Progesterone biosynthesis: from cholesterol to P4 through StAR, CYP11A1, and HSD3B

Project: L7-1
Stage: REPORT_EN
Line: L7 — Progesterone and neurosteroid metabolomics
Evidence cutoff date: 2026-07-29
Primary population: nonpregnant women, approximately 35–55 years old, from the late reproductive stage through the menopausal transition
Scope: mechanistic research and hypothesis generation; this is not clinical advice, diagnosis, or prescription
Status: long-form scientific report in English, prepared for BILINGUAL_QA

Scientific delta of the report

This report replaces the linear narrative age → less StAR/CYP11A1/HSD3B2 → less progesterone → worse longevity with a hierarchical causal architecture. Human evidence locates the best-demonstrated change in the existence and trajectory of luteal tissue—ovulation, corpus luteum duration/mass, and LH support—not in an already proven uniform loss of the three enzymatic steps. The donor-aware reanalysis of GSE202601 weakened an independent OXPHOS decline and did not support a pathway-specific lesion; the most coherent pattern involved granulosa identity/composition. In addition, the high processivity of CYP11A1 shows that free hydroxylated substrates do not by themselves localize an endogenous lesion and requires blinded assay qualification against known partial perturbations before reproductive stages are compared.

The connection to health is reorganized into five gates: production, secretion/exposure, distribution/clearance, tissue conversion, and functional sensitivity. This separation explains why a single P4 measurement, PdG, serum allopregnanolone, sleep, FMD, BMD, and longevity cannot substitute for one another. No direct human chain was verified from ovarian StAR–CYP11A1–HSD3B2 flux to healthspan or survival. The most mature health link remains a contextual association with the complete ovulatory state; the Pharma opportunity remains assay-enabled discovery, not target validation.

1. Executive scientific abstract

Progesterone (P4) in the nonpregnant ovary is produced mainly after ovulation. LH/hCG signaling sustains a steroidogenic luteal identity; cholesterol must be available and reach the inner mitochondrial membrane with help from StAR; CYP11A1, coupled to FDX1/FDXR and NADPH, converts cholesterol into pregnenolone through three highly processive oxidations; HSD3B2, the predominant gonadal isoform, converts pregnenolone into P4. The necessity of StAR and CYP11A1 is established by severe human natural experiments, and the function of HSD3B2 by human tissue. Biochemical necessity, however, is not equivalent to a mechanism of acquired aging.

Two repeatedly observed human phenomena coexist during the menopausal transition: anovulatory cycles increase, and PdG also decreases within apparently ovulatory cycles. No verified study causally decomposed the contribution of each process. P4 responds to LH within minutes to hours and can fall before StAR, CYP11A1, or HSD3B change in abundance; a single serum sample correlates weakly with integrated luteal exposure. In older IVF granulosa cells, StAR did not decline and CYP11A1 increased, together with apoptosis and altered identity. The simple form “less enzyme with age” is therefore weakened.

Pseudobulk reanalysis of eight ovaries from GSE202601 confirmed the threat of pseudoreplication. In granulosa, nuclear OXPHOS changed from an extreme nucleus-level difference to a small and uncertain donor-aware effect (old−young = −0.211; descriptive CI95% −0.836 to 0.413; exact p 0.371). The steroidogenic module was incomplete and inseparable from identity. The most ordered signal was a granulosa-identity module, limited to three filtered genes and not surviving module-level FDR. This supports exploratory cellular remodeling, not reduced flux.

The extension to health is even less mature. P4 and allopregnanolone are not proportional across the cycle; a serum ratio measures neither enzymatic activity nor the brain. Sleep studies are small and contextual; vascular signals are not P4-specific; the closest bone study did not separate P4 from ovulation, E2, FSH, and energy availability. No source connected ovarian flux, a tissue mediator, function, a clinical event, and longevity.

Three hypotheses govern validation:

1. Primary — L7-1-AR-H1 v2: ovulatory–luteal architecture explains a larger contribution to the decline in P4 exposure than per-cell capacity.
2. Competing — L7-1-AR-H2 v2: a dynamic pre-pregnenolone loss exists within comparable cells and is detectable only by a previously qualified flux assay.
3. Translational — L7-1-AR-HT1 v2: only after the gates are validated will a reproducible dynamic component add out-of-sample information; the ex vivo component is removed if the cellular hypothesis dies.

The smallest decisive experiment is D0A: a blinded benchmark with partial perturbations of StAR, CYP11A1, HSD3B2, and a common bioenergetic failure. If the assay cannot distinguish the classes, age/stage is not compared and all inference is reduced to “global steroidogenic reserve.” If it passes, D0B tests stage within a comparable cellular state; D1 then decomposes ovulation, corpus luteum, exposure, and—only when discordance remains—clearance. Healthspan remains subordinate to future organ-specific mediation.

2. Scientific question and relevance

Question

In nonpregnant women transitioning from the late reproductive stage to perimenopause, is there an intrinsic and localizable cellular reduction in flux:

available cholesterol → StAR-dependent mitochondrial transport/access → pregnenolone through CYP11A1–FDX1/FDXR → P4 through HSD3B2,

independent of ovulation, luteal mass and identity, LH support, mitochondrial function, and cycle timing; and, if it exists, does it alter P4/neurosteroid exposure and a proximal function before contributing to health or longevity?

Relevance

The question matters because “low progesterone” mixes distinct mechanisms. It may mean that a corpus luteum did not form, lasted less time, received less LH support, contains fewer competent cells, produces less P4 per cell, clears P4 faster, or was sampled at a noncomparable time. Each explanation predicts a different experimental intervention and Pharma opportunity. Collapsing them creates false targets.

The relevance to female longevity is potential but unproven. P4 and its derivatives can act in the brain, endometrium, and other tissues; ovulatory state is associated with bone and vascular processes. Nevertheless, a healthspan mechanism requires precedence and mediation: localized flux → validated exposure → tissue effector → function → organ-specific outcome → disease-free years. This project finds firm support only for the earliest biochemical steps and contextual support for some functional steps.

3. Scope, population, and life stage

Primary human population

Nonpregnant adult women with ovaries, approximately 35–55 years old, followed from the late reproductive stage through the menopausal transition. Stage must be defined using STRAW+10 criteria, menstrual history, and repeated evidence of ovulation, not age or a single hormone.

Required strata

• late reproductive stage with ovulatory cycles;
• early transition with documented ovulation;
• late transition, separating ovulatory and anovulatory cycles;
• natural postmenopause as a boundary, not as a substitute for luteal physiology;
• phase dated by LH/ovulation, not calendar day alone;
• age, adiposity, thyroid function, energy status, inflammation, liver/kidney function, sleep/circadian state, medications, and glucocorticoid exposure as modifiers or confounders.

Experimental systems

• primary, nontumor human granulosa-lutein cells;
• human corpus luteum tissue when provenance is traceable;
• animal models only for architecture and conservation;
• in-vitro/reconstituted systems for kinetics and localization;
• public computational data only for state/composition, never as substitutes for flux.

Exclusions from the causal core

Pregnancy and postpartum because of the placental source; exogenous steroids during analytical windows; oophorectomy and primary ovarian insufficiency in the primary analysis; advanced adrenal, liver, or kidney disease; a single P4 sample without phase; and direct extrapolation from IVF, placenta, adrenal, or brain to the natural corpus luteum.

4. Background knowledge and mechanistic map

4.1 The biochemical chain

1. LH/LHCGR–cAMP/PKA. Ovulation generates and sustains the corpus luteum. P4 follows LH pulses and changes before the transcriptional machinery.
2. Cholesterol availability. Lipoprotein uptake, local synthesis, lipid droplets, de-esterification, and intracellular trafficking determine accessible substrate.
3. StAR. Facilitates acute cholesterol access to CYP11A1 at the inner mitochondrial membrane. Severe human mutations profoundly reduce steroidogenesis, although residual StAR-independent flux exists.
4. CYP11A1. Executes three oxidations and side-chain cleavage. It depends on FDX1/FDXR, NADPH, oxygen, and mitochondrial organization. The first C-22 oxidation is the slowest; intermediates remain bound in a highly processive manner [1].
5. HSD3B2. Converts pregnenolone into P4 with NAD⁺. It is the predominant gonadal/adrenal isoform; HSD3B1 must not be treated as equivalent [5].
6. Secretion and exposure. P4 is secreted pulsatilely; plasma, urinary PdG, and production are not the same object.
7. Downstream conversion. P4 → 5α-DHP → allopregnanolone. Serum abundance identifies neither source, flux, nor brain access.
8. Receptor and function. P4 acts through progesterone receptors; allopregnanolone modulates GABA(_A). Tissue, receptor, prior exposure, and rate of change modify response.

4.2 Two-layer architecture

Layer A — ovulatory–luteal architecture. Determines whether a corpus luteum exists, how much steroidogenic tissue is present, for how long, and with what LH support. This layer has the strongest longitudinal human evidence.

Layer B — dynamic per-cell capacity. A cell may preserve basal StAR/CYP11A1/HSD3B2 abundance yet respond less to LH because of cholesterol logistics, CYP11A1–redox coupling, or bioenergetics. This layer is plausible but has not been measured by human stage with tracers.

4.3 Five gates between ovary and health

Longevity would be a sixth longitudinal layer, not an implicit property of P4.

5. Evidence method

All prior project artifacts were integrated: scoping, evidence map, verification, mechanism, computational decision and execution, hypothesis generation, adversarial review, and experimental design. Decisive claims were traced to primary sources.

For each study, when available, the following were extracted: design, population, sex/life stage, species/tissue, independent unit, size, exposure, comparator, outcome, effect, uncertainty, DOI, and PMID/PMCID. Human genetic, human longitudinal/experimental, human ex-vivo/in-vitro, animal, computational, and inferred evidence were separated.

Null results and contrary explanations were actively sought. The audit corrected the DOI for Devoto 2001, SWAN denominators, the PMDD study design, pseudoreplication in the nine-woman bone study, and HSD3B1/HSD3B2 conflation. A final search of primary sources confirmed CYP11A1 processivity, the biological 4+4 size of the ovarian atlas, and the two-cycle design in 20 trained women [1,16,22].

The absence of a stage-specific flux study is interpreted as a gap in the located evidence set, not proof that no such study exists.

6. Evidence map

6.1 Biochemistry and cell

6.2 Menopausal transition and metrology

The most defensible inference is coexistence of cycle reconfiguration and residual ovulatory decline. It is invalid to state what proportion belongs to each.

6.3 Computational layer

The GSE202601 reanalysis used ovary/donor as the unit (4+4) and exact permutation of the 70 balanced assignments.

HSD3B2 appeared in three of eight donors, with 99.8% of its counts in one ovary; LHCGR appeared in only one donor. The dataset does not evaluate the complete pathway. When nuclei were treated as replicates, several modules yielded p<2.2×10⁻¹⁶; when donors were used, the signal weakened materially. The result supports exploratory identity/composition remodeling, not flux or computational H2 [16].

6.4 Downstream and health

7. Contradictory evidence and null findings

1. StAR/CYP11A1 are necessary but not uniformly reduced with age. Congenital deficiencies prove necessity; older IVF granulosa showed similar StAR and higher CYP11A1, with altered identity/apoptosis [10].
2. P4 can fall before the machinery. After LH withdrawal, P4 fell earlier and more than StAR in humans and CYP11A1/HSD3B mRNA in macaque [8,14]. This permits rapid regulation and uncoupling; it does not prove that the proteins are functionally intact.
3. More CYP11A1 does not imply more P4. Murine overexpression altered differentiation and mitochondria, with low early P4 [15].
4. Anovulation and ovulatory decline coexist. The first explanation does not eliminate the second; data do not permit a declaration of dominance.
5. The OXPHOS signal depends on the analytical unit. Extreme nucleus-level significance did not survive robustly by donor.
6. P4 and allopregnanolone are not linear. The ratio changes by phase, but this does not demonstrate saturation, conversion, or source.
7. More P4/allopregnanolone does not always equal better function. Physiological sleep may remain stable; in a selected affective population, higher follicular allopregnanolone anticipated greater symptom burden [21,22].
8. Bone/vasculature do not isolate P4. PeKnO did not separate co-hormones; FMD did not specifically follow P4; postmenopausal oral P4 was largely null [25–27].
9. Null results do not prove equivalence. Several samples are small. Their intervals require “inconclusive” when they include both the null and a material effect.

8. Multiscale mechanistic synthesis

Molecular/organelle

The pathway operates as a module:

LH/LHCGR–cAMP/PKA × luteal identity × cholesterol pool/mobilization × processed/localized StAR × CYP11A1–FDX1–FDXR–NADPH–O₂ × mitochondrial integrity × HSD3B2–NAD⁺ × viability.

A reduction in any term can lower P4. The abundance of three proteins does not measure the product of the module.

Cellular/tissue

The transition may change cell proportions, luteinization, apoptosis, perfusion, and CL duration. Tissue with fewer competent cells produces less P4 even if each cell retains capacity. A common bioenergetic failure can reduce all substrates and mimic a pathway lesion. The only way to separate the models is to measure flux within previously defined states and demonstrate a reproducible selective rescue.

Systemic

Circulating exposure depends on pulsatile secretion, tissue mass, distribution, and clearance. Integrated P4/PdG improves dating but does not equal production. If cellular capacity, imaging, and LH do not explain persistent discordance, a clearance substudy must be activated rather than imputing an ovarian lesion.

Target tissue

Function depends on local conversion and receptors. Plasma does not represent the brain. A temporal P4–allopregnanolone–function association is insufficient to attribute GABA(_A); it requires a receptor readout or orthogonal perturbation. For bone and vasculature, E2, FSH, energy, ovulation, and stage are explicit rivals.

Healthspan trajectory

The candidate pathway:

architecture/flux → P4 AUC → tissue effector → proximal function → organ-specific outcome → disease-free years

contains unobserved gates. Current evidence does not permit concatenating them. Longevity cannot be inferred from fertility, age at menopause, sleep, FMD, or BMD in isolation.

9. Computational layer

Question

Did the OXPHOS/steroidogenic signal from the human atlas persist when donors, not nuclei, were treated as replicates?

Input and method

Public counts and metadata from GSE202601, granulosa pseudobulk across eight ovaries, edgeR, prespecified modules, exact permutation, leave-one-donor-out, and sensitivity analyses by nucleus count. No Lua or private data were used.

Decision

The computational layer weakened the priority of a specific mitochondrial lesion and did not strengthen a StAR–CYP11A1–HSD3B2 lesion. It preserved an exploratory association with identity/composition. No embedding, sequence model, or structure model can correct n=4+4, stage confounding, sparsity, or the absence of labeled product; BioNeMo was therefore not used.

What cannot be inferred

Cholesterol transport, StAR processing, CYP11A1/HSD3B2 activity, flux, secretion, allopregnanolone, function, healthspan, or longevity.

10. Primary hypothesis — L7-1-AR-H1 v2

Falsifiable statement

In a within-woman decomposition of stage-related P4/PdG change, the contribution of forming and maintaining functional luteal tissue—ovulation probability, duration/mass/perfusion, and LH support—will exceed the contribution of a per-cell capacity difference; cellular capacity will fall within a justified equivalence margin.

Mechanism

HPO/LH dynamics → ovulation probability → CL mass × duration × perfusion → cell-time executing the pathway → P4/PdG AUC → contextual association with functions, initially dominated by the complete cycle.

Evidence for

• longitudinal human cohorts with more anovulation and changing PdG;
• rapid and pulsatile LH dependence;
• absence of a robust specific cellular lesion in omics;
• bone associations that do not separate P4 from ovulatory state.

Evidence against

• PdG also declines in apparently ovulatory cycles;
• “functional CL time” is an unvalidated construct;
• volume/perfusion do not equal steroidogenic mass;
• per-cell capacity and clearance were not measured in the same women.

Predictions

1. Separating ovulation probability and conditional AUC will explain distinct components of the decline.
2. CL imaging/vascularity/duration, measured without P4, will precede and explain AUC better than age.
3. The stage contrast will decrease among cycles with comparable architecture.
4. In qualified D0B, per-cell capacity will fall within the material margin.
5. If clearance changes, an exposure–capacity discordance will persist.

Kill criteria

• specific cellular capacity contributes more on a common scale;
• the cellular defect precedes the architectural change;
• the CL construct does not improve out-of-sample fit/calibration;
• attenuation disappears after correcting ovulation error, attrition, or clearance.

Status and maturity

weakened_narrowed, still primary. Confidence 0.52. H1 for architectural change; H0 for causal dominance and cellular equivalence.

11. Competing hypothesis — L7-1-AR-H2 v2

Falsifiable statement

After a qualified assay distinguishes partial perturbations of StAR, CYP11A1, and HSD3B2, luteal cells from a more advanced stage will show a lower LH-evoked cholesterol-to-product flux response within the same cellular state, not explained by global mitochondrial function.

Mechanism

The candidate lesion lies before pregnenolone: cholesterol logistics/access, localized StAR, or CYP11A1 catalysis/coupling. It is an evoked reserve, not a basal decline in abundance.

Evidence for

• P4 falls before StAR/CYP11A1/HSD3B;
• CYP11A1 requires redox/mitochondrial coupling;
• the hypothesis produces flux patterns and is refutable.

Evidence against

• no human stage-specific flux study exists;
• 22R-OHC bypasses logistics and the first CYP11A1 oxidation;
• free intermediates do not reproduce high processivity;
• HDL adds an independent steroidogenic signal;
• donor-aware omics did not reveal a robust pathway-specific signal.

Predictions

1. Negative stage × LH × time interaction for labeled product from cholesterol.
2. The pattern will appear in two readouts/preparations and reproduce a D0A signature.
3. StAR, CYP11A1, and HSD3B2 perturbations will generate blindly classifiable signatures.
4. Global failure will reduce several arms and accompany low ATP/OCR/viability.
5. The effect will persist within pretreatment state and comparable lipoproteins.

Kill criteria

• D0A cannot distinguish targets or the pattern depends on dose/permeation;
• basal and stimulated flux falls within equivalence with adequate power;
• the effect disappears when equivalent states are compared;
• no orthogonal rescue exists;
• only RNA/protein or unlabeled product changes.

Status and maturity

weakened_narrowed, proposed. Confidence 0.22. Maturity H0.

12. Translational hypothesis — L7-1-AR-HT1 v2

Falsifiable statement

Only after each gate separately passes metrology and causality will one frozen component per gate improve external error and calibration over phase, ovulation, and integrated P4/PdG; ex-vivo flux will be excluded if AR-H2 is refuted.

Mechanism

Variation is divided into two gates: architecture/production/clearance → exposure, and conversion/receptor → function. A large vector would be vulnerable to phase, batch, and missingness, so only one component per gate is allowed.

Predictions

1. Each assay will pass analytical precision, stability, and within-state repeatability.
2. The component will be selected before the confirmatory outcome.
3. It will improve external error/calibration, not merely phase classification.
4. The architecture will simplify if a gate is null.
5. It will not be called a healthspan biomarker without replication and incremental utility.

Evidence for

A single P4 measurement has inadequate metrology; P4 and allopregnanolone are not proportional; functions are context-dependent.

Evidence against

The component does not yet exist, may be impractical, and ex-vivo flux may not transport. Predicting a proximal function does not prove health or utility.

Kill criteria

• metrology failure;
• gain only in-sample or through phase, batch, center, or missingness;
• failure to transport beyond IVF/selected phenotype;
• null proximal function or no precedence;
• complexity incompatible with reproducibility.

Status and maturity

parked, confidence 0.05, maturity H0; HUMAN_QA_REQUIRED before translational evaluation.

13. Falsifiable predictions and kill criteria

Equivalence must be justified using analytical error, between-donor variation, and a minimum biologically material difference. An interval crossing both nullity and relevance is inconclusive.

14. Discriminating experiment

D0A — the first decisive gate

Question: can the assay blindly distinguish a partial perturbation of StAR, CYP11A1, or HSD3B2 from a common bioenergetic failure?

System: primary, nontumor human granulosa-lutein cells, with donor as the unit. D0A-0 begins with at least four donors for feasibility; D0A-1 proposes 16 independent donors for a paired benchmark of large signatures.

Perturbations:

• two nonoverlapping reagents for STAR, CYP11A1, and HSD3B2;
• nontargeting control;
• reagent-resistant rescue when feasible;
• general mitochondrial-impairment control, titrated without dominant toxicity.

Inputs/readouts:

• stably labeled cholesterol → labeled pregnenolone/P4;
• 22R-OHC → P4, with HDL/LDL fixed;
• labeled pregnenolone → labeled P4;
• cholesterol import/access, processed/localized StAR;
• CYP11A1 activity, FDX1/FDXR;
• HSD3B2 and HSD3B1 separately;
• ATP, OCR, potential, redox, and viability.

Prespecified success:

• balanced accuracy ≥0.75;
• sensitivity per class ≥0.75;
• bootstrap lower bound above the chance level for four classes;
• agreement of two reagents;
• rescue of ≥50% of the specific loss;
• independence from viability, plate, and batch.

Failure: if StAR and CYP11A1 cannot be separated, the conclusion is reduced to “pre-pregnenolone reserve.” If targets are confused with general failure, localization dies and stage comparison does not begin.

D0B — stage × state

After freezing the assay, compare the late reproductive stage and early transition within common support of pretreatment identity/state. Pilot with 12 donors per stage; proposed maximum expansion to 32 per stage. Endpoint: log-AUC of labeled P4 from cholesterol/viable cell. The final margin is not fixed until D0A error and between-donor variance are observed.

AR-H2 is strengthened if a specific D0A signature persists within state and does not depend on lipoproteins, OCR, or viability. AR-H3, the state/composition rival, wins if state explains flux and stage falls within equivalence.

D1 — ovulatory–luteal decomposition

After D0: metrology pilot of 20 participants/two cycles; proposed confirmation of 120 completers, 60 per stage, three cycles. Measure daily LH/PdG, serial P4/E2, and P4-independent 3D ultrasound/Doppler of the CL. A hierarchical hurdle model separates ovulation probability from conditional AUC. G-computation places contributions on a common scale.

Clearance is activated only with persistent discordance among capacity, imaging/LH, and exposure. Any human tracer requires independent ethical and human review; no dose is specified.

Conditional D2 and D3

D2: P4, 5α-DHP, allopregnanolone, and E2 by LC–MS/MS, with one proximal neurophysiological readout and external validation. Tests incremental value, not GABA(_A).

D3: one organ, initially bone because PeKnO is available. Woman-level reanalysis modeling E2, FSH, stage, BMI, energy/exercise, and attrition. The 49 completers filter large mediation effects; they do not prove equivalence.

15. Biomarkers and stratification

There is no validated biomarker of StAR–CYP11A1–HSD3B2 capacity or healthspan.

Research candidates

• Dynamic cellular capacity: AUC of labeled product per cell under LH; only after D0A.
• Luteal architecture: ovulation + duration + P4-independent volume/perfusion; construct under validation.
• Exposure: serial LH-aligned P4/PdG; exposure, not pure production.
• Downstream conversion: P4, 5α-DHP, and allopregnanolone by MS; do not use a ratio as activity.
• Cellular state: pretreatment identity independent of pathway genes/products; do not turn isolated FSHR or OXPHOS into a biomarker.

Required stratification

STRAW+10 stage, ovulatory/anovulatory cycle, IVF/nonstimulated provenance, cellular state/competence, gonadotropin protocol, BMI/energy, thyroid function, liver/kidney function, medications, and susceptibility defined before the outcome. Ancestry and metabolic context require replication; no Mexico/LATAM-specific conclusion exists.

16. Individual variability

Variability is not a single type of noise. It arises from:

• ovulation probability and timing;
• CL mass, duration, perfusion, and LH support;
• pulsatility;
• cellular identity and reserve;
• lipoprotein availability and bioenergetics;
• production and clearance;
• conversion to 5α-DHP/allopregnanolone;
• receptor, prior exposure, and susceptibility;
• E2, FSH, energy, adiposity, thyroid function, inflammation, and rhythms;
• analytical and preanalytical error.

The same serum P4 can correspond to different fluxes; the same flux can produce different exposures; the same exposure can generate different functions. This double source–effector uncoupling is a central and falsifiable inference: covariance between levels should improve only when the intermediate gates are added.

17. Pharma relevance and maturity

Current level

Assay-enabled discovery. There is no validated target, candidate, safety window, or clinical readiness.

Research opportunities

1. Phenotypic rescue screen in qualified D0A. This is the nearest opportunity because it does not presuppose the rate-limiting step.
2. Cholesterol logistics/mitochondrial coupling. SCARB1, LXR/SREBP, StAR, and the redox system are plausible nodes with high pleiotropy.
3. Direct CYP11A1/HSD3B2. Less attractive because of adrenal/gonadal essentiality, isoform selectivity, and risk of steroid imbalance.
4. Neurosteroids/GABA(_A). Potential tractability in other contexts, but it does not validate ovarian biosynthesis or longevity.

Minimum pre-lead package

• replication across multiple donors;
• concentration–effect anchored to plausible exposure;
• target engagement/deconvolution;
• LC–MS/MS steroid panel, not P4 alone;
• viability, ATP, OCR, and stress;
• adrenal counter-screen and HSD3B2/HSD3B1 selectivity;
• PR/GR/MR and off-target panel;
• basic metabolism/stability;
• replication in a preparation close to the natural corpus luteum.

Pharma kill criteria

Supraplausible concentration, indiscriminate steroid amplification, toxicity/bioenergetic impairment, failure to replicate across donors, loss of effect outside a stimulated preparation, discordant target engagement, or rescue explained by common cellular state.

Pharma maturity: pre-target/assay-enabling, H0; HUMAN_QA_REQUIRED.

18. Limitations

1. No human stage comparison with isotopic flux per luteal cell exists.
2. Accessible tissue comes from IVF, surgery, or postmortem contexts and does not represent natural transition.
3. Granulosa, granulosa-lutein, and corpus luteum are not equivalent.
4. Stage is coupled to age, reserve, ovulation, provenance, and composition.
5. Longitudinal studies measure PdG/exposure, not cellular production.
6. P4 is pulsatile; PdG adds metabolism, kidney function, and dilution.
7. HSD3B1/HSD3B2 were conflated in part of the literature.
8. 22R-OHC and free intermediates are not perfect scalpels because of processivity and permeation.
9. OXPHOS/identity may be cause, consequence, mediator, selection, or artifact.
10. GSE202601 has four donors per group, unequal depth, and sparse pathway genes.
11. Plasma does not measure brain or GABA(_A) adaptation.
12. Sleep, affect, FMD, and bone studies are small, selected, or confounded.
13. A low-powered null does not demonstrate equivalence.
14. Transportability by ancestry or LATAM context has not been demonstrated.
15. No branch has demonstrated clinical events, composite healthspan, or longevity.
16. This report does not evaluate efficacy, safety, or clinical use of progesterone/progestogens.

19. Conclusions

The StAR–CYP11A1–HSD3B2 pathway is indispensable for progesterone production, but a uniform intrinsic loss of the pathway has not been shown to drive the menopausal transition. The strongest human evidence lies upstream, in ovulatory–luteal architecture and LH support; a decline in PdG also coexists within apparently ovulatory cycles and remains unlocalized.

Omics does not resolve that residual. When donor was used as the unit, the presumed OXPHOS/pathway decline became uncertain and shared with identity/composition. Nor is a 22R-OHC bypass sufficient: CYP11A1 is highly processive and the substrate bypasses more than one step; HDL adds a signal not attributable solely to cholesterol delivery. The first validation must demonstrate that the assay recognizes known perturbations.

The health connection is plausible only through gates. P4–allopregnanolone nonproportionality, contextual sleep results, and bone/vascular data preclude a universal protective narrative. No direct human evidence supports mediation from ovarian flux to healthspan or longevity.

Decisive progress is sequential: qualify flux, compare stage within state, decompose architecture and exposure, validate a proximal function, and only then test organ-specific mediation. The Pharma opportunity is a discovery platform; any language of target, biomarker, treatment, or longevity would be premature.

20. References

1. McCarty KD, Liu L, Tateishi Y, Wapshott-Stehli HL, Guengerich FP. The multistep oxidation of cholesterol to pregnenolone by human cytochrome P450 11A1 is highly processive. J Biol Chem. 2024;300(1):105495. DOI: 10.1016/j.jbc.2023.105495. PMID: 38006947; PMCID: PMC10716780.
2. Lin D, Sugawara T, Strauss JF III, et al. Role of steroidogenic acute regulatory protein in adrenal and gonadal steroidogenesis. Science. 1995;267:1828–1831. DOI: 10.1126/science.7892608. PMID: 7892608.
3. Bose HS, Sugawara T, Strauss JF III, Miller WL. The pathophysiology and genetics of congenital lipoid adrenal hyperplasia. N Engl J Med. 1996;335:1870–1878. DOI: 10.1056/NEJM199612193352503. PMID: 8948562.
4. al Kandari H, Katsumata N, Alexander S, Rasoul MA. Homozygous mutation of P450 side-chain cleavage enzyme gene (CYP11A1) in 46,XY patient with adrenal insufficiency, complete sex reversal, and agenesis of corpus callosum. J Clin Endocrinol Metab. 2006;91:2821–2826. DOI: 10.1210/jc.2005-2230. PMID: 16705068.
5. Havelock JC, Smith AL, Seely JB, Dooley CA, Rodgers RJ, Rainey WE, Carr BR. The NGFI-B family of transcription factors regulates expression of 3β-hydroxysteroid dehydrogenase type 2 in the human ovary. Mol Hum Reprod. 2005;11(2):79–85. DOI: 10.1093/molehr/gah139.
6. Du Y, et al. Expression of genes and enzymes involved in ovarian steroidogenesis in relation to human follicular development. Front Endocrinol. 2023;14:1268248. DOI: 10.3389/fendo.2023.1268248. PMID: 37964966; PMCID: PMC10641382.
7. Ragoobir J, Abayasekara DRE, Michael AE. Stimulation of progesterone production in human granulosa-lutein cells by lipoproteins: evidence for cholesterol-independent actions of high-density lipoproteins. J Endocrinol. 2002;173:103–115. DOI: 10.1677/joe.0.1730103. PMID: 11927389.
8. Devoto L, et al. Expression of steroidogenic acute regulatory protein in the human corpus luteum throughout the luteal phase. J Clin Endocrinol Metab. 2001;86. DOI: 10.1210/jcem.86.11.7982. PMID: 11701746.
9. Filicori M, Butler JP, Crowley WF Jr. Neuroendocrine regulation of the corpus luteum in the human. Evidence for pulsatile progesterone secretion. J Clin Invest. 1984;73. DOI: 10.1172/JCI111370.
10. Wu YT, et al. Aging-related premature luteinization of granulosa cells is avoided by early oocyte retrieval. J Endocrinol. 2015. DOI: 10.1530/JOE-15-0246. PMID: 26264981.
11. Santoro N, et al. Factors related to declining luteal function in women during the menopausal transition. J Clin Endocrinol Metab. 2008;93. DOI: 10.1210/jc.2007-2165. PMID: 18285413; PMCID: PMC2386686.
12. O’Connor KA, et al. Progesterone and ovulation across stages of the transition to menopause. Menopause. 2009;16. DOI: 10.1097/gme.0b013e3181aa192d. PMID: 19568209; PMCID: PMC2783957.
13. Santoro N, et al. Daily luteal serum and urinary hormone profiles in the menopause transition: Study of Women's Health Across the Nation. Menopause. 2020. DOI: 10.1097/GME.0000000000001453. PMID: 31794501; PMCID: PMC7050767.
14. Ravindranath N, et al. Role of luteinizing hormone in the expression of cholesterol side-chain cleavage cytochrome P450 and 3 beta-hydroxysteroid dehydrogenase, delta 5-4 isomerase messenger ribonucleic acids in the primate corpus luteum. Endocrinology. 1992;131. DOI: 10.1210/endo.131.5.1425410. PMID: 1425410.
15. Chien Y, et al. Misregulated progesterone secretion and impaired pregnancy in Cyp11a1 transgenic mice. Biol Reprod. 2013. DOI: 10.1095/biolreprod.113.110833. PMID: 23966322.
16. Jin C, Wang X, Yang J, et al. Molecular and genetic insights into human ovarian aging from single-nuclei multi-omics analyses. Nat Aging. 2025;5:275–290. DOI: 10.1038/s43587-024-00762-5. PMID: 39578560. Data: GSE202601.
17. Hamidovic A, et al. Trajectories of allopregnanolone and allopregnanolone to progesterone ratio across the six subphases of menstrual cycle. Biomolecules. 2023;13:652. DOI: 10.3390/biom13040652. PMID: 37189398; PMCID: PMC10135782.
18. Kimball A, et al. Allopregnanolone and progesterone across menstrual cycle and menopause. Psychoneuroendocrinology. 2020;112:104512. DOI: 10.1016/j.psyneuen.2019.104512. PMID: 31780185; PMCID: PMC6935417.
19. Schmidt PJ, et al. Premenstrual dysphoric disorder symptoms following ovarian suppression: triggered by change in ovarian steroid levels but not continuous stable levels. Am J Psychiatry. 2017. DOI: 10.1176/appi.ajp.2017.16101113. PMID: 28427285; PMCID: PMC5624833.
20. Nguyen TV, et al. The steroid metabolome in women with premenstrual dysphoric disorder during GnRH agonist-induced ovarian suppression: effects of estradiol and progesterone addback. Transl Psychiatry. 2017. DOI: 10.1038/tp.2017.146. PMID: 28786978; PMCID: PMC5611719.
21. Kimball A, et al. Neuroactive steroid levels are elevated in the follicular phase and predict premenstrual depression and anxiety symptom severity in women with MRMD. Arch Womens Ment Health. 2025. DOI: 10.1007/s00737-024-01532-3. PMID: 39601877.
22. Pearson MA, Weakley JJS, Johnston RD, et al. Variations in estrogen, progesterone, and nocturnal physiology are not associated with performance and sleep within and between menstrual cycles. Med Sci Sports Exerc. 2026. DOI: 10.1249/MSS.0000000000003966. PMID: 41747284.
23. Shechter A, et al. Circadian variation of sleep during the follicular and luteal phases of the menstrual cycle. Sleep. 2010;33:647–656. DOI: 10.1093/sleep/33.5.647. PMID: 20469807; PMCID: PMC2864880.
24. Schüssler P, et al. Progesterone reduces wakefulness in sleep EEG and has no effect on cognition in healthy postmenopausal women. Psychoneuroendocrinology. 2008. DOI: 10.1016/j.psyneuen.2008.05.013. PMID: 18676087.
25. Hashimoto M, et al. Modulation of endothelium-dependent flow-mediated dilatation of the brachial artery by sex and menstrual cycle. Circulation. 1995;92:3431–3435. DOI: 10.1161/01.CIR.92.12.3431. PMID: 8521564.
26. Prior JC, et al. Progesterone therapy, endothelial function and cardiovascular risk markers in healthy postmenopausal women. PLOS One. 2014;9:e84698. DOI: 10.1371/journal.pone.0084698. PMCID: PMC3897380.
27. Starrach T, Santl A, Seifert-Klauss VR. Perimenopausal bone loss is associated with ovulatory activity—results of the PeKnO Study (Perimenopausal Bone Density and Ovulation). Diagnostics. 2022;12:305. DOI: 10.3390/diagnostics12020305. PMID: 35204396.
28. Gorgels WJMJ, et al. Urinary sex hormone excretions in premenopausal women and coronary heart disease risk: a nested case-referent study in the DOM-cohort. J Clin Epidemiol. 1997. DOI: 10.1016/S0895-4356(96)00367-8. PMID: 9120526.
29. Kohen P, et al. The steroidogenic response and corpus luteum expression of the steroidogenic acute regulatory protein after human chorionic gonadotropin administration at different times in the human luteal phase. J Clin Endocrinol Metab. 2003. DOI: 10.1210/jc.2002-021916. PMID: 12843197.
30. Moravek MB, et al. HCG-mediated activation of mTORC1 signaling plays a crucial role in steroidogenesis in human granulosa lutein cells. Endocrine. 2016. DOI: 10.1007/s12020-016-1065-8. PMID: 27503318; PMCID: PMC5071160.

Final report claim registry

• L7-1-REPORT-EN-C1: human evidence supports a two-layer architecture: demonstrated ovulatory–luteal reconfiguration and still-hypothetical dynamic cellular reserve.
• L7-1-REPORT-EN-C2: StAR/CYP11A1/HSD3B2 abundance, a single P4 measurement, and downstream serum ratios are inadequate substitutes for flux, exposure, and function.
• L7-1-REPORT-EN-C3: donor-aware reanalysis weakens a specific OXPHOS/steroidogenic lesion and retains only exploratory identity/composition remodeling.
• L7-1-REPORT-EN-C4: CYP11A1 processivity and HDL effects preclude using 22R-OHC as a unique StAR localizer; D0A must precede every stage comparison.
• L7-1-REPORT-EN-C5: no direct human evidence links ovarian P4 flux to healthspan or longevity; translation requires organ-specific mediation and competing risks.
• L7-1-REPORT-EN-C6: the Pharma opportunity is a qualified phenotypic discovery platform, not a validated target or candidate.