Behavioral and endocrine correlates of reproductive failure in social aggregations of captive wolverines (Gulo gulo) pot

Behavioral and endocrine correlates of reproductive failure in social aggregations of captive wolverines ( Gulo gulo ) F. Dalerum 1 , S. Creel 2 & S. B. Hall 3 1 Department of Zoology, Stockholm University, Sweden 2 Department of Ecology, Montana State University, Bozeman, MT, USA 3 Division of Recovery, Endangered Species Division, US Fish and Wildlife Service, Portland, OR, USA Keywords social stress; reproductive success; sociobiology; carnivore; endocrinology. Correspondence Fredrik Dalerum, Department of Zoology, Stockholm University, SE-106 91 Stockholm, Sweden. Tel: +46-8-16-40 01; Fax: +46-8-16-77-15 Email: fredrik.dalerum@zoologi.su.se Received 19 April 2005; accepted 1 November 2005 doi:10.1111/j.1469-7998.2006.00116.x Abstract Sociality in mammals is often viewed as a dichotomy, with sociality contrasted against solitariness. However, variation within these broad categories may have strong effects on individual fitness. For example, reproductive suppression of social subordinates is generally associated with group living, but suppression may also occur in solitary species if the behavioral and physiological processes involved can be modulated by the demographic environment. To investigate whether behavioral and physiological traits that normally are associated with group living might be latent even in a solitary species, we explored the level of sociality and investigated causes and mechanisms of reproductive failure in female wolverines Gulo gulo that experienced a highly aggregated social environment in captivity. Behaviorally, females showed low levels of aggression and intermediate levels of social interactions. Reproductive failure seemed to have been related to low social rank and to have occurred between ovulation and implantation in 13 out of 15 breeding attempts. However, three of eight females observed to mate produced offspring, indicating that no individual female fully managed to monopolize breeding. Reproductive failure was not related to elevated levels of glucocorticoid stress hormones. Rather, elevated glucocorticoid levels during the mating season were associated with successful reproduction. We suggest that social tendencies and physiological mechanisms mediating reproductive suppression may be viewed as reaction norms to the social environment. We further suggest that the social flexibility of solitary carnivores might be greater than is commonly observed, due to ecological constraints that may limit aggregation. Introduction Sociality in mammals is often viewed as a dichotomy, with various forms of sociality contrasted to solitary living. However, even individuals of solitary species engage in social interactions, although their interactions are less com- mon and elaborate than in species that form cohesive social bonds (Leyhausen, 1965). Hence, some authors favor an alternative approach, which begins by describing the spatial structures of individuals, and then treating social interac- tions as a function of these spatial structures (Sandell, 1989). There are many examples of intraspecific variation of the spatial organization of individuals. In these species, the spatial organization within populations is affected by many variables, such as the amount and spatial distribution of resources, competitors and predators (Moehlman, 1989; Johnsson, Macdonald & Dickman, 2000). Such social flex- ibility may exist even in species where it is not observed, if sociality is constrained by ecological factors. Many of the behavioral and physiological traits found in complex social societies might then be found in species with solitary social structures, perhaps in less well-developed forms. Reproductive suppression of socially subordinate indivi- duals is commonly found among group-living mammal females (Wasser & Barash, 1983; Jennions & Macdonald, 1994). Such suppression of subordinates can be directly caused by dominant individuals killing offspring (Rood, 1990) or inhibiting mating opportunities (Creel et al., 1992; Clutton-Brock et al., 1998a; Ebensberger, 1998). Suppres- sion can also be caused by physiological mechanisms that disrupt the endocrine events that control ovulation and implantation (Creel, 1996; Faulkes & Abbott, 1997) or by mechanisms that terminate established pregnancies during gestation (Hackl ¨ ander, M ¨ ostl & Arnold, 2003). We will here use the term ‘physiological suppression’ for these latter two cases, that is when a subordinate is physiologically pre- vented from reproducing, either pre- or post-mating, due to the presence of dominants. Physiological suppression is, however, rarely observed in solitary societies, and is in some cases assumed to be absent (e.g. Creel & Macdonald, 1995). Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London 527 Journal of Zoology. Print ISSN 0952-8369 However, if individuals of solitary species experience aggre- gated social situations, it is possible to test if the physiologi- cal mechanism that mediates suppression involves latent behavioral or physiological traits that can be modulated by the social environment. From a proximate standpoint, glucocorticoids (GCs) have been suggested as a mediator of physiological suppression (Schoech, Mumme & Moore, 1991; Blanchard et al., 1995). GCs are steroid hormones produced by the adrenal cortex and are released in response to physical or physiological stressors (Sapolsky, 2002). Among other effects, long-term elevation in GC levels can suppress reproduction (Dobson & Smith, 2000). However, long-term elevations in GCs also have a wide range of negative effects, such as suppressed immune function and increased blood pressure. It has there- fore been argued that non-GC-mediated mechanisms of reproductive suppression would be evolutionarily favored in species for which reproductive suppression is a part of the normal social organization (Creel, Creel & Monfort, 1996). This argument is supported by field data, because there is an increasing array of evidence that it is more common for elevated GC levels to be a cost of dominance than of subordination in free-living populations (Creel, 2001). How- ever, if subordinates or normally solitary species cannot escape the presence of dominant individuals, stress responses might drive reproductive suppression. This situation might arise, for example, in a high-density rodent population. In such situations, GC-driven mechanisms are a plausible, and even expected, mediator of reproductive suppression. The wolverine Gulo gulo is a large mustelid that inhabits the northern boreal and arctic zones of North America and Eurasia (Pasitschniak-Arts & Larivie ` re, 1995). It is a solitary species with reflex ovulation and a delayed implantation period of c. 5–6 months (Mead et al., 1993; Banci, 1994). In the wild, wolverine males spatially overlap both other males and several females, whereas territories of reproductive females only overlap territories with males and recent off- spring (Powell, 1979; Hornocker & Hash, 1981; Magoun, 1985; Banci & Harestad, 1990). Social groupings other than mating pairs and mother and infants are very rarely observed. To investigate whether behavioral and physiological traits that normally are associated with group living might be latent even in a solitary species, we explored the level of sociality and investigated causes and mechanisms of repro- ductive failure in female wolverines that experienced a highly aggregated social environment in captivity, with surplus food and no predation. We predicted that social interactions would generate a social hierarchy, and that reproduction would be heavily skewed toward specific, top- ranked individuals. We further predicted that the lack of possibility for subordinates to escape dominant individuals would cause an elevated stress response in subordinate individuals, and that this would result in subordinate reproductive failure. We pursued the following specific questions: (1) How much time do females engage in social interactions? (2) Is there individual variance in female reproductive success? (3) Is any such variance related to social rank of the females? (4) If reproduction fails, at what stage of the reproductive process does it occur? (5) If reproduction fails, is reproductive failure related to an endocrine stress response (i.e. elevated GC levels)? Methods Animal housing and monitoring of reproduction We conducted the study at a private facility in Washington State, USA. The animals at the facility were mainly or- phaned individuals that were housed in captivity for huma- nitarian reasons. The study was conducted under a license from the State of Washington, USA. Animals were kept in three outdoor enclosures, each c. 1500 m 2 in size. The enclosures consisted of mature conifer and hardwood forest enriched with water bodies and logs for climbing. The animals were fed commercial mink food mixed with offal provided by local elk and deer hunters. The animals had ad lib access to food and water, and were occasionally provided with elk and deer heads, hoofs and legs to play with and chew upon. During the mating seasons (May–August) of 1995, 1997 and 2001, study groups of 12–15 females and five to six males were housed in one of the 1500 m 2 enclosures. During the mating season of 1996, two groups were enclosed (again in two of the 1500 m 2 enclosures), each with five to six females and three to four males. Animals were normally housed under these conditions and not specifically grouped together for the study. Animals that required special atten- tion for medical reasons or that were believed to be pregnant (see below) were housed in individual pens, 3 Â 6 m in size. These pens were situated in semi-open buildings (i.e. with a solid roof but only wire mesh walls) immediately adjacent to each of the enclosures, so that separated individuals were within sight, smell and sound of both each other and the remaining group, as well as of environmental stimuli such as photoperiod. These pens had concrete floor covered with sawdust and were cleaned daily. Each pen had a small sleeping box of c.1m 3 , padded with straw. All but two of the females included in the study were born wild in Canada. These two were born at the facilities, one by a female not included in the study and one by an included study female (Table 1). All wild-born females were caught at different locations and can thus be assumed to have been unrelated. During each mating season, females were ob- served daily for signs of mating activities. Except for beha- vioral data on one female (see Behavioral data), we only included data on females that had been observed to copu- late, as several of the other females were of unknown age and could have been sexually immature. Females observed to have copulated were separated during the fall (from Septem- ber to December) and housed in individual pens as described above but with a slightly larger nest box, which was subdivided into one interior and one exterior compartment. Each nest box had an infrared camera that was fitted to the interior compartment and linked to a monitor in an adjacent Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London528 Reproductive failure in wolverines F. Dalerum, S. Creel and S. B. Hall building. At the time of expected parturition, nest boxes were inspected daily through the remote sensing system to determine day of parturition. We followed the reproductive outcome of 20 breeding attempts where copulation was confirmed in eight females during four breeding seasons, 1995/1996, 1996/1997 and 1997/1998 and 2001/2002 (Table 1). Throughout the paper, we have used the term ‘breeding attempt’ to describe the entire time period from mating season (in which copulations had been observed) through d elayed implantation, gestation and lactation. Behavioral data During the 1996 mating season we collected behavioral data on five females, of which two were housed in one enclosure and three in another. During 1997, we collected data on four of these females. In this mating season all were housed in the same enclosure. During 1996, we collected data from 18 May to 25 August, and during 1997 from 10 May to 20 July. This provided a total of 135 h of behavioral observa- tions. Two females observed in 1996 produced kits the subsequent spring, and one produced kits also after the mating season 1997. None of the other three females pro- duced kits, although two were observed to copulate. The female not observed to copulate was regarded as an unsuc- cessful reproducing female in the analyses because at the time she was at least 3 years old, and hence should have been sexually mature (Persson, 2003). During 1996, both of the successfully reproducing females were housed in the same enclosure. Each animal was observed for 1 h three times a week, using scan sampling at 30 s intervals. Each 1 h scan sampling session was carried out either between 9:00 and 12:00 h or between 18:00 and 21:00 h. We classed behaviors into five categories: (1) inactive (any inactive behavior such as sleeping or resting), (2) moving (any movement, such as walking, running or pacing), (3) social interactions (any clear social interaction between two or more individuals), (4) grooming (grooming behavior), (5) other, including dig- ging, carrying objects and swimming (any other behavior). These definitions were based on pilot observations of the study animals. In addition to scan sampling, we recorded all occurrences of social behavior and scent marking during focal hours. For social interactions, we recorded the gender, and if possible the identity, of the initiator of the interaction. Scent marking was divided into four categories: defecation, urina- tion, pelage rubbing and abdominal rubbing. Pelage rubbing is a distinct behavior in which the animal vigorously rubs one of its sides on an object or on the ground. Abdominal rubbing is an equally distinct behavior in which the animal drags or rocks its abdomen against a surface or object. Fecal collection and steroid extraction We collected fecal samples from 15 breeding attempts by eight females for endocrine analyses (Table 1). The number of analyzed fecal samples per breeding attempt ranged from 6 to 54 (mean =21, SD =11). The data were collected during the 1995/1996, 1996/1997 and 2001/2002 breeding seasons. During the 2001/2002 breeding season, we only collected fecal samples from January to May. During periods when females were communally housed, we collected feces oppor- tunistically whenever a female was observed to defecate. In addition, we hand-fed females with food items labeled with colored plastic pellets and other identifying markers to facilitate identification of feces. During periods when fe- males were housed in individual pens, feces were collected as part of the daily maintenance of the facilities. Each scat was frozen immediately after collection and stored in À20 1C until further analyses. We derived endocrine data non-invasively from enzyme- linked immunosorbent assays (EIA) and radioimmuno assays (RIA) on the fecal material. We validated the assays according to standard criteria, described below for each assay (Cekan, 1975). A total of 580 feces was used for these analyses. We extracted steroids using previously published meth- ods (Creel et al., 2002). Each sample was homogenized, and a subsample was dried in a vacuum evaporator. The dried sample was pulverized, and a subsample of c. 0.2 g was boiled for 20 min in 10 mL of absolute ethanol. After Table 1 Female wolverines Gulo gulo included in the study, their origin and year of birth, number of breeding attempts followed (with copulation confirmed), number of litters produced, number of mating seasons with behavioral data and number of breeding attempts with endocrine data Female Year born Origin Number of breeding attempts Number of litters produced Number of seasons with behavioral data Number of breeding attempts with endocrine data 52 1985 Wild caught 3 2 a 22 54 1984 Wild caught 2 0 0 2 56 1994 Wild caught 3 0 0 2 62 1986 Wild caught 3 1 2 2 66 Before 1995 Wild caught 3 0 2 2 9406 b 1994 Captive born 0 0 2 0 9500 Before 1995 Wild caught 4 3 1 3 01-1 Before 1984 Wild caught 1 0 0 1 01-2 c 1996 Captive born 1 0 0 1 a Gave birth to three litters, but killed her single kit in one of these cases. b Daughter to female not included in study. c Daughter to female #9500. Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London 529 Reproductive failure in wolverinesF. Dalerum, S. Creel and S. B. Hall centrifuging at 1500 r.p.m (504 g) for 15 min and decanting the supernatant (containing extracted steroids), the residual pellet was weighed as a measure of indigestible matter. The recovered supernatant was evaporated under air, rinsed in 1.5–2.5 mL of absolute ethanol, sonicated, vortexed and re- dried. The re-dried samples were reconstituted in 1.0 mL of absolute methanol, sonicated for 45 s and then vortexed for 1 min to free particles adhering to the tube walls. We then transferred the extracts to 2 mL cryovials with silicone O-rings for long-term storage. The methanol extracts were stored at À80 1C until analyses. Fecal progesterone assay We measured fecal progesterone using an EIA kit from R&D Systems Inc. (Minneapolis, MN, USA), with low reported cross-reactivity to other steroids (3.5% for 17-OH- progesterone, 0.77% for corticosterone and o0.10% for the other steroids tested). Serial dilutions of standards and fecal extracts gave parallel antibody binding for 7 points from 16-fold to 16 000-fold dilution. The recovery of known amounts of progesterone to fecal extracts (50 mLat 1.95–2000 pg mL À1 ) was 88 Æ 18% (y =À60+1.04x, SE b =0.02, r 2 =0.999). Fecal extracts were assayed in 200-fold dilutions following the protocol provided with the assay kit. Intra-assay variation was 8.4%, and inter-assay variation was 11.3 and 5.7% for pooled high and low controls, respectively. Sensitivity was 0.195 pg (per determi- nation, or 3.9 pg mL À1 of extract). Fecal estrogen assay We measured fecal estrogens using a commercially available double antibody 125 I RIA kit for total estrogens from ICN Pharmaceuticals (Costa Mesa, CA, USA), previously vali- dated for fecal assays with other carnivores (e.g. Creel et al., 1997). This antibody does not distinguish between estradiol- 17 a and estrone, but shows little cross-reactivity with other steroids (9.0% for estriol, 7.0% for estradiol-17 b , 2.5% for equillin and o0.10% for the other steroids tested). Serial dilutions of standards and fecal extracts gave parallel anti- body bindings for 6 points from a two-fold to a 64-fold dilution. All unknowns were run within this range of dilution. The recovery of known amounts of estrogens to fecal extracts (50 mL at 5–200 pg mL À1 ) was 88.7 Æ 8.54% (y=À0.90+0.85b, SE b =0.03, r 2 =0.993). Aliquots of 10 mL of sample extract (equivalent to a 10-fold dilution) were evaporated under air and reconstituted with dilutent buffer. Aliquots of 0.5 mL of the reconstituted samples were assayed following the instructions supplied with the assay kit. Intra-assay variation was 3.5% and inter-assay varia- tion was 9.1%. Sensitivity was 2.5 pg tube À1 . Fecal corticosterone assay We measured fecal GCs using a commercially available double antibody 125 I RIA kit for corticosterone from ICN Pharmaceuticals (Costa Mesa, CA, USA), previously vali- dated for fecal assays with other carnivores (e.g. Creel et al., 2002). The antibody shows little cross-reactivity with other steroids (0.34% for desoxycorticosterone, 0.10% for testos- terone and less than 0.10% for the other steroids tested). Serial dilutions of standards and fecal extracts gave parallel antibody bindings for 8 points from a fivefold to a 640-fold dilution. All unknowns were run within this range. The recovery of known amounts of corticosterone added to fecal extracts (50 mL at 12.5–250 pg mL À1 ) was 67 Æ 15% (y=À34.6+0.88x, SE b =0.05, r 2 =0.979). Fecal extracts were assayed in 100-fold dilutions, following the supplied protocol for the kit, with the exception of halving volumes. Intra-assay variation was 5.3% and inter-assay variation was 7.7 and 4.3% for pooled high and low controls, respectively. Sensitivity was 2.5 pg tube À1 . We evaluated the biological validity of the corticosterone assay by measuring the GC response to injections of 25 IU Cortrosyn s (Amphastar, Rancho Cucamanga, CA, USA) injected intramuscularly into the quadriceps muscle. We injected four wolverines (three females and one male) and collected all feces from 48 h before the injection until 72 h after the injection. There was a significant increase in fecal corticosterone between 12 and 72 h post-injection (pre- injection, mean Æ SE: 8.82 Æ 1.03 mgg À1 dry feces; post-injec- tion, mean Æ SE: 12.65 Æ 2.26 mgg À1 dry feces; t 3 =3.14, P=0.05, calculated on individual means), although the timing of the corticosterone peak varied among individuals. Although the GC response was significant, we believe that the measured response to the ACTH challenge was probably muted by elevated baselines in this trial, because all animals had to be shifted to new, isolated housing just before the experiment for logistical reasons. Statistics To account for non-independence of different behavioral frequencies during the scan sampling, we log-ratio trans- formed the frequencies of each behavior (i.e. number of scan events with the behavior), and used this transformed vari- able in an ANOVA to analyze behavioral time budget data (Aitchison, 1986). For social and scent marking data (col- lected by all-occurrences sampling), basic behavioral rates were expressed as observations of each behavior per obser- vation hour. We square root transformed this variable to approximate normality and used this transformed value as a response variable in an ANOVA (Crawley, 2002). In ana- lyses of behavioral data, we used breeding attempt as the smallest sample unit. We used mixed linear models to test for differences in fecal steroid levels between successful and unsuccessful breeding attempts during different phases of the breeding cycle. One female that was observed to kill her newborn kit was regarded as a successful breeder in these analyses, because she was not physiologically suppressed according to the definition described in the introduction. We divided each breeding attempt into four phases: mating, diapause (recall that wolverines show delayed implantation), gesta- tion and lactation. The mating season was defined as May, June and July, because mating was recorded during this Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London530 Reproductive failure in wolverines F. Dalerum, S. Creel and S. B. Hall entire period. Diapause was defined from the beginning of August until 50 days before parturition or expected parturi- tion (i.e. after mating and before gestation). Gestation was defined as the time from 50 days before and until the time of parturition or expected parturition, and lactation was de- fined from parturition or expected parturition until the end of April. The expected time of parturition for unsuccessful breeding attempts was calculated as the average parturition date for each specific breeding season or as the average date for all breeding seasons if a season was without any success- ful breeding attempts. To avoid temporal pseudoreplication, we used days until parturition or expected parturition as a random covariate for successful and unsuccessful breeders, respectively. Further, to avoid pseudoreplication related to multiple breeding attempts by the same female, we nested each breeding attempt within each female and added this as a random factor in the models. We included scat water content and per cent indigestible matter (indexed as weight of centrifuged pellet divided by amount of extracted dried sample) as fixed methodological covariates. The models were fitted using restricted maximum likelihood (REML). We evaluated main and interaction effects by a conditional F-test, which is appropriate when fitting models using REML (Pinheiro & Bates, 2000). For interaction effects, we used t-tests on breeding attempt means to test for differences between successful and unsuccessful breeders within each time period. Analyses of progesterone and estrogen concentrations were conducted on log-transformed endocrine data to meet the assumption of homoscedasticity. To approximate normality, we arcsine transformed scat water content (expressed as per cent water of sample before being dried) and used the transformed variable in the models (Zar, 1996). Values are presented as mean Æ SE, unless otherwise noted. All test statistical tests were two-tailed. Statistical analyses were performed with the software R, version 1.8.1 for Linux (http://www.r-project.com). Results Time budget of females The wolverines spent most of their time inactive (1996: 33.6 Æ 6.2% of scans, n=5 individual females; 1997: 54.7 Æ 3.9%, n=4) or just moving (1996: 21.3 Æ 4.6%; 1997: 27.0 Æ 4.9%; Fig. 1). They were involved in social interactions 10–20% of total time observed (1996: 19.1 Æ 3.5%; 1997: 9.9 Æ 3.1%) and 20–36% of active time (1996: 36.6 Æ 7.2%; 1997: 21.2 Æ 5.6% of the scans when animals where active). Social interactions mostly consisted of chasing and various forms of playing, and were generally accompanied by various types of vocalizations. Aggression was overall rare among the observed females. Because no clear dominant or submissive behaviors were observed, we could not utilize the data on social behavior to derive relative rank between the observed individuals. Dif- ferences in behavior for females that bred successfully or failed to reproduce are discussed below. Distribution of reproductive success among females Overall, 20 breeding attempts by eight females were mon- itored. Of these eight females, three produced offspring during the study (Table 1). Seven out of 10 breeding attempts resulted in offspring for these three females (one out of three, three out of three and three out of four breeding attempts, respectively). However, one female killed her single kit 5 h after birth, limiting the total number of successful attempts to six out of the total 20. Behavioral correlates of reproductive failure We observed no open aggression involving any of the three females that bred successfully. However, successful females showed a significantly higher rate of scent marking than unsuccessful females (F 1,9 =11.9, Po0.001), particularly abdominal and pelage rubbing and urination (Fig. 2). Reproductive success did not seem to affect the time budget of the females (F 5,18 =0.52, P=0.75; Fig. 1) nor the amount of social interactions either with other females (F 1,6 =0.10, P=0.75) or with males (F 1,6 =2.86, P =0.14). Further, reproductive success did not affect whether females initiated social interactions, either with females (F 1,6 =0.25, P=0.63) or with males (F 1,6 =0.41, P =0.54). Endocrine correlates of successful reproduction The total gestation time (measured from copulation to parturition) was 252 Æ 7 days (mean Æ SD of four breeding 70 60 50 40 30 20 10 0 % of scans 1996 All females Unsuccessful Successful 70 60 50 40 30 20 10 0 % of scans 1997 All females Unsuccessful Successful Inactive Moving Social Groom Other Figure 1 Behavior of female wolverines during the 1996 (n= 5) and the 1997 (n =4) breeding seasons. The number of successful breeders was two in 1996 and one in 1997. There was a significant difference between the behavioral categories recorded (F 5,18 =3.43, P =0.02), but reproductive success did not seem to affect the time budget of the females (F 5,18 =0.52, P= 0.75). The figure represents means and standard errors of per cent of 30 s scans. Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London 531 Reproductive failure in wolverinesF. Dalerum, S. Creel and S. B. Hall attempts where the date of mating was known). Parturition occurred on 25, 26, 28 February and 3 March. These breeding attempts came from three females. Successful breeding attempts (seven breeding attempts by three fe- males) were characterized by baseline progesterone levels during mating (5.78 Æ 0.77 mgg À1 dry feces) and diapause (6.01 Æ 0.77 mgg À1 dry feces), elevated progesterone levels during the gestation period (14.90 Æ 3.80 mgg À1 dry feces; here defined as the period between implantation and parturition) and baseline levels throughout lactation (6.59 Æ 53.00 mgg À1 dry feces) (Fig. 3a). Estrogen levels were elevated during the mating season (32.59 Æ 3.93 ng g À1 dry feces) and lactation (31.09 Æ 3.41 ng g À1 dry feces), but low during diapause (22.81 Æ 1.39 ng g À1 dry feces) and gesta- tion (22.60 Æ 3.41 ng g À1 dry feces) (Fig. 3b). Progesterone started to increase 50 days before parturition, indicating implantation (Fig. 4a). There was no detectable change in estrogen levels during implantation and gestation (Fig. 4b). Endocrine correlates of reproductive failure Progesterone profiles during the nine unsuccessful breeding attempts closely resembled profiles from the six successful ones, with one exception (Figs 3a, b and 4c–f). The excep- tional case did not show any endocrine signals of either implantation or gestation (Fig. 4e). Excluding this breeding attempt, there were no significant differences between suc- cessful and unsuccessful breeding attempts in either fecal progesterone (F 1,7 =0.04, P =0.85) or fecal estrogens (F 1,7 =0.82, P=0.39), nor significant interaction effects between breeding period and breeding outcome (proges- terone: F 3,283 =1.14, P=0.33; estrogens: F 3,347 =1.77, P=0.15). Thus, with one exception, sex-steroid profiles did not distinguish successful breeding attempts from failed. However, there was a significant interaction between breed- ing outcome and breeding period for fecal corticosterone (F 2,252 =3.37, P =0.036; Fig. 5). Corticosterone levels were significantly higher during the mating season for success- ful breeders than for unsuccessful breeders (t 8 =2.74, P=0.025), while there were no statistically significant differences during either diapause or gestation (diapause: t 9 =0.77, P=0.46; gestation: t 8 =0.64, P =0.54). Discussion Female behavior and social flexibility The amount of social interactions among the studied wol- verines approached the levels found in group-living carni- vores (e.g. Biben, 1983; Macdonald, 1996). Although our behavioral data were limited to only a few individuals, it still highlights that wolverines possess behavioral traits that permit them to participate in frequent social interactions. This indicates that sociality in this solitary carnivore is probably more flexible than previously believed. Traditionally, solitary social systems have been regarded as the original social system in carnivores, from which more complex forms of sociality have evolved (e.g. Packer, 1986; Gittleman, 1989). An alternative interpretation might be that social flexibility was the original state, from which the full range of patterns found across the Carnivora today has radiated. This could have happened, for instance, by ecolo- gical constraints in aggregating or in living solitary. Such an approach offers an evolutionary interpretation that may explain how a species that normally does not form social groups manages dense social aggregations such as those 2 1.5 1 0.5 0 Scentmarks/h Successful Unsuccessful TOTAL AR+PR DEF UR Figure 2 Frequency of scent marking of different types (AR, abdom- inal rubbing; PR, pelage rubbing; DEF, defecation; UR, urination) for successful (n =2) and unsuccessful female breeders (n=3), ex- pressed as scent marks per hour of observation. The difference in total scent marking rates between successful and unsuccessful females is statistically significant (F 1,9 =11.9, Po0.001). The figure represents means and standard errors. 20 15 10 5 0 Progesterone Mating Diapause Gestation Lactation Mating Diapause Gestation Lactation 40 35 30 25 20 15 Estrogens Successful Unsuccessful (a) (b) Figure 3 Concentrations of (a) fecal progester- one (mgg À1 dry feces) and (b) fecal estrogens (ng g À1 dry feces) during mating, embryonic diapause, true gestation and lactation in suc- cessful (n=7) and unsuccessful (n=9) breed- ing attempts. There are no statistically significant differences between successful and unsuccessful breeding attempts (proges- terone: F 1,7 =0.04, P=0.85; estrogens: F 1,7 =0.82, P= 0.39). The figure represents means and standard error of breeding attempt means. Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London532 Reproductive failure in wolverines F. Dalerum, S. Creel and S. B. Hall found in this study, and shows behavioral patterns similar to those found in traditionally group-living species. Female reproductive failure and social rank Of the eight females that were observed to copulate, only three gave birth. The reproductive success of these females, compared with the absence of successful reproduction in the other females, highlights a fairly substantial variation in reproductive success among individual females. This varia- tion indicates either that reproductive failure was related to social rank or that the successful females better adapted to non-social factors in the captive environment that could have impaired reproductive function. Although we could not use our data on social behavior to derive rank relation- ships between individuals, there are three lines of evidence that support the first case, that is social suppression. First, scent marking rates, which have been shown to increase with social rank in many carnivores (Ralls, 1971; Ough, 1982; Gorman & Trowbridge, 1989; Clutton-Brock et al., 1998b), were substantially higher in the successful breeders than in the unsuccessful ones. Second, the endocrine profiles of the unsuccessful breeding attempts imply that the endocrine functions in all but one of these females were intact and not impaired by environmental effects. Third, a potential stress response to the captive environment itself should have resulted in elevated GC levels in females that failed to breed successfully. However, this was not observed. In fact, the opposite was true, where GC levels were actually elevated at the time of mating for successful breeders. However, no individual female fully managed to monop- olize breeding. Two separate models have been suggested to explain such incomplete reproductive skew (i.e. the extent to which reproduction is monopolized by dominant indivi- duals): ‘optimal skew’ models and ‘limited control’ models (Clutton-Brock, 1998; Reeve, Emlen & Keller, 1998). Opti- mal skew models suggest that dominant individuals have full control over reproduction, but grant reproductive con- cessions to subordinates to keep them in the social group. Limited control models, on the other hand, suggest that subordinate breeding occurs because of the lack of domi- nant control. There are no a priori reasons to predict that these captive wolverine females would benefit from granting reproductive concessions, because these are normally attrib- uted to benefits in terms of increased group size or kin selection (Clutton-Brock, 1998). Hence, we suggest that the intermediate reproductive skew among these wolverine females is due to either the lack of control of a single, dominant female or a more complex, non-linear rank hierarchy that divides individuals into high- or low-ranking 40 30 20 10 40 30 20 10 40 30 20 10 –100 –50 0 50 –100 –50 0 50 –100 –50 0 50 –100 –50 0 50 –100 –50 0 50 –100 –50 0 50 ProgesteroneEstrogens 75 50 25 75 50 25 75 50 25 Days from birth Successful Unsuccessful Female 52 (a) (c) (e) (b) (d) (f) Figure 4 Profiles of fecal progesterone (mgg À1 dry feces) and fecal estrogens (ng g À1 dry feces) during the period of true gestation for pooled successful breeding attempts (a, b; n=7), pooled unsuccessful breeding attempts (c, d; n=8) and one unsuccessful breeding attempt (e, f) that showed no signs of ovarian activity at the time of expected implantation. Samples were pooled over 5 consecutive days, and the figure represents means and standard error of breeding attempt means for each 5-day period. 4 3.5 3 2.5 2 Fecal corticosterone (g/g dry feces) Mating Diapause Gestation Successful Unsuccessful Figure 5 Fecal corticosterone levels during mating, embryonic dia- pause and gestation. The difference between successful (n=7) and unsuccessful (n= 9) breeding attempts is statistically significant dur- ing the mating season (t 8 =2.74, P= 0.025), but not during either the diapause or gestation (diapause: t 9 =0.77, P =0.46; gestation: t 8 =0.64, P =0.54). The figure represents means and standard error of breeding attempt means. Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London 533 Reproductive failure in wolverinesF. Dalerum, S. Creel and S. B. Hall individuals rather than into a dominant and subordinates (which has been suggested for ungulates, e.g. Veiberg et al., 2004). Apart from the difference in scent marking rates, we found no other behavioral differences between successful and unsuccessful females. Because there was no evidence for strong, overt assertion of social rank among these females, maintenance of social rank in this population is not clear. However, it could be through olfactory means or by beha- viors not identified by our observations. Timing of reproductive failure All females that failed to give birth, except one, exhibited endocrine profiles similar to each other. In the only female that showed an inconsistent progesterone profile compared with the others, the lack of a progesterone response during the period of expected implantation and gestation agrees with endocrine patterns related to ovulation failure in this species (Mead et al., 1993; F. Dalerum, S. Creel & S. B. Hall, unpubl. data). For the other females, the increase in proges- terone levels during the time of expected implantation and gestation indicates that they ovulated, because a major part of this progesterone response is produced by the corpus luteum (Hodgen & Itskovitz, 1988). Hence, either obligate pseudopregnancies occurred or we observed actual pregnan- cies but failed to record parturition in the majority of the females. We feel that the latter case is highly unlikely with our nest box monitoring program. In primates, both abor- tion and re-absorption is followed by a sharp decline in gonadotropin and sex steroid levels, typically to below baseline (e.g. Kuehl, Kang & Siler-Khodr, 1992; Fortman et al., 1993; Steinetz, Randolph, & Mahoney, 1995). We did not find any such endocrine evidence for gestation failure in any of our failed breeding profiles. Gestation failure has been observed in some cases with intact progesterone profiles (Kuehl et al., 1992). Hence, we cannot entirely rule out the possibility that gestation failure occurred in our case. However, without any endocrine indications of gestation failure, we believe that, except for one female that showed strong signs of ovulation failure, reproductive failure in most cases was related to either a failure in fertilization or a failure in maintaining the embryo during diapause. Female reproductive failure and endocrine stress Our results contradicted the fact that elevated GC levels mediated reproductive failure. On the contrary, these results are in accordance with recent studies on group-living ani- mals that have opposed the hypothesis that elevated GC levels in subordinates mediate reproductive failure (see the review in Creel, 2001). Hence, not only behavioral tenden- cies among wolverine females appeared similar to group- living species, but also physiological. It is worth noting that these results were derived under experimental conditions, that is an abnormally crowded social environment for an essentially solitary species, similar to earlier studies that were used to support the hypothesis that GC levels mediated subordinate reproductive failure (e.g. Bronson & Elefther- iou, 1964; Louch & Higginbotham, 1967; Manogue, 1975). Considering the very low levels of aggressive behaviors observed, it is in fact surprising that we found any differ- ences between successful and unsuccessful breeders in the endocrine stress response at all. As there does not seem to be a strong behavioral assertment of social status among these females, it is unclear as to what causes the elevated stress response in successful breeders during the mating season. Conclusions In this solitary species, when housed in a communal setting, we found similarities in the level of social interactions to what is found in many obligate group-living species. Repro- ductive success seemed to be related to social rank, but reproduction was not entirely monopolized by a single individual female. Physiological correlates of reproductive failure appeared similar to what is found in many group- living species. This indicates that both the behavioral and physiological mechanisms observed could be viewed as reaction norms to the social environment, rather than as fixed traits that have coevolved with various forms of social structures. We suggest that sociality in solitary carnivores may be more flexible than commonly observed, and that a solitary social system may be maintained by ecological constraints that limit the formation of social aggregatations. Acknowledgements This project was supported by the Mt Hood National Forest, Bureau of Land Management, Oregon Zoo, North- west Trek Wildlife Park, Center for Wildlife Conservation and grants from the private foundations of J. A. Ahlstrand and Roland Nilsson. We are grateful to Dale Pedersen and Audrey Magoun for logistical support and constructive discussions, to Shane Iverson for support during the data collection and to Kathleen Cooper for help with early stages of the laboratory work. Anders Angerbj ¨ orn, Bertil Borg and Magnus Tannerfeldt improved earlier drafts of the manu- script. References Aitchison, J. (1986). The statistical analysis of compositional data. London: Chapman & Hall. Banci, V. (1994). Wolverine. In American marten, fisher, lynx and wolverine in the western United States: 99–127. Rug- giero, L.F., Aubry, K.B., Buskirk, S.W., Lyon, L.J. & Zielinski, W.J. (Eds). Fort Collins: US Department of Agriculture, US Forest Service, Technical Report RM-254. Banci, V.A. & Harestad, A.S. (1990). Home range and habitat use of wolverines Gulo gulo in Yukon, Canada. Holarct. Ecol. 13, 195–200. Biben, M. (1983). Comparative ontogeny of social behaviour in three South American canids, the maned wolf, Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London534 Reproductive failure in wolverines F. Dalerum, S. Creel and S. B. Hall crab-eating fox and bush dog: implications for sociality. Anim. Behav. 31, 814–826. Blanchard, D.C., Spencer, R.L., Weiss, S.M., Blanchard, R.J., McEwen, B. & Sakai, R.R. (1995). Visible burrow system as a model of chronic social stress: behavioural and neuroendocrine correlates. Psychoneuroendocrinology 20, 117–134. Bronson, F.H. & Eleftheriou, B.E. (1964). Chronic physio- logical effects of fighting in mice. Gen. Comp. Endocrinol. 4, 9–14. Cekan, Z. (1975). Assessment of reliability of steroid radio- immunoassays. J. Ster. Biochem. 6, 271–275. Clutton-Brock, T.H. (1998). Reproductive skew, concessions, and limited control. Trend. Ecol. Evol. 13, 288–292. Clutton-Brock, T.H., Brotherton, P.N.B., Smith, R., McIl- rath, G.M., Kansky, R., Gaynor, D., O’Riain, M.J. & Skinner, J.D. (1998a). Infanticide and expulsion of females in a cooperative mammal. Proc. Roy. Soc. Lond. Ser. B 265, 2291–2295. Clutton-Brock, T.H., Gaynor, D., Kansky, R., MacColl, A.D.C., McIlrath, G.M., Chadwick, P., Brotherton, P.N.B., O’Riain, M.J., Manser, M. & Skinner, J.D. (1998b). Cost of cooperative behaviour in suricates (Sur- icata suricatta). Proc. Roy. Soc. Lond. Ser. B 265, 185–189. Crawley, M.J. (2002). Statistical computing an introduction to data analysis using S-plus. Chichester: John Wiley & Sons Ltd. Creel, S. (1996). Behavioural endocrinology and social organization in dwarf mongooses. In Carnivore behaviour, ecology and evolution, Vol. 2: 46–77. Gittleman, J.L. (Ed.). Ithaca: Cornell University Press. Creel, S., Creel, N.M. & Monfort, S.L. (1996). Social stress and dominance. Nature 379, 212. Creel, S. (2001). Social dominance and stress hormones. Trend. Ecol. Evol. 16, 491–497. Creel, S., Creel, N.M., Mills, M.G.L. & Monfort, S.L. (1997). Rank and reproduction in cooperatively breeding African wild dogs: behavioural and endocrine correlates. Behav. Ecol. 8, 298–306. Creel, S., Creel, N.M., Wildt, D.E. & Monfort, S.L. (1992). Behavioural and endocrine mechanisms of reproductive suppression in Serengeti dwarf mongooses. Anim. Behav. 43, 231–245. Creel, S., Fox, J.E., Hardy, A., Sands, J., Garrot, B. & Peterson, R.O. (2002). Snowmobile activity and glucocor- ticoid stress responses in wolves and elk. Conserv. Biol. 16, 809–814. Creel, S. & Macdonald, D. (1995). Sociality, group-size, and reproductive suppression among carnivores. Adv. Stud. Behav. 24, 203–257. Dobson, H. & Smith, R.F. (2000). What is stress, and how does it affect reproduction? Anim. Repr. Sci. 60–61, 743–752. Ebensberger, L. (1998). Strategies and counterstrategies to infanticide in mammals. Biol. Rev. Cambridge Philos. Soc. 73, 321–346. Faulkes, C.G. & Abbott, D.H. (1997). Proximate regulation of a reproductive dictatorship: a single dominant female controls male and female reproduction in colonies of naked mole rats. In Cooperative breeding in mammals: 302–334. Salomon, N.G. & French, J.A. (Eds). Cambridge: Cambridge University Press. Fortman, J.D., Herring, J.M., Miller, J.B., Hess, D.L., Verhage, H.G. & Fazlebas, A.T. (1993). Chorionic-gona- dotropin, estradiol, and progesterone levels in baboons (Papio anubis) during early pregnancy and spontaneous abortion. Biol. Repr. 49, 737–742. Gittleman, J.L. (1989). Carnivore group living: comparative trends. In Carnivore behaviour, ecology and evolution: 183–207. Gittleman, J.L. (Ed.). Ithaca: Cornell University Press. Gorman, M.L. & Trowbridge, B.J. (1989). The role of odor in the social lives of carnivores. In Carnivore behaviour, ecology and evolution: 57–88. Gittleman, J.L. (Ed.). Ithaca: Cornell University Press. Hackl ¨ ander, K., M ¨ ostl, E. & Arnold, W. (2003). Reproduc- tive suppression in female Alpine marmots, Marmota marmota. Anim. Behav. 65, 1133–1140. Hodgen, G.D. & Itskovitz, J. (1988). Recognition and main- tenance of pregnancy. In The physiology of reproduction, Vol. 2. 2nd edn. 1995–2021. Knobil, E. & Neill, J. (Eds). New York: Raven Press Ltd. Hornocker, M.G. & Hash, H.S. (1981). Ecology of the wolverine in northwestern Montana. Can. J. Zool. 59, 1286–1301. Jennions, M.D. & Macdonald, D.W. (1994). Cooperative breeding in mammals. Trend. Ecol. Syst. 9, 89–93. Johnsson, D.D.P., Macdonald, D.W. & Dickman, A.J. (2000). An analysis and review of models of the socio- biology of the Mustelidae. Mamm. Rev. 30, 171–196. Kuehl, T.J., Kang, I.S. & Siler-Khodr, T.M. (1992). Preg- nancy and early reproductive function in the baboon. Am. J. Primat. 28, 41–48. Leyhausen, P. (1965). The communal organization of solitary mammals. Symp. Zool. Soc. Lond. 14, 249–263. Louch, C.D. & Higginbotham, M. (1967). The relation between social rank and plasma corticosterone levels in mice. Gen. Comp. Endocrinol. 8, 441–444. Macdonald, D.W. (1996). Social behaviour of captive bush dogs (Speotus venaticus). J. Zool. (Lond.) 239, 525–543. Magoun, A.J. (1985). Population characteristics, ecology and management of wolverines in northwestern Alaska. PhD thesis, University of Alaska, Fairbanks. Manogue, K.R. (1975). Dominance status and adrenocortical reactivity to stress in squirrel monkeys (Saimiri sciureus). Primates 14, 457–463. Mead, R., Bowles, M., Starypan, G. & Jones, M. (1993). Evidence for pseudopregnancy and induced ovulation in captive wolverines (Gulo gulo). Zoo Biol. 12, 353–358. Moehlman, P. (1989). Intraspecific variation in canid social systems. In Carnivore behaviour, ecology and evolution: Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London 535 Reproductive failure in wolverinesF. Dalerum, S. Creel and S. B. Hall 143–163. Gittleman, J.L. (Ed.). Ithaca: Cornell University Press. Ough, W.D. (1982). Scent marking by captive raccoons. J. Mamm. 63, 318–319. Packer, C. (1986). The ecology of sociality in felids. In Ecological aspects of social evolution : 429–452. Rubestein, D.I. & Wrangham, R.W. (Eds). New Jersey: Princeton University Press. Pasitschniak-Arts, M. & Larivie ` re, S. (1995). The wolverine. Mamm. Spec. 499, 1–10. Persson, J. (2003). Population ecology of Scandinavian wolver- ines. PhD thesis, Swedish University of Agricultural Science, Ume ˚ a, Sweden. Pinheiro, J.C. & Bates, D.M. (2000). Mixed-effects models in S and S-plus. Berlin: Springer Verlaag. Powell, R.A. (1979). Mustelid spacing patterns: variation on a theme by Mustela. Z. Tierpsych. 50, 153–165. Ralls, K. (1971). Mammalian scent marking. Science 171 , 443–449. Reeve, H.K., Emlen, T.R. & Keller, L. (1998). Reproductive sharing in animal societies: reproductive incentives or incomplete control by dominant breeders. Behav. Ecol. 9, 267–278. Sandell, M. (1989). The mating tactics and spacing patterns of solitary carnivores. In Carnivore behaviour, ecology and evolution: 164–182. Gittleman, J.L. (Ed.). Ithaca: Cornell University Press. Sapolsky, R.M. (2002). Endocrinology of the stress response. In Behavioural endocrinology. 2nd edn. 409–451. Becker, J.B., Breedlove, S.B., Crews, D. & McCarthy, M. (Eds). Cambridge, MA: Massachusetts Institute of Technology Press. Schoech, S.J., Mumme, R.L. & Moore, M.C. (1991). Repro- ductive endocrinology and mechanisms of breeding inhibi- tion in cooperatively breeding Florida scrub jays, Aphelocoma c. coerulescens. Condor 93, 354–364. Steinetz, B.G., Randolph, C. & Mahoney, C.J. (1995). Pat- terns of relaxin and steroids in the reproductive-cycle of the common marmoset (Callithrix jacchus) – effects of prosta- glandins-F2-alpha on relaxin and progesterone secretion during pregnancy. Biol. Repr. 53, 834–839. Veiberg, V., Loe, L.E., Mysterud, A., Langvatn, R. & Stenseth, N.C. (2004). Social rank, feeding, and winter weight loss in red deer: any evidence for interference competition. Oecologia 138, 135–142. Wasser, S.K. & Barash, D.P. (1983). Reproductive suppres- sion among female mammals: implications for biomedicine and sexual selection. Quart. Rev. Biol. 58, 513–537. Zar, J.H. (1996). Biostatistical analysis. 3rd edn. New Jersey: Prentice-Hall. Journal of Zoology 269 (2006) 527–536 c 2006 The Authors. Journal compilation c 2006 The Zoological Society of London536 Reproductive failure in wolverines F. Dalerum, S. Creel and S. B. Hall . Behavioral and endocrine correlates of reproductive failure in social aggregations of captive wolverines ( Gulo gulo ) F. Dalerum 1 , S. Creel 2 & S. B. Hall 3 1 Department of Zoology,. reproductive failure in most cases was related to either a failure in fertilization or a failure in maintaining the embryo during diapause. Female reproductive failure and endocrine stress Our. identity, of the initiator of the interaction. Scent marking was divided into four categories: defecation, urina- tion, pelage rubbing and abdominal rubbing. Pelage rubbing is a distinct behavior in