|Home | About | Journals | Submit | Contact Us | Français|
Previous research with female sheep indicates that exposure to excess testosterone for 60 days (from Gestational Days 30–90 of the 147-day gestation) leads to virilized genitalia, severe neuroendocrine deficits, as well as masculinization and defeminization of sexual behavior (T60 females). In contrast, 30 days of testosterone exposure (Gestational Days 60–90) produce animals with female-typical genitalia, less severe neuroendocrine alterations, and variable gender patterns of sexual behavior (T30 females). Variation in adult sexual behavior of male ungulates is influenced by early social experience, but this has never been tested in females. Here we investigate the influence of rank in the dominance hierarchy on the expression of adult sexual behavior in females. Specifically, we hypothesized that juvenile rank would predict the amount of male- and female-typical mating behavior exhibited by adult female sheep. This hypothesis was tested in two treatment groups and their controls (group 1: T60 females; group 2: T30 females). Dominance hierarchies were determined by observing competition over resources. Both groups of prenatal testosterone-treated females were higher ranking than controls (T60: P = 0.05; T30: P < 0.01). During the breeding season, both T60 and T30 females exhibited more male-typical mating behavior than did controls; however, the T30 animals also exhibited female-typical behavior. For the T60 group, prenatal treatment, not juvenile rank, best predicted male-typical sex behavior (P = 0.007), while juvenile rank better predicted male mating behavior for the T30 group (P = 0.006). Rank did not predict female mating behavior in the hormone-treated or control ewes. We conclude that the effect of prenatal testosterone exposure on adult male-specific but not female-specific mating behavior is modulated by juvenile social experiences.
Differentiation of the male central nervous system (CNS) is caused primarily by exposure to pre- and/or postnatal testosterone that is aromatized to estrogen. Several animal models (e.g., rodents, ferret, and rhesus monkey) have been used to examine the relationship between early androgen exposure, phenotypic virilization, and the development of male-typical adult behaviors [1–3]. Differentiation of male external genitalia requires dihydrotestosterone (DHT; the 5α-reduced product of testosterone), although sufficiently high levels of testosterone without conversion can also cause complete virilization . Female brains are also readily masculinized and defeminized by exposure to excess testosterone during development [5–7]. However, the timing of prenatal testosterone effects on differentiation of CNS and external genitalia are not the same (e.g., Rhees et al. ). Genital changes typically begin and end earlier than CNS masculinization. Furthermore, timing of organization of neural circuitry regulating behavior may differ from that regulating ovarian cyclicity. This is exemplified by the fact that exposure to low levels of prenatal testosterone in rats for varying amounts of time can increase male-typical behaviors (levels of aggression, dominance, and activity levels) without affecting estrous cyclicity, fecundity, and maternal behavior [9, 10].
The critical prenatal period for sexual differentiation of the hypothalamic-pituitary-gonadal axis of sheep spans Days 30–90 of their 147-day gestational period [11–14]. Females treated with testosterone for 60 days during the entire critical period, gestational days (GD) 30–90 (T60 females), are born with completely virilized genitalia, that is, with a functioning pseudopenis and an empty scrotal sac in place of a vaginal opening. The virilized genitalia are accompanied by gradual development of multiple reproductive neuroendocrine and ovarian disruptions [12–14]. Typically T60 females generate normal progestogenic cycles at the onset of puberty and throughout the first breeding season . In contrast, treatment lasting 30 days between GD60 and GD90 (T30 females), spanning the second half of the critical period, yields females born with essentially normal external genitalia . However, these females display a masculinized urinary posture and a small but significant delay in estradiol positive feedback response [16, 17], suggesting that the brain is either defeminized or masculinized. The reproductive neuroendocrine and ovarian cycles are otherwise normal in the first breeding season. Disruption of the CNS resulting in animals that exhibit both male and female behaviors is also apparent in studies with ferrets and nonhuman primates [2, 3, 18–20].
A detailed analysis of adult reproductive behavior in the T60 and T30 treatment groups yielded intriguing results . Both T60 and T30 females exhibited more masculinized reproductive behavior than did controls during monitoring of natural behavioral estrus in the flock and hormonally controlled behavioral tests. Interestingly, the T30 animals also exhibited feminine proceptive behavior. The T30 group produced highly variable patterns of reproductive behavior. They range from entirely female typical to entirely male typical and include mixed male and female patterns of sex behavior. In contrast, there was much less variation in the behavioral patterns of the T60 females (no female-typical behavior) and the controls (very little male-typical behavior ). While these results suggest that variation in the timing or duration of exposure to prenatal testosterone during a critical period for masculinization can have variable effects on defeminization, they fail to explain the variation in reproductive behaviors produced within treatment, as all individuals in each treatment received the same prenatal exposure.
In males, early social experience can influence reproductive behavior , and dominance behaviors associated with rank predict which animals are able to have access to females and therefore the opportunity to exhibit male-typical sex behaviors. Within stable social groups, dominance hierarchies are used to minimize the chance of injury and decrease energy expenditure during competitive interactions because low-ranking individuals learn to avoid challenging high-ranking competitors. As a result, dominant animals often control resources (e.g., food and mates), while subordinate animals often lose in competitions over resources and can become socially isolated to avoid aggressive interactions . In ungulates, dominance hierarchies form in early adulthood and are generally stable throughout life . Less dominant rams exhibit little male-male aggression or copulation, whereas high-ranking rams engage in both behaviors. It is unclear whether the variability in male- and female-specific behavior found in T30 adult females may also be a function of postnatal social interactions and development of dominance hierarchies interacting with the shorter duration of testosterone exposure. Because dominance hierarchies form early and influence almost all aspects of ungulate life and prenatal testosterone treatment increased aggressive interactions among female lambs , we hypothesized that juvenile rank that develops prior to the first breeding season would predict the amount of male- and female-typical mating behavior exhibited by adult female sheep, such that higher-ranked females would be more masculinized and defeminized than same-treated, lower-ranked females within each treatment group. This hypothesis could also predict that higher-ranking control ewes would also exhibit more male-typical behaviors than low-ranking control ewes, although the relationship may be dependent on some testosterone exposure during the prenatal critical period of development.
All animal procedures were approved by the University of Michigan Committee for the Use and Care of Animals. Two prenatal testosterone-treated groups of ewes were raised in separate small flocks with control females. For breeding purposes, adult Suffolk ewes were purchased from a local breeder and moved to a farm inspected and approved by both the U.S. Department of Agriculture and the University of Michigan Department of Laboratory Animal Medicine. In order to determine the date of mating during breeding, a mixture of paint and grease was applied daily to the chests of rams in the flock (raddled rams). Daily monitoring allowed visual confirmation of marks left by the rams, and the date of mating was recorded. If a female was marked again after the original mating date, it was considered a rebreed, and the mating date was changed accordingly. Once bred, ewes were maintained on pasture and supplemented with 1.25 kg alfalfa/brome mix hay per ewe.
For generation of T60 females, pregnant ewes received twice-weekly 2-ml i.m. injections of 100 mg testosterone propionate (Sigma-Aldrich Corp., St. Louis, MO, USA) in cottonseed oil for the 60 days between GD 30 and GD90 of the 147-day gestational period. Control females included those that were bred in the facility with the T60 females and additional animals that were purchased from the breeder who provided the ewes for generating the experimental females. The purchased controls were brought to the facility at 10 wk of age (weaning). No significant differences were observed in the growth trajectory, cyclicity, or behaviors of control females born at the facility and those purchased for the first group . The T30 females were generated by treating pregnant ewes with testosterone for 30 days from GD 60 to GD90. In the absence of behavioral differences between controls born at the facility and those that were purchased in previous studies, all controls used for comparison with T30 females (n = 12) were purchased from the same breeder and brought to the facility after weaning at 10 wk of age. The T60 females had anogenital distances equivalent to that of males, while T30 females did not differ from control females (T60, 26; T30, 15).
Beginning 6 wk prior to expected lambing, pregnant ewes were fed an enriched diet consisting of 0.5 kg shelled corn, 2 kg alfalfa hay, and 250 mg chlortetracycline (aureomycin) ewe−1 day−1, a common husbandry practice to protect against infection-related abortions. Lambing occurs each year between early March and early April . Throughout this time lactating ewes were provided a ration of 1 kg shelled corn and 2–2.5 kg of alfalfa hay ewe−1 day−1. Lambs were weaned at approximately 8 wk of age. After weaning, female lambs were moved to the University of Michigan's Sheep Research Facility, where they were kept in their birth group that included treatment and control animals, always able to freely interact with each other. The lambs were given commercial sheep food pellets (Shur-Gain, Elma, NY) and alfalfa hay, available ad libitum, until they reached a body weight of 40 kg. At that time they were provided a diet containing 15% crude protein until 6 mo of age. All lambs and ewes were provided water and minerals ad libitum and were treated regularly with antihelmintics to minimize parasitic infection. Throughout the testing period, all lambs were kept at the Sheep Research Facility.
For studies involving the T60 females, data were collected from 18 T60 females and 15 control females. For studies involving the T30 females, 12 T30 and 12 control females were studied. The T60 and corresponding control female lambs were housed together from birth with 10 males that were vasectomized at 3–3.5 mo of age. Prior to the onset of the breeding season, all but three of the males were removed. The T30 and their corresponding controls were housed together without males until 24 wk of age when one vasectomized ram joined the group. Free-living behaviors of all lambs in both groups were observed prior to and throughout their first breeding season.
Because dominance hierarchies arise from the competition for resources, we monitored their behavior at the time daily fresh hay and alfalfa pellets were provided (sheep are provided with daily ration at 0700 h that allowed ad libitum feeding throughout most of the day). Each day, despite more-than-adequate food and space, individuals compete for positions at the feed bins. Displacement behavior, supplanting of an animal from its position at the feeding bin, was monitored once per week for at least 1 h, or until all animals leave the feeders, from 5 wk of age to the end of the breeding season (about 48 wk of age). The frequency of displacements is high because individuals continually move from one position to another, displacing each other, even after they have initially won a position. Pushing, head butting, and other aggressive behaviors commonly occur as some sheep resist the incoming sheep. An ethogram of specific displacement behaviors developed is provided in Table 1.
Displacement was recorded on all interactions during feeding bouts. All observers were blind to treatment groups. For each interaction, the animals involved, the type of behavior used in the displacement, the initiator, the winner, and the loser of the interaction were recorded. The initiator is the animal that instigates the displacement; this animal can be either the winner or the loser, depending on the success of the displacement. If the initiator is successful in displacing the targeted animal, it is the winner of the interaction, and the targeted animal is labeled as the loser. If the initiator is not successful in displacing the targeted animal, the initiator is labeled as the loser, and targeted animal is the winner of the interaction. In the case where more than one animal is targeted by the initiator (e.g., the initiator pushes in between two targets), the interaction is coded as two separate events with a note that another target was involved.
Competition over resources includes competition for other resources in addition to food; therefore, we also collected displacement data on individuals during weekly focal observations. Each sheep was individually observed for approximately 10 min three times per week from 5 wk of age until the end of the breeding season. All the interactions with other sheep were recorded, except displacement behaviors were used in this analysis (see Table 1 for an ethogram). Animals engaged in similar displacement behavior at a lower frequency, at water sources, at favored resting sites, and at a mineral lick, for example.
To determine juvenile social rank prior to the onset of the breeding season, data collected over a 2-mo, prepubertal period from 1 July to 1 September were used. Puberty in female sheep occurs in early October in our flock. The Bradley-Terry model (reviewed in Boyd and Silk ) was used to generate a score for each animal and create the dominance hierarchy. This score was then used to assign rank to the sheep without adjusting for the covariates of gender, treatment, or weight. The ranking was generated with respect to individuals' chances of winning compared to their total interactions. Rank was determined within each group, resulting in one hierarchy for each group (T60, T30). To get a clearer idea of how the animals distributed themselves, we examined the hierarchy in thirds, that is, the top third, middle third and bottom third. Chi-square analysis was used to test whether the distribution differed between the treatments within each group. In addition, behavior data were analyzed separately by treatment (testosterone or control) for each group; t-tests were used to determine whether treatment differences occurred on the following measures: number of wins, losses, initiations, times that they were initiated against, and total number of interactions. Finally, chi-square analysis was used to test whether treatment affected the percent wins of total number of interactions as well as the sex and treatment of the opponent.
Because body size/weight is thought to be a factor in rank determination among males (e.g., Berenbaum ), we examined whether the average weekly body weight for the month of September differed between control and testosterone-treated sheep in each group; whether weight differed between the top, middle, and lower third of the hierarchy; and whether body weight correlated with rank.
A subset of animals from the first group (9 T60 and 10 control females, randomly chosen from larger cohort) and all animals of the second group (12 T30 with 12 control) were used in reproductive behavior tests near the end of their first breeding season to investigate effects of prenatal treatment on adult reproductive behavior. Methodology and details of results from these tests have been published . Here we utilize a summary of adult male and female reproductive behavior only from the behavioral test that was conducted in the free-living social group to test the hypothesis that individual variation in reproductive behavior can be predicted by earlier rank. Briefly, this test examined reproductive behavior after all females in a group were administered 20 mg prostaglandin F2α (PGF2α, 5 mg/ml Lutalyse; Pfizer Animal Health, Kalamazoo, MI) twice, 11 days apart, to synchronize estrus and facilitate behavioral observations. Behavior was recorded between 0730 h and 1730 h on three consecutive days after the second PGF2α injection. For both groups, interactions with the males and other females in the group were observed and recorded using a mating ethogram developed from Banks  and our own observations (see Roberts et al. ). For purposes of relating rank development in this study to later reproductive behavior, a total count of male- and female-typical behaviors was generated for each animal by collapsing all male- and female-typical behaviors, respectively, produced during the testing period.
Linear regression was used to test the predictability of male- and female-typical reproductive behavior using juvenile rank. Each group was analyzed separately, with rank and treatment included sequentially as predictors of maleness or femaleness.
The chi-square analysis revealed that testosterone-treated (T60) females of group 1 were higher ranking than controls (χ2 = 4.45, P = 0.05; Fig. 1). The majority of the animals in the top third of the hierarchy are T60 females, whereas controls are the majority in the bottom third. Additionally, the total number of wins (P = 0.05) was significantly greater in T60 females than controls (Table 2). The number of losses during displacement, initiations by, initiations against, and percent wins of total interactions did not differ between treatment groups.
In the second group (T30), testosterone-treated females were also found to be higher ranking than controls (χ2 = 13.0, P < 0.01; Fig. 1). Total number of wins (P = 0.032; Table 2) and percent wins (P < 0.001) were also significantly higher in the T30 females than controls. The number of losses tended to be less for T30 than controls (P = 0.066).
In both groups, the sex and treatment of the opponent played no role in the win percentage for an animal initiating the interaction. There was also no treatment effect found in either group on the total number of interactions. The T30 females won a higher percent of interactions than their controls (P = 0.018; Table 2), and this difference was nearly significant in the T60 group (P = 0.061).
Consistent with data collected on previous cohorts of testosterone-treated females , we found no difference in mean body weight (in kg) between T60 females and their respective controls at the end of the period when rank is determined (T60 = 62.4 ± 1.8, C60 = 59.4 ± 1.3), nor did we find a significant difference in the T30 group (T30 = 52.4 ± 1.1, C30 = 53.0 ± 1.3). Mean body weight also did not differ between the top, middle, and lower thirds of the hierarchy. Rank and mean body weight also were statistically uncorrelated (r = −0.167 and −0.164, for T60 and T30 groups, respectively). Thus, body weight during the development of the hierarchy did not seem to play a significant role in determining rank in our groups. It is possible that the correlations previously described in flocks with adults over a year old are the result of high-ranking animals gaining better access to resources and therefore growing bigger, or size may only be relevant in rank development of male sheep.
T60 females produced significantly more male-typical mating behavior than controls (t = −3.221, P = 0.003; Table 3). T60 females produced no female typical behavior, resulting in a nearly significant difference from the female-typical behavior produced by controls (t = 1.670, P = 0.094; Table 3). T30 females also produced significantly more male-typical behavior than controls (t = −3.266, P = 0.004; Table 3). However, the display of female sexual behavior did not differ between the T30 and control groups (t = 0.926, P = 0.364; Table 3).
In a regression that included rank and treatment variables, T60 juvenile rank did not predict adult male-typical sex behavior (standardized beta coefficient [β] = −0.235; t = −1.366, P = 0.181; Fig. 2); however, treatment did predict adult male sex behavior (β = 0.465; t = 2.888, P = 0.007). In contrast, juvenile rank in the T30 group significantly predicted adult male-typical sexual behavior (β = −0.541; t = −3.014, P = 0.006; Fig. 2). Treatment also predicted male-typical behavior in T30 females, but only if rank was not included in the model (β = 0.571; t = −3.266, P = 0.004). Adult female-typical sexual behavior was not predicted by rank or treatment in either group (T60: t = −1.093, P = 0.282; T30: t = 0.674, P = 0.508).
Female sheep exposed to either 30 or 60 days of testosterone during a critical period for sexual differentiation of physiology and behavior demonstrate significantly more masculine sex behavior than control females (Roberts et al.  and this study). However, the T60 group exhibits no female-typical sex behavior, while some animals in the T30 group exhibit robust female proceptive and receptive behaviors, as described in Roberts et al . Interestingly, the amount of male-typical behavior exhibited when the animals were in their social group varied among the individuals within both treatments, as it does among males. Among rams and testosterone-treated ewes, exposure to an estrous female in the absence of other sheep results in all males, all T60 ewes, and most T30 ewes exhibiting male reproductive behavior. However, when living in a group in which a hierarchy has been formed, typically with males at the top of the hierarchy, only the top-ranked males engage in courtship and mating behaviors while actively guarding estrous females from other high-ranking males. Low-ranking males demonstrate little interest unless the high-ranking male is removed. Thus, we hypothesized that the variability in the amount of male or female-specific behavior in ewes might correlate positively for male behavior and negatively for female behavior with hierarchical rank. We found that prenatal testosterone treatment significantly shifts females to a higher rank. Regression analysis with rank and treatment included showed that male-typical behavior of the less variable T60 group was best predicted by testosterone treatment, not rank. The more variable behavior of T30 group was better predicted by rank than treatment, such that high-ranking animals were more likely to demonstrate male-typical behavior (Fig. 2). In both groups, prenatal testosterone altered rank. In neither group did rank or treatment predict the amount of female-specific behavior. And in neither group did rank correlate with male- or female-specific behaviors in control ewes.
The ability to predict the amount of female-specific behavior is poor because very little female-specific behavior was generated by the T60 females and their respective controls. The results, however, suggest that prenatal testosterone treatment for 60 days is a good predictor of low female-specific sex behavior. Subsequent testing of animals individually with a ram found that controls generated considerably more proceptive and receptive behavior , while T60 animals did not, supporting this conclusion. In contrast, both the controls and T30 females produced substantial amounts of female-specific behavior, particularly proceptive behavior. There was considerable variability in whether the females allowed male courtship, mounting, or copulation, and it was therefore somewhat surprising that rank did not predict female-specific behavior as well as male-specific behavior.
The development of hierarchical rank in sheep is a process that requires many weeks. After weaning, through struggles over resources and during play, lambs sort themselves out such that most males are higher ranking than most females, males have a linear hierarchy, and females have a mixed linear and layered pattern with rank ties. This pattern held in our study groups. However, while males did outrank the females, when multiple ewes are in estrus simultaneously, the male is unable to guard all females. This provided the lower-ranking testosterone-treated females opportunity to exhibit male-typical sexual behavior. Unlike many primates, female sheep do not inherit their rank from their mothers, and its importance is uncertain [30, 31]. Male sheep, like male primates, suffer significant loss in reproductive opportunity if they are not high ranking and are therefore highly motivated early in life to “win” contests even though resources are plentiful for all.
We recognize that domestication may have altered the costs and benefits of dominance hierarchies for our sheep compared to wild ungulates. However, the animals still compete for resources, resulting in the formation of a hierarchy similar to that seen in wild species [32, 33]. In addition, nothing we do forces this competition; it is not necessary for them to compete to get food, as there is more than enough for all individuals. Finally, we do see these behaviors (although at a lower frequency) even when they are in pasture and at water sources. For example, in the T60 group, controls succeed in 36.8% of their initiated displacements and testosterone-treated 46.2% at nonfood locations. This difference between treatments is similar in size to the effect at the food sites, although both groups win more often when attempting a displacement at food. Here we argue that hierarchy is one juvenile predictor of adult reproductive behavior, but further investigation of such patterns is needed to clarify this relationship in wild populations of ungulates.
Interestingly, the motivation to strive for high rank in the flock seems to be primarily programmed by prenatal testosterone exposure. Because males undergo early pubertal development, one might imagine that rising testosterone levels during the summer prior to the first breeding season drive much of the development of these behaviors. However, the prenatal testosterone-treated females are engaging in the same behaviors in the absence of significant gonadal hormones, as juvenile rank formation is complete prior to female puberty and the onset of the first breeding season. However, some gonadal steroid production may be necessary, and, therefore, further experiments are needed to test their activational role in the development or maintenance of a hierarchy.
Findings from this study provide support for the hypothesis that behavior early in life, prior to puberty, can predict preferred sex partner choice. The T30 females exhibited variation in sex behavior, including entirely male, mixed male and female, and entirely female sex behavior when living in a social group that included a ram and other estrous females. Rank, developed during the juvenile period between weaning and puberty, predicted whether females would show more or less male-specific behavior. The high-ranking T30 females are fully capable of allowing copulation , but most prefer to take a male sex role .
There are some human conditions in which fetuses are exposed prenatally to excess androgens and manifest virilization of female genitalia and masculinization of some aspects of behavior. For example, some prepubertal behaviors (i.e., play, cognitive interests, and social interactions) are altered in girls with congenital adrenal hyperplasia [34, 35] who are exposed to high levels of adrenal androgens during fetal life. Postpubertal gender identity, partner choice, and sexual behavior are far more variable in this group of women than in the general population (e.g., Dittman et al. ), suggestive of gene and environment (prenatal and postnatal) interactions.
The data from prenatal testosterone-treated sheep raise the interesting possibility that the variable outcomes for adult behavior of human women exposed to prenatal androgens is the consequence of differential social interactions during development. Human social interactions are multilayered and very complex and include the family environment, peer-group environment, and wider society interactions and expectations. The data presented here for sheep suggest that peer-group interactions may be important in determining relative rank within the peer group and thereby in “allowing” more or less male-typical behavior to be exhibited should the individual be biologically inclined in that direction. Since human children spend a great deal of their time in peer-group activities, peer interactions may be a good place to look for variation that may predict adult behavior.
We wish to thank Doug Doop for assistance with breeding, lambing, and excellent care of the sheep, particularly the very young lambs; Carol Herkimer, James Lee, and Mohan Manikkam for help with prenatal treatment; and Dr. Morton Brown for advice on determining rank. We also are indebted to the many University of Michigan undergraduates who spent many hours watching and recording sheep behavior. They include Dominica Groom, Tiffany Chao, Joseph Wehri, Lindawati, Jessica Bogart, John Meixner, Yan Iuan Ho, and Lisa Lothamer.
1Supported by USPHS grants P01-HD44232 and R01 HD41098.