Jeffrey E. Lane, Centre d’Ecologie Fonctionelle et Evolutive, Unité Mixte de Recherche 5175, Centre National de la Recherche Scientifique, 1919 Route de Mende, Montpellier Cedex 5, 34293, France. Tel.: +33 (0) 4 67 61 32 27; fax: +33 (0) 4 67 41 21 38; e-mail: email@example.com
The life history schedules of wild organisms have long attracted scientific interest, and, in light of ongoing climate change, an understanding of their genetic and environmental underpinnings is increasingly becoming of applied concern. We used a multi-generation pedigree and detailed phenotypic records, spanning 18 years, to estimate the quantitative genetic influences on the timing of hibernation emergence in a wild population of Columbian ground squirrels (Urocitellus columbianus). Emergence date was significantly heritable [h2 = 0.22 ± 0.05 (in females) and 0.34 ± 0.14 (in males)], and there was a positive genetic correlation (rG = 0.76 ± 0.22) between male and female emergence dates. In adult females, the heritabilities of body mass at emergence and oestrous date were h2 = 0.23 ± 0.09 and h2 = 0.18 ± 0.12, respectively. The date of hibernation emergence has been hypothesized to have evolved so as to synchronize subsequent reproduction with upcoming peaks in vegetation abundance. In support of this hypothesis, although levels of phenotypic variance in emergence date were higher than oestrous date, there was a highly significant genetic correlation between the two (rG = 0.98 ± 0.01). Hibernation is a prominent feature in the annual cycle of many small mammals, but our understanding of its influences lags behind that for phenological traits in many other taxa. Our results provide the first insight into its quantitative genetic influences and thus help contribute to a more general understanding of its evolutionary significance.
Hibernation is a seasonal period of depressed behavioural and metabolic activity, widely assumed to be an adaptation to long-term energy deficits (Wang, 1989; Humphries et al., 2003; but see Lovegrove, 2000; Liow et al., 2009 for alternative hypotheses). It involves an interval of prehibernal energy storage, followed by an extended bout of metabolic depression (torpor), punctuated by periodic arousals to euthermic body temperature (Geiser & Ruf, 1995). During torpor, body temperature is maintained at a drastically reduced set point (e.g. as low as −2.9 °C in arctic ground squirrels, Urocitellus parryii; Barnes, 1989), resulting in pronounced metabolic savings (torpid metabolic rate is, on average, 5% of euthermic basal metabolic rate; Geiser & Ruf, 1995). Hibernation lasts several weeks to months (Geiser & Ruf, 1995), with animals in temperate regions typically entering (immerging) in the autumn and emerging in the following spring.
Hibernating mammals potentially face important energy-mediated trade-offs between the timing of hibernation emergence and reproduction. Emergence from hibernation must be early enough to allow for sufficient time to complete reproduction in the subsequent year and for both mothers and offspring to subsequently allocate the necessary energetic resources prior to immergence for the following winter. Synchronizing the periods of highest energetic need (i.e. lactation and juvenile weaning) with peaks in resource abundance has thus been hypothesized as an ultimate influence on emergence dates (Michener, 1983). However, the energetic costs of reproduction (Speakman, 2008; Lane et al., 2010) may not be supported early in the season, and early emergence may thus negatively influence survival probabilities (Michener, 1979). The ecological significance of the date of emergence from hibernation has recently been shown in yellow-bellied marmots (Marmota flaviventris). In years of early snowmelt, marmots emerge from hibernation earlier. This, in turn, yields a longer growing season (for the marmots) and allows individuals to achieve a larger body mass prior to immergence. Over-winter survival is consequently greater and population size has increased in recent years (Ozgul et al., 2010).
In general, the date of emergence should be an important life history trait in hibernating mammals. Relatively few attempts, however, have been made at explaining individual variation in hibernation timing and, to our knowledge, no studies to date have estimated its heritability or genetic associations with other life history traits. For example, a positive genetic correlation between emergence date and reproductive timing is an intrinsic, but yet untested, assumption of the hypothesis that the former has evolved so as to synchronize the latter with the temporal distribution of resource availability. In addition, we would predict a negative covariance between emergence date and energy stores, as the individuals with the greatest stores should be more able to fuel the metabolic costs of early emergence whereas those with diminished stores should remain in hibernation until environmental resources become more plentiful. Under this scenario, understanding the causes of variation in energy stores is necessary for predicting variation in emergence date. To date, however, the associations between energy stores and hibernation and reproductive phenology have only been addressed at a phenotypic, and not genetic, level (Dobson, 1988; Neuhaus, 2000; Humphries et al., 2003).
The aim of this study was to examine the quantitative genetic influences on hibernation phenological traits in a wild population of Columbian ground squirrels (Urocitellus columbianus). Columbian ground squirrels are herbivorous (Ritchie & Belovsky, 1990; Elliott & Flinders, 1991), and the short growing seasons of plants in the alpine and subalpine regions that they inhabit require that they spend the majority of the year hibernating (8–9 months, depending on sex and age-class; Dobson et al., 1992). The timing of emergence should thus be especially important in this species as reproduction and preparation for hibernation must all be completed during the short 3–4 months of activity. We estimated the heritabilities of emergence date from hibernation for both sexes separately, as well as genetic correlations between the sexes; to our knowledge, this is the first test for heritable genetic variation underlying hibernation phenology in a wild mammal. We also estimated the sex-specific heritabilities and between-sex genetic correlation of body mass at emergence. Individuals in this population are thought to rely entirely on accumulated fat stores to supply the energetic resources during hibernation (Young, 1990), and total body mass has been shown to correlate strongly with total fat content in the congeneric Belding’s ground squirrel (Urocitellus beldingi; Morton & Tung, 1971). We thus used body mass as a noninvasive proxy for total fat content. For females, we also estimated the heritability of oestrous date and its genetic correlation with the date of emergence. Finally, we estimated the phenotypic and genetic correlations among all of the traits within each sex.
Study area and population
Columbian ground squirrels are small (< 1 kg), ground-dwelling mammals resident in the Rocky Mountains of North America (Elliott & Flinders, 1991). We studied a free-ranging population in Sheep River Provincial Park, Alberta (50° N, 114° W and 1500 m a.s.l.), from 1992–2009. Details of the study area and population have been provided previously (Skibiel et al., 2009; Viblanc et al., 2010). Briefly, the population inhabits a 1.5-ha mixed-species grassy meadow [common species: tufted hair grass (Deschampsia caespitosa), old man’s whiskers (Geum triflorum) and dandelion (Taraxacum spp.)] on the banks of the Sheep River, surrounded by a forest of white spruce (Picea glauca), lodgepole pine (Pinus contorta) and trembling aspen (Populus tremuloides).
Beginning in the early spring (prior to emergence of the first individual), we conducted daily observations from 3-m-high observation towers that permitted full coverage of the study site to survey for emergent squirrels. Upon emergence from hibernation, squirrels were caught in live traps (Tomahawk Live Trap, Tomahawk, WI, USA) that were baited with peanut butter and placed in the immediate vicinity of the exit tunnel from the hibernaculum. Columbian ground squirrels hibernate individually (Shaw, 1925; Young, 1990) and are relatively sedentary until females enter oestrus in the week following emergence. This allowed us to be certain that we captured the focal animal for which we observed emergence date. Upon capture, individuals were weighed (to the nearest 5 g, using a Pesola spring scale) and received unique dye marks on their dorsal pelage (Clairol commercial hair dye) to facilitate subsequent observation from a distance. Most individuals (N = 1127) were originally handled as juveniles and at that time received unique alphanumeric eartags. Adults in the first year of the study (1992; N =32) and any immigrating animals (N = 51), similarly, received eartags on first capture. All animals were subsequently followed until either natural death or emigration from the study population. Therefore, any resident animals surviving for two or more years provided repeated measurements of traits (Table 1).
Table 1. Sample sizes and parental information for the quantitative genetic analysis of three phenotypic traits (emergence date from hibernation, body mass at emergence and oestrous date) for Columbian ground squirrels.
Number of animals with both parents known
Number of individuals with known maternity only
Average (and range) number of measurements per individual
Total number of measurements
*Any yearlings surviving to an age of at least two years will be counted in both the yearling and adult data sets.
Body mass at emergence
Oestrous dates of females were determined through regular trapping and behavioural observation. Females in oestrus enter burrows with males for periods of up to 2 h, after which most are aggressively mate-guarded by the males (Manno et al., 2007). These events occur for only a few hours on a single day each year, and backdating from known parturition dates (gestation length = 24 days; Shaw, 1925; Murie & Harris, 1982) confirms that this is when mating occurs (F.S. Dobson and J.O. Murie, unpublished).
Nest burrows were located towards the end of the gestation period by following the protocols of Murie et al. (1998). From 2001 to 2009, females were removed from the field 22 days after mating (ca. 2 days preparturition) and transported to a nearby (< 1 km) laboratory facility. Females were housed in polycarbonate cages (48 × 27 × 20 cm; Allentown Caging Equipment Company, Allentown, NJ, USA) and provided with wood chip bedding and newspaper or paper towel nest material. To obscure vision from neighbouring females and to simulate the burrow environment, cages were covered with vented black plastic bags. Females were fed lettuce and apple twice daily and a high-protein horse feed (oats, barley, wheat and compressed vegetable material in a molasses mix; 13% crude protein) ad libitum. Within 12 h following parturition, neonates were marked individually by removing a small tissue biopsy from an outer hind digit or the tail with sterile scissors. Tissues were subsequently individually preserved in 95% ethanol. Following litter processing, the female and her litter were transported back to the field and released into the nest burrow following protocols of Murie et al. (1998). Juveniles emerged from the nest burrow ca. 27 days post-parturition. Individuals of neither sex disperse as juveniles (Boag & Murie, 1981); therefore, all animals surviving their first winter were captured as yearlings the following year. Subsequently, yearling males are more likely to disperse than yearling females (Neuahus, 2006), and our adult data set is consequently female-biased (Table 1). All protocols were approved by the Life and Environmental Sciences Animal Care Committee at the University of Calgary as well as the Animal Care and Use Committee at the University of Alberta (1992–1999) or the Institutional Animal Care and Use Committee at Auburn University (1999–2009).
Molecular analyses, paternity assignment and pedigree reconstruction
DNA was extracted from preserved tissue using DNeasy tissue extraction kits (Qiagen, Venlo, The Netherlands), and polymerase chain reaction amplification was performed for a panel of 13 microsatellite loci. Prior to 2001, maternity was determined through behavioural observation at the nest burrow. Confidence of maternity assignment is assured through individual females occupying separate nest burrows and targeted behavioural observations and trapping effort on and around estimated dates of offspring emergence. For laboratory births, maternity assignment was guaranteed because females were housed in separate cages. Paternity was assigned at 95% to 99% tri-confidence (assumed dam–sire–offspring relationship) using cervus 3.0 (Marshall et al., 1998; Kalinowski et al., 2007). Analyses were conducted for each year (2001–2009) separately. Further details of microsatellite loci isolation and paternity assignment are provided elsewhere (Raveh et al., 2010).
Maternities have been determined since the inception of the project (first litters documented in 1992), whereas paternities have been determined since 2001. Consequently, our pedigree consists of a greater number of maternal than paternal links. In total, our pedigree consists of 1210 individuals (416 with both parents known, 711 with known maternity but not paternity and 83 base individuals). Of these individuals, 732 did not survive their first winter and consequently did not provide any relevant phenotypic data. Sample sizes and parental information for each of the phenotypic traits are provided in Table 1.
As a first step, we evaluated the influence of age-class (yearling and adult) and year as, respectively, two- and 18-level factors on emergence dates and body masses of each sex separately by fitting them as fixed effects, with individual as a random effect, in a linear mixed-effects model in the ‘nlme’ package (ver. 3.1-96; Pinheiro et al., 2009) in r (ver. 2.9.2; R Core Development Team). We did not include yearling data in the analyses of oestrous date because only 15 yearlings entered oestrus over the duration of the study. Consequently, these analyses only include year as a fixed effect. Variance components, heritabilities and trait correlations for hibernation emergence date, body mass and oestrous date were subsequently estimated using restricted maximum likelihood ‘animal models’ in asreml 2.0 (Gilmour et al., 2006). The phenotypes of each individual were split into the relevant fixed and random effects using a linear mixed animal model of the form:
where y is the vector of observed phenotypic values, b is the vector of fixed effects, a, m and pe are the vectors of additive genetic, maternal and permanent environment effects, respectively, and X and Z1–3 are the corresponding design matrices (Lynch & Walsh, 1998; Kruuk, 2004).
The total phenotypic variance (VP) in each trait was thus partitioned as:
where VP is the total phenotypic variance, comprising the additive genetic variance (VA), the maternal effect variance (VM), the permanent environment effect variance (VPE) and the residual variance (VR). The maternal effect variance arises due to maternal effects shared by offspring of the same mother. The permanent environment effect arises due to differences between individuals other than those due to additive genetic or maternal effects and can be estimated from repeated measures of a trait on the same individual (Kruuk & Hadfield, 2007). Both were fitted to prevent the variance associated with them being mistaken for additive genetic variance (Lynch & Walsh, 1998; Kruuk & Hadfield, 2007).
Narrow-sense heritabilities (h2), maternal effects (m) and permanent environment effects (pe) were calculated as h2 = VA/VP, m = VM/VP and pe = VPE/VP, respectively. Phenotypic (rP) and genetic (rG) correlations were calculated from bivariate animal models. Due to problems with model convergence and because maternal effects were small and nonsignificant in all univariate analyses (see below), we did not include maternal effects in the bivariate analyses. The multivariate phenotypic variances for traits x and y were partitioned first into individual and residual variance components with the added covariance components (individual [covID(xy)] and residual [covR(xy)]). In those cases, where the individual covariance component was significantly different from zero, we then attempted to partition this component into the constituent genetic [covG(xy)] and permanent environment covariances [covPE(xy)]. Between-sex genetic correlations were calculated by estimating the covariances between the sex-specific traits. Because these traits are measured in different individuals, we cannot evaluate the phenotypic correlations. Fixed effects included in the univariate models [age-class (where appropriate) and year] were likewise included in the bivariate models.
We used likelihood ratio tests (−2× difference in log-likelihood between hierarchical models) to test for the significance of variance and covariance components. Models were reduced either by omitting random effects sequentially or by fitting the covariances to 0. However, to avoid pseudoreplication, the permanent environment variance component was retained in all models. The likelihood ratio was tested against a chi-square distribution with the number of d.f. that corresponded to the difference in the number of variance or covariance components estimated. For univariate analyses, we follow Self & Liang (1987) and provide the P-values from one-tailed tests (because we are testing the hypothesis that the variance component is greater than 0). For bivariate analyses, we provide the P-values from two-tailed tests (because we are testing the hypothesis that the covariance component is either significantly less or more than 0).
Females cannot breed before they emerge from hibernation; some degree of positive correlation between these two traits is thus the null expectation. To evaluate whether the phenotypic and genetic correlations between emergence and oestrous dates that we estimated were significantly different from the null expectation, therefore, in addition to the likelihood ratio tests, we also ran a more conservative permutation analysis (sensu Charmantier et al., 2006). Specifically, for each emergence date, we randomly selected (with replacement and within each year) an oestrous date from the data set. To accommodate the biological constraint that females must emerge from hibernation before mating, we set a rule in the re-sampling procedure that oestrous date must be greater than emergence date. We permuted the data set 500 times to create the re-sampled data sets and estimated the phenotypic and genetic correlations for each with asreml-r (Butler et al., 2007). The r code for this analysis is available, upon request, from J.E.L, A.C. or M.B. Means, variance components, heritabilities and phenotypic and genetic correlations are presented as ± 1 standard error (SE) throughout, with SEs for the variance components, heritabilities, covariances and correlation estimates being provided by asreml.
The mean Julian date of emergence from hibernation of adult females and males, respectively, was 116.9 ± 0.3 (ca. 27 April; N = 472 records) and 112.7 ± 0.6 (ca. 23 April; N = 195 records). By comparison, the mean Julian date of emergence from hibernation of yearling females and males, respectively, was 123.1 ± 0.6 (ca. 3 May; N = 199 records) and 122.4 ± 0.7 (ca. 2 May; N = 212 records). Yearlings of both sexes emerged significantly later than their adult counterparts [F1,441 = 113.77, P <0.0001(female); F1,137 = 116.88, P <0.0001 (male)] (Fig. 1). Emergence dates also differed significantly across the 18 years of the study in both females (F17,441 = 20.03, P <0.0001) and males (F17,137 = 5.42, P <0.0001).
Within a year, there was an average 19.1 ± 2.1 (female) and 18.4 ± 2.4 (male) day interval between the emergences of the first and last adults [N = 18 years; range, 5–37 days (female) and 5–35 days (male)]. There was also a 17.2 ± 2.2 (female) and 22.0 ± 2.7 (male) day interval between the emergences of the first and last yearlings [N = 18 years (female) and 17 years (male; no yearling males emerged in 2003); range, 5–34 days (female) and 2–41 days (male)]. The phenotypic variance in emergence date was comprised of a significant additive genetic variance component in both sexes (Table 2), resulting in heritabilities of 0.22 ± 0.05 (female) and 0.34 ± 0.14 (male). Between sexes, there was a positive genetic correlation in emergence date (rG = 0.76 ± 0.22; COVA = 10.45 ± 3.77; χ21 = 3.73, P =0.05).
Table 2. Variance components, heritabilities and maternal and permanent environment effects (±SEs) of hibernation emergence date, body mass at emergence and oestrous date in Columbian grounds squirrels. Estimates from both the full models (including all variance components) and the reduced models (including only those retained as significant at the P <0.05 level) are provided.
The mean body mass at emergence of adult females and males, respectively, was 426.17 ± 2.32 g (N = 464 records) and 530.80 ± 6.58 g (N = 195 records). By comparison, the mean body mass at emergence of yearling females and males, respectively, was 253.73 ± 2.37 g (N = 193 records) and 270.34 ± 3.04 (N = 208 records). Yearlings of both sexes emerged at a significantly lower mass than their adult counterparts [F1,428 = 2374.09, P <0.0001 (female); F1,135 = 1130.29, P <0.0001 (male)]. Body masses at emergence also differed significantly across the 18 years of the study in both females (F17,428 = 22.25, P <0.0001) and males (F17,135 = 9.04, P <0.0001).
Within a year, there was an average difference of 182.50 ± 10.69 g (female) and 255.56 ± 16.26 g (male) in body mass between the lightest and heaviest emerging adults [N =18 years; range, 70.00–275.00 g (female); 125.00–410.00 g (male)]. There was also a 83.89 ± 10.15 g (female) and 119.42 ± 12.88 g (male) difference between the body masses of the lightest and heaviest emerging yearlings [N = 18 years (female) and 17 years (male; no yearling males emerged in 2003); range, 5.00–155.00 g (female) and 45.00–225.00 g (male)]. The phenotypic variance in body mass was comprised of a significant additive genetic variance component in females (h2 = 0.23 ± 0.09) but not in males (h2 = 0.02 ± 0.15; Table 2). Between sexes, there was a positive genetic correlation in body mass (rG = 0.74 ± 0.22; COVA = 866.66 ± 288.41; χ21 = 5.32, P = 0.02), but this must be interpreted with caution due to the lack of significant heritability for body mass in males.
Adult females entered oestrus, on average, 3.6 ± 0.1 (N = 444 records) days after emerging from hibernation [mean oestrous date = 120.2 ± 0.3 (ca. 30 April)]. Patterns in oestrous date thus correspond closely to those of emergence dates (Fig. 2a). However, as the later-emerging females were less likely to breed, there was less phenotypic variance in oestrous, relative to emergence, date (Table 2). Oestrous dates differed significantly across the 18 years (F17,294 = 14.76, P <0.0001). Within a year, there was an average of 15.8 ± 1.8 days (N = 18 years; range, 1–32 days) interval between the oestrous dates of the first and last adult females. The estimate of additive genetic variance in oestrous date was not significantly different from zero (Table 2), but the heritability of oestrous date (h2 = 0.18 ± 0.12) was very similar to that of emergence date.
Within-sex correlations between traits
There was significant phenotypic covariance between female emergence date and oestrous date (COVP = 25.68 ± 2.24; χ21 = 412.42, P <0.0001; Fig. 2a) that was comprised of significant underlying additive genetic covariance (COVA = 7.70 ± 2.17; χ21 = 15.8, P <0.0001) (Table 3). We then compared these values with the distribution of covariances expected under a null hypothesis of no association between dates (but with the proviso that emergence date should precede oestrous date). The mean phenotypic and genetic covariances estimated from the simulated data sets were, respectively, COVP = 18.49 ± 0.08 and COVA = 6.09 ± 0.05 (N = 424 data sets; 76 models did not converge). The phenotypic and genetic covariances estimated from the real data are equal to or larger than all (424 of 424) of the COVP estimates and 93% (393 of 424) of the COVA estimates.
Table 3. Individual, permanent environment, genetic, residual and phenotypic (maternal effects were not included) covariances (above the diagonal) and correlations (below the diagonal) between hibernation emergence date, body mass at emergence and oestrous date for adult female Columbian ground squirrels. Estimates are provided for two models. In Model 1, we combined the permanent environment and additive genetic components into one variable (individual). If the individual covariance was significantly different from zero, we then attempted to partition it into its constituent additive genetic and permanent environment parts in Model 2. The phenotypic covariance represents the sum of the covariance components from Model 1 or, if the individual covariance was significantly different from zero, Model 2. Values are reported as ±1 SE.
In contrast, neither the phenotypic (COVP = −18.17 ± 11.71; χ21 = 2.28, P =0.13; Fig. 2b) nor the individual (COVID = −4.45 ± 10.89; χ21 = 0.16, P =0.69) covariances between emergence date and body mass at emergence was significantly different from zero (Table 3). Because the individual correlation was not significantly different from zero, we did not attempt to partition into its constituent genetic and permanent environment covariances. Although there was a significant negative phenotypic covariance (COVP = −41.31 ± 14.57; χ21 = 10.76, P =0.001; Fig. 2c) between oestrous date and body mass, only the residual covariance (COVR = −20.10 ± 7.94; χ21 = 9.42, P =0.002) was significantly different from zero, whereas the individual (COVID = −21.22 ± 14.31; χ21 = 2.12, P =0.15) covariance was not (Table 3). We therefore also did not attempt to partition the individual correlation.
In males, there was a significant negative phenotypic covariance between emergence date and body mass at emergence (rP = −0.13 ± 0.06; COVP = −70.97 ± 29.92; χ21 = 11.16, P <0.001). This covariance was largely due to a significant residual covariance (rR = −0.36 ± 0.07; COVR = −112.08 ± 24.59; χ21 = 21.70, P <0.0001) as the individual covariance (rID = 0.18 ± 0.15; COVID = 41.11 ± 32.00; χ21 = 1.66, P =0.20) was not significantly different from zero. We did not attempt to partition the genetic and permanent environment correlations.
Predictable energy shortfalls are a regular feature of the life cycles of most organisms and have selected for a diversity of physiological and behavioural coping adaptations. Many birds migrate to more favourable climates, perennial plants become dormant and many invertebrates, amphibians and mammals hibernate during temperate zone winters. Although the quantitative genetic influences on avian migration and plant phenological traits have been estimated for a number of species (e.g. Rathcke & Lacey, 1985; Pulido et al., 2001), we are unaware of any similar attempts with hibernation phenological traits. We found that in Columbian ground squirrels the date of emergence from hibernation was heritable in both males and females and that there was a positive genetic correlation between the two sexes in emergence date. Oestrous date displayed a similar (albeit, not statistically significant) heritability as emergence date and the two traits were strongly phenotypically and genetically correlated. Body mass at emergence was also significantly heritable in females. Although it must be interpreted with caution, due to a lack of heritability for body mass in males, there was a positive genetic correlation in body mass between the sexes. There were, however, no significant genetic correlations between body mass and emergence date in either sex or body mass and oestrous date in females.
There was considerable phenotypic variation in emergence dates in our population. For example, the last adult female to emerge in a given year did so, on average, almost three weeks after the first. The compressed growing season at our Rocky Mountain field site, and correspondingly short active season, renders this variation (equal to almost 20% of the, on average, 99-day active season; Neuhaus, 2000) even more striking. Approximately 20% of this phenotypic variation was explained by an additive genetic variance component in females. In addition to our results, to our knowledge, the heritabilities of phenological traits in free-ranging mammals have been estimated in populations of three, nonhibernating species: North American red squirrels (Tamiasciurus hudsonicus) in the southwest Yukon Territory of Canada, red deer (Cervus elaphus) on the Isle of Rum, UK, and Soay sheep (Ovis aries) on the Island of Hirta, UK. Parturition date in both red squirrels (h2 = 0.16; Réale et al., 2003) and Soay sheep (h2 = 0.19; Kruuk & Hadfield, 2007) exhibited significant heritable genetic variation, and heritabilities across eight phenological traits in red deer ranged from 0.05 (female oestrous date) to 0.26 (both female coat change date and male rut start date) (Clements et al., 2010). Although the number of mammalian studies is still limited and an appropriate taxonomic comparison is thus premature at this point, available evidence suggests that many mammalian phenological traits are heritable and that estimates often overlap with those obtained from, the more often studied, bird populations [range, 0.16–0.45 (Sheldon et al., 2003 and references therein)].
All of the traits that we analysed exhibited pronounced annual variation. Including year as a fixed effect in our analyses does not markedly influence levels of additive genetic variance (e.g. VA in female emergence date with year as a fixed vs. random effect is respectively 8.04 vs. 8.65). However, because controlling for annual variation influences levels of explained phenotypic variation (e.g. VP in female emergence date with year as a fixed vs. random effect is respectively 36.85 vs. 56.20), heritability estimates are influenced (e.g. h2 in female emergence date with year as a fixed vs. random effect is respectively 0.22 vs. 0.15). Consequently, our heritability estimates need to be interpreted as the proportion of phenotypic variation explained by an additive genetic component, after controlling for the influence of annual variation (sensuWilson, 2008). Annual phenological variation likely results from annual variation in climatic conditions, and previous studies have identified the influence of soil temperature and snow depth (Murie & Harris, 1982; Michener, 1992). We believe it is unlikely, however, that a single environmental variable will explain the majority of the variation in hibernation traits. Hibernacula are typically 60 cm below the soil surface (Young, 1990), meaning that ambient temperature, snow depth and thermal inertia of the soil may all interact to influence the conditions experienced by hibernating squirrels.
Microevolutionary trajectories are predicted not only by the heritability of the focal trait but also by any genetic correlations between it and other traits, as well as the adaptive landscape associated with the multivariate phenotype (Lande, 1982; Kruuk et al., 2008). For hibernating species, the associations between hibernation and reproductive timing and energy stores should be important vectors of life history variation. Specifically, the date of emergence from hibernation in ground squirrels has been hypothesized to have evolved so as to synchronize the subsequent reproduction with the peak in vegetation resources (Michener, 1983). In addition to the heritability of emergence date, this hypothesis relies on the assumption of a genetic correlation between it and reproductive timing. We have shown here that oestrous dates, indeed, exhibited strong phenotypic and genetic correlations with female emergence date. Some degree of caution needs to be exercised in this case because, although the estimated heritabilities of emergence date and oestrous date were of similar magnitude, we are not able to conclude definitively that the heritability of oestrous date was statistically significant. We believe, however, that this limitation is due to the reduced statistical power resulting from a smaller sample size (arising from the fact that the oestrous date analyses did not include yearling data) and slightly smaller levels of additive genetic variation (Table 3). On the whole, therefore, we believe that the genetic evidence supports the synchrony hypothesis. Indeed, the strength of the genetic correlation (rG = 0.98) indicates that the two phenological traits are likely influenced by the same set of genes and that the ability of females to adaptively adjust the timing of reproduction is fundamentally linked to their ability to adjust the timing of hibernation emergence. Undoubtedly, there are numerous assumptions underlying the synchrony hypothesis that remain to be tested in this, and other, species (e.g. the correspondence between the ground squirrel and plant phenologies, and the strength and shape of selection on variation in synchrony). Our results, however, provide a necessary first step that can be built upon with subsequent investigations that test these assumptions.
We found no evidence that body mass was genetically correlated with emergence dates (in either sex) or oestrous dates in females. The pronounced metabolic savings achieved through hibernation should be expected to create a covariance between the energetic resources of individuals at emergence and their dates of emergence. Specifically, early emergences should coincide with greater energetic stores (necessary to fuel the elevated metabolic costs whilst active), leading to a negative covariance. There was a negative phenotypic covariance between male emergence date and body mass, as well as female oestrous date and body mass, but this resulted primarily from a negative residual (or ‘within-individual’) correlation. This suggests that the environmental conditions associated with early emergences or oestruses (e.g. favourable climatic conditions during the hibernation period) are also associated with emergences at higher mass. We chose to analyse body mass as a noninvasive proxy for total body fat (sensu Morton & Tung, 1971). Admittedly, the former represents a somewhat coarse-grained measure of the latter as it also incorporates variation in lean body mass (Peig & Green, 2010). In Belding’s ground squirrels, however, body mass correlated strongly with total body fat (r = 0.87; Morton & Tung, 1971), and in our population, the rank scores of body mass at emergence correlate strongly with the rank scores of body condition [estimated from the sex-specific regression of body mass on zygomatic arch breadth; statistics not shown (see also Dobson, 1992)]. We did not analyse body condition as our sample size of zygomatic arch breadths (N = 89 females and N = 43 males) is currently unsuitable for a quantitative genetic analysis. This analysis could be performed in future, however, and may provide improved detail into the relationship between energetic stores and hibernation phenology for this species.
If the level of energetic stores at emergence truly is genetically uncorrelated with emergence and oestrous dates, however, it does suggest that the latter two traits could evolve independently of the former (Lande, 1982). Mothers that emerge from hibernation in better condition tend to give birth to more offspring and, overall, heavier litters (Dobson et al., 1999; Broussard et al., 2003, 2005), suggesting that condition is likely subjected to positive directional selection. By comparison, previous work on Columbian ground squirrels has shown a reproductive advantage to early reproduction (Neuhaus, 2000). Unlike migratory birds, therefore, in which a reproductive/somatic investment trade-off is imposed by the energetic challenges of completing migration (Schmidt-Wellenburg et al., 2008), it is feasible that earlier emergences (and therefore reproduction) could evolve in parallel with higher body masses in this species.
To what extent phenological traits can evolve along sex-specific trajectories will also depend on the magnitude of the between-sex genetic correlation in the traits. In Columbian ground squirrels, a strong between-sex genetic correlation in emergence date implies that such sex-specific evolution will be constrained. In red deer, by contrast, although female oestrous date exhibits a significant genetic correlation with male rut start date, in general there is little evidence for significant genetic correlations between different phenological traits either within or across the sexes (Clements et al., 2010). Recent studies have suggested that elevated levels of protandry (i.e. the earlier arrival of males relative to females to the breeding population; Morbey & Ydenberg, 2001) may arise in response to climate change (Møller, 2004). However, this result has not been observed in other studies (Rainio et al., 2007). The presence of, and interpopulation variation in, between-sex genetic correlations in phenological traits may provide one causal explanation for these divergent results (Lane et al., 2011).
The earth is in the midst of a pronounced warming trend (IPCC, 2007), and there is growing concern as to how natural populations will respond to these rapidly changing conditions (Berteaux et al., 2004). Because an evolutionary response to climate change is assumed to be necessary for the long-term viability of natural populations (Visser, 2008), disentangling the genetic and environmental influences on life history traits is taking on increased applied relevance (Gienapp et al., 2008). Adjustments of phenological traits are the most commonly reported response to climate change (Parmesan, 2006), and the ability to adaptively respond to a changing climate has been shown to influence population dynamics in multiple species (Both et al., 2006; Møller et al., 2008), including hibernating mammals (Ozgul et al., 2010), making them especially pertinent. We have provided evidence for a necessary prerequisite to microevolution, heritable genetic variation, for hibernation phenological traits in Columbian ground squirrels (Falconer & Mackay, 1996). To what extent hibernation phenology may evolve in response to ongoing climate change is currently unknown but, quite independent of these issues, an improved understanding of this phenological trait should provide important insight into a fundamental component of the ecology of natural populations: the temporal linkages between organisms and their environment.
We thank the numerous volunteers and assistants who have helped with fieldwork on this population for the past 18 years and Philip Jones, Erica Kubanek and Jamie Gorrell for assistance with genotyping tissue samples. We are grateful to the University of Calgary’s Biogeosciences Institute (E. Johnson, Director) for providing us with accommodation and workspace at the R. B. Miller Research Station (J. Buchanan-Mappin, Station Manager). Fieldwork was supported Natural Sciences and Engineering Research Council of Canada grants (J.O.M.) and by a National Science Foundation grant (DEB-0089473) (F.S.D.). The Natural Science and Engineering Research Council of Canada (D.W.C.) also helped support genetic analyses. J.E.L. and L.E.B.K. were supported by the Royal Society of London and A.C. and F.S.D. were supported by a grant from the Agence Nationale de la Recherche of France (ANR-08-JCJC-0041-01). The ANR also supported J.E.L. during portions of the writing of this manuscript. We also thank two anonymous referees for helpful comments on the manuscript.