High total water loss driven by low-fat diet in desert-adapted mice

Abstract

Availability of food resources is an important driver of survival. Populations must either relocate or adapt to persist in environments where food availability is changing. An optimal diet balances energy gain, water regulation, and nutrition. We used flow-through respirometry to characterize metabolic phenotypes of the desert-adapted Cactus Mouse (Peromyscus eremicus) under diurnally variable environmental conditions that mimic that of the Sonoran Desert. We treated mice with 2 different energetically equivalent diets, a standard diet and a low-fat diet, and measured energy expenditure, water loss rate, respiratory quotient, weight, and electrolyte levels. Mice fed the low-fat diet lost significantly more water than those on the standard diet. Despite being desert-adapted, our results suggest that cactus mice may have limited capacity to tolerate water deprivation if optimal foods become less abundant. Given that climate change is predicted to modify the distribution of food items, understanding these links may have important implications for long-term population viability for desert and non-desert-adapted animals alike.

Food is one important way in which animals manage water. However, food varies in both composition and seasonal availability, forcing animals to make decisions that could have a substantial effect on their survival. To understand ways in which the desert-adapted Cactus Mouse, Peromyscus eremicus, responds to variation in dietary fat content, mice were fed to 1 of 2 diets: a higher fat diet or a lower fat diet. Mice fed the lower fat diet had higher rates of water loss and differences in serum electrolyte values indicating dehydration. In environments where water scarcity is extreme, even small differences in water balance could have substantial effects on survival and reproductive success.

One of the most critical challenges faced by desert mammals is the regulation of water. Water balance is maintained through a cascade of hormones that elicit a response designed to increase water retention and decrease water use in the excretion of nitrogenous wastes. Dehydration, in contrast, is an outcome of inadequate water balance. This results in a decrease in blood volume and an increase in osmolality, the latter an effect driven largely by the rise in serum sodium (Thornton 2010; Leib et al. 2016) and eventually leading to death if water homeostasis is not achieved. While hormonal regulation of water balance is essential for survival, behavioral, physiological, and molecular responses are also critical.

Dietary intake of food is one important way in which animals manage water. There are 3 main macromolecules that animals ingest as food: carbohydrates, fats, and proteins—all of which vary in energy and water content as well as in metabolic cost. Energy potential of macromolecules varies greatly, with carbohydrates and proteins yielding 16.74 kJ/g, while fats yield more than 37.66 kJ/g (Sánchez-Peña et al. 2017). The amount of water released during endogenous catabolism of carbohydrates, fats, or proteins and metabolic water production also varies dramatically (Schmidt-Nielsen and Adolph 1964; Frank 1988; Orr et al. 2015). Oxidation of carbohydrates, for example, yields 0.60 g of metabolic water per gram and fats yield 1.07 g per gram due to the greater oxygen requirements of lipid metabolism, while proteins yield the least metabolic water (0.41 g) of any macronutrient and even require water loss for excretion of nitrogenous waste (Mellanby 1942; Davidson and Passmore 1963; Schmidt-Nielsen and Adolph 1964).

In desert environments, where extrinsic water is limited, both preformed (dietary) water and endogenous water production are key to survival. Dietary intake represents the raw material used to produce metabolic water. Food catabolism then results in some water loss through respiration, urination, and fecal production (Schmidt-Nielsen 1975). Specific dietary composition is often unknown and highly variable, particularly for wild animals. However, energy potential (Pyke et al. 1977; Pyke 1984), nutrition, and capacity for water production (Schmidt-Nielsen and Adolph 1964) all depend on diet composition. In deserts, vegetation, seeds, and insects comprise the bulk of rodent diets (Reichman 1975) with each resource varying in composition of fat, protein, and carbohydrate as well as seasonal availability (Wolf and del Rio 2003; Orr et al. 2015). In response to natural variation in resource availability, animals may be forced to make dietary decisions that have a substantial effect on their internal water economy.

Mice in the genus Peromyscus have the widest distribution of any North American mammal and are considered model organisms in evolutionary biology due to their unparalleled habitat diversity (Bedford and Hoekstra 2015), extensive genomic resources (Tigano et al. 2020; Colella et al. 2021b), abundance of historical and contemporary collections in natural history museums (Pergams and Lawler 2009; Pardi et al. 2020), and ability to live and breed under laboratory conditions (Crossland et al. 2014). The desert specialist Cactus Mouse, Peromyscus eremicus, is endemic to the southwestern United States and exhibits behavioral (Murie 1961; Veal and Caire 1979), physiological (Macmillen 1965; Kordonowy et al. 2017; Colella et al. 2021a), and molecular (MacManes 2017; Tigano et al. 2020) adaptations to arid environments, making it an interesting natural experimental model to examine mechanisms of adaptation to warmer, drier environments. Peromyscus eremicus are omnivorous and opportunistic in their diet, utilizing seeds, arthropods, and green vegetation seasonally (Bradley and Mauer 1973; Meserve 1976) but shifting to consuming cactus seeds and/or fruit pulp during summer months (Orr et al. 2015).

We examine the effect of dietary fat content on the metabolic physiology of the desert-adapted Cactus Mouse to understand the physiological consequences of food availability and dietary variation. To accomplish this, we extend previously characterized circadian metabolic patterns for males and females (Colella et al. 2021a) to estimate rates of metabolism and water loss in animals fed an experimental diet low in fat—but comparable in terms of energy composition and levels of macronutrients—to those on a standard laboratory diet. This experimental framework allows us to test the relationships between dietary fats, energy expenditure (EE), and water balance, which have direct connections to fitness.

Materials and methods

Animal care and experimental model

We worked with a captive colony of cactus mice maintained at the University of New Hampshire. Captive animal care procedures followed guidelines established by the American Society of Mammalogists (Sikes et al. 2016), American Veterinary Medical Association (Leary et al. 2013), and were approved by the University of New Hampshire Institutional Animal Care and Use Committee (IACUC, #210604).

Sexually mature, nonreproductive healthy adult mice (n = 28 males, n = 28 females) between 3 and 9 months of age and bred from wild-derived lines at the Peromyscus Genetic Stock Center at the University of South Carolina (Columbia, South Carolina) were used in this study. Mice were assigned at random to 1 of 2 feeding groups: a standard diet group (SD; LabDiet 5015*, 26.10% fat, 19.752% protein, 54.15% carbohydrates, energy 15.02 kJ/g, food quotient [FQ, the theoretical respiratory quotient {RQ} produced by the diet based on macronutrient composition; Westerterp 1993] 0.89) or a low-fat diet group (LFD; Modified LabDiet 5015 [5G0Z], 6.6% fat, 22.8% protein, 70.6% carbohydrates, energy 14.31 kJ/g, FQ 0.92). All food was stored in the desert chamber to control for water content in the diets. Animals were housed individually in 9.5-L animal chambers inside a larger, environmentally controlled room built to simulate temperature, humidity, and photoperiodic conditions of the Sonoran Desert (Kordonowy et al. 2017; Colella et al. 2021a). Each animal chamber contained dried cellulose-based bedding and animals were provided food and water ad libitum. Mice were acclimated to the LFD for 1 month and to experimental cages for 24 h prior to the beginning of metabolic measurements. Animals were weighed to the nearest 10th of a gram on a digital scale before being housed in an experimental chamber for 4 days during metabolic data collection. Animals were weighed again at the end of the experiment.

Environmental temperatures and relative humidity (RH) followed a natural, diurnal pattern. Temperature and RH were fixed at 32 °C and 10% RH during the light phase for 11 h (06:00 to 20:00) and dropped to 24 °C and increased to 25% RH during a 1-h transition starting at 20:00. Temperature and RH then remained constant during the dark phase for 7 h, before temperature increased again and RH fell to return to light phase conditions during a 3-h transition that started at 06:00. Photoperiod was regular with 16 h of light and 8 h of dark. See Fig. 1A for a visual representation of environmental conditions.

Fig. 1.

Seventy-two hours of environmental (A) and respirometry (B to D) data split by sex for 28 adult males and 28 adult females Peromyscus eremicus plotted over a 24-h window to display circadian patterns for each group. Organismal responses to standard diet are in gray and low-fat diet in blue. Vertical blocks shaded in gray indicate the dark phase when animals are active, and unshaded blocks indicate light phase when animals are inactive. (A) Room temperature (°C) is indicated by a solid red line. (B) 24-hour measurements of EE (kJ h⁻¹), (C) WLR (H₂O g h⁻¹), and (D) RQ, for females (F, left) and males (M, right). Statistical significance for diet in the 24 h GAMMs is denoted with 1 or more · along the y axis label. Statistical significance for diet from GAMMs on the subsetted light and dark data is denoted with 1 or more * along the top of the plots. The 2 symbols (*/·) signify the same significance values (*P ≤ 0.05, ***P ≤ 0.001) “ns” indicates nonsignificance.

Metabolic phenotyping

Metabolic phenotyping data were collected starting at 12:00 using a previously described (Colella et al. 2021a) commercially available field metabolic system (FMS) purchased from Sable Systems International (SSI). Physiological data were collected following the methods outlined in Whitfield et al. (2015) and Colella et al. (2021a). Oxygen consumption (VO₂ mL min⁻¹), carbon dioxide production (VCO₂ mL min⁻¹), and water vapor pressure (WVP, kPa) lost via urine, feces, cutaneous, and respiratory pathways were measured continually during 72-h trials. The experiment was repeated twice for each sex under each diet (SD and LFD), resulting in 3-day experiments performed sequentially for 8 batches of 7 mice (and 1 empty, baseline chamber) per 3-day experiment. Batches alternated between females and males.

Airstreams were pulled from flow-through chambers at a constant rate of 1600 mL min⁻¹ (96 L h⁻¹) by an SSI SS-4 subsampler pump (1 for each chamber), multiplexed through an SSI MUXSCAN, subsampled at 250 mL min⁻¹ into an FMS where WVP, CO₂, and O₂ were analyzed with no scrubbing. The FMS was zeroed, and sensors were spanned between each 72-h experiment following the SSI Instrument Settings and Calibration manual using dry gases of known CO₂ and O₂ concentrations.

To generate accurate individual measurements, chambers were measured in a pseudorandom order, alternating between a baseline (empty) chamber measurement for 120 s and a random animal chamber measurement for 120 s, measuring each chamber approximately twice every hour for 3 days. Using methods described in Lighton (2018), we used the most stable 50% of each 120-s period to produce a single averaged value for the rate of O₂ consumption (mL min⁻¹), rate of CO₂ production (mL min⁻¹), and water loss rate (WLR, mg h⁻¹) per measurement window. Lag time between analyzers was corrected so that measurements represent the same time periods (Lighton 2018), resulting in a highly repeatable set of measurements. VO₂ and VCO₂ were calculated using equations 10.5 and 10.6, respectively, from Lighton (2018). RQ was calculated as the ratio of VCO₂ to VO₂ (Lighton 2018). EE (kJ h⁻¹) was calculated as in Lighton (2018, eq. 9.15). Response variable measurements that were more than 3 standard deviations away from the mean, which often represented times when animal care staff entered the room, were considered outliers and were removed from downstream analysis.

Electrolytes

At 12:00 Eastern Standard Time (EST), immediately following the conclusion of the experiment, animals were weighed and euthanized with an overdose of inhaled isoflurane follow by decapitation. Within 1 min of death, 120 µL of trunk blood was collected for serum electrolyte measurement using an Abaxis i-STAT Alinity machine and CHEM8+ cartridges (Abbott Point of Care Inc., Abbott Park, Illinois), and the Abaxis calibrated automatically with each cartridge run. We used CHEM8+ cartridges to measure the concentration of sodium (Na), potassium (K), creatinine (Cr), blood urea nitrogen (BUN), hematocrit (Hct), glucose (Glu), and ionized calcium (iCa), which are expected to vary in response to hydration status and renal function. Finally, using Na, Glu, and BUN, we calculated osmolality using the formula in Rasouli (2016). After assessing for normality, we used a Student’s 2-tailed t-test in R (stats::t.test) to test for significant (P < 0.05) differences in weight and electrolytes within and between the treatments and sexes for each experimental group.

Statistical analysis

All statistical analyses were conducted in R v 4.0.3 (R Core Team 2020) unless otherwise specified. To perform analyses on the shape of the temporally variable curves, we modeled the effect of time in days, diurnal time, EE, RQ, and/or WLR on EE, RQ, and WLR using generalized additive mixed models (GAMMs) with the gamm function in the “mgcv” R package (Lin and Zhang 1999; Wood 2017). To assure independence, we used experimental batches and mice within each experimental batch as random effects. The R function gamm then simultaneously estimated variances of these random effects as well as the treatment effects, thus implicitly adjusting treatment and covariate effects for the effect of correlated, nested groups of measurements. As a result, treatments do not explain differences between individual mice, but rather the average difference between groups of mice, because we were interested in the diet treatment and the effect of sex as well as in the response over the full 72-h period and the diurnal cycle. Our models included 2 fixed effects, diet and sex, and 4 nonlinear smoothing regression terms: time in days as a nonlinear trend; diurnal cycle as a circular spline function; and 2 of the 3 respirometry response variables (EE, RQ, and/or WLR). Using the effective degrees of freedom (edf), we were able to determine the degree of nonlinearity of each curve (Wood 2017). To test for physiological differences between mice fed each diet during the light and dark phases, we subsetted data collected during the light phase and dark phase for each response variable and ran GAMMs as described above.

To test for differences in the total volume of water lost by each animal over the course of the experiment, rather than the shape of the curve, we estimated the amount of water lost per individual for each hour of the experiment and summed those values. These values were then used as the response variable in an analysis of covariance (ANCOVA; car::anova; Fox and Weisberg 2019). The model consisted of 2 independent variables (sex and diet) and the interaction between them. Weight at the start of the experiment was used as the covariate. Pairwise comparisons were made using the Tukey separation of adjusted means test (emmeans::emmeans and emmeans::contrast; Lenth, 2023).

Results

Data from 56 sexually mature (but nonreproductive) male and female mice were collected as described above. No health issues were detected by veterinary staff during the duration of the experiment, nor were there significant differences in weight between males and females—mean of 22.20 g (range: 15.40 to 29.10 g) versus 21.2 g (range: 16.91 to 29.34 g), or between diet groups.

Metabolic phenotyping

In total, we recorded 8,125 120-s intervals: 4,091 female and 4,034 male. We measured metabolic variables for 4,105 120-s intervals under the LFD: 2,053 female and 2,052 male measurements. For the SD treatment, we recorded 4,020 total observations (2,038 female and 1,982 male). After removing outliers, we retained a total of 7,908 measurements: 3,945 female measurements (1,977 SD and 1,968 LFD); and 3,963 male measurements (1,945 SD and 2,018 LFD). Peromyscus eremicus show physiological patterns that are in phase with photoperiod and room temperature for both diets (Fig. 1), as described in Fox and Weisberg (2019), and Colella et al. (2021a).

Energy expenditure

Consistent with previous work (Colella et al. 2021a), males and females show diurnal patterning of EE on both diets, with the highest EE occurring during the dark phase when environmental temperature is lowest and animals are active. The lowest EE occurred during the light phase, when environmental temperature was greatest and animals were inactive (Fig. 1B). Using a GAMM, we modeled EE with diet and sex as fixed effects, as well as other predictors such as day, time, WLR, and RQ and found that neither sex nor diet influenced EE (diet P = 0.56, sex P = 0.18; Supplementary Data SD1). When analyzing light and dark phases separately, there were no differences in EE between males and females nor between the 2 diets (light phase, diet P = 0.550, sex P = 0.293; dark phase, diet P = 0.571, sex P = 0.182; Supplementary Data SD2 and SD3). There was a weakly nonlinear but significant relationship between EE and light phase (Supplementary Data SD1 and SD4; edf = 1.98, P < 0.01). Additionally, the relationship between time and EE was statistically significant but nonlinear (Supplementary Data SD1 and SD4; edf = 7.98, P < 0.01).

Water loss rate

Both treatment groups, SD and LFD, and both sexes exhibited diurnal patterning of WLR and had the highest WLR during the light phase when environmental temperature was highest, even though animals were inactive then (Fig. 1C; Supplementary Data SD1 and SD4; edf = 7.95, P < 0.01). Peak loss occurred between 09:00 and 11:00. By fitting a 24-h time-continuous GAMM (Supplementary Data SD4), we showed that both diet and sex had a significant effect on the WLR (diet P < 0.01, sex P < 0.01; Supplementary Data SD1), with both males and females fed an LFD losing more water than those fed the SD. When modeling water loss during the light phase only, we showed that both diet and sex had a significant effect on WLR (diet P < 0.01, sex P < 0.01; Supplementary Data SD2). During the dark phase when animals are active, WLR was significantly different between males and females (P = 0.015; Supplementary Data SD3), but WLR was not different within sex between different diets (P = 0.797; Supplementary Data SD3). The relationship between WLR and experimental day was weakly nonlinear, but significant (Supplementary Data SD1 and SD4; edf = 1.99, P < 0.01).

Both diet and sex affected the amount of water lost over the course of the experiment when controlling for individual body weight (P < 0.01 and P = 0.017; Supplementary Data S5D), but the interaction term was not significant (P = 0.633; Supplementary Data SD5). Further investigation into the sex:diet interaction showed significant differences in both the female (P = 0.043; Supplementary Data SD5) and male (P = 0.005, Supplementary Data SD5) diet comparison.

Respiratory quotient

RQ showed circadian rhythms for both diets and sexes with the highest RQ occurring during the light phase and the lowest during the dark phase (Colella et al. 2021a; Fig. 1D; Supplementary Data SD6). By fitting a 24-h GAMM, we found no significant difference in RQ between sexes or diets (sex P = 0.83, diet P = 0.52; Supplementary Data SD1) but we did find that day and time were significant (P < 0.01, P < 0.01, respectively). RQ had a complex relationship with time of day (Supplementary Data SD1 and SD4; edf = 7.69) and experimental day had a weak nonlinear relationship with RQ (Supplementary Data SD1 and SD4; edf = 1.59). During the light phase, RQ rose above FQ, with females fed the SD having the highest RQ at the start of the light phase and then stabilizing at an RQ > 1 around 14:00. RQ did not differ by sex nor diet (sex P = 0.740, diet P = 0.543; Supplementary Data SD2). When modeling RQ during the dark phase, values were comparable to FQ of the respective diet for both sexes (FQ LFD = 0.92, FQ SD = 0.89) and RQ was significantly different between the diets (P = 0.017; Supplementary Data SD3) but not between males and females (P = 0.739; Supplementary Data SD3).

Electrolytes

Several serum electrolyte values were significantly different between diets for males and females (Na P = 0.22 and P = 0.01; K P = 0.02 and P = 0.01; BUN P = 0.03 and 0.71; Hct P = 0.04 and 0.02; iCa P = 0.001 and 0.001; osmolality P = 0.01 and 0.05; Fig. 2). When comparing males and females on the same diet, Na (P = 0.02) and osmolality (P = 0.03) differed significantly for mice on the SD and no electrolyte measurements were different between males and females on the LFD.

Fig. 2.

Violin plots showing the distribution of serum electrolyte measurements: Na = sodium (mmol/L), K = potassium (mmol/L), BUN = blood urea nitrogen (mmol/L), Hct = hematocrit (% PCV), iCa = ionized calcium (mmol/L), Glu = glucose (mmol/L), and osmolality (mmol/L) for female (F, left) and male (M, right) Peromyscus eremicus fed standard (gray) and low-fat (blue) diets. Observations (28 adult females and 28 adult males, n = 14 of each treatment, total n = 56) are represented by black dots. P-values from pairwise t-tests are reported above the brackets.

Discussion

We explored the effect of dietary fat intake on water balance in a desert-adapted rodent and found that mice fed a diet lower in fat lost more water than mice fed a diet higher in fat. In environments where water scarcity is extreme, even small differences in water balance could have substantial effects on cognitive function and physical performance, and therefore survival and reproductive success (Boogert et al. 2018). Our results help clarify the physiological consequences of food availability and dietary choices.

Energy expenditure

Not all diets provide animals with the same amount of energy per unit mass (Withers 1982). Different macronutrients have different energy potentials (Sánchez-Peña et al. 2017). Diet composition influences EE (Acheson 1990; Westerterp 2004; Secor 2009; Beale et al. 2018), as do other intrinsic (i.e., body mass, temperature, evolutionary history, reproductive stage, activity, or energy intake) and extrinsic (i.e., habitat, climate—including environmental temperature, or social environment; Speakman 1997) variables. Many plants native to the Sonoran Desert have a fat content comparable to that of the LFD used in this study—5.49% to 9.88% for leaf tissues and branches (McArthur 1994); 0.90% to 3.60% for roots and seeds (Castle et al. 2020). In nature, dietary switches tend to coincide with seasonal changes that affect resource quantity and/or quality (Noble et al. 2019). Previous studies have recorded cactus mice shifting diet seasonally, consuming arthropods during the winter (Hope and Parmenter 2007) and transitioning to the consumption of cactus seeds and/or fruits during the summer (Orr et al. 2015; Hope and Parmenter 2007).

In the current study, both diets had equivalent amounts of energy per gram of weight. We designed the experiment to control for several intrinsic and extrinsic factors that represent sources of variation in EE, but not in individual activity. We showed that experimental manipulation of diet did not result in differences in EE when food was provided ad libitum. EE and activity are tightly coupled (Garland et al. 2011; Kaiyala et al. 2012). Therefore, we expect at most limited differences in activity given limited differences in EE. For both diets, animals likely ate enough to satisfy energy needs, a hypothesis further supported by the fact that there was no net change in weight over the course of the experiment. Consistent with other studies, we recorded higher EE during the dark phase of the 24-h cycle when nocturnal animals were active (Colella et al. 2021a), further supporting that metabolic rate is complex and may be modulated by the environment, activity, or other extrinsic/intrinsic factors.

Water loss rate

Thermoregulation and water loss

Evaporative water loss is one way for organisms to cool themselves (e.g., via latent heat of vaporization), which results in a decrease in body temperature and loss of body mass (Porter and Gates 1969). While body temperature was not measured as a part of this study, numerous studies have shown that desert animals reduce respiratory water loss compared to non-desert-adapted animals (Schmidt-Nielsen and Schmidt-Nielsen 1951; Hart 1971; MacMillen 1972) which could be an adaptive physiological mechanism for maintaining water balance in hot, arid desert environments. Indeed, a fine balance must exist between using water for thermoregulation, for other critical metabolic processes, and dehydration. Exaggerated use of water for thermoregulation may be harmful when water is limited. Instead, other less water-intensive mechanisms of heat loss such as conduction or behavioral changes including estivation and nocturnality may be used to maintain a homeostatic body temperature but were not quantified in this study.

Despite the consequences of limited water supply, desert Peromyscus have significant capacity for evaporative cooling for thermoregulation (Ramirez et al. 2022). In our study, we found that animals fed the LFD lost more water than animals fed the SD (Fig. 1C). While that loss may be related to differences in thermoregulatory performance, it also affects water balance. Are these patterns suggestive of differences in physiological performance? Is the consumption of a diet lower in fat related to a reduction in physiological performance and, consequently, fitness? While the current experiment cannot definitively answer these questions, there is some evidence (e.g., electrolytes, water loss) to suggest that physiological performance is impaired in animals consuming the LFD and that this performance, particularly when water is scarce, may have fitness consequences.

Potential mechanisms for increased water loss on LFD

Across all comparisons, there is a strong and statistically significant relationship between WLR and dietary fat content, with individuals fed the LFD losing more water. While identifying the potential mechanisms underlying these findings will require future experiments, several can be ruled out. First, the maintenance of water balance is a process regulated by a cascade of hormones including arginine vasopressin and the renin–angiotensin–aldosterone system pathway (Aisenbrey et al. 1981; Greenleaf 1992). While some hormones are lipid-based (e.g., aldosterone) and their production and secretion are linked to dietary composition (i.e., sex hormones and severe low body fat), the degree to which dietary fats are reduced for mice fed the LFD is not at the level at which hormonal imbalances are seen (Staszkiewicz et al. 2007; Abdel-Rahman 2010). Second, while there is a known relationship between ketogenic diets and dehydration (Wheless 2001; Freeman et al. 2006), those diets are associated with enhanced diuresis. We observed the opposite with animals fed the SD; they lost less water than animals fed the LFD. Third, diets higher in protein require more water for the removal of nitrogenous wastes (Calder and Braun 1983), but higher protein does not account for the observations herein. To approximate equivalent energy availability, there is a small difference in the amount of dietary protein in each dietary treatment (18.9% in SD vs. 19.5% in LFD), however, that small difference is unlikely to result in the observed differences in water loss and hydration status.

While food and water intake were not measured directly, a potential mechanism for the observed differences could be that animals fed the LFD simply eat more than those fed a diet higher in fat. While the diets are closely matched in terms of available energy per unit mass, the LFD is slightly less energy dense. Additionally, it is possible that there are substantial differences in the nonnutritive fraction of the diet that should be explored in future studies. These differences suggest that animals fed the LFD may be ingesting more food mass to maintain nutritional homeostasis, which may also increase their water intake (Bachmanov et al. 2002). While we have no evidence for this, an increased rate of consumption would potentially result in an increased rate of urine production and the production of water-containing fecal material and urine. Given that animals fed the LFD seem to be more dehydrated than animals fed the SD (increased Na, osmolality, Hct; Fig. 2), this hypothesis requires that animals fed the LFD consume more water, but not enough to satisfy their physiological needs. Given that the animals have free access to water, an unmet need is irreconcilable without further explanation.

Another intriguing hypothesis linked to the concept of unmet hydrational needs has to do with thirst pathways—thirst is under tight neural control. Osmo- and baroreceptors in the vasculature are responsible for monitoring water balance and, when triggered, set into motion a hormonal cascade that simultaneously enhances renal water retention and stimulates thirst (Thornton 2010; Leib et al. 2016). While former aspects of the response are crucial to many physiological processes, the latter response (thirst) could be responsible for the observations described above. While purely a hypothesis, there are several examples of species losing aspects of their sensory systems via natural selection, including loss of sight (Gore et al. 2018), decreases in certain taste sensitivities or loss of taste entirely (Jiang et al. 2012), and loss of specific olfactory functions (Kishida et al. 2015). For animals in dry deserts, where water is scarce, could well-developed pathways leading to thirst be similarly changed? While further research is required, perhaps via a comparative genomics study, this hypothesis could explain why animals with free access to water remain significantly dehydrated.

Respiratory quotient

RQ, the ratio of CO₂ produced to O₂ consumed, can inform differential fuel utilization. RQ values typically fall between 0.7 and 1.0, with catabolism of fats yielding 0.7, carbohydrates 1.0, and proteins at intermediate values (Kleiber 1975). RQ values can exceed 1.0 when anabolism of fat exceeds catabolism, i.e., de novo lipogenesis (DNL; Benedict 1937; Abreu-Vieira et al. 2015; Levin et al. 2017) or during anaerobic exercise (Whipp 2007; Zagatto et al. 2012). While carbohydrates, fats, or proteins are all burned as fuel, if present in excess, they can also be converted to fat which can be stored for future use (Barboza et al. 2009). DNL is 1 mechanism for converting nonfat energy into a form that can be stored. The conversion of glucose to fats prior to entry into the citric acid cycle is exothermic and results in an energetic cost, representing a metabolically inefficient use of dietary substrate in cold habitats (Solinas et al. 2015), but a potentially efficient mechanism of heat regulation in hot deserts. RQ was greater than 1.0 (with recordings as high as ~1.6; Supplementary Data SD1) during the inactive period in animals on both diets suggesting that DNL is not limited by fat intake in this species. Here and in previous experiments on P. eremicus (Colella et al. 2021a), RQ is only greater than 1.0 during the light phase when animals are inactive, which strongly suggests that DNL is occurring. Further, fat storage could be a way for desert mammals to maintain water homeostasis during uncertain water availability because fat oxidation releases metabolic water (Mellanby 1942; Davidson and Passmore 1963; Schmidt-Nielsen and Adolph 1964). Given this, the conversion of other fuel sources to fats may serve as an important reservoir of metabolic water. Previous studies have shown that DNL can occur under high-carbohydrate and high-fat diets (Strable and Ntambi 2010) which is supported by our results because both groups had an RQ greater than 1.0 during the light phase and RQ values at or below FQ during the dark phase, which indicates a shift to the supplied dietary substrate as the main oxidative fuel during cooler, dark periods.

Electrolytes

Electrolytes provide information on the overall metabolic state, renal function, and other core physiological functions of an organism (Kutscher 1968) including hydration status (Cheuvront et al. 2010) or disease state. Therefore, to further understand the effects of diet on physiological function of small mammals in desert environments, we examined serum electrolytes. We established diet-specific differences for serum Na, K, Cr, BUN, Hct, iCa, and osmolality values for P. eremicus. Notably, synthetic markers of pathological renal impairment, BUN and Cr, were not significantly different between dietary treatments. Previous studies have shown that the electrolyte–water balance is frequently unaffected when desert rodents are water-deprived for long periods of time (Heimeier et al. 2004; Heimeier and Donald 2006; Boumansour et al. 2021). However, other desert rodent studies have demonstrated evidence of dehydration in the absence of exogenous water without renal impairment (Kordonowy et al. 2017; Boumansour et al. 2021). Therefore, even with a significant amount of water lost through respiration, dietary fat content does not significantly impair the kidney function of P. eremicus when free water is available.

Interestingly, other key electrolytes that are sensitive to hydration status differed significantly between the dietary treatment groups (Fig. 2). Serum Na, osmolality, and Hct are expected to be elevated in clinical dehydration and were elevated in animals fed the LFD. Although some contrasts were not statistically significant, measurements were consistently higher for both males and females, a strong sign of differences in hydration between treatment groups. These differences are particularly interesting because while animals fed the LFD had higher rates of water loss than those fed the SD, all animals had access to water ad libitum. Elevated serum Na, osmolality, and Hct in the context of normal renal function are evidence of dehydration, but dehydration in neuro-intact animals with free access to water is unusual. While dietary fats may be responsible for changes in solute balance (Friedman et al. 2012), typically the relationship is in the opposite direction to what we found (Abdel-Rahman 2010). Again, 1 possible hypothesis is that there may be an impairment of the physiological mechanisms related to thirst. Thirst is a response to changes in blood chemistry, i.e., increase in osmolality (Gilman 1937; Wolf 1950; Heimeier et al. 2004) or volume (Fitzsimons 1961; Stricker 1966) both of which are monitored by the lamina terminalis circumventricular organs (Bourque 2008; Zimmerman et al. 2017). Activation of these brain regions motivates an organism to find water. Previous studies have found that the expression of 2 genes, Rxfp1 and Pdyn, in relevant cell types correlates with neural activation under osmotic and hypovolemic thirst in lab mice (Pool et al. 2020). Further studies should examine these genes and others in desert organisms to understand the evolutionary basis of thirst pathways. Regardless of the process, elevated serum Na may have important and complex physiological consequences for, e.g., elevated blood pressure and increased fluid retention as a result of elevated Na levels, both having obvious implications for survival in desert environments and deserve future study.

In conclusion, we showed that dietary fat is important for water regulation in cactus mice through the analysis of interacting physiological traits: (1) high rates of WLR during the warmer, drier light phase and (2) a difference in serum electrolyte values indicating dehydration. Given that fat catabolism yields more water and energy compared to carbohydrates or proteins, our results suggest that an LFD could limit the capacity of desert animals to tolerate limited access to free water, as is common in arid environments. The identification of short-term, physiological responses of P. eremicus associated with dietary fat composition demonstrates that adaptations evolved over long evolutionary timescales will affect survival during the accelerated timescales of climate change. In light of global climate change and increased desertification, investigating the range and mechanisms of plastic responses employed by desert-adapted species could provide insights into physiological responses to increasingly erratic climate.

Supplementary data

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1. Generalized additive mixed models (GAMMs) 24-h results. Statistical models and results for (A) water loss rate (WLR, H₂O g h⁻¹), (B) energy expenditure (EE, kJ h⁻¹), and (C) respiratory quotient (RQ).

Supplementary Data SD2. Generalized additive mixed models (GAMMs) results for only data collected during the light phase. Statistical models and results for (A) water loss rate (WLR, H₂O g h ⁻¹), (B) energy expenditure (EE, kJ h ⁻¹), and (C) respiratory quotient (RQ).

Supplementary Data SD3. Generalized additive mixed models (GAMMs) results for only data collected during the dark phase. Statistical models and results for (A) water loss rate (WLR, H₂O g h ⁻¹), (B) energy expenditure (EE, kJ h ⁻¹), and (C) respiratory quotient (RQ).

Supplementary Data SD4. Contributions of model terms to energy expenditure (EE, kJ h⁻¹, A to D), water loss rate (WLR, g h⁻¹, E to H), and respiratory quotient (RQ, I to L) for female and male Peromyscus eremicus across 2 diet groups (standard diet and low-fat diet) in general additive mixed models (GAMMs). The smoothing curves for each response variable included 2 fixed effects: diet (low-fat vs. standard) and sex, 2 random effects: mouse identification number and date of data collection, and 4 regression terms: time in days, diurnal cycle, and 2 of the 3 respirometry response variables (EE, RQ, and/or EWL) as regression terms. For each graph, the y axis is the effect of the x axis on the respirometry response variable estimated by a multivariable GAMM. Blue-shaded areas are 95% confidence intervals; black points are residuals.

Supplementary Data SD5. ANCOVA results for total water loss.

Supplementary Data SD6. Respirometry means for split by sex for 28 adult males and 28 adult females, diet treatment, and time of day (t1 = 3 h temperature increase from 24 °C to 32 °C and relative humidity [RH] decrease from 25% to 10%, day = 11 h fixed temperature at 32 °C and 10% RH, t2 = 1 h temperature drop from 32 °C to 24 °C and RH increase from 10% to 25%, night = 7 h fixed temperature at 24 °C and 25% RH). Respirometry measurements are energy expenditure (EE, kJ h⁻¹), water loss rate (WLR, g h⁻¹), and respiratory quotient (RQ).

Author contributions

Conceptualization, MDM; methodology, DMB, MDM; formal analysis, DMB, EL; investigation, DMB, JPC; resources, MDM; writing—original draft, DMB; writing—review & editing, DMB, JPC, EL, MDM; visualization, DMB, EL; supervision, MDM; project administration, MDM; funding acquisition, MDM.

Funding

This work was supported by the National Institute of Health National Institute of General Medical Sciences (R35 GM128843 to MDM).

Conflict of interest

None declared.

Data availability

Raw ExpeData (SSI) files are available through Zenodo: https://zenodo.org/record/6422231#.YlXSD9PML0o. Macro processing files, raw and processed respirometry data, and cage sampling scheme files are also available on Zenodo. All R scripts used in this project are available through GitHub at: https://github.com/DaniBlumstein/Diet_paper.