CO₂ production, rapid spikes in T_b, or T_b increasing to 43°C (McKechnie and Wolf, 2019). Our HTL values for Peromyscus and Neotoma ranged from a T_a of 42.5 to 45.7°C, showing no clear seasonal pattern or effect of site. Our estimated heat tolerance values may be somewhat conservative because we ended experimental trials when T_b=43°C, though some Neotoma have been observed to survive a T_b of 44°C (Brown, 1968; Nelson and Yousef, 1979). Our exposure times (<1 h) to the highest temperatures were substantially shorter than those used in previous studies (1–7 h), which found lethal values (Brown, 1968; Lee, 1963; Nelson and Yousef, 1979). The long exposure time used in earlier studies may have resulted in HTL values defined by dehydration state as opposed to acute heat stress. Our data suggest that Neotoma held at these high T_a would have lost 10–14% M_b of water during a 5–7 h trial (Table 3), which would have significantly affected plasma volume and hindered cooling ability (Zurovski et al., 1991). The effect of dehydration resulting from prolonged heat exposure helps explain why our HTL values for N. lepida (43–46°C) are well above the previous values of other studies (37–43°C) (Brown, 1968; Lee, 1963; Nelson and Yousef, 1979). Determination of the values for thermal limits is also affected by how T_b measurements are obtained. It is very difficult to obtain continuous T_b measurements without telemetry or implants because of the tendency of rodents to chew temperature leads. Most commonly, single-point T_b measurements have been obtained rectally in animals after experimental trials (e.g. McNab and Morrison, 1963; Brown, 1968). Our values were obtained continuously via subcutaneous PIT tag implants which were checked against rectal T_b measurements both before and after each experimental trial (see Figs S1 and S2).

In general, our metabolic values and defined critical limits agree with earlier studies on Mojave Peromyscus and Neotoma (Table 4). We note, however, that our metabolism measurements and estimates of critical temperatures were made under experimental conditions that differ from those of most earlier studies (e.g. Murie, 1961; McNab and Morrison, 1963). These earlier studies focused on obtaining minimum metabolism values during long trials at ‘steady state’, in fasted rest-phase animals that often had been held in captivity at a T_a of ∼25°C for weeks or months. These values are unlikely to represent an accurate picture of animal performance in nature. In this study, by necessity and design, we focused on making measurements of wild-caught, non-fasted animals. For animals exposed to T_a near and above T_b, we conducted measurements over shorter periods where animals were held at a single temperature for 30–60 min to limit total water loss and changes in the hydration state of the animal that would have diminished its thermoregulatory performance (Sawka et al., 2001). As a consequence, we urge some caution in directly comparing our values with those from earlier studies (Table 4).

As chamber T_a increased towards T_b, all animals showed increased MR as a result of their efforts to actively increase evaporative heat loss. Across species, seasons and sites, T_uc values on average ranged from 34 to 37°C with no consistent seasonal or site-based difference in any species and were broadly similar to those reported by other researchers, ranging from 33 to 35°C for Neotoma and 35 to 36°C for Peromyscus (Table 4) (Brown, 1968; Lee, 1963; Murie, 1961; McNab and Morrison, 1963; Nelson and Yousef, 1979). Increases in metabolism at the T_uc were closely followed by marked increases in the rates of EWL, with T_a for EWL at inflection of about ∼0.4°C above the T_uc in Neotoma and 1.5 and 2.4°C above the T_uc in P. eremicus and P. crinitus, respectively. Animals were observed salivating at a chamber T_a of 38–39°C and the steep increases in metabolism above the T_uc represent the costs of increasing evaporative cooling rates to maintain pace with increasing environmental heat loads. Unfortunately, increases in metabolic heat production also increase the total heat load that must be dissipated through evaporative cooling (Lee, 1963; Schmidt-Nielsen and Schmidt-Nielsen, 1950a; Nelson and Yousef, 1979). Overall, metabolic heat production increased by 150–250% from the T_uc to the HTL, increasing the large burden on the evaporative heat cooling system and driving animals towards lethal T_b.

At T_a within the TNZ, EWL in rodents is almost evenly divided between respiratory and cutaneous avenues. This represents the minimum obligatory water loss associated with lung ventilation and skin permeability (Baudinette, 1972; Chew, 1955; Edwards and Haines, 1978; Yousef, 1980). Our values for thermoneutral rates of EWL among species compare well with those of earlier studies (Table 4). In all species, we observed significant reductions in thermoneutral rates of EWL during the summer that likely reduced total water requirements by 10–15% during a period where free water is scarce and increased temperature/decreased humidity potentially drive increased water demands. This may be largely due to reduced MR during the summer, which would tend to reduce lung ventilation rates and respiratory EWL (Schmidt-Nielsen, 1964).

Evaporative heat loss can be critical to survival when T_a approaches and exceeds T_b, and, in small rodents, pathways for evaporation may include panting, salivation (including extensive wiping of saliva on the animal's surface) and increases in cutaneous water loss (Edwards and Haines, 1978; Hainsworth, 1967, 1968; Lee, 1963). For rodents exposed to environmental temperatures exceeding T_b, saliva spreading is the dominant mechanism used to increase evaporative cooling and cope with potentially lethal heat loads (Hainsworth, 1967, 1968; Hainsworth and Epstein, 1966; Hainsworth et al., 1968; Horowitz and Mani, 1978; Nelson and Yousef, 1979). Few studies have measured rates of EWL in cricetids at T_a exceeding 35°C, and this study is the first to look across multiples species at different sites and seasons (Chew, 1955; Layne and Dolan, 1975; MacMillen, 1965; Yousef, 1980). Our EWL data for these higher T_a overlap with those for the few earlier studies (Table 4) that have measured water loss in Neotoma (Nelson and Yousef, 1979) and desert Peromyscus (MacMillen, 1965). We also found seasonal differences in rates of water loss at the highest exposure temperatures. In Neotoma, maximum rates of EWL were as much as 20% higher in the winter and were accompanied by higher HTL values, reflecting the greater cooling abilities of winter animals. Similarly, we found that maximum rates of EWL increased in the winter in both P. eremicus and P. crinitus by as much as 96% and 23%, respectively, and sometimes reflected modestly higher HTL values.

Across all species and populations of Neotoma and Peromyscus, we observed maximal evaporative scope ranging from 16 in Neotoma to 24 in Peromyscus, which suggests a significant capacity for evaporative cooling (Table 3). These values for rodents are greater than the EWL scope values observed in desert songbirds (Passeriformes) (Smith et al., 2017; McKechnie et al., 2017; Whitfield et al., 2015). However, the HTL for cricetid rodents (∼43–46°C) was significantly lower than that observed in songbirds (∼48–52°C). These differences in actual cooling capacity across taxa can be illustrated by comparing relative evaporative heat loss (EWL scope) with the maximum T_a–T_b gradient that can be maintained by an animal exhibiting maximal rates of water loss. When comparing EWL scope with the maximum T_a–T_b gradient maintained in desert songbirds (summer) from North America, southern Africa and Australia with our values for Mojave cricetid rodents during the summer (Smith et al., 2017; McKechnie et al., 2017; Whitfield et al., 2015), we found that, strikingly, cricetid rodents have very high evaporative scope – almost twice that of most songbirds. However, high evaporative scope does not predict the relative capacity for coping with high heat loads: N. lepida, P. eremicus and P. crinitus can only withstand a T_a of ∼1°C greater than T_b. Most songbirds that show relatively low EWL scope can withstand T_a that averages 5.4°C greater than T_b. Ecologically, this makes sense because we are comparing diurnally active desert birds with nocturnal desert rodents. The conundrum here is that the cricetid rodents vastly increase water loss rates at high T_a with apparently minimal cooling benefit. Songbirds pant to increase evaporative heat loss when exposed to high T_a. In contrast, few cricetid rodents are known to pant and most use salivation, which may be accompanied by wiping saliva on the fur to enhance evaporative heat loss (Hainsworth, 1967, 1968; Horowitz and Mani, 1978). Based on our data, salivation is a very inefficient means of evaporative heat loss. By comparing total maximum rates of EWL and the maximum T_a–T_b gradient between taxa, we found that cricetid rodents (this study) evaporate an average of 42.2 mg H₂O g⁻¹ h⁻¹ and maintain an average maximum T_a–T_b gradient of 1.3°C, which contrasts with Australian songbirds that evaporate an average 44.6 mg H₂O g⁻¹ h⁻¹ and maintain an average maximum T_a–T_b gradient of 4.0°C (Fig. 5) (McKechnie et al., 2017). These efficiency differences among birds and rodents are striking and probably greatly overestimate salivation efficiency as a mechanism of EHD, because in all cases when rodents were removed from trials after being exposed to T_a>T_b, their fur was noticeably wetted. Our measurements of EWL only account for water that was evaporated and carried in the airstream to our analyzers. The wetted pelts represent an unmeasured water loss, indicating that water loss is likely much higher and that the cooling benefit is very low (Hainsworth, 1967, 1968; Nelson and Yousef, 1979; Brown, 1968). Our data suggest that cricetid rodents have few physiological tools for coping with T_a that approaches T_b in the wild.

The lack of effective or efficient cooling mechanisms suggests that cricetid thermoregulation and survival are mediated almost entirely through behavioral processes. The cricetid rodents we examined are nocturnal species that persist and thrive in hot deserts such as Death Valley, CA, USA, where daytime T_a regularly exceeds 55°C, despite their limited capacity to cope with T_a>T_b. Although desert rodents seek the thermal refuges provided by burrows or nests during the day, as small endotherms their high energy and water demands require that they forage regularly. Some smaller species such as P. eremicus may use daily torpor to provide temporary relief from environmental extremes (MacMillen, 1965). However, periods of high temperature extremes are becoming more frequent, intense and prolonged (IPCC, 2014), and night-time T_a in these regions is increasing. Despite the stability of the rodent community in the Mojave Desert over the last century (Riddell et al., 2021), the future of rodent populations in a rapidly warming world is uncertain. Our study indicates that even modest increases in environmental temperatures could importantly impact thermoregulatory costs and limit activity. Murray and Smith (2012), using animals collared with T_a loggers, found that N. lepida did not emerge from nests or burrows until T_a<42°C, that larger animals emerged later than smaller animals, and that the active periods were reduced by nearly 50% at daily maximum T_a>41°C. Our data suggest that N. lepida can tolerate short exposures of T_a>T_b and respond by increasing evaporative heat loss via copious salivation and spreading, but the losses of water conservatively range from 2% to 6% M_b h⁻¹ which makes lethal dehydration a risk (Table 3). Although burrows may provide a thermal refuge from extreme heat for desert rodents, burrow temperatures currently can exceed 35°C (Tracy and Walsberg, 2002). Higher diurnal air and burrow temperatures will increase overall water requirements, while at the same time higher minimum T_a during summer nights will increasingly limit nightly activity because of thermoregulatory constraints (Murray and Smith, 2012; Tracy and Walsberg, 2002; Walsberg, 2000). The current climate trends are not promising; the number of days with a minimum night-time T_a≥35°C at Furnace Creek in Death Valley was 3, 3 and 5 days in 1962, 1990 and 2000, respectively; in 2020, there were 22 days with minimum T_a≥35°C (NOAA/NCEI). A changing climate may also lead to changes in the availability of water-rich succulent food for desert rodents by altering the plant community (Abatzoglou and Kolden, 2011; Notaro et al., 2012). Finally, even adaptive shifts in body size are very unlikely to keep pace with the high rate of warming (Smith et al., 2014; Quintero and Wiens, 2013). Our data suggest that a general lack of ability to withstand T_a approaching or exceeding T_b and the associated high costs of responding to even modestly higher heat loads will increasingly constrain desert rodents under future climate scenarios.

We thank the National Park Service and the Desert Studies Center field station of CSU, especially Jason Wallace and Norma Charest. For their hard work gathering data in the Mojave heat, we thank Rosa Brasmer, Bill Talbot, Dylan Conradson, Britney Coe, Malik Alqawasmi and Armando Aispuro. Finally, we thank Dr Andrew McKechnie for feedback on the manuscript.