Witnessed trauma exposure induces fear in mice through a reduction in endogenous neurosteroid synthesis

Abstract

Neurosteroids have been implicated in the pathophysiology of post-traumatic stress disorder (PTSD). Allopregnanolone is reduced in subsets of individuals with PTSD and has been explored as a novel treatment strategy. Both direct trauma exposure and witnessed trauma are risk factors for PTSD; however, the role of neurosteroids in the behavioral outcomes of these unique experiences has not been explored. Here, we investigate whether observational fear is associated with a reduced capacity for endogenous neurosteroidogenesis and the relationship with behavioral outcomes. We demonstrated that mice directly subjected to a threat (foot shocks) and those witnessing the threat have decreased plasma levels of allopregnanolone. The expression of a key enzyme involved in endogenous neurosteroid synthesis, 5α-reductase type 2, is decreased in the basolateral amygdala, which is a major emotional processing hub implicated in PTSD. We demonstrated that genetic knockdown or pharmacological inhibition of 5α-reductase type 2 exaggerates the behavioral expression of fear in response to witnessed trauma, whereas oral treatment with an exogenous, synthetic neuroactive steroid gamma-aminobutyric acid-A receptor positive allosteric modulator with molecular pharmacology similar to allopregnanolone (SGE-516 [tool compound]) decreased the behavioral response to observational fear. These data implicate impaired endogenous neurosteroidogenesis in the pathophysiology of threat exposure, both direct and witnessed. Further, these data suggest that treatment with exogenous 5α-reduced neurosteroids or targeting endogenous neurosteroidogenesis may be beneficial for the treatment of individuals with PTSD, whether resulting from direct or witnessed trauma.

1 |. INTRODUCTION

Previous trauma exposure is a major risk factor for psychiatric illnesses, including depression, anxiety, and post-traumatic stress disorder (PTSD).¹ PTSD is characterized by repeated intrusive memories of the trauma, dissociative episodes, and heightened negative reactions to stimuli associated with the trauma, among other symptoms.² Those diagnosed with PTSD are also often afflicted by negative changes in mood and an increase in generalized anxiety. However, not everyone who is exposed to a traumatic event will go on to develop PTSD; current estimates find that 25%–30% of those who experience a trauma develop PTSD, putting its prevalence in the population at 7.8%.^1,3–6 The factors contributing to the individual differences in the vulnerability to developing PTSD are not yet fully understood.

Both direct trauma exposure and witnessed trauma exposure have been shown to increase risk for PTSD, and while the incidence of PTSD following witnessed trauma is lower, the overall prevalence is larger making it a significant societal issue.¹ Various animal models, such as observational fear (OF), have been developed to study the mechanisms through which witnessed trauma increases the risk for psychiatric illnesses.^7,8 OF learning has been employed in rodents and involves pairing a vicarious unconditioned stimulus through observation of another conspecific receiving foot shocks within a specific context. This paradigm has been suggested to model fear transference and affective empathy.⁹ These studies have shown that witnessing trauma in the absence of physical trauma is sufficient to produce long-term behavioral and biochemical changes associated with depression and anxiety in rats and mice.^10,11 Here, we employ this paradigm to examine the mechanisms contributing to negative behavioral outcomes following OF.

Neuroactive steroids (NASs), which act as positive allosteric modulators (PAMs) of gamma-aminobutyric acid-A receptors (GABA_ARs), have been implicated in the pathophysiology of PTSD in both clinical and preclinical studies.^12,13 Accumulating evidence suggests that altered levels of allopregnanolone may contribute to the pathophysiology of PTSD in at least some subpopulations.^12,14,15 For example, allopregnanolone levels are reduced in the cerebrospinal fluid of women with PTSD, which is inversely correlated with PTSD symptoms and negative mood and is most prominent in individuals with comorbid major depressive disorder.¹⁶ NASs have also been suggested to have therapeutic potential for individuals with PTSD. Unfortunately, treatment with ganaxolone, an allopregnanolone analog, failed to perform better than placebo in a phase 2, proof-of-concept clinical trial. This may have been attributable to the lower than anticipated therapeutic levels, suggesting potential underdosing of ganaxolone in this study.¹⁷ Despite this setback, interest remains in the pathophysiological role of neurosteroids in PTSD.

Endogenous allopregnanolone can be synthesized de novo from cholesterol or through the metabolism of the precursor progesterone by the sequential actions of two enzymes, 5α-reductase and 3α-hydroxysteroid dehydrogenase (for review, see Diviccaro et al.¹⁸). Two isoforms of 5α-reductase, type 1 and type 2, are the rate-limiting enzymes involved in neurosteroidogenesis.¹⁹ However, few studies have investigated the potential pathophysiological role of deficits in endogenous neurosteroidogenesis in mood disorders. Inhibition of 5α-reductase with finasteride is associated with long-term deficits in mood, including anxiety and depression, termed postfinasteride syndrome, which is thought to involve deficits in endogenous neurosteroidogenesis.²⁰ Further, a polymorphism in the gene encoding for 5α-reductase type 2, SRD5A2, has been associated with an increased risk for PTSD in males.²¹ These correlational studies implicate deficits in endogenous neurosteroids in the pathophysiology of PTSD; however, additional studies are required to directly link deficits in endogenous neurosteroidogenesis and PTSD.

Our group has recently demonstrated that impaired neurosteroid synthesis in the basolateral amygdala (BLA) is a key driver in facilitating behavioral responses to stress. We demonstrate that reducing the capacity for endogenous neurosteroidogenesis specifically in the BLA is sufficient to mimic the behavioral deficits associated with chronic stress, whereas enhancing endogenous neurosteroidogenesis in the BLA improves behavioral outcomes following chronic stress.²² Further, infusion of allopregnanolone directly into the BLA decreases avoidance behaviors and stress-induced helplessness.²³ These data pinpoint neurosteroidogenesis in the BLA as a potential hub for behavioral changes associated with stress. Here, we investigate whether OF is associated with a reduced capacity for endogenous neurosteroidogenesis in the BLA and its relationship with behavioral outcomes.

2 |. MATERIALS AND METHODS

2.1 |. Animals

Adult (12 weeks old) male C57BL/6J mice (Jackson Labs 000664) were used for these studies. Males were selected because female mouse behavior has not been as well validated in this OF paradigm and due to the evidence that a polymorphism in the gene encoding for 5α-reductase type 2, SRD5A2, has been associated with an increased risk for PTSD in males.²¹ Animals were housed on a 12:12 h light/dark cycle at Tufts University School of Medicine in a temperature and humidity-controlled environment with ad libitum access to food and water. All experiments were performed between 10:00 and 16:00. Observer, naïve, and demonstrator mice were housed with their respective experimental groups throughout the duration of the experiment. All animal procedures were handled according to the protocols approved by the Tufts University Institutional Animal Care and Use Committee (IACUC). Floxed Srd5a2 mice were generated by GenOway (Figure 2).

These mice were genotyped using the following primers (Table 1):

TABLE 1.

List of primers used for genotyping.

Gene	Forward primer (5’ –3’)	Reverse primer (5’ –3’)
Srd5a2	AACTTGTAAATCTTTGCCACTCTGCGG	AGCAGGAGAAATGTAGGTGAGCAGGG

2.2 |. Observational fear paradigm

The OF paradigm has been widely used as an animal model of fear transference as well as empathy.^4,8,9,24 Our model was adapted from the studies by Jeon and Shin²⁵ and Davis et al.²⁶ and involves three consecutive days of shock sessions in which the observer mouse witnesses a demonstrator mouse undergoing a series of foot shocks. The animals are separated by a clear plastic barrier, and thus, the observer can receive visual, auditory, and olfactory cues from the demonstrator throughout the threat exposure. Animals were randomly assigned to be either naïve, observers, or demonstrators. Naïve mice were handled by the experimenters similarly to the observers and demonstrators but not exposed to threat exposure (either witnessed or direct) and were placed in the shock chamber only to record control freezing data. The OF paradigm used, described in Figure 1A, consists of three shock sessions on days 1, 2, and 3, a recall session on day 4 (24 h after the final shock session), and a recall session on day 7 (96 h after the final shock session). For all shock sessions, the observer and demonstrator mice were placed on separate sides of a fear-conditioning box separated by a clear plastic barrier. Following a 30-s habituation period, 2 s foot shocks of 0.7 mA were delivered every 28 s to the demonstrator mouse only, lasting for a total of 6 min. This was repeated with each observer-demonstrator pair once per day over three consecutive days. During recall timepoints, each observer mouse was placed back into the same half of the fear-conditioning box without the demonstrator present. For all shock and recall sessions, video recordings of the trials were analyzed using EthoVision XT software. Freezing was quantified using the movement parameter in EthoVision. The software calculated cumulative time moving and not moving, the latter of which was used as the total time freezing. To account for individual differences between mice, the recall session freezing value for each animal was divided by its freezing during the initial shock sessions, and this number is presented as the normalized freezing value.

FIGURE 1.

Observational threat exposure produces a contextual fear response and impairs endogenous neurosteroid signaling. (A) Diagram and schedule of the observational fear paradigm. (B) Normalized freezing times of naïve (N), observer (O) at the 24- and 96-h timepoints, and demonstrator (D) mice at the 24-h time point. n = 6–9 mice per experimental group. (F (3,27) = 37.47, R2 = 0.8063) using a multiway Analysis of Variance (ANOVA) followed by Tukey’s post hoc test. (C) Serum levels of allopregnanolone in observer (O), demonstrator (D) and naïve (N) mice after observational fear exposure. n = 4–13 mice per experimental group. (F (2,23) = 3.555, R2 = 0.4156) using a multiway ANOVA followed by Tukey’s post hoc test. (D) Transcript levels of Srd5a2 in the BLA in observer (O) and demonstrator (D) mice after undergoing observational fear. Fold change difference were calculated using (F (3, 23) = 0.7399, R2 = 0.5147) using a multiway ANOVA followed by Tukey’s post hoc test. (E) Transcript levels of Gabrd in the BLA in observer (O) and demonstrator (D) mice after undergoing observational fear. (F (3, 22) = 4.695, R2 = 0.5351) using a multiway ANOVA followed by Tukey’s post hoc test. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 using a one-sample t-test.

2.3 |. Drug administration

Finasteride (Sigma-Aldrich F1293) was dissolved in vehicle (15% [2-hydroxypropyl]-β-cyclodextrin [Sigma-Aldrich H107] in ddH2O) to make a 5 mg/mL solution. Mice were then dosed intraperitoneally at 5 mg/kg according to body weight 24 h prior to the first shock session and the 24-h recall session. SGE-516 (450 mg/kg/chow) was administered chronically via chow beginning 1 week prior to the first shock session and throughout the duration of the OF protocol. This dose and route of administration of SGE-516 have been previously demonstrated to achieve 72.5 ± 17.4 ng/mL of allopregnanolone in the plasma and 86.6 ± 21.1 ng/g in the brain.²⁷

2.4 |. qRT-PCR

After the final OF recall time point, mice were anesthetized using isoflurane and sacrificed by decapitation with a guillotine. Brains were then extracted immediately, flash-frozen in liquid nitrogen, and stored at −80°C until sectioning. Tissue punches of the BLA were collected with a 0.5 mm sterile biopsy needle on a cryostat at −20°C according to anterior–posterior BLA coordinates from the Allen Mouse Brain Atlas. DNA was extracted and RNA samples were prepared using the SuperScript^™ III Platinum^™ SYBR^™ Green One-Step qRT-PCR Kit (Invitrogen 11736059) as previously described by our laboratory.^22,28 RNA integrity and purity were determined by measuring the ratio of absorbance at 260 and 280 nm (Table S1). Two replicates per animal were measured using the primers in Table 2. Samples were run using the Mx3000P qPCR system and analyzed with the MxPro software. All samples were run in duplicate with an average CT value normalized to β-actin and relative transcript levels were calculated in accordance with the 2^−ΔΔCT method,²² with the relative expression in naive mice set to 1. This is also now clarified in the text. Representative melt curves are shown in Figure S1. Fold-change from naïve mice was then calculated to determine the changes in RNA transcript levels following OF in both observers and demonstrators.

TABLE 2.

List of IDT custom primers used for qRT-PCR.

Gene	Forward primer (5’ –3’)	Reverse primer (5’–3’)
Srd5a1	GAGATATTCAGCTGAGACCC	TTAGTATGTGGGCAGCTTGG
Srd5a2	ATTTGTGTGGCAGAGAGAGG	TTGATTGACTGCCTGGATGG
Gabrd (GABAA-δ)	AGGAACGCCATCGTCCTTTT	CTTGACGACGGGAGATAGCC
β-actin	GGCTGTATTCCCCTCCATCG	CCAGTTGGTAACAATGCCATGTC

Note: Quantified mRNA levels of SRD5A1, Srd5a2, and Gabrd were normalized to measured levels of β-actin.

2.5 |. Western blot

Western blots were performed as previously described.²² Briefly, mice were anesthetized with isoflurane, decapitated using a guillotine, and brains rapidly extracted into ice-cold homogenization buffer (150 mM NaCl, 10 mM Tris, 0.1% SDS, 1% sodium deoxycholate, 0.5% TritonX-100, pH 7.4) with mini cOmplete protease inhibitors and PhosSTOP phosphatase inhibitors(Roche). Samples were briefly sonicated before being placed on a heat block at 100C for 30-min. Samples were then centrifuged at 13.2×g for 20-min at 4°C. The supernatant was collected into fresh tubes before protein quantification using a DC Assay.

A total of 25 μg of each sample was prepared with 2× Laemmli sample buffer and separated on a 12% SDS-polyacrylamide gel. Next, proteins were transferred to a nitrocellulose membrane, blocked overnight in 10% nonfat milk, and subsequently probed with goat polyclonal SRD5A2 (1:1000, Abcam, ab2746) for 2-h at room temperature. Blots were then incubated in donkey anti-goat HRP (1:2000, 154960) for 2-h at room temperature. Blots were developed and imaged using Pierce ECL (ThermoFisher Scientific, P132106) on a ChemiDoc MP(Bio-Rad). Optical density measurements were calculated with Image J Software (NIH) and normalized to β-tubulin levels (1:10,000, Sigma T8328).

2.6 |. Allopregnanolone measurements

For allopregnanolone measurements, immediately after OF exposure at the 96-h time point, mice were decapitated using a guillotine and trunk blood was collected into ethylenediaminetetraacetic acid-coated microcentrifuge tubes on ice. Samples were centrifuged at 1500×g for 15-min and the supernatant containing serum was then transferred into a fresh microcentrifuge tube and stored at −80°C until further processing. Allopregnanolone levels were measured by liquid chromatography–tandem mass spectrometry by PharmaCadence as previously described.²² Allopregnanolone levels in the range of 0.2–200 ng/mL were measured against reference material purchased (Tocris Cat No. 3653) and a penta-deuterated stable isotope-labeled internal standard (Tocris Cat No. 5532). Samples were prepared using liquid–liquid extraction with chlorobutane and diluted 1:1 with deionized water prior to extraction with 1 mL of chlorobutane. The organic layer was extracted, dried under a stream of nitrogen, reconstituted in a 1:1 mixture of deionized water and acetonitrile, and detection was accomplished using a Sciex 5500 Qtrap mass spectrometer operated in positive ion mode. Concentrations were calculated using the integrated peak area and the response factor determined by 1/concentration² weighted, least squares regression of the calibration curve.

2.7 |. Immunohistochemistry

Immunohistochemistry was performed as previously described.²² In brief, mice were deeply anesthetized with isoflurane sacrificed using a guillotine. Brains were rapidly extracted directly into 4% paraformaldehyde and fixed overnight at 4°C before subsequently being cryo-protected in successive overnight saturation in 10%–30% sucrose. Brains were then frozen in chilled isopentane, and 40-micron sections were collected on a cryostat. Free-floating brain sections were blocked in 10% normal goat serum in 0.3% Triton-phosphate buffered saline for 2-h then incubated in rabbit anti-RFP (1:250 Rockland Cat# 600–401-379) overnight. Sections were then incubated in goat anti-rabbit 647(1:200 Invitrogen Cat# A32733) for 2-h. All sections were then mounted on microscope slides and immunofluorescence was imaged on a Leica SP8 confocal microscope.

2.8 |. Statistical tests

Statistical analysis was performed using GraphPad Prism 9 software. A Kolmogorov–Smirnov test for normality was performed, and the data are normally distributed (Figure S2). Unpaired Student’s t-tests were used to compare two experimental groups and one-way Analysis of Variance (ANOVAs) were used to compare more than two experimental groups. Tukey’s post hoc test was employed to analyze individual group differences. One-sample t-tests were used to analyze fold change values. Values in this paper are written as mean ± SEM. p-values of <.05 were considered statistically significant and denoted as p ≥ .05 = n.s., p < .05 = *, p < .01 = **, p < .001 = ***, and p < .0001 = ****. A comprehensive list of statistical tests and outcomes for all figures is provided in Table S2.

3 |. RESULTS

3.1 |. Observational threat exposure produces a contextual fear response

The OF paradigm consists of three consecutive days of shock sessions in which observer mice witness a demonstrator mouse undergo a series of foot shocks. Mice are separated by a clear plastic barrier, receiving visual, auditory, and olfactory cues from the demonstrator throughout the paradigm (Figure 1A). In accordance with previously published data,⁸ we found that observer mice show significantly increased freezing behavior (0.47 ± 0.03 normalized to shock session baselines) when returned to the chamber for contextual recall as compared to naïve mice (0.15 ± 0.02) (p < .0001, Tukey’s test, Figure 1B). Interestingly, the behavioral expression of fear following witnessed trauma is nearly the same magnitude as the animals that directly experienced the foot shocks (demonstrator: 0.65 ± 0.04 normalized to shock session baselines) (p = .0035, Tukey’s test, Figure 1B). Thus, this model is valuable for examining the neurobiological processes contributing to the pathophysiological consequences of witnessed trauma exposure.

3.2 |. Observational threat impairs endogenous neurosteroid signaling

To investigate the hypothesis that altered endogenous neurosteroids may contribute to the behavioral expression of OF, we measured the levels of endogenous allopregnanolone in plasma from mice subjected to OF compared with naïve mice and demonstrators. Plasma levels of allopregnanolone are reduced in both observer (1.138 ± 0.113 ng/mL) and demonstrator (1.044 ± 0.157 ng/mL) mice compared with naïve controls (2.035 ± 0.205 ng/mL) (observers: p = .042, demonstrators: p = .0028, one-way ANOVA, Figure 1C). To determine whether endogenous neurosteroid levels in the brain may also be impaired, we measured the expression levels of two key neurosteroidogenic enzymes: 5α-reductase type 1 and type 2. We found a significant decrease in the transcript expression of Srd5a2 in the BLA of observer mice (0.393 ± 0.099-fold change from naïve) (p = .0005, one sample t-test) compared with naïve (dotted line normalized to 1) (Figure 1D). Interestingly, transcript levels of Srd5a2 in the BLA were not significantly different between observer (0.393 ± 0.099-fold change) and demonstrator mice (0.174 ± 0.049) (p = .209, one-way ANOVA). However, we found no significant change in the levels of Srd5a1 transcript in the BLA in observer (0.709 ± 0.297-fold change from naïve) or demonstrator (0.843 ± 0.166-fold change) mice (p = .911, one way ANOVA and one sample t-test, Figure S3) compared to naïve. We found that transcript levels of the GABA-δ receptor (Gabrd), the pre-dominant site of action for allopregnanolone, were reduced in the BLA of observer (0.288 ± 0.047-fold change from naïve) as well as demonstrator (0.200 ± 0.022) mice compared with naïve (dotted line normalized to 1) (p < .0001, one sample t-test, Figure 1E). These data suggest that endogenous neurosteroid synthesis and signaling are compromised following OF exposure.

3.3 |. Impaired endogenous neurosteroidogenesis alters the contextual observational fear response

To directly investigate the role of endogenous neurosteroid signaling on the behavioral expression of OF, we generated a novel mouse model, which enables us to manipulate 5α-reductase type 2 expression (Figure 2A). Based on the impairment of Srd5a2 expression following OF exposure, we crossed floxed Srd5a2 mice generated by genOway with CaMKII-Cre mice to knockout this key neurosteroidogenic enzyme in excitatory neurons (Figure 2B–D), which was sufficient to reduce expression of the reporter (Figure 2D) and Srd5a2 protein levels in the BLA (0.343 ± 0.025 normalized to β-tubulin) compared with controls (Cre-) (0.508 ± 0.043 normalized to β-tubulin) (Figure 2E; p = .090, ANOVA). Mice lacking Srd5a2 in CaMKII neurons (Srd5a2/CaMKII) exhibited increased behavioral expression of observed fear, increasing freezing behavior during recall (1.121 ± 0.146 normalized to shock session baselines) compared with controls (0.590 ± 0.076) (Figure 2F; p = .017, Welch’s t-test).

FIGURE 2.

Impaired endogenous neurosteroidogenesis alters the contextual observational fear response. (A) Illustration of CRISPR Cas9-mediated homologous recombination used to introduce a tdTomato reporter and lox p sites between Exon 3 and Exon 5 in the floxed Srd5a2 (fl5a2) mouse line. (B) Schematic of the generation of the fl5a2 × CaMKII-Cre mouse strain. (C) Representative image of Srd5a2 (RFP) expression in the BLA of the fl5a2 knock-in mouse. (D) Representative image of CaMKII-Cre (GFP) and 5a2 (RFP) expression in the BLA of the fl5a2 × CaMKII-Cre mouse. (E) Protein expression of 5α-reductase type 2 in the BLA of heterozygous floxed (floxed) mice compared to wild type (Cre-) controls (Ctrl). n = 3 animals per group and *p < 0.05 using a one-sample t-test. (F) Floxed Srd5a2 × CamKII mice exhibit an increase in freezing behavior during 24 h recall following observational fear compared to controls. n = 5–13 animals per group and *p < 0.05 using unpaired t-test.

Finasteride is a pharmacological inhibitor of 5α-reductase type 2, resulting in a reduction in the synthesis of allopregnanolone.²⁰ The impact of finasteride on the behavioral expression of OF was measured to further explore the role of endogenous neurosteroid synthesis in the behavioral sequelae of witnessed trauma. Mice treated with finasteride (5 mg/kg, Figure 3A)²⁹ exhibited an increase in the behavioral expression of OF, increasing their freezing behavior (0.646 ± 0.046 normalized to shock session baselines) compared with vehicle-treated mice (0.439 ± 0.027) (Figure 3B; p = .002, Welch’s t-test). Thus, both genetic and pharmacological inhibition of 5α-reductase is sufficient to increase the behavioral expression of OF, suggesting that the observed reduction in 5α-reductase may contribute to the fear response following OF exposure.

FIGURE 3.

Impaired or enhanced endogenous neurosteroid activity alters the contextual observational fear response. (A) Diagram of the experimental paradigm with schedule of finasteride treatment relevant to observational fear exposure. (B) Mice treated with finasteride also exhibit an increase in freezing behavior during 24 h recall compared to controls treated with vehicle injections following observational fear. n = 8–9 mice per experimental group. (C) Schematic of the time course of SGE-516 administration. (D) The average freezing behavior of observer mice dosed with vehicle during the 24- and 96-h recall sessions in vehicle-treated observer mice (O 24 and O 96) and observers treated with SGE-516 (O 24 SGE and O 96 SGE). n = 14 mice per experimental group. (F (2,8) = 0.3580, R2 = 0.02375) using a multiway ANOVA followed by Tukey’s post hoc test. *p < 0.05 and **p < 0.01.

3.4 |. Exogenous neurosteroid treatment decreases the observational fear response

Given that our data indicates that OF is correlated with decreased allopregnanolone synthesis and/or signaling, it raises the question of whether treatment with a GABA_AR PAM may reduce the behavioral expression of OF. To test this hypothesis, we treated mice with chow containing SGE-516 (450 mg/kg/chow, developed by SAGE Therapeutics) for 1 week prior to shock sessions (Figure 3C). Mice treated with SGE-516 exhibited decreased freezing at both recall timepoints (0.501 ± 0.049 at 24 h; 0.462 ± 0.047 at 96 h) compared with controls (0.749 ± 0.075 at 24 h; 0.745 ± 0.073 at 96 h) (Figure 3D; p = .010 at 24 h; p = .0031 at 96 h; unpaired t-tests). These data suggest that NAS GABA_AR PAMs, such as SGE-516, may have therapeutic utility for witnessed trauma exposure.

4 |. DISCUSSION

4.1 |. Trauma exposure-induced impaired neurosteroid synthesis enhances the behavioral expression of fear

Deficits in allopregnanolone levels have been associated with PTSD.¹⁶ However, these studies were largely correlational and the direct role of endogenous neurosteroids in the pathophysiology of PTSD remains unclear. Here, we demonstrate that both witnessed and direct trauma exposure reduces endogenous allopregnanolone levels in mice. These deficits in endogenous allopregnanolone levels are associated with a reduction in the expression of 5α-reductase type 2, a key neurosteroidogenic enzyme, in the BLA. While these changes are still correlational, these alterations are consistent with those observed in individuals with PTSD.^13,14 To directly interrogate the role of endogenous neurosteroids on the behavioral sequelae of witnessed trauma, we assessed OF responses in mice with a knockdown of 5α-reductase type 2. Mice with a reduced capacity for endogenous neurosteroid synthesis, either genetic or pharmacological, exhibited an exaggerated behavioral expression of fear following OF exposure. These data suggest that deficits in endogenous neurosteroids may directly contribute to the behavioral deficits associated with witnessed trauma. Additionally, we demonstrate that treatment with a synthetic NAS GABA_AR PAM is sufficient to reduce the behavioral expression of fear, further implicating neurosteroid deficits in the pathophysiology of witnessed trauma. The ability of NAS GABA_AR PAMs to improve behavioral outcomes following OF may be due to the ability to modulate BLA network states,²³ which have been demonstrated to govern the behavioral expression of fear.^25,30 Taken together, these findings indicate that the OF response is at least partially driven by inhibition of neurosteroidogenesis in the BLA, and that targeting this mechanism may have therapeutic benefit.

4.2 |. Chronic stress and threat exposure impair endogenous neurosteroidogenesis

The impact of stress on neurosteroids is related to the magnitude and the duration of the stressor. Acute stress has been shown to increase circulating levels of allopregnanolone,^31–33 which involves at least in part endogenous neurosteroid synthesis in the brain.³³ Although few studies have explored the impact of chronic stress on endogenous neurosteroid levels, it is thought that chronic stress and deficits in endogenous neurosteroids contribute to psychiatric illnesses, such as anxiety and depression (for review, see Almeida et al.¹² and Zorumski et al.³⁴). Protracted social isolation^35,36 and chronic unpredictable stress²² have both been shown to decrease allopregnanolone levels, demonstrating that chronic stress alters endogenous neurosteroid levels. Chronic stress-induced impairment in endogenous neurosteroidogenesis is also supported by the evidence that social isolation and chronic unpredictable stress decrease the expression of 5α-reductase in the brain.^22,35,36 Impaired endogenous neurosteroid synthesis following chronic stress has been implicated in stress-related behavioral impairments.²² Consistent with the current findings, the reduction in 5α-reductase expression following social isolation has been associated with an increase in the behavioral expression of fear.³⁷

Previous work from our laboratory has also demonstrated that chronic stress decreases the expression of the GABA_AR δ subunit, the primary site of action for neurosteroid modulation, in the BLA.²² Similarly, chronic stress has been shown to impair tonic GABAergic inhibition in principal neurons in the BLA.³⁸ The impact of stress on GABAergic signaling is also related to the extent and magnitude of the stress exposure and is brain region-specific (for review, see Maguire,³⁹ Maguire and Mody,⁴⁰ and Mody and Maguire⁴¹). Related to neurosteroid signaling, acute stress and chronic stress have been shown to increase GABA_AR δ subunit and tonic GABAergic inhibition in the hippocampus^42–46 in contrast to the observed decrease in the BLA.²² Knockdown of the GABA_AR δ subunit in the BLA is sufficient to recapitulate the behavioral deficits associated with chronic stress.²² Related to the current study, mice that lack the δ subunit of the GABA_AR exhibit an enhancement of trace fear conditioning⁴⁷ and allopregnanolone decreases the behavioral expression of contextual fear.⁴⁸ These data are consistent with our current findings demonstrating a decrease in 5α-reductase expression following OF and that treatment with a NAS GABA_AR PAM decreases the behavioral expression of witnessed fear.

4.3 |. Endogenous neurosteroids and affective tone

Finasteride is a selective inhibitor of 5α-reductase and was approved for the treatment of lower urinary tract symptoms associated with benign prostatic hyperplasia in 1992 and androgenetic alopecia in early 2000 (for review, see Traish et al.⁴⁹). The number of annual prescriptions of finasteride in the United States has exceeded over 2.4 million in 2020. This is alarming given the well-documented and serious side effects of finasteride, including sexual dysfunction, depression, diabetes, high-grade prostate cancer and vascular disease (for review, see Diviccaro et al.²⁰ and Traish et al.⁴⁹). The affective side effects of these compounds are so severe that there was a request to the FDA to pull the drug from the market, which was rejected but now requires a disclosure about the increased suicide risk.

Reports of depression and anxiety are the most common psychiatric complications associated with the use of finasteride (for review, see Alhetheli et al.,⁵⁰ Hirshburg et al.,⁵¹ and Traish⁵²). Several studies have demonstrated an association between finasteride use and the incidence of depression^53–56 (for review, see Hirshburg et al.⁵¹). Alarmingly, former finasteride users had a marked increase in suicidal thoughts (44%) compared to nonusers (3%),⁵⁴ which may be particularly relevant for a subset of patients with a history of mood disorders.⁵⁷ Fewer studies have explored the relationship between anxiety and finasteride usage, but self-reports indicate a comparable number of patients experiencing anxiety (73%) as depression (74%).⁵⁸ Preclinical studies have demonstrated that treating rodents with finasteride increases depressive and anxiety phenotypes.^59,60 Despite the patient outcry,⁵⁸ accumulating evidence (for review, see Alhetheli et al.,⁵⁰ Hirshburg et al.,⁵¹ and Traish⁵²), and high potential cost to benefit of the use of these compounds, there remain skeptics and advocates for their continued use⁶¹ when in fact this is a significant issue for patients and a “surmountable challenge for clinicians”.⁵²

Consistent with these clinical findings, preclinical studies demonstrate that finasteride treatment alters valence processing. Finasteride treatment increases negative valence processing, increasing stress-induced helplessness in the forced swim test and increasing avoidance behaviors in the open field and elevated plus maze.^62–64 Finasteride treatment also impairs positive valence processing, inducing anhedonia in rats.⁶³ Like the effects of pharmacological inhibition of 5α-reductase, genetic knockdown of 5α-reductase in the BLA is sufficient to increase negative valence processing, increasing stress-induced helplessness and increasing avoidance behaviors.²² Conversely, overexpression of 5α-reductase is sufficient to prevent the behavioral deficits, including increased negative valence processing, associated with chronic stress exposure.²² These data provide empirical evidence supporting the clinical associations between 5α-reductase inhibition and psychiatric complications and are consistent with our current findings demonstrating that impaired endogenous neurosteroid synthesis (either genetic or pharmacological) increases negative valence processing. Further, it demonstrates a role for endogenous neurosteroid signaling in setting a baseline affective tone.

4.4 |. Therapeutic potential of harnessing endogenous neurosteroidogenesis

The current study implicates impaired endogenous neurosteroid synthesis in the pathophysiology of witnessed trauma. Previous studies also demonstrated a reduced capacity for endogenous neurosteroid synthesis in the behavioral deficits associated with chronic stress.²² Interestingly, these studies also demonstrated that increasing endogenous neurosteroidogenesis was capable of preventing the behavioral deficits following chronic stress,²² suggesting that enhancing endogenous neurosteroidogenesis may have therapeutic potential. Treatment with exogenous allopregnanolone has demonstrated clinical antidepressant effects and in the current study is sufficient to reduce the behavioral expression of fear following witnessed trauma. Rather than administering exogenous allopregnanolone, a superior therapeutic approach may be to target 5α-reductase, implicated in the underlying neurobiology, to enhance endogenous neurosteroid synthesis. In addition, 5α-reductase may serve as a useful biomarker for identifying those at risk for developing PTSD as well as those that may be amenable to neurosteroid-based treatments. Further studies are required to explore the translational potential of targeting 5α-reductase for treatment.

DATA AVAILABILITY STATEMENT

Raw data used in this study are available from the lead contact upon reasonable request. All custom Python scripts for analysis and visualization are available from the lead contact upon reasonable request.