Preparing for hibernation in ground squirrels: adrenal androgen production in summer linked to environmental severity in winter


Correspondence author. E-mail:


1. At high latitudes, evolutionary adaptations focus on those that maximize survival, with hibernation being a major one used by many smaller mammals. Typically, mammalian hibernators overwinter in sites that are ≈0°C. However, in arctic regions, such sites do not exist, necessitating hibernation at sites well below 0°C. Lipid, the normal fuel of most hibernators, may not provide sufficient glucose needed by certain tissues to permit survival, with muscle breakdown being required. Critical to enhancing muscle stores are high concentrations of anabolic androgens prior to hibernation when the gonads are inactive.

2. We compare and contrast androgen levels in arctic ground squirrels (AGS) (Urocitellus parryii Richardson) from the Yukon and Columbian ground squirrel (CGS) (U. columbianus Ord) from southern Alberta.

3. In males, changes in testes mass over the active season were similar between AGS and CGS. In contrast, during the breeding and the nonbreeding, pre-hibernation periods, androgen levels in AGS were 6–10 and 20–25 times, respectively, those of CGS. From the breeding to the pre-hibernation periods, androgen levels declined 41% in AGS, but 86% in CGS. In females, androgen levels in AGS were high throughout the active season and, prior to hibernation, were 24 times those in CGS.

4. In pre-hibernating AGS, we determined the source of these high androgen levels from two studies. First, using a hormonal challenge protocol to probe the hypothalamic-pituitary-adrenal axis, we found that androgen levels in both females and males decreased ≥10% in response to suppression by dexamethasone (an artificial glucocorticoid that inhibits ACTH release) and increased ≥18% in response to direct adrenal stimulation by adrenocorticorticotropic homone (ACTH). Second, by sequential hormonal injections in males of gonadotropin-releasing hormone (GnRH) and of ACTH coupled with gonadectomy (GDX), followed by adrenalectomy (ADX), we found that GnRH had no effect, whereas ACTH stimulated androgen levels by >40%, both before and after GDX. After ADX, levels fell by 80%. Thus, the adrenals, not the gonads, are the source of these androgens.

5. We hypothesize that pre-hibernating AGS have evolved a solution seen in no other known mammal: exploiting the benefits of high adrenal androgen levels prior to hibernation to build muscle that is then catabolized overwinter. The unknown is how AGS have mitigated the costs of these high androgen levels.


The severity of high-latitude arctic and alpine habitats presents special challenges for mammals that live out all or some of their lives there. Given that habitat is the template that shapes and constrains the evolution of life history traits (Southwood 1977), these extreme habitats may require unique behavioural and physiological solutions. They are characterized by long, cold winters with permanent snow cover separated by short, cool summers with a burst of primary productivity. As reproduction is usually restricted to the summer, evolutionary adaptations for winter focus on those that maximize survival. The principal threats to winter survival are starvation and low temperatures, both of which may be aggravated by high predation risk. The options for surviving high-latitude winters in mammals are limited – either endure, leave, remain active under the snow or hibernate (Boonstra 2004). Superficially, hibernation would seem to be the ideal strategy but this ignores the fact that arctic hibernators cannot get below the frost line and thus must overwinter in frozen ground well below 0°C. Unlike the thermal conditions experienced by most other hibernators in nature where soil temperatures are at or near 0°C (e.g. Young 1990; Arnold et al. 1991), those of arctic ground squirrels (Urocitellus parryii; AGS, Fig. 1) can reach −23°C (Buck & Barnes 1999a). To survive these subzero temperatures necessitates much higher metabolic rates (Buck & Barnes 2000). Lipid, the primary metabolic fuel for most hibernators, provides the needed energy from the catabolism of the fatty acids but provides insufficient glucose (from glycerol). Certain tissues and functions require glucose for survival (brain, heart, erythrocytes and nonshivering thermogenesis of brown adipose tissue –Gibbons, Denton & McCormack 1985; Cooney et al. 1986; Dienel & Hertz 2001), and to provide it, gluconeogenesis from muscle breakdown is required (Galster & Morrison 1975; Buck & Barnes 2000). The problem for AGS is how to build the needed muscle tissue prior to hibernation.

Figure 1.

 Adult male arctic ground squirrel (Urocitellus parryi) foraging in the alpine region of south-western Yukon in August 2000. Photograph taken by Tim Karels.

The key anabolic steroid required for an increase in muscle mass is testosterone (Mooradian, Morley & Korenman 1987). It can be provided from the blood either directly (via testosterone from the testes) or indirectly (via the conversion within the tissue of androgen precursors such as androstenedione which is usually produced by the adrenals, King et al. 1999). Testosterone is critical for the behavioural and physiological changes necessary for breeding in vertebrates (Baum 2002). In breeding males, testosterone normally orchestrates a suite of co-evolved traits and is key in trading off self-maintenance for reproduction (Hau 2007). In nature, the benefits of high testosterone levels (aggressive and sexual behaviour, spermatogenesis, secondary sex characteristics, muscle hypertrophy and reduced response to pain) come with costs (immunosuppression, higher energetic demands and reduced paternal behaviour) (Wingfield, Lynn & Soma 2001). As a consequence, many seasonally breeding vertebrates limit high testosterone levels to the breeding season.

The AGS is the most northerly hibernator in North America (Banfield 1974) and differs from the typical testosterone pattern in two key respects. First, in late summer all males are in nonbreeding condition, yet published evidence indicate high plasma testosterone levels comparable to those seen during the breeding season (Barnes 1996; Boonstra, McColl & Karels 2001a). Second, nonbreeding males respond to ACTH (the hormone produced in the pituitary in response to a stressor) injections with a pronounced increase, not a decrease, in androgen levels (Boonstra, McColl & Karels 2001a). However, normally ACTH, acting by increasing adrenal glucocorticoids such as cortisol, suppresses testosterone (Wingfield & Sapolksy 2003). We are unaware of any other nonbreeding male mammal that shows this response. However, ACTH is known to stimulate the production and release of adrenal androgens such as androstenedione, dehydroepiandrosterone and its conjugates (Odell & Parker 1984). Thus, an adrenal origin for these androgens in AGS was suggested. The robust production of the adrenal androgens (but not testosterone) has been regarded as primarily a primate characteristic (Conley, Pattison & Bird 2004).

Here, we first document the pattern of androgen and testes changes in AGS males over the entire active season. Second, as females are faced with the same conditions as males during hibernation, we evaluated whether they too have a pattern of androgen changes similar to that of males. We contrast this pattern with that seen in males and females of a closely related species, the Columbian ground squirrel (U. columbianus, CGS), a mid-latitude hibernator overwintering in ground ≈0°C (Elliot & Flinders 1991). Third, to test whether pre-hibernating AGS females also respond positively to ACTH as do the male AGS, we use a hormonal challenge protocol [dexamethasone (DEX) suppression – ACTH stimulation]. Fourth, we determine the site of production of androgens in pre-hibernating male AGS (testes vs. adrenals).

Methods and materials

Study species

Critical to understanding the pattern of production of androgens in ground squirrels and its underlying cause is knowledge of their life history. First, it is important to know that all the literature on the species we discuss refer to the genus as Spermophilus, but that in a recent revision of the genus (Helgen et al. 2009), both AGS and CGS were reclassified into a closely related group of holarctic ground squirrels in the genus Urocitellus. AGS are found throughout the arctic tundra in North American and, in the north-west, in alpine areas and forest meadows (latitudinal range ca. 58–70°N, Banfield 1974). They are obligate hibernators, emerging above-ground from 8–9 months of hibernation in early to mid-April, with males appearing 1–2 weeks before females. As virtually all male and female yearlings breed (except for a small proportion of nonmaturing yearling males), there are essentially two age classes in summer after the breeding season: adults and juveniles. By midsummer, all are actively foraging to acquire sufficient mass to survive hibernation. Males, but not females, cache seeds underground (McLean & Towns 1981). Males eat these seeds the next spring when they arouse and become continuously euthermic up to 3 weeks prior to their emergence above-ground. This allows them to grow their testes and complete spermatogenesis prior to female emergence, and to replace all the mass lost during hibernation (Buck & Barnes 1999b). Buck & Barnes (1999a) found that AGS hibernated just above the permanently frozen ground at an average depth of 97 cm (N = 20 burrows, range 80–104 cm) and that hibernacula temperatures from October to April over 3 years averaged −8·9°C, with an average minimum of −15·5°C (range −8·7 to −23·4°C).

CGS are found in the alpine and subalpine meadows of the mountains of western North America (latitudinal range ca. 43–55°N, Elliot & Flinders 1991). Their life history is roughly similar to that of AGS; though, they are longer lived. Hibernation also lasts 8–9 months, but males mature at two and females at 3 years of age resulting in four age categories in summer prior to hibernation: mature adults, subadult females 1–2 years old, subadult males 1-year old and juveniles (Murie & Harris 1984). Young (1990) found that CGS hibernated at an average depth of 60 cm (N = 19, range 33–94) and from mid-December to mid-April soil temperatures measured at this depth were stable at −1°C.

We trapped AGS in two valleys in the Yukon: near Kluane Lake (61°N, 138°W) (Boonstra, McColl & Karels 2001a; Boonstra et al. 2001b) from 1987–1999 and near the Pelly River (63°N, 137°W) in spring 2007 (Delehanty & Boonstra 2011). We trapped CGS in two nearby valleys in south-western Alberta: in the Sheep River Wildlife Reserve (ca. 50°38′N, 114°38′W) (see Young 1990) in July 2005 and in the Kananaskis area (51°N, 115°W) in spring 2007.

Field protocol

The trapping technique was similar at all sites. We trapped squirrels in live traps baited with peanut butter (14 × 14 × 40 cm or 16·5 × 16·5 × 48 cm, Tomahawk Live Trap Co., Tomahawk, WI, USA) or in home-made burrow traps (Wobeser & Leighton 1979), restrained them in a netted bag, tagged them with monel #1005-1 tags (National Band and Tag Co., Newport, KY, USA) in both ears and aged, weighed, sexed and determined their breeding condition (males: testes scrotal or abdominal; females: pregnant, lactating or nonreproductive). We set livetraps outside burrow entrances between 7:30 and 8:30 h and checked them several times during the morning and then locked them closed.

On capture, we transported squirrels to a central site for processing and collection of blood samples. We did not know exactly when the squirrels entered the traps, but the maximum amount of time between capture and blood collection was <4 h. They were allowed to habituate for ca. 1 h before being processed. Anaesthesia was induced within 30–60 s and a blood sample (∼200–500 μL) collected within 1 min with a heparinized Pasteur pipette from the suborbital sinus. Prior to 1997, we used the bell-jar technique with the inhalant anaesthetic methoxyflurane at 3% and thereafter, we used a purpose-built portable anaesthetic unit delivering isoflurane at 3·5% (Boonstra et al. 2008). We collected blood samples within 1 min of induction, immediately centrifuged it (8 min at 8800 g) and froze the plasma (initially at −20°C in the field and at −80°C in Toronto). Our procedures follow Canadian Council of Animal Care guidelines and were approved under a University of Toronto Animal Use Protocol.

Study 1: seasonal changes in testes and androgen levels in males

To determine the seasonal changes in male androgen levels and testes mass, we obtained samples from both species over the active season (see Table 1 for sample sizes). Breeding males had enlarged, descended testes over the period from pre-breeding (from emergence to appearance of the females) to post-breeding phases (most females had been bred by this time); all others were nonbreeding with abdominal testes. We had three age categories in AGS males: adult breeding, adult nonbreeding (>1-year old), and juvenile nonbreeding (<4 months old); and four categories in CGS males: adult breeding, adult nonbreeding (>2-year old), yearling nonbreeding (1-year old) and juvenile nonbreeding (<4 months old). In AGS males, the seasonal overview of the pattern of change in androgen levels was a synthesis of samples collected in the Yukon throughout the active season over a number of years. Samples were divided into four periods. In the Pelly River area, we captured pre-breeding and breeding males from 9 to 12 April and 27 to 30 April 2007, respectively. In the Kluane area, we captured post-breeding and pre-hibernation males from 3 to 11 May and 25 July to 27 August 1996–99, respectively. The sample sizes for testes mass and plasma were not identical as some of the males were part of another study and were released back at their home burrow.

Table 1.   Samples sizes of males and female ground squirrels from field samples for studies 1 and 2
 Arctic ground squirrelColumbian ground squirrel
  1. Both the Kluane and the Pelly Rivers are in the Yukon, Canada, and both the Kananaskis and the Sheep Rivers are in south-western Alberta, Canada.

  2. *Testes masses for these two classes came from Elliott (1980).

Study 1: Males
  Pre-breeding adultPelly River9–12 Apr 20071217Kananaskis22–24 Apr 20071616
  Breeding adultPelly River27–30 Apr 20071418Kananaskis7–9 May 20071616
  Post-breeding adultKluane3–11 May 1996–991810    
  Pre-hibernation adultKluane25 Jul – 27 Aug 1996–9917 9Sheep River8–27 July 2005 5*12
  Pre-hibernation yearling    Sheep River8–27 July 2005  6
  Pre-hibernation juvenileKluane25 Jul – 27 Aug 1996–991910Sheep River8–27 July 2005 4* 2
  Total  8064  4152
Study 2: Females
    Plasma  Plasma
  PregnantKluaneLate Apr–May 1987–96 11   
  LactatingKluaneLate May–June 1987–96  5   
  Post-breeding adultKluaneJuly 1987–96 10   
  Pre-hibernation adultKluaneAug 1987–96 13Sheep River8–27 July 2005 13
  Early summer juvenileKluaneJuly 1987–96 10   
  Pre-hibernation subadults    Sheep River8–27 July 2005 13
  Pre-hibernation juvenileKluaneAug 1987–96 10Sheep River8–27 July 2005 11
  Total   59  37

In CGS males, the overview was a synthesis of samples collected over two years and these were divided into three periods (a post-breeding sample was not collected). In the Kananaskis area, we captured pre-breeding and breeding males from 22 to 24 April and 7 to 9 May 2007, respectively. In the Sheep River Reserve, we captured pre-hibernation samples obtained from 8 to 27 July 2005. Although we had plasma samples from all three periods, we had testes masses only from the first two (collection of animals was not permitted in the Reserve). Hence, for pre-hibernation values we used published testes masses collected on adult and juvenile nonbreeding males on 18 August 1978 from the Idaho Primitive Area (Elliott 1980).

Study 2: seasonal changes in androgen levels in females

To determine the seasonal changes in female androgen levels, we obtained blood samples from both species (see Table 1 for sample sizes). In AGS, we obtained samples were over the entire active season and the overview of the pattern was a synthesis of samples collected from 1987 to 1996 in the Kluane area. We determined reproductive condition by examination and sampled six categories: pregnant (late April–May, determined from knowledge of oestrus and copulation dates, visually near the end of pregnancy, or back-calculated given evidence of lactation), lactating (late May–June), adult – July (>1-year old, post-reproductive), adult – August (>1-year old, post-reproductive), juvenile – July (<3 months) and juvenile – August (3–4 months). In CGS, we obtained samples only from the pre-hibernation, nonreproductive period from 8 to 27 July 2005 in the Sheep River Reserve. We had three categories: adult (post-reproductive >2 years), subadult (1–2 years) and juvenile (<3 months).

Study 3: response to hormonal challenge

To probe the stress response in pre-hibernating AGS to a standardized hormonal challenge, we used the dexamethasone (DEX) suppression test followed by the ACTH stimulation test.

This protocol overrides the immediate stress effects of capture and handling, by standardizing the present to an integrated picture of stress axis responsiveness and what the animal has been experiencing in the recent past (days to weeks) (see Boonstra et al. 1998 for a discussion). The DEX suppression test is a method to assess whether the brain is registering glucocorticoid levels correctly (DEX is an artificial glucocorticoid) and making the necessary negative feedback adjustment by reducing ACTH and then adrenal steroid production. The ACTH stimulation test directly probes the capacity of the adrenals to produce and release steroids (normally glucocorticoids). Although our initial goal focussed primarily on changes in cortisol (Boonstra, McColl & Karels 2001a), given the evidence that the adrenals might be producing androgens, here we present only the changes in androgens. We have already reported the changes in androgens in pre-hibernating adult and juvenile AGS males (see Boonstra, McColl & Karels 2001a for additional details), but when it became apparent that females were also producing large amounts of androgens, we analysed the female plasma for changes in these hormones. Here, we contrast the pattern in females with that in males. The challenge was carried out on nonbreeding adult males (n = 10) and females (n = 7) and on juvenile males (n = 10) and females (n = 8) captured in late July and August 1995 and 1996 and transferred to the field laboratory at Kluane Lake. Each squirrel was bled five times with the first (the Base bleed, ∼500 μL) used to obtain baseline estimates of androgens. Immediately thereafter, squirrels were injected with 0·4 mg kg−1 DEX sodium phosphate (Sabex, Montreal, Canada). The second blood sample (the DEX bleed, 300 μL) was collected 2 h later and was immediately followed by an intramuscular injection in the thigh of 4 IU kg−1 of synthetic ACTH (Synacthen Depot; CIBA, Ontario, Canada). Subsequent blood samples were collected at 30, 60 and 120 min post-ACTH injection (called the P30, P60 and P120 bleeds, respectively, and all 300 μL). We determined the amounts of DEX and ACTH injected based on the mass obtained when the animals were trapped.

Study 4: source of androgen in nonbreeding male AGS

To determine the site of androgen production in nonbreeding, pre-hibernating males, we injected them sequentially with hormones to assess the responses of the testes and then of the adrenals when the animals were intact and when they were not (the organs were removed in a sequential fashion). We live-trapped adults (N = 12) from 25 to 29 July 1999 near Kluane Lake. They were housed at the Arctic Institute Base in large wire cages (66 × 23 × 23 cm) covered with burlap, provided with a nest box and cotton for warmth, and supplied with water and commercial rabbit chow (16% protein, Shur-Gain; Maple Leaf Foods Inc., Edmonton, AB, Canada) supplemented daily with fresh forbs, herbs and shrubs. Squirrels were allowed to habituate to the laboratory surroundings for at least 1 day before the start of the experiment. To control for individual differences among animals, all individuals served as their own controls. Half received the treatment on day 1 and half received saline; this was reversed on day 2. Each animal was bled twice per day: a Base bleed on removal from the cage between 09.30 and 11.00 h to assess resting levels of androgens followed immediately by an intramuscular thigh injection (either saline or treatment), followed by a second bleed 2 h later to assess the impact of injection. Table 2 gives the chronology of the experiment. Sequentially, the first treatment was GnRH (gonadotropin-releasing hormone) (10 IU of serum gonadotrophin – Folligon, Intervet Australia Pty Ltd, Lane Cove, NSW, Australia), the second ACTH (4 IU of synthetic ACTH – Synacthen Depot, CIBA), the third gonadectomy (GDX) followed by an ACTH injection and finally, adrenalectomy (ADX) of half; the other half were sham operated. We weighed all animals at first capture, at ADX and at killing, and the testes when the animals were gonadectomized. From initial capture to the date of ADX, males continued to gain weight (mean ± 1 SE: 3·4 ± 1·03 g day−1). All animals were killed on completion of the experiment and autopsied.

Table 2.   Chronology of experimental treatments in male arctic ground squirrels in 1999. Twelve males were captured and split into two groups
  1. All were initially bled and then half immediately received an intramuscular injection of saline on day 1 and the other half received the treatment injection. The order was reversed on day 2. Response to the injections was assessed by a second bleed 2 h later on each day.

July 25–29Capture at burrow sites
July 30–31Injected with either saline or GnRH
August 1–2Injected with either saline or ACTH
August 2Gonadectomy of all males
August 3–4Recovery from surgery
August 5–6Injected with either saline or ACTH
August 8–9Adrenalectomy of half; sham surgery of half
August 11Collected all surviving adrenalectomized squirrels
August 12Collected all sham operated squirrels

We carried out GDX and ADX surgical procedures following Lang (1976), with animals being anaesthetized with an intramuscular ketamine injection of 300 μL (100 mg mL−1; Parnell Laboratories, Silverwater, NSW, Australia) and supplemented as needed with isoflurane. To perform the GDX, a mid-abdominal incision was made through the abdominal wall to enable removal of the testes. Blood vessels were heat cauterized to prevent bleeding. To perform the ADX, a 30-mm mid-abdominal incision was made along the linea alba and the abdominal viscera from the pyloric region to the caecum were withdrawn and slid into a sterile powderless latex examination glove (Conform latex exam gloves; Ansell Medical, Dothan, AL, USA). During removal, amoxicillin sodium (Ibiamox; Protea Pharmaceuticals, Thornleigh, NSW, Australia) was puffed onto the viscera and which were kept moistened with sterile saline. The adrenal glands were dissected free from the connective tissue and fat, ligated and then removed carefully to ensure that no adrenal tissue remained. Abdominal viscera were carefully replaced and both the abdominal wall and the skin were sutured. Animals were observed at regular intervals during recovery and for the next 24 h after surgery. ADX animals were maintained on a normal diet but given isotonic saline (0·9g dL−1) to drink, which is sufficient for maintenance when renal function is not impaired. Two ADX animals died 2 days after surgery, but on autopsy there were was no sign of infection or internal bleeding.

Androgen hormone assays

All samples were analysed by radioimmunoassay with a testosterone (T) antibody (P43/11, Croze & Etches 1980) that we have used in all our studies (see Boonstra, McColl & Karels 2001a; b for RIA details). This antibody has low cross-reactivity to corticosterone (<0·9%), progesterone (<0·4%), estradiol (<0·8%) and dehydroepiandrostereone (DHEA) (<0·8%, Boonstra et al. 2008), but higher cross-reactivity to androstenedione (15·7%) and dihydrotestosterone (DHT) (62%) (cross-reactivities to 11 other androgen variants were between 0·7% and 9·0%). The protocol for this radioimmunoassay was a double diethyl ether extraction of duplicate plasma samples. Prior to extraction, each plasma sample (25 μL) was treated with 20 μL NH4OH to saponify triglycerides. Two quality control samples (a low and a high concentration of T) were run in duplicate with each assay. Blank values of solvents were also run with each assay and did not differ significantly from zero. The mean recovery of [1,2,6,7-3H] testosterone added to plasma was 96·5% (SE = 0·7%; range 92–102%). The assay was sensitive to 10 pg/25 μL plasma. The intra-assay coefficient of variation was 5%. The androgen data we report here came from multiple assays conducted over a 16-year period, and over that time, inter-assay coefficient of variation for the low quality control was 7·1% and for the high 6·2%. Thus, this assay was a robust, highly repeatable one giving accurate androgen values reflective of actual plasma concentrations.

To establish whether other major plasma androgens (other than T) are present in ground squirrel plasma, we analysed a number of samples with highly specific commercial antibodies. As little plasma remained from many of the samples used in these four studies, we analysed additional representative samples. To assess the DHT levels, we used a RIA kit (DSL-9600; Diagnostic Systems Laboratories, Webster, TX, USA) with the antibody having cross-reactivities of 0·2% to T and 1·9% to androstenedione. All samples listed below were analysed in a single assay, and thus, there was no inter-assay variation. The kit intra-assay coefficient of variation was <4·6% for higher sample quality controls (>0·29 ng mL−1). We found that DHT levels in AGS breeding males (mean ± 1 SE: 1·53 ng mL−1 ± 0·38, N = 15), nonbreeding, pre-hibernation males (1·28 ± 0·18, N = 10), and nonbreeding, pre-hibernation females (1·09 ± 0·239, N = 6) tended to be higher than those in pregnant (0·46 ± 0·14, N = 6) and lactating females (0·15 ± 0·03, N = 3) (F4,35 = 2·58, P = 0·054). In CGS, DHT levels were uniformly low and about 16–20% of those in comparable classes in AGS (breeding males 0·24 ± 0·02, N = 11; nonbreeding, pre-hibernation males 0·24 ± 0·01, N = 10, and nonbreeding, pre-hibernation females 0·22 ± 0·10, N = 14). Thus, plasma levels of DHT in AGS may account for up to 20% of androgen levels in breeding males and pre-hibernation males and females (Figs 2 and 3).

Figure 2.

 Seasonal changes in testes mass (a) and androgen levels (b) in male arctic and Columbian ground squirrels (means ± 1 SE).

Figure 3.

 Seasonal changes in androgen levels (means ± 1 SE) in female arctic and Columbian ground squirrels as a function of reproductive condition and age class. (Significant differences within Arctics are indicated by different lower case superscripts and within Columbians by different upper case superscripts).

To assess the androstenedione levels, we used an enzyme immunoassay kit (DRG EIA-3265; DRG International, Marburg, Germany) with the antibody having cross-reactivities of 0·1% to T and <0·01% to DHT. All samples listed below were analysed in a single assay, and thus, there was no inter-assay variation. The kit intra-assay coefficient of variation was 9·1% for low sample quality controls (0·3 ng mL−1) and <5·7% for higher samples quality controls (>2·6 ng mL−1). In AGS, androstenedione levels in breeding males (4·46 ng mL−1 ± 0·55, N = 10), nonbreeding, pre-hibernation males (4·60 ± 0·39, N = 24), and nonbreeding, pre-hibernation females (3·14 ± 0·60, N = 17) were significantly higher than those in pregnant (0·46 ± 0·141, N = 6) and lactating females (0·38 ± 0·13, N = 6) (F4,58 = 10·57, < 0·0001). In CGS, levels were uniformly low and about 4–10% of those in comparable classes in AGS (breeding males 0·18 ± 0·03, N = 9; nonbreeding, pre-hibernation males 0·22 ± 0·65, N = 4; and nonbreeding, pre-hibernation females 0·32 ± 0·61, N = 6). Thus, plasma levels of androstenedione in AGS may account for about 20% of the androgen levels in breeding males and pre-hibernation males and females (Figs 2 and 3). Given that a significant proportion of the androgen steroid measured by the original T antibody was not only androstenedione and DHT but also T, we will refer to these steroids collectively as androgens.

Statistical analysis

Data are reported as means ± 1 SE. The androgen data were transformed by log (x + 1) to make the variances homogeneous; testes mass was not transformed. We carried out the following analyses: (i) analyses of variance (anova) to test for differences in androgen levels and testes masses with season; (ii) a repeated measures anova in males and in females to test for the impact of the hormonal challenge on androgen levels; (iii) one-sample t-test to assess the impact of the injection (whether the difference in concentrations between the injection and the Base bleed were different from zero); and (iv) because of reduced sample size, a Mann–Whitney test to test for differences in androgen levels between sham + GDX and ADX + GDX males. We used the Tukey–Kramer multiple comparison post hoc test to examine for significance differences between main effects. StatView software was used for all analyses (version 5.0.1; SAS Institute Inc., Cary, NC, USA).


Study 1: Seasonal changes in males

Testes mass changed markedly over the active season in both species [AGS (entire active season): F4,75 = 204·84, < 0·0001; CGS (analysed only the pre-breeding and breeding periods): t23 = −10·04, < 0·0001] (Fig. 2a). Within the period from emergence to ca. 2 weeks later near the end of the breeding period, testes mass declined 52% and 68% in AGS and CGS, respectively. By the late summer, testes (now regressed and abdominal) in adult AGS and CGS only 9% and 7%, respectively, of their mass at emergence. In late summer, testes mass of juvenile AGS and CGS were 49% and 12%, respectively, of that in adults.

We were only able to make statistical comparisons in testes masses between species for the pre-breeding and breeding periods and carried out a 2-way anova (species × period). There was no species effect (F1,48 = 0·60, P = 0·80), a period effect (F1,48 = 207·11, < 0·0001) and an interaction effect (F1,48 = 6·15, P = 0·017). The period effect was a result of the decline in testes mass between the pre-breeding and breeding periods already examined above. The interaction effect occurred because although both species were similar in testes mass in the pre-breeding period (t22 = −1·58, P = 0·13), AGS had marginally larger testes in the breeding period (t26 = 2·00, P = 0·06) (Fig. 2a). Although we could not compare the two species in late summer (having to rely on published data from another site for CGS), the testes masses of adult CGS were 86% of that in adult AGS (juveniles were 22%). Thus, the testes masses and their changes over the active season were broadly similar in the two species.

Androgen levels also changed markedly over that active season in both species (AGS: F4,59 = 12·64, P < 0·0001; CGS: F4,47 = 16·76, P < 0·0001) (Fig. 2b). Within ca. 2 weeks of emerging, levels in AGS had declined only 13% from the pre-breeding to the breeding period and a further 18% to the post-breeding period. In contrast, levels in CGS had declined 47% from the pre-breeding to the late breeding period. Levels in AGS were exceptionally high, being 6 and 10 times those of CGS during the pre-breeding and breeding periods, respectively. By late summer, these marked differences remained, with levels in AGS adults still being 59% of what they were at emergence, whereas in CGS adults, levels were only 14% of what they were at emergence. Differences between juveniles in the two species prior to hibernation were similar to those between adults, with levels being 53% and 16% in AGS and CGS, respectively, of what levels were in pre-breeding adult males (CGS yearlings were similar to juveniles).

To make comparisons between species (2-way anova– species × period), we pooled pre-hibernating yearling and juvenile categories in CGS as there was no difference between these two pre-reproductive classes and sample size in juveniles was low. There was a marked species effect (F1,93 = 1090·74, P < 0·0001), a period effect (F3,93 = 39·42, P < 0·0001) and no interaction effect (F3,93 = 1·92, P = 0·13). Over the entire active season, androgen levels in AGS were on average about 10 times those in CGS (Fig. 2b). The differences between the species were marked during the pre-hibernation period when adult and juvenile male AGS had levels 25·8 and 20·1 times those of adult and juvenile male CGS, respectively (comparable differences between AGS vs. CGS pre-breeding and breeding males was 6·1 and 10·1 times, respectively). Thus, male AGS had much higher androgen levels than male CGS throughout the active season.

Study 2: Seasonal changes in females

Androgen levels in AGS varied across the active season (F5,54 = 10·59, P < 0·0001), with levels always being high in all classes (>1·25 ng mL−1), with the highest being in pre-hibernation juveniles in August (>5·75 ng mL−1) (Fig. 3). An unexpected finding was that pregnant and lactating females also had high levels. Levels in CGS (obtained only in July just prior to hibernation) were low in all classes (<0·22 ng mL−1) and differing among classes (F2,34 = 6·28, P = 0·005), being lowest in subadults (Fig. 3).

To compare between species, we carried out a 2-way anova [species × age class (adults, juveniles)]. We examined only those females just prior to hibernation (August in AGS and July in CGS). In addition, in CGS we pooled subadults and juveniles [although they differed statistically (Fig. 3), these differences were minor (average difference of 0·16 ng mL−1) and functionally, both were pre-reproductive]. There was a marked species effect (F1,67 = 441·96, P < 0·0001), an age effect (F1,67 = 20·40, < 0·0001) and an interaction effect (F1,67 = 27·22, P < 0·0001). Overall, AGS females had androgen levels 24 times those of CGS females (3·46 ± 0·38 ng mL−1, N = 34 vs. 0·14 ± 0·02, N = 37). The significant effect of age class and the interaction between species and age class were a consequence of AGS juveniles having levels ca. 2·3 times those of adult AGS whereas CGS juveniles plus subadults showed the opposite relationship, being only 0·7 times those of adult CGS. Thus, AGS females have high androgen levels throughout the active season, but particularly prior to hibernation, whereas CGS females have extremely low levels prior to hibernation.

Study 3: Response to hormonal challenge in pre-hibernating AGS

We carried out a repeated measures anova to assess whether adults and juveniles differ in their response to the hormone challenge (Fig. 4). Females and males were analysed separately as gender affects adrenal androgen production (Conley, Pattison & Bird 2004). In females, juveniles had overall levels 2·2 times those of adults (6·20 ± 0·36 ng mL−1 vs. 2·82 ± 0·18; F1,13 = 16·61, P = 0·001). The within-subject effect indicated a significant response to the injections (F4,52 = 2·86, P = 0·03), but no interaction effect (F4,52 = 1·94, P = 0·11). DEX caused a 16% decline from the Base sample and ACTH a 17·9% increase from the DEX to the P30 sample. Given the variation, we assessed whether the decline with DEX and the increase with ACTH were real by carrying out additional analyses on these changes specifically. The decline from Base to DEX was significant (repeated measures anova: adults vs. juveniles F1,13 = 26·83, P = 0·0002; Base vs. DEX F1,13 = 9·58, P = 0·0085; interaction F1,13 = 0·007, P = 0·93) as was the increase from DEX to P30 (adults vs. juveniles F1,13 = 19·03, P = 0·0008; DEX vs. P30 F1,13 = 6·19, P = 0·027; interaction F1,13 = 2·68, P = 0·13).

Figure 4.

 Late summer changes in androgen levels (means ± 1 SE) in nonbreeding arctic ground squirrels in response to a hormonal challenge protocol: Base denotes initial levels, dexamethasone (DEX) indicates levels 2 h after a DEX injection, and P30, P60 and P120 indicate levels 30, 60 and 120 min after the adrenocorticotropic hormone (ACTH) injection.

In males, juveniles had overall levels that were similar to those of adults (6·20 ± 0·27 ng mL−1 vs. 7·27 ± 0·37, respectively) (F1,18 = 1·12, P = 0·30). The within-subject effect indicated a significant response to the injections (F4,72 = 4·05, P = 0·005), but no interaction effect (F4,52 = 0·14, P = 0·97). DEX caused a 10·4% decline from the Base sample and ACTH a 19·4% increase from the DEX to the P30 sample. To assess whether the decline with DEX and the increase with ACTH were real, we carried out an analysis on these changes. The decline from Base to DEX was significant (repeated measures anova: adults vs. juveniles F1,18 = 1·37, P = 0·26; Base vs. DEX F1,18 = 10·76, P = 0·004; interaction F1,18 = 0·85, P = 0·37) as was the increase from DEX to P30 (adults vs. juveniles F1,18 = 1·43, P = 0·25; DEX vs. P30 F1,18 = 7·76, P = 0·012; interaction F1,18 = 0·13, P = 0·72). Thus, the response of androgens in both females and males was inhibition by DEX and stimulation by ACTH. In addition, adult females had levels about half those of the other three groups.

Study 4: Source of androgens in pre-hibernating male AGS

GnRH had no effect on androgen levels relative to saline (Fig. 5a, repeated measures anovaF1,11 = 1·44, = 0·25). This lack of androgen increase could indicate changes at the level of the pituitary (the gonadotropes in the pituitary are not perceiving the GnRH signal and thus releasing luteinizing homone [LH] and follicle-stimulating hormone [FSH]), the testes (the Leydig cells are not perceiving the LH signal) or both. Prior to GDX ACTH increased androgen levels significantly (by 44·3%) relative to saline (F1,11 = 20·37, = 0·0009). After GDX, ACTH increased levels by 41·6% relative to saline (F1,9 = 40·15, = 0·0001). The within-subject response to ACTH was similar before and after GDX (F1,9 = 3·42, P = 0·10). We then adrenalectomized (ADX) half the males. Shams + GDX had five times the androgen levels of the ADX + GDX animals (Fig. 5b, Z = −2·134, = 0·033). Thus, the adrenals, not the testes, must be the source of androgens.

Figure 5.

 Changes in plasma androgen levels (means ± 1 SE) in nonbreeding, pre-hibernating male arctic ground squirrels in response to experimental manipulations. All saline injections and the gonadotropin-releasing hormone (GnRH) injection (a) resulted in nonsignificant changes to androgen levels. The ACTH injection resulted in a significant increase both before and after all males were gonadectomized (GDX) (a) (N = 12). Adrenalectomy (ADX) (b) of half of these animals (two died before a blood sample could be obtained) resulted in a significant androgen decline (GDX + sham ADX, N = 6; GDX + ADX, N = 4).


The exceptionally high androgen levels seen in AGS males at all times (Fig. 2) have no parallel in any other ground squirrel species (nor any other previously studied mammal). During the breeding season, when androgen levels (T normally) are typically high, those in AGS males were between 2–12 times those in other ground squirrel species (Ellis, Palmer & Balph 1983; Holekamp & Talamantes 1992; Barnes 1996; Millesi et al. 2002; Strauss et al. 2007; Delehanty & Boonstra 2009). During the pre-hibernation period, when androgen levels (T normally) are typically low, those in AGS males were between 10 and 200 times those in other species (Fig. 2; Nunes et al. 1999; Strauss et al. 2007; Barnes 1996). In females (Fig. 3), no one working on other hibernating ground squirrel species has analysed androgen levels during the pre-hibernation period. Only Nunes et al. (2000) assayed testosterone levels in adult breeding female U. beldingi and found levels 40–100 times lower than those in adult breeding female AGS. Thus, the pattern of high androgen levels in AGS at all times, but particularly prior to hibernation, is unique among the hibernating mammals and raises two questions. First, are the high levels in breeding season related functionally to those in the pre-hibernation period? Second, why are levels so high in the pre-hibernation period? We address both of these after we identify the source as both may be related.

There is one caveat that must be addressed first before we discuss the results. Did the trapping and handling protocol prior to bleeding affect the androgen levels in these squirrels and, more importantly, did it compromise our findings? The answer to the first question is yes and, we argue, to the second no. There are now numerous wildlife studies showing that plasma glucocorticoid (GC) levels increase significantly 2–5 min after capture (for references, see Sheriff et al. 2011). The typical consequence of this GC increase is an inhibition of the reproductive axis, with a decline in gonadal steroids (Wingfield & Sapolksy 2003). For example, in breeding male Richardson’s ground squirrels (U. richardsonii), a species closely related to the two we studied, Delehanty & Boonstra (2009), have elucidated the physiological impact of such a trapping and handling protocol. Along with changes in other blood variables, it caused androgen levels to decline 43% from true basal levels (from 9·42 to 5·41 ng mL−1, with the latter animals being bled a maximum of 4 h after capture). We suggest that given the results on CGS above (Figs 2 and 3), our capture protocol in CGS also caused androgen levels to be less than they would have been prior to capture, at least in breeding males. In contrast, in AGS the evidence indicates the opposite – that the stress of capture caused androgen levels to increase. From breeding male AGS collected in spring either instantaneously (by shooting) or after they had been stressed by trapping and handling (Boonstra et al. 2001b), androgen levels increased 47% from true basal levels to trap-stressed levels (9·65 to 14·20 ng mL−1). This is consistent with the evidence we report here. The ACTH challenge in nonbreeding animals in all sex and age groups (Study 3, Fig. 4) caused an increase in androgens by ≥18% by 30 min post-injection. The ACTH injections in nonbreeding males (Study 4, Fig. 5a) caused an increase in androgens by ≥41% by 2 h post-injection. In summary, the evidence suggests that our capture protocol would have stressed the squirrels, but in so doing, served to exacerbate the differences between AGS and CGS (i.e. the levels in AGS would be higher than those prior to capture whereas in CGS they would be lower). However, the capture protocol served to reinforce our interpretation that AGS have an androgen profile and a response to stressors that is singular among mammals.

During the pre-hibernation season, both studies 3 and 4 indicate that the source of the high levels of androgens in AGS males must be the adrenals, not the testes. It is clear that the testes are quiescent at this time, as neither GnRH (which should prompt the testes to release T) nor GDX affected androgen levels (Fig. 5a). ACTH stimulation of androgen production (Fig. 4) is consistent with the adrenals being the source of the androgens. The evidence cited above (on shot vs. trapped breeding male AGS, Boonstra et al. 2001b) suggests that the adrenals may also be the source for a high portion of the androgen levels seen in breeding males (Fig. 2) and possibly in breeding females (Fig. 3). ACTH is known to stimulate the adrenal androgens, such as androstenedione and dehydroepiandrosterone (DHEA) (Odell & Parker 1984), but we know of no study that implicates ACTH in stimulating adrenals to produce significant amounts of plasma T (Vinson, Bell & Whitehouse 1976). In GDX guinea pigs (Cavia aperea f. porcellus) but not rats (Rattus norvegicus), ACTH caused adrenal T to increase, but plasma T remained nondetectable (i.e. T is converted within the adrenal before release into the blood: Bélanger et al. 1990). However, in AGS the picture is complicated by two other results. First, DEX, which inhibits ACTH release by the pituitary because of its negative feedback effects, could cause androgen levels to fall, but these decline only an average of 12% (Fig. 4). In contrast, for the same males presented in Fig. 4, DEX caused free plasma cortisol to decline >95% (Boonstra, McColl & Karels 2001a; Fig. 3). For the females presented in Fig. 4, the free cortisol decline was 98% (R. Boonstra, unpublished data). Thus, ACTH must have declined markedly in response to DEX. Our failure to find a marked decline in androgens following DEX is consistent with findings in humans (Odell & Parker 1984) and rhesus monkeys (Zhou et al. 2005), indicating that adrenal androgens respond very slowly to the lack of ACTH stimulation. Second, although ADX + GDX resulted in a significant reduction in androgen levels (Fig. 5b), it did not cause them to fall to nondetectable values. In contrast, following ADX + GDX in rats, T levels declined to nondetectable levels within 12 h (Kniewald, Danisi & Martini 1971). There are two possibilities. First, sex hormone binding protein (SHBP) in plasma may sequester T (and other androgens) and preventing degradation by the liver and reducing metabolic clearance rates for this class of steroids. The presence of this binding protein in mammals varies, with some rodents having SHBP (such as hamsters –Mesocricetus auratus) but others not (rats and guinea pigs) (for a review see Petra 1991). Evidence from our laboratory (B. Delehanty, et al. unpublished data) indicates that ground squirrels have SHBP. Second, fat tissue is known to be a steroid hormone reservoir (Deslypere, Verdonk & Vermulen 1985) and, as fat constitutes up to 30–35% of the mass of ground squirrels prior to hibernation (e.g. Young 1990; Buck & Barnes 1999a), it may buffer androgen disappearance.

To our knowledge, in no other mammal do the adrenals function the way they do in AGS. However, relative to other mammals, this ability is more one of degree, rather than potential. The zona reticularis of the adrenal cortex is the site of production of adrenal androgens and can potentially produce a range of them [androstenedione, T, dehydroepiandrosterone (DHEA) and its sulphated form, DHEA-S; Vinson, Whitehouse & Hinson 2007]. Typically, the levels of plasma T having an adrenal origin are very low (Vinson, Bell & Whitehouse 1976). When high concentrations of adrenal androgens are produced, as they are in humans, some primates and at least one rodent (i.e. red squirrels –Tamiasciurus hudsonicusBoonstra et al. 2008), it is DHEA. However, the low cross-reactivity of DHEA to the T antibody in our study precludes this. The obvious question with respect to AGS is what ecological need has prompted the evolution of this androgen solution? There are at least two potential explanations for these high levels during the pre-hibernation period. First, it may underlie the late summer–autumn aggressive behaviour needed to defend resources (hibernacula and seed caches) from competitors. Second, it may act as the anabolic steroid needed to lay down muscle mass in summer that is subsequently required to sustain protein catabolism and permit glucose production overwinter during hibernation under deep freeze conditions.

First, Carl (1971) reported increased territoriality in defence of the hibernaculum among both sexes and age classes of AGS in the fall on the north coast of Alaska. However, in the Kluane area, only males were reported to defend hibernacula (Lacey 1991). In the case of males, maintaining territories and hibernacula near females will advantageously position them for the ensuing competition in the next spring (Carl 1971; Barnes 1996; Boonstra, McColl & Karels 2001a). In addition, males, but not females, are defending their underground seed cache in late summer. However, males of other ground squirrel species have the same needs for the defence of their hibernacula and seed caches (Michener 1990), yet this defence is not dependent on high androgen levels. In the case of AGS males, high androgen levels may be necessary for pre-hibernation aggressive behaviour, but it is not sufficient, nor does its presence motivate the Kluane females to be similarly aggressive. Thus, we think this explanation is unlikely to be correct.

Critical to the second explanation is that muscle mass increases markedly prior to hibernation and then declines during hibernation. The mass dynamics in adult males cannot be used, because although they increase their lean body mass by 22·4% over the active season (C.L. Buck & R.W. Fridinger, unpublished 2011), they appear to lose nothing overwinter (Buck & Barnes 1999b). This is a direct result of the consumption of their seed cache when they arouse and become continuously euthermic up to 3 weeks prior to their emergence, replacing all the mass lost during hibernation. In contrast, adult female AGS increase their lean body mass by 28·6% prior to hibernation (C.L. Buck & R.W. Fridinger, unpublished 2011) and this is matched by a 33·6% loss in lean mass overwinter (Buck & Barnes 1999b). In contrast, in adult male and female CGS, Young (1988) estimated that the percentage protein loss overwinter (which then necessitates replacement in the active season) is negligible (0·9–5·4%). In addition, neither U. beldingi (Morton 1975) nor Callospermophilus lateralis (Jameson & Mead 1964) shows any change in the adult lean body mass over the active season. Thus, of those species examined, only AGS catabolize significant amounts of muscle during hibernation that must then be replaced during the active season.

The evidence that the second explanation is most likely is as follows. First, at hibernation temperatures of −16°C the metabolic rate in AGS is 16 times that at 0°C (Buck & Barnes 2000; see Karpovich et al. 2009 as well). No other hibernating ground squirrel is exposed to such temperatures, and of those exposed to <0°C, some individuals die (e.g. in C. lateralis, all animals exposed to down to −2°C died –Wit & Twente 1983 whereas in C. saturatus, all survived down to −2°C –Geiser & Kenagy 1988). Thus, the energetic demands on AGS are much greater. To sustain hibernation under these conditions, adult female loose an average of 0·5 g of lean mass (protein) and 0·6 g of fat per day (Buck & Barnes 1999a). Second, there is a shift in the respiratory quotient in AGS from 0·7 when only lipid is being catabolized to >0·85 at low temperatures (−16 to −20°C) when nonlipid is being catabolized (Buck & Barnes 2000; Karpovich et al. 2009). Finally, although lipid catabolism, the primary fuel of fat-storing hibernators (Dark 2005), can produce sufficient glucose from the glycerol molecule of the triglyceride to meet glucose needs under normal hibernacula temperatures of 0°C, glucose production from gluconeogenesis through protein catabolism is required at sub-zero temperatures (Galster & Morrison 1975, 1976). The brain, heart, erythrocytes and nonshivering thermogenesis in brown adipose tissue all require glucose to survive (Gibbons, Denton & McCormack 1985; Cooney et al. 1986; Dienel & Hertz 2001). We propose that the most parsimonious explanation is that high androgen levels in AGS during the active season are necessary for the growth of the lean muscle mass that is required to sustain hibernation under arctic winter conditions. However, testing this hypothesis will require experiments that show that adrenal androgen production indeed causes an increase in lean mass and that this enhances overwinter survival. One possible approach involves androgen antagonists (e.g. flutamide or cyproterone acetate) that prevent androgens from having their effect.

The higher androgen levels in juvenile than adult females (Figs 3 and 4) may be related to their greater need to build up protein stores prior to hibernation. First, their overwintering hibernacula are colder than that of adult females (Buck & Barnes 1999a). Second, although adult and juvenile females attain a similar body weight just prior to immergence into hibernation (Buck & Barnes 1999b), the body mass of adult females increase about 1·7 times between spring emergence and autumn immergence whereas the body mass of juvenile females increase three times between emergence from their natal burrow and autumn immergence. It takes juvenile females about 1–1·5 months longer to accumulate this mass (McLean & Towns 1981).

It is clear that AGS, but not CGS, are producing significant quantities of a variety of androgens, particularly prior to hibernation, and some may have tissue effects directly (T and DHT) and one when enzymatically converted to T (androstenedione – a pro-hormone). In pre-hibernating AGS, high T could compromise winter survival with AGS being less able to cope with disease threats and by expending energy at a time when the conservation of resources is critical to meet the demands of the coming winter. Possibly, there is a down-regulation of the androgen receptor in most tissues at that time, so only the muscles are sensitive to the high plasma androgen levels. Consistent with this explanation is that in pre-hibernating adult males and, to a lesser extent, in juvenile males, immunosuppression does not occur in the face of high androgen levels (Boonstra, McColl & Karels 2001a). Additional research will be required to determine whether we are correct.

The second line of research that would be to assess the universality of this solution to sub-zero soil winter temperatures by examining other species hibernating under similar conditions. There are three potential candidates. The Alaska marmot Marmota broweri is found in the Brooks Range and nearby areas (latitude of ca. 68°N). From a preliminary study (Lee, Barnes & Buck 2009; temperatures assessed at one hibernaculum overwinter), it experiences similar conditions to those of AGS. However, unlike AGS, it hibernates communally and this may ameliorate the metabolic demands. In arctic Russia, hibernating M. camtschatica can experience soil temperatures at hibernacula level down to −22°C (Vasil’ev 2000). Finally, in North America the least chipmunk (Tamias minimus) has a distributional range extending from west central Yukon to the shores of the Hudson Bay and south to north-eastern Arizona (Verts & Carraway 1991). In the north, it may experience similar conditions to those of AGS and thus may have evolved a comparable solution, whereas in the south, it may resemble that of CGS. Such population-specific adaptations are not unusual in nature. Virtually all species are collections of genetically differentiated populations that have evolved to deal with unique environments (with respect to either climatic differences between regions or their interactions with other species in the local area; reviewed in Thompson 2005). In addition, chipmunks meet their winter energy requirements with a seed cache and this may mitigate the need to catabolize internal protein.

In conclusion, we propose that AGS have solved the problem of hibernating in sub-zero temperatures by ramping up the adrenal production of key androgens. We hypothesize that at the tissue level it must ultimately be T, as only T stimulates muscle growth (Mooradian, Morley & Korenman 1987) and muscle protein is required for gluconeogenesis to supplement that coming from the glycerol when lipid is catabolized. This trait possibly originated sometime during the Pleistocene in response to the repeated cycles of glaciations that commenced about 1·8 million years ago, producing permafrost soils. Molecular evidence indicates that AGS separated from related, southern species between 1·3 and 2·5 million years ago (Harrison et al. 2003). Unlike all the other ground squirrel species that were driven south with each glacial advance (Pielou 1991), AGS may have persisted in the ice-free Beringian refugia (Eddingsaas et al. 2004).


We thank Loren Buck for access to the unpublished Ph. D. thesis of R. Fridinger; S. Boutin, T. Jessop, T. Karels and K. Soma for comments; Hugh and Dick Bradley at the Pelly River Ranch (Yukon), Andy Williams at the Arctic Institute of North America at Kluane Lake (Yukon), and Ed Johnson and Judy Mappin of the Biogeosciences Institute, University of Calgary (Alberta) for use of field facilities; and J. Castillo, A. Hubbs, E. Lacey, L. Lu, C. McColl and L. Desantis for assistance in the field and laboratory. This article benefitted greatly from the critical comments of three anonymous reviewers and of Keith Sockman, associate editor. This research was supported by grants from the Natural Sciences and Engineering Research Council of Canada (RB), the Department of Indian Affairs and Northern Development (BD) and the University of Queensland (AJB).