Prenatal Androgen Exposure Induces Anxiety-Like Behavior in Ewes Free

Exposure to excessive androgens during pregnancy has been linked in humans to several conditions, such as polycystic ovary syndrome (PCOS), autism spectrum disorder, and attention deficit hyperactivity disorder [1‒4]. Those pathologies include neuropsychiatric symptoms like anxiety, cognitive deficiencies, and a higher prevalence of depression [5, 6]. Many women with PCOS experience psychological symptoms such as depression, anxiety, or eating disorders [7, 8]. Moreover, in women with PCOS, the risk of developing depressive disorders is reported to be two-fold higher than in healthy women [9] around 15% of PCOS women exhibit increased anxiety as compared to 1–2% of the general women population [7, 10]. Finally, a recent study showed that daughters of women with PCOS are at a 78% higher risk of being diagnosed with anxiety disorders compared with children born to women without PCOS [11]. Despite the burden of anxiety and depression on women’s wellbeing, the mechanisms underpinning the development of anxiety disorders in PCOS are poorly understood. While preclinical studies in mice [12] and rats [13] have demonstrated that maternal androgen excess impacts brain development and elevates the likelihood of anxiety and affective disorders in adult offspring [11, 13], whether this phenomenon extends to larger mammals like sheep, which bear closer resemblance to human physiology, remains uncertain.

Ewes and humans exhibit similar body weights both at birth and in adulthood. In terms of reproductive physiology, ewes share with women a low ovulation rate and a luteal phase that is comparable. A notable distinction lies in the level of steroidogenesis of the foeto-placental unit, which generates higher levels of steroids in humans [14]. The prenatal androgenized (PNA) sheep represents a well-recognized model to understand the etiology of PCOS [4], as these animals recapitulate most of the reproductive, neuroendocrine, and metabolic features of the syndrome. Furthermore, these mammals exhibit intricate biobehavioral capacities and share several similarities with human emotions.

Indeed, unlike rodents, sheep are mono-ovulatory species with a circumvolved brain, and their fetal development bears greater resemblance to humans than rodents, particularly in terms of gestational length, maternal-fetal exchanges [15], and the timing of key steps of the development [16]. Moreover, similarly to women with PCOS, PNA ewes present polycystic ovarian morphology and an important number of reproductive and metabolic symptoms overlapping with those of women with PCOS [4, 17]. Thus, the present study aimed to investigate the effects of prenatal androgenization on anxiety and depressive-related behaviors in PNA and control sheep offspring.

Ewes were androgenized from gestational day 65 to 105 (term at gestational day 147) using a protocol previously validated to induce a PCOS-like phenotype in sheep [4, 17, 18]. After weaning, control, and PNA ewes were submitted to different behavioral tests.

Our findings illustrate that prenatal androgenization in sheep leads to a notable rise in anxiety-related behaviors compared to control animals, while leaving their cognitive functions and learning abilities unaffected. These findings underscore the significance of this animal model for further research endeavors aimed at unraveling the molecular mechanisms behind anxiety in PCOS.

Experiments were carried out between December 2018 and October 2019 in Nouzilly, France (latitude 47° 32 N and longitude 0° 46 E). A total of 43 Ile-de-France ewes (Ovis aries), 20 control and 23 prenatally exposed to androgens (PNA group) born between 7 and September 17, 2018 (mean date of birth, 11 September ± 0.5 days). Animals were housed indoors under natural lighting and temperature conditions.

PNA ewes were generated by injecting intra-muscularly their mothers twice a week with 100 mg of testosterone propionate diluted in sesame oil, the control group received sesame oil only. This prenatal androgenization protocol has been validated by previous studies [18, 19]. In order to prevent significant anatomical malformations that could jeopardize the survival of the lambs, we administered testosterone propionate to the mothers during the second third of gestation, specifically from the 65th day to the 105th day of gestation [18, 19]. Within 24 h after birth, animals were separated from their mother and reared artificially with free access to artificial milk. After weaning at 2 months, young females were grouped in the same pen and were fed daily with barley straw, lucerne hay, and commercial concentrate, with free access to water and mineral blocks. All procedures were performed in compliance with the European directive 2010/63/EU on the protection of animals used for scientific purposes and were approved by the Local Ethical Committee for animal experimentation (CEEA VdL, Tours, France, No. 2016062917335667).

At 2.5 months old, all female subjects underwent two distinct “novelty tests” over three consecutive days [20]. The first 2 days they were submitted to an open field test in a pen of 2 m × 7.5 m with solid walls of 1.5 m enabling total visual isolation of the animal tested toward other animals housed in the same building. On the third day, animals were introduced in the same pen as the first 2 days but an unknown object (an orange construction cone with a mop on the top) was placed in the middle of the pen.

For each test, each animal was left alone in the test pen for 10 min, behaviors were recorded using a SONY Handycam video camera and were quantified using the Behavioral Observation Research Interactive Software (BORIS). For the open field test, we monitored the number of vocalizations (bleats), defecations, flight attempts, and we calculated the time the animal spent standing, walking, and running. For the novel object test, we calculated the latency to touch the novel object for the first time.

To determine the basal level of cortisol in our two groups of sheep, 2 weeks after the last day of the test, blood samples were collected from 10 control and 10 PNA ewes by jugular venipuncture in 5 mL tubes containing heparin. Three samples were collected from each sheep, with a delay of 30 min between two samplings. Venipunctures were performed by experienced and competent staff to minimize the stress level that could be induced by this procedure. After collection, blood samples were spun for 30 min at 3,500 g, and plasma collected was analyzed using the radioimmunoassay method described and validated for ovine species with a sensitivity of 4 ng/mL [21]. The data obtained from the three plasmas were pooled.

At the end of the experiment, the animals were sacrificed by an overdose of pentobarbital in a registered slaughterhouse by a certified butcher, and the head was immediately perfused via carotid arteries with 2 L of 1% sodium nitrite followed by 4 L of 4% paraformaldehyde in 0.1m phosphate buffer (pH 7.4). The brain was removed and the diencephalon was dissected between the optic chiasma and the mammillary bodies. Tissue samples were postfixed for 24 h in the same fixative and cryoprotected in 15% sucrose and 0.1% sodium azide in 0.1 m phosphate buffer (pH 7.4) at 4°C. Free-floating frontal sections (30 μm) were cut on a freezing microtome and placed in 0.01 m phosphate buffer saline (pH 7.4), containing 0.3% Triton X-100 and 0.1% sodium azide (phosphate buffered saline-TA) at 4°C. Immunohistochemical staining was performed for AVP and CRH using specific antibodies against AVP, raised in the rabbit and previously used in sheep [22] and against ovine-CRH, raised in rabbit (gift of Dr. G. Tramu) and also previously used in sheep [23]. Two series of one-in-ten sections (through the preoptic area and rostral hypothalamus) were incubated for 4 days with each primary antibodies following a previously described procedure [22]. After washing, the primary antibodies were detected by immunofluorescence by incubating the sections overnight in a solution of donkey immunoglobulins raised against rabbit immunoglobulin (Alexafluor 488^®, Molecular Probes – Invitrogen). Primary and secondary antisera were diluted in phosphate buffered saline containing 0.3% Triton X-100, 0.1% sodium azide, and 0.1% human serum albumin. Both primary antisera were diluted 1/10,000 and the fluorescent antiserum was diluted 1/2,000. The specificity of the anti-CRH and the anti-AVP has been previously demonstrated (CRH [24]; AVP [22]), in addition, the specificity of the immunostaining was checked by incubating sections without the primary antisera.

The immunofluorescence staining was observed through a confocal microscope (Olympus^®) fitted with laser pens (argon and helium-neon) and filters to observe Alexa Fluor 488^® dyes. Immunostained sections were then scanned using a Zeiss Axioscan Z1 equipped with an ORCA Flash 4.0 V2 camera (2,048 × 2,048, pixel size: 6.5 µm). Image analysis was performed with ImageJ software [25]. Regions of interest, of constant area, were manually marked over the paraventricular nucleus (PVN), and the threshold (mini and maxi) was adjusted manually to select a range of values that were optically identified positive for staining. Immunostaining intensity for CRH and AVP in the PVN correspond to the positive pixel area within each ROI.

At 10 months of age, animals’ cognitive performances and perseveration were tested using a detour task test [26, 27]. To minimize the stress of our animals, this test was performed in a part of their home pen ( Fig. 1a), while the other sheep remained in a second part of their home pen called the “Stock pen.” The Stock and Test pens were separated by a “Security pen” used to avoid the escape of sheep during the transfer to the test pen and to isolate the subject tested between two trials. All fences separating those areas were solid, ensuring that the tested animal was visually isolated from other sheep but could still maintain olfactory and auditory contact with them. During the weeks leading up to the detour task, the animals were familiarized with the bucket and accustomed to eating the commercial concentrate inside. Some sheep had to be removed from this test as they continually refused to approach the bucket during the familiarization or failed on the habituation phase of the test (Fig. 1a, step 1); hence, 10 control and 11 PNA ewes were finally tested.

During this habituation phase, the bucket was placed inside the “V” designed by the two openwork fences (height: 120 cm) placed in the middle of the test pen. Ewes had a maximum of 6 min to eat inside the bucket; females had to pass at least two out of the three trials to participate in the detour task test. Before the start of each trial, the ewe was led by the experimenter to a defined position of the pen to standardize the starting point (Fig. 1a).

After the habituation phase, animals were submitted to a classical detour task, the bucket was disposed just behind the openwork fences, at the tip of the “V” (Fig. 1a, step 2). This test was repeated 3 times and females had 6 min to solve the task, and their latency to reach the bucket and the side chosen was noted. The side chosen at least 2 times over the three tests was defined as their preferred side.

After the classical detour task, females were submitted to a reversed detour task [26]. At this point, the preferred side used to reach the bucket in the preceding phase was blocked by an openwork fence (Fig. 1a, step 3). This test was repeated 3 times and females had 6 min to solve it, and their latency to reach the bucket and the number of times they try to go to their preferred side was noted.

At 1-year old, the influence of a prenatal androgenization on the judgment of ewes was assessed using a spatial learning task previously validated in sheep [28]. A test pen of 3 m × 2 m was constructed using solid wooden walls of 1.5 m high visually isolating the sheep tested from other sheep (Fig. 1b). To reduce the stress linked to transportation before tests, the test pen was constructed in the same building as their home pen, hence animals tested remained in olfactory and auditory contact with other sheep.

Sheep underwent a habituation session to familiarize themselves with the apparatus; they were let inside by groups of two to three for 10 min and received a food reward. Then, during three sessions, sheep were introduced alone in the test arena and were trained to associate one side with a positive reward (food); each training session consisted of five trials. If animals ate or approached the bucket (less than 30 cm) within 25 s, the trial was considered to be successful. Ewes were randomly assigned to associate the right (n = 6 for both groups) or the left (control, n = 7; PNA, n = 6) corner to the reward.

After those three sessions, we introduced the negative event, an empty bucket was placed on the corner opposite to that considered positive and if the sheep approached it closer than 30 cm, a panel on the wall was lifted to reveal a blower that was turned on (Fig. 1b). The blower was surrounded with plastic to ensure it would induce a withdrawal behavior and therefore avoidance of the bucket. The blower was stopped as soon as the sheep were distant enough from the bucket (more than 30 cm). As the negative bucket was empty and to ensure ewes did not just discriminate the empty bucket from the full bucket using olfactory clues, the two buckets were stacked and tied together with the bottom one containing commercial concentrate, the food reward used in this test, ensuring that the bucket smells of food even during negative trials. This training lasted for 13 sessions over 2 months, with five trials per session where three positive and two negative trials were alternated. The last trial was always positive to avoid a negative association with the test pen. To be considered trained sheep had to reach at least 75% of good responses over three training sessions, i.e., approaching the positive buckets (GO response) and not approaching the negative buckets (NO-GO response). All animals were trained for 13 sessions regardless of when they reached the learning criteria.

Once all sheep were trained, they were tested in the same arena for two consecutive days. During these two test sessions, the bucket was presented in three ambiguous locations in addition to the positive and negative locations. Those three locations were as follows: middle (M) at an equal distance (1 m) from the positive and the negative location, near positive (NP) at 50 cm from the positive location, and near negative (NN) at 50 cm from the negative location (Fig. 1b). The bucket was empty when presented at ambiguous locations. During the two test sessions, the first two bucket locations were respectively, positive and negative to validate that the sheep had learned to discriminate between these two positions. Then ewes were exposed to ambiguous locations on this order NP, NN, M on day 1 and M, NP, NN on day 2. As during training sessions, if the sheep went close to the bucket (less than 30 cm) within 25 s it was considered to be a GO response otherwise it was considered a No-GO response. The percentage of sheep with a GO response for each bucket and the latency to approach the bucket was analyzed using the Behavioral Observation Research Interactive Software (BORIS) on video recorded by a SONY Handycam video camera. We attributed a latency of 26 s to the sheep with a No-GO response.

All statistical analyses were performed using R version 3.4.4 [29], and the significance level was set at p < 0.05.

As open field data did not reach assumptions of normality and variance homogeneity, we used nonparametric tests. Unpaired Mann-Whitney tests were used to compare between groups for each of the 2 days of the test and paired were used to compare within a group between the 2 days of the open field test. Between day 1 and day 2, individual evolution of some behaviors (bleats, flight attempt, and time spent running) was determined to assess the proportions of animals habituating to the apparatus and so exhibiting less of those behaviors. The proportion of those animals was compared between groups using Fisher’s exact tests. Cortisol and latency to touch the novel object were compared between groups using an unpaired Mann-Whitney test.

Latency data did not reach assumptions of normality and variance homogeneity, we chose to perform a log transformation of those data and the statistical analysis using a linear model (“lmer” function) was performed on the transformed measures. Hence, we used an ANOVA to analyze latencies to reach the bucket, the model contained the group and the number of the trial and their interaction as a fixed effect and the animal identity as a random effect. Multiple comparisons were then performed using the “emmeans” package with a Tukey correction.

To compare the preferred side used to solve the detour task between groups we used Fisher’s exact tests. Perseveration was analyzed by comparing groups for the number of times each ewe tried to go to her preferred side. Comparisons were performed for each day using Mann-Whitney tests as data did not reach assumptions of normality and variance homogeneity.

The percentage of GO response was analyzed using a generalized mixed-effects model with a binomial distribution and logit link function [30]. The model contained four fixed effects: the group, the bucket location, the day, and the order in which sheep passed the test. Multiple comparisons were then performed using the “glht” function (“multcomp” package) with a Tukey correction. Latency to reach the bucket was analyzed using a generalized mixed-effects model with a quasi-Poisson distribution and logit link function. The model contained four fixed effects: the group, the bucket location, the day, and the order in which sheep passed the test. Multiple comparisons were then performed using the “glht” function (“multcomp” package) with a Tukey correction.

Statistical analyses of the AVP and CRF immunostaining data were carried out using R software (version 4.2.2). Missing values were imputed using the loess method (span = 0.75 and degree = 2) using values from the same samples in order to obtain a continuous curve for each. Then, nonparametric Kruskal-Wallis tests were performed on the average number of neurons to identify differences between the two groups (p value <0.05).

To assess the anxiety levels between control and PNA ewes at 2.5 months, 20 control and 23 PNA females underwent two novelty tests: an open field test and a novel object test, all conducted within the same pen. The open field test was performed on two consecutive days. No difference in the locomotion parameters, the number of bleats and the number of flight attempts was observed between control and PNA animals during the first day of the test (Fig. 2a, c, e). When comparing the results between day 1 and day 2 within each group, we noted a significant decrease in the number of vocalizations in the control group (Fig. 2a) and a reduction in the time spent running during the test (Fig. 2e). Additionally, there was a tendency for the number of attempts to flee to decrease on day 2 compared to day 1 (Fig. 2c). No differences in behavioral response were observed in PNA ewes when comparing the 2 days of the test (Fig. 2c).

By analyzing the individual evolution of these behaviors between day 1 and day 2, we observed that most of PNA ewes (13 of 23) spent more time running on day 2 than day 1 while only 3 out of the 20 control ewes showed a similar behavior (Fig. 2f). Consequently, there is a significant difference in the evolution of running behavior between the two groups. Similarly, half of the PNA ewes (11 of 23) increased their flight attempts on day 2 and only 4 of the 20 controls displayed such increase (Fig. 2d). Nevertheless, comparison between groups failed to reach significance. In terms of vocalizations, only 8 PNA and 4 control ewes bleat more on day 2 in comparison to day 1 (Fig. 2b).

The day after the last open field test, ewes were submitted to a novel object recognition test. PNA ewes took 4 times longer than control ewes to touch the unknown object, with median latencies of 100 s and 25 s, respectively (Fig. 3a). Basal cortisol levels of ewes were assessed 2 weeks after the anxiety tests and showed that PNA ewes have an increased cortisolemia as compared to control animals (Fig. 3b).

The immunoreactivity of CRH and AVP, two hypophysiotropic neuropeptides known to stimulate the corticotrope axis in mammals, was assessed in the PVN of the hypothalamus. Both peptides were found distributed among large populations of neurons (Fig. 4A, B) throughout the PVN, consistent with prior descriptions [31, 32]. While the immunoreactivity of AVP showed no difference between the PNA and control groups, CRH immunoreactivity was observed to be increased in the PNA group compared to the control group (Fig. 4C–E).

The detour task test was employed to assess the cognitive abilities of both control and PNA females and to identify potential perseverative behaviors in these animals. Concerning the latency to reach the bucket, a significant effect of the number of trials was observed, indicating that ewes were learning. No significant effect of the treatment was detected, nor was there any interaction between the treatment and the number of trials (Fig. 5a). There was no discernible difference between the PNA and control groups in terms of the occurrence of perseverative behaviors, as evaluated by comparing the number of errors in the reversed detour task. During the last trial (Trial 6), most of the ewes in both groups made no errors.

During the classical detour task, the side chosen by the ewes to solve the task was noted but no difference in the lateralization according to the group could be observed. The left was the preferred side of 60% of control ewes and 54% of PNA ewes.

The judgment bias test was used to detect a possible effect of the prenatal androgen exposure on depressive-like behaviors in ewes. This test enables to discriminate ewes with an optimistic or pessimistic state of mind [27]. The learning curve was highly similar between the control and PNA groups (Fig. 5b). Data from the judgment bias test were analyzed utilizing a GLM model incorporating the subsequent fixed effects: group, bucket location, test day (day 1 or day 2), and the sequence in which sheep completed the test. Comparisons of models with or without the “order” and “day” factors were not significant. Therefore, these factors were excluded from the analysis and data for days 1 and 2 were combined.

Looking at the percentage of sheep exhibiting a GO or a NO-GO response for the different buckets locations, we detected an interaction between the treatment and the position of the bucket (Fig. 5c). A strong effect of the buckets’ locations was also detected while there was no significant effect of the treatment and in the percentage of GO responses. A dramatic drop in the number of ewes reaching the bucket was detected between the P-NP-M buckets and the NN-N ones (Binomial GLM, Tukey’s post hoc test, p < 0.01).

A strong effect of the buckets’ location was also detected for the latency to reach the bucket (Fig. 5d). In contrast to control ewes, PNA ewes appeared to exhibit a greater level of discrimination between buckets P, NP, and M. Specifically, PNA ewes took significantly longer to reach the M bucket compared to the positive bucket. Additionally, their latency to reach the NP bucket also tended to be prolonged in comparison to the P bucket. Similar to the observations with control ewes, the latencies to reach the N and NN buckets did not exhibit variance, yet they were notably higher than the latencies to reach the remaining three buckets.

To our knowledge, this is the first study investigating in ewes the consequences of prenatal exposure to androgen excess on cognitive functions and anxiety-related behaviors. We demonstrated that a prenatal androgenization underpins anxiety-like behavior in sheep and a slight modification of their judgment in a judgment bias test without affecting their cognitive functions or learning abilities. Those are interesting observations as women with PCOS have an increased prevalence of anxiety and depression symptoms [5, 7], or more generally of psychiatric disorders [33] independently of body mass index. In a recent investigation, Risal et al. demonstrated that daughters born to mothers with PCOS face a 78% higher likelihood of being diagnosed with anxiety disorders [10]. Importantly, female offspring of PNA mice also develop anxiety-like behavior [11].

In humans, anxiety can be defined as an emotional state induced by the perception of a possible threat, increased fearfulness, and avoidance of novelty [34]. In animals, anxiety can be measured using the locomotor activity and the reaction of animals when confronted with an unknown environment or object [34]. We started our examination of the behavior of our sheep by the open field test as it is commonly used to assess fear and anxiety in rodents [35, 36], and it has also been used in sheep for several decades [37]. In our experiment, we did not observe any difference between PNA and control ewes when exposed for the first time to the open field test. This can be a consequence of the high level of stress induced by the combination of social isolation and exposure to a novel environment [20]. Within-group comparison of the stress behavior between the first session and the second one indicates a significant decrease of anxiety-related behaviors in the control group, suggesting that these animals habituate to the new environment. Consistently, we also observed that the time spent running tended to decrease on the second day in control animals, while no effect was observed for the PNA ewes, regardless of the training session. Moreover, individual analysis of the open field test showed that the number of PNA sheep exhibiting an increase in all parameters examined on day 2 is more than twice as compared to controls. Confirmation of anxious behavior in PNA sheep occurred during the third session when they were introduced to an unfamiliar object. Significantly, longer durations were observed before they made contact with this object in comparison to control ewes. Indeed, avoidance of novelty is described as a compound of anxiety in both humans and animals [34, 38]. These behavioral observations find support in physiological data, as PNA ewes exhibit elevated basal cortisol levels even in the absence of social isolation or novelty. Cortisolemia is known to be increased in humans [39] and animals [40, 41] that exhibit increased anxiety. The influence of testosterone on anxiety-like behaviors varies depending on the sex and appears to differ across species and the timing of testosterone exposure (i.e., pre or postnatal). Prenatal androgenization may be responsible for developing anxiety disorders in female offspring [12, 13], and our results in prenatally androgenized sheep are consistent with these data. The prenatal androgenization has been commonly used to induce a PCOS phenotype in female offspring [4], and the androgenization protocol used in this study has been previously validated to induce PCOS-like symptoms in sheep [16]. The rise in cortisolemia observed in PNA animals may be associated with the elevated CRH immunoreactivity, suggesting activation of the entire corticotrope axis rather than solely the production of cortisol in the adrenals. In addition, the CRH neurons could be a target of testosterone action during pregnancy. The expression of androgen receptors in CRH-containing neurons remains uncertain in sheep, with only an 8–11% presence reported in goats [42], and no colocalization noted in mice [43]. The limited putative colocalization in sheep implies a more indirect impact of androgenization on the corticotrope axis. Fetal exposure to abnormally high levels of testosterone has also been linked to autism [2, 44]. Autism is a syndrome involving numerous symptoms including increased anxiety [6], impaired inhibition but also an increased perseveration in a detour task [45]. As we observed an effect of the prenatal androgenization on anxiety-like behavior, we investigated other two relevant parameters (inhibition and perseveration) using a modified detour task, previously validated in sheep [26]. We did not observe any impairment of the cognitive functions in PNA ewes when compared to control animals. Moreover, the perseverance assessed by the number of errors in the reversal detour task was similar between groups. We did not observe any influence of maternal androgenization on the cognitive abilities of ewes. In autism spectrum disorder, symptoms appear to manifest differently based on sex. While boys typically exhibit more aggressive, repetitive, or hyperactive behaviors, girls often experience heightened emotional symptoms such as increased anxiety and depression [46, 47]. Therefore, future investigations should focus on evaluating the performance of PNA rams on a reversal detour task. Autism and PCOS are associated to prenatal exposure to elevated levels of testosterone and women of both disorders face increased risk of depression [7, 47]. Testing depression in sheep is not an easy task because some behavioral tests, such as the forced swim test, are difficult to transpose to large animals. Thus, we used a judgment bias test as the first attempt to examine these biobehavioral responses in prenatally androgenized sheep. This test enables to discriminate between optimistic and pessimistic animals, and it has been previously used in different species, including sheep and cows [28, 30, 48]. We did not find any evidence that a prenatal androgenization could affect the depressive-like behavior in sheep. Our results are consistent with previous findings in rodent models of PCOS [11, 12]. This difference between human and other species may be attributed to the elevated level of placental steroidogenesis in humans [14], potentially leading to a heightened sensitivity to depression in humans.

The potential mechanism by which high circulating testosterone in the mother ultimately changes emotionality in the offspring could be due to the organizational effects of testosterone during fetal development, which may alter brain connectivity and subsequently give rise to altered behavioral responses. However, future studies in PNA sheep are granted to address whether changes in the anatomy and gene expression of the neuroendocrine and limbic cell populations implicated in the control of anxiety-like behavior might exist like previously described in mice [11]. The regulation of other putative candidates that are known to be involved in anxiety, stress, and mood regulation, such as the adrenocorticotropic hormone, dopamine, thyroid hormones, or even vasopressin and oxytocin, should be investigated also to improve our understanding of this phenomenon.

In conclusion, prenatal androgen excess induces anxiety-like behavior in ewes without affecting their cognitive functions and learning abilities. Despite the effect on anxiety, no significant difference was observed on depressive-like behavior using a judgment bias test. Our study confirms the feasibility and the interest of using non-rodent models in the investigation of the behavioral consequences of a physiological dysfunction during fetal development.

We thank Olivier Lasserre, Didier Dubreuil, and all the shepherds of the UEPAO INRAE experimental unit (UE No. 1297, EU0028) for providing care to the animals and for their help for the installation of test pens. We also thank Anne-Lyse Lainé, Corinne Laclie, and Dominique Gennetay from the hormonal assay laboratory for the cortisol assay. We also thank Julie Lemarchand for her help in performing the open-field test and Chantal Moussu for her help during the blood sampling of animals. We thank Gaëlle Lefort (UMR Physiologie de la Reproduction et des Comportements, INRAE, CNRS, IFCE, Université de Tours, Nouzilly, France) for her help in the statistical analysis of the immunohistochemical labeling. Sesame oil was kindly provided by ADM-SIO, Saint Laurent-Blangy – 62223 France.

All procedures were performed in compliance with the European directive 2010/63/EU on the protection of animals used for scientific purposes and were approved by the Local Ethical Committee for animal experimentation (Comité d’Ethique en Expérimentaion Animale (CEEA) Val de Loire No. 19 Tours, France, No. 2016062917335667) under the reference number: APAFIS#17271-2018102509368756 v4.

The authors have no conflicts of interest to declare.

This work was supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (ERC-2016-CoG to P.G., Grant agreement No. 725149/REPRODAMH); Institut National de la Santé et de la Recherche Médicale (INSERM), France (Grant No. U1172 to P.G.).

M.C: conceptualization, methodology, analysis, writing – original draft. R.F: methodology, acquisition, and analysis. P.G: conceptualization, funding acquisition, supervision, and writing – review and editing. Y.T: conceptualization, supervision, and writing – review and editing.

Paolo Giacobini and Yves Tillet contributed equally to this work.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author: [email protected] or [email protected].