Developmental delay in shivering limits thermogenic capacity in juvenile high-altitude deer mice (Peromyscus maniculatus)

At birth, altricial mammals lack the capacity to regulate body temperature through metabolic heat production (Pembrey, 1895). Instead, they develop homeothermic endothermy over the early postnatal period as the primary thermo-effector organs and their corresponding regulatory systems mature. This dynamic life stage is characterized first by the rapid maturation of brown adipose tissue (BAT) and non-shivering thermogenesis (NST; Lagerspetz, 1966). In contrast, the skeletal muscles of altricial mammals are extremely immature at birth (Dubowitz, 1963). In these species, it takes several weeks before muscle phenotype is fully established (e.g. Goldspink and Ward, 1979; Adams et al., 1999; Gokhin et al., 2008; Agbulut et al., 2003). The relatively slow postnatal maturation of skeletal muscle corresponds with a gradual onset of the capacity to engage in shivering (Arajamma and Lagerspetz, 1978). For example, shivering was not observed in the white-footed mouse (Peromyscus leucopus) until approximately 2 weeks after birth (Hill, 1976).

In adult endotherms, a well-developed capacity for shivering thermogenesis is especially important in the unremitting low ambient temperatures characteristic of high-altitude (HA) ecosystems. This environment is particularly challenging for small obligate endotherms as their high rates of heat loss increase the demand for aerobic heat production in the face of low O₂ availability. As a result of these combined selective pressures, many birds and mammals native to high alpine regions have evolved highly vascularized and aerobic adult skeletal muscle phenotypes (León-Velarde et al., 1993; Hepple et al., 1998; Mathieu-Costello et al., 1998; Kayser et al., 1991; Sheafor, 2003; Scott et al., 2009; Lui et al., 2015; Lau et al., 2017). This high muscle aerobic capacity is accompanied by an elevated capacity for lipid oxidation in adult HA native deer mice (Peromyscus maniculatus; Cheviron et al., 2012, 2014; Lau et al., 2017), presumably to support high rates of shivering. These underlying adaptations have allowed HA deer mice to evolve a high thermogenic capacity (cold-induced V̇_O2,max), which contributes directly to fitness by improving survival (Hayes and O'Connor, 1999).

Given that the skeletal muscle of altricial mammals is immature at birth, the aerobic muscle fiber type, metabolic capacity and capillarity characteristic of this adaptive HA muscle phenotype may be absent early in postnatal development of HA deer mice. It is unclear at what age HA native deer mice develop this phenotype which is so critical for adult performance. The early postnatal period, during which muscle phenotype is established, is characterized by extremely high mortality rates (50–95%) in wild rodent populations (e.g. Bendell, 1959; Howard, 1949). Thus, the physiological systems that develop during this period are likely subject to intense natural selection (Hill, 1983). It is unknown how natural selection in extreme environments, such as HA, may have altered the postnatal maturation of skeletal muscle and/or the capacity for shivering thermogenesis to allow juveniles to cope with the constant cold and hypoxia at high elevations. Given the importance of muscle phenotype for adult thermoregulation, the timing of the maturation of this system likely has major performance consequences for the thermoregulatory ability of juveniles at HA. Additionally, we have previously demonstrated that in the first 10 postnatal days, activation and regulation of NST in BAT is delayed in HA native deer mouse pups compared with lowlanders (Robertson et al., 2019). As a result, HA pups are unable to adequately thermoregulate during this early developmental window, inconsistent with their superior thermogenic capacity as adults (Cheviron et al., 2014). Given the delay in BAT maturation, HA juveniles may rely more heavily on skeletal muscle-based thermogenesis. Therefore, natural selection may have accelerated the development of this tissue at high elevations.

To assess whether HA adaptation has altered the developmental program of skeletal muscle and shivering thermogenesis, we used low-altitude (LA) and HA North American deer mice (P. maniculatus) and the closely related but strictly LA white-footed mice (P. leucopus) born and raised in common garden conditions at LA. We tested the hypothesis that adaptation to HA accelerates the maturation of skeletal muscle, allowing juveniles to induce shivering thermogenesis in the cold. If so, we predict that HA pups will utilize shivering earlier in development than their LA counterparts. We further predict that the thermoregulatory capacity of HA juveniles will only exceed that of LA mice once they have developed a more aerobic muscle phenotype.

Juvenile mice used in this study were the second generation (G₂) laboratory-born progeny of a captive breeding colony of HA and LA native deer mice, P. maniculatus, and white-footed mice, P. leucopus (Rafinesque 1818). In both 2013 and 2014, the HA wild-caught breeding stock (P. maniculatus rufinus) were trapped at the summit of Mount Evans, CO, USA (4350 m a.s.l.), and the LA natives (P. maniculatus nebrascensis and P. leucopus) were trapped at Nine Mile Prairie, NE, USA (430 m a.s.l.), as previously described (Cheviron et al., 2012). Wild-caught mice were transported to McMaster University, Canada (∼90 m a.s.l.), and bred within their respective populations under common garden conditions (24°C, 760 mmHg, 14 h:10 h light:dark cycle, with rodent chow and water ad libitum). First generation (G₁) mice were mated within their respective populations to produce the G₂ offspring used in this study (Robertson et al., 2019). All pups were weaned at postnatal day 21 (P21) and housed with their same-sex littermates post-weaning. Given that P. leucopus are closely related to P. maniculatus but are found exclusively at LA, we used this species as a LA outgroup (Velotta et al., 2018). All procedures were approved by the McMaster University Animal Research Ethics Board.

At postnatal days P0, P2, P4, P6, P8, P10, P14, P21 and P27, one pup per G₁ breeding pair was sampled by an overdose of isoflurane followed by cervical dislocation. The gastrocnemius muscle from one hindlimb was blunt dissected, weighed and freeze-clamped using two liquid N₂-cooled aluminium plates. We used the gastrocnemius as a representative muscle as previous studies have shown it is involved in shivering thermogenesis (Oufara et al., 1987) and that phenotypic differences between LA and HA adult mice correlate strongly with whole-animal thermogenic capacity (Cheviron et al., 2012, 2014). Muscle tissue was stored at −80°C for future molecular analysis. From the other hindlimb, the triceps surae (consisting of gastrocnemius, plantaris and soleus) was frozen in embedding medium (Cryomatrix, Thermo Scientific) for histological analysis.

We assessed developmental changes in the numerical density of aerobic muscle fibers and capillarity as previously described for adult deer mice (Lui et al., 2015; Mahalingam et al., 2017). Briefly, frozen triceps surae was sectioned transversely (10 µm) at −20°C using a cryostat (Leica Biosystems CM1860). Capillaries were identified by staining for alkaline phosphatase activity. Capillary area was quantified as previously described (Mathieu-Costello, 1987) using NIS-Elements Imaging Software v.4.30 (Laboratory Imaging, Prague, Czechia). Briefly, capillary areal density (CAD) was determined as the ratio of capillary area (μm²) to transverse muscle area (μm²). As it has previously been suggested that skeletal muscle capillaries of HA deer mice are more tortuous than those of LA deer mice (Lui et al., 2015; Scott et al., 2015), we used capillary length density (CLD) as a measure of tortuosity. CLD is the quotient of areal density and the transverse area of the smallest 10% of capillaries (Dawson et al., 2018). Aerobic muscle fibers were identified by staining using succinate dehydrogenase (SDH) activity (Lui et al., 2015). The numerical density of oxidative fibers was determined as the number of oxidative fibers divided by the total number of fibers. Images were randomized and analyzed blind using ImageJ software. Muscle fiber size was determined in LA P. leucopus and HA P. maniculatus pups using hematoxylin and eosin (H&E) staining. Briefly, slides were fixed in 95% ethanol then incubated in Gills II hematoxylin for 30 s and eosin for 5 s (Leica Biosystems).

The apparent V_max of citrate synthase (CS), lactate dehydrogenase (LDH) and β-hydroxyacyl-CoA dehydrogenase (HOAD) was assayed to assess capacity for the TCA cycle, anaerobic glycolysis and fatty acid oxidation, respectively, using assay conditions previously described (Cheviron et al., 2012; Lui et al., 2015; Lau et al., 2017).

At P14, P21 and P27, individual pups were used to determine basal metabolic rate (BMR), NST or maximum thermogenic capacity (cold-induced V̇_O2,max) by indirect calorimetry as described below.

Juveniles were removed from their nests and fasted for 4 h to achieve a post-absorptive state prior to BMR trials, after which time they were placed in respirometry chambers (475 ml) maintained within the thermoneutral zone (Hill, 1983) at 28°C using a Peltier Cabinet (Sable Systems, Las Vegas, NV, USA). Outside air was dried and stripped of CO₂ using ascarite and soda lime, before flowing through the 475 ml chambers at a rate of 1000 ml min⁻¹ using a mass flow controller (Sable Systems). Excurrent air was subsampled, dried with magnesium perchlorate and passed through CO₂ (CA-10A) and O₂ (FC-1A) analyzers (Sable Systems) for determination of oxygen consumption (V̇_O2). BMR was calculated with equation 3b from Withers (1977) using the average of the three lowest stable (5 min) O₂ traces over 2 h.

Mice were placed in metabolic chambers (475 ml) at 28°C and baseline V̇_O2 measurements were obtained after 45 min in normoxia as described above. Noradrenaline (norepinephrine) was then injected subcutaneously at a standardized dose (Wunder and Gettinger, 1996) in a total of 500 µl of saline, and mice were returned to the respirometry chamber for 1 h or until V̇_O2 returned to baseline. We determined in a preliminary study that any change in V̇_O2 due to subcutaneous injection of saline alone returned to baseline V̇_O2 after ∼10 min, twice as fast as with intraperitoneal injections. Thus, we determined maximal NST (NST_max) as the highest stable (5 min) V̇_O2 with subcutaneous noradrenaline injection after this time point. NST was then calculated as NST_max−BMR.

To determine maximal cold-induced V̇_O2, juveniles were placed in respirometry chambers at −5°C and either normoxic (20% O₂) or hypoxic (12% O₂) heliox (O₂ with He) was flowed through the chambers at a rate of 1000 ml min⁻¹. Subsampled air was dried before analysis for changes in O₂ and CO₂ (Fox Box Respiratory System, Sable Systems). Data acquisition was performed using LabChart software (ADI instruments) and cold-induced V̇_O2,max was calculated as the highest stable (10 s) V̇_O2 during a 10 min trial (Tate et al., 2017). Rectal body temperature was taken before and after each trial using a rectal probe (RET-4, Physitemp) to ensure that the trial was sufficient to induce hypothermia.

Shivering thermogenesis was estimated in normoxia as V̇_O2,max−(NST+BMR) (Wunder and Gettinger, 1996).

We used two-way repeated measures (RM) ANOVA to test the effect of population and age on mean pup mass, and one-way ANOVA to test the effect of population on growth rate. For these data, we used family as a replicate, using the mean of all offspring from a G₁ breeding pair as a single measurement. We used two-way ANOVA to test the effects of population and age on tissue mass, oxidative fiber numerical density, capillary area, enzyme activity and percentage V̇_O2,max. Percentage V̇_O2,max data were arcsine square-root transformed and values of 0% were assigned a value of 1/4X, where X=V̇_O2,max, prior to analysis. For measures of V̇_O2 (V̇_O2,max, NST, BMR), we used two-way ANCOVA to test the effects of population and age with body mass as a covariate. When significant interactions (P<0.05) were found between population and age, we used Holm–Šidák post hoc analysis to perform pairwise comparison between groups. Data are available from figshare (10.6084/m9.figshare.8979842).

We tracked the growth of deer mouse (P. maniculatus) and white-footed mouse (P. leucopus) pups over the first 30 days of postnatal development and found that P. leucopus were larger and grew faster than either population of P. maniculatus (age×population F_16,176=12.851, P<0.001; Fig. 1A). While P. maniculatus pups were being provisioned by their mothers, their size and growth rates were similar regardless of altitude ancestry; post-weaning (P21) growth rate accelerated in HA and LA P. leucopus pups (population F_2,22=14.487, P<0.001), resulting in a significant size differential between all three groups by P27 (Fig. 1, Table 1). Growth patterns of the gastrocnemius muscle also differed between populations (age×population F_12,184=6.684, P<0.001; Fig. 1B). For the first 2 weeks of development, the size of this hindlimb muscle was similar for all three groups. However, after P14, the gastrocnemius of HA pups diverged and was smaller than that of either LA group. This altitude difference in muscle size was further exacerbated when expressed relative to body mass, with HA pups showing disproportionately small muscles at P21 and P27 (age×population F_12,184=3.516, P<0.001; Fig. 1C). The relative size of the gastrocnemius muscle was identical in LA pups after P14, despite the large interspecies difference in body size.

At birth, muscle fibers of the hindlimb were small and undifferentiated in all groups. By P14, all pups showed a small number of oxidative fibers in these muscles as determined by SDH staining. However, the numerical density of oxidative fibers rapidly increased from P14 to P21 in all groups. Between P21 and P27, there was a further increase in oxidative fiber numerical density but only in HA juveniles (age×population F_4,47=3.298, P=0.018; Fig. 2). At P21 and P27, the muscle fibers of HA pups were generally smaller than those of LA P. leucopus (age×population F_7,75=3.434, P=0.003; Fig. 2B, inset), which is characteristic of a higher proportion of Type I and Type IIa fibers relative to the larger Type IIb fibers.

This change in aerobic capacity of the muscle was reflected in gastrocnemius enzyme activity. The V_max of CS, a marker of mitochondrial density, increased similarly in all groups from P14 to P21 (age F_2,90=8.173, P<0.001) and was higher in HA pups at P27 relative to both LA populations (population F_2,90=14.451, P<0.001; Fig. 3A). V_max of LDH also increased with age (age F_2,92=11.820, P<0.001) but was lower in HA pups than in LA P. maniculatus (population F_2,92=36.337, P<0.001; Fig. 3B). Interestingly, HOAD activity was generally consistent between populations and across ages. It was slightly lower in LA P. maniculatus at P14 relative to that of HA pups but increased by P21 and remained consistent from this point onwards (age×population F_4,90=4.170, P=0.004; Fig. 3C).

In addition to immature muscle fibers, the skeletal muscle of LA and HA pups was poorly vascularized at birth. We were unable to detect capillaries in the hindlimb muscles using alkaline phosphatase staining at P2 and found negligible vascularization at P8 regardless of altitude ancestry. By P14, the capillary area was similar between all groups. This level remained constant in LA pups for the remainder of development but increased further by P21 in HA pups such that capillary area in the hindlimb was 2- to 3-fold higher than that of LA pups for the remainder of development (age×population F_4,53=3.494, P=0.013; Fig. 4A). Capillary area was larger in HA pups, in part, because the capillaries were more tortuous as capillary length density followed an identical trend (age×population F_4,51=2.577, P=0.048; Fig. 4B).

We tracked the whole-animal thermogenic performance of deer mouse (P. maniculatus) and white-footed mouse (P. leucopus) pups over the age range where muscle phenotype diverged (P14–P27). We found that mass-corrected thermogenic capacity (cold-induced V̇_O2,max) did not change across the second 2 weeks of postnatal development in LA pups of either species. However, HA pups showed a different pattern with age (age×population F_4,47=10.756, P<0.001). At P14, the thermogenic capacity of HA pups was lower than that of either LA group but increased continuously over the next 2 weeks. By P21, the thermogenic capacity of HA pups was equivalent to that measured in LA native mice and exceeded that of LA mice by P27 (Fig. 5A).

The estimated proportional contribution of shivering thermogenesis to thermogenic capacity changed differentially with age and population (age×population F_4,48=28.93, P<0.001). For P. leucopus, the contribution of shivering thermogenesis remained constant at ∼30% V̇_O2,max, while in both LA and HA P. maniculatus the contribution of shivering thermogenesis increased with age to ∼55% V̇_O2,max at P27. However, at P14 the contribution of shivering thermogenesis in HA pups was dramatically lower than that in either LA P. maniculatus or P. leucopus (post hoc P<0.001), accounting for only 3.8% of V̇_O2,max in this population (Fig. 5B). In fact, in most HA individuals tested, there was no evidence of shivering thermogenesis measured at P14, with combined BMR and NST meeting or exceeding cold-induced V̇_O2,max. Neither absolute NST nor BMR changed with age or population (Table 2). The intrascapular BAT (iBAT) grew between P14 and weaning (P21) and then showed a post-weaning decline in size (P27) in all populations (age F_2,52=11.050, P<0.001; data not shown). However, this change was not proportional to body size (age×population F_4,52=2.760, P=0.036). At P14, LA P. maniculatus had a higher proportion of iBAT than either P. leucopus or HA P. maniculatus. However, from P21 onwards there was no difference between the P. maniculatus populations regardless of altitude ancestry (Table 2).

Finally, we measured cold-induced V̇_O2 under hypoxia (12% O₂) to determine the extent to which thermogenic capacity during postnatal development was limited by O₂ availability. Generally, both HA and LA P. maniculatus had a greater hypoxic V̇_O2,max than LA P. leucopus (population F_2,25=16.065, P<0.001; Fig. 6). Hypoxic V̇_O2,max decreased slightly with age (age F_1,25=46.085, P<0.001), which was driven by statistically non-significant decreases in both LA populations (P. maniculatus P=0.089, P. leucopus P=0.17) but not HA P. maniculatus.

The main objective of this study was to determine whether altitude ancestry influences the postnatal maturation of skeletal muscle and the onset of shivering thermogenesis in deer mice. In contrast to our predictions, we found that the postnatal maturation of the lower hindlimb muscles and the onset of shivering thermogenesis were not accelerated in HA deer mice. In fact, HA mice showed a substantial delay in the developmental onset of shivering and a lower overall capacity for thermogenesis early in postnatal development, compared with LA conspecifics and LA P. leucopus. It was not until P21 that HA deer mice developed the capacity to shiver, and not until P27 that their skeletal muscle matured to establish the specialized aerobic phenotype characteristic of HA adults. Thermogenic capacity in HA juveniles surpassed that of lowlanders only after this more aerobic muscle phenotype was fully developed, and the capacity for shivering increased. To our knowledge, this is the first study to show that the postnatal maturation of skeletal muscle and of shivering thermogenesis is altered with adaptation to HA.

Skeletal muscles of altricial mammals, such as mice, fully mature and assume their adult phenotype sometime in the first weeks after birth (Dubowitz, 1963). However, the timing of key steps in muscle postnatal development is poorly understood in all but a few species (White et al., 2010; Schiafino and Reggiani, 2011), and to our knowledge has never been studied in relation to adaptive variation in wild populations. We found that at birth, regardless of altitude ancestry, the lower hindlimb muscles of deer mice were small and underdeveloped, consistent with data from laboratory mice (Wirtz et al., 1983). Muscles in both LA and HA mice grew proportionately with increases in body size over the first 10 days of development, after which muscle growth accelerated. Postnatal muscle growth in mice is generally attributed to hypertrophy of muscle fibers (Gokhin et al., 2008; White et al., 2010) and we found that muscle fiber size increased significantly in both LA and HA mice over the first 27 days after birth. However, later in development (P21–P27), HA juveniles had smaller muscle fibers compared with LA mice. This may be indicative of a higher proportion of oxidative fibers, which tend to be smaller than glycolytic fibers, an adaptation thought to minimize O₂ diffusion distance from the capillaries to the mitochondria (van Wessel et al., 2010). Overall, these smaller fibers likely contributed to smaller gastrocnemius muscles in HA juveniles after P21, even after considering any differences in body mass between experimental groups. These smaller muscle fibers and overall muscle size are also seen in adult HA deer mice (Lui et al., 2015).

We found that many of the traits that are characteristic of a specialized HA muscle phenotype were established in HA mice over the first month of postnatal development, but interestingly not all traits matured at a similar pace. For example, muscle capillarity diverged between HA and LA pups at P21, suggesting that HA mice have an increased capacity for O₂ delivery to muscle by this age. Evidence for a fiber-type difference at this age is supported by the higher capacity for pyruvate to lactate exchange in LA pups (LDH activity) relative to HA pups over this same time period (P21–P27). However, differences in muscle aerobic fiber numerical density (SDH staining) were not apparent until 1 week later at P27. This contrasts with adult skeletal muscle, where capillarity tends to vary in close association with muscle fiber aerobic capacity (e.g. Sullivan and Pittman, 1987; Hoppeler and Kayar, 1988). Another important mitochondrial protein, the citric acid cycle enzyme CS, also showed greater activity in HA mice by P27. CS is often used as an indirect marker of mitochondrial volume and these data suggest that it may take at least 4 weeks of postnatal development to establish the higher mitochondrial volume density seen in adult HA deer mice (Mahalingam et al., 2017). In skeletal muscle of adults there is a tight relationship between mitochondrial volume density and muscle fiber capillarity, which allows maximal O₂ extraction from blood (Hepple, 2000). This developmental pattern, where increases in capillarity precede increases in mitochondria, may be a mechanism to ensure that muscle cells have an adequate O₂ supply prior to increasing aerobic capacity. This may prevent a mismatch between O₂ supply and demand in the maturing muscle of HA mice.

Interestingly, there was no change in the capacity for lipid oxidation (indexed by HOAD activity) across development in HA juveniles, despite the importance of lipid metabolism in adults of this population (Cheviron et al., 2012). This may represent a temporal mismatch in the development of mitochondrial quantity versus quality and indicates that not all aspects of the HA adult muscle phenotype are established by P27. Taken together, these data serve as an important reminder that maturation of adult muscle phenotype is a combination of many independently developed functional parameters (e.g. mitochondrial volume, capillarity, myosin heavy chain isoform) that eventually preferentially cluster together in the mature animal (Schiaffino and Reggiani, 2011). The underlying mechanisms responsible for this developmental program are unclear but the regulatory gene networks that drive the dynamic changes during ontogeny of skeletal muscle in HA mice are likely complex and this is an important area for future study.

It is notable that the developmental patterns for the traits we measured were similar in muscle of both LA P. maniculatus and P. leucopus. This suggests that the developmental trajectory seen in HA muscle is a derived phenotype that has evolved at higher elevations. Given our common garden experimental design, we attribute the observed differences in LA and HA skeletal muscle to genetically based changes in the developmental program. It is also possible that evolved differences in environmentally induced plasticity influence the development of skeletal muscle in the wild. However, adult deer mouse skeletal muscle is generally resistant to hypoxia-induced phenotypic flexibility (Lui et al., 2015; Lau et al., 2017; Mahalingam et al., 2017), and rearing deer mice in chronic hypoxia had little effect on adult muscle phenotype (Nikel et al., 2018). Thus, the pattern of muscle maturation we observed likely reflects the developmental trajectory of wild deer mice. What then are the functional and fitness consequences of mismatched developmental timing of important muscle traits on whole-animal thermal performance in juvenile mice?

Like other altricial mammals, the development of homeothermic endothermy in young deer mice first involves the maturation of BAT and its regulation early in the postnatal period (P0–P10; Robertson et al., 2019). Any further increase in thermogenic capacity after this point may involve either increases in the capacity for NST and/or the developmental onset of shivering. In altricial mammals, there is a gradual postnatal onset of shivering, which occurs at irregular bouts in response to cold, and with a much lower amplitude compared with that in adults (direct quantification using electromyography in guinea pig: Brück and Wünnenberg, 1965; golden hamster: Hissa and Lagerspetz, 1964; Norwegian lemming: Hissa, 1968; Egyptian fruit bat: Hissa, 1964; and laboratory mouse: Arajamma and Lagerspetz, 1978). We found that LA deer mice were able to shiver at P14, with shivering thermogenesis contributing ∼35% to cold-induced V̇_O2,max. Surprisingly, at this same age, stimulation of NST in HA pups could account for 100% of cold-induced V̇_O2,max, suggesting an inability to shiver. Lack of shivering likely led to the 30% lower thermogenic capacity of HA mice compared with LA mice at P14. Importantly, regardless of altitude ancestry, at P14 all measures of muscle phenotype were similar. Yet, LA juveniles could shiver robustly in response to cold while HA juveniles could not. This implies that the whole-animal performance differences at this age are not due to differences in muscle phenotype. It has been suggested that shivering occurs later in ontogeny relative to NST because of the slow development of the muscles compared with BAT (Lagerspetz, 1966). However, the gradual onset of shivering in altricial mammals is also caused by the postnatal development of the neural pathways that regulate heat production (Arajamma and Lagerspetz, 1978). Thus, it is possible that HA mice have delayed maturation of the regulatory systems that govern the shivering reflex. This is supported by our previous study, where delayed activation of BAT in HA pups over the first 10 days of postnatal development appeared to be driven by a delay in the sympathetic innervation of the tissue (Robertson et al., 2019). Given that BAT in the HA mice appears to be functional at P14, the shared neural circuitry of the shivering and non-shivering pathways upstream of the motor neurons is likely intact (Madden and Morrison, 2019). Therefore, the delay in shivering may be due to underdeveloped efferent activation of muscle in response to cold.

Between P14 and P27, V̇_O2,max steadily increased in HA pups, while LA pups showed no significant change in thermogenic capacity over the same period. During this time, absolute rates of NST did not change; therefore, the resulting increase in thermogenic capacity must be due to the onset of shivering in HA pups. Despite the delay in the onset of shivering thermogenesis, the capacity for shivering increased continuously from 2 to 4 weeks of postnatal development in HA deer mouse pups and their V̇_O2,max surpassed that of LA pups by P27, consistent with population divergence in muscle phenotype. While the specialized aerobic phenotype of HA muscle is not required for shivering thermogenesis, once established it likely drives the improved thermogenic performance of HA relative to LA mice.

It is notable that cold-induced V̇_O2,max was only higher in 27 day old HA juveniles when tested under normoxic conditions. In contrast, when tested under hypoxia, thermogenic capacity was severely limited and did not differ between LA and HA mice. As previously discussed, HA mice at this age do not yet have the high capacity for lipid oxidation that is seen in adults. In adult mice, this gastrocnemius muscle metabolic profile directly correlates with thermal performance under hypoxia (Cheviron et al., 2012, 2014) and could explain the lower performance of juveniles under hypoxia at this age. Additionally, thermogenesis is limited not just by muscle performance but also by O₂ availability (di Prampero, 1985). Many HA natives, including deer mice, have a suite of respiratory adaptations (hemoglobin affinity, breathing pattern, etc.) that increase their capacity for O₂ delivery (McClelland and Scott, 2019). Many of these traits are also established postnatally and could lag behind skeletal muscle development, limiting thermogenesis under hypoxia. Importantly, these findings emphasize that even at this more mature age, HA juveniles are likely not proficient thermoregulators in their native environment.

If an elevated cold-induced V̇_O2,max is a critical adaptation for adult HA deer mice, why then do they delay the development of shivering thermogenesis and overall thermogenic capacity, particularly under hypoxia? The answer may lie in the high energetic cost of thermoregulation at HA. For example, field metabolic rate data suggest that adult HA deer mice routinely operate much closer to their V̇_O2,max than LA deer mice in the wild (Hayes, 1989). During the postnatal period, pups can avoid this energetic cost by relying on their mothers as an external heat source (Hill, 1983). It is notable that HA thermogenic capacity only begins to surpass that of LA pups after weaning at P21. This is also the point where many of the phenotypic traits of the skeletal muscle diverge. HA pups may not require the full development of thermogenic capacity before this point if they rely more heavily on maternal care than their LA conspecifics. However, female HA P. maniculatus give birth to larger litters of pups than either LA P. maniculatus or P. leucopus, both in the wild and in captivity (Robertson et al., 2019; Halfpenny, 1980). Large litter size increases intra-litter competition for food, limiting growth rate in rodents (Kaufman and Kaufman, 1987). Wild and captive-bred LA Peromyscus mothers compensate for large litter sizes by increasing food intake (Miller, 1975, 1979; Glazier, 1985), allowing postnatal growth rates of their pups to be maintained. This response is exacerbated in the cold, where rodent mothers must further increase food intake to meet their own energetic needs (Hammond and Kristan, 2000; Paul et al., 2010). To provision their large litters in cold conditions, female HA deer mice may spend substantially more time foraging for food than their LA counterparts, leaving their pups unattended. HA pups would therefore experience longer or more frequent bouts of cold. Thus, HA pups may have evolved to prioritize the maintenance of postnatal growth rates under these conditions by limiting their capacity for energetically costly thermogenesis over the nursing period. We see evidence for this in the similar pre-weaning growth rates of HA pups and their LA conspecifics, despite the HA mice having greater intra-litter competition for milk. Instead of attempting to maintain body temperature when their mothers are absent from the nest, juveniles may save energy by allowing body temperature to drop. Whether this is accomplished actively, by purposefully entering a torpor-like state, or passively as a result of an immature thermoregulatory system remains to be seen (Geiser, 2008). This growth versus thermoregulation trade-off likely manifests throughout postnatal development at HA as we have also found that the maturation of BAT-based NST is delayed in HA deer mice (Robertson et al., 2019).

We have shown for the first time that the aerobic muscle phenotype, characteristic of many HA native endotherms, is established over the first month after birth in HA deer mice. Additionally, we have shown that juvenile HA deer mice have a limited capacity to independently thermoregulate early in postnatal development because of a delayed ability to engage in shivering thermogenesis relative to their LA conspecifics. This delay is likely a result of the regulatory systems that activate shivering, rather than the maturation of the skeletal muscle itself. Previously, we have shown that the earlier postnatal onset of NST is also delayed in HA deer mice because of a similar latency in BAT activation (Robertson et al., 2019). Taken together, these findings suggest that HA mice repress thermogenesis after birth until weaning. While these findings may seem paradoxical given the superior thermogenic capacity of adult mice native to high elevations, it is likely that this is a critical adaptation that allows pups and juveniles to preserve energy and maintain high postnatal growth rates in this extreme environment.

The authors thank Leanne Zubowski, Debbie Chan, Wafa Khoja and Kwasi Nkansah for their invaluable assistance in data collection. Additionally, we sincerely thank Sulaymon Lyons for his assistance with histological analysis.