INSR/IRS1 Polymorphisms and Insulin Resistance in PCOS: Why the Same Diet Does Not Work Equally

Stage: REPORT_EN
Cutoff date: 2026-07-22
Project: L6-4 · Hormonal nutrigenomics
Status: complete long-form scientific report; QA required before promotion
Overall maturity: H1 for the non-cell-autonomous state model; H0–H1 for acquired persistence and modification by IRS1 G972R; H3 observational for the PCOS/insulin-resistance → diabetes/events bridge, not for the allele
Scope: mechanistic and translational research; not diagnosis, prescription, or individual diet selection

1. Executive scientific abstract

The available evidence does not support the popular explanation that a common INSR or IRS1 variant determines which diet “works” for a woman with polycystic ovary syndrome (PCOS). The most defensible conclusion is more complex and scientifically more useful: the same dietary prescription can produce different responses because it does not guarantee the same consumed exposure or the same internal metabolic dose, and because insulin action depends on a dynamic, nonlinear, tissue-specific state. Candidate genetics may modify that state, but it is not currently the dominant explanation.

The historical INSR rs1799817 variant is synonymous—NM_000208.4:c.3255C>T; p.His1085= in the current annotation—lacks allele-specific functional validation, and shows incompatible PCOS associations across populations. One of the most frequently cited positive studies deviates markedly from Hardy–Weinberg equilibrium in controls. Large 2025–2026 GWAS support a polygenic architecture and, in 2025, a region near INSR, but they do not make rs1799817 a causal variant. Locus, gene, and variant are not equivalent.

IRS1 rs1801278, p.Gly971Arg in current RefSeq and G972R in the historical nomenclature, does have biochemical plausibility. Recombinant fragments and overexpression systems show abnormal association with INSR, reduced IRS1 phosphorylation, reduced p85–PI3K recruitment, lower AKT signaling, glucose transport, and glycogen synthesis. Nevertheless, the human association with PCOS or insulin resistance is not stable, the allele does not emerge as a principal determinant in recent GWAS, and the verified evidence set contains no endogenous, isogenic, rescued demonstration in female human muscle.

Human ex-vivo evidence more strongly favors a state dependent on the extracellular milieu. In small historical series, a receptor-phosphorylation defect could be transferred or corrected without an INSR mutation. In later studies, insulin resistance and certain muscle signatures present in vivo were not retained after myotube expansion and differentiation. This elevates an adversarial explanation: adiposity, hyperinsulinemia, androgens, lipids, adipokines, and other systemic signals may sustain the defect while present without leaving durable cellular memory.

The most plausible endocrine bridge is tissue selectivity. Muscle and adipose tissue may require more insulin to move glucose, while ovarian theca retains sufficient steroidogenic signaling. Compensatory hyperinsulinemia may then amplify androgen production without a receptor mutation. This chain is supported by separate human links, but it has not been measured end to end, in paired tissues, or by genotype within the same participants.

There is insufficient human evidence for a chronic INSR/IRS1 × diet interaction specific to PCOS. The closest study administered acute glucose or fructose loads to only four carriers and four noncarriers with PCOS per genotype cell, without a sufficiently powered formal interaction contrast. Longer dietary trials were conducted outside PCOS, used another variant near IRS1 or a multigene panel, and produced context-dependent or null signals. There is no scientific basis for assigning a diet from rs1799817 or G972R.

The connection to long-term women's health currently belongs to the metabolic phenotype, not the SNP. Cohorts of women with PCOS link insulin resistance, adiposity, and glycemia to type 2 diabetes and cardiocerebrovascular events. The 2026 GWAS adds genetic pleiotropy: susceptibility to PCOS is associated with later natural menopause and adverse cardiometabolic outcomes. That does not establish two already separable components or a net “longevity” valence; later menopause is not equivalent to healthspan, functional ovarian reserve, or survival.

The candidate mechanistic hypothesis is that G972R increases the persistence of functional impairment after a defined metabolic exposure. The competing hypothesis—and the one currently better supported—is that the defect requires continuous exposure or, secondarily, persists independently of the allele. The smallest discriminating experiment, D0-AR, uses female human myotubes, two genetic backgrounds, reciprocal Gly/Gly ↔ Gly/Arg editing, a palmitate challenge frozen before genotype unblinding, continuous exposure versus washout and recovery, glucose uptake as the primary endpoint, and isotope tracing to distinguish functional memory from lipid retention.

Concrete scientific delta of REPORT_EN

This English scientific twin preserves and consolidates the project's corrected causal hierarchy: continuous non-cell-autonomous state > persistent acquired state > G972R allelic persistence. It carries forward the discriminable explanation of internal lipid-dose retention, relegates pAKT to a mechanistic readout, and treats absence of post-washout persistence as a positive result for the non-cell-autonomous hypothesis. The central implication is that the question “which diet belongs to which genotype” is premature; it must first be shown that the allele modifies functional flux at endogenous dosage and that the modification survives control of exposure, state, tissue, and genetic background.

2. Scientific question and relevance

Question

In adult premenopausal women with PCOS, do common INSR or IRS1 variants causally modify the insulin-sensitivity response to an equivalent dietary exposure through a proximal and tissue-coherent alteration of INSR→IRS1→PI3K→AKT signaling; or is the heterogeneity better explained by effective exposure, adiposity, beta-cell compensation, acquired phosphoregulation, and polygenic architecture?

Why it matters

The question combines three problems that are often conflated:

1. PCOS susceptibility: whether a variant increases the probability of meeting diagnostic criteria.
2. Metabolic severity: whether it modifies sensitivity, secretion, or compensation within PCOS.
3. Differential dietary response: whether it changes the causal effect of one composition versus another after energy, adherence, and weight are equalized.

An association with PCOS does not prove insulin resistance. An association with HOMA does not identify muscle, liver, or beta cell. A different response within a single diet does not prove composition specificity. And a change in an intermediate metabolic endpoint is not equivalent to greater longevity.

The relevance to women's health arises from the convergence of metabolism and reproduction. Peripheral resistance can increase insulin demand; ovarian theca may retain a steroidogenic response; hyperinsulinemia can amplify the androgenic environment; and the resulting cycle may extend a trajectory of diabetes and vascular risk beyond reproductive age. Identifying which part is genetic, acquired, and reversible changes experimental priority and the plausibility of any pharmacological translation.

3. Scope, population, and life stage

Primary population

The core population is women approximately 18–40 years old, premenopausal, with PCOS characterized under a recognized framework and with its diagnostic components reported separately: hyperandrogenism, ovulatory dysfunction, and ovarian morphology. A broad range of adiposity and insulin sensitivity is required; BMI must not substitute for body composition or central adiposity. Women without PCOS, comparable in age, ancestry, and adiposity, are necessary to test syndrome specificity.

Adolescence is outside the primary analysis because of overlap among normal puberty, menstrual irregularity, and PCOS criteria. Pregnancy, lactation, and the early postpartum period are excluded because insulin sensitivity is physiologically reprogrammed. Perimenopause and postmenopause enter only through the life-course bridge when PCOS history is well characterized; reproductive-age dietary response is not extrapolated to those stages.

Rare pathogenic INSR variants, lipodystrophies, and monogenic severe-insulin-resistance syndromes do not estimate the effect of common variants. Mexico and Latin America are priorities for transportability, but no frequency or effect magnitude may be inferred from an ethnic label. Genetic ancestry, relevant LD, dietary context, and regional replication are required.

What “the same diet” means

The phrase must be separated into four layers:

Only the last two layers approximate the biological dose. A valid nutrigenetic trial therefore requires at least two diets, quantified exposure, stable weight or explicit weight modeling, a dynamic measure of sensitivity, and a formal genotype×diet term. “Same recommendation” does not identify a genetic interaction.

4. Background knowledge and mechanistic map

Canonical signaling

Insulin binding to INSR promotes tyrosine autophosphorylation and IRS1 recruitment/phosphorylation. Phosphorylated IRS1 recruits p85–PI3K, increases PIP3, and activates AKT. In muscle and adipocytes, this network facilitates GLUT4 trafficking, glucose uptake, and storage. In liver, insulin suppresses endogenous glucose production and participates in SHBG regulation. In beta cells, secretion compensates for loss of peripheral sensitivity. This biochemistry is stable background knowledge; it does not attribute a defect to PCOS or to an allele.

Defensible chain

Polygenic/regulatory architecture + acquired metabolic state
→ tissue-dependent INSR–IRS1–PI3K–AKT dose–time curve
→ variable muscle/adipose uptake and hepatic suppression
→ compensatory insulin demand
→ hyperinsulinemic euglycemia or glycemic excursion according to beta-cell reserve
→ ovarian-theca exposure to insulin with preserved steroidogenic response
→ amplification of androgens and ovulatory dysfunction
→ persistent metabolic trajectory
→ type 2 diabetes and vascular events
→ potential reduction in disease-free years.

The first nodes are connected by biochemical and human ex-vivo evidence. Full multitissue coexistence is inferred because the links were measured in different studies and participants. The step to total longevity is not established.

Three causal layers explaining heterogeneity

1. Exposure: the dose of glucose, fat, and energy that reaches tissues.
2. Signaling slope: how much insulin is needed to achieve one unit of flux.
3. Compensation/selectivity: how much the beta cell can secrete and which tissues retain different outputs at the same insulin level.

A SNP is causal for dietary response only if it reproducibly modifies one of these layers at endogenous dosage and the interaction persists after the other two are controlled.

5. Evidence method

The scoping, evidence-map, verification, synthesis, compute-decision, hypothesis-generation, adversarial-review, and experiment-design artifacts were integrated in full. Decisive sources were verified against primary articles, PubMed, publisher pages, and supplements when available. Reviews and meta-analyses were used to locate heterogeneity, not to replace audit of primary studies.

Study design, population, sex/life stage, species or cell type, sample size, exposure/comparator, result, uncertainty, DOI/PMID, and role in the causal chain were extracted. Variant nomenclature, Hardy–Weinberg equilibrium, cohort independence, multiplicity, experimental unit, temporality, control selection, ancestry, and endpoint validity were audited.

Evidence hierarchy

• Human in vivo: association, physiology, intervention, or follow-up in participants.
• Human ex vivo: biopsy, primary culture, theca, muscle, or human islets.
• In vitro: animal or human cell lines, overexpression, protein, or recombinant fragment.
• Animal: organism-level causality not directly transportable. No animal result is decisive for this report's central estimand.
• Computational/genetic: GWAS, PRS, colocalization, and variant-to-gene assignment.
• Inferred: connection of results across levels; never presented as direct observation.

HOMA and QUICKI are treated as fasting proxies. They do not replace clamp, FSIVGTT, insulin-suppression testing, or tracers. pAKT is a proximal readout; it does not replace glucose uptake. Natural menopause, ovulation, weight, and HOMA do not replace healthspan or survival.

6. Evidence map

6.1 Human genetics and candidate variants

The balance rules out rs1799817 as a robust, transportable determinant. G972R may modify a metabolic subphenotype, but it does not demonstrate general PCOS susceptibility.

6.2 Molecular function and cellular state

Functional convergence retains G972R in the experimental program but does not justify attributing clinical heterogeneity to it. Adversarial evidence that the phenotype is erased in culture ranks the continuous systemic state above a universal cellular memory.

6.3 Tissue selectivity and ovary

Nestler et al. showed in human PCOS theca that insulin increased testosterone and that blocking INSR, but not IGF1R, abolished the effect; the number of donors could not be verified in the accessible text, and the magnitude must not be generalized [17]. Yen et al., in cells from 11 women with PCOS and 10 controls, observed theca-specific remodeling of IRS1/IRS2/IRS4 relative to granulosa, without proving that abundance equals signaling [18]. Tosi et al., in nine women with PCOS, found that experimental hyperinsulinemia amplified the ovarian steroid response to GnRH without a parallel increase in gonadotropins [19].

These results support the possibility of selective resistance: metabolically resistant muscle and an ovary with preserved steroidogenic output. They do not show that both states coexisted in the same women or depended on a SNP.

6.4 Interaction with diet

No chronic RCT was located that combined two diets, women with PCOS, rs1799817 or G972R, measured exposure, and clamp/FSIVGTT. This is a conditional search result, not logical proof of global nonexistence.

6.5 Healthspan and life course

Gambineri et al. followed 255 Italian women with PCOS for a mean of 16.9 years; 42 incident diabetes cases occurred, and BMI and baseline glucose/AUC predicted the trajectory [29]. Ryu et al. compared 137,416 Korean women with PCOS with 411,984 controls; the adjusted HR for the cardiovascular/cerebrovascular composite was 1.224, with limitations from administrative coding, residual confounding, and a median follow-up of 54 months [30].

These cohorts connect PCOS/insulin resistance to diabetes and events, not to the candidate SNPs. Moolhuijsen 2026 adds polygenic associations with T2D/CAD and later age at menopause, but not individual follow-up from allele→diet→event or an already demonstrated causal partition.

7. Contradictory evidence and null findings

7.1 INSR: local positives versus large-scale genetics

rs1799817 was positive in some Indian or Saudi samples and null in Croatian, Iranian, Kashmiri, Han-family, and meta-analytic evidence. Rival explanations are ancestry-dependent LD, control selection, adiposity subgroups, orientation/nomenclature, multiplicity, or false positives. HWE failure in the Saudi study and absence as a lead signal in large GWAS prevent using the SNP as a stable cause.

7.2 G972R: coherent biochemistry versus inconsistent clinical evidence

The variant perturbs signaling in models but is not consistently associated with PCOS and does not appear as a major locus in 2025–2026. This may reflect a small human effect, context dependence, IRS2 compensation, low frequency, publication bias, or overexpression artifact. A true molecular function does not guarantee clinical magnitude.

7.3 “Adverse” allele versus better response on high CHO

The small Marín crossover produced a direction opposite to the deterministic narrative. Greater room for improvement, adaptation, chance, simultaneous compositional changes, or crossover structure remain open. The result requires that dietary direction not be deduced from a biochemical lesion.

7.4 Intrinsic defect versus erasure in culture

Dunaif and Corbould found intrinsic or transferable components; Glintborg and Moreno-Asso showed in-vivo phenotypes that disappeared in myotubes. The contradiction may represent PCOS heterogeneity, culture protocols that erase real states, or genuine milieu dependence. It must not be resolved by calling everything “epigenetic.”

7.5 Normal pAKT versus altered flux

A comparable maximum response can coexist with a shifted EC50, abnormal temporal signaling, or a distal lesion in GLUT4 trafficking. Conversely, pAKT may change without altering uptake. D0-AR therefore prioritizes flux and uses signaling only for localization.

7.6 Resistant muscle versus responsive ovary

Selectivity is plausible, but the findings come from different studies and participants. The smallest future discriminant requires paired curves, not inference from receptor expression.

7.7 Later menopause versus worse cardiometabolic health

The 2026 GWAS is compatible with balanced or heterogeneous pleiotropy. It does not show a reproductive “longevity advantage” or two independent components. Age at menopause is subject to selection, surgery, and competing risks and does not measure oocyte quality, frailty, cancer, or survival.

8. Multiscale synthesis of the mechanism

Sequence level

rs1799817 does not alter the canonical protein sequence and remains outside perturbation until transancestry fine-mapping with cellular QTL or a splicing/expression assay prioritizes it. The INSR region remains an open locus. G972R does change IRS1 and has a plausible experimental interface, but its effect at endogenous dosage remains unproven.

Protein and signaling level

G972R could reduce signaling reserve through two nonexclusive routes: less phosphorylation of p85-recruiting motifs or abnormal association with the kinase domain that reduces autophosphorylation. The relevant consequence may be altered sensitivity and recovery rather than abolition of the maximum.

In parallel, kinases, phosphatases, lipids, or associated proteins may increase inhibitory serine phosphorylation or uncouple receptor from substrate without a sequence change. If the state depends on the milieu, it will disappear after washout; if persistence occurs, it could reflect slow turnover, organelles, metabolites, phosphoregulation, or chromatin. Only the last possibility would justify an epigenetic label, and it requires further evidence.

Cellular level

In muscle, the decisive result is how much glucose enters at a submaximal insulin dose. In beta cells, secretory reserve separates two responses:

• sufficient compensation: relatively stable glucose with high insulin;
• limited compensation: greater glycemic excursion for similar resistance.

HOMA mixes both dimensions and does not localize tissue. G972R can affect both in models, complicating interpretation of an oral load.

Tissue level

Muscle and adipose tissue determine much of glucose disposal and storage; liver regulates endogenous production and SHBG; theca converts part of insulin exposure into steroidogenesis. Their curves, thresholds, and branches differ. Restoring muscle signaling and directly activating ovarian signaling are therefore not equivalent interventions.

Systemic level and endocrine loop

A small reduction in peripheral reserve can be amplified:

lower peripheral action → greater insulin demand → greater ovarian exposure → altered androgens/SHBG → more adverse substrate distribution and adiposity → greater resistance.

The full loop is inferred. It can explain why a small genetic effect emerges only near an adiposity or hyperinsulinemia threshold, but also why reducing energy balance can improve the state regardless of genotype.

Life course

The relevant trajectory is cumulative time with resistance, hyperinsulinemia, glycemia, and vascular risk. PCOS does not metabolically end when reproductive life ends. Nevertheless, rs1799817 and G972R have not been shown to modify that trajectory, and genotype-guided diet has not been shown to change events. Maturity of the allelic longevity bridge is H0/not established.

9. Computational layer

Decision: skip.

No bioinformatics or BioNeMo analysis was run because no currently available output would change the variant, tissue, or experimental order. An expression atlas does not measure phosphorylation or flux. A static G972R structure does not decide endogenous effect or interaction with the environment. rs1799817 is synonymous and must not be modeled as a protein change. A dietary reanalysis without a second diet, exposure, PCOS, and a dynamic endpoint would repeat the causal error.

The layer will be reactivated only with one of two packages:

1. complete ancestry-specific GWAS statistics, matched LD, and dense QTL from a relevant cell type for multicausal fine-mapping/colocalization; or
2. individual-level data from at least two diets, genotype, PCOS, adherence, weight change, and clamp/FSIVGTT to estimate genotype×diet×PCOS.

The decision adds no H2 computational maturity. D0-AR remains smaller and more informative.

10. Primary hypothesis

L6-4-AR-H1 v1 — Contextual allelic persistence of IRS1 G972R

Falsifiable statement: under endogenous heterozygosity and one prespecified metabolic environment, IRS1 Gly/Arg will leave greater or longer-lasting post-washout functional impairment than Gly/Gly, and reciprocal editing will normalize that persistence in two independent female human genetic backgrounds.

Mechanism: G972R reduces IRS1–PI3K reserve and favors stabilization or slow reversal of an inhibitory state when the system operates near its limit.

Status: weakened, proposed.
Maturity: H0–H1.
Qualitative confidence: low; 0.20 in the adversarial ranking.

Evidence for:

• convergent biochemistry in transfection, rat L6, and recombinant fragments;
• small associations with metabolic severity in some subgroups;
• plausibility that reduced reserve modifies recovery.

Evidence against:

• absence as a major locus in recent GWAS;
• inconsistent clinical associations;
• lack of endogenous editing/rescue;
• PCOS muscle phenotypes that do not always persist in culture;
• paradoxical or null external dietary signals.

Specific predictions:

1. genotype×preexposure×time interaction on uptake, not merely a main effect;
2. greater post-washout area of deviation in Gly/Arg;
3. coherence with GLUT4 and, secondarily, pAKT/IRS1–p85;
4. Gly/Gly→Gly/Arg editing reproduces and Gly/Arg→Gly/Gly editing rescues;
5. consistent direction in two backgrounds;
6. an effect only in Arg/Arg does not pass the transportability gate.

11. Competing hypothesis

L6-4-AR-H2N v1 — Non-cell-autonomous state dependent on continuous exposure

Falsifiable statement: functional impairment will appear during metabolic exposure in both genotypes but will be rapidly erased after washout, so the extracellular milieu and systemic state will explain more than G972R or stable cellular memory.

Mechanism: circulating and multitissue signals—substrates, insulin, androgens, adipokines, or others—sustain the defect; when they are removed, the cell recovers function.

Status: strengthened and favored.
Maturity: H1.
Qualitative confidence: intermediate; 0.62 in the adversarial ranking.

Evidence for:

• in-vivo resistance and pioglitazone effects erased in myotubes;
• biopsy signature not retained in culture;
• distributed genetic architecture and weak candidates;
• exposure, adiposity, and baseline state explain substantial human heterogeneity.

Evidence against:

• Dunaif demonstrated a transferable and correctable defect;
• Corbould found intrinsic components;
• expansion/differentiation may erase true memory.

Predictions:

1. robust functional deterioration during continuous exposure;
2. rapid recovery, operationally by the first post-washout measurements at the latest;
3. equivalence between Gly/Gly and Gly/Arg;
4. reexposure reproduces the defect;
5. variation across milieus/backgrounds exceeds the allelic difference.

Nested competitor: L6-4-AR-H2P v1

If functional loss persists after washout with equivalent magnitude across alleles, a persistent acquired state independent of the allele wins. Its maturity is H0–H1. It will not be called phosphoregulatory, epigenetic, or lipid-mediated until those mechanisms are distinguished. The ED-H2L explanation predicts that persistence will track intracellular palmitate and derived lipids and disappear when that burden is equalized or depleted.

12. Translational hypothesis

L6-4-AR-HT1 v1 — Reliability of a dynamic phenotype

Falsifiable statement: a single challenge–recovery parameter, frozen before recruitment, will reach prespecified reliability; only then can it be tested for added external prediction beyond adiposity, baseline sensitivity, consumed diet, and weight change.

Role: research metrology hypothesis, not a clinical biomarker.
Status/maturity: weakened, H0; HUMAN_QA_REQUIRED.

Predictions:

1. ICC ≥0.75 with lower bound ≥0.65, technical CV ≤10%, and batch drift below 10% of the minimum experimental change;
2. the parameter is defined before validation and not learned in the same sample;
3. it improves external error/calibration beyond covariates;
4. it does not arise from mathematical coupling to the endpoint or reduce to BMI/fitness;
5. rs1799817 adds no performance; G972R could add only if AR-H1 passes D0-AR.

Kill criteria: low reliability, drift, in-sample-only signal, loss after controlling adiposity/exposure, exclusive prediction of HOMA/weight, or operational infeasibility.

Parked systemic hypothesis: L6-4-AR-H3 v1

Genetic PCOS–menopause and PCOS–T2D/CAD covariance may be explained better by partially independent components than by a single pleiotropic factor. The 2026 evidence justifies testing this but does not establish the partition. It remains H0–H1, weakened, and parked until an independent female cohort, competing-risk data, and ancestral replication are available.

13. Falsifiable predictions and kill criteria

Kill criteria for the dietary claim

Even if AR-H1 is positive in cells, the claim “G972R modifies a diet in PCOS” is killed if an adequately powered human crossover finds the genotype×diet×PCOS term within equivalence, if the effect is explained by weight change/exposure, or if there is no coherent dynamic measure. A cellular result never substitutes for this gate.

14. Discriminating experiment

D0-AR — continuous exposure, acquired persistence, or allelic persistence

Nature: preclinical proposal. It does not represent a diet, prescribe, or authorize human recruitment.

D-1: blinded qualification

Two unrelated female human iPSC lines are used. In each background, scarless reciprocal Gly/Gly ↔ Gly/Arg editing creates two clones per edit, mock controls, and revertants. Arg/Arg is only a secondary control. Allelic phase, local copy number, karyotype/low-pass genome, priority off-targets, identity, and mycoplasma must be verified.

The myotube must express INSR, IRS1, AKT, TBC1D4, and GLUT4; show stable fusion/maturity; and increase 2-DG uptake by at least 1.4× with insulin, with within-assay CV ≤15%. A curve is estimated and one common dose near EC40–EC60 is frozen, not a genotype-specific dose.

The challenge is selected in parental/mock cells without revealing the allelic contrast: palmitate–BSA at 0.2, 0.3, or 0.4 mM for 12 h. The lowest concentration that reduces stimulated uptake by 20–35%, with viability affected by <10% and without material stress/toxicity, is frozen. If none passes, the branch fails; a different mixture is not selected post hoc.

Confirmatory D0-AR

Three arms:

1. time-matched vehicle;
2. continuous exposure;
3. 12-h pulse, validated washout, and recovery.

Time points: 0, 6, 24, and 48 h. Each includes basal conditions and an acute challenge with the common submaximal dose.

Primary endpoint: area under the deviation in stimulated 2-deoxy-D-glucose uptake over 0–48 h relative to time-matched vehicle. Nonspecific uptake is defined with cytochalasin B, and the readout is normalized to cellular mass and myotube count.

Confirmatory readouts: surface GLUT4; pAKT Ser473/Thr308; pTBC1D4 Thr642; IRS1–p85; IRS1 Tyr612/Ser312; total proteins; glycogen. None rescues a null functional endpoint.

Internal dose: 5% [U-13C16]palmitate within the total, after showing that it does not alter the phenotype; LC–MS of labeled palmitate, TAG, DAG, and ceramides in cell, medium, and washout. If burden differs by >15% by genotype or tracks recovery, the finding is classified first as substrate handling/retention.

Experimental unit: independent differentiation; three wells are averaged. Wells and fields are not biological n. Two clones are not equivalent to two people; attribution requires two backgrounds.

Size and analysis: ten initial blocks, blinded variance re-estimation after five, maximum fourteen. Minimum experimental change 20% and equivalence margin ±15%, conditional on D-1. Mixed model:

response ~ genotype × arm × time + background + rescue terms + hierarchical effects of clone, block, and plate.

Equivalence requires the entire CI90% within the margin; “not significant” is not equivalent to null. AR-H1 requires a CI95% exceeding the minimum change, concordant direction by background, and reciprocal rescue.

Decision matrix

Subsequent sequence

If AR-H2N wins, exposure–withdrawal is replicated in primary myotubes and myobundles and the extracellular factor is identified causally. If AR-H2P/ED-H2L wins, time-resolved phosphoproteomics/lipidomics and at most two orthogonal rescues are used. If AR-H1 wins, it is replicated in a third background and in 3D before beta cells, theca, or humans.

Only after allelic function, rescue, and replication would a controlled-feeding, isocaloric, weight-stable crossover be justified, with women with and without PCOS prospectively recruited by G972R, clamp/FSIVGTT, and the genotype×diet×PCOS estimand.

15. Biomarkers and stratification

There is no validated biomarker of INSR/IRS1 dietary response in PCOS.

Research candidates, not clinical biomarkers

• Genotype: rs1799817 is unsuitable; G972R is only an experimental candidate.
• Dynamic phenotype: sensitivity, compensation, and recovery might integrate more biology than a SNP, but reliability is unproven.
• Proximal state: pINSR/pIRS1/pAKT or a phosphoregulatory signature cannot be used without repeatability, relevant tissue, and a relationship to flux.
• Internal dose: tracers and lipidomics are mechanistic readouts, not clinical classifiers.
• Systemic stratification: adiposity, baseline sensitivity, consumed exposure, weight change, phase/cycle, sleep/timing, medications, and thyroid function must be modeled as context; they must not be fused into an unvalidated score.

The correct sequence is: repeatability → construct validity → out-of-sample gain → external replication → human evaluation of utility. No layer may be skipped on molecular plausibility.

16. Individual variability

Variability may arise from:

1. ancestry and LD, which change which variant tags a locus;
2. PCOS phenotype and diagnostic criteria;
3. central adiposity, muscle mass, and baseline state;
4. beta-cell compensation;
5. muscle–adipose–liver–ovary selectivity;
6. hormonal and androgenic milieu;
7. meal timing, sleep, and circadian disruption;
8. microbiota, absorption, and substrate kinetics;
9. exposure/adherence and weight change;
10. polygenic architecture and cellular regulation;
11. medications, thyroid function, and other comorbidities;
12. measurement error and regression to the mean.

This does not mean that everything matters equally. The current ranking prioritizes exposure/functional state and distributed architecture; G972R remains a possible contextual modifier; rs1799817 has low priority. Ancestral diversity is not addressed by stratifying on ethnic labels, but through genotype, principal components, LD, and replication.

17. Pharma relevance and maturity

The Pharma opportunity is a conditional discovery program, not a ready target.

Tissue-selectivity risk

A systemic agent that increases insulin signaling could improve muscle and simultaneously amplify thecal steroidogenesis. Before a target is considered, improved muscle flux without increased androgen output must be demonstrated in a relevant ovarian model. Cytotoxicity, mitochondrial stress, altered contractility, and proliferative signaling must also be excluded.

Maturity

• AR-H2N: H1.
• AR-H2P: H0–H1.
• AR-H1: H0–H1.
• ED-H2L: H0.
• AR-H3: H0–H1, parked.
• AR-HT1: H0, HUMAN_QA_REQUIRED.
• PCOS/IR→T2D/events bridge: H3 observational.
• Allele→diet→healthspan/longevity: H0/not established.

No branch reaches H5 or partnering readiness. Every human protocol requires ethical review and human QA.

18. Limitations

1. The candidate literature is small, old, and vulnerable to selection, multiplicity, HWE problems, publication bias, and heterogeneous PCOS definitions.
2. GWAS identify loci and pleiotropy, not cells, causal variants, or dietary mechanisms.
3. Assignment of the 19p13 region to INSR remains weak; rs1799817 has no allele-specific function.
4. G972R evidence relies on overexpression, nonhuman lines, and fragments; endogenous female editing is missing.
5. Islet studies include only two donors per genotype; islet replicates do not increase human n.
6. Human theca provides a plausible bridge, but the sample size of one key study was not recovered, and there are no paired multitissue measurements.
7. HOMA dominates the human literature and does not separate sensitivity from secretion.
8. The verified set contains no chronic PCOS RCT with two diets, the candidate genotype, and a dynamic endpoint.
9. Palmitate in D0-AR is a cellular challenge, not a diet or a PCOS-specific state.
10. iPSC myotubes may be immature, and culture may erase true states.
11. Persistence for up to 48 h does not prove strict hysteresis, chromatin involvement, or mitotic inheritance.
12. Two backgrounds detect obvious artifacts but do not estimate population distribution or ancestry interaction.
13. Isotope retention localizes internal dose but requires perturbation to establish causality.
14. Event cohorts do not contain genotype/diet and are subject to confounding, coding limitations, and selection.
15. The 2026 GWAS is predominantly European, and the pleiotropic partition is not validated.
16. Mexico/Latin America is underrepresented; frequencies or effects are not transported without evidence.
17. Later menopause, ovulation, weight, HOMA, and pAKT are not longevity proxies.
18. This report does not evaluate individual prescription or clinical utility.

19. Conclusions

1. rs1799817 does not explain why a diet works differently. It is synonymous, inconsistent, and lacks allele-specific function; the INSR GWAS region does not rehabilitate it.
2. G972R retains plausibility, not clinical validation. Its biochemistry justifies an endogenous experiment, not diet selection.
3. State dominates the current ranking. Human ex-vivo evidence favors a defect sustained by the extracellular milieu or, secondarily, an acquired state; allelic persistence is less likely but falsifiable.
4. Tissue selectivity is the central endocrine bridge. Peripheral resistance and hyperinsulinemia can coexist with a thecal response, amplifying androgens without a causal SNP.
5. “Same diet” is causally incomplete. Consumption, internal dose, weight, baseline sensitivity, and temporality must be equalized before genotype×diet is estimated.
6. There is no chronic PCOS-specific evidence. The acute pilot and non-PCOS trials do not cover the decisive link.
7. Healthspan currently belongs to the phenotype. PCOS/insulin resistance connects to T2D and events; the alleles do not connect to longevity.
8. D0-AR is the rational next step. One experiment distinguishes continuous exposure, acquired persistence, lipid retention, and allelic persistence with explicit rescue and failure criteria.
9. A null result will be informative. Rapid recovery and allelic equivalence would eliminate one concrete version of G972R×environment and strengthen the non-cell-autonomous state.
10. Maturity requires restraint. No hypothesis authorizes personalized diet, a clinical biomarker, or a ready Pharma program.

20. References

1. Mukherjee S, et al. Association of polymorphism in the insulin receptor gene with PCOS. European Journal of Endocrinology. 2009;160:855–862. DOI: 10.1530/EJE-08-0932. PMID: 19211708.
2. Daghestani MH. INSR rs1799817 and PCOS susceptibility in Saudi women. Journal of Medical Biochemistry. 2020;39:149–159. DOI: 10.2478/jomb-2019-0023. PMID: 33033446.
3. Xu Y, et al. Association study of insulin receptor gene polymorphisms with PCOS in Han Chinese families. Reproductive Biology and Endocrinology. 2011;9:76. DOI: 10.1186/1477-7827-9-76. PMID: 21645371.
4. Škrgatić L, et al. INSR, IRS1 and INS-VNTR variants in women with PCOS. Collegium Antropologicum. 2013;37:141–146. PMID: 23697264.
5. Rasool SUA, et al. INSR rs1799817 in women with PCOS. International Journal of Endocrinology. 2021:7522487. DOI: 10.1155/2021/7522487. PMID: 34912452.
6. Rasool SUA, et al. IRS1 rs1801278 and metabolic features in PCOS. Genes. 2022;13:1463. DOI: 10.3390/genes13081463. PMID: 36011374.
7. Shi Y, et al. Genome-wide association study identifies eight new risk loci for polycystic ovary syndrome. Nature Genetics. 2012;44:1020–1025. DOI: 10.1038/ng.2384. PMID: 22885925.
8. Day F, et al. Large-scale genome-wide meta-analysis of polycystic ovary syndrome suggests shared genetic architecture across diagnostic criteria. PLoS Genetics. 2018;14:e1007813. DOI: 10.1371/journal.pgen.1007813. PMID: 30566500.
9. Zhao H, et al. Multi-ancestry genome-wide association analyses of polycystic ovary syndrome. Nature Genetics. 2025;57:2669–2681. DOI: 10.1038/s41588-025-02393-x. PMID: 41188533.
10. Moolhuijsen LME, et al. Genomic analyses implicate hormonal and metabolic dysregulation in polycystic ovary syndrome. Nature Genetics. 2026;58:1040–1050. DOI: 10.1038/s41588-026-02543-9. PMID: 42026183.
11. Ait El Mkadem S, et al. Association of the Gly972Arg IRS1 polymorphism with insulin resistance in PCOS. Diabetes. 2001;50:2164–2168. DOI: 10.2337/diabetes.50.9.2164. PMID: 11522686.
12. Villuendas G, et al. Association between IRS1/IRS2 variants and insulin resistance in hyperandrogenic women. Human Reproduction. 2005;20:3184–3191. DOI: 10.1093/humrep/dei205. PMID: 16037106.
13. Dunaif A, et al. Excessive insulin receptor serine phosphorylation in cultured fibroblasts and in skeletal muscle. Journal of Clinical Investigation. 1995;96:801–810. DOI: 10.1172/JCI118126. PMID: 7635975.
14. Corbould A, et al. Insulin resistance in the skeletal muscle of women with PCOS involves intrinsic and acquired defects in insulin signaling. American Journal of Physiology-Endocrinology and Metabolism. 2005;288:E1047–E1054. DOI: 10.1152/ajpendo.00361.2004. PMID: 15613682.
15. Eriksen M, et al. Insulin resistance is not conserved in myotubes established from women with PCOS. PLoS One. 2010;5:e14469. DOI: 10.1371/journal.pone.0014469. PMID: 21209881.
16. Moreno-Asso A, et al. Non-cell autonomous mechanisms control mitochondrial gene dysregulation in skeletal muscle of women with PCOS. Journal of Molecular Endocrinology. 2022;68:103–116. DOI: 10.1530/JME-21-0212. PMID: 34752415.
17. Nestler JE, et al. Insulin stimulates testosterone biosynthesis by human thecal cells through its own receptor. Journal of Clinical Endocrinology & Metabolism. 1998;83. DOI: 10.1210/jcem.83.6.4886.
18. Yen HW, et al. Differential expression of insulin receptor substrates in theca and granulosa cells from women with PCOS. Molecular Human Reproduction. 2004. DOI: 10.1093/molehr/gah066.
19. Tosi F, et al. Hyperinsulinemia amplifies ovarian steroid response in women with PCOS. Journal of Clinical Endocrinology & Metabolism. 2012. DOI: 10.1210/jc.2011-2939.
20. Almind K, et al. A common amino acid polymorphism in insulin receptor substrate-1 causes impaired insulin signaling. Journal of Clinical Investigation. 1996;97:2569–2575. DOI: 10.1172/JCI118705.
21. Hribal ML, et al. The common Gly972Arg polymorphism in insulin receptor substrate-1 affects glucose metabolism in skeletal muscle cells. Journal of Clinical Endocrinology & Metabolism. 2000;85:2004–2013. DOI: 10.1210/jcem.85.5.6608. PMID: 10843189.
22. McGettrick AJ, et al. Human IRS-1 polymorphism G972R causes IRS-1 to associate with the insulin receptor and inhibit receptor autophosphorylation. Journal of Biological Chemistry. 2005;280:6441–6446. DOI: 10.1074/jbc.M412300200. PMID: 15590636.
23. Porzio O, et al. The Gly972Arg IRS1 variant impairs insulin secretion in pancreatic beta cells. Journal of Clinical Investigation. 1999;104:357–364. DOI: 10.1172/JCI5870. PMID: 10430617.
24. Marchetti P, et al. Insulin secretory defects in human islets carrying the Gly972Arg IRS1 polymorphism. Diabetes. 2002;51:1419–1424. DOI: 10.2337/diabetes.51.5.1419. PMID: 11978638.
25. Sir-Petermann T, et al. Gly972Arg polymorphism of IRS-1 in women with polycystic ovary syndrome. Nutrition. 2004;20:905–910. DOI: 10.1016/j.nut.2004.08.017. PMID: 15474880.
26. Marín C, et al. IRS1 Gly972Arg modifies insulin sensitivity after diets differing in fat and carbohydrate. Molecular Nutrition & Food Research. 2011;55:368–378. DOI: 10.1002/mnfr.201000235. PMID: 20824664.
27. Qi Q, et al. IRS1 rs2943641, dietary macronutrients and changes in insulin resistance and weight loss. Circulation. 2011;124:563–571. DOI: 10.1161/CIRCULATIONAHA.111.025767. PMID: 21747052.
28. Gardner CD, et al. Effect of a nutrient genotype score on response to a healthy low-fat versus healthy low-carbohydrate diet. Nature Communications. 2023. DOI: 10.1038/s41467-023-41969-1. PMID: 37813841.
29. Gambineri A, et al. Diabetes mellitus in polycystic ovary syndrome: incidence, prevalence and predictors in a long-term prospective study. Diabetes. 2012;61:2369–2374. DOI: 10.2337/db11-1360. PMID: 22698921.
30. Ryu KJ, et al. Risk of cardiovascular and cerebrovascular disease in Korean women with PCOS. Scientific Reports. 2024. DOI: 10.1038/s41598-023-50650-y. PMID: 38212642.
31. Day F, et al. Causal mechanisms and balancing selection inferred from genetic associations with polycystic ovary syndrome. Nature Communications. 2015;6:8464. DOI: 10.1038/ncomms9464. PMID: 26416764.
32. Iovino S, et al. Genetic insulin resistance is recapitulated in human iPSC-derived skeletal muscle cells. Proceedings of the National Academy of Sciences USA. 2016;113:1889–1894. DOI: 10.1073/pnas.1525665113. PMID: 26831110.
33. Mäkinen S, et al. Palmitate-induced insulin resistance in primary human myotubes. Endocrine Connections. 2017;6:331–339. DOI: 10.1530/EC-17-0039. PMID: 28584168.
34. Green CJ, et al. Palmitate-induced insulin resistance and donor-dependent metabolic responses in human myotubes. Journal of Physiology. 2013;591:4621–4635. DOI: 10.1113/jphysiol.2013.251421. PMID: 23774280.
35. Henry RR, et al. Insulin action and glucose transport in cultured human skeletal muscle cells. Diabetes. 1995;44:936–946. PMID: 7622000.
36. Kondash ME, et al. Insulin-responsive human skeletal muscle myobundles. Tissue Engineering and Regenerative Medicine. 2020;17:801–813. DOI: 10.1007/s13770-020-00242-y. PMID: 32200516.

REPORT_EN closure status

• Complete canonical structure: yes.
• Three principal hypotheses—primary, competing, and translational—with lineage, predictions, and kill criteria: yes.
• Adversarial and null evidence preserved: yes.
• Smallest experiment, controls, size, analysis, and advancement rules: yes.
• Pharma relevance and maturity: yes, conditional.
• Complete primary references for every decisive claim: yes.
• Human, human ex-vivo, in-vitro, computational, and inferred evidence separated: yes.
• Prescription, current Lua data, product proposals, personas, and marketing: absent.
• Next stage: BILINGUAL_QA.