Steroid 5α-Reductases in the Brain: From Neurosteroidogenesis to Therapeutic Implications

Abstract

Steroid 5α-reductases (5αRs) catalyze the irreversible, NADPH-dependent reduction of Δ⁴-3-ketosteroids. This process regulates major steroid signaling pathways, including neurosteroid biosynthesis, androgen activation, and glucocorticoid metabolism. By governing these processes, 5αRs influence GABAergic and dopaminergic neurotransmission, and behavioral responses to stress and reward. Although the two principal isoenzymes, 5αR1 and 5αR2, have overlapping functions, they are increasingly recognized as functionally distinct. 5αR1 supports constitutive neurosteroid synthesis in several brain regions. Conversely, 5αR2, long known for converting testosterone to dihydrotestosterone during male sexual differentiation, has emerged as the enzyme that mediates rapid allopregnanolone production in the prefrontal cortex of males during acute stress. These advances may provide a framework for understanding the complex neurobehavioral profile of 5αR inhibitors such as finasteride and dutasteride. These drugs have been associated with depression, anxiety, suicidality, sexual dysfunction, and, in some individuals, persistent sexual, somatic, and neuropsychiatric symptoms after discontinuation. At the same time, 5αR inhibition may have therapeutic value in conditions characterized by excessive or maladaptive neurosteroid signaling, including tic disorders, impulse-control disorders, and substance use disorders. In this comprehensive review, we synthesize pharmacological, genetic, and behavioral evidence on the distinct contributions of 5αR1 and 5αR2 to steroid signaling across stress, sex, and development. Finally, we outline several open questions in the biology of 5αRs, the resolution of which will be essential to advancing toward mechanistically precise and clinically actionable interventions.

1. Introduction

Steroid 5α-reductases (5αRs) are a family of enzymes that catalyze the irreversible reduction of the Δ⁴ double bond in steroids such as testosterone, progesterone, and glucocorticoids. By converting these hormones into 5α-reduced metabolites with distinct biological activities, this reaction redirects steroid signaling toward physiological programs essential for sexual differentiation, neuroendocrine regulation, and stress adaptation.

The first 5α-reduced steroid was identified in 1931, when Adolf Butenandt isolated androsterone from male urine (Butenandt, 1931). The biochemical basis for the formation of 5α-reduced steroids, however, remained unresolved until 1951, when Schneider and Horstmann detected 5αR activity in rat liver preparations during studies of deoxycorticosterone (DOC) metabolism (Schneider and Horstmann, 1951). Subsequent work showed that this activity was NADPH-dependent and could reduce multiple steroid substrates, establishing 5α-reduction as a general enzymatic pathway rather than an isolated substrate-specific reaction (Tomkins, 1957). Yet the physiological significance of this pathway did not become clear until the late 1960s, when steroid 5α-reduction was linked to the conversion of testosterone into dihydrotestosterone (DHT). This conceptual transition was driven by evidence that DHT generates a more potent androgen-receptor signal than testosterone in key target tissues and accumulates preferentially within prostatic nuclei (Anderson and Liao, 1968; Bruchovsky and Wilson, 1968a, 1968b). Studies in androgen-responsive tissues, particularly the prostate (Bruchovsky and Wilson, 1968b) and perineal skin (Wilson and Walker, 1969), then established 5α-reduction as a tissue-selective mechanism for the local amplification of androgen signaling.

The developmental relevance of this mechanism was defined soon thereafter. In rabbit and rat embryos, 5αR activity was enriched in the primordia of the prostate and external genitalia before the onset of virilization, whereas it remained undetectable in Wolffian duct derivatives until later stages of differentiation (Wilson and Lasnitzki, 1971). These findings showed that 5α-reduction is not only tissue-selective but also developmentally regulated. They also led to a decisive prediction: hereditary disruption of this pathway should dissociate Wolffian-derived, testosterone-dependent differentiation from DHT-dependent development of the external genitalia. This prediction was confirmed shortly thereafter by independent clinical reports of 46,XY individuals born with ambiguous external genitalia who underwent marked virilization at puberty (Imperato-McGinley et al., 1974; Walsh et al., 1974). The disorder, originally termed pseudovaginal perineoscrotal hypospadias and later reclassified as 5αR type 2 deficiency, features normal circulating testosterone but markedly reduced DHT, indicating a selective impairment in testosterone 5α-reduction rather than a generalized defect in androgen production. These cases established a central principle of androgen biology: DHT is required for masculinization of the external genitalia and prostate, whereas testosterone is sufficient for Wolffian duct differentiation and several aspects of pubertal virilization (Wilson et al., 1993). The syndrome also revealed a clinically important pattern: affected males typically develop markedly hypoplastic prostates and appear largely protected from benign prostatic hyperplasia (BPH).

A further refinement in 5αR biology was the recognition that activity previously attributed to a single enzyme arises from multiple isoenzymes. Progress was long constrained by the lack of genetic probes, a limitation overcome in 1989 through expression cloning in Xenopus laevis oocytes (Andersson et al., 1989). This approach yielded a cDNA from rat liver encoding a 5αR, which was used to clone a human ortholog by cross-hybridization with a prostate cDNA library (Andersson and Russell, 1990). Strikingly, this gene was normal in subjects with pseudovaginal perineoscrotal hypospadias (Jenkins et al., 1992). The cloned enzyme therefore could not account for the developmental disorder, pointing instead to a second, functionally nonredundant isoenzyme. This hypothesis was confirmed with the identification of a second gene (Andersson et al., 1991). The two distinct genes were named SRD5A1 and SRD5A2 and correspond to the two main 5αR isoenzymes, 5αR1 and 5αR2. Subsequent work delineated biochemical and functional divergence of these isoenzymes, including differences in inhibitor sensitivity, pH optima, kinetic properties, developmental regulation, and tissue distribution (Jenkins et al., 1992; Russell and Wilson, 1994; Silver et al., 1994; Thigpen et al., 1993a; Thigpen et al., 1993b). These discoveries also made 5αR inhibition a rational strategy for reducing local androgenic tone in DHT-dependent tissues. Finasteride, a 5αR2-preferential inhibitor developed by Merck in the late 1980s (Gormley et al., 1990; Rittmaster et al., 1989; Stoner, 1990), and the dual inhibitor dutasteride (Clark et al., 2004; Roehrborn et al., 2002) entered clinical use for BPH and androgenetic alopecia (AGA).

A decisive expansion of 5αR biology came from Luciano Martini’s work on neuroendocrine steroid metabolism. This research helped establish that the brain and pituitary are not merely targets of circulating steroids but sites of local steroid conversion. In the rat hypothalamus and anterior pituitary, testosterone was shown to be metabolized to DHT and related 5α-reduced products (Massa et al., 1972), providing early evidence for 5α-reduction within neuroendocrine tissues. This concept was later extended across the central nervous system (CNS), where 5αR activity was found in multiple brain regions and shown to be particularly enriched in white matter and purified myelin (Celotti et al., 1987; Melcangi et al., 1988). Thus, CNS 5α-reduction can be understood as a local steroidogenic system capable of shaping neurosteroidogenesis, neurotransmission, and stress responsivity.

In the brain, 5αR converts progesterone and DOC into 5α-dihydroprogesterone (DHP) and 5α-dihydrodeoxycorticosterone (DHDOC). These are then 3α-hydroxylated to the neuroactive steroids 3α,5α-tetrahydroprogesterone (allopregnanolone; AP) and 3α,5α-tetrahydrodeoxycorticosterone (THDOC). The 5α-reduction step of this pathway was defined by studies showing the conversion of progesterone to DHP in the rat hypothalamus (Karavolas et al., 1976). The concept that the brain can synthesize steroid metabolites independently of peripheral endocrine organs was subsequently formalized with the term “neurosteroid” (Baulieu, 1998; Corpéchot et al., 1981). The identification of AP and THDOC as potent positive allosteric modulators of GABA-A receptors (Majewska et al., 1986) positioned 5αR-dependent neurosteroidogenesis as a proximal determinant of inhibitory neurotransmission. The finding that acute stress can rapidly elevate brain AP and THDOC (Barbaccia et al., 1996; Purdy et al., 1991), while repeated or chronic stress may dysregulate this response (Bortolato et al., 2011; Dong et al., 2001; Serra et al., 2000), further implicated 5αR activity as a key gatekeeper of stress responsivity.

Over the past two decades, accumulating evidence has linked 5αR inhibitors (5αRIs) to neuropsychiatric adverse effects, including mood and cognitive disturbances, with disrupted 5αR-dependent neurosteroidogenesis one plausible mechanism (Irwig, 2012a; Melcangi et al., 2017; Traish et al., 2015). Conversely, preclinical and clinical studies suggest 5αRs may be therapeutic targets, although evidence remains largely preliminary. Inhibition of these enzymes has shown potential efficacy in substance use disorders, impulse control disorders, premenstrual dysphoric disorder (PMDD), and hyperkinetic movement disorders (Bourque et al., 2024; Martinez et al., 2016; Paba et al., 2011).

Taken together, these observations indicate that 5αRs should not be regarded as ancillary steroid-metabolizing enzymes, but rather as regulatory nodes that determine how steroid signals are amplified, redirected, or constrained across tissues and physiological states. This review synthesizes current knowledge of 5αR biology in the CNS. We first examine the enzymology and molecular architecture of the 5αR family, emphasizing gene organization, phylogeny, and isoenzyme convergence and divergence (Section 2). We then consider the metabolic roles of 5αRs across major steroidogenic pathways, including neurosteroid and androgen signaling (Section 3), and their contribution to neurotransmission and stress responsivity (Section 4). We next review the pharmacology and toxicology of 5αRIs, focusing on neuropsychiatric effects (Section 5), and their therapeutic potential across neuropsychiatric disorders (Section 6). Section 7 discusses evidence that 5αR1 and 5αR2 subserve distinct, dynamic, and sex-dependent functions, with 5αR1 supporting a constitutive tonic baseline and 5αR2 implicated in phasic AP synthesis during acute stress in males. Finally, Section 8 considers how this isoenzyme-resolved framework may clarify the behavioral effects of 5αRIs and help resolve open questions, including sex differences in vulnerability to stress-related psychopathology.

2. The 5αR Family: Enzymology, Phylogeny, and Isoenzymes

2.1. Enzymology

As noted in the introduction, 5αR [3-oxo-5α-steroid 4-dehydrogenase (NADP⁺); EC 1.3.1.22] catalyzes the irreversible, NADPH-dependent saturation of the Δ⁴ double bond in Δ⁴-3-ketosteroid substrates (3-oxo-Δ⁴-steroids), yielding the corresponding 5α-reduced products. The reaction involves a stereospecific reduction of the Δ⁴ bond, with transfer of a hydride anion (H⁻) from NADPH to C5 on the α (below-plane) face (Russell and Wilson, 1994). The overall reaction can be written as:

$$3-\text{oxo}-{\Delta }^4-\text{steroid}+\text{NADPH}+H^{+}\to 3-\text{oxo}-5\text{α}-\text{steroid}+{\mathrm{NADP}}^{+}$$

Saturation of the Δ⁴ bond abolishes its conjugation with the C3 ketone and yields a trans-fused A/B ring junction with the C5 hydrogen in the α orientation (Fig. 1). 5αRs do not catalyze 5β-reduction; this is instead performed by the structurally unrelated aldo-keto reductase AKR1D1, which generates the bent, cis-fused 5β configuration with distinct geometry and properties (Di Costanzo et al., 2008) (Fig. 1). Two crystallographic studies clarified the structural basis of 5α-reduction. The first, resolving human 5αR2 in complex with NADPH and finasteride, revealed a seven-transmembrane fold enclosing an hourglass-shaped cavity with two adjacent sites, a hydrophobic ketosteroid pocket and an NADPH pocket, with the nicotinamide ring being poised to deliver a hydride to C5 of the substrate (Xiao et al., 2020). The second showed that this fold is retained in a bacterial homolog, indicating conservation of the catalytic architecture from prokaryotes to humans (Han et al., 2021). Despite its seven-helix topology, the enzyme bears no structural relationship to G protein-coupled receptors; its membrane-embedded active site instead allows hydrophobic substrates to enter laterally from the lipid bilayer. The reaction follows an ordered bi-bi mechanism (Levy et al., 1990): NADPH binds first, from the cytoplasmic pool, inducing conformational changes that create a competent steroid-binding site, after which the steroid enters from the endoplasmic reticulum (ER) membrane (Han et al., 2021; Xiao et al., 2020). In the proposed catalytic mechanism, glutamate 57 and tyrosine 91 polarize and stabilize the C3 carbonyl, facilitating hydride transfer from NADPH to C5, generating a C3 enolate-like intermediate that is subsequently protonated at C4 to yield the saturated 5α-reduced product (Xiao et al., 2020).

Figure 1. Stereochemistry of 5α-reduction and other metabolic pathways of Δ⁴-3-ketosteroids .

The central 3-oxo-Δ⁴-steroid substrate undergoes irreversible 5α- or 5β-reduction catalyzed by 5α-reductase (5αR) or aldo-keto reductase 1D1 (AKR1D1), respectively, yielding 3-oxo-5α- and 3-oxo-5β-steroids with distinct A/B ring junction geometries (trans for 5α, cis for 5β). Each 5-reduced intermediate then serves as a substrate for reversible C3-keto reduction by 3α- or 3β-hydroxysteroid dehydrogenase (3αHSD, 3βHSD), producing four stereochemically distinct 3-hydroxy metabolites. Colored highlights denote the A-ring region affected by each transformation. R represents the variable substituent at C17, which determines steroid class (e.g., androstanes, pregnanes, corticoids). Single arrows indicate irreversible reactions; double arrows indicate reversible, bidirectional interconversion.

The reaction is thermodynamically favorable and essentially irreversible under physiological conditions: no mammalian enzyme has been identified that catalyzes the reverse desaturation at appreciable rates (Azzouni et al., 2012; Russell and Wilson, 1994). Consequently, once a steroid undergoes 5α-reduction, it is irrevocably diverted from the parent precursor pool, and its subsequent fate depends largely on downstream metabolism of the 5α-dihydro intermediate. These downstream transformations include reduction at C3 by 3α- and 3β-hydroxysteroid oxidoreductases (3α-HSOR, 3β-HSOR) to yield metabolites with distinct receptor-binding profiles and biological activities (Penning et al., 2000; Russell and Wilson, 1994). It should be noted that, in the literature, these enzymes are often designated 3α- and 3β-hydroxysteroid dehydrogenases (3α-HSD, 3β-HSD); although that convention names only the oxidative direction, both sets of terms are used interchangeably throughout this article. Unlike 5α-reduction, this step is reversible, and the back-oxidation of 3α-hydroxy metabolites to their 3-oxo precursors provides a mechanism for the dynamic regulation of neuroactive steroid levels. Further transformations include reduction at C20 by 20α-hydroxysteroid dehydrogenase (20αHSD) (Penning et al., 2000; Usami et al., 2002), and conjugation with sulfate or glucuronyl moieties to facilitate elimination (Bélanger et al., 2003; Mueller et al., 2015).

2.2. Gene Superfamily and Phylogeny

In mammals, the 5αR superfamily comprises five evolutionarily related membrane-embedded proteins encoded by SRD5A1, SRD5A2, SRD5A3, TECR, and TECRL. All share a conserved catalytic domain and a common structural organization, and those whose activity has been characterized are NADPH-dependent reductases; TECRL, however, has no demonstrated catalytic function and is assigned to the family on the basis of sequence homology alone. Phylogenetic analyses support descent from an ancestral reductase whose domain architecture likely predates the diversification of eukaryotic lineages, consistent with the bacterial homolog described above, and which underwent early duplications, generating three principal subfamilies: SRD5A1/2/DET2, SRD5A3/PPRD, and TECR/CER (Langlois et al., 2010). Representatives of these subfamilies are found across major eukaryotic lineages, including protists, plants, fungi, and animals, consistent with deep evolutionary conservation of this catalytic strategy. A recent comprehensive phylogenomic analysis further supports the presence of all major 5αR subgroups in early protist lineages and identifies lineage-specific diversification, including a plant-specific DET2-like (DET2L) subclass, suggestive of neofunctionalization (Ali et al., 2024). 5αR1 and 5αR2 diverged later, likely with the emergence of early vertebrates (Langlois et al., 2010), and both are conserved across vertebrates, although lineage-specific duplications and losses indicate substantial evolutionary plasticity.

Of the five superfamily members, only SRD5A1 and SRD5A2 encode canonical steroidogenic enzymes responsible for the 5α-reduction of Δ⁴-3-ketosteroids (see Table 1 for their enzymological and molecular characteristics). Despite catalyzing the same overall reaction, 5αR1 (259 amino acids, ~29.5 kDa) and 5αR2 (254 amino acids, ~28.4 kDa) share only 50% sequence identity within a species, whereas each isoenzyme is markedly better conserved between orthologs (approximately 60% identity between human and rat 5αR1, and 77% between human and rat 5αR2; Azzouni et al., 2012). This asymmetry indicates an ancient duplication followed by sustained purifying selection on each paralog independently, in line with the distinct biochemical properties and expression patterns detailed below. The two genes have a similar architecture, with five exons separated by four introns, but lie on different chromosomes (SRD5A1 at 5p15; SRD5A2 at 2p23).

Table 1. Characteristics of 5αR1 and 5αR2 in humans.

Gene coordinates, spans, transcript models, and exon/intron counts refer to the canonical MANE transcripts on GRCh38 (NCBI RefSeq/MANE; Ensembl). Gene-level annotation reports more exonic segments (7 for SRD5A1, 9 for SRD5A2) and, for SRD5A2, a much larger gene-level locus (~140.5 kb). Protein accessions from UniProt. Abbreviations: AP, allopregnanolone; DOC, deoxycorticosterone; DSD, disorder of sex development; ER, endoplasmic reticulum; MANE, Matched Annotation from NCBI and EMBL-EBI; 7-TM, seven transmembrane domains.

Property	5αR1	5αR2
Gene	SRD5A1	SRD5A2
Chromosome	5p15.31	2p23.1
Canonical transcript	NM_001047.4 (ENST00000274192.7)	NM_000348.4 (ENST00000622030.2)
Gene span	40,947 bp	58,459 bp
Exons/Introns	5 / 4	5 / 4
Protein (aa)	259 aa, ~29.5 kDa	254 aa, ~28.4 kDa
UniProt	P18405	P31213
Topology	7-TM, ER membrane	7-TM, ER membrane
pH optimum	neutral (6.0–8.5)	acidic (5.0–5.5)a
Km (testosterone)	~3 μM	~0.7 μMb
Km (progesterone)	~ 0.3 μM	~0.1 μMb
Vmax (steroid conversion)	Higher	Lowerc
Primary substrates	Progesterone, testosterone, DOC, corticosterone	Testosterone (higher affinity), progesterone
Main tissues	Skin, liver, brain (widespread), adrenals	Prostate, epididymis, genital skin; brain
Brain function	Basal (tonic) neurosteroid synthesis	Phasic, stress-induced AP synthesis (males)
KO phenotype	Parturition defects; altered glucocorticoid metabolism; reduced bone mass and muscle strength (mouse)	Reduced prostate; male-specific stress-coping deficits
Human deficiency	Not described	46,XY DSD (ambiguous genitalia, pubertal virilization)

^aAcidic optimum may reflect in vitro assay conditions; both isoenzymes may operate near neutral pH in native membranes (Iehlé et al., 1995; Levy et al., 1995).

^bDirectly isoenzyme-assigned human recombinant value (Iehlé et al., 1995; Levy et al., 1995). Testosterone Km is strongly assay-dependent; purified, reconstituted-enzyme assays report higher values (Peng et al., 2020).

^cFrom purified, liposome-reconstituted human enzymes (Peng et al., 2020): 5αR2 shows a lower Km than 5αR1 for testosterone, progesterone, androstenedione, and 17-hydroxyprogesterone, whereas 5αR1 has the higher Vmax. Progesterone Km values are approximate and assay-dependent.

The remaining three members of the superfamily serve distinct metabolic functions. SRD5A3 was initially proposed to encode a third 5αR (5αR3) capable of converting testosterone into DHT (Uemura et al., 2008), but this interpretation was revised when 5αR3 was shown to function primarily as a polyprenol reductase required for dolichol biosynthesis and N-linked glycosylation (Cantagrel et al., 2010). TECR encodes a trans-2,3-enoyl-CoA reductase that catalyzes the final step of each very-long-chain fatty acid elongation cycle (Moon and Horton, 2003) and additionally acts in the sphingosine-1-phosphate degradation pathway (Wakashima et al., 2014). TECRL encodes a related paralog whose enzymatic function is uncertain; it was named by homology and has been associated genetically with an inherited arrhythmia syndrome with features of both catecholaminergic polymorphic ventricular tachycardia and long QT syndrome (Devalla et al., 2016). Neither TECR nor TECRL has an established role in steroid metabolism.

2.3. 5αR1 (SRD5A1)

The enzyme 5αR1, encoded by SRD5A1 at chromosomal locus 5p15.31, was the first 5αR isoenzyme to be molecularly characterized (Andersson et al., 1989; Andersson and Russell, 1990). The gene spans 40,947 bp (GRCh38.p14: NC_000005.10:6,633,440–6,674,386) and retains the five-exon architecture described in Section 2.2, conserved between humans and rodents. Genetic variation in SRD5A1 is extensive, with over 550 reported SNPs, but it is overwhelmingly concentrated in synonymous, intronic, and untranslated region positions: no coding variant producing loss of enzymatic function has been conclusively documented. Several non-missense variants have nonetheless been associated with androgen-dependent and affective phenotypes (Table 2), including reduced severity of premenstrual symptoms (rs501999; Adams and McCrone, 2012), baseline positive affect (rs472402; Hart et al., 2012), and alcohol dependence (rs248793; Milivojevic et al., 2011).

Table 2.

Key functionally and clinically characterized SNPs in SRD5A1

SNP ID	Amino Acid Change	Location	Variant Type	Functional Effect and Clinical Associations
CODING VARIANTS
rs248793	Thr160Thr (synonymous)	Exon	Synonymous	Associated with altered DHT/T ratio (Ellis et al., 2005) Alcohol dependence (protective minor C allele) (Milivojevic et al., 2011) Peripheral artery disease (GG genotype risk) (Barresi et al., 2014; Signorelli et al., 2008)
rs3736316	Synonymous	Exon	Synonymous	May affect progesterone metabolism. Breast cancer risk modification with combined menopausal hormone therapy (significant after Bonferroni correction) (Hein et al., 2012) Associated with the risk of high grade prostate cancer in the PCPT (Price et al., 2016)
INTRONIC / REGULATORY VARIANTS
rs518673	—	Intron	Intronic	Protective allele (T) reduces prostate cancer BCR risk in combination with SRD5A2 rs12470143 (Audet-Walsh et al., 2011) Associated with PSA levels (Poniah et al., 2015)
rs166050	—	Intron	Intronic	Risk allele (C) associated with higher prostatic DHEA (+60%) and androsterone (+100%) (Lévesque et al., 2014) Posttreatment change in total prostate volume in BPH patient treated with 5α-reductase inhibitors and α-adrenergic receptor antagonists (Gu et al., 2013)
rs1691053	—	Intron	Intronic	AG or GG genotype associated with increased prostate cancer risk (OR increased, especially combined with HSD3B1 AA genotype) (Setlur et al., 2010) Associated with prostate-specific antigen levels at PCa diagnosis (Henríquez-Hernández et al., 2015) independently associated with glycometabolism (Liu et al., 2020)
rs3822430	—	Intron	Intronic	Associated with prostate-specific antigen levels at PCa diagnosis (Henríquez-Hernández et al., 2015) Associated with the risk of high grade prostate cancer in the PCPT (Price et al., 2016)
rs501999	—	Intron	Intronic	CC genotype protected women against severe premenstrual syndrome (Adams and McCrone, 2012)
rs6884552	—	Intron	Intronic	Significantly associated with baseline I-PSS (Gu et al., 2013)
rs3797177	—	Intron	Intronic	Significantly associated with baseline I-PSS (Gu et al., 2013) Carriers of the A allele showed lower serum insulin levels compared with GG homozygotes. (Rył et al., 2017)
rs472402	—	Intron	Intronic	GG genotype reduced the risk of developing high-grade cancer when assigned to finasteride in PCPT (Dai et al., 2019) Associated with the risk of high grade prostate cancer (Price et al., 2016) Associated with baseline positive affect (Hart et al., 2012)

List of abbreviations: BCR, biochemical recurrence; BPH, benign prostatic hyperplasia; DHEA, dehydroepiandrosterone; DHT, dihydrotestosterone; HSD3B1, gene encoding 3β-hydroxysteroid dehydrogenase type 1; I-PSS, International Prostate Symptom Score; OR, odds ratio; PCa, prostate cancer; PCPT, Prostate Cancer Prevention Trial; PSA, prostate-specific antigen; T, testosterone.

The protein comprises 259 amino acids, with a calculated molecular mass of approximately 29.5 kDa, and is highly lipophilic and non-glycosylated (Andersson and Russell, 1990; Russell and Wilson, 1994). Its three-dimensional structure has not been solved, but hydropathy analysis predicts multiple membrane-spanning segments (Andersson and Russell, 1990), and the ~50% identity it shares with 5αR2 makes it likely that it adopts the same seven-transmembrane architecture defined crystallographically for that isoenzyme (Xiao et al., 2020). Human and rat orthologs share approximately 60% amino acid identity, with the highest conservation in the predicted substrate- and cofactor-binding regions, indicating strong selective constraint on the catalytic mechanism alongside greater divergence in regulatory and species-specific features (Andersson and Russell, 1990).

Within the central nervous system (CNS), it is found throughout the cortex, hippocampus, hypothalamus, and cerebellum (Azzouni et al., 2012; Pelletier et al., 1994). Outside the CNS, high levels of 5αR1 are found in the skin (Chen et al., 1996; Thiboutot et al., 1995) and in the liver (Nixon et al., 2012; Toews et al., 2021). It is also present in both subcutaneous and visceral depots of adipose tissue ((Fouad Mansour et al., 2016) and osteo-blasts (Issa et al., 2002), Prostatic levels are comparatively low under physiological conditions but increase in castration-resistant prostate cancer (Mostaghel and Nelson, 2008). In the female reproductive tract 5αR1 is expressed in the endometrium, where it participates in vascular remodeling during decidualization (Shaw et al., 2022).

At the subcellular level, 5αR1 is an integral membrane protein localized primarily to the ER and the contiguous nuclear envelope (Azzouni et al., 2012; Russell and Wilson, 1994); this arrangement allows the membrane-embedded active site to draw lipophilic substrates laterally from the surrounding bilayer while the adjacent cofactor site is supplied from the cytosolic NADPH pool. Kinetically, 5αR1 accepts a broad range of substrates with affinities that differ by substrate: Michaelis constants for testosterone lie in the low micromolar range (~1–5 μM), whereas those for progesterone are submicromolar (Table 1). Affinity for NADPH is likewise modest, and the pH optimum is broad (pH 6.0–8.5) in cell-free assays (Russell and Wilson, 1994). These estimates derive from detergent-solubilized preparations, vary severalfold with assay conditions, and are therefore best read as relative rather than absolute. One implication is nonetheless robust: because tissue concentrations of Δ⁴-3-ketosteroids are typically nanomolar, and thus well below the Km for every characterized substrate, 5αR1 operates in the first-order regime, where product formation scales linearly with substrate supply rather than being capped by enzyme capacity. Flux is accordingly set by the product of catalytic efficiency and enzyme abundance rather than by affinity. Together with the constitutive expression described above, this regime is consistent with a generalist role in the continuous processing of multiple substrates in proportion to their availability.

Regulation of 5αR1 occurs principally at the level of gene expression and is influenced by developmental and hormonal factors, with post-translational regulation remaining largely uncharacterized. In rodents, hepatic expression increases postnatally and is female-predominant, reflecting the continuous, female-type pattern of growth hormone secretion; thyroid hormone and glucocorticoids likewise act as positive modulators, whereas androgens are inhibitory (El-Awady et al., 2004; Toews et al., 2021).

The behavioral significance of 5αR1 follows from this profile: an enzyme that is broadly distributed, constitutively expressed, and unresponsive to acute demand is positioned to contribute tonically to neurosteroid-dependent signaling rather than to govern a discrete behavioral output. The human variants noted above link 5αR1 to affective and addiction-related phenotypes, including premenstrual symptom severity, positive affect, and alcohol dependence, associations that are individually modest and mechanistically indirect but that are collectively consistent with a constitutive role in neurosteroid-dependent emotional regulation. In keeping with the absence of characterized loss-of-function variants, no 5αR1-deficiency syndrome has been described, and Srd5a1-knockout mice display no overt developmental phenotype, the most striking defect being in female parturition (Mahendroo et al., 1999, 1996), although metabolic and skeletal alterations have also been reported (Livingstone et al., 2015; Windahl et al., 2011). This may reflect modest consequences of 5αR1 loss, isoenzyme compensation, or phenotypes too subtle for clinical recognition.

2.4. 5αR2 (SRD5A2)

The enzyme 5αR2, encoded by SRD5A2 at chromosomal locus 2p23.1, was cloned shortly after SRD5A1 from a prostate cDNA library (Andersson et al., 1991). The gene spans 58,459 bp (GRCh38.p14: NC_000002.12:31,522,480–31,580,938). It shares with SRD5A1 the five-exon architecture and common ancestry described in Section 2.2 (Labrie et al., 1992), although it spans roughly 40% more genomic sequence, a disparity accounted for almost entirely by intronic expansion. The two genes differ more consequentially in their mutational landscapes: whereas no coding variant in SRD5A1 has been linked to a Men-delian phenotype, loss-of-function mutations in SRD5A2 cause a well-defined disorder of sex development. The clinical syndrome of 5αR2 deficiency was first described in kindreds from the Dominican Republic (Imperato-McGinley et al., 1974); the gene was identified, and its mutations characterized, only after cloning (Andersson et al., 1991), which also revealed large structural defects such as deletion of most of SRD5A2 in a Papua New Guinea kindred. The Human Gene Mutation Database catalogs roughly 120 pathogenic SRD5A2 variants, more than 50 of them missense (see Table 3 for a list of representative variants), most clustering in exons 1 and 4. Among common polymorphisms, only a subset measurably affects activity: in vitro, the V89L variant reduces activity by approximately 30% (Makridakis et al., 1997), whereas A49T increases Vmax roughly fivefold (Makridakis et al., 2000), although the disease associations of both have proved inconsistent across studies (Cussenot et al., 2007; Ntais et al., 2003).

Table 3.

Key functionally and clinically characterized SNPs in SRD5A2

SNP ID	Amino Acid Change	Location	Variant Type	Functional Effect and Clinical Associations
CODING / MISSENSE VARIANTS
rs9282858	A49T (Ala→Thr)	Exon 1	Missense (gain-of-function)	~5-fold increase in enzyme activity; increased DHT production (Makridakis et al., 2000) Contradictory results on PCa risk (Ayeni et al., 2024; Fang et al., 2017; Li et al., 2010) Correlated with susceptibility to BPH (Zeng et al., 2017)
rs523349	V89L (Val→Leu)	Exon 1	Missense (loss-of-function)	Reduction in enzyme activity (Makridakis et al., 2000, 1997) 46XY DSD (Eren et al., 2016; Hackel et al., 2005) Associated with testosterone and free testosterone serum levels (Allen et al., 2001) Significant correlation with androstane-3α,17β-diol-3G in the serum (Jiang et al., 2010) and with circulating levels of 3α-diol-3G (Lévesque et al., 2014) Associated with reduced BCR after radical prostatectomy (Audet-Walsh et al., 2011) Associated with high-stage PCa (Fernandez et al., 2012) Correlated with durability of tumor response to ADT and overall survival in metastatic prostate cancer (Shiota et al., 2015) Associated with baseline prostate volume in BPH patients and posttreatment I-PSS change (Gu et al., 2013). Associated with SHBG levels and incidence rate of MetS among BPH patients (Rył et al., 2017). Correlated with metabolic syndrome in testicular cancer survivors (Boer et al., 2016) Correlated to hypospadias risk (Diposarosa et al., 2024; Sun et al., 2019; van der Zanden et al., 2012) The Leu allele was associated with protection against PCOS (Goodarzi et al., 2006; Graupp et al., 2011) Associated with PTSD symptoms in males but not females (Gillespie et al., 2013) Associated with ALTs in a general population (Zettergren et al., 2013)
rs9332964	R227Q (Arg→Gln)	Exon 4	Missense (loss-of-function)	Reduction in enzyme activity (Makridakis et al., 2000). Correlated with micropenis incidence (Sasaki et al., 2003) and 46,XY DSD (Fan et al., 2020; Nie et al., 2011). Associated with ASD (Bui et al., 2024)
rs9332967	R246Q (Arg→Gln)	Exon 4	Missense (loss-of-function)	Reduced enzyme activity (Makridakis et al., 2000; Thigpen et al., 1992; Wigley et al., 1994; Wilson et al., 1993), attributed to an increased NADPH binding Km (Avendaño et al., 2018). Reduced external masculinization and higher cryptorchidism incidence in 46,XY DSD (Eunice et al., 2008; Jia et al., 2018; Ko et al., 2010; Nagaraja et al., 2010; Nie et al., 2011; Rawal et al., 2025; Seo et al., 2023)
rs121434250	G196S (Gly→Ser)	Exon 4	Missense (loss-of-function)	Reduced NADPH binding Km (Avendaño et al., 2018) 46,XY DSD (Bertelloni et al., 2007; Hackel et al., 2005; Nagaraja et al., 2010)
—	Q6X	Exon 1	Nonsense	Premature stop codon; truncated protein with complete loss of enzymatic activity due to the lack of both the testosterone- and the NADPH-binding domains. 46,XY DSD (compound heterozygous with H232R) (Li et al., 2022a; Nie et al., 2011)
—	H232R	Exon 4	Missense	Disrupts NADPH binding domain. 46,XY DSD (compound heterozygous with Q6X) (Li et al., 2022a).
—	Complete gene deletion	Whole gene	Gross deletion	Complete loss of function. First reported SRD5A2 mutation (Papua New Guinea kindred); classical 5αR2 deficiency (Andersson et al., 1991).
INTRONIC / REGULATORY VARIANTS
rs9332975	—	3’-UTR	Non-coding	Associated with baseline prostate volume in BPH patients (Gu et al., 2013) and with hypospadias (Carmichael et al., 2014)
rs632148	—	5’-UTR	Non-coding	Associated with syndromic hypospadias (Diposarosa et al., 2024) Associated with sperm motility (Zhao et al., 2012) Associated with testicular volume (Peters et al., 2010)
rs12470143	—	Intron	Intronic	Significant correlation with androstane-3α,l7β-diol-3G in the serum in Chinese population (Jiang et al., 2010) Associated with testicular volume (Peters et al., 2010) and with hypospadias (Carmichael et al., 2014). Associated with reduced BCR after radical prostatectomy (Audet-Walsh et al., 2011). Associated with a change HDL and FT levels, as well as with incidence rate of MetS among BPH patients (Rył et al., 2017). Associated with high circulating androstane-3α, 17β-diol-17G-glucuronide (Lévesque et al., 2014)
rs2208532	—	Intron	Intronic	Associated with reduced BCR after radical prostatectomy (Audet-Walsh et al., 2011) Associated with circulating 3α-androstanediol-glucuronide (Ahn et al., 2009) and with intraprostatic testosterone levels (Lévesque et al., 2014). Associated with hypospadias (Carmichael et al., 2014). PCa clinical stage (Choi et al., 2015)
rs13395648	—	Intron	Intronic	Associated with semen volume (Zhao et al., 2012). Associated with hypospadias (Carmichael et al., 2014).
rs4952197	—	Intron	Intronic	Associated with reduced BCR after radical prostatectomy (Audet-Walsh et al., 2011)
rs676033	—	Intron	Intronic	Strong LD with rs523349 (V89L); Associated with higher circulating levels of 3α-diol-3G and with enhanced intraprostatic DHT and 3β-diol (Lévesque et al., 2014)
rs522638	—	Intron	Intronic	Associated with syndromic hypospadias (Diposarosa et al., 2024)
rs508562	—	Intron	Intronic	PCa risk, PSA level, and clinical stage (Choi et al., 2015)
rs11675297	—	Intron	Intronic	PCa risk, PSA level, and clinical stage (Choi et al., 2015)
rs7594951	—	Intron	Intronic	AA carriers showed higher serum DHT levels (Setlur et al., 2010). Associated with hypospadias (Carmichael et al., 2014).
MICROSATELLITE / REPEAT VARIANTS
(TA)n repeat	—	3’ UTR	Dinucleotide repeat	Longer TA repeats are associated with lower enzymatic activity (Bharaj et al., 2000) Correlated with breast cancer prognosis (Bharaj et al., 2000) Variable repeat length (ranging from 0 to 18 repeats) with ancestry-dependent prevalence (Bharaj et al., 2000; Kantoff et al., 1997; Samtani et al., 2015).

List of abbreviations: 3α,17β-diol-3G, 3α,17β-diol-3glucuronide; 3α-diol-3G, 3α-diol-3glucuronide; ADT, androgen deprivation therapy; ALTs, autistic-like traits; ASD, autism spectrum disorder; BCR, biochemical recurrence; DSD, disorder of sex development; FT, free testosterone; HDL, high-density lipoprotein; LD, linkage disequilibrium; MetS, metabolic syndrome; PCa, prostate cancer; PCOS, polycystic ovary syndrome; SHBG, sex hormone-binding globulin.

The 5αR2 protein shares ~50% sequence identity with 5αR1 (Section 2.2) and is comparable in size (254 amino acids), lipophilicity, and ER localization. Kinetically, however, the two isoenzymes differ sharply: 5αR2 has substantially higher substrate affinity, with testosterone Km values in the submicromolar range (~0.1–1.0 μM), roughly 4-fold lower than 5αR1 (Russell and Wilson, 1994; Thigpen et al., 1993a), and an acidic pH optimum (pH 5.0–5.5) in cell-free assays, subject to the same assay caveats noted above. The acidic optimum sits awkwardly with the near-neutral pH of the cytosol, from which the active site draws its cofactor, and its physiological significance remains unclear. These differences point to the possibility of complementary roles, with 5αR2 as the higher-affinity, lower-capacity isoenzyme and 5αR1 as its lower-affinity, higher-capacity counterpart (Table 1).

In the CNS, 5αR2 was long considered largely restricted to fetal and neonatal stages (Poletti et al., 1998; Torres and Ortega, 2003). This view was partially revised by immuno-histochemical mapping of the adult rat brain, which revealed a widespread but markedly heterogeneous distribution (Castelli et al., 2013). Immunoreactivity was most intense in the neocortex, peaking in frontal and somatosensory areas, and was also dense in the olfactory bulb, hippocampal CA3 field, basolateral amygdala, thalamic nuclei, locus coeruleus, and cerebellar Purkinje cell layer, with moderate to low levels in the hypothalamus, midbrain dopaminergic and raphe nuclei, dorsal caudate, and nucleus accumbens (NAc), where the shell labeled more strongly than the core. Notably, this protein-level distribution is difficult to reconcile with the earlier transcript data, and the discrepancy has yet to be resolved. In the cortex, amygdala, and NAc, the pattern places 5αR2 within circuits governing emotion, motivation, and cognition. At the cellular level, cortical 5αR2 immunoreactivity was localized to neuronal somata, mainly the larger pyramidal neurons of layers II, III, and V, co-localizing with NeuN but not GFAP (glial fibrillary acidic protein) and with only a minority of GAD67-positive GABAergic neurons (Castelli et al., 2013); more recent work in the rat medial PFC localizes the enzyme to the somata and neurites of pyramidal neurons (Cadeddu et al., 2025). Because GFAP is an astrocytic marker, however, these studies did not test for oligodendroglial expression. Preliminary analyses of publicly available human single-cell datasets indicate that cortical expression is largely restricted to two cell populations: pyramidal cells and oligodendrocytes (M. Bortolato, unpublished observations). Consistent with an oligodendroglial role, knocking down 5αR2 in the rat medial PFC alters the transcriptome chiefly in oligodendrocytes, which show broad upregulation of biosynthetic and metabolic pathways (Cadeddu et al., 2025), in line with foundational studies identifying oligodendrocytes as key sites of 5αR expression (Baulieu and Robel, 1990; Hu et al., 1987; Jung-Testas et al., 1989; Melcangi et al., 1998, 1988).

Outside the CNS, 5αR2 is the principal isoenzyme of classical androgen-target tissues. Its expression predominates in the male urogenital tract, including the prostate (along with its hyperplastic and neoplastic forms), the seminal vesicles, epididymis, and genital skin, where it generates the high local DHT concentrations on which these tissues depend (Andersson et al., 1991; Thigpen et al., 1993b). It is also expressed in the liver, although only postnatally (Thigpen et al., 1993b). Although 5αR1 is the predominant scalp isoenzyme, 5αR2 is detectable within the hair follicle, in the inner root sheath, the inner layer of the outer root sheath, and the infundibulum (Bayne et al., 1999). This concentration in androgen-dependent tissues contrasts with the broad, constitutive distribution of 5αR1 and is consistent with a more specialized, high-affinity role for 5αR2 in amplifying androgen signaling where it is required.

Regulation of 5αR2 occurs predominantly at the level of expression and is shaped by developmental and hormonal factors. Transcript levels peak in fetal and neonatal urogenital tissues and follow a similar developmental course in the CNS, declining from an early peak to low levels in adulthood (Poletti et al., 1998; Thigpen et al., 1993b; Torres and Ortega, 2003). A further layer of control is epigenetic. Interestingly, roughly one-third of benign adult prostates lack detectable 5αR2 (Niu et al., 2011), reflecting hypermethylation of the SRD5A2 promoter, which increases with age and inflammatory signaling (Ge et al., 2015). Whether comparable epigenetic regulation operates in the brain remains unknown.

Genotype-phenotype correlations in 5αR2 deficiency reveal considerable variability in clinical presentation. Although the two most frequent missense variants, R227Q and R246Q, both cause 46,XY disorders of sex development (DSD), R227Q is associated with a comparatively more masculinized phenotype and a lower incidence of cryptorchidism (Fan et al., 2020; Nie et al., 2011), whereas carriers of the cofactor-binding variant R246Q exhibit reduced masculinization and higher rates of cryptorchidism (Seo et al., 2023). Beyond monogenic disorders, SRD5A2 variations may also contribute to complex neuropsychiatric phenotypes, as illustrated by the association of the low-activity V89L polymorphism (rs523349) with post-traumatic stress disorder (PTSD) symptom severity in a large cohort, with sex-dependent effects and internal replication (Gillespie et al., 2013).

3. 5αR-Dependent Metabolic Pathways

The 5α-reduction reaction described in Section 2.1 is best understood as a family of irreversible Δ⁴-3-ketosteroid transformations whose biological meaning depends on substrate, isoenzyme distribution, and cellular context. The substrate scope is broad: besides testosterone and progesterone, characterized substrates include androstenedione, DOC, cortisol, and corticosterone (Azzouni et al., 2012; Russell and Wilson, 1994) (Fig. 2), plus aldosterone, 17α-hydroxyprogesterone, androstadienone, and the 11-oxygenated C19 steroids. The following sections outline these pathways, emphasizing their significance for the CNS.

Figure 2. Catalytic functions of 5α-reductase (5αR).

(A) Under conditions of reduced 5αR activity, testosterone (TEST) remains the primary activator of androgen receptors (AR), while corticosterone in rodents and cortisol in humans (CORT) activate glucocorticoid receptors (GR). Conversion of progesterone (PROG) to dihydroprogesterone (DHP) and subsequently to allopregnanolone (AP) is limited. (B) When 5αR activity is intact, TEST is converted to the more potent AR agonist dihydrotestosterone (DHT), which can be further metabolized to 5α-androstane-3α, 17β-diol (3α-DIOL). PROG is converted to DHP, promoting AP synthesis. Together, 3α-DIOL and AP act as positive allosteric modulators of GABA-A receptors. In parallel, conversion of CORT to dihydrocorticosterone (DHB) or dihydrocortisol (DHF) alters GR activation, thereby reducing the impact of hypothalamic–pituitary–adrenal (HPA) axis signaling on neuronal function. Adapted from Servier Medical Art (https://smart.servier.com).

3.1. The GABAergic Axis

Among the most consequential CNS functions of 5αRs is the conversion of progesterone and DOC into the 5α-reduced pregnane intermediates DHP and DHDOC, respectively (Fig. 3). In the brain, the reduction of progesterone to DHP is the first committed step of local GABAergic neurosteroid synthesis. DHP has a pharmacological profile that only partially overlaps with that of its downstream metabolites. It binds the classical nuclear pro-gesterone receptor (nPR) (Iswari et al., 1986; Scholtz et al., 2014; Wierer et al., 2012) and, to a lesser extent, the membrane progesterone receptor alpha (mPRα) (Kelder et al., 2022). DHP is also anticonvulsant at doses well below those producing sedation (Lonsdale et al., 2006; Lonsdale and Burnham, 2003; Wu and Burnham, 2018). Because DHP lacks the 3α-hydroxyl required for potent GABA-A modulation, this activity likely reflects, at least in part, local conversion to AP, although the dose-response profile is not fully accounted for by progesterone or AP alone.

Figure 3. The GABAergic axis: 5α-reduction of progesterone and deoxycorticosterone (DOC) and downstream metabolic transformations.

5αRs catalyze the irreversible 5α-reduction of progesterone and DOC, yielding 5α-dihydroprogesterone (DHP) and 5α-dihydrodeoxycorticosterone (DHDOC), respectively. These 5α-reduced intermediates retain both a C3 ketone, which can be reduced to either of two epimeric alcohols by 3α- or 3β-hydroxysteroid dehydrogenase (3αHSD or 3βHSD, respectively), and a C20 ketone, which can be independently reduced by 20α-hydroxysteroid dehydrogenase (20αHSD). The 3α-reduced metabolites, 3α,5α-tetrahydroprogesterone (allopregnanolone; AP) and 3α,5α-tetrahydrodeoxycorticosterone (THDOC), act as potent positive allosteric modulators of GABA-A receptors. The 3β-reduced epimers (isoAP and isoTHDOC) lack this activity; isoAP (sepranolone) instead functions as a selective antagonist of AP-mediated GABA-A modulation. Single-headed arrows indicate irreversible reactions; double-headed arrows indicate reactions that are formally reversible but typically favor reduction under physiological conditions.

The same enzymatic step reduces DOC to DHDOC, but because this conversion occurs mainly in the adrenal gland, the amount of DHDOC produced depends on adrenal DOC output and therefore tracks hypothalamic-pituitary-adrenal (HPA) axis activation. DHDOC’s intrinsic actions remain largely uncharacterized, with the exception of reported anticonvulsant properties (Perez-Cruz et al., 2006). DHP and DHDOC can each be reduced at two positions, with three distinct functional outcomes. At C3, 3α-reduction generates the GABA-A receptor positive modulators AP and THDOC, whereas 3β-reduction yields their epimers; at C20, reduction diverts intermediates away from the neuroactive branch. Each of these fates is considered separately below.

3α-reduction: generation of AP and THDOC.

Reduction at C3 to the 3α-hydroxy epimer is mediated primarily by 3α-HSORs, chiefly AKR1C-family aldo-keto reductases. Among these, AKR1C2 is among the most abundant brain isoenzymes, whereas AKR1C3 and AKR1C4 contribute mainly in peripheral tissues, particularly liver and prostate (Penning et al., 2000; Steckelbroeck et al., 2004). AKR1C2 acts as a 3-ketosteroid reductase on pregnane substrates, generating AP from DHP and THDOC from DHDOC (Usami et al., 2002); although bidirectional in vitro, it operates reductively in intact cells, where the high cytosolic NADPH/NADP+ ratio drives net flux toward the 3α-hydroxy products.

AP and THDOC share a highly convergent pharmacological profile. Both are potent positive allosteric modulators of GABA-A receptors, acting at a transmembrane site at the β–α subunit interface that is distinct from the benzodiazepine site (Chen et al., 2018; Laverty et al., 2017; Miller et al., 2017). Both potentiate receptor responses at synaptic γ2-containing and extrasynaptic δ-containing receptor populations (Belelli and Lambert, 2005; Majewska et al., 1986; Reddy and Rogawski, 2002). Their action on δ-containing extrasynaptic receptors, enriched in dentate gyrus granule cells, thalamic relay neurons, and cortical inter-neurons, enhances tonic inhibition and contributes substantially to their anxiolytic and stress-adaptive effects. Their action on synaptic γ2-containing receptors contributes to phasic inhibition, whereas sedative and anticonvulsant effects involve both receptor populations (Carver and Reddy, 2013). At micromolar concentrations, both steroids can also directly gate GABA-A receptors in the absence of GABA. Through these combined actions, AP and THDOC produce anxiolytic, anticonvulsant, sedative, and antidepressant-like effects and contribute to behavioral and neuroendocrine adaptation to stress (Barbaccia et al., 1996; Pinna et al., 2006b; Purdy et al., 1991).

Despite this pharmacological convergence, AP and THDOC have partially divergent origins. THDOC depends almost entirely on adrenal DOC, and therefore rises rapidly after acute stress in parallel with ACTH-driven corticosterone or cortisol production (Barbaccia et al., 1996; Purdy et al., 1991). Thus, the same HPA axis drive that generates glucocorticoids also produces a GABAergic neurosteroid with rapid non-genomic inhibitory effects, providing a temporally coordinated counter-regulatory signal to stress-induced excitation. AP, by contrast, draws on three sources: adrenal progesterone, gonadal progesterone, and local synthesis from cholesterol in the brain through the classical steroidogenic cascade. Beyond the stress-linked adrenal component it shares with THDOC, AP also contributes to tonic inhibitory regulation, sex-hormone-dependent modulation, and longer-term stress adaptation (Barbaccia et al., 1996; Pinna et al., 2006b). Dynamic regulation of this pathway is most evident in the stress response: acute stressors rapidly elevate brain AP and THDOC, largely through increased precursor supply and adrenal output (Barbaccia et al., 1996; Purdy et al., 1991). Conversely, chronic stress can reduce cortical AP due to downregulation of 5αR1 expression (Dong et al., 2001).

3β-reduction: generation of isoAP and isoTHDOC.

The same C3 ketone can be reduced to the 3β-hydroxy epimer, yielding metabolites with opposing effects on GABA-A signaling. This reaction, catalyzed by 3β-HSOR, is distinct from the activity of the classical 3βHSD/Δ⁵-Δ⁴ isomerase, which acts on Δ⁵ substrates upstream of 5α-reduction. While the specific brain isoenzyme remains incompletely characterized, AKR1C members display substantial 3β-reductase activity alongside their 3α-activity and are plausible candidates (Steckelbroeck et al., 2004). isoAP can also be generated from AP itself: the bifunctional 3(α→β)-hydroxysteroid epimerase, an NAD(H)-dependent enzyme expressed in brain, oxidizes the 3α-hydroxyl to DHP and then stereoselectively reduces it to isoAP, converting a positive modulator into its own antagonist in two steps (Huang and Luu-The, 2000).

The 3β-epimers antagonize the potentiating actions of their 3α-counterparts. Compounds with this activity are termed GABA-A receptor modulating steroid antagonists (GAMSAs) (Bäckström et al., 2022; Lundgren et al., 2003; Wang et al., 2002). The most well-characterized GAMSA is isoAP, which functions as an AP antagonist at both α1β2γ2L synaptic and α4β3δ extrasynaptic receptor subtypes (Strömberg et al., 2006). Analogous antagonism of THDOC by isoTHDOC has also been reported, although its receptor-subtype specificity is less firmly established (Strömberg et al., 2006). Mechanistically, 3β-hydroxypregnane steroids act not by competition at the potentiation site but as noncompetitive blockers resembling sulfated neurosteroids, a parallel supported by the α1-V256S mutation, which reduces both sulfate and 3β-hydroxysteroid block (Wang et al., 2002). Because the 3β-epimers are generated from the same 5α-reduced precursor as their 3α-counterparts, and can also be formed from those counterparts directly, the 3α/3β balance may be as important as the absolute level of either metabolite in setting GABAergic tone.

20α-reduction: diversion from the neuroactive branch.

Reduction at C20 diverts 5α-reduced pregnanes away from the neurosteroidogenic 3α pathway. AKR1C1, the dominant 20αHSD on pregnane substrates, converts DHP to inactive 20α-hydroxy products lacking GABAergic activity (Usami et al., 2002). Although AKR1C1 and AKR1C2 differ by only seven residues (Dufort et al., 1996), and both are among the most abundant AKR1C isoforms in human brain (Penning et al., 2000), they form predominantly opposing arms of GABAergic regulation: AKR1C1 diverts DHP away from the 3α pathway, whereas AKR1C2 sustains neurosteroid tone by 3α-reduction. This competition is substrate-dependent: the adjacent C21 hydroxyl of DHDOC appears to prevent productive reduction at C20, restricting it to C3 and protecting the THDOC pathway from 20α-diversion (Usami et al., 2002).

3.2. The Androgenic Axes

Testosterone enters the 5α-reduced pathway via reduction of the Δ4 double bond by 5αR, yielding DHT (Bruchovsky and Wilson, 1968a). As with the GABAergic (progestin) axis, this irreversible “frontdoor” reaction makes 5αR the committed entry point: once testosterone is 5α-reduced, its downstream fate is determined by enzymes acting on DHT (Fig. 4). Because DHT is not an aromatase substrate, 5α-reduction also withdraws testosterone from the estrogenic pool generated in the same neurons and channels it toward AR signaling.

Figure 4. Frontdoor androgen metabolism pathway and neuroactive steroid signaling.

In the frontdoor pathway, 5α-reductase (5αR) converts testosterone into dihydrotestosterone (DHT), the principal high-affinity ligand of the androgen receptor (AR). Compared with testosterone, DHT binds AR with greater affinity and dissociates more slowly, resulting in a more sustained transcriptionally active AR state. Subsequent C3 metabolism of DHT generates either 5α-androstane-3α,17β-diol (3α-diol), a positive allosteric modulator of GABA-A receptors with anxiolytic, anticonvulsant, and pain-modulatory effects, or 5α-androstane-3β,17β-diol (3β-diol), an estrogen receptor β (ERβ) agonist that dampens hypothalamic–pituitary–adrenal (HPA) axis stress reactivity. Single-headed arrows indicate irreversible 5αR-mediated reaction, whereas double-headed arrows represent reversible hydroxysteroid dehydrogenase (HSD)-mediated conversions.

For a brain-focused account, the key question is the origin of the testosterone that enters this pathway in the CNS. Most brain testosterone is likely gonadal in origin, with a further contribution from adrenal C19 precursors, and reaches the brain through the circulation. However, the brain also expresses elements of the upstream steroidogenic machinery and may generate a fraction of its androgenic substrate locally. The committed enzyme for de novo androgen synthesis, CYP17A1 (P450c17, 17α-hydroxylase/17,20 desmolase/lyase), is expressed in neurons of the developing CNS (Compagnone et al., 1995) and has been detected in several adult brain regions. The adult rodent hippocampus synthesizes sex steroids, including estradiol, from pregnenolone through P450c17 and aromatase localized in neurons (Hojo et al., 2004), and Cyp17a1 transcripts are present throughout the mesocorticolimbic system, where testosterone concentrations exceed those in blood and remain detectable after gonadectomy, consistent with local synthesis (Tobiansky et al., 2018). Functional evidence reinforces this interpretation. Intracerebroventricular administration of the selective CYP17A1 inhibitor abiraterone reproduces the antidopaminergic effects of systemic treatment on sensorimotor gating (Frau et al., 2014). Systemic abiraterone also enhances prefrontal-dependent behavioral flexibility while increasing tyrosine hydroxylase immunoreactivity in the medial prefrontal cortex (Tomm et al., 2022), implicating brain CYP17A1 in the androgenic modulation of dopaminergic circuits. Whether such activity is widespread in the adult brain nonetheless remains contested, as some studies have not detected the protein or its activity (Le Goascogne et al., 1991; Mellon and Deschepper, 1993).

Taken together, these findings suggest that locally synthesized neuroandrogens can be generated within at least some neural contexts and may influence neurophysiology and behavior partly independently of circulating hormone. To the extent that this local route operates, a portion of the androgenic substrate for 5α-reduction in the brain is of intrinsic rather than peripheral origin.

DHT is the most potent endogenous androgen: A-ring saturation increases AR affinity roughly two-fold, slows receptor-complex dissociation, and stabilizes the AF2 coactivator surface, yielding a 3- to 10-fold greater transactivation potency than testosterone (Askew et al., 2007; Deslypere et al., 1992; Grino et al., 1990). These properties, together with the concentration of 5αR2 in androgen-target tissues (Section 2.4), underlie the DHT-specific roles of 5α-reduction in external genital masculinization, prostate development, and pubertal virilization not supported by testosterone alone (Wilson et al., 1993).

DHT is further reduced at C3 by 3α-HSOR or 3β-HSOR to yield 5α-androstane-3α,17β-diol (3α-diol) or 5α-androstane-3β,17β-diol (3β-diol). As in the progestin axis, the 3α step is reversible: among the AKR1C isoenzymes, only AKR1C2 is capable of regenerating DHT from 3α-diol (Penning et al., 2000), so 3α-diol is a potential reservoir rather than an obligate terminus, and AKR1C2 gates the androgenic and GABAergic outputs of 5α-reduction alike.

3α-diol is a positive allosteric modulator of GABA-A receptors and contributes to anxiolytic, anticonvulsant, pain-modulatory, and cognitive-enhancing effects (Edinger and Frye, 2004; Frye et al., 2008; Reddy, 2004; Reddy and Jian, 2010), and facilitates sexual motivation in females when infused into the NAc shell (Sánchez Montoya et al., 2010). In practice, the reductive bias of the AKR1C isoenzymes in intact cells (Section 3.1) makes 3α-diol the dominant product: its glucuronide conjugate, 3α-androstanediol glucuronide, is a major urinary metabolite of DHT and a peripheral biomarker of 5αR activity, markedly reduced in congenital 5αR2 deficiency (Azzouni et al., 2012; Imperato-McGinley and Zhu, 2002).

3β-diol functions as an agonist of estrogen receptor β (ERβ) (Handa et al., 2008; Lund et al., 2004) and, through this receptor, attenuates HPA axis responses to stress: corticosterone and ACTH elevations following restraint are reduced in an estrogen-receptor-dependent manner (Lund et al., 2006, 2004). This ERβ-mediated mechanism is distinct from the GAMSA activity of isoAP and operates independently of the GABA-A receptor, illustrating that the 3β-reduced arm of 5αR metabolism does not uniformly generate GABA-A antagonists but can instead engage entirely separate signaling pathways. The androgenic axis thus returns to estrogen receptor signaling by an alternative route: 5α-reduction forecloses aromatization, whereas 3β-reduction of DHT leads to ERβ activation.

Alternative entry routes to 5α-reduced androgens.

The frontdoor pathway is not the only route to 5α-reduced androgens. Three alternatives bypass testosterone entirely. Two of them, the backdoor and the 5α-dione pathway, converge on DHT itself (Fig. 5); the third, originating from adrenal 11-oxygenated C19 steroids, yields a chemically distinct set of androgens with potent AR activity (Fig. 6). Across these routes 5αR remains the committed step, but its position within the sequence varies: it acts last on testosterone in the frontdoor pathway, and first, before any androgen has been formed, in the backdoor.

Figure 5. Pathways of androgen metabolism through 5αR: frontdoor, backdoor, and 5α-dione.

5αR catalyzes the irreversible 5α-reduction of multiple Δ⁴-3-ketosteroid androgens, occupying the committed step of androgen activation across three convergent pathways. (i) The frontdoor pathway (red) reduces testosterone directly to dihydrotestosterone (DHT). (ii) The backdoor pathway (blue) bypasses testosterone entirely: 17α-hydroxyprogesterone (17-OHP) is 5α-reduced by 5αR to 5α-pregnan-17α-ol-3,20-dione, which is sequentially reduced by 3αHSD (AKR1C2/AKR1C4) to 5α-pregnane-3α,17α-diol-20-one (Pdiol), cleaved by the 17,20-lyase activity of CYP17A1 to androsterone, reduced by 17βHSD to 5α-androstane-3α,17β-diol (3α-diol), and finally oxidized by 3αHSD acting in reverse to yield DHT. This route is critical during fetal male sexual development and operates predominantly in placenta, fetal liver, and adrenal glands, with 5αR1 as the primary isoenzyme. (iii) The 5α-dione pathway (yellow) 5α-reduces androstenedione to 5α-androstanedione, which is converted to androsterone by 3αHSD, reduced by 17βHSD to 3α-diol, and oxidized to DHT. This pathway is particularly significant in castration-resistant prostate cancer (CRPC), where 5αR1 enables sustained AR activation despite androgen deprivation therapy. The three pathways converge at DHT or its immediate metabolic neighbors but originate from distinct precursors and operate in tissue- and developmental-stage-specific contexts. Single-headed arrows denote irreversible reactions; double-headed arrows denote reactions that are formally reversible.

Figure 6. Biosynthetic pathway of 11-oxygenated androgens from androstenedione.

Androstenedione is hydroxylated at C11 by CYP11B1 (steroid 11β-hydroxylase) to generate 11β-hydroxyandrostenedione (11OHA4). 11OHA4 can either undergo irreversible 5α-reduction by 5α-reductase (5αR) to form 11β-hydroxy-5α-androstanedione or be oxidized by 11β-hydroxysteroid dehydrogenase type 2 (11βHSD2) to produce 11-ketoandrostenedione (11KA4). 11KA4 is subsequently converted to 11-ketotestosterone (11KT) by AKR1C3, while 5αR catalyzes the irreversible conversion of 11KT to 11-ketodihydrotestosterone (11KDHT), the most potent androgen of the 11-oxygenated androgen pathway. Double arrows indicate reversible reactions, whereas single arrows indicate irreversible enzymatic steps.

In the backdoor pathway, 17α-hydroxyprogesterone is 5α-reduced and then processed by AKR1C2/AKR1C4, the 17,20 desmolase/lyase activity of CYP17A1, and 17βHSD to androsterone and 3α-diol, which is finally oxidized to DHT, bypassing testosterone entirely (Auchus, 2004; Fukami et al., 2013; Miller and Auchus, 2019). This route is especially important during fetal male sexual development, where androsterone is the principal circulating androgen during genital masculinization (O’Shaughnessy et al., 2019).

Tissue-specific transcript analyses show that backdoor intermediates are produced mainly in placenta, fetal liver, and adrenal glands rather than in the testes, implicating 5αR1 as the primary isoenzyme driving backdoor androgen synthesis (Fukami et al., 2013; O’Shaughnessy et al., 2019). Genetic defects in backdoor pathway enzymes (AKR1C2, AKR1C4) cause 46,XY DSD with undervirilization, confirming the physiological importance of the route (Flück et al., 2011; Fukami et al., 2013). In the brain, the backdoor pathway may provide an additional source of androgenic and neuroactive metabolites independent of gonadal testosterone.

In the 5α-dione pathway, 5αR reduces androstenedione to 5α-androstanedione, which is then converted directly to DHT by AKR1C3 (Chang et al., 2011; Sharifi, 2013). This pathway is particularly significant in castration-resistant prostate cancer (CRPC), where intratumoral DHT synthesis sustains AR signaling despite gonadal androgen deprivation. Androstenedione, which remains elevated in the castrate environment, serves as substrate for intratumoral 5αR1, the dominant isoenzyme in CRPC tissue (Chang et al., 2011). Whether an analogous intracrine route operates in the brain to augment the local neuroandrogen pool remains unclear.

A third androgenic route originates from adrenal 11-oxygenated C19 steroids (Fig. 6). Through successive reactions involving CYP11B1, 11βHSD2, AKR1C3, and 5α-reduction, the adrenal precursor 11β-hydroxyandrostenedione (11OHA4) is converted to the potent AR agonists 11-ketotestosterone (11KT) and 11-ketodihydrotestosterone (11KDHT), whose transcriptional activities are comparable to testosterone and DHT, respectively (Pretorius et al., 2016; Storbeck et al., 2013; Turcu et al., 2018). While 11OHA4 and 11KA4 are 5α-reduced by both isoenzymes, the conversion of 11KT to 11KDHT is selective for 5αR2 (Barnard et al., 2020; Storbeck et al., 2013). The alternative fate of 11KT is A-ring reduction by AKR1D1, so this steroid stands at the fork between the 5α- and 5β-reduced configurations introduced in Section 2.1: 5αR2 activates it, whereas AKR1D1 in-activates it.

11-oxygenated androgens constitute a significant fraction of circulating androgens in women and children, and elevated levels have been implicated in hyperandrogenic disorders including polycystic ovary syndrome, congenital adrenal hyperplasia (particularly 21-hydroxylase deficiency), and premature adrenarche (O’Reilly et al., 2017; Rege et al., 2018; Turcu et al., 2016). In CRPC, they represent an adrenal-derived reservoir of AR ligands that sustains AR signaling despite androgen deprivation therapy: CRPC tissues express the enzymes needed to convert circulating 11-oxygenated precursors to active androgens, and elevated levels have been detected in patient plasma and tumor tissue (Pretorius et al., 2016). Whether 11KT and 11KDHT are produced or act within CNS circuits, and whether they interact with neurosteroid pathways, remains unexamined, with implications for adrenarchal brain development and stress responsivity.

3.3. The Glucocorticoid Axis

5αR also catalyzes the 5α-reduction of the principal glucocorticoids, cortisol (in humans) and corticosterone (in rodents), to their respective 5α-dihydro metabolites (5α-dihydrocortisol, DHF; and 5α-dihydrocorticosterone, DHB). These products can be further reduced by 3αHSOR to the corresponding tetrahydro metabolites, 5α-tetrahydrocortisol (THF) and 5α-tetrahydrocorticosterone (THB) (Fig. 7). This pathway has traditionally been regarded as a mechanism of irreversible glucocorticoid inactivation. However, 5α-reduced glucocorticoid metabolites retain glucocorticoid receptor (GR) binding and can activate GR-dependent transcription (McInnes et al., 2004). Their signaling profile is nevertheless dissociated: THB weakly induces metabolic genes, including tyrosine aminotransferase and phosphoenolpyruvate carboxykinase, while retaining anti-inflammatory activity in vitro and in vivo at doses that do not produce the adverse effects of conventional glucocorticoids, such as weight loss, hypertension, thymic involution, and hyperinsulinemia (Livingstone et al., 2024; McInnes et al., 2004; Yang et al., 2011). These findings derive from the corticosterone series; whether 5α-reduced cortisol metabolites exhibit a similar dissociation remains unknown.

Figure 7. The glucocorticoid inactivation axis: 5α-reduction of cortisol and corticosterone.

5αR catalyzes the 5α-reduction of cortisol (in humans) and corticosterone (in rodents) to 5α-dihydrocortisol (DHF) and 5α-dihydrocorticosterone (DHB), respectively. These intermediates are further reduced at C3 by 3α-hydroxysteroid dehydrogenase (3αHSD) to yield 5α-tetrahydrocortisol (THF) and 5α-tetrahydrocorticosterone (THB). Although this pathway has traditionally been regarded as a mechanism of irreversible glucocorticoid inactivation, recent evidence indicates that 5α-reduced metabolites, particularly THB, retain partial GR binding affinity and can suppress inflammation without inducing the metabolic toxicity associated with conventional glucocorticoids, suggesting a qualitative shift in signaling profile rather than simple inactivation. Single-headed arrows indicate irreversible reactions; double-headed arrows indicate formally reversible reactions that typically favor reduction under physiological conditions.

Topical THB suppresses inflammation in models of irritant dermatitis with potency approaching that of hydrocortisone, yet chronic administration does not cause skin thinning, adrenal atrophy, or other adverse effects associated with conventional glucocorticoids at equipotent anti-inflammatory doses (Gastaldello et al., 2017; Livingstone et al., 2024). The underlying mechanism remains unclear, and available evidence argues against canonical GR-mediated transrepression. The GR antagonist RU486 attenuates the anti-inflammatory effects of corticosterone but not those of THB in models of dermatitis and angiogenesis (Abernethie et al., 2022; Gastaldello et al., 2017), and THB does not alter the expression of classical GR target genes, including Fkbp51 and Hsd11b1 (Abernethie et al., 2022). Thus, rather than simply inactivating glucocorticoids, 5α-reduction appears to redirect their signaling, preserving anti-inflammatory efficacy while uncoupling it from both the metabolic actions of the parent hormone and, apparently, canonical GR signaling. The receptor or receptors mediating these effects remain unidentified (Nixon et al., 2012).

Although direct evidence from the brain remains limited, the relationships among these substrates suggest a coordinated functional role. DOC is the immediate precursor of corticosterone, and both increase under ACTH stimulation. Thus, 5αR acts on both the precursor and product of the same adrenal biosynthetic step, generating the GABAergic neurosteroid THDOC from DOC (Section 3.1) while redirecting glucocorticoid signaling through corticosterone metabolism. A single enzyme may therefore convert related components of the stress response into two complementary anti-excitatory outputs. Two findings bear directly on this possibility. Lower urinary 5α-reductase activity in trauma-exposed individuals predicted nonresponse to psychotherapy for PTSD and correlated with avoidance-symptom severity (Yehuda et al., 2009), although this measure primarily reflects hepatic rather than central 5α-reduction. In addition, DHB modulates hippocampal long-term potentiation (Dubrovsky et al., 1987, 1985), establishing that 5α-reduced glucocorticoid metabolites are neuroactive, although the responsible receptors were not identified.

Outside the brain, altered glucocorticoid 5α-reduction has been characterized most extensively in metabolic disease, where human and murine findings initially appear discordant. Increased 5αR activity, reflected by elevated urinary 5α-THF:5β-THF ratios, has been associated with visceral obesity, insulin resistance, and hypertension in both sexes (Andrew et al., 1998; Rask et al., 2002, 2001; Upreti et al., 2014). Patients with glucose intolerance likewise show increased excretion of 5α-reduced cortisol metabolites alongside enhanced central and peripheral glucocorticoid sensitivity (Andrews et al., 2002). Conversely, disruption of 5αR1 in mice impairs hepatic clearance of the parent hormone, causing glucocorticoid accumulation, insulin resistance, hepatic steatosis, and relative adrenal insufficiency through enhanced negative feedback on the HPA axis (Livingstone et al., 2014, 2017).

These findings can be reconciled if the increased 5αR activity observed in humans is compensatory rather than causal. This interpretation is supported by the tissue-specific pattern of dysregulation in obesity: hepatic 5αR activity is increased while hepatic 11βHSD1 is reduced, potentially accelerating cortisol clearance and eliciting compensatory HPA-axis activation; by contrast, adipose 11βHSD1 is upregulated, promoting local cortisol regeneration (Livingstone et al., 2000; Nixon et al., 2012; Rask et al., 2002, 2001). The metabolic phenotype of 5αR1-deficient mice is more pronounced in females, a finding interpreted as evidence that it primarily reflects altered glucocorticoid rather than androgen metabolism (Livingstone et al., 2017). However, because hepatic 5αR1 expression is itself female-predominant in rodents (Section 2.3), the sex difference may also reflect a greater absolute loss of enzymatic activity. In critical illness, suppression of cortisol-metabolizing enzymes, including 5αR, contributes to hypercortisolemia and corticotropin suppression (Boonen et al., 2013).

Across these contexts, 5α-reduction is a major route of glucocorticoid clearance, yet its products are biologically active rather than inert. The quantitative importance of this pathway in the brain, and whether it is coordinated with neurosteroid synthesis to generate an integrated stress response, remain important unresolved questions.

3.4. The Chemosensory/Pheromone Axis: 16-Androstene Steroids

A less widely recognized function of 5αR concerns the biosynthesis of the 16-androstene steroids, C19 compounds named for their common Δ¹⁶ double bond, with pheromonal activity in several mammalian species and putative chemosensory functions in humans. The pathway proceeds from pregnenolone through the andien-β-synthase (16-ene synthase) activity of CYP17A1, the classical 3β-HSD/Δ⁵-Δ⁴ isomerase, an obligatory 5αR step, and 3α-HSOR, yielding androstenone and ultimately 3α-androstenol (Fig. 8). Because the andien-β-synthase reaction leaves C17 unsubstituted, these steroids lack the oxygen required for androgen receptor binding and are volatile rather than hormonal. This axis is therefore distinctive in two respects. First, the 5αR substrate androstadienone possesses no cognate nuclear receptor; accordingly, the reaction proceeds within a branch already committed to non-hormonal output, without diverting a ligand from a competing pathway. Second, its products act upon the olfactory receptors of conspecifics, and not upon the intracellular receptors of the organism in which they are formed, thereby extending the reach of 5αR from steroid metabolism into social communication.

Figure 8. The chemosensory/pheromone axis: biosynthesis of 16-androstene steroids.

Pregnenolone is converted by the andien-β-synthase activity of CYP17A1 to 5,16-androstadien-3β-ol (androstadienol), which is oxidized by 3β-hydroxysteroid dehydrogenase (3βHSD) to androsta-4,16-dien-3-one (androstadienone). 5αR then catalyzes the irreversible 5α-reduction of androstadienone to 5α-androst-16-en-3-one (androstenone), the committed step into the active 16-androstene metabolite pool. 3αHSD subsequently reduces androstenone to 5α-androst-16-en-3α-ol (3α-androstenol). Androstenone is an established porcine pheromone; androstadienone and 3α-androstenol have been reported to exert chemosensory and neuroactive effects in humans (see text for details). Single-headed arrows indicate irreversible reactions; double-headed arrows indicate formally reversible reactions.

Where localized directly, the pathway is principally gonadal. In the boar, 16-androstenes are made in testicular Leydig cells, including the obligatory 5α-reduction of androstadienone to androstenone (Meadus et al., 1993; Sinclair et al., 2005), and the human testis generates androsta-5,16-dien-3β-ol as a major metabolite of pregnenolone (Smals and Weusten, 1991; Weusten et al., 1987). Since 16-ene synthase is absent from rat and monkey testis, the axis has been demonstrated only in pig and human, precluding rodent models (Weusten et al., 1990, 1987). Human testis, however, shows no 5α-reductase activity toward the 16-androstenes, so the odorous 5α-reduced steroids must arise extragonadally, with skin proposed given its high 5αR content (Smals and Weusten, 1991).

In the human axilla, however, the reaction appears at least partly microbial: 16-androstenes are scarce in sterile apocrine secretion (Labows et al., 1979), antibacterial washing reduces androstenone output (Bird and Gower, 1982), and isolated axillary bacteria produce both 5α- and 5β-dihydrotestosterone (Nixon et al., 1984); formation of the 5β epimer is incompatible with mammalian 5αR. Host catalysis is nonetheless required elsewhere, since urinary 3α-androstenol shows the sex- and gonadotrophin-dependence of an endogenous product rather than of surface flora (Brooksbank et al., 1972; Brooksbank, 1962).

Androstenone is the best characterized mammalian pheromone, facilitating the standing reflex in estrous sows (Claus, 1979), and its perception in humans varies with polymorphisms in the olfactory receptor gene OR7D4 (Keller et al., 2007; Wysocki and Beauchamp, 1984). Human work has centered on androstadienone, the substrate rather than the product of the 5α step, whose pheromonal status remains contested (Lundström and Olsson, 2005; Wyart et al., 2007). 5α-Reduction thus determines the composition of the emitted 16-androstene profile: because this step occurs peripherally, local 5αR activity sets the ratio of androstadienone to its 5α-reduced products at the site of emission.

The downstream product 3α-androstenol is also a neuroactive steroid, carrying the 3α-hydroxyl, 5α-reduced pharmacophore required for positive allosteric modulation of GABA-A receptors (Lambert et al., 2003; Majewska et al., 1986). It potentiates GABA-activated currents in cerebellar granule cells and in recombinant α1β2γ2 and α2β2γ2 receptors, with an EC50 in the submicromolar to low-micromolar range, and systemic administration elicits anxiolytic-, antidepressant-, and anticonvulsant-like effects in rodents, all stereospecific to the 3α-epimer (Kaminski et al., 2006). Male axillary extracts containing 3α-androstenol likewise modulate luteinizing hormone pulse frequency during the follicular phase, although the active compounds were not identified (Preti et al., 2003).

3.5. Additional pathways

The axes considered so far bear directly or plausibly on the brain. For completeness, we note three additional 5α-reductive functions with little established CNS relevance. 5αR converts aldosterone to the antinatriuretic 5α-dihydroaldosterone, a pathway stimulated by low dietary sodium (Gorsline et al., 1988; Kenyon et al., 1983); as with the glucocorticoids, the 5α-reduced metabolite is not inert but retains antinatriuretic activity, so the reaction redirects rather than terminates mineralocorticoid signaling. Whether it operates in the brain, where mineralocorticoid receptors are abundantly expressed, has not been examined. 5α-Reduction also participates in bile acid biosynthesis, where the AKR1D1-generated 5β-isomers are the obligate precursors of the primary bile acids and the 5α-reduced allo-isomers a minor branch that becomes prominent when 5β-reduction is lost (Russell, 2003). Finally, the stereochemical outcome of A-ring reduction shapes erythropoiesis: 5α-reduced androgens such as DHT stimulate renal erythropoietin, whereas 5β-reduced steroids act directly on erythroid progenitors independently of erythropoietin (Gordon et al., 1970; Shahidi, 1973).

4. 5αR in Neurotransmission and Stress Response

4.1. GABAergic neurotransmission

The GABAergic axis defined in Section 3.1 acts chiefly through modulation of GABA-A receptors and is the best-characterized route by which 5αR activity influences brain function and behavior. AP and THDOC are positive allosteric modulators that potentiate inhibitory currents at nanomolar concentrations and directly gate GABA-A receptors at micromolar concentrations; 3α-diol acts through the same mechanism with lower potency.

Synaptic and extrasynaptic receptors are affected differently, and the distinction is quantitative as well as anatomical. The effect of neurosteroids on GABA-A receptors is conventionally expressed as fractional potentiation, the ratio of the GABA-evoked current in the presence of the steroid to that evoked by the same GABA concentration alone. Extrasynaptic δ-containing receptors, typically assembled as α4βδ or α6βδ, bind GABA with high affinity but gate it with low intrinsic efficacy; neurosteroids relieve this constraint by stabilizing open and bursting states, yielding the largest fractional potentiation among GABA-A receptor subtypes (Wohlfarth et al., 2002). The magnitude of this enhancement depends on the receptor’s gating efficiency: because fractional potentiation is calculated relative to baseline activity, the effect is most pronounced where intrinsic gating is weakest. Consistent with this, raising δ-receptor efficacy experimentally attenuates the enhancement, whereas driving γ2-containing receptors into a low-efficacy mode confers it (Bianchi and Macdonald, 2003). Because high affinity and limited desensitization allow these receptors to respond to ambient GABA, neurosteroids at physiological concentrations selectively enhance the resulting tonic current (Stell et al., 2003), positioning endogenous signaling to regulate network excitability thresholds. At synaptic γ2-containing receptors, by contrast, peak currents evoked by saturating GABA are largely unchanged, whereas AP prolongs the IPSC decay phase by extending burst duration and promoting channel re-openings (Belelli and Lambert, 2005; Bianchi and Macdonald, 2003; Puia et al., 1990).

Conversely, pharmacological 5αR inhibition reduces endogenous AP and produces the opposite effects: sIPSC decay time and total charge transfer fall, and both are restored by exogenous AP (Puia et al., 2003). The same holds behaviorally: the anesthetic and sedative effects of progesterone depend largely on its conversion to AP, with AP levels, rather than progesterone itself, predicting behavioral outcomes (Korneyev and Costa, 1996).

5αR-dependent neurosteroidogenesis exerts region- and circuit-specific effects on inhibitory transmission. In the somatosensory thalamus of mature mice, acute 5αR inhibition reduces the amplitude and frequency of phasic currents as well as tonic inhibition in the ventrobasal nucleus, but in the thalamic reticular nucleus alters only the decay and amplitude of phasic currents, sparing both their frequency and tonic inhibition (Christian, 2020). In the auditory midbrain, 5α-reduced neurosteroids likewise support both stimulus-induced and constitutive inhibitory control: systemic 5αR inhibition blocks habituation of click-evoked potentials in the inferior colliculus and raises their baseline amplitude (Disney and Calford, 2001), and after loss of an inhibitory input, spared circuits mount a rapid compensatory potentiation accompanied by local accumulation of 3α,5α-reduced neurosteroids (Saalmann et al., 2006). A hypothalamic instance of this circuit specificity, in which δ-containing receptors on corticotropin-releasing hormone (CRH) neurons of the paraventricular nucleus (PVN) gate HPA-axis output, is considered in Section 4.3.

Chronic alterations in neurosteroid availability induce adaptive remodeling of GABA-A receptor composition. Sustained reductions in AP, as observed in socially isolated rodents, elicit coordinated changes in GABA-A receptor subunit expression across corticolimbic regions. Specifically, α1, α2, and γ2 subunits are downregulated, while α4 and α5 subunits are upregulated in the frontal cortex and amygdala (Pinna et al., 2006a). Concordantly, the hippocampus shows increased α4 and δ subunits alongside decreased γ2 subunits (Pibiri et al., 2008; Serra et al., 2006). This shift toward extrasynaptic, benzodiazepine-insensitive assemblies produces a distinctive pharmacological phenotype: socially isolated mice show reduced benzodiazepine recognition-site binding and resistance to the sedative effects of diazepam and zolpidem (Pinna et al., 2006a), while the resulting α4/δ-enriched composition confers greater neurosteroid sensitivity (Locci and Pinna, 2017). This dissociation suggests that neurosteroid-based interventions may retain activity precisely where benzodiazepines fail (see Sections 6.1 and 6.4).

The same machinery serves adaptive plasticity across physiological states in which progesterone-derived neurosteroids fluctuate, although the direction of that plasticity depends on timescale: across the ovarian cycle, 5αR inhibition abolishes cycle-dependent changes in tonic, but not phasic, currents in the dorsal motor nucleus of the vagus (Littlejohn et al., 2019) and δ expression tracks progesterone, gating limbic seizure susceptibility (Maguire et al., 2005; Wu et al., 2013). Sustained elevation, by contrast, produces homeostatic downregulation: extrasynaptic receptor function is remodeled across pregnancy and the postpartum period (Maguire and Mody, 2008; Mostallino et al., 2009), mirroring in the opposite direction the remodeling that follows sustained depletion. Such remodeling can invert the behavioral valence of neurosteroid signaling: in the pubertal hippocampus, increased expression of α4β2δ receptors switches the effect of AP from anxiolytic to anxiogenic (Shen et al., 2007), the developmental counterpart of the pharmacologically induced withdrawal state discussed in Section 6.1. This context-dependent reversal shows that the behavioral significance of 5αR-dependent neurosteroid output is not fixed by neurosteroid concentration alone, but is shaped by receptor composition, hormonal history, and circuit state.

4.2. Other neurotransmitter systems

The effects of 5αR-dependent neurosteroidogenesis on neurotransmitter systems other than GABA are less well understood. The best-characterized non-GABAergic target is the neuronal nicotinic acetylcholine receptor (nAChR). In mouse synaptosomal preparations, both AP and DHP act as noncompetitive allosteric inhibitors of nicotinic responses, reducing agonist-evoked ⁸⁶Rb⁺ efflux in thalamus (attributable primarily to α4β2-type receptors) as well as [³H]dopamine release from striatal terminals, originally ascribed to an α3-containing receptor but subsequently reassigned to α6β2*-containing subtypes (Bullock et al., 1997; Champtiaux et al., 2003). Inhibition is slow in onset and slowly reversible and occurs without changes in the rate or equilibrium of [³H] nicotine binding, consistent with an allosteric rather than orthosteric mechanism (Bullock et al., 1997). The subtype range of this action, however, depends on A-ring reduction.

It is worth noting that the available evidence provides only limited support for two actions frequently attributed to 5α-reduced steroids: direct modulation of NMDA receptors and inhibition of 5-HT₃ receptors. Indeed, NMDA receptor inhibition is primarily a property of the sulfated 5β-pregnane pregnanolone sulfate (Park-Chung et al., 1994; Petrović et al., 2005), while 5-HT₃ antagonism by AP (Wetzel et al., 1998) requires supraphysiological concentrations and is shared with structurally unrelated steroids (Oz et al., 2002).

A separate and more extensively characterized line of evidence concerns the role of 5αR in dopaminergic regulation, arguably the most translationally important of the downstream consequences of neurosteroid signaling. The mechanisms involved extend beyond canonical GABA-A modulation and remain incompletely defined. Finasteride and other 5αRIs, including dutasteride and SKF-105111, prevent prepulse inhibition (PPI) deficits and hyperlocomotion induced by dopaminergic agonists such as apomorphine and amphetamine, with efficacy comparable to haloperidol and clozapine but without catalepsy (Bortolato et al., 2008). Subsequent anatomical studies localized these effects principally to the NAc: intra-accumbal 5αR inhibition (shell and core), like intracerebroventricular administration, restored PPI, whereas equivalent infusions into dorsal caudate, basolateral amygdala, ventral hippocampus, and medial PFC were ineffective, although a statistical trend emerged following medial PFC infusion (Devoto et al., 2012). Notably, dopamine efflux in the NAc was increased by systemic but not intra-accumbal finasteride, indicating that the gating-ameliorating effect rests on postsynaptic mechanisms rather than on altered dopamine release (Devoto et al., 2012). Receptor-specific analyses across rodent species and strains indicate preferential involvement of D₁-mediated signaling: 5αR inhibition attenuates behavioral effects driven by D₁ activation while leaving D₂-mediated responses intact, both in C57BL/6 mice (Frau et al., 2013) and in Long-Evans rats (Frau et al., 2016). In Sprague-Dawley rats, finasteride additionally countered the reduction in percent PPI induced by D₃, but not D₂, stimulation, although this effect was not observed in absolute PPI values (Frau et al., 2016). Direct pharmacological evidence for the causal involvement of 5α-reduced steroids comes from studies showing that AP is required for D₁-induced PPI deficits (Mosher et al., 2019), and that subthreshold doses of intra-medial PFC AP and a D₁ receptor agonist synergistically disrupt PPI in Sprague-Dawley rats, a strain otherwise relatively insensitive to D₁ agonists (Frau et al., 2023). AP is thus necessary for D1-driven gating deficits and, at subthreshold levels, sufficient to lower the threshold at which D₁ activation becomes disruptive. Because acute stress raises prefrontal AP, and intra-medial PFC AP infusions are sufficient to reproduce the gating deficits caused by stressors (Cadeddu et al., 2022), these findings support a model in which stress-induced increases in neurosteroid levels modulate D₁-dependent cortical signaling, thereby influencing cognitive and sensorimotor processes.

The mechanisms linking D₁ receptors and AP remain unresolved, but the relation is causal rather than correlative: beyond the necessity and sufficiency data above, a strain concordance is hard to explain otherwise. The strains susceptible to D₁ agonist-induced PPI deficits, C57BL/6 mice and Long-Evans rats (Mosher et al., 2016; Ralph-Williams et al., 2003), are precisely those in which systemic AP reduces PPI, whereas Sprague-Dawley rats are resistant to both manipulations when either is applied alone (Cadeddu et al., 2022). Irrespective of the molecular process, this interaction carries therapeutic implications. Given that 5αRIs oppose dopaminergic activation without inducing catalepsy (Bortolato et al., 2008), the 5αR-neurosteroid pathway may offer a means of modulating D₁-dependent signaling without extrapyramidal liability. This possibility, however, remains untested: the available evidence pertains exclusively to sensorimotor gating, and no study has yet examined whether 5αR inhibition modifies D₁-dependent cognitive performance.

D₃ receptors figure in a second, mechanistically separate context, in which the readout is impulse control rather than sensorimotor gating and the evidence is anatomical and pharmacological. In rats pretreated with low-dose reserpine (a regimen unmasking otherwise undetectable effects) the D₃-preferring agonist pramipexole increases the propensity for highly disadvantageous probabilistic choices, an effect paralleled by a selective upregulation of D₃, but not D₂, receptor membrane expression in the NAc (Orrù et al., 2020; Pes et al., 2017). Finasteride (25 mg/kg) partially countered this behavioral alteration (namely, the only significant effect was detected at the most disadvantageous winning-probability block) and reversed the accumbal D₃ upregulation, without affecting D₂ levels or either receptor in the caudate-putamen (Floris et al., 2022). Several caveats temper this interpretation: finasteride intrinsically reduced probability discrimination, and a higher dose (50 mg/kg) elicited an amotivational response consistent with this drug’s depressogenic liability; furthermore, neither D₂ nor D₃ antagonism reverses the effects of pramipexole in this paradigm (Orrù et al., 2020), suggesting that D₃ normalization may be a correlate rather than the mediator of finasteride’s action. These findings nevertheless supply a plausible mechanistic substrate for the earlier clinical observation that finasteride attenuated pathological gambling in two PD patients treated with pramipexole (Bortolato et al., 2012).

The relevance of 5αR-dependent neurosteroidogenesis to dopaminergic regulation is further underscored by studies using isolation rearing (IR), a neurodevelopmental model of chronic psychosocial stress that produces schizophrenia-related behavioral and neurochemical alterations, including PPI deficits (Geyer et al., 1993; Wilkinson et al., 1994). In Sprague-Dawley rats, IR from weaning through adulthood produces a marked reduction in the protein expression of both 5αR1 and 5αR2 in the NAc and medial PFC, accompanied by decreases in cortical AP (−60%) and THDOC (−47%) levels; plasma AP was not significantly affected, leading the authors to conclude that the central decrement reflects reduced brain neurosteroidogenesis rather than diminished HPA output, although an earlier report found AP reduced in both brain and plasma of IR rats (Bortolato et al., 2011; Serra et al., 2000). These changes are accompanied by elevated extracellular dopamine and DOPAC in the NAc shell, but not in the medial PFC. Finasteride’s effects on dopamine efflux are regionally dissociated in a manner that maps onto this pattern: the drug increased accumbal dopamine in socially reared but not IR rats, whereas it elicited equivalent increases in prefrontal dopamine and in accumbal DOPAC in both groups (Bortolato et al., 2011). Rather than a global loss of 5αR function, this dissociation suggests that IR selectively alters GABA–dopamine coupling in the NAc but not the medial PFC. This is consistent with the circuit-dependent localization of 5αR1 to principal output neurons, which are glutamatergic in cortex but GABAergic in the striatal complex (Agís-Balboa et al., 2006). Notably, neurosteroid levels could not be quantified in the NAc itself, so the link between accumbal 5αR downregulation and local AP depletion remains inferential. Despite this downregulation, finasteride (25–100 mg/kg) dose-dependently reverses IR-induced PPI deficits, without affecting gating integrity in socially reared controls (Frau et al., 2015); across dopaminergic models, this therapeutic profile occurs without catalepsy in either the bar or paw test (Bortolato et al., 2008).

Complementary NAc proteomics in IR rats (Roncada et al., 2009) and finasteride-treated unstressed animals flagged further candidate targets, most consistently a reduction in GABA transaminase that could influence local GABA availability and inhibitory tone (Soggiu et al., 2016). These links remain untested, however, and both datasets rest on small samples (n = 4-5 per group). Taken together, these findings expose a paradox: acute stress transiently elevates 5αR expression and neurosteroid content and disrupts gating, whereas persistent inescapable stress progressively downregulates them and disrupts gating as well; yet 5αR inhibition is therapeutic in both states (Bortolato et al., 2011), arguing against a simple monotonic relation between AP tone and gating integrity.

4.3. Regulation of the stress response by 5αR-derived neurosteroids

5αR-derived neurosteroids constitute a rapid, largely membrane-initiated buffering system that converges with the glucocorticoid inactivation pathway described in Section 3.3 and operates faster than classical genomic feedback. As noted in Section 3.1, acute stressors elevate AP and THDOC in plasma and brain, with increases detectable within ~10 min and peaking near 30 min; the magnitude ranges from two- to several-fold and varies by steroid, brain region, sex, and stressor (Barbaccia et al., 1996; Purdy et al., 1991). This response depends substantially on peripheral steroidogenesis: adrenalectomy largely abolishes it, reflecting stress-induced adrenal output of progesterone and DOC (the respective precursors of AP and THDOC), although the relative contributions of peripheral 5α-reduction with subsequent CNS uptake versus central conversion of peripherally derived precursors remain incompletely resolved. In humans, acute-stress neurosteroid responses are considerably less consistent (Crowley and Girdler, 2014).

Under basal conditions, 5α-reduced neurosteroids enhance GABAergic inhibition of PVN CRH neurons and restrain HPA-axis activation. THDOC selectively potentiates the δ-mediated tonic current of these neurons and reduces their firing, both effects abolished in Gabrd^−/− mice, and intra-PVN THDOC lowers circulating corticosterone in unstressed animals (Sarkar et al., 2011). Convergent pharmacology supports this restraint: AP, THDOC, or progesterone attenuate stress-induced ACTH and corticosterone release (Owens et al., 1992; Patchev et al., 1996), AP reduces Crh mRNA expression in the PVN (Patchev et al., 1994), and immunoneutralization of AP enhances the corticosterone response to acute stress (Guo et al., 1995).

Under acute challenge, this relationship inverts. After restraint stress, THDOC instead increases CRH neuron firing, and intra-PVN infusion exacerbates the corticosterone response (Sarkar et al., 2011). The reversal follows collapse of the chloride gradient in CRH neurons, traced to stress-induced dephosphorylation of KCC2 at Ser940 and loss of surface KCC2, such that GABA-A responses become depolarizing and GABA-A blockade lowers rather than raises firing (Hewitt et al., 2009; Sarkar et al., 2011). δ-containing receptors are required: the corticosterone response is blunted in Gabrd^−/− mice, finasteride prevents both the corticosterone elevation and the ensuing anxiety-like behavior, and KCC2 Ser940 phosphorylation tracks corticosterone across animals (Sarkar et al., 2011). The sign of 5αR-derived neurosteroid action at the PVN is therefore set by the chloride homeostatic state of the target neuron. This anxiolytic-like effect of 5αR inhibition under acute stress contrasts with the anxiogenic consequences of AP withdrawal after sustained blockade (Section 6.1), providing another example of the divergence between the acute and chronic effects discussed below.

Beyond this GABAergic mechanism, 5α-reduction gates HPA output through additional routes. In late pregnancy, AP suppresses the HPA and oxytocin responses to immune challenge via an opioid-linked, non-GABAergic pathway (Brunton et al., 2012, 2009). A third route is androgenic. In gonadectomized males, testosterone suppresses the ACTH and corticosterone responses to stress, but this effect is blocked by finasteride given into the third ventricle. The equivalent effect of DHT is unaffected by finasteride. Because testosterone requires central 5α-reduction to act while DHT does not, brain 5α-reduction appears to be necessary for androgenic inhibition of the HPA axis (Handa et al., 2013). Downstream transduction of this arm may involve further reduction of DHT to 3β-diol acting at ERβ-expressing hypothalamic neurons (Lund et al., 2006), extending the gatekeeping role of 5αR beyond GABAergic neurosteroidogenesis. A caveat applies throughout: finasteride blocks the 5α-reduction of progesterone, DOC, testosterone, and glucocorticoids concurrently and without isoform selectivity, so attribution of these effects to loss of a single metabolite remains uncertain.

Chronic stress produces a different phenotype. As described in Section 4.1, sustained reductions in corticolimbic neurosteroids accompany GABA-A subunit remodeling toward extrasynaptic, benzodiazepine-insensitive but neurosteroid-hypersensitive configurations, in which the behavioral effects of AP depletion are amplified and rescued by restoring neurosteroid tone (Locci and Pinna, 2017). Clinically, GABAergic 3α-reduced steroids (AP plus pregnanolone) are reduced in major depressive disorder, where fluoxetine or fluvoxamine normalizes CSF levels in correlation with symptom improvement (Uzunova et al., 1998), and in premenopausal women with PTSD (Rasmusson et al., 2006). Notably, in both cohorts the immediate 5αR product DHP was unchanged, localizing the clinical block to 3αHSD.

A related but mechanistically distinct disruption occurs in cortical circuits. Prolonged sleep deprivation (72 hours in rats, 24 hours in mice) selectively elevates 5αR expression, 5αR activity, and AP levels in the PFC while impairing sensorimotor gating (Cadeddu et al., 2022; Frau et al., 2017a, 2008). In parallel, sleep deprivation reduces membrane trafficking of KCC2 via Thr1007 phosphorylation (rather than the Ser940 dephosphorylation seen in the PVN), driving intracellular chloride from ~14 to ~36 mM in medial PFC pyramidal neurons and depolarizing E_GABA; this arm is BDNF–TrkB-dependent (Frau et al., 2026). The two pathways dissociate pharmacologically: finasteride rescues gating without normalizing chloride, and neither AP nor KCC2 blockade alone impairs PPI, whereas their combination reproduces the deficit (Frau et al., 2026). The cortical case therefore represents a conductance-by-driving-force interaction: neurosteroid potentiation of an already-depolarized GABA-A current degrades inhibitory precision rather than strengthening inhibition. Together with the PVN findings, these results indicate that the sign and functional consequence of neurosteroid action are set by the prevailing chloride gradient, departing from the classical view of neurosteroids as uniformly anxiolytic.

5. 5αR Inhibitors (5αRIs): Pharmacology and Toxicology

5.1. The pharmacological landscape of 5αRIs

As discussed in Section 1, the clinical pharmacology of 5αRIs is primarily defined by finasteride and dutasteride, two 4-azasteroid compounds approved for therapeutic use (Fig. 9). Finasteride, a synthetic 4-azasteroid structurally derived from testosterone, was developed for its capacity to inhibit 5αR (George et al., 1989; Rittmaster et al., 1989). It received approval from the U.S. Food and Drug Administration (FDA) in 1992 for the treatment of BPH at a dose of 5 mg/day, and subsequently in 1997 for AGA at 1 mg/day (Aggarwal et al., 2010; Kaufman et al., 1998; Lowe et al., 2003; Nickel, 1998). Dutasteride, a dual inhibitor targeting both 5αR isoenzymes (Bramson et al., 1997), was approved by the FDA in 2001 for BPH. Despite its substantial effectiveness in AGA, this drug received formal approval for this indication only in a few countries (Japan, South Korea); however, it is used off label for this purpose elsewhere (Arif et al., 2017). In addition to BPH and AGA, both finasteride and dutasteride were also evaluated for prostate cancer chemoprevention. Nevertheless, despite positive results in large randomized trials (Andriole et al., 2010; Thompson et al., 2003), regulatory approval was not granted, owing to concerns about an increased long-term incidence of high-grade tumors. For finasteride, this signal has been largely attributed to detection bias: reduced prostate volume increases biopsy sensitivity in treated men. This interpretation is supported by long-term follow-up showing no excess mortality (Goodman et al., 2019; Redman et al., 2008).

Figure 9.

Chemical structures of representative 5α-reductase inhibitors (5αRIs) classified according to isoenzyme selectivity. Compounds are grouped as dual 5αR1/5αR2 inhibitors, selective 5αR1 inhibitors, or selective 5αR2 inhibitors. Finasteride and dutasteride, the two clinically approved 4-azasteroid inhibitors, are included among the 5αR2-selective and dual compounds, respectively, together with experimental inhibitors developed to target specific 5αR isoforms. Structural diversity across classes includes steroidal and nonsteroidal scaffolds, reflecting distinct mechanisms of enzyme binding and selectivity.

Both finasteride and dutasteride act as mechanism-based (“suicide-substrate”) inhibitors. Consistent with the ordered bi-bi catalytic mechanism described in Section 2.1, the enzyme catalyzes reduction of the inhibitor using NADPH as a cofactor, generating a highly stable NADP-dihydroinhibitor adduct that dissociates extremely slowly (apparent Ki ≤ 1 × 10⁻¹³ M for finasteride), effectively rendering the enzyme catalytically inactive (Bull et al., 1996; Han et al., 2021). Consequently, restoration of enzymatic activity depends on de novo protein synthesis rather than ligand dissociation. This pharmacodynamic property has important implications for persistent central effects: even after systemic clearance of the drug, 5αR activity in the brain may remain suppressed until sufficient enzyme is newly synthesized and properly localized to ER membranes, a prediction consistent with the isoenzyme-specific recovery kinetics described below.

The pharmacokinetic properties of finasteride and dutasteride differ: whereas finasteride has a short half-life of 5-6 hours, dutasteride exhibits a prolonged terminal half-life of approximately five weeks, leading to sustained pharmacodynamic effects that can persist for months after discontinuation. Their selectivity also diverges. In humans, finasteride is markedly more selective for 5αR2 than for 5αR1: in prostatic tissue from finasteride-treated patients, type 2 enzymatic activity is suppressed roughly 100-fold, whereas type 1 activity is reduced only about 3-fold and may continue to contribute to DHT production (Span et al., 1999). Dutasteride, by contrast, inhibits both isoenzymes potently (IC₅₀ ≈ 7 nM and 6 nM for types 1 and 2, respectively; Bramson et al., 1997). Both compounds also potently inhibit 5αR3 (Yamana et al., 2010), although the physiological relevance of this activity is uncertain, given that this enzyme functions primarily as a polyprenol reductase (see Section 2). These pharmacological differences translate into distinct profiles. Finasteride (5 mg/day) reduces serum DHT levels by approximately 65% and lowers intraprostatic DHT to less than 15% of control levels (McConnell et al., 1992), whereas dutasteride achieves near-complete suppression of serum DHT (approximately 95% at the 0.5 mg clinical dose; Clark et al., 2004) and reduces intraprostatic DHT by more than 90% (Gleave et al., 2006).

It should be noted that the 5αR2 selectivity of finasteride is not universal across species. Rat 5αR1 is roughly 100-fold more sensitive to finasteride than its human counterpart (Ki ≈ 3–5 nM vs. ≥300 nM), a difference mapped by chimeric-enzyme analysis to a four-residue segment inferred to contribute to the substrate-binding domain (Thigpen and Russell, 1992). Because finasteride is a potent inhibitor of both isoenzymes in rats, preclinical studies in this species effectively assess dual inhibition rather than the 5αR2-selective profile observed in humans. In line with a dual mechanism, subchronic finasteride treatment in rats reduced both 5αR1 and 5αR2 protein levels in the brain, accompanied by decreased Fos immunoreactivity and a non-significant reduction in Fos phosphorylation. Four weeks after discontinuation, 5αR1 expression, Fos phosphorylation, and serum DHT had all recovered, whereas 5αR2 remained persistently suppressed (Choi et al., 2025).

An incompletely addressed question concerns the possibility that finasteride and dutasteride may exert effects that are not fully accounted for by inhibition of 5αRs. Finasteride has been shown to inhibit human steroid 5β-reductase (AKR1D1), acting as a competitive inhibitor with low micromolar affinity rather than as a mechanism-based inactivator. This potency is several orders of magnitude weaker than its subnanomolar functional inactivation of 5αR2, and finasteride is therefore unlikely to engage this enzyme substantially at therapeutic exposures (Drury et al., 2009). More recently, proteome-wide in silico screening, validated in vitro and in rats, identified phenylethanolamine N-methyltransferase (PNMT), the enzyme catalyzing the terminal step in epinephrine biosynthesis, as an off-target of finasteride (Giatti et al., 2021), a finding of potential relevance to the autonomic and affective symptoms discussed in Section 5.3. Dutasteride has been proposed for repositioning as an anti-neuroinflammatory agent: in lipopolysaccharide (LPS)-stimulated BV2 microglial cells it reduced IL-6 and TNF-α secretion, and in an LPS-challenged mouse model it lowered hippocampal and plasma IL-6, attenuated microglial activation, and mitigated cognitive deficits (Luo et al., 2021). This anti-inflammatory profile, however, does not necessarily indicate a 5αR-independent mechanism. As the authors themselves acknowledge, by suppressing the conversion of testosterone and progesterone to DHT and DHP, 5αR inhibition redistributes substrate flux within target tissues, increasing local availability of these precursors and, downstream of testosterone, of 17β-estradiol (Favilla et al., 2021; Litim et al., 2015; Olsen et al., 2006), which restrain microglial activation and exert anti-inflammatory and neuroprotective actions (Barreto et al., 2007). The anti-inflammatory effect of dutasteride may therefore represent an indirect, on-target consequence of a shift in steroid flux rather than evidence of a discrete additional target.

Other neuroprotective actions of 4-azasteroids are more difficult to reconcile with 5αR inhibition and point to possible off-target mechanisms that converge on mitochondria. In a chemical-proteomic reprofiling study, affinity probes derived from 4-azasteroids captured adenine nucleotide translocator-1 (ANT-1), a regulator of the mitochondrial permeability transition; consistent with engagement of this target, dutasteride inhibited pore opening and increased autophagosomal structures in human cell lines, and reduced β-amyloid plaque burden in a model of cerebral amyloidosis (Soskić et al., 2008). Independently, an unbiased screen of 570 compounds in primary cortical neurons, designed to isolate agents acting downstream of NMDA receptor signaling, identified both dutasteride and finasteride as neuroprotective against NMDA-induced excitotoxicity, with dutasteride attenuating the associated collapse of mitochondrial membrane potential (Merz et al., 2020). While neither study establishes that this neuroprotection is ANT-1-mediated, the findings collectively suggest that, beyond their canonical enzymatic target, 4-azasteroids may stabilize mitochondrial function under excitotoxic and proteotoxic stress, a property whose therapeutic and toxicological implications warrant further investigation.

Although no alternative compounds have achieved clinical use comparable to finasteride and dutasteride, the pharmacological landscape of 5αRIs includes other agents. Selective 5αR1 inhibitors such as the 4-azasteroid MK-386 (Ellsworth et al., 1996; Schwartz et al., 1997) and the nonsteroidal benzoquinolinone bexlosteride (LY300502), the active enantiomer of LY191704 (Hirsch et al., 1993), showed acceptable safety but limited efficacy; consequently, these agents did not advance for their intended clinical applications (acne and BPH, respectively). Among 5αR2-selective compounds, epristeride is notable for its distinct mechanism, acting as an uncompetitive inhibitor that binds the enzyme-NADP⁺ complex (Levy et al., 1994); it is currently approved for BPH only in China (Cerra and Gioiello, 2025). Several additional 4-azasteroids that inhibit mainly 5αR2, including MK-434 and MK-963, have been described but did not reach clinical use (Azzouni et al., 2012; Cerra and Gioiello, 2025). Among dual inhibitors, the 4-azasteroid 4-MA is a potent inhibitor of both isoenzymes but was withdrawn from development because of hepatotoxicity and off-target inhibition of 3βHSD (Azzouni et al., 2012).

Numerous herbal and botanical products inhibit 5αRs in vitro, including phytosterols, flavonoids such as quercetin and myricetin, epigallocatechin gallate, and polyunsaturated fatty acids, though their potency, selectivity, and clinical relevance remain poorly characterized (Azizi et al., 2021). Screening studies have likewise identified 5αR1 as a target of diverse environmental contaminants (Guo et al., 2017; Li et al., 2024; Zhu et al., 2024), effects documented only in vitro and of unestablished significance at realistic exposures.

5.2. Adverse effects of 5αRIs during treatment

Sexual dysfunction is the most consistently reported adverse effect associated with 5αRIs, with pooled analyses indicating modest absolute risk increases of approximately 1-3% over placebo for decreased libido, erectile dysfunction (ED), and ejaculatory disturbances (Trost et al., 2013). This association is strongly supported by multiple randomized controlled trials (Andriole et al., 2010; Fertig et al., 2017; Kaufman et al., 1998; McConnell et al., 1992; Moinpour et al., 2007; Nickel et al., 1996; Roehrborn et al., 2011; Wessells et al., 2003), with ejaculatory dysfunction occurring in approximately 4% of treated patients during long-term therapy, more frequently with dutasteride than with finasteride (Herberts et al., 2016; Kaplan et al., 2012). Reproductive effects include reversible reductions in sperm count, concentration, semen volume, and motility (Amory et al., 2007; Carreño-Orellana et al., 2016), as well as rare cases of azoospermia (Glina et al., 2004). In pharmacovigilance databases, finasteride is the most frequently reported drug in association with male-factor infertility (Baldini et al., 2023). Although the dose used for AGA (1 mg/day) does not appear to impair spermatogenesis in healthy men (Overstreet et al., 1999), adverse outcomes have been documented in men with pre-existing subfertility (Liu et al., 2008; Ricci et al., 2012).

Several lines of evidence, however, qualify the magnitude of this signal. In a large Japanese post-marketing surveillance cohort, spontaneously reported adverse events occurred in 0.7% of cases (Sato and Takeda, 2012), a figure that likely reflects the limited sensitivity of passive reporting (see below). In long-term treatment for BPH, incident sexual adverse events are concentrated in the first year and the difference from placebo diminishes thereafter, with the PLESS (Proscar Long-term Efficacy and Safety Study) trial showing convergence to an incidence of approximately 7% in both the treatment and placebo arms between years 2 and 4 (Tacklind et al., 2010; Wessells et al., 2003). Increased risk is more evident in men treated for BPH than for AGA (Liu et al., 2016), although this contrast is confounded by the older age, greater comorbidity, and higher dose of the BPH population; nonetheless, an increased risk of ED has also been documented in the AGA setting (Mella et al., 2010). Gynecomastia occurs more frequently in treated than placebo groups and more with dutasteride than finasteride (Kaplan et al., 2012; Thompson et al., 2003), while isolated case reports of male breast carcinoma (Lee and Ellis, 2004; Shenoy and Prabhakar, 2010) have not been borne out as a causal association in large trials and cohort studies (Bird et al., 2013; Fertig et al., 2017; Kjærulff et al., 2019; McConnell et al., 2003; Thompson et al., 2003).

Systemic effects span multiple organ systems. Reversible hepatotoxicity (Martínez de Guzmán and Martínez-Crespo, 2006), insulin resistance and increased incidence of type 2 diabetes (Upreti et al., 2014; Wei et al., 2019), and an increased risk of osteoporosis diagnosis (Lin et al., 2015) have all been reported, with dutasteride increasing body fat (Carreño-Orellana et al., 2016; Upreti et al., 2014). Cardiovascular findings are heterogeneous: an unexpected post hoc rise in a composite cardiac-failure category accompanied dutasteride in one chemoprevention trial without any difference in overall cardiovascular events or deaths (Andriole et al., 2010); separately, exposure to 5αRIs has been linked to an elevated risk of acute coronary syndrome in an observational cohort (Chou et al., 2015). Conversely, preclinical studies suggest potential cardioprotective effects (Zwadlo et al., 2015).

Finasteride and dutasteride were classified as FDA pregnancy category X owing to their teratogenic potential, particularly the risk of ambiguous genitalia in male fetuses (BinJadeed and Alajlan, 2021; Lechuga Sancho et al., 2004; Sallout and Al Wadi, 2009; Teichert et al., 2017); although the letter categories have since been superseded, both agents remain contraindicated in pregnancy. The safety profile of finasteride in women more broadly remains poorly characterized (Hirshburg et al., 2016; Seale et al., 2016).

Two additional mechanisms warrant consideration for their potential relevance to CNS dysfunction. First, 5αR inhibition impairs myelination in cultured cerebellar slices in a manner reversible by AP (Ghoumari et al., 2003), suggesting that sustained inhibition may compromise myelination. Second, rare thrombotic events have been described, including stroke in a young man receiving finasteride 1 mg (Tsuji et al., 2014), potentially mediated by increased estrone and estradiol levels; a similar mechanism may underlie the case of cerebral venous sinus thrombosis reported with dutasteride (Choi et al., 2020).

Psychiatric outcomes were not recognized as adverse effects in early clinical trials of 5αRIs; however, accumulating evidence has since identified a range of neuropsychiatric consequences. The first report of such sequelae came from the observation of moderate-to-severe depression within 9-19 weeks of finasteride treatment at 1 mg/day, with complete resolution upon discontinuation; two individuals who underwent rechallenge experienced recurrence within two weeks, providing suggestive evidence of causality (Altomare and Capella, 2002). A subsequent study with 128 men treated with finasteride for two months found statistically significant, albeit clinically modest, increases in depression scores; notably, none of the patients reported depressive symptoms spontaneously, indicating that passive reporting likely underestimates these effects (Rahimi-Ardabili et al., 2006). Large-scale observational studies further support this association: a Canadian cohort linked 5αRI exposure to increased self-harm and depression, with self-harm risk confined to the first 18 months of exposure and depression risk persisting beyond that interval, although no increase in completed suicide was detected (Welk et al., 2017), and a French national database of over 200,000 individuals corroborated associations with depression and suicidal ideation (Laanani et al., 2023). Pharmacovigilance analyses and meta-analyses converge on the same signal, with increased reporting of depression, anxiety, suicidal ideation, and completed suicide among 5αRI users (Deng et al., 2020; Nguyen et al., 2021; Pompili et al., 2021). However, disproportionality measures are susceptible to notoriety bias, as heightened regulatory and media attention can itself increase adverse event reporting. Regulatory responses have evolved accordingly, including FDA label modifications in 2011-2012, a warning issued by the UK Medicines and Healthcare products Regulatory Agency (MHRA) in 2023, and a 2025 decision by the European Medicines Agency (EMA) Pharmacovigilance Risk Assessment Committee (PRAC) requiring updated labeling and patient information highlighting risks of depressed mood and suicidality.

The most plausible mechanism for these depressive symptoms is a reduction in AP levels. In line with this interpretation, depressed individuals exhibit reduced CSF and serum AP levels, which normalize with antidepressant treatment (Uzunova et al., 2006, 1998). Consistent with this, postmortem studies demonstrate decreased 5αR1 mRNA expression in Brodmann area 9 of depressed subjects (Agis-Balboa et al., 2014). These observations establish an association between deficient AP signaling and depression rather than demonstrating that pharmacological AP depletion is depressogenic; the causal direction is supported instead by the preclinical evidence discussed below. Additionally, preliminary evidence points to other mechanisms that may contribute to the depressogenic effects of 5αRIs, including suppression of hippocampal neurogenesis (Römer et al., 2010), neuroinflammation, and alterations in gut microbiota composition (Diviccaro et al., 2019).

Several factors appear to modulate individual vulnerability to these adverse outcomes, including baseline levels of androsterone sulfate and epiandrosterone sulfate (Surendran et al., 2022), baseline androgenic status (Marberger, 1998; Motofei et al., 2020), age of AGA onset (Rossi et al., 2011), and other genetic or population differences (Kaufman et al., 1998; Sato and Takeda, 2012). The contribution of nocebo effects has also been examined. Informing patients about potential sexual side effects increased their reporting rates from 15.3% to 43.6% (Mondaini et al., 2007). Nevertheless, a purely nocebo-based explanation is inconsistent with objective neurosteroid alterations, preclinical evidence, and epidemiological data demonstrating significant associations with sexual dysfunction after controlling for comorbidities (Lauck et al., 2024). In animal models, repeated finasteride administration produces time-dependent depression-like behavior, with increased immobility in the forced swim test emerging after several days (Sasibhushana et al., 2019), alongside anxiety-like phenotypes, impaired hippocampal-dependent learning, and reduced long-term potentiation (Nayana et al., 2025; Sasibhushana et al., 2024).

5.3. Persistent adverse effects after 5αR inhibitor discontinuation

Beginning in 2011, a series of reports documented that adverse effects may persist for months or years after discontinuation of finasteride or dutasteride, particularly in young men treated for AGA (Ali et al., 2015; Cecchin et al., 2014; Di Loreto et al., 2014; Irwig, 2015, 2012b, 2012a; Irwig and Kolukula, 2011; La Marra et al., 2012; Traish et al., 2015, 2011), although several of these cohorts were recruited through patient advocacy forums and are therefore subject to selection bias. These observations led to the formulation of post-finasteride syndrome (PFS), a clinical entity encompassing sexual, somatic, and neuropsychiatric symptoms that endure despite drug cessation. Consistent with this tripartite structure, a pharmacovigilance analysis of FAERS (FDA Adverse Event Reporting System) data grouped the reported adverse events into sexual, physical, and psychological domains and identified them at both the 1 mg (alopecia) and 5 mg (BPH) doses (Baas et al., 2018). Although PFS was catalogued by the US NIH Genetic and Rare Diseases Information Center in 2015, this listing conferred no formal disease status, and the syndrome remains insufficiently addressed in dermatological, urological, and psychiatric guidelines. Proposed diagnostic criteria for PFS include prior exposure to a 5αRI, sexual dysfunction that persists for at least three months after discontinuation, and the absence of pre-existing sexual dysfunction or of any current or prior medical condition, medication, or substance use that could otherwise account for the symptoms; cognitive impairment, depression, and suicidality are recognized as associated, non-diagnostic features that may arise independently of the sexual symptoms (Healy et al., 2022).

Persistent peripheral manifestations have been characterized. In a retrospective cohort of 79 men aged 22-50, Chiriacò and colleagues reported high rates of loss of penile sensitivity, reduced ejaculatory force, and decreased ejaculate volume, accompanied by markedly elevated sexual dysfunction scores relative to retrospectively recalled pre-treatment status (p < 0.001); additional somatic findings included loss of muscle tone and gynecomastia (Chiriacò et al., 2016). Patient self-report datasets compiled through adverse-event reporting portals describe a constellation of persistent symptoms, including ED, decreased libido, genital anesthesia, reduced ejaculate, testicular atrophy, reduced penile size and gynecomastia (Healy et al., 2018; Walf et al., 2018). Health-record data provide the only prevalence estimate for these outcomes and indicate that they are uncommon: persistent ED lasting more than 90 days after discontinuation occurred in 1.4% of exposed men overall, and in 0.8% among men aged 16-42 receiving low-dose finasteride, with a median duration exceeding three years and significantly higher risk associated with longer exposure (Kiguradze et al., 2017). Isolated reports describe additional features, including penile shortening (Garreton et al., 2016).

While these accounts rest on self-report and clinical observation, objective evidence of organic dysfunction is emerging. Abnormal pudendal nerve somatosensory evoked potentials provide evidence of altered neural conduction, with ED severity correlating with these electrophysiological abnormalities rather than with mood disturbance (Melcangi et al., 2017). Increased AR expression has also been documented in penile tissue of affected individuals (Di Loreto et al., 2014).

Persistent neuropsychiatric symptoms are equally prominent: roughly half of patients meet criteria for major depressive disorder (Melcangi et al., 2017), 44-63% report suicidal ideation (Ganzer et al., 2015; Irwig, 2012a), and nearly three-quarters experience anhedonia (Chiriacò et al., 2016); these proportions refer to men with persistent symptoms rather than exposed men generally, and the cohorts overlap in recruitment through patient advocacy networks. In a controlled comparison, symptomatic finasteride users showed impaired sexual function, elevated depression and negative affectivity, cognitive complaints, and depression-consistent fMRI BOLD (Blood Oxygen Level-Dependent) responses to both erotic and non-erotic stimuli, yet without androgen deficiency or persistent peripheral 5αR inhibition (Basaria et al., 2016). Pharmacovigilance data show higher reporting with the 1 mg than the 5 mg dose across all PFS domains (Baas et al., 2018), possibly reflecting differential reporting in younger users or heightened sensitivity to neurosteroid disruption. In women, PFS has rarely been reported, and the few described cases have been transient, although exposure in this population is limited and ascertainment correspondingly poor (Fiuk et al., 2016; Mervis et al., 2018).

Clinical heterogeneity is a defining feature of PFS, likely reflecting interindividual differences in neurosteroid responses to prolonged 5αR inhibition. Genetic factors have been proposed to modulate vulnerability: androgen receptor (CAG)n repeat length influences symptom onset, with shorter repeats associated with more rapid development of adverse effects, and carriers of short (GGN)n<23 repeats reported higher frequencies of somatic symptoms (Cauci et al., 2017; Cecchin et al., 2014). These associations should be regarded as preliminary: the studies are small, and none compared symptomatic patients with exposed men who remained asymptomatic, the group required to distinguish susceptibility to persistent effects from characteristics of treated men generally. Pre-existing psychiatric illness may also represent a relevant clinical consideration, as over half of men with PFS reported a prior psychiatric diagnosis, whereas personality traits did not differ from normative expectations (Ganzer and Jacobs, 2018). Although depression and anxiety were common in this cohort, the cross-sectional design precludes determining whether these symptoms preceded finasteride exposure or developed as part of the syndrome.

The therapeutic strategies for PFS remain largely empirical. Given the absence of consistent evidence for androgen deficiency or persistent peripheral enzyme inhibition, androgen replacement therapy is unlikely to be effective (Basaria et al., 2016; Trüeb et al., 2021). Current management focuses on psychotherapy and pharmacological treatment tailored to specific symptoms (Trüeb et al., 2021, 2019). Symptomatic approaches for persistent sexual dysfunction have included phosphodiesterase inhibitors (sildenafil, vardenafil) and centrally acting agents (buspirone, trazodone, mirtazapine), although clinical efficacy appears limited (Laporte et al., 2018). Preventive strategies have been proposed, although none has been validated: these include psychological screening prior to treatment initiation (Li et al., 2022b; Trüeb et al., 2019), the use of topical formulations with lower systemic absorption (Gupta et al., 2026; Gupta and Talukder, 2022; Piraccini et al., 2022), and genetic profiling to identify individuals at increased risk (Leliefeld et al., 2025; Li et al., 2022b).

Mechanistically, the leading explanation for PFS centers on persistent alterations in neurosteroidogenesis. Chronic depletion of neurosteroids may induce compensatory remodeling of GABA-A receptors, resulting in a self-sustaining state of altered receptor sensitivity and dysregulated neurosteroid production that persists after drug discontinuation through transcriptional or epigenetic mechanisms. The receptor component of this model is supported by preclinical evidence that finasteride-induced AP withdrawal produces marked upregulation of the GABA-A α4 subunit together with altered pharmacological sensitivity (Section 6.1), although comparable remodeling has not been demonstrated in patients. Supporting the neurosteroidogenic component, reduced AP levels have been demonstrated in both CSF and plasma years after cessation, with the critical observation that CSF profiles diverge from plasma measures: pregnenolone is decreased centrally but increased peripherally, indicating that circulating levels are an unreliable proxy for brain neurosteroid status (Melcangi et al., 2017, 2013). Epigenetic alterations provide a plausible substrate for persistence, as increased methylation of SRD5A2 in CSF has been associated with reduced pregnenolone, DHT, and DHP, alongside elevated testosterone (Melcangi et al., 2019).

The gut-brain axis has been proposed as an additional component of PFS pathophysiology. In rats, finasteride withdrawal induces colonic inflammation, hypothalamic neuroinflammation, and blood-brain barrier disruption. While AP treatment ameliorates inflammatory parameters, it only partially restores mucosal and barrier structure (Diviccaro et al., 2025). Separately, patients with PFS exhibit reduced gut microbial α-diversity (Borgo et al., 2021). Mechanistically, decreased colonic AP levels during treatment and withdrawal are associated with increased IL-1β and TNF-α expression, elevated serotonin, and reduced dopamine, all of which are reversed by AP in a manner attributed to modulation of GABA-A receptors (Diviccaro et al., 2022).

6. Targeting 5αR-dependent neurosteroid signaling in neuropsychiatric disorders

The mechanistic framework outlined in Section 4 provides a biologically coherent rationale for considering 5αR-dependent neurosteroid signaling as a therapeutic target in selected neuropsychiatric disorders, through either enzyme inhibition or downstream modulation of the AP-GABA-A interaction. With few exceptions, the evidence for therapeutic benefit derives from animal models, case reports, small proof-of-concept studies, or exploratory pharmacoepidemiologic analyses. Thus, while the indications discussed below should not be regarded as clinically established, they point to conditions in which 5αR-dependent neurosteroid signaling may contribute to pathophysiology and warrant further validation. A common feature across several candidate indications is impaired behavioral control arising from corticostriatal circuit dysfunction, in which 5αR-derived neurosteroids potentiate dopaminergic signaling, particularly through D₁ receptor pathways. On this account, conditions associated with excessive or dysregulated increases in AP are the most plausible candidates for 5αR inhibition. The relation is not monotonic, however: 5αR inhibition also restores sensorimotor gating in models of chronic stress, in which neurosteroid levels are reduced rather than elevated (Section 4.2). Any selection rule based on AP tone alone is therefore provisional, and the therapeutic logic in some indications may rest on precursor redistribution rather than AP depletion.

6.1. Mood, anxiety, and stress-related disorders.

Within mood disorders, the therapeutic rationale for 5αR inhibition appears to be limited and is most plausibly relevant to premenstrual dysphoric disorder (PMDD). Chronic stress-related conditions point in the opposite direction: depression, anxiety, and PTSD are consistently associated with deficient AP signaling, and the corticolimbic remodeling that accompanies chronic neurosteroid depletion is reversed by restoring neurosteroid tone rather than by further reducing it (Pibiri et al., 2008; Pinna et al., 2006b, 2003). These findings argue against 5αRIs in such contexts and instead support strategies aimed at restoring AP tone.

PMDD represents a distinct, though still incompletely validated, case because symptoms appear to track cyclical luteal-phase changes in AP, particularly the rise and subsequent withdrawal of AP, rather than a sustained excess or deficit. Thus, any benefit of 5αR inhibition in PMDD would be expected to derive from stabilizing AP dynamics rather than simply reducing AP levels. Smith and colleagues modeled the withdrawal component of this fluctuation using finasteride-induced neurosteroid withdrawal in mice. Three days of 5αR blockade reduced hippocampal AP by approximately 50% and produced an eightfold increase in α4 subunit expression in CA1, accompanied by benzodiazepine insensitivity and enhanced responsiveness to the δ-preferring agonist THIP (Smith et al., 2006). Under these conditions, pregnanolone (the 5β epimer of AP and likewise a positive GABA-A modulator) lost its anxiolytic effect and became anxiogenic when paired with an aversive stimulus; this effect was absent in δ-knockout mice, indicating dependence on δ-containing GABA-A receptor signaling (Smith et al., 2006), with an equivalent reversal shown for AP at α4β2δ receptors at puberty (Shen et al., 2007). Thus, this paradigm reproduces the disorder rather than treating it, a distinction relevant to interpreting the clinical findings below.

Preliminary clinical evidence suggests that suppressing the rising limb of the AP cycle may attenuate PMDD symptoms. In a randomized, double-blind, placebo-controlled crossover trial, dutasteride at 2.5 mg/day prevented the luteal increase in plasma AP and reduced premenstrual symptoms in women with PMDD, whereas 0.5 mg/day was ineffective and neither dose affected asymptomatic controls (Martinez et al., 2016). However, this finding should be interpreted as proof-of-concept rather than established therapeutic validation. The apparent paradox that AP withdrawal is anxiogenic, whereas suppression of AP synthesis may be beneficial, is reconciled if the pathogenic variable in PMDD is the cyclical fluctuation in AP rather than its absolute level. Abrupt withdrawal can precipitate symptoms, whereas sustained inhibition that prevents the luteal rise and fall of AP may blunt them. This distinction is also pharmacokinetic: the intermittent blockade used to model withdrawal was achieved with finasteride, whose short half-life permits abrupt AP decline on cessation, whereas the sustained suppression achieved with dutasteride precludes it. Further studies are needed to define clinical utility, safety, patient selection criteria, and long-term consequences of this approach. Notably, any such development faces substantial constraints: the effective dose exceeded the licensed dose fivefold, and dutasteride is contraindicated in women of childbearing potential, in whom its protracted elimination sustains exposure for months after discontinuation (Section 5.2). These considerations may favor downstream strategies that stabilize AP signaling without enzyme inhibition.

6.2. Schizophrenia and psychotic disorders.

Evidence supporting 5αR inhibition in schizophrenia and related psychotic disorders remains preliminary, derived predominantly from preclinical models. As detailed in Section 4, 5αR inhibition exerts antipsychotic-like actions across dopaminergic and developmental-stress models, reversing PPI deficits induced by dopaminergic agonists, IR, and sleep deprivation without inducing catalepsy (Bortolato et al., 2011, 2008; Frau et al., 2017a, 2015). These findings implicate dysregulated 5αR-dependent neurosteroidogenesis in cortical information-processing deficits. Human data, by contrast, are limited to a single case report describing improvement in general and negative symptoms after adjunctive finasteride in a man with chronic psychosis (Koethe et al., 2008), an observation that does not distinguish neurosteroidogenic from androgenic mechanisms. Because finasteride suppresses the conversion of testosterone to DHT and can modestly increase circulating testosterone in men (Gormley et al., 1990; McConnell et al., 1992; Rittmaster et al., 1989), an alternative endocrine explanation is biologically plausible. However, whether this testosterone elevation improves negative symptoms remains unproven. Indeed, the biomarker literature in schizophrenia more consistently supports heterogeneous abnormalities of neuroactive steroids, including AP-related pathways, pregnenolone, and DHEA, than any reproducible testosterone-defined target (Marx et al., 2006; Ritsner et al., 2007). Adding further complexity, finasteride increases suicide-related aggression and abolishes clozapine-induced beneficial effects in an animal model of schizophrenia (Maurice-Gélinas et al., 2018), a finding with immediate practical implications given that 5αRIs are prescribed for AGA to the same young male population in which psychotic disorders typically emerge. Accordingly, 5αR-dependent steroid signaling remains mechanistically relevant to psychosis, but current findings do not establish 5αRIs as therapeutic agents for schizophrenia; any translational significance is likely to be state- and sex-dependent, as the male preponderance and earlier onset of the disorder may reflect the pubertal rise in 5αR activity that favors androgenic over estrogenic testosterone metabolism (Godar and Bortolato, 2014).

6.3. Substance use disorders.

Emerging evidence implicates 5αR-dependent neurosteroidogenesis in the neurobiology of substance use disorders, with the most extensive data involving alcohol. Finasteride dose-dependently blocked the acquisition of ethanol drinking and prevented ethanol preference in C57BL/6J mice, with attenuated consumption persisting two weeks post-treatment despite full recovery of brain AP, suggesting persistent behavioral modification (Ford et al., 2008, 2005). This dissociation between drug clearance and behavioral persistence is mechanistically notable, in that it parallels the phenomenon underlying PFS (Section 5.3) and may reflect a common substrate. It should be noted, however, that these studies addressed the acquisition of drinking rather than its maintenance in established dependence. In a randomized, double-blind, placebo-controlled crossover study in 27 healthy social drinkers, high-dose finasteride attenuated the stimulant and anesthetic subjective effects of alcohol on the ascending limb of the breath alcohol curve, without altering breath alcohol concentrations; these effects were largely confined to homozygotes for the A-allele at GABRA2 (n = 7), and self-reported liking and intoxication were unaffected (Pierucci-Lagha et al., 2005). Two features of this study limit its clinical extrapolation. First, the participants were healthy social drinkers, since any history of substance abuse or dependence was an exclusion criterion; second, the dose administered (200 mg over 24 h) was forty times the standard daily dose for BPH, chosen to compensate for finasteride’s weak inhibition of human 5αR1, the predominant brain isoform. Whether clinically used doses engage this mechanism centrally therefore remains untested.

These findings support a biologically plausible mechanism whereby ethanol, despite acting as a partial NMDA receptor antagonist, paradoxically enhances 5α-reduced neurosteroid immunoreactivity in CA1 pyramidal neurons through activation of the remaining unblocked receptors. This effect is abolished by finasteride and dutasteride, which also restore long-term potentiation (Tokuda et al., 2011). 5αR inhibition therefore attenuates this neurosteroid component of alcohol’s pharmacology without directly blocking ethanol’s actions at GABA-A or other receptors, representing a mechanistically distinct approach to alcohol pharmacotherapy. However, it should be observed that the effect emerged at high but not moderate ethanol concentrations (60 mM versus 20 mM in hippocampal slices), and the same authors note that in vivo ethanol additionally drives neurosteroidogenesis through ACTH-dependent adrenal StAR induction, so local NMDAR-mediated synthesis is unlikely to be the sole route.

Finasteride also reduces opioid self-administration in zebrafish and rat models without affecting locomotion or feeding, and attenuates physical signs of opioid withdrawal without altering antinociception (Bosse et al., 2021). Pharmacoepidemiologic evidence has subsequently extended these preclinical findings to humans: in an exploratory analysis of male Medicare beneficiaries receiving prescription opioid medications, 5αRI exposure was associated with reduced risk of opioid use disorder (Tawfik et al., 2026). This design cannot exclude confounding by indication, since men receiving 5αRIs differ systematically from untreated men in comorbidity, prescribing patterns, and healthcare contact.

Taken together, these findings position 5αR inhibition as a candidate strategy for both alcohol and opioid use disorders. Two constraints, however, bear directly on this indication. First, the neurosteroid rationale derives from acute intoxication, whereas AP tone is reduced rather than elevated during withdrawal and early abstinence, the phase in which pharmacotherapy is typically initiated. Second, substance use disorders carry high rates of comorbid depression and suicide, the outcomes most consistently associated with 5αRI exposure (Section 5.2). Any therapeutic development in this population would therefore need to address this risk directly.

6.4. Tourette syndrome (TS) and tic disorders.

TS is a neurodevelopmental disorder characterized by multiple motor and at least one phonic tic, with typical onset between ages 4 and 8, a 3-4:1 male preponderance, and a characteristic waxing-and-waning course in which tic severity worsens with psychological and physical stress (Leckman, 2002; Robertson, 2000). The pathophysiology involves dysfunction of cortico-striato-thalamo-cortical circuits, with dopaminergic hyperinnervation of the striatum and impaired GABAergic inhibition in the basal ganglia.

The connection between 5αR and TS was established through several convergent observations. Finasteride and dutasteride attenuate D₁-mediated behavioral responses (Frau et al., 2016, 2013) , raising the possibility that 5αR inhibition could suppress tics by dampening neurosteroid-dependent potentiation of D₁ signaling in the striatum and medial PFC. Acute stress aggravates tic-like behaviors in the D1CT7 mouse model of TS, which is characterized by D₁ receptor potentiation, with a concurrent rise in striatal AP (Mosher et al., 2017). Both stress-induced and pharmacologically induced increases in AP in the PFC and striatum are sufficient to exacerbate tic-like behaviors, and finasteride blocks this exacerbation (Cadeddu et al., 2023, 2022, 2020).

The developmental trajectory of TS is broadly compatible with a 5αR-based framework. The male preponderance is consistent with a contribution from androgen-derived neurosteroids, and the temporal course aligns with the ontogeny of their production: tic onset spans adrenarche, when adrenal androgen precursors rise; peak severity in early adolescence coincides with gonadarchal increases in testosterone and its 5αR-dependent metabolite DHT; and the frequent tic attenuation in late adolescence parallels the protracted maturation of prefrontal inhibitory circuits. This alignment may reflect a developmental window in which 5αR-dependent neurosteroidogenesis shapes tic expression before prefrontal control matures. Unlike purely dopaminergic or GABAergic accounts, this hypothesis offers a principled explanation for the age-dependence of the disorder, although it awaits direct testing (Bortolato et al., 2013; Branca and Bortolato, 2024).

Clinical translation is underway through two complementary strategies. Direct 5αR inhibition has shown preliminary evidence for potential benefit (Bortolato et al., 2007; Muroni et al., 2011), though no placebo-controlled trial has been completed. Downstream targeting of the AP-GABA-A interaction has been examined with the GAMSA sepranolone (isoAP). In an open-label, randomized, multicenter Phase IIa trial in pediatric and adult individuals with TS (NCT05434546), twice-weekly injections of sepranolone (10 mg) added to standard of care for twelve weeks reduced the Yale Global Tic Severity Scale total score by 28.0%, compared with 12.6% in patients receiving standard of care alone. This difference did not reach the prespecified threshold for significance on the primary endpoint (p = 0.051). However, secondary endpoints, including quality of life, overall impairment, and premonitory urge, improved in the sepranolone group. It should be noted that the trial was unblinded and compared against standard of care rather than placebo, a design in which these outcome measures are susceptible to expectancy; the secondary endpoints, following a primary that did not reach significance, are hypothesis-generating.

6.5. Iatrogenic complications of dopaminergic therapy in Parkinson’s disease (PD).

Two distinct iatrogenic complications of dopaminergic therapy in PD have emerged as candidate translational targets for 5αR inhibition, both plausibly grounded in the 5αR-dependent modulation of dopaminergic signaling described in Section 4.2.

The first is levodopa-induced dyskinesias (LIDs), a motor complication of chronic dopaminergic therapy that involves neuroplastic changes in the dopamine-denervated striatum. In 6-OHDA-lesioned rats, finasteride dampens both the development and expression of LIDs across acute, chronic, preventive, and concomitant regimens, and at the lower effective dose does so without compromising the ability of levodopa to improve forelimb use in the stepping test (Frau et al., 2017b). The antidyskinetic action appears to engage both D₁- and D₂/D₃-receptor function, as finasteride also attenuates abnormal involuntary movements (AIMs) elicited by SKF-82958 and ropinirole (Frau et al., 2017b), with subsequent work implicating modulation of D₁ receptor signaling in the lesioned striatum (Fanni et al., 2019). 5αRIs may additionally be neuroprotective here: dutasteride, but not finasteride, protects dopamine neurons against MPTP-induced depletion in male mice, suggesting that dual 5αR1/5αR2 inhibition is required (Litim et al., 2017, 2015). Notably, pregnenolone also reduces LIDs in female parkinsonian monkeys (Bourque et al., 2025). This result is not readily accommodated by a purely AP-dependent account, since pregnenolone is an upstream precursor of AP. One possibility is that the two interventions converge: by blocking the reduction of progesterone, 5αR inhibition redistributes substrate flux toward upstream precursors (Section 5.1), and a resulting rise in pregnenolone may contribute to the antidyskinetic effect of finasteride. This interpretation requires that pregnenolone act through routes independent of its conversion to AP, and it sits uneasily with the reduced CSF pregnenolone documented after prolonged 5αR inhibition in humans (Section 5.3). Additivity or occlusion between the two agents in the same model would discriminate between a shared and a parallel mechanism.

The second complication lies in impulse-control disorders secondary to use of dopamine agonists (particularly pramipexole and ropinirole), including pathological gambling, which appear in a substantial subset of patients treated with these medications for PD (Soileau et al., 2024). In the reserpine-plus-pramipexole model of probability discounting (arguably relevant to gambling) detailed in Section 4.2, finasteride normalizes the accompanying accumbal D3 upregulation and partially reverses the impulsive phenotype (Floris et al., 2022), consistent with preliminary clinical observations that finasteride reduces pathological gambling severity in PD patients on pramipexole (Bortolato et al., 2012). The convergence of LID and dopamine-agonist-induced impulse-control disorders on 5αR-dependent modulation of dopaminergic signaling, albeit through different receptor subtypes, makes these two iatrogenic complications among the more tractable repurposing opportunities for 5αRIs in clinical neurology. Any such use, however, would have to be weighed against the neuropsychiatric liability of 5αRIs, a concern heightened by the prevalence of depression in PD and by the impaired behavioral restraint that defines the impulse-control subgroup.

7. A Sex-Biased Role of 5αR2 in Stress Reactivity

The preceding sections establish that the two steroidogenic 5αRs are not redundant enzymes but functionally distinct regulators of neurosteroid signaling, neuroendocrine integration, and behavior. This section examines a particularly consequential recent finding: a tonic/phasic division of labor in which 5αR1 maintains a neurosteroidogenic baseline common to both sexes, while 5αR2 enables phasic AP synthesis during acute challenge, a response observed thus far only in males (Cadeddu et al., 2025). This distinction refines the stress-related framework developed in Section 4.3, according to which 5αR-derived neurosteroids both restrain HPA-axis activation under basal conditions and participate in the acute response to challenge. Acute stress robustly stimulates AP and THDOC synthesis (Barbaccia et al., 1996; Purdy et al., 1991) and enhances GABAergic inhibition across prefrontal, hippocampal, and amygdalar circuits (Brunton, 2015; Pinna et al., 2006a).

The evidence for this dissociation derives primarily from a recent systematic study of both isoenzymes in the medial PFC of rats. Forced swim and footshock selectively increased 5αR2 protein in the male medial PFC within 30 min, without altering 5αR1; neither isoenzyme changed in females. Immunofluorescence localized 5αR2 to the somata and neurites of prelimbic and infralimbic pyramidal neurons. AAV5-shRNA experiments revealed distinct functional contributions of the two isoenzymes: 5αR1 knockdown produced minimal behavioral effects, whereas 5αR2 knockdown increased forced-swim immobility and reduced defensive withdrawal, exploration, olfactory arousal, social interaction, and sucrose preference, without impairing locomotion. These deficits were confined to the medial PFC and males, were rescued by systemic AP, and were reproduced in constitutive CRISPR/Cas9 5αR2-knockout rats. By contrast, 5αR2 knockdown in the NAc reduced only social interaction and sucrose preference (Cadeddu et al., 2025).

Neurosteroid measurements provided mechanistic support for this distinction. At baseline, 5αR1 knockdown reduced the AP/progesterone ratio, consistent with impaired constitutive conversion, whereas 5αR2 knockdown lowered both steroids without altering their ratio, suggesting an indirect effect on substrate availability. After stress, however, only 5αR2 knockdown abolished the AP rise, leaving progesterone unconverted (Cadeddu et al., 2025). This pattern accords with the complementary kinetics and regulation of the two enzymes (Table 1). 5αR1 is constitutively expressed and displays lower affinity but higher capacity, a profile well suited to tonic conversion in proportion to substrate availability. Conversely, 5αR2 is scarcer, displays higher affinity, and, most critically, is inducible, so that its rise within 30 min of stress ensures that the adrenal progesterone surge encounters an expanded enzyme pool. Accordingly, the phasic character of this response is likely to follow from induction, with affinity plausibly contributing to conversion efficiency (Russell and Wilson, 1994). Single-nucleus RNA sequencing further indicated that the 5αR2-dependent signal is required for an adaptive cortical stress response and that its loss redirects, rather than merely suppresses, this response. Forced swim upregulated translational and biosynthetic programs across neuronal and glial populations in controls but abolished or reversed them after knockdown, downregulating mitochondrial ATP-synthesis and oxidative-phosphorylation modules and upregulating cytoskeletal and synaptic-organization pathways. The extension of these changes to astrocytes, which lacked detectable 5αR2, suggests that pyramidal neuron-derived AP acts in a paracrine manner (Cadeddu et al., 2025).

Complementary evidence supports the tonic arm of the model. Chronic stress and social isolation downregulate 5αR1 in corticolimbic pyramidal neurons (Agís-Balboa et al., 2007). Constitutive 5αR1 loss also disrupts HPA-axis regulation: knockout mice show impaired glucocorticoid clearance, hepatic corticosterone accumulation, reduced hypothalamic Crh, decreased pituitary Crhr1, and a 44% reduction in the corticosterone response to restraint (Livingstone et al., 2017, 2014). The phenotype is more pronounced in females, which has been interpreted as evidence that it reflects altered glucocorticoid rather than androgen metabolism; as noted in Section 3.3, however, the female predominance of hepatic 5αR1 expression in rodents provides an alternative account of the same asymmetry. GABA-A subunit composition, chloride homeostasis, and circuit context further shape the interaction between the two arms, such that chronic stress, pharmacological inhibition, and genetic loss may generate qualitatively distinct forms of dysregulation.

This framework extends beyond the PFC. In a witnessed-trauma paradigm, both direct and vicarious threat selectively reduced Srd5a2, but not Srd5a1, in the basolateral amygdala, alongside reduced Gabrd expression. Conditional deletion of Srd5a2 in CaMKII-positive neurons exaggerated contextual fear in male mice (Evans-Strong et al., 2024). Although the direction of regulation differs (acute stress induces 5αR2 in the medial PFC, whereas sustained threat depletes it in the basolateral amygdala), both findings implicate 5αR2 as the stress-responsive isoenzyme. Together, they suggest that phasic, 5αR2-dependent neurosteroidogenesis may be a broader property of corticolimbic stress circuits.

A central feature of the model is the sexual dimorphism of its phasic arm. Stress-induced 5αR2 induction occurs only in males (Cadeddu et al., 2025), consistent with evidence that acute stress mobilizes brain AP more reliably in males (Pisu et al., 2022; Sze and Brunton, 2020). Stressors, CRH, and ACTH rapidly raise central AP in males (Barbaccia et al., 1996; Purdy et al., 1991), whereas in females they raise plasma but not brain AP, and footshock often fails in either compartment despite intact corticosterone (Pisu et al., 2022). This dissociation separates the neurosteroidogenic limb from HPA-axis activation, mirroring the lost AP surge after male 5αR2 knockdown. Baseline tone may provide a second determinant: brain and plasma AP are two- to sixteen-fold higher in females, plausibly reflecting greater progesterone supply, and raising male AP to female levels with progesterone abolishes the stress-induced rise (Pisu et al., 2022). The female pattern may thus combine an absent inducible arm with an elevated baseline that occludes on-demand synthesis. The knockout reproduces this asymmetry: male knockout rats recapitulate the full medial PFC phenotype, whereas females show only reduced social interaction (Cadeddu et al., 2025). Furthermore, mice with the orthologous mutation show a milder, dominance-related deficit (Mosher et al., 2018) resistant to the antidepressant effects of exercise (unpublished observations).

These findings suggest that the phasic AP signal may be protective only within an optimal range, at least in males. Insufficient signaling, modeled by 5αR2 knockdown or finasteride administration, produces passive coping, anhedonia, and social withdrawal, features characteristic of the internalizing pole. Conversely, excessive signaling, particularly when recruited by acute stress in vulnerable circuits, may potentiate GABA-A currents on prefrontal pyramidal neurons, weaken top-down control over subcortical motor and motivational networks, and promote tics and impulsive choice. This possibility is supported by observations that these behaviors can be reversed by finasteride and by evidence that intra-PFC AP appears to worsen information processing and increase the propensity for tic-relevant behaviors (Cadeddu et al., 2023, 2022, 2020; Floris et al., 2022; Mosher et al., 2017; see Sections 4.2, 6.4, and 6.5). Thus, the same response that buffers acute stress under physiological conditions may become maladaptive when either insufficient or excessive.

Because the phasic arm appears to be male-restricted, its excessive activation may contribute to sex differences in psychopathology. Women and men show higher rates of internalizing and externalizing disorders, respectively (Martel, 2013; Seedat et al., 2009). Latent-variable analyses support this distinction, identifying internalizing and externalizing liability dimensions with largely invariant structure but sex-divergent mean levels (Eaton et al., 2012; Kramer et al., 2008; Krueger, 1999; Krueger et al., 2001). If comparable stress-related arousal is preferentially expressed through inward- or outward-directed coping modes (approximating, at their extremes, “tend-and-befriend” and “fight-or-flight”) the tonic–phasic model offers a potential biological substrate (Cannon, 1932; Taylor et al., 2000). Both sexes are exposed to the internalizing route through erosion of tonic 5αR1-dependent neurosteroidogenesis (Sections 4.3 and 5.2), whereas males may additionally possess a phasic pathway whose overactivation favors externalizing outcomes. Under this formulation, comparable overall liability combined with an additional male-biased pathway could produce divergence in clinical expression rather than total burden. These patterns are further shaped by sexually dimorphic HPA-axis and CRH signaling (Bangasser and Valentino, 2014; Crowley and Girdler, 2014). Consistent with the tonic arm of the model, women with major depression, PMDD, or PTSD show reduced basal AP and a blunted or absent AP response to laboratory stressors despite preserved cortisol responses, and a prior history of depression alone is sufficient to impair this response (Pisu et al., 2022).

This framework may also help explain the affective liabilities of 5αRIs. Because finasteride preferentially inhibits human 5αR2 (Section 5.1), its central effects at clinical exposures should fall primarily on the inducible, phasic component, producing selective loss of the stress-evoked AP surge rather than a global failure of 5α-reduction. Finasteride may therefore shift the system toward reduced phasic AP signaling: this may benefit externalizing phenotypes, consistent with its anti-tic efficacy (Section 6.4), while shifting stress adaptation toward an internalizing, depressive state in susceptible individuals. Its anti-tic efficacy and depressogenic liability could thus arise from the same action on the same isoenzyme, expressed in different biological and clinical contexts. The rodent behavioral studies cited above, however, used doses at which finasteride inhibits both isoenzymes; the isoenzyme-specific attribution therefore rests on genetic, not pharmacological, evidence.

The depressive phenotype associated with 5αR2 loss is unlikely to reflect nonspecific behavioral suppression. Increased forced-swim immobility, social withdrawal, anhedonia, and blunted arousal were all rescued by systemic AP (Cadeddu et al., 2025). This pattern aligns with depressive symptoms reported after finasteride exposure and provides a potential mechanism by which a 5αR2-preferring drug could precipitate depression rather than merely cause sedation. The same model may also account for symptoms that persist after discontinuation. In PFS, SRD5A2, but not SRD5A1, shows increased methylation accompanied by altered neurosteroid profiles (Melcangi et al., 2019). Recovery after subchronic finasteride exposure in rats is similarly asymmetric: 5αR1 expression, Fos signaling, and serum DHT normalize within four weeks, whereas 5αR2 remains suppressed (Choi et al., 2025). Thus, the isoenzyme preferentially inhibited by finasteride is also selectively methylated in patients and persistently suppressed in animals, providing a plausible molecular correlate of symptoms that outlast treatment.

These claims require caution. The central premise rests primarily on a single rat study, and the stress-inducible role of 5αR2 has not been replicated across other regions or species. Moreover, the rodent work models experimentally imposed 5αR2 loss rather than the clinical syndrome, and the human evidence remains largely associative. The framework should be regarded as a mechanistic hypothesis rather than an established explanation. It nonetheless generates testable predictions: finasteride-associated depression should involve isoenzyme-selective impairment rather than global suppression of 5α-reduction; persistent symptoms should preferentially track 5αR2 dysregulation; and residual deficits should, in principle, respond to restored AP signaling. The model therefore provides a mechanistically specific alternative to explanations based solely on androgen deficiency or nocebo effects and defines the findings that direct validation in human tissue, peripheral biomarkers, and translational models would need to demonstrate (Section 5.3).

8. Open Questions and Concluding Remarks

The framework in Section 7 proposes an isoenzyme-resolved model of brain neurosteroidogenic stress responses, comprising a tonic 5αR1-dependent baseline and a phasic, male-biased 5αR2-dependent response. This distinction may reconcile the divergent expression and regulation of both isoenzymes, sex-dependent AP responses to stress, the sex-biased prevalence of internalizing and externalizing psychopathology, and the affective liabilities of 5αR2-preferring inhibitors. Because evidence remains limited, the model should be regarded as a working hypothesis. The questions below highlight the principal gaps that must be addressed to define the roles of 5αRs in brain function and behavior.

• How are 5αR1 and 5αR2 regulated in the brain? Acute stress in males is proposed to recruit a phasic 5αR2-dependent response (Cadeddu et al., 2025), whereas chronic stress may weaken tonic 5αR1-dependent neurosteroidogenesis (Agís-Balboa et al., 2007; Pibiri et al., 2008) and reduce both isoenzymes in male corticolimbic regions (Bortolato et al., 2011). Future studies should define the upstream regulatory mechanisms (including stress-induced epigenetic modifications of SRD5A1 and SRD5A2) that govern these changes across cell types, brain regions, sexes, and stress paradigms, and determine whether they are transient or persistent.

• Which downstream metabolic pathways mediate the differential effects of 5αR activity? 5αR catalysis simultaneously drives GABAergic neurosteroidogenesis from progesterone and DOC (Majewska et al., 1986; Purdy et al., 1991), amplifies androgen signaling (Bruchovsky and Wilson, 1968a; Russell and Wilson, 1994), participates in corticosteroid metabolism (Livingstone et al., 2015), and, at least in some species, generates pheromonal 16-androstenes (Claus, 1979). The relative behavioral contributions of these pathways, and how they vary across regions, cell types, sexes, and physiological states, remain unresolved.

• What are the cellular architecture and developmental timing of 5αRs? Studies based on bulk tissue, systemic pharmacology, or constitutive knockouts cannot distinguish neuronal, glial, autocrine, and paracrine mechanisms (Cadeddu et al., 2025; Castelli et al., 2013; Melcangi et al., 1993, 1990). The limitation is most acute during development, when neurosteroids exert organizational effects on circuit assembly (Baulieu, 1998; Corpéchot et al., 1981). Conditional deletion, single-nucleus sequencing, spatial transcriptomics, and developmental time-course studies should clarify how 5αR expression shapes circuit assembly and how it influences vulnerability to neuropsychiatric disorders (Pinna and Bortolato, 2025).

• How can rapid neurosteroid dynamics be monitored in vivo? Current methods, including tissue extraction, CSF sampling, and peripheral assays, remain poorly suited to resolving rapid, region-specific neurosteroid fluctuations in vivo. Although microdialysis detected insult-evoked AP increases in the extracellular fluid of the fetal sheep brain (Nguyen et al., 2004), it has not become a routine method for monitoring neurosteroids during stress in rodents because of low extracellular concentrations, high lipophilicity, poor recovery, co-elution, and ion suppression (Dury et al., 2015; Higashi and Ogawa, 2016; Lionetto et al., 2017). Stable-isotope derivatization and magnetic dispersive extraction have enabled picogram-per-milliliter quantification in rat blood microdialysate (Xu et al., 2019). Extending these methods to brain microdialysis could reveal phasic AP and THDOC dynamics and help distinguish local synthesis from the uptake of peripheral precursors.

• Does 5αR3 have any direct role in steroid metabolism or neurosteroid signaling? Although SRD5A3 was initially proposed to encode a steroid-metabolizing enzyme (Uemura et al., 2008; Yamana et al., 2010), it was subsequently identified as a polyprenol reductase in dolichol synthesis and N-glycosylation (Cantagrel et al., 2010). Whether 5αR3 retains direct steroidogenic activity or instead modulates steroid-responsive signaling through glycosylation remains unresolved. Targeted steroidomics and glycoproteomic profiling should distinguish these possibilities.

• Could 5αR1-preferring inhibitors be repurposed? The translational implications of 5αR biology remain underdeveloped. 5αR1-preferring compounds such as bexlosteride and MK-386 were generally tolerable but insufficiently efficacious in androgen-dependent conditions (Ellsworth et al., 1996; Hirsch et al., 1993; Leyden et al., 2004; Schwartz et al., 1997). Given the emerging role of 5αR1 in neurosteroidogenesis, however, these compounds may warrant evaluation in neurobehavioral conditions associated with excessive neurosteroid production, as discussed in Section 6.

• Can peripherally restricted 5αR inhibitors be developed to avoid the neurobehavioral adverse effects of current 5αR inhibitors? Finasteride and dutasteride suppress peripheral androgen metabolism effectively (Clark et al., 2004; Gormley et al., 1992; McConnell et al., 1992) but cross the blood-brain barrier and disrupt central neurosteroidogenesis, with persistent sequelae reported in susceptible individuals (Healy et al., 2022; Irwig and Kolukula, 2011; Melcangi et al., 2017; Traish et al., 2011). Such agents therefore represent an important unmet need. Their rational use will also require longitudinal phenotyping, pharmacogenomic stratification (Cauci et al., 2017; Gillespie et al., 2013; Leliefeld et al., 2025), and assays capable of resolving individual steroid species.

Two priorities emerge from these questions. The first is to establish the causal roles of 5αRs in psychiatric disorders through isoenzyme-, cell-type-, circuit-, and sex-resolved analyses. The model predicts that phasic 5αR2 signaling contributes preferentially to male-biased externalizing pathology, whereas disruption of the tonic 5αR1 baseline is more relevant to AP-deficient internalizing states. Testing these predictions will require parallel male and female cohorts, acute and chronic stress paradigms, isoenzyme-selective perturbations, and behavioral measures that resolve specific functional domains.

The second priority is therapeutic translation. The approval of brexanolone and zuranolone for postpartum depression (Deligiannidis et al., 2021; Meltzer-Brody et al., 2018), together with the development of related GABA-A receptor modulators, establishes the AP pathway as clinically tractable. Nevertheless, a uniform therapeutic strategy is unlikely to be effective across disorders. Restoring an impaired tonic baseline may benefit AP-deficient internalizing states, whereas excessive phasic neurosteroidogenesis may require attenuation. Future trials should therefore incorporate sex-resolved designs, steroidomic biomarkers, measures of stress reactivity, and circuit-state markers such as indices of chloride regulation. The objective should be to match treatment to isoenzyme, circuit state, sex, and clinical phenotype rather than treating neurosteroid enhancement or suppression as universally beneficial.

Taken together, the available evidence argues against viewing 5α-reduction solely as a peripheral, androgen-activating process. Instead, brain 5αR1 and 5αR2 should be regarded as functionally distinct contributors to neurosteroidogenesis whose roles vary by cell type, circuit, and sex. The tonic/phasic framework should be regarded as a working hypothesis rather than a settled conclusion; from this perspective, its main value lies in organizing a fragmented literature and identifying the gaps that must be closed before it can be properly tested. What is already apparent, however, is that the two isoenzymes are not functionally redundant. We fully expect that resolving their distinct, sex-dependent contributions will prove essential to understanding how neurosteroidogenic capacity shapes vulnerability and resilience across stress-related disorders.

Abbreviations:

11βHSD1–2

11β-hydroxysteroid dehydrogenase type 1–2

11KA4

11-ketoandrostenedione

11KDHT

11-ketodihydrotestosterone

11KT

11-ketotestosterone

11OHA4

11βhydroxyandrostenedione

17βHSD

17β-hydroxysteroid dehydrogenase

17-OHP

17α-hydroxyprogesterone

20αHSD

20α-hydroxysteroid dehydrogenase

3α-diol

5α-androstane-3α,17β-diol

3αHSD

3α-hydroxysteroid dehydrogenase

3α-HSOR

3α-hydroxysteroid oxidoreductase

3β-diol

5α-androstane-3β,17β-diol

3βHSD

3βhydroxysteroid dehydrogenase

3β-HSOR

3β-hydroxysteroid oxidoreductase

5αR1

5α-reductase type 1

5αR2

5α-reductase type 2

5αR3

5α-reductase type 3

5αRIs

5α-reductase inhibitors

5-HT3

5-hydroxytryptamine type 3 receptor

6-OHDA

6-hydroxydopamine

ACTH

adrenocorticotropic hormone

AF2

activation function-2

AGA

androgenetic alopecia

AIMs

abnormal involuntary movements

AKR1C1–AKR1C4

aldo-keto reductase family 1, members C1 to C4

AKR1D1

aldo-keto reductase family 1 member D1 (steroid 5β-reductase)

ANT-1

adenine nucleotide translocator-1

AP

allopregnanolone (3α,5α-tetrahydroprogesterone)

AR

androgen receptor

BDNF

brain-derived neurotrophic factor

BPH

benign prostatic hyperplasia

CaMKII

Ca²⁺/calmodulin-dependent protein kinase II

CRH

corticotropin-releasing hormone

CRHR1

CRH receptor 1

CRPC

castration-resistant prostate cancer

CYP11B1

cytochrome P450 11B1 (11β-hydroxylase)

CYP17A1

cytochrome P450 17A1 (17αhydroxylase/17, 20-lyase, P450c17)

DDET2 / DET2L

deetiolated 2 / DET2-like

DHB

5α-dihydrocorticosterone

DHDOC

5α-dihydrodeoxycorticosterone

DHEA

dehydroepiandrosterone

DHF

5α-dihydrocortisol

DHP

5α-dihydroprogesterone

DHT

5αdihydrotestosterone

DOC

deoxycorticosterone

DOPAC

3,4-dihydroxyphenylacetic acid

ED

erectile dysfunction

ER

endoplasmic reticulum

ERβ

estrogen receptor β

Fkbp51

FK506-binding protein 51

GABA

γ-aminobutyric acid

GABA-A

γ-aminobutyric acid type A receptor

GABRA2

gene encoding GABA-A receptor α2 subunit

GABRB3

gene encoding GABA-A receptor β3 subunit

GABRD

gene encoding GABA-A receptor δ subunit

GAD67

glutamate decarboxylase 67

GAMSA

GABA-A receptor modulating steroid antagonist

GFAP

glial fibrillary acidic protein

GR

glucocorticoid receptor

HPA

hypothalamic-pituitary-adrenal (axis)

HSD11B1

gene encoding 11βHSD1

IL-1β

interleukin-1β

IL-6

interleukin-6

IPSC

inhibitory postsynaptic current

IR

isolation rearing

isoAP

isoallopregnanolone

isoTHDOC

isotetrahydrodeoxycorticosterone

KCC2

K + −Cl - cotransporter 2

LIDs

levodopa-induced dyskinesias

LPS

lipopolysaccharide

mPRα

membrane progesterone receptor α

MPTP

1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine

NAc

nucleus accumbens

nAChR

nicotinic acetylcholine receptor

NMDA

N-methyl-Daspartate

nPR

nuclear progesterone receptor

OR7D4

olfactory receptor 7D4

PD

Parkinson’s disease

Pdiol

5α-pregnane-3α,17α-diol-20-one

PFC

prefrontal cortex

PFS

post-finasteride syndrome

PMDD

premenstrual dysphoric disorder

PNMT

phenylethanolamine N-methyltransferase

PPI

prepulse inhibition

PPRD

polyprenol reductase

PTSD

post-traumatic stress disorder

PVN

paraventricular nucleus

shRNA

short hairpin RNA

sIPSC

spontaneous inhibitory postsynaptic current

SNP

single-nucleotide polymorphism

snRNA-seq

single-nucleus RNA sequencing

SRD5A1

gene encoding 5αreductase type 1

SRD5A2

gene encoding 5α-reductase type 2

SRD5A3

gene encoding 5α-reductase type 3 (polyprenol reductase)

StAR

steroidogenic acute regulatory protein

TECR

trans-2,3-enoyl-CoA reductase

TECRL

trans-2,3-enoyl-CoA reductase-like

THB

3α,5αtetrahydrocorticosterone

THDOC

3α,5α-tetrahydrodeoxycorticosterone

THF

3α,5α-tetrahydrocortisol

TNF-α

tumor necrosis factor-α

TrkB

tropomyosin receptor kinase B

TS

Tourette syndrome