Sex differences in resilience at puberty depend upon divergent effects of a stress steroid at α4βδ GABA_A receptors

Abstract

Resilience is a critical skill that lessens the adverse effects of stress which are increasingly reported in adolescents, where sex differences are noted. It is not known how this process develops in adolescence when unique changes in neuronal properties occur. For this study, pubertal mice were tested for their coping ability following 2-week restraint. We show here that this predictable stress produced resilience in pubertal female mice where time immobile in the forced swim test (FST) decreased by ~50 % (P < 0.0001), an effect that extended into adulthood, and increased escape behavior 8-fold (P = 0.01). This effect was not seen in pubertal male or adult female mice. This process required the stress steroid 3α-OH,5α-pregnan-20-one (THP, allopregnanolone) and its primary target, α4βδ GABA_A receptors (GABARs). These receptors emerge at puberty in prelimbic prefrontal cortex (PL PFC) and basolateral amygdala (BLA), which play a pivotal role in resilient behavior. Stress-induced release of THP decreased anxiety-like behavior (increasing open arm time in the elevated plus maze) and enhanced PFC-dependent learning (temporal order recognition) after 1d restraint in pubertal female, but not male, mice while differentially altering α4βδ expression in PL and BLA. This divergent THP-induced effect ultimately increased and decreased mushroom spine density in PL and BLA, respectively, to produce a circuit optimized for resilient behavior in the pubertal females. These findings demonstrate a novel mechanism for the development of resilience unique to the pubertal period. The results from the present study may suggest therapeutic strategies for adolescent stress which would impact mental health in adulthood.

1. Introduction

Increases in post-pandemic stress have resulted in increasing reports of mental disorders, including depression (Shorey et al., 2022). Adolescence, in particular, is the most vulnerable time for adverse stress responses (Hare et al., 2008) where 1 in 5 teens report adverse mood (Lewinsohn et al., 1993, Shorey et al., 2022). This can result in depression that extends into adulthood (Bhasin et al., 2010, Grant et al., 2003, Pine et al., 1998) and sometimes leads to suicide (Miron et al., 2019). Teenage girls are at the highest risk (Shorey et al., 2022; Moksnes et al., 2010). There is a need to understand the process of developing stress-coping strategies which would reduce the risk of stress-related dysphoria. Despite the numerous studies investigating resilience in adulthood (de Kloet and Joels, 2024, Nestler and Waxman, 2020), to date there is little insight into this learning process at puberty. Potential sex differences in the resilience response have been reported in humans and rodents during adolescence but there are conflicting data because overlapping developmental periods are often merged (Suo et al., 2013, Eiland et al., 2012, Kendig et al., 2011, McCormick et al., 2008, Griffith et al., 2000; Eschenbeck et al., 2007) and underlying mechanisms are rarely identified. Although studies have reported that increased plasticity of pubertal CNS circuits have the potential to affect stress responses during adolescence (Romeo, 2017, Buchanan et al., 1992) when the stress reactivity of the hypothalamic–pituitary–adrenal (HPA) is changing (Romeo and Sciortino, 2021), none have considered the impact of the unique inhibitory α4βδ GABA_A receptors (GABARs) which emerge at puberty that have the potential to facilitate a stress coping process.

The onset of puberty is associated with a 10 d spike in expression of α4βδ GABARs from nearly undetectable pre-pubertal levels in areas involved in emotional regulation (Shen et al., 2007; Aoki et al., 2012), including the prelimbic (PL) prefrontal cortex (PFC) of both sexes (Evrard et al., 2021), which along with the basolateral amygdala (BLA), is implicated in resilience (Covington et al., 2010, Barbour et al., 2020). These receptors do not express at the GABA synapse (Shen et al., 2010). They have a high sensitivity to GABA (Brown et al., 2002), little desensitization and are directly activated by the low concentrations of ambient GABA (100 nM) which are maintained around the neuron due to regulation by the GABA transporters and spillover from GABAergic synapses (Wu et al., 2001). This produces a low amplitude tonic, continuous current (Stell and Mody, 2002) in contrast to the much greater but rapid, intermittent phasic synaptic current. At puberty they express on the dendritic spine, adjacent to the excitatory synapse, when they impair learning (Shen et al., 2010). They are also the most sensitive target for the neurosteroid 3α-OH, 5α-pregnan-20-one (THP or allopregnanolone) (Brown et al., 2002, Belelli et al., 2002), a metabolite of progesterone which can also be formed directly in the brain (Purdy et al., 1991) and is released in response to restraint stress (Cozzoli et al., 2014). This stress-induced release exhibits sex differences (Cozzoli et al., 2014, Sze et al., 2018). Acute stress produces THP levels 5-fold higher in the female PFC compared to the male, and 8-fold higher in the female BLA where the male has no response (Sze et al., 2018). This provides a basis for a potential sex difference in the behavioral response to stress in adolescence.

THP is a GABAR modulator, which can enhance GABA-gated current (Brown et al., 2002, Bianchi and Macdonald, 2003, Belelli et al., 2002), resulting in increased inhibition and reduced anxiety (Bitran et al., 1999) in contrast to other stress hormones which increase anxiety (Reul and Holsboer, 2002; Mitra and Sapolsky, 2008) suggesting that THP may be important for adaptation to stress. However, at puberty, THP can also enhance desensitization of α4βδ GABARs under certain conditions (Shen et al., 2007) and reduce their cell surface expression (Shen et al., 2007), leading to decreased inhibition. These atypical effects of the stress steroid can impact mood and facilitate learning uniquely at puberty, when learning is normally impaired due to α4βδ-induced inhibition of excitatory synapses (Shen et al., 2010). Both actions could impact the development of resilience.

Learning a resilient response to stress during adolescence, assessed as reduced time immobile in the forced swim test (FST) or reduced risk-taking behavior (decreased open arm time on the elevated plus maze (EPM)), could impact later development of depressive-like and anxiety-like behavior, respectively. Although some studies suggest that the FST does not always reflect the anti-depressant effects of some monoamines (Almeida et al., 2020), it closely reflects the anti-depressant effects of THP (Almeida et al., 2018), which is the mode of action of the novel anti-depressant brexanolone, used to treat post-partum depression (Cornett et al., 2021), and is relevant for the hypothesis we are testing here. To further assess active coping, we also employed a modified version of the FST with escape (Nishimura et al., 1988). The FST, FST with escape and EPM were used in the present study rather than measures of anhedonia, which are dependent upon the dopamine reward system (Medel-Matus et al., 2017). We tested for potential sex differences in the development of resilience in response to predictable mild stress in pubertal and adult mice and the role of the GABAergic stress steroid THP in mediating these differences as a novel mechanism unique to puberty. Both adaptive (coping) and maladaptive stress responses are reported in adolescence which are a function of predictability, circuit plasticity and stress severity.

2. Results

2.1. Sex differences in the response to predictable mild stress at puberty: Forced swim test – Stress coping

Male and female mice were subjected to restraint stress (2 h/day for 2 weeks) beginning at the onset of puberty (vaginal opening ~ PND 35, female; preputial separation ~ PND 38, male) (Deboer and Li, 2011) and tested on the forced swim test (FST) the day following the final restraint stress session. A reduction in the time spent immobile in the forced swim test is typically considered a measure of improved coping (Almeida et al., 2020; Walton et al., 2023). There were significant differences in the interaction of sex by stress state (Fig. 1A, F(1,71) = 7.43, P = 0.0081). Post-hoc analysis revealed that pubertal female, but not male, mice exhibited a ~50 % reduction in time immobile following restraint compared to control mice that did not undergo restraint stress, suggesting that pubertal females uniquely develop resilience to predictable stress (q value = 9.117, P < 0.0001). This resilience response was maintained into adulthood, where time immobile was significantly reduced in the female mice restrained at puberty (Fig. 1B, t(34) = 4.54, P < 0.0001). However, for female mice restrained as adults the time immobile was not different from control (Fig. 1B, P = 0.86) suggesting that restraint did not improve coping. In fact, restraint resulted in a change in weight (Table 1) significantly different from controls for both adult females (t(17) = 2.99, P = 0.008) and pubertal males (t(13) = 3.56, P = 0.003), which would be considered an adverse outcome of stress (Hyldelund et al., 2022). In contrast, the weight change of restrained pubertal females was not significantly different from control (Table 1, t(22) = 0.78, P = 0.44). Unlike chronic predictable stress, chronic unpredictable stress (2 weeks) did not reduce time immobile in the FST of the pubertal female mice (Suppl. Fig. 1A) unless combined with predictable restraint stress (Suppl. Fig. 1B; t(18) = 2.65, P = 0.016).

Fig. 1. Chronic predictable stress during adolescence generates resilience only in pubertal female mice.

Inset, Timeline for restraint – 14 d beginning at puberty (lower bar) or adulthood (upper bar) and testing (1–2 d later or in adulthood). PND, post-natal day; Restr, restraint; Con, control. In this and the following figures, data are represented as diamonds (pubertal females), up triangles (pubertal males) or down triangles (adult females). A, Average time immobile (% of total) on the forced swim test (FST) following pubertal restraint stress for female (left) and male (right) mice. F(3,71) = 18.83, P < 0.0001, sex, F(1,71) = 8.92, P = 0.0039, stress, F(1,71) = 35.4, P < 0.0001, sex x stress interaction, F (1,71) = 7.43, P = 0.0081: *P < 0.0001 versus Con, N = 16–21 mice/group. B, Time immobile for female mice tested as adults, following restraint at puberty (left) or restraint as adults (right). *t(34) = 4.54, P = 0.00007 vs. Con, N = 17–19 mice/group. C, Distance scaled (% of total) on the escape device (plastic straw) presented after the 6 min FST for pubertal female (left), pubertal male (center) and adult female (right) mice after restraint versus control. *Mann Whitney U test: U; 31, Z: 2.41, *P: 0.016. N = 5–7 mice/group. D, Open arm time (% of total) on the elevated plus maze (EPM) for female and male mice after pubertal restraint. F(3,72) = 3.4, P = 0.02; sex, F(1,72) = 4.37, P = 0.04. There were no significant interactions, q value = 0.53, P = 0.98, female con vs. restraint; q value = 3.34, P = 0.09, male con vs. restraint. N = 18–21 mice/group. E, Open arm time, EPM for adults restrained as adults. N = 14 mice/group.

Table 1. Change in weight across the 2-week restraint period for the experimental groups.

Mean ± S.E.M. indicated for the change in weight (grams) from day 1 to day 14 of restraint (Res) or no restraint (control, Con). Left, sex and age/pubertal status of experimental animals, Top, group designations. Each row represents a different comparison. The female pub group was the only group to not change weight during restraint.

Sex/age	Group	Group	P value
Female pub	Con	Res
	1.37 ± 0.22	1.17 ± 0.13	0.44
Male pub	Con	Res
	2.86 ± 0.56	0.88 ± 0.16	0.003*
Female adult	Con	Res
	0.47 ± 0.16	1.12 ± 0.14	0.008**

^*t(13) = 3.56;

^**t(17) = 2.99 (N = 7–12 mice/group).

2.1.1. Forced swim test with escape – Stress coping

A modified version of the FST was implemented to assess escape behavior as an additional form of active coping following the 2-week restraint period. For this test, a plastic straw is secured on the rim of the water cylinder following the 6 min FST (Nishimura et al., 1988), and the distance climbed is recorded for the following 2 min. Pubertal female mice that were restrained climbed significantly further than controls (Fig. 1C, 80 ± 20 % of the total versus 8.57 ± 8 %, control, Mann-Whitney U test, U: 31, Z: 2.41, P: 0.016) and were more likely to escape (Table 2, 80 % versus 0 %, control, Fisher’s exact test, P = 0.0101), suggesting that restraint improved active coping. For this group, the time immobile in the FST was strongly inversely correlated with distance climbed on the escape straw (Pearson’s r = −0.89, P < 0.0001). Restraint did not significantly alter climbing distance (P = 0.87, pubertal male; P = 0.48, adult female) or escape attempts (P = 1, pubertal male; P = 1, adult female) for the other groups, consistent with the FST results (Fig. 1A,B).

Table 2. Percentage of escape attempts after the 6 min forced swim.

Left, sex and age/pubertal status of experimental animals, Top, group designations. Con, control; Res, restraint. Each row represents a different comparison made using the binary Fisher’s exact test. Only pubertal female mice had a significant effect of restraint on escape behavior (N = 5–7 mice/group).

Sex	Group	Group	P value
Female pub	Con	Res
	0 %	80 %	*0.0101
Male pub	Con	Res
	28.6 %	33.3 %	1
Female adult	Res-VEH	Res-FIN
	17 %	40 %	1

2.2. Elevated plus maze – Risk-taking behavior

In contrast to the FST results, there were no differences in the interaction of sex by stress state for open arm time on the elevated plus maze (EPM, Fig. 1D, F(1,72) = 3.82, P = 0.05) or distance traveled (Table 3, F(1,72) = 0.68, P = 0.41) of pubertal mice. These results suggest that predictable stress during adolescence does not alter risk-taking behavior. Similarly, open arm time (Fig. 1E, P = 0.54) and distance traveled (Table 3, P = 0.31) were not changed in adult, female mice restrained as adults.

Table 3. Distance traveled for the elevated plus maze experiments.

Mean ± S.E.M. indicated for elevated plus maze experiments. Left, age during the 2-week restraint stress, Top, group designations. Con, control; Res, restraint; no Res, no restraint; VEH, Vehicle; FIN, Finasteride; KO, α4−/−. Each partial row across 3 columns represents a different comparison. No comparisons were significantly different.

Age at Restraint			P value			P value
Pubertal	Female WT-Con	Female WT-Res		Male WT-Con	Male WT-Res
	2794 ± 94	2644 ± 69.8	0.23	2597 ± 93.5	2583 ± 64	0.23
Pubertal	Female WT-Res-VEH	Female WT-Res-FIN		Male WT-Res-VEH	Male WT-Res-FIN
	2977 ± 75	2831 ± 94	0.24	2578 ± 56	2470 ± 76	0.26
Pubertal	Female KO-Con	Female KO-Res		Male KO-Con	Male KO-Res
	2951 ± 123	2772 ± 148	0.36	2712 ± 102	2741 ± 169	0.89
Pubertal	Female WT-VEH- no Res	Female WT-FIN-no Res		Male WT-VEH- no Res	Male WT-FIN-no Res
	2565 ± 159	2668 ± 91	0.58	2500 ± 165	2398 ± 184	0.69
Adult	Female WT-Con	Female WT-Res
	2507 ± 106	2388 ± 48	0.21
Adult	Female WT-Res-VEH	Female WT-Res-FIN
	3372 ± 134	3581 ± 162	0.35
Adult	Female KO-Con	Female KO-Res
	2566 ± 77	2484 ± 144	0.63
Adult	Female WT-VEH- no Res	Female WT-FIN-no Res
	2554 ± 90	2645 ± 129	0.56

2.3. Resilience response to stress in the FST – Role of THP at puberty

In order to test whether the stress steroid THP played a role in the resilience response, mice were injected with the 5α-reductase blocker finasteride, which prevents THP synthesis (Smith et al., 2006), or with vehicle, prior to and during the 2-week restraint stress protocol. Although this drug also blocks formation of dihydrotestosterone, circulating levels of this androgen are negligible by the onset of puberty (Jean-Faucher et al., 1985). In the absence of THP production, the time immobile on the FST for pubertal female mice was increased by ~ 50 % compared to vehicle-injected mice (Fig. 2A, t(28) = 4.98, P < 0.0001) suggesting that stress-induced release of THP is necessary for this resilient response. Finasteride had no effect on time immobile when not combined with restraint at puberty (Table 4, % time immobile, finasteride-no restraint, 54.5 ± 1.8; vehicle-no restraint, 57.2 ± 3.7; P = 0.58). In contrast, finasteride had no effect on % time immobile in the FST in restrained pubertal males (Fig. 2B, P = 0.74) or adult females (Fig. 2C, P = 0.35) nor did it have any effect in non-restrained mice (Table 4).

Fig. 2. Blockade of THP formation and knock-down of α4 prevent the development of resilience in female mice at puberty.

Inset, Timeline for restraint and injections – 14 d beginning at puberty or adulthood and testing (1–2 d later). Upper bar, WT mice: restraint + FIN or VEH; lower bar, α4 knock-out (KO) mice: restraint or no restraint control. (FIN, finasteride; VEH, vehicle. A-C, Time immobile, forced swim test (FST), restraint plus FIN or VEH (left) or after α4 knock-out (KO, right) in pubertal female (A), pubertal male (B) or adult female (C) mice. Pubertal female, FIN, *t(28) = 4.98, P < 0.0001 vs. VEH, N = 15 mice/group (A); adult female, *t(16) = 2.13, P = 0.049 (C). D-F, Open arm time, elevated plus maze (EPM), restraint plus FIN or VEH (left) or after α4 knock-out (KO, right) in pubertal female (D), pubertal male (E) or adult female (F) mice. Pubertal female, FIN, *t(28) = 2.52, P < 0.018 vs. VEH, N = 15 mice/group (D); KO, *t(28) = 3.24, P = 0.0031, N = 15 mice/group (D).

Table 4. Effects of 2-week finasteride without restraint on behavioral outcomes.

Mean ± S.E.M. for % time in the open arm on the elevated plus maze (left) and % time immobile on the forced swim test (right) after 2-week administration of finasteride (FIN) or vehicle (VEH) without restraint to the indicated groups of wild-type mice. Left, sex and age of experimental animals, Top, group designations. Pub, pubertal. Each row represents a different comparison within the behavioral sub-group. No comparisons were significantly different (7–10 mice/group).

Elevated plus maze Forced swim test
Sex	VEH	FIN	P value	VEH	FIN	P value
Female pub	36.1 ± 2.4	34.3 ± 1.4	0.53	57.2 ± 3.7	54.5 ± 1.8	0.58
Male pub	35.7 ± 2.6	36.5 ± 2.1	0.82	60.3 ± 4.7	66.9 ± 2.3	0.25
Female adult	31.0 ± 2.9	36.4 ± 2.4	0.18	59.2 ± 4.1	55.3 ± 2.4	0.44

2.4. Resilience response to stress in the FST – Role of α4βδ GABARs at puberty

Pubertal female α4 knock-out mice subjected to the restraint protocol did not develop resilience as evidenced by a similar time immobile as the controls (Fig. 2A, P = 0.14), suggesting that α4βδ GABARs, the primary target for THP, are also necessary for the resilient response to stress. Time immobile in the FST was not altered by restraint of pubertal male α4 knock-out mice (Fig. 2B, P = 0.14). However, restraint of adult female α4 knock-out mice reduced the time immobile significantly compared to control (Fig. 2C, t(16) = 2.13, P = 0.049).

2.5. Elevated plus maze – Effect of THP and α4βδ GABARs during adolescence

Blockade of THP with finasteride concomitant with restraint reduced open arm time on the EPM of wild-type pubertal female mice by 15 % (Fig. 2D, t(28) = 2.52, P = 0.018), while α4 knock-out reduced open arm time by ~ 30 % (Fig. 2D, t(28) = 3.24, P = 0.0031) without altering distance traveled (Table 3, Fin, P = 0.24; α4 KO, P = 0.36) suggesting that THP effects at α4βδ GABARs exert a protective effect against stress-induced anxiety-like behavior during adolescence. In contrast, finasteride administered without restraint had no effect on open arm time (Table 4). Finasteride treatment did not alter open arm time of pubertal male mice (Fig. 2E, P = 0.63) or adult female mice (Fig. 2F, P = 0.06) nor did restraint of pubertal male α4 knock-out mice (Fig. 2D, P = 0.22) or adult female α4 knock-out mice (Fig. 2F, P = 0.33). Finasteride without restraint had no effect in these groups (Table 4). Distance traveled was similarly not affected (Table 3, Fin, P = 0.26; α4 KO, P = 0.89).

2.6. Sex differences in stress-induced effects of THP at puberty

Several studies suggest that stress-induced release of THP is greater in females (Cozzoli et al., 2014, Sze et al., 2018). Thus, we examined whether there were sex differences in the effect of THP acutely released by stress in the present study. To this end, we tested whether blockade of the THP formed during 1 d of 2 h restraint stress using the 5α-reductase blocker finasteride would have different outcomes on open arm time in the EPM for males and females tested 1d later when anxiety-reducing effects have been reported (Yoshizawa et al., 2017). There were significant differences in the sex by drug state interaction (Fig. 3A, F(1,34) = 19.5, P < 0.0001). Post-hoc analysis revealed that finasteride resulted in a ~37 % reduction in open arm time on the EPM in pubertal female, but not male, mice following one day of restraint (female, q value = 6.73, P = 0.00024 vs. vehicle; male, q value = 1.76, P = 0.61 vs. vehicle). These results suggest that stress-induced release of THP reduces anxiety only in females following one day of restraint. In contrast, administration of finasteride alone, without restraint, had no effect for either sex (Fig. 3B, female, P = 0.67; male, P = 0.22).

Fig. 3. Restraint stress-induced release of THP reduces open arm time and improves PFC-dependent learning only in pubertal female mice.

Inset, Timeline, Upper bar (A, B, E), day 1: 2 h restraint stress (or no restraint control), preceded by finasteride (FIN) or vehicle (VEH) injections; day 2: FIN or VEH followed by testing (EPM (A, B) or temporal order recognition (TOR) task (E)) 2 h later. Lower bar (C, D), day 1: 2 h restraint stress (or no restraint control); day 2: TOR task. A, Open arm time on the EPM after 1 d restraint plus FIN or VEH for pubertal female (left) and male (right) mice. F(3,35) = 10.79, P < 0.0001, sex, F(1,35) = 7.26, P = 0.011, drug, F(1,35) = 7.71, P = 0.009, sex x drug interaction, *F(1,35) = 19.8, P < 0.0001: female, q value = 6.73, *P = 0.00024 versus VEH, N = 8–10 mice/group. B, Open arm time on the EPM after FIN or VEH without restraint for pubertal female (left) and male (right) mice. No comparisons were significant. N = 8–10 mice/ group. C, D, Discrimination ratio on the TOR task after 1 d restraint for female (C) and male (D) mice. Female, *t(16) = 2.94, P = 0.01 vs. Con. N = 10 mice/group (B). E, Discrimination ratio on the TOR task after 1 d restraint plus FIN or VEH (left) or FIN/VEH without restraint (right) to female mice, Restraint, FIN vs. VEH, *t(16) = 7.73, P < 0.0001, N = 8–10 mice/group.

2.7. PFC-dependent learning is improved by restraint stress-induced release of THP in pubertal females

Our previous studies have shown that the emergence of α4βδ GABARs at puberty impairs learning, an effect that can be reversed by administration of the stress steroid THP (Shen et al., 2010). Thus, we tested whether restraint stress would improve learning on the PFC-dependent temporal order recognition (TOR) task (Barker et al., 2007). One day of restraint resulted in a 5-fold increase in the discrimination ratio (Fig. 3C, t(16) = 2.94, P = 0.01 vs. control) for pubertal females. This stress-induced enhancement of learning was prevented by blockade of THP synthesis using finasteride (Fig. 3E, t(16) = 7.73, P < 0.0001 vs. vehicle) suggesting that stress-induced release of THP was responsible for enhancing learning in pubertal females. Finasteride administered without restraint had no effect on the discrimination ratio compared to vehicle (Fig. 3E, P = 0.72). In addition, restraint had no effect on TOR performance in males (Fig. 3D, P = 0.29). There were no changes in the number of approaches for any groups (Table 5), indicating no difference in general locomotor activity.

Table 5. Number of approaches for the temporal order recognition task experiments.

Mean ± S.E.M. indicated for the temporal order recognition (TOR) task. Left, sex of experimental animals, Top, group designations. Con, control; Res, restraint; no Res, no restraint; VEH, Vehicle; FIN, Finasteride. Each row represents a different comparison. No comparisons were significantly different (N = 7 – 10 mice/group).

Sex	Group	Group	P value
Female pub	Con	Res
	34.1 ± 1.3	30.2 ± 2.1	0.14
Male pub	Con	Res
	32 ± 1.9	28 ± 1.9	0.16
Female pub	Res-VEH	Res-FIN
	28.8 ± 2.25	26.75 ± 0.92	0.45
Female pub	no Res-VEH	no Res-FIN
	32.95 ± 1.16	30.86 ± 1.27	0.24

2.8. Changes in α4 expression in female prefrontal cortex and amygdala during chronic restraint stress during adolescence

We have shown previously that THP can decrease expression of α4βδ GABAR during their period of high pubertal expression in the female mouse hippocampus (Shen et al., 2007). Therefore, we examined levels of α4 expression on layer 5 pyramidal cells in PL and BLA on days 2, 7 and 14 of the 2-week restraint stress protocol to see whether stress-induced THP release altered this parameter. Compared to control, expression of α4 was significantly decreased on days 2 and 7 of restraint in PL (Fig. 4A–C, day 2, t(30) = 2.44, P = 0.04; day 7, t(36) = 2.58, P = 0.03), but unchanged at day 14 (PND 48, P = 0.93) when control levels of α4 expression were decreased by half from their elevated pubertal levels (F(2,67) = 82.9, P < 0.0001 vs. days 2 and 7). In contrast, expression of α4 was significantly increased in BLA on days 2 and 7 of restraint (Fig. 4D–F, day 2, t(36) = 2.4, P = 0.04; day 7, t(40) = 2.23, P = 0.04), but returned to control levels by day 14 (P = 0.13). Control levels of α4 expression were more than 3-fold higher on PND 41 compared to PND 36 and 48 (F(2,62) = 316.4, P < 0.0001) also reflecting the pubertal increase in α4 expression in BLA. The restraint stress-induced alterations in α4 expression were prevented by administration of finasteride to block THP formation during the first 2d of restraint. In the absence of THP, restraint stress instead significantly increased α4 expression in PL (Fig. 5A–C, t(44) = 2.5, P = 0.04) and decreased α4 expression in BLA (Fig. 5D–F, t(44) = 3.7, P = 0.006) compared to vehicle-treated restrained controls. These data show that altered α4 expression was due to stress-induced release of THP.

Fig. 4. Restraint stress alters expression of the α4 GABAR subunit in PL and BLA of pubertal female mice.

Inset, Timeline of IHC assessment of α4 expression after 2, 7 or 14 D of restraint. A, PL, Averaged data, fluorescence intensity of α4 immunostaining as a percent of control, 2D, *t(8) = 2.44, P = 0.04, 7D, *t (8) = 2.58, P = 0.03. B, Representative images, PL, Con (top) and Restr (bottom), 2D restraint. Boxes, soma magnified in C, 2D. C, Representative soma, 2, 7 and 14 D restraint. D, BLA, Averaged data, Fluorescence intensity as a percent of control, 2D, *t(8) = 2.4, P = 0.04, 7D, *t(10) = 2.23, P = 0.04. E, Representative images, BLA, Con (top) and Restr (bottom), 2D restraint. Boxes, soma magnified in F, 2D. F, Representative soma, 2, 7 and 14 D restraint. N = 5–6 mice/group, 3–4 neurons/mouse.

Fig. 5. Blockade of THP formation prevents the changes in α4 expression produced by 2 d of restraint stress of pubertal female mice.

Inset, timeline of vehicle (VEH) or finasteride (FIN) injection and α4 assessment. A, PL, Averaged data, fluorescence intensity of α4 immunostaining as a percent of control, *t(8) = 2.45, P = 0.04. B, Representative images, PL, Con (top) and Restr (bottom), Boxes, soma magnified in I. C, Representative soma. N = 5 mice, 4–6 neurons/mouse. D, BLA, Averaged data, fluorescence intensity as a percent of control, *t(8) = 3.7, P = 0.006. E, Representative images, BLA, Con (top) and Restr (bottom). Boxes, soma magnified in L. F, Representative soma. N = 5 mice/group, 3–8 neurons/mouse.

2.9. Pubertal restraint stress alters mushroom spine density in female PL and BLA due to stress-induced release of THP

Mushroom spine density of layer 5 PL pyramidal cells of the adolescent female mouse exhibited a ~55 % increase following the 2-week restraint period (Fig. 6A,C, t(65) = 4.6, P < 0.0001) with no effect on other spine types or on total spine density. In contrast, restraint decreased mushroom spine density on BLA pyramidal neurons by 50 % (Fig. 6B,D, t(65) = 3.99, P = 0.0018) with an overall 20 % decrease in total spines (t(65) = 2.29, P = 0.04), but no change in other spine types. These changes were prevented when finasteride was administered during the 2-week restraint period. In the absence of THP, restraint decreased mushroom spine density in the PL by ~30 % (Fig. 7A,C, t(48) = 4.7, P < 0.0001) and increased mushroom spine density in the BLA by more than 2-fold (Fig. 7B,D, t(42) = 12.58, P < 0.0001) compared to vehicle-treated restrained mice, resulting in mushroom spine densities similar to non-restrained controls for each group. These data suggest that THP was necessary for alterations in mushroom spine density following restraint stress in both CNS regions.

Fig. 6. Restraint stress during adolescence alters mushroom spine density in the PL and BLA of pubertal female mice.

A, B, Averaged densities for thin, mushroom, stubby and total spines/10 μm in the PL (A) and BLA (B) of control or restrained (2 weeks) female mice. A, PL, Averaged data, mushroom spines, *t(65) = 4.6, P < 0.0001 vs. Con. B, BLA, Averaged data, mushroom spines, *t(65) = 3.99, P = 0.0018 vs. Con; total spines, *t(65) = 2.29, P = 0.04 vs. Con. C-D, Representative images for PL (C) and BLA (D). N = 7 mice, 4–5 dendrites/mouse/group.

Fig. 7. Blockade of THP formation prevents the changes in mushroom spine density produced by restraint stress in pubertal female mice.

A, B, Averaged densities for thin, mushroom, stubby and total spines/10 μm in the PL (A) or BLA (B) of female mice injected with finasteride (FIN) or vehicle (VEH) during the 2-week restraint. A, PL, Averaged data, mushroom spines, *t(48) = 4.7, P < 0.0001 vs. Con. B, BLA, Averaged data, mushroom spines, *t(42) = 12.58, P < 0.0001 vs. Con; total spines, *t(42) = 3.71, P = 0.0006 vs. Con. C,D, Representative images for PL (C) and BLA (D). N = 5 mice, 4–7 dendrites/mouse/group.

In contrast, restraint stress was unaccompanied by changes in the density of any spine types in PL and BLA of adolescent male mice (Fig. 8, A,C, PL, P = 0.73; B,D, BLA, P = 0.62).

Fig. 8. Restraint stress does not alter dendritic spine density of pubertal male mice.

A, B, Averaged densities for thin, mushroom, stubby and total spines/10 μm in the PL (A) and BLA (B) for control or restrained (2 weeks) male mice. C, D, Representative images for PL (C) and BLA (D). N = 5 mice, 3–4 dendrites/mouse/group.

3. Discussion

The results from this study suggest that only pubertal female mice exhibit resilience, but not pubertal male or adult female mice, following 2 weeks of predictable, mild stress as reflected by a reduced time immobile in the forced swim test and an increase in escape frequency. This resilient response was dependent upon a novel mechanism: release of the stress steroid THP and its effect at its primary target, α4βδ GABARs, which transiently increase expression in the PL of both sexes at the onset of puberty (Evrard et al., 2021). This outcome was independent of any changes in locomotor activity.

In this study, we show that restraint stress reduces time immobile in the FST, one type of active coping which has been documented in many reports (Almeida et al., 2020). This is in contrast to passive coping, where mice experience an increase in time immobile in this test. Both have been suggested as adaptive responses to stress (Almeida et al., 2020; de Kloet and Molendijk, 2016). Passive coping would conserve energy in an inescapable stress paradigm which could be advantageous (de Kloet and Molendijk, 2016). However, active coping is proactive and would optimize alternate escape opportunities which are not immediately apparent. This possibility was borne out in the present study when an escape route was presented at the end of the 6 min inescapable FST protocol. In this case, only pubertal female mice undergoing restraint were more likely to escape than the controls. We also show that % time immobile and climbing distance have a strong inverse correlation (Pearson’s r = −0.89, P < 0.0001), suggesting that the decrease in time immobile for the FST experiments without escape would represent an increased potential for escape. Two potential confounds (Almeida et al., 2020).for considering a decrease in time immobile as adaptive resilience have been eliminated in this study, locomotor activity (which did not change) and anxiety (which did not change). Although some studies suggest that time immobile on the FST is not an accurate predictor of the anti-depressant potential of some monoamines (Nestler and Hyman, 2010), this parameter tested in rodent studies is strongly correlated with the anti-depressant effects of THP (Almeida et al., 2018; Walton et al., 2023), which is administered to humans as brexanolone for post-partum depression (Cornett et al., 2021), a finding especially relevant for the present study. In fact, THP levels in the CSF are reduced in depressed individuals (Uzunova et al., 1998), further underscoring its role in facilitating adaptive coping.

Restraint did not alter time immobile in the FST or escape behavior for pubertal males and adult females, suggesting that their coping ability was not altered. In addition, both of these groups exhibited a maladaptive stress response as evidenced by their body weight change across the 2-week restraint period which was significantly different from control. The body weight of restrained pubertal females, in contrast, was not different from control, suggesting that this group was better able to adapt to the restraint stress. Numerous studies have shown that both increases and decreases in appetite can result from chronic stress as maladaptive responses that lead to altered body weight in humans (Block et al., 2009; Coccurello, 2019; Hyldelund et al., 2022).

Resilience is the process of successfully adapting to difficult life experiences (Nestler and Russo, 2024). Both human and rodent studies suggest that this is associated with activation of the PL subregion of the medial PFC (mPFC) and inhibition of the BLA (Keynan et al., 2019, Phan et al., 2005, Milad and Quirk, 2002) through a learning process. Indeed, activation of the mPFC does not result in resilience unless accompanied by exposure to stressful stimuli (Amat et al., 2006), and is prevented by protein synthesis blockers (Amat et al., 2006), suggesting that it involves selective learning. The results from the present study suggest that pubertal female mice, but not male mice, exhibit markedly improved PFC-dependent learning as a result of restraint stress-induced release of THP, concomitant with THP-induced decreases in anxiety, during the first week of restraint when the process of learning to cope would need to take place. Both end-points were prevented by the 5α-reductase blocker finasteride which blocks formation of THP (Mukai et al., 2008). Although this drug also blocks formation of dihydrotestosterone, circulating levels of this androgen are negligible by the onset of puberty (Jean-Faucher et al., 1985).

The predictable nature of the chronic restraint stress would create a repetitive experience consistent with optimal “Hebbian” learning (Oberauer et al., 2015) which was maximized by stress-induced release of THP during adolescence (Shen et al., 2010). This is in contrast to chronic unpredictable stress (CUS), which typically decreases resilience (Wankhar et al., 2020, Walton et al., 2023). In the present study, CUS during adolescence had no effect on time immobile in the FST unless it was paired with the predictable restraint stress which led to a decrease in time immobile. This suggests that the learning of coping skills in response to the repetitive restraint episodes was robust enough to increase resilience even in the presence of unpredictable stress for pubertal female mice. It is noteworthy that combined predictable and unpredictable stress during puberty increased resilience because, in contrast, this combination of stressors has been shown to decrease resilience in adult mice, increasing time immobile in the FST to a greater degree than each stressor alone (Qiao et al., 2020).

There are many studies suggesting that adaptation to stress during adolescence may benefit from the potential for increased plasticity at this time (Romeo and McEwen, 2006, Romeo, 2015, Buchanan et al., 1992). Pre-pubertal mice have greater levels of synaptic plasticity and learning (Shen et al., 2010), likely due in part to their greater expression (Sans et al., 2000) of the slowly deactivating NMDA 2B receptor (Santucci and Raghavachari, 2008), which is associated with enhanced synaptic plasticity/learning (Plattner et al., 2014). However, pubertal mice do not display this optimal learning because plasticity is held in check by inhibitory α4βδ GABARs which emerge at puberty on the dendritic spine (Shen et al., 2010, Evrard et al., 2021), where they impair the activation of NMDA receptors necessary for synaptic plasticity and learning (Shen et al., 2007). The expression of these inhibitory receptors is reduced by the stress steroid THP (Shen et al., 2010), as shown here in the PL, during the first week of restraint. This reduction in inhibition would facilitate the depolarization necessary for robust synaptic plasticity (Shen et al., 2010), and is evidenced in the present study by a more than 5-fold improvement in PFC-dependent learning after restraint stress-induced release of the steroid in the female because it was blocked by finasteride. Finasteride administered without restraint did not alter learning, suggesting that stress-induced release of THP was the mediating factor. In contrast, restraint stress did not alter learning in the male which exhibits a markedly reduced THP response (Sze et al., 2018).

Expression of α4βδ GABARs was increased in the BLA after THP release during the first week of restraint, most likely as a result of the high levels of locally produced estradiol here (~60 pg/ml), which are 9-fold greater than in PL (Barker and Galea, 2009) and which would increase α4βδ expression in the presence of THP (Keating et al., 2019). The resultant increase in inhibition of the BLA would reduce anxiety as seen in the present study. These divergent effects of THP – to facilitate learning, by reducing inhibition via reduced α4βδ GABARs in the PL, and to reduce anxiety, by increasing inhibition via increased α4βδ GABARs in the BLA – would both facilitate the development of resilience. These actions are in contrast to those of other stress hormones, including CRH and corticosterone, which increase anxiety (Reul and Holsboer, 2002, Mitra and Sapolsky, 2008). The actions of THP at puberty are unique due to this transient period of increased α4βδ expression; THP impairs learning at other ages, (Johansson et al., 2002, Frye and Sturgis, 1995), when it also reduces anxiety (Bitran et al., 1999).

THP has additional divergent actions acutely at puberty due to its unique polarity-dependent effects at α4βδ GABARs (Shen et al., 2007). When Cl- flux gated by this receptor is inward, as reported for activated layer 5 pyramidal cells of the PL (Rosenkranz and Grace, 2002, Kim et al., 2021), THP enhances its desensitization (Shen et al., 2007) and thus decreases its inhibitory current, which would facilitate learning. The polarity-dependent enhancement of receptor desensitization by THP at α4βδ is reflective of the reported polarity-dependent rates of desensitization for the homologous α6β3δ (Bianchi et al., 2002). In the case of α4βδ, this polarity-dependent effect is limited to the transient period of its high pubertal expression (Evrard et al., 2021). However, when Cl- flux gated by this receptor is outward, as reported for the pyramidal cells of the BLA (Rosenkranz and Grace, 2002), THP enhances this inhibitory current (Shen et al., 2007), an effect which would suppress BLA activity, thereby reducing anxiety. This was verified by the increase in anxiety observed after blockade of THP formation during acute restraint stress in the female. These divergent acute effects of the steroid may also underlie the distinct effects of THP to decrease or increase α4βδ expression in PL and BLA, respectively, as we have shown (Kuver et al., 2012, Kuver and Smith, 2016).

The 2-week restraint period was ultimately associated with increases in the density of mushroom spines, thought to represent stable “memory” spines (Bourne and Harris, 2007), in layer 5 of the PL in the resilient females. This may be a result of PFC-dependent learning of coping skills because mushroom spine density decreased when THP formation was blocked by finasteride. Similar to results from the present study, other studies have shown that predictable stress increases dendritic spines in basilar dendrites of the mPFC in resilient and adolescent female rats (Yang et al., 2015, Melo et al., 2020), while unpredictable stress that does not result in resilience decreases mushroom spine density here (Woo et al., 2021).

In contrast to the PL, mushroom spine density of BLA neurons decreased as a result of stress-induced release of THP. This is likely due to the increase in α4βδ GABARs and their THP-induced potentiation which would enhance the pruning of these synapses, as we have shown in other areas, including PL, during adolescence of both sexes (Evrard et al., 2021). In the present study, the decrease in mushroom spine density in the BLA would reduce the impact of excitatory input to this area, which would also facilitate the development of resilience.

In human studies, increased activity of the left mPFC and reduced activity of the amygdala is reported in resilient subjects (Phan et al., 2005, Davidson, 2004, Eaton et al., 2022). Biofeedback therapies which reduce blood oxygen levels in the amygdala have been reported to result in better stress coping (Keynan et al., 2019). In contrast, increases in amygdala responsivity and decreases in mPFC blood flow are reported in subjects with low resilience (Barbour et al., 2020, Phan et al., 2005). In rodent studies, optogenetic stimulation of the mPFC (Covington et al., 2010), and specifically the PL (Moshe et al., 2016), induces resilience in response to stress and in a depressive rat line, while lesions of the BLA decrease the fear response (LaBar and LeDoux, 1996).

Other rodent studies have shown that adolescent restraint stress can result in resilience but there is no consensus on potential sex differences nor has an underlying CNS mechanism been identified (Hong et al., 2012, Suo et al., 2013, McCormick et al., 2008) in contrast to prenatal stress which has greater effects in females (Thomas and Becker, 2019). Many studies use adolescent periods which span both pre-pubertal and pubertal stages (Eiland et al., 2012, McCormick et al., 2008, Suo et al., 2013), frequently comparing pubertal females to pre-pubertal males, which would have different levels of α4βδ expression and thus different responses to THP (Shen et al., 2007). The results from the present study suggest that resilience in adolescent females is dependent upon the stress-induced release of this steroid because preventing THP formation with the 5α-reductase blocker finasteride prevented the development of resilience after 2 weeks of restraint stress. Finasteride alone without restraint had no effect on time immobile in the FST suggesting that basal levels of THP do not have a significant effect on this coping behavior.

THP is released 20–45 min after the onset of sustained stress (Purdy et al., 1991), and high levels of the steroid persist for 24 h, when they can reduce anxiety (Yoshizawa et al., 2017). Our results show that only pubertal females exhibited long-lasting stress-induced THP release, because finasteride administration during restraint decreased open arm time 24 h after 2 h of restraint stress in females but not in males. Finasteride without restraint had no effect suggesting that stress-induced release of THP specifically reduced anxiety-like behaviour. Up to 8fold greater stress-induced increases in THP levels have been reported in females in plasma and frontal cortex as well as in amygdala, where males do not have a stress-induced increase (Sze et al., 2018). Neither α4 knock-out nor blockade of THP synthesis altered coping or risk-taking behavior in the male in the present study consistent with the lack of stress-induced THP release in the male. The anxiety-reducing effect of THP at α4βδ GABARs may have also prevented increases in anxiety-like behavior after 2-week restraint in pubertal female mice because α4 knock-out and blockade of THP formation during restraint decreased open arm entries post-pubertally. This suggests that the stress-induced release of THP may also be a protective mechanism for mitigating stress-induced anxiety responses during adolescence. In addition, restraint decreased time immobile in the FST after knock-out of α4βδ GABARs in adult female mice, suggesting that α4βδ-mediated inhibition may be detrimental for the development of resilience in adulthood possibly due to its typical effect to impair learning at this time (Johansson et al., 2002, Frye and Sturgis, 1995).

The development of coping skills in response to CUS is also related to THP and its primary GABAR target (Walton et al., 2023). A recent study suggests that CUS reduces levels of both THP in BLA and δ-GABAR expression on parvalbumin + GABAergic interneurons in BLA in association with decreases in time immobile, an outcome which could be reversed by increasing endogenous THP in the BLA or by treatment with a THP analog (Walton et al., 2023). This study (Walton et al., 2023) also demonstrates that THP plays a pivotal role in improving coping behavior, similar to the present study. However, unlike the present results, its δ-GABAR target (α1βδ) is localized to interneurons (Glykys et al., 2007) rather than the pyramidal cells which express α4βδ GABARs (Shen et al., 2010). This suggests that a different underlying process is required for the development of resilience in response to predictable vs. unpredictable stress.

The HPA axis is also involved in the stress response. In particular, increases in levels of CRH (corticotropin releasing hormone) may contribute to anxiety-like behavior and the reduction in coping behavior associated with CUS as well as with social defeat stress (Dedic et al., 2019, Veenit et al., 2014). Restraint stress does not activate CRH neurons in the amygdala (De Francesco et al., 2015). However, administration of THP reduces hypothalamic CRH increased by restraint stress only in females (Boero et al., 2022), which may have contributed to the sex difference in the acute effects of restraint stress on anxiety-like behavior in the present study. However, since local infusion of THP into the hippocampus reduces anxiety (Bitran et al., 1999) due to its potentiation of GABAergic inhibition, additional indirect mechanisms are not required for THP effects on anxiety-like behavior.

Additional factors which may contribute to resilience in adults include serotonin, dopamine, and/or corticosterone (Willmore et al., 2022, de Kloet and Joels, 2024, Puglisi-Allegra and Andolina, 2015) as well as other genetic factors (Nestler and Waxman, 2020). The development of resilience in response to social defeat stress, in particular, is dependent upon the dopamine reward system (Krishnan et al., 2007), where distinct populations of BLA neurons exhibit robust discrimination between reward choices, an effect not seen in susceptible animals (Xia et al., 2025).

3.1. Limitations and alternate conclusions

Although the most parsimonious explanation for the mechanism underlying the increase in active coping in female adolescent mice is via THP effects at α4βδ GABARs, we cannot rule out other contributing factors including the effects mediated by the HPA axis and dopamine or serotonin effects on limbic circuits. Additional behavioral measures would indicate whether other types of stress responses (such as anhedonia or social defeat stress) exhibit sex differences in adolescence. One alternate conclusion could be that THP effects at other limbic CNS sites which project to the PL and BLA mediate the observed effect. Investigation of local manipulations in THP and α4βδ following restraint stress is beyond the scope of the present study but will be the focus of future investigations.

3.2. Conclusions

Adolescent girls are reported to respond more to stress and are more likely to experience depressive-like symptoms than are teenage boys (Mosknes et al., 2010). The results from the present study may suggest a factor, the experience of predictability, which could mitigate against this outcome, even in the presence of unpredictable stress, and also describe a novel mechanism unique to puberty dependent upon the stress steroid THP and its primary target, α4βδ GABARs. These findings may suggest an additional therapeutic use for THP, which is the mode of action of brexanolone, recently shown to ameliorate post-partum depression (Cornett et al., 2021). The present results may lead to novel therapeutic approaches for stress-related disorders in adolescence to prevent future psychiatric disorders in adulthood.

4. Experimental procedures

4.1. Analytic plan

1. Hypothesis 1: There will be sex differences in the active coping response to predictable restraint stress during adolescence because of sex differences in the THP response to stress.

   Statistical analysis: FST and EPM: 2-factor ANOVA (sex by stress state); FST with escape: non-parametric tests: Mann Whitney U test for climbing distance and Fisher’s exact test for escape/no escape; Weight change before and after restraint: t test

2. Hypothesis 2: Increases in the active coping response to predictable restraint stress will not be observed in adult females because they express low levels of α4βδ GABARs.

   Statistical test: 2-sample t-test (control and predictable restraint stress)

3. Hypothesis 3: Chronic unpredictable stress (CUS) will not increase active coping in the pubertal female.

   Statistical analysis: 2-sample t-test (control and CUS)

4. Hypothesis 4: The addition of chronic predictable stress (restraint) along with CUS will restore the increased active coping response to stress in pubertal females.

   Statistical analysis: 2 sample t-test (control and predictable stress + CUS).

5. Hypothesis 5: Increased active coping in response to predictable stress in the pubertal female is due to the stress steroid THP and α4βδ GABARs.

   Statistical analysis: 2-sample t-test on individual comparisons of the effect of finasteride (to block THP) and α4 knock-out.

6. Hypothesis 6: There are sex differences in the anxiolytic effect of 1 day of restraint stress due to sex differences in release of the stress steroid THP.

   Statistical analysis: 2 factor ANOVA (sex by drug state) – 1 day restraint – finasteride – EPM.

7. Hypothesis 7: PFC-dependent learning is facilitated only in pubertal female mice due to stress-induced release of THP.

   Statistical analysis: 2-sample t-test on measures of PFC-dependent learning with or without finasteride (to block THP release).

8. Hypothesis 8: Chronic predictable restraint stress increases mushroom spines in female PFC, associated with increases in learning and memory, but not in male PFC.

   Statistical analysis: Nested t-test comparing spine types after predictable restraint stress versus control in males and females, with or without finasteride to block formation of stress-induced release of THP.

9. Hypothesis 9: Stress-induced release of THP decreases α4 GABAR expression in PFC, but increases its expression in BLA in the pubertal female.

   Statistical analysis: Nested t-test across the 14 day restraint period versus control in pubertal females, with or without finasteride to block formation of stress-induced release of THP.

4.2. Animals

Female and male wildtype C57BL/6 mice (Jackson Labs, Bar Harbor, Maine) were subjected to restraint stress at puberty onset (for females, post-natal day (PND) 35 as confirmed by vaginal opening or for males, PND 38, based on average age of preputial separation) (Deboer and Li, 2011) and tested post-pubertally (PND 49–53) or later in adulthood (PND 75–76) (Fig. 1, inset). In some cases, female mice were subjected to restraint stress in adulthood (PND 75–77). In most cases, mice were tested for 2 d following the stress regimen unless otherwise indicated. In most cases, they were housed 5 animals per cage throughout the experiment and illumination was maintained in a 12:12-h reverse light–dark cycle (light on 23:00–11:00). Estrous cycle stage was determined by the vaginal cytology in 8-week-old animals with established regular cycles, and these mice were not used in the stage of proestrus when GABAR expression changes (Sabaliauskas et al., 2014). Pubertal mice typically do not exhibit estrous cyclicity (Hodes and Shors, 2005). This study was carried out following the principles of the National Institute of Health Office of Laboratory Animal Welfare (OLAW) and recommendations of the SUNY Downstate Institutional Animal Care and Use Committee (IACUC). The protocol was approved by the SUNY Downstate IACUC before the study was initiated.

In some cases mice were injected with finasteride (50 mg/kg, i.p.) using a dosing regimen shown to reduce hippocampal THP levels by ^~50 % (Smith et al., 2006). Finasteride treatment was used for its action as a 5α-reductase type II inhibitor that prevents the conversion of progesterone to THP. Injections were given daily 1 h prior to restraint. This time was chosen based on results from other studies showing a > 90 % reduction in THP levels after restraint (Mukai et al., 2008). Although this drug also blocks formation of dihydrotestosterone, circulating levels of this androgen are negligible by the onset of puberty (Jean-Faucher et al., 1985).

In another experiment, mice with deletions of the GABAR α4 subunit were used. The background strains for α4 knock-out (KO) mice are C57BL/6J and Strain 129S1/X1. The initial mutation inserted Cre recombinase-activated LoxP (locus of X-over P1) sites flanking exon 3 of the GABRA4 gene (in 129S1/X1 mice) (Chandra et al., 2006). These mice were bred with a Cre expressing mouse strain to delete exon 3 of the GABRA4 gene, and then the Cre was bred out. The α4 gene is transcribed but cannot be translated in these mice because of the frame shift caused by the exon 3 deletion. These mice were back-crossed for 3 generations to C57BL/6J mice, yielding a 99.8 % genetic similarity. The mice continue to be back-crossed to C57BL/6J mice every 5 generations. α4 KO mice were bred on site from α4+/− mice originally supplied by G. Homanics (Univ. of Pittsburgh). Because α4 +/+ mice exhibit similar results in the behavioral protocols as wild-type C57BL/6J mice, results were combined. Genotyping of the tails was used to identify mice that were homozygous α4 KO (Transnetyx, Cordova, TN). α4 KO mice are functional δ knock-outs (Sabaliauskas et al., 2012). They were used rather than δ KO to spare the α1βδ present on interneurons (Glykys et al., 2007).

4.3. Stress protocols

4.3.1. Chronic, predictable restraint stress

Restraint tubes (SUNY Downstate; Kent Scientific Corportation, Torrington, CT) were round slotted Plexiglas cylinders with sliding plugs to allow adjustment of the tube length for each animal’s size. Cylinders measured 6.5 in. long with an internal diameter of 1 in. Animals were placed in the tubes for 2 h/day beginning at the start of their dark cycle, between 11:00 and 13:00 h for 2 weeks (Mukai et al., 2008). All mice were returned to their home cages after restraint. Control mice were brought from the animal facility to the lab space but stayed in a different room from experimental animals. In some cases, mice were weighed on day 1 and day 14 of the restraint stress protocol. A significant weight change different from control is consistent with a poor adaptation to stress (Hyldelund et al., 2022).

4.3.2. Chronic, unpredictable stress (CUS)

Rodents were exposed on a daily basis to a randomized series of mild environmental stressors with a high degree of unpredictability and uncontrollability for a 2-week period (Borrow et al., 2018). Specifically, 2 out of 6 possible stressors were administered each day, with variations in the duration and time of onset of the stressor. All mice experienced the same schedule. Stressors used included: restraint; tilted cage; cold, wet bedding; empty cage; predator scent (fox urine); and cold room. All mice were returned to their cages of 5 mice after restraint. Control mice were brought from the animal facility to the lab space but stayed in a different room from experimental animals. In a second series of experiments, mice were administered 1 session of CUS and 2 h of chronic predictable restraint stress each day for 2 weeks to determine if combining predictable and unpredictable stress would have a different outcome.

4.4. Behavioral protocols

4.4.1. Forced swim test (FST)

Mice were tested the day following the final day of the stress protocol. The FST was conducted by gently placing mice inside individual glass cylinders (24 cm height x 13 cm diameter) filled with water (28 ° C) to a 15 cm height, for 6 min (Can et al., 2012). Then, the mice were removed from the cylinders, dried with paper towels, and placed in heated cages for 30 mins. The FST was videotaped for 6 min, and the latter 4 min were later scored by an experienced observer blinded to the experimental condition. Behaviors scored included (1) Immobility—when the mouse floated in the water without struggling and performed only those necessary movements to keep the head above the water, and (2) swimming—when the mouse performed more active movements, such as moving around the cylinder. A decrease in immobility is considered a measure of active coping, consistent with a decrease in depressive-like behavior.

4.4.2. Forced swim test (FST) with escape

Mice from each group (pubertal female, pubertal male, adult female) were tested for their escape behavior after administration of the FST as described above. At the end of the 6 min FST, a straw was secured on the rim of the cylinder as described in Nishimura et al. (1988). The distance traveled on the straw was recorded for each mouse as was escape, and the restraint and control groups compared. Initial studies verified that mice were able to climb the straw when the FST was not implemented. This escape behavior is an additional measure of active coping.

4.4.3. Elevated plus maze (EPM)

Mice were tested the day following the forced swim test protocol or in some cases, following 1 d of restraint. The EPM consists of two opposed open arms and two opposed enclosed arms, each 8 × 35 cm, which are connected by a central platform and elevated 57 cm above the floor (Smith et al., 2006). The “closed” arms have 33-cm walls, and the “open” arms have no walls, but are partially bordered by small rails (5 × 15 cm) extending to the proximal half of the arm. Each animal was initially acclimated to the behavior room for 1 h before being placed into the center of the plus maze, oriented to the open arm opposite to the experimenter, and were allowed to explore the maze for 5 min. The percent of time spent in the open arms and number of entries to the open arms were recorded using the video-tracking system Ethovision XT (Noldus, Leesburg, VA). Greater time spent in the open arms compared to controls would indicate decreased anxiety-like (i.e., increased risk-taking) behavior.

4.4.4. Temporal order recognition (TOR) task

Mice were tested on this mPFC-dependent task which assesses their ability to recognize the temporal order of object presentation (Barker et al., 2007). Mice were handled for 1 week and then habituated to the test cage (10–15 min/day for 2 d). On the day of testing, mice were placed in the test cage containing 2 similar objects (O1) for 4 min. Mice were removed from the test cage for 1 h, and placed back in the cage containing 2 similar objects (O2) different from the original ones for 4 min. Mice were removed from the test cage for 3 h. When they were placed back in the test cage (for 3 min) it contained 1 O1 (less recent/ “novel”) and 1 O2 object (more recent/ “familiar”). The time spent exploring O1 versus O2 was determined. When temporal order memory is optimal, mice will spend more time exploring O1 (novel object) that was seen less recently compared to O2 (familiar object) that was seen more recently. A discrimination ratio was calculated as a measure of temporal order memory: (time spent-O1 – time spent-O2)/(time spent-O1 + time spent-O2). The total number of approaches was also calculated as a measure of locomotor activity.

4.5. Immunohistochemistry and fluorescent Microscopy

4.5.1. Preparation of brains

Mice were administered urethane (0.1 ml 40 %, i.p.) anesthesia and perfused transcardially using a peristaltic pump, first with 100–300 ml of heparinized saline over a 1-min period followed by 4 % paraformaldehyde in 0.1 M phosphate buffer (PB), set at pH 7.4, over a 3-min period. Brains were post-fixed for 24 h before sectioning at 40 μm with a vibratome (Leica VT1200S). Sections included the PL mPFC and BLA, which were stored at 4 °C in saline (0.9 % NaCl), buffered by 0.01 M phosphate salts (PBS, pH 7.4) and with 0.05 % sodium azide.

4.5.2. Immunohistochemistry

Sections were washed 2x (10 mins each) with 0.01 M PBS supplemented with 0.05 % Triton prior to blocking with buffer (2.5 % donkey serum and 2.5 % goat serum in 0.05 % Triton PBS) for 2 h (Evrard et al., 2021). Sections were then blocked with 2.5 % goat anti-mouse IgG F(ab) Fragment (Jackson Immunoresearch, AB_2338476) in buffer for 2 h. Following blocking, sections were incubated with mouse anti-GABAR α4 (NeuroMab N398A/34, 1:500) and rabbit anti-Map-2 (microtubule associated protein 2) (Abcam ab32454, 1:500), to detect α4 GABA_ARs and Map-2, respectively, diluted in the blocking solution overnight at 4°C. Selectivity of the anti-GABARα4 has been demonstrated previously (Evrard et al., 2021). After washing 3x (20 mins. each), sections were incubated overnight at 4° C with fluorescent secondary antibodies goat anti-mouse (Invitrogen, Alexafluor 647, 1:500) and donkey anti-rabbit (Jackson Immunoresearch, Alexafluor 488, 1:500). Sections were washed again 3x (20 mins. each), before mounting on superfrost slides (Fisher Scientific) with Prolong antifade glass (Invitrogen).

Prior to experiments, cross-reactivity between primary antibodies was analyzed with an EMBOSS sequence alignment (Madeira et al., 2022). Scores under 80 % are not considered cross-reactive, and the two antibodies used in this protocol had an alignment score of 70.5 % (data not shown). Since the GABAR α4 antibody was raised in mice, species cross-reactivity was minimized by blocking with goat anti-mouse IgG F (ab) fragment. F(ab) antibody fragments are used to block endogenous immunoglobulins on cells and can minimize potential cross-reactivity between antibodies and cell-surface receptors.

4.5.3. Fluorescent microscopy

Images were taken with a View TM FV1000 confocal Olympus Fluo inverted microscope with objective UPLSAPO 60x NA:1:35 (Olympus, Tokyo, Japan) after adjusting the laser intensity to minimize background (Evrard et al., 2021). Z-stack projection photomicrographs of ten sections, roughly 3 μm each were taken. Pyramidal cell bodies from merged images of the z-stacks (excluding the upper and lower 10 μm to eliminate slices with artificially enhanced luminosity) were analyzed for luminosity at 488 λ (reflecting Map-2 expression) and 647 λ (reflecting α4 expression) using the region of interest (ROI) program of Fiji (Image J) software (NIH). Each merged image was between 10 and 15 μm thick.

4.5.4. Neuron selection and analysis

Map-2 is a structural protein that helps maintain neuroarchitecture (Johnson and Jope, 1992). As such, it was used to identify and select pyramidal neurons based on their size and shape. Two brain regions, layer 5 PL and BLA, were analyzed at 3 time points during the 2-week restraint protocol: after 2, 7, and 14 d (control and restraint) or after 2 d (vehicle and finasteride). Mice were perfused 24 h after the last restraint session. 5–6 mice in each condition were analyzed in each brain region and at each time point. 3–8 neurons were analyzed per mouse.

4.6. Golgi staining and spine density Measurements

Whole brains from euthanized animals were processed for Golgi impregnation using the FD Rapid Golgi Stain kit (FD Neurotechnologies, Columbia, MD) (Afroz et al., 2016). Coronal sections were prepared using a vibratome (Leica VT1200s) set to a thickness of 100 μm. Pyramidal cells from L5 PL and BLA were identified using The Mouse Brain in Stereotaxic Coordinates (Paxinos and Franklin, 2001) and the Allen Brain Institute’s Mouse Brain Atlas (http://mouse.brainmap.org). The layer 5 PL neurons were approximately 1.7 mm ventral from the dorsal surface and the cell bodies were 500–700 μm from the medial surface. BLA pyramidal cells were 1.1 mm ventral from the dorsal surface and 3.2 mm from the medial surface. The neurons were viewed with a 100x oil objective on an Olympus BX51 upright light microscope. Z-stack projection photomicrographs (0.1 μm steps) were taken with a Nikon DS-U3 camera mounted on a Nikon Eclipse Ci-L microscope using a 100x oil objective and analyzed with NIS-Elements D 4.40.00 software. Each dendrite segment (middle 80 %) was ~ 1 μm thick, 20–50 μm in length and was taken from a 2° or 3° order dendrite. Spines were typed and counted in 2 limbic brain regions (layer 5 PL and BLA). Briefly, stubby spines had a length to width ratio of ~ 1, mushroom spines were identified by a > .35 mm head width, with a head dia: neck > 2, while thin spines were classified if the head dia: neck < 1.2 and a length: width > 3 (Arellano et al., 2007; Evrard et al., 2021). All spine density and morphology assessments were made with the investigator blinded to the condition of the animals tested. Five mice in each group and brain region were analyzed, with 4–9 dendrites per animal.

4.7. Statistics

All statistical analyses were performed with Origin (OriginLab, Northampton, MA) or GraphPad Prism (GraphPad Software Inc., San Diego, CA, USA). All data are represented as the mean ± standard error of the mean (SEM). Initially, the Kolmogorov–Smirnov test was used to identify data that followed a normal distribution. In these cases, parametric statistical tests were used. The statistical significance of sex differences in behavior (except for the escape studies) was analyzed with a 2-way ANOVA and post-hoc Tukey tests (Origin). Significant differences between other behavioral groups were analyzed with student t-tests (two groups) or 1-way ANOVAs with post-hoc Tukey tests for multiple comparisons (more than two groups) (Origin). Significant differences in IHC and dendritic spine data were analyzed with nested t-tests (GraphPad Prism).

Data from the escape studies did not follow a normal distribution and thus, non-parametric statistical tests were used. The Mann Whitney U test was used to compare climbing distance for restrained versus control mice in the FST with escape test. The binary Fisher’s exact test was used to compare escape frequency for restrained versus control mice. Linear regression was used to establish a correlation between distance climbed and % time immobile using the Pearson’s r. For all statistical tests, the level of significance was determined to be P < 0.05.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.brainres.2025.149874.

Glossary

BLA

Basolateral amygdala

Cre

Cre recombinase

CRH

Corticotropin releasing hormone

CUS

Chronic unpredictable stress

EPM

Elevated plus maze

FIN

Finasteride

FST

Forced swim test

GABAR

GABA_A receptor

HPA

Hypothalamic-pituitary-adrenal

KO

Knock-out

PFC

Prefrontal cortex

PL

Prelimbic

THP

3α-OH,5α-pregnan-20-one

THP

Allopregnanolone

TOR

Temporal order recognition

Data availability

Data will be made available on request.