NAD+ and ovarian aging: SIRT1/SIRT3, compartments, and mitochondrial quality

Stage: REPORT_EN
Evidence cutoff: 2026-09-09
Project: L9-1
Classification: NAD+ and ovarian function
Scope: mechanistic and translational research; this is not a prescription, diagnosis, or individual recommendation.

1. Executive scientific abstract

The popular hypothesis—that NAD+ declines with age, switches off SIRT1/SIRT3, damages mitochondria, and thereby ages the ovary—is biochemically plausible but has not been demonstrated as a causal chain in women. The closest human evidence is fragmented across different studies. Metabolomics of immature oocytes obtained after assisted reproduction suggested less NAD+ with age, but relied on three pools of ten oocytes per category, and most metabolites did not survive multiple-testing correction. A separate cohort of 72 women found a functional signature compatible with lower SIRT3/GDH function in granulosa and cumulus cells with advanced age or reduced reserve, but did not measure NAD in the same material. No human study establishes the temporal sequence compartmental NAD → SIRT1/SIRT3 activity → mitochondrial quality-control flux → follicular function.

In rodents, increasing NAD+ or perturbing its synthesis and consumption changes reserve, ovulation, oocyte competence, or embryo development. This perturbability is real, but it does not identify the mechanism. Experiments commonly use systemic precursors, constitutive knockouts, few animals, and proxies such as ROS, membrane potential, mtDNA, or PINK1/Parkin proteins. Null findings are also informative: whole-ovary NAD(H) may not decline even when oocyte NAD(P)H changes; NMN showed a non-monotonic reproductive response; oocyte SIRT1 was dispensable for reserve and maturation but required for post-fertilization competence during aging; SIRT3 was dispensable in young and obese mice in one study, whereas its loss accelerated senescence in another aged model; and CD38 absence increased reserve mainly through neonatal programming without preserving a fecundity advantage when reproduction began late.

The most rigorous synthesis is a bidirectional, compartmental, follicular-state-dependent circuit, not a linear cascade. Age, follicular state, atresia, mitochondrial damage, DNA-damage response, inflammation, and cellular composition can alter both NAD supply/consumption and mitochondrial function. SIRT1 and SIRT3 are better interpreted as contextual buffers on distinct branches: nuclear/cytoplasmic SIRT1 participates in survival and maternal programs, whereas mitochondrial SIRT3 regulates matrix-enzyme acetylation, redox, and respiratory capacity. Redox pathways, dehydrogenases, and consumers such as PARP/CD38 may respond to NAD independently of both sirtuins.

Two recent findings refine this architecture. First, a 2026 mouse preprint shows that an intact cumulus–oocyte complex favors conversion of exogenous NMN into oocyte NAD+; denudation or carbenoxolone reduces that conversion, and NMN does not rescue the oocyte without cumulus. This does not identify the transferred molecule, because cumulus also supplies adenylates, energy, signals, and architecture. In 23 human cumulus samples from the same work, NAD+ did not decline with age. Second, SLC25A51 transports intact NAD+ into the matrix in non-ovarian cells and can uncouple total cellular NAD from mitochondrial NAD and respiration, but its follicular function, age regulation, and epistasis with SIRT3 have not been demonstrated.

The project's central scientific delta is to convert these gaps into a falsification sequence. The first gate is not another precursor administration: a tracer must be loaded exclusively into donor cumulus, extracellular tracer must be removed, and an identified species must appear in the recipient oocyte in a Cx37/GJA4-dependent manner. Only then should investigators ask whether SLC25A51 limits matrix NAD, whether rescue requires SIRT3 or SIRT1 catalysis, and whether mitochondria-to-lysosome flux changes. Selection/state is the leading explanation for human associations; damage or atresia first, with NAD as an amplifier, is the best-supported biological competitor. The hypothesis of early sirtuin-independent rescue remains open.

The link to women's health and longevity remains broken. Reproductive function, ovarian endocrine output, distal-organ health, and survival are different outcomes. Human genetics of age at menopause shows potential bone and metabolic benefits alongside greater hormone-dependent cancer risk and no clear causal effect on longevity. There is no evidence that correcting ovarian NAD, SIRT1, SIRT3, or a mitochondrial proxy increases female healthspan or lifespan. Integrated maturity is H0–H1; no target is ready for an indication, clinical biomarker, or pharmaceutical program.

2. Scientific question and relevance

The exact question is: in natural ovarian aging and within the same follicular cell type and state, does a reduction in nuclear or mitochondrial NAD precede and limit the catalytic activity of SIRT1 or SIRT3, impair mitochondrial-maintenance flux, and reduce a function of the granulosa–cumulus–oocyte unit; or does low NAD appear after damage, atresia, consumption, or cellular selection?

Relevance to women's health does not follow automatically from fertility. The ovary participates in four domains that must remain separate:

1. Follicular quantity: initial reserve, recruitment, and atresia.
2. Gamete and maternal competence: meiosis, fertilization, and early embryo development.
3. Endocrine function: ovulation, steroidogenesis, and inhibins.
4. Systemic exposure and life course: potentially opposing organ-specific effects and shared systemic causes.

An NAD–mitochondria mechanism could affect one domain without changing the others. Oocyte SIRT1 and CD38 already show this modularity in mice. Therefore, AMH, recovered oocytes, blastocyst, age at menopause, and a mitochondrial marker are not accepted as substitutes for healthspan or longevity.

3. Scope, population, and life stage

The target human population is women aged 35–55 years undergoing natural reproductive aging, from late reproductive stages STRAW+10 −3b/−3a through transition −2/−1 and early postmenopause +1a/+1b. Women aged 20–34 years provide a range reference, not a substitute for an intraindividual trajectory. The human statistical unit is the donor; follicles, COC, oocytes, and cells are nested. The preclinical causal unit is the animal, not the number of oocytes, images, or organelles.

The following contexts remain separate:

• natural aging versus primary ovarian insufficiency, gonadotoxic injury, oophorectomy, and autoimmune or genetic etiologies;
• assisted-reproduction samples versus the general population, because infertility, stimulation, luteinization, and selection of recoverable follicles alter the estimand;
• PCOS, endometriosis, complicated obesity, diabetes, and uncontrolled thyroid disease as modifiers, not equivalents of physiological aging;
• pregnancy, postpartum, and exogenous hormone exposure as distinct endocrine states;
• human, animal, ex vivo, in vitro, computational, and inferred evidence, without automatically transporting magnitude or causality across levels.

The project uses no current Lua user data. Predominantly European, North American, or ART-center evidence is not assumed to represent Mexican or Latin American populations. Ancestry, adiposity, metabolic exposure, clinical protocols, and access to assisted reproduction are transportability variables, not post hoc explanations.

4. Background knowledge and mechanistic map

4.1 NAD is not a single pool

NAD+ has two relevant and non-interchangeable functions. As the NAD+/NADH pair, it transfers reducing equivalents through glycolysis, the TCA cycle, and the respiratory chain. As a cosubstrate, it is consumed by sirtuins, PARP, CD38, and other enzymes. Its final abundance reflects de novo synthesis from tryptophan, Preiss–Handler and nicotinamide/NR/NMN salvage pathways, transport, redox interconversion, and consumption.

Nuclear, cytosolic, and mitochondrial compartments can diverge. Whole-ovary LC–MS mixes oocyte, granulosa, theca, stroma, endothelium, immune cells, and different follicular stages. NAD(P)H autofluorescence mixes NADH and NADPH and does not measure NAD+. A normal bulk measurement therefore does not exclude matrix deficiency, and a local signal does not establish decline of the entire organ.

SLC25A51/MCART1 is the best-established mechanism for importing intact NAD+ into the mammalian mitochondrial matrix. In non-ovarian cells, its loss reduces mitochondrial NAD, TCA flux, complex I function, and respiration with less effect on total NAD; WT rescue restores function, whereas K91A and R278A mutants do not. Conservation of this function in the ovary is an H0 hypothesis, not an established fact.

4.2 SIRT1 and SIRT3: branches, not a serial relay

SIRT1 is located mainly in the nucleus/cytoplasm. Candidate substrates include p53, FOXO, and PGC-1α, allowing modulation of survival, stress, transcription, biogenesis, and the maternal program. In recent porcine atresia, lower NMNAT1/NAD was associated with lower SIRT1 deacetylation, hyperacetylated p53, and apoptosis; this demonstrates a pathway in porcine granulosa, not its order in human aging. In mice, deletion of the SIRT1 catalytic domain in growing oocytes left reserve, maturation, ATP, and young fertility normal but worsened cleavage and fecundity with age. The evidence favors a resilience/post-fertilization competence function rather than a universal reserve gate.

SIRT3 acts in the matrix and deacetylates metabolic and antioxidant enzymes. The closest human association found greater acetylation and lower GDH activity in follicular cells from women with advanced age or reduced reserve. In preclinical models, SIRT3 modifies ROS, respiration, and competence under IVM, aging, AGEs, or circadian disruption; yet its loss can also be phenotypically tolerated in young mice and under diet-induced obesity. An age-dependent phenotype is compatible with a threshold or capacity reserve, but does not demonstrate redundancy with SIRT1. Parallelism, compensation, and epistasis are different claims.

4.3 Mitochondrial quality requires flux

Four modules are distinguished:

• energy/redox: NAD+/NADH, OCR, coupling, TCA, and ATP;
• damage/defense: ROS, SOD2, peroxidation, mtDNA, and repair;
• quantity/biogenesis: mass, mtDNA copy number, and protein/organelle synthesis;
• selection/turnover: fusion–fission, lysosomal delivery, mitophagy, and replacement.

ROS, potential, mtDNA, TOM20, PINK1, Parkin, LC3, or morphology alone do not demonstrate quality control. Marker accumulation may mean induction or blockade. A time series with a ratiometric reporter, lysosomal manipulation, and a functional readout is required. No source integrates that flux with compartmental NAD and SIRT1/SIRT3 epistasis in natural aging.

4.4 Revised causal mechanism

age/stage + inflammation + DDR + metabolic/circadian stress
          │                    │
          │                    ├── PARP/CD38/other consumers
          │                    │
          ├── state/atresia ───┴── NAD synthesis and salvage
          │                                  │
          │               ┌──────────────────┴──────────────────┐
          │               │                                     │
          │        nuclear/cytosolic NAD              mitochondrial NAD
          │               │                          (candidate SLC25A51)
          │             SIRT1                                  SIRT3
          │      p53/maternal program              matrix/redox/respiration
          │               └──────────────────┬──────────────────┘
          │                                  │
          └──────── damage/turnover/apoptosis/energy ───────────┘
                                             │
                                  granulosa/cumulus ↔ oocyte
                                             │
                 reserve | steroidogenesis | embryo competence
                                             │
                  endocrine exposure and non-univocal organ outcomes

Causality can run in both directions. Primary loss of salvage may initiate deterioration; CoQ/ETC injury or atresia may lower NAD secondarily; each may reinforce the other. The experimental question is not whether NAD “participates,” but which event occurs first, in which compartment, and which rescue component requires a catalytically competent sirtuin.

5. Evidence method

All project artifacts were integrated: scoping, evidence map, source-by-source verification, mechanistic synthesis, computational decision, hypothesis generation, adversarial review, and experiment design. Primary human and genetic studies, animal or cellular perturbations with rescue, null results, and sources capable of refuting the mechanism were prioritized. Narrative reviews were not used as final evidence.

Evidence classes are HUM, human observational; GEN, human genetic; ANI, animal in vivo; EXV, primary ex vivo material; INV, in vitro; COMP, computational; and INF, unmeasured inference across levels. Certainty reflects design, biological unit, metrology, bias, consistency, and transportability, not statistical significance alone.

Sirtuin activity claims required catalytic deacetylation or the state of a qualified substrate; RNA/protein was insufficient. Mito-QC required temporal elimination/biogenesis flux with blockade controls; static inventory was insufficient. Donor or animal was treated as n; nested cells, embryos, and organelles did not increase the independent count.

Recent and uncertain 2025–2026 sources were checked against their articles, preprint, PubMed/PMC, or canonical DOI record. Mihalas et al. remains explicitly a non-peer-reviewed preprint. Piasecki et al. reports operational heterogeneity by FAMS, not rejuvenation or flux. Wang et al.'s porcine atresia study demonstrates association and perturbation in that model, not direction in human aging. Ying et al.'s circadian-disruption study approximates several links, but uses young rats under an 18:6 photoperiod and does not demonstrate that SIRT3 is required for rescue.

6. Evidence map

Human evidence supports only separate, selected signals for oocyte NAD and follicular SIRT3 function. Animal evidence demonstrates system perturbability and multiple convergence points. The inference joining those levels remains the weakest link.

7. Contradictory evidence and null findings

Null findings are not noise to remove; they define the mechanism that can still survive.

1. Oocyte versus bulk ovary. Lower oocyte NAD(P)H coexisted with no detectable decline in whole-ovary NAD(H). Explanations are compartment, NADPH contribution to autofluorescence, tissue composition, or power. Absolute NAD+/NADH in cells from the same ovary plus matrix/nuclear sensors would discriminate them.
2. Discordant human NAD. Smits suggested lower oocyte NAD+; the 2026 preprint did not observe lower NAD+ in cumulus from 23 women, although NAM, meNAM, and adenosine did decline. Different tissue, preanalytics, stimulation, and selection compete with true non-conservation.
3. SIRT3 dispensability and harm when absent. E10/E11 cannot be reconciled by invoking compensation without measuring it. Age/threshold, background, colony, diet, power, or KO construction remain rivals.
4. SIRT1 affects quality, not quantity. Catalytic oocyte loss impairs embryo development with age without changing reserve, ovulation, or fertilization. Reserve and competence are not one scale.
5. CD38 changes initial state, not necessarily rate. Larger reserve in KO animals may arise from prolonged neonatal formation; the advantage does not persist when testing begins late. High NAD and high reserve do not demonstrate slower adult aging.
6. NMN is neither monotonic nor specific. A lower dose improved live births while a higher one did not; rescue under sirtinol had incomplete target qualification; the precursor changes redox, dehydrogenases, PARP/CD38, and sirtuins simultaneously.
7. Damage may precede NAD. Complex II, CoQ, AGE, photoperiod, or atresia injury produces the same ROS, respiration, and functional pattern. This permits a loop but contradicts NAD as a universal initiator.
8. Organelle heterogeneity is not rejuvenation. NMN moved mitochondrial subpopulations in opposite directions. A “normalized” mean or bin does not demonstrate better function or selective elimination.
9. Mito-QC has not been measured adequately. PINK1/Parkin/LC3, mtDNA, potential, or crista changes can reflect inventory, damage, blockade, or selection.
10. Later menopause does not imply uniform longevity. Human genetics shows opposing organ-specific effects and no clear causal lifespan signal.

The current hierarchy is: (1) selection/state for human associations; (2) damage/atresia first as a biological mechanism; (3) redox/sirtuin-independent rescue; (4) cumulus support and SLC25A51 partitioning as discovery hypotheses; and (5) the simple serial chain as a refuted formulation.

8. Multiscale mechanism synthesis

8.1 From somatic metabolism to oocyte NAD

The oocyte is not metabolically autonomous. Cumulus supports its environment through transzonal projections and gap junctions. E22 shows that COC integrity is required to convert exogenous NMN efficiently into oocyte NAD+. The article's strongest explanation is somatic–germline cooperation, but it does not identify the cargo. If labeled NMN enters the oocyte but is not converted, ATP/adenylates or NMNAT activity may be limiting; if labeled NAD appears after only cumulus is loaded, precursor, product, or an intermediate may have transferred; if the effect appears only with denudation/CBX, nonspecific injury or blockade may explain it.

The oocyte Cx37/GJA4 interface and somatic-network Cx43/GJA1 enable a more specific experiment. Label must be donor-restricted, absent from the last wash, and cross to the recipient with coherent precursor→product kinetics. Only connexin-specific loss with viability, ATP, cumulus load, and architecture inside margins can assign gap-junction dependence. Until then, the correct formulation is “cumulus support for oocyte NAD synthesis,” not “NAD transfer.”

8.2 From cellular NAD to matrix NAD

If the oocyte receives or synthesizes cytosolic NAD, matrix entry is the next potential bottleneck. SLC25A51 can explain why total NAD fails to track SIRT3/OCR. OCR, however, also responds to dehydrogenases, TCA, and complex I. The required sequence is SLC25A51 loss → matrix NAD↓ with equivalent total NAD and mass → OCR↓ → rescue by WT but not transport mutants. SIRT3 is added only afterward. An OCR change without a matrix-NAD change opens another hypothesis; it does not validate the transporter.

8.3 From matrix NAD to SIRT3 and function

Only after establishing the matrix boundary can the fraction requiring SIRT3 be estimated. Loss should be acute to limit developmental compensation, and rescue must compare SIRT3 WT with catalytically inactive H248Y at equivalent expression/localization. GDH, LCAD, or SOD2 deacetylation is interpreted according to tissue expression and activity; global acetylation is insufficient. The primary estimand is the fraction of NAD functional rescue lost when SIRT3 catalysis is removed.

8.4 The somatic/maternal SIRT1 branch

SIRT1 should not be inserted into a SIRT3 chain. It is tested separately in cumulus/granulosa and oocyte. Acute loss and SIRT1 WT versus H363Y rescue distinguish catalysis from structural functions. Acetylated p53 and a second nuclear substrate qualify engagement. Function must separate somatic apoptosis/steroidogenesis from oocyte-to-embryo competence. Correcting the oocyte matrix may rescue respiration yet still fail when cumulus is compromised; that would confirm convergent branches, not redundancy.

8.5 From organelle to ovary

Atresia, recruitment, reserve, steroidogenesis, and embryo competence follow different architectures. Age may increase the proportion of compromised COC without changing NAD inside each healthy state; conversely, low NAD may mediate transition in some states. To avoid adjusting away a mediator as though it were a confounder, total age effect and within-state effect must both be reported, together with recovery weighting. Selection of “analyzable oocytes” is part of the phenomenon and must remain in the denominator.

8.6 From ovary to health and longevity

Three routes compete:

• reproductive mediation: improved gamete competence or early development;
• ovarian endocrine mediation: changes in ovulation/E2/P4/inhibins affecting a specific organ;
• shared systemic cause: inflammation, metabolism, circadian disruption, or mitochondrial damage affecting ovary and distal organ in parallel.

Only the second is ovarian mediation toward healthspan, and even it can generate simultaneous benefits and risks. Current evidence does not distinguish these routes for NAD/SIRT. Any future extension must select one organ outcome, demonstrate exposure and mediator, and model competing risks before addressing disease-free years or survival.

9. Computational layer

Decision: skip. Optional computation was not run because no evaluated public dataset jointly measures donor, follicular state/atresia, total and compartmental NAD, SIRT1/SIRT3 activity, and mitochondrial flux. GSE202601 was already analyzed donor-aware in L7-1: four younger and four older donors; the OXPHOS signal was imprecise and sensitive to nucleus number/identity, whereas identity/composition was more ordered. Post hoc reanalysis of SLC25A51, SIRT1, or SIRT3 in the same eight people would not provide independent validation, and expression would not identify transport or activity.

GSE270748 pools patients and precludes donor-level inference. GSE329417 is postmenopausal and lacks the follicle–oocyte unit. Smits human metabolomics cannot be disaggregated by donor. Geneformer or sequence models cannot recover missing variables: an embedding could classify age, batch, or identity, but could not demonstrate NAD import, deacetylation, or turnover. The decision can reopen if a cohort appears with individual donors, follicular state, and at least paired metabolomics/functional proteomics or a rescued SLC25A51/SIRT perturbation.

10. Primary hypothesis

L9-1-HG-H1 v2 — cumulus support and candidate matrix gate

Falsifiable statement: in intact, non-atretic COC, an identifiable metabolic cargo transferred from cumulus supports oocyte NAD synthesis; if that gate passes, SLC25A51 limits matrix NAD and respiration, with a fraction requiring catalytic SIRT3, while somatic SIRT1 is tested as a separate branch.

Lineage: L9-1-SCOPE-H1 v3 → L9-1-MECH-H9/H10 → L9-1-HG-H1 v1 → adversarial pruning L9-1-HG-H1 v2.

Evidence for: COC dependence of NMN→NAD conversion in mouse ex vivo; SLC25A51 transport validated by loss/rescue in non-ovarian systems; associative human SIRT3/GDH signature; context-dependent SIRT1/SIRT3 functions in mice.

Evidence against: unknown intercellular cargo; nonspecific denudation/CBX; null human NAD in 23 cumulus samples; SLC25A51 not validated in ovary; SIRT3 not universally required; no integrated mito-QC flux.

Status: degraded, gated proposal.
Maturity: mouse cumulus support H1–H3; specific transfer, ovarian SLC25A51, and sirtuin epistasis H0; human H0–H1.
Comparative confidence: 0.24.

11. Competing hypothesis and adversarial rivals

L9-1-HG-H2 v2 — damage or atresia first; NAD as amplifier

Falsifiable statement: during natural aging, mitochondrial injury/DDR or atretic transition precedes the fall in compartmental NAD; restoring NAD while injury persists rescues less than removing the injury at matched initial NAD.

Mechanism: damage/DDR/atresia → PARP/CD38 + lower NAMPT/NMNAT1 → low NAD → reduced redox/SIRT margin → amplified apoptosis and dysfunction.

Evidence for: 3-NPA and Pdss2/CoQ injury generate the phenotype without NAD as initial lesion; porcine atresia lowers salvage/NAD; CD38 may modify initial state more than adult rate.
Evidence against: Ido1/Qprt show that a biosynthetic deficit can precede phenotype; damage models are not human aging; atresia may be downstream of low NAD.

Status: retained, narrowed competitor.
Maturity: animal damage→phenotype H2–H4; order in human aging H0.
Comparative confidence: 0.48.

Identification rival L9-1-AR-H2S v1 — selection/state

Statement: within follicles/COC comparable for stage, viability, atresia, stimulation, and baseline competence, the age–NAD/SIRT/mitochondria association will fall within a small margin; age mainly changes the frequency of observed states.

This rival explains associations and sampling bias; it makes no molecular-order claim. It is currently the leading explanation for human data (confidence 0.68, observational H1, causal H0). It is killed if a large age effect precedes atresia and persists within state, with replication and recovery weighting.

Mediator rival L9-1-HG-H3 v2 — early sirtuin-independent rescue

Statement: with NAD restored to the same range, a prespecified fraction of early functional rescue will persist after inducible catalytic SIRT1+SIRT3 loss; the sirtuins will contribute more to later resilience.

This is a contrary H0–H1 hypothesis with confidence 0.31. It is supported by NAD's broad biochemistry, rescue under incompletely qualified sirtinol, and contextual nulls. It is killed if catalytic loss eliminates all early rescue and only WT—not dead mutants—restores it at equivalent exposure, NAD, and viability.

These explanations must not be merged: selection/state can create an association without causal order; damage-first orders events inside one unit; sirtuin independence assigns the mediator of a rescue after NAD changes.

12. Translational hypothesis

L9-1-HG-HT1 v2 — parked, non-destructive dynamic assay

Falsifiable statement: only after proximal-chain validation, a non-destructive dynamic measure of cumulus–oocyte capacity or matrix response will improve out-of-sample error and calibration for ex vivo competence over age, stage, atresia, stimulation, reserve, and static NAD.

The hypothesis remains parked because the most informative measurements—LC–MS, OCR, and invasive sensors—consume or alter the unit whose competence is meant to be followed. A cumulus aliquot is not the oocyte, and repeated provocation may change the system. Even a reproducible proximal ART assay would not be a healthspan biomarker.

Predictions: metrology above prespecified thresholds; compatible follow-up; at most two frozen components; improved donor-level error and calibration at another center; gain independent of state/atresia.
Kill: death of H1, destructive measurement, inadequate ICC/CCC, preanalytic drift, no external gain, or dependence on pools/post hoc subgroups.
Status/maturity: parked, H0, confidence 0.03, HUMAN_QA_REQUIRED.

13. Falsifiable predictions and kill criteria

The global criterion that kills the NAD–SIRT narrative is: with compartmental NAD and engagement confirmed, catalytic SIRT1/SIRT3 loss does not modify rescue of flux/function, or damage/atresia reproducibly precedes the NAD decline. NAD may remain a modulator but ceases to be the principal relay.

14. Discriminating experiment

Sequence Q0 → D0 → D1/D2 → D3 → D4 → D4b → HUM

Not all layers run in parallel. Every negative gate prevents construction of causality on a nonexistent link.

Q0-METROLOGY

COC from 3–4-month-old C57BL/6 mice under brief meiotic arrest. Targeted LC–MS/MS for NAD+, NADH, NMN, NAM, adenosine, AMP/ADP/ATP, and isotopologues of d4-NMN, d3-NAD+, and 13C5-adenosine/AMP. Cumulus and oocyte are processed separately. Linearity, LLOQ, recovery, matrix effect, carry-over, stability, and time to quench are qualified.

Reaggregation of GV oocyte with primary cumulus is validated by adhesion, TZP, dye transfer, and Cx37 blockade. Gates: CV ≤15%, recovery 80–120%, last wash <LLOQ, quench ±30 s, viability/ATP/competence ≥85% of intact COC, and Cx37 blockade reducing coupling ≥50% without losing >15% viability. If tracer cannot be spatially restricted, microinjection or local uncaging is used; if that fails, D0 does not begin.

D0-TRANSFER — priority minimal experiment

System: young mouse COC; 16 mice in the paired confirmatory module; nested COC; randomization and blinded analysis; two batches.

D0a: load d4-NMN only into donor cumulus, wash exhaustively, reaggregate with unlabeled oocyte, and measure precursor/NAD in donor and recipient at 5, 15, 30, and 60 min.
D0b: if negative/ambiguous, repeat separately with 13C5-adenosine and 13C5-AMP.
D0c: confirm the positive module by microinjection/uncaging and a second isotopic standard when available.

Controls: acute/specific Cx37/GJA4 loss plus perturbation control; transwell/conditioned medium; last wash; recovery spike; blank; intact COC; denudation/CBX only as nonspecific comparators; ATP/ADP/AMP, viability, TZP, and donor load.

Primary: oocyte molar isotopic enrichment/donor enrichment. Pass requires signal >LLOQ and ≥5 times last wash/no-contact, Cx37 reduction ≥50% with the upper IC90% bound of the blockade/sham ratio <0.80, viability within margin, and independent replication. If label does not appear, specific transfer dies; indirect somatic support survives.

D1-STATE — selection versus within-state lesion

Young 3–4-month versus middle-aged 10–12-month mice, at the same stage and circadian window. A pre-NAD classifier frozen in Q0 uses stage/size, cumulus/TZP integrity, Annexin V/caspase, viability, and baseline competence. Every recovered and lost follicle is recorded. Total age effect, within-state effect, and recovery weighting are estimated.

The pilot uses 8 animals/age and prospectively recalculates 16–30/age for equivalence; a null requiring >30 is inconclusive. H2S gains support if the age residual lies within ±Δ, attenuates ≥50%, and replicates. Total and direct effects are not conflated because atresia may mediate age.

D2-ORDER — NAD first versus damage first

Two initially orthogonal perturbations are calibrated:

• NAD-first: FK866 titrated to a 20–30% NAD fall before OCR/ATP/viability changes >10%, followed by inducible Nampt confirmation and biochemical rescue;
• damage-first: two distinct ETC/CoQ injuries reducing OCR 20–30% while initial NAD remains within ±10%.

Times are 0, 15, and 60 min, 4 h, and 24 h. Readouts are compartmental NAD, OCR, ATP/adenylates, orthogonal ROS, PAR, γH2AX, caspase/Annexin, TZP, and viability. Within each family, insult, NAD replenishment with insult maintained, and insult removal at comparable initial NAD are contrasted. Order and cross-rescue decide; simultaneity indicates a circuit or insufficient resolution, not permission to choose a narrative.

D3-SLC — matrix transport

This gate opens only after positive D0 or a reproducible compartmental deficit in D2. SLC25A51 expression/localization is confirmed. Acute, cell-specific loss is performed separately in oocyte and cumulus, with WT and K91A/R278A rescue at equivalent expression/localization. Primary endpoint: free matrix NAD by a qualified sensor benchmarked to rapid mitochondrial LC–MS. Secondary endpoints: total NAD, NAD+/NADH, OCR, ATP, potential, mass, and viability. SIRT3 is not yet part of the estimand.

D4-SIRT and D4b-MITOQC

SIRT3 is perturbed in the cell in which D3 showed the matrix effect; SIRT1 is studied separately. Acute loss, substrate-based engagement, WT versus H248Y/H363Y rescue, and distinct functions are required. Dual loss opens only if single losses leave residual rescue and editing stress lies within equivalence. The primary endpoint is the fraction of early NAD rescue persisting without catalysis; ATP/redox are not adjusted in the primary model because they may mediate it.

Mito-QC opens only after causal function. mt-Keima or tandem mito-QC is used with pH calibration and lysosomal blockade, a biogenesis pulse–chase, and OCR/ATP. Greater lysosomal delivery with worse function is not called rejuvenation.

HUM-PROX

Only after D0–D4: 72 ART donors, 24 in each 20–29, 30–37, and 38–42-year band, with continuous age, protocol, diagnosis, reserve, stage, luteinization, atresia, BMI, smoking, and time to quench recorded. One preselected proximal marker is replicated; no pools; a second center is mandatory. This gate tests association and transportability, not causality in healthy women.

15. Biomarkers and stratification

There is no validated biomarker. These are experimental measurement candidates, each conditional on a context of use:

Scientific stratification precedes interpretation: follicular stage, atresia, cumulus/TZP integrity, stimulation and ART diagnosis, reserve, continuous age, batch, and preanalytics. Age must not be converted into three bins to select a favorable contrast afterward. No candidate advances to clinical use without reliability, stability, incremental value, and external replication.

16. Individual variability

Expected variability has at least six sources:

1. Biological state: initial reserve, recruitment, atresia, luteinization, and baseline competence.
2. Cellular compartment: oocyte, cumulus, mural granulosa, theca, stroma, immune cells, and endothelium may have different pools and consumers.
3. Life history: age, parity, gonadotoxic exposure, metabolic/inflammatory disease, and circadian disruption.
4. Clinical context: infertility diagnosis, stimulation protocol/dose, and selection of residual material.
5. Genetics/ancestry: DDR pathways and age at menopause may vary; European-study magnitudes do not transport to Latin America without replication.
6. Metrology: time to quench, temperature, denudation, pooling, sensor sensitivity, and analytical batch may dominate small differences.

Intrafollicular mitochondrial heterogeneity is a real measurement problem but must not automatically become biological subtypes. Bins defined by potential or size—variables changed by intervention—can create selection/collider bias. Analysis must retain animal/donor as the unit and validate that a subpopulation adds out-of-sample function.

17. Pharma relevance and maturity

Pharma opportunity criteria are causality in two models, cellular/ovarian target engagement, WT–dead rescue, exposure–response, independent function, donor-aware human material, and a window over systemic tissues. No-go criteria are dependence only on precursor, CBX/denudation, RNA/protein, a mitochondrial proxy, or ART association. No TPP, indication, dose, clinical biomarker, or trial readiness exists. Any advancement requires HUMAN_QA_REQUIRED.

Global maturity:

• rodent NAD–ovary perturbability: H1–H4 by component;
• integrated compartmental NAD–SIRT1/SIRT3–mito-QC chain: H0–H1;
• specific cumulus→oocyte support/transfer: support H1–H3, cargo H0;
• ovarian SLC25A51: H0;
• proximal human association: H0–H1;
• specific female healthspan/longevity: H0.

18. Limitations

1. Human evidence comes almost entirely from ART, discarded oocytes, or luteinized cells; selection and stimulation limit generalization.
2. Pooling and pseudoreplication reduce certainty. Hundreds of oocytes or organelles do not replace few animals or donors.
3. Many studies use whole-body/constitutive KO and systemic exposure, mixing development, immunity, metabolism, and ovary.
4. The chain depends on rarely measured compartments. NAD(P)H, total NAD, and matrix NAD are not equivalent.
5. Sirtuin activity is often inferred from abundance or one substrate; non-catalytic functions and compensation remain open.
6. “Mitochondrial quality” commonly rests on static proxies. Absence of dynamic flux is a central, not semantic, limitation.
7. Recent 2025–2026 studies include a preprint and work with small n or stress models rather than natural aging.
8. SIRT3 and CD38 discrepancies have not been replicated with harmonized background, age, construction, and endpoints.
9. Available public computation does not identify metabolite, compartment, catalysis, or flux.
10. Distal-health links are inferred from menopause age or mouse transplantation and may suffer pleiotropy, shared cause, and competing risks.
11. Most human evidence does not represent Mexican/Latin American populations or fully cover ages 35–55/STRAW+10.
12. An ex vivo perturbation may reveal capacity without showing that the same route limits in vivo aging.

19. Conclusions

NAD+ probably participates in the loss of follicular resilience, but cannot be presented as a universal initiator or longevity therapy. Evidence demonstrates modules: NAD supply/consumption changes rodent phenotypes; SIRT1 protects one component of maternal competence with age; SIRT3 modulates matrix function in some contexts; primary mitochondrial damage can produce the same phenotype; and follicular state and composition distort human signals. The relay commonly assumed between these modules is precisely what remains missing.

The strongest discovery hypothesis proposes a conditional architecture: cumulus supports oocyte NAD synthesis; an unknown intercellular cargo may cross through Cx37; SLC25A51 may limit the matrix pool; SIRT3 and SIRT1 may mediate separate fractions. Selection/state is the best explanation for human associations, and damage/atresia first is the strongest biological competitor. A third explanation—sirtuin-independent redox/energy rescue—prevents assignment of a precursor benefit to SIRT1/SIRT3.

The highest-information experiment is Q0 → D0-TRANSFER: a donor-cumulus-restricted tracer, demonstrated washout, Cx37 dependence, and precursor→product kinetics. A negative result would close the transfer boundary without denying indirect somatic support. Only a positive result justifies SLC25A51; only positive ovarian transport justifies SIRT3 epistasis; only positive catalysis justifies mito-QC; only then should donor-aware human replication begin. This sequence converts an appealing narrative into five genuine opportunities for refutation.

The connection to women's health and longevity must remain conditional. First validate proximal ovarian function, then endocrine exposure, then an organ-specific outcome with mediation, and finally disease-free years or survival with competing risks. Menopause genetics predicts trade-offs, not uniform benefit. At the evidence cutoff, there is no basis to prescribe, clinically stratify, or claim that raising ovarian NAD extends women's healthspan or lifespan.

20. References

1. Smits MAJ, et al. Human ovarian aging is characterized by oxidative damage and mitochondrial dysfunction. Human Reproduction. 2023;38:2208–2220. DOI: 10.1093/humrep/dead177. PMID: 37671592.
2. Pacella-Ince L, Zander-Fox DL, Lane M. Mitochondrial SIRT3 and its target glutamate dehydrogenase are altered in follicular cells of women with reduced ovarian reserve or advanced maternal age. Human Reproduction. 2014;29:1490–1499. DOI: 10.1093/humrep/deu071. PMID: 24771001.
3. Zhao HC, et al. Role of Sirt3 in mitochondrial biogenesis and developmental competence of human in vitro matured oocytes. Human Reproduction. 2016;31:607–622. DOI: 10.1093/humrep/dev345. PMID: 26787646.
4. Bertoldo MJ, et al. NAD+ repletion rescues female fertility during reproductive aging. Cell Reports. 2020;30:1670–1681.e7. DOI: 10.1016/j.celrep.2020.01.058. PMID: 32049001.
5. Yang Q, et al. Increasing ovarian NAD+ levels improve mitochondrial functions and reverse ovarian aging. Free Radical Biology and Medicine. 2020;156:1–10. DOI: 10.1016/j.freeradbiomed.2020.05.003. PMID: 32492457. Corrigendum: DOI 10.1016/j.freeradbiomed.2021.04.014, quantitative content remains quarantined.
6. Yang Q, et al. Deletion of enzymes for de novo NAD+ biosynthesis accelerated ovarian aging. Aging Cell. 2023;22:e13904. DOI: 10.1111/acel.13904. PMID: 37332134.
7. Yang Q, et al. NADase CD38 is a key determinant of ovarian aging. Nature Aging. 2024;4:110–128. DOI: 10.1038/s43587-023-00532-9. PMID: 38129670.
8. Perrone R, et al. CD38 regulates ovarian function and fecundity via NAD+ metabolism. iScience. 2023;26:107949. DOI: 10.1016/j.isci.2023.107949. PMID: 37822499.
9. Iljas JD, et al. Sirt1 sustains female fertility by slowing age-related decline in oocyte quality required for post-fertilization embryo development. Aging Cell. 2020;19:e13204. DOI: 10.1111/acel.13204. PMID: 32729989.
10. Iljas JD, Homer HA. Sirt3 is dispensable for oocyte quality and female fertility in lean and obese mice. FASEB Journal. 2020;34:6641–6653. DOI: 10.1096/fj.202000153R. PMID: 32212196.
11. Zhu J, et al. Sirt3 deficiency accelerates ovarian senescence without affecting spermatogenesis in aging mice. Free Radical Biology and Medicine. 2022;193:511–525. DOI: 10.1016/j.freeradbiomed.2022.10.324. PMID: 36336229.
12. Ben-Meir A, et al. Coenzyme Q10 restores oocyte mitochondrial function and fertility during reproductive aging. Aging Cell. 2015;14:887–895. DOI: 10.1111/acel.12368. PMID: 26111777.
13. Guo L, et al. Decrease in ovarian reserve through the inhibition of SIRT1-mediated oxidative phosphorylation. Aging. 2022;14:2335–2347. DOI: 10.18632/aging.203942. PMID: 35275845.
14. Ying Y, et al. Circadian rhythm disruption impairs ovarian follicular development via NAD+ metabolic reprogramming. eBioMedicine. 2026;126:106200. DOI: 10.1016/j.ebiom.2026.106200. PMID: 41797048.
15. Wang Z, et al. Investigation of NAD+ metabolism alterations during follicular atresia in porcine ovaries and their regulatory mechanisms in granulosa cell apoptosis. Free Radical Biology and Medicine. 2025;239:257–269. DOI: 10.1016/j.freeradbiomed.2025.07.049. PMID: 40752633.
16. Li Y, et al. SIRT3 attenuates AGEs-induced senescence in human granulosa cells through enhancing mitophagy. Cell Biology International. 2026. DOI: 10.1007/s10565-026-10138-7.
17. Mihalas BP, et al. Aging disrupts cumulus–oocyte NAD+ homeostasis. bioRxiv. 2026-07-23. Preprint, not peer reviewed. DOI: 10.64898/2026.07.22.739977.
18. Piasecki AJ, et al. Mitochondrial subpopulations in oocytes and cumulus cells exhibit distinct age-associated changes and selective plasticity in response to NMN supplementation. Journal of Ovarian Research. 2026. DOI: 10.1186/s13048-026-02223-6.
19. Luongo TS, et al. SLC25A51 is a mammalian mitochondrial NAD+ transporter. Nature. 2020;588:174–179. DOI: 10.1038/s41586-020-2741-7. PMID: 32906142.
20. Kory N, et al. MCART1/SLC25A51 is required for mitochondrial NAD transport. Science Advances. 2020;6:eabe5310. DOI: 10.1126/sciadv.abe5310.
21. Veitch GI, et al. Selective assembly of connexin37 into heterocellular gap junctions at the oocyte/granulosa cell interface. Journal of Cell Science. 2004. DOI: 10.1242/jcs.01124.
22. Gittens JEI, Kidder GM. Differential contributions of connexin37 and connexin43 to oogenesis revealed in chimeric reaggregated mouse ovaries. Journal of Cell Science. 2005. DOI: 10.1242/jcs.02624.
23. Li TY, et al. Rescue of oogenesis in Cx37-null mutant mice by oocyte-specific replacement with Cx43. Journal of Cell Science. 2007. DOI: 10.1242/jcs.03488.
24. Richard S, Baltz JM. Prophase I arrest of mouse oocytes mediated by natriuretic peptide precursor C requires GJA1 and GJA4 gap junctions in the antral follicle. Biology of Reproduction. 2014. DOI: 10.1095/biolreprod.114.118505.
25. Ahn BH, et al. A role for the mitochondrial deacetylase Sirt3 in regulating energy homeostasis. PNAS. 2008. DOI: 10.1073/pnas.0803790105.
26. Hirschey MD, et al. SIRT3 regulates mitochondrial fatty-acid oxidation by reversible enzyme deacetylation. Nature. 2010. DOI: 10.1038/nature08778.
27. Solomon JM, et al. Inhibition of SIRT1 catalytic activity increases p53 acetylation but does not alter cell survival following DNA damage. Molecular and Cellular Biology. 2006;26:28–38. DOI: 10.1128/MCB.26.1.28-38.2006.
28. Pfister JA, et al. Opposing effects of sirtuins on neuronal survival: SIRT1-mediated neuroprotection is independent of its deacetylase activity. PLoS ONE. 2008;3:e4090. DOI: 10.1371/journal.pone.0004090.
29. Sun N, et al. Measuring in vivo mitophagy. Molecular Cell. 2015;60:685–696. DOI: 10.1016/j.molcel.2015.10.009.
30. Sun N, et al. A fluorescence-based imaging method to measure in vitro and in vivo mitophagy using mt-Keima. Nature Protocols. 2017. DOI: 10.1038/nprot.2017.060.
31. Ruth KS, et al. Genetic insights into biological mechanisms governing human ovarian ageing. Nature. 2021;596:393–397. DOI: 10.1038/s41586-021-03779-7. PMID: 34349265.
32. Cargill SL, et al. Age of ovary determines remaining life expectancy in old ovariectomized mice. Aging Cell. 2003;2:185–190. DOI: 10.1046/j.1474-9728.2003.00049.x. PMID: 12882411.

Scientific delta of this report

The report consolidates a concrete causal correction: the NAD+–SIRT1/SIRT3–mitochondria chain cannot be considered established; it must be decomposed into somatic–germline transfer, matrix partitioning, catalytic necessity, and mito-QC flux, each with an independent kill gate. The highest-information first test is a donor-cumulus-restricted, Cx37-dependent tracer. This sequence distinguishes indirect support, state selection, damage/atresia first, and sirtuin-independent rescue before attributing health or longevity to an ovarian mechanism.