Age-Dependent Hysteresis in Human Macrophage Polarization

Determine whether primary monocyte-derived macrophages from older adults exhibit a stable, history-dependent inflammatory state whose entry and exit thresholds differ from those of younger donors under a defined bidirectional polarization-control protocol.

Central claim: Aged primary human macrophages exhibit quantitatively wider, persistent stimulus-history-dependent hysteresis between inflammatory-state entry and regulatory-state exit than young macrophages.

Success requires, in paired same-donor cultures:

1. Reproducible direction-dependent transition thresholds separated by >1.0 log10 unit in aged donors.
2. Mean aged hysteresis width at least 1.5-fold that in young donors at two-sided alpha=0.05, in a separately powered confirmatory cohort.
3. Concordant viable-cell flow cytometry, cytokine secretion, and NF-kB localization.
4. Persistence after matched ligand withdrawal/rest and identical rechallenge, excluding an acute ligand-specific response.

This 24-donor design is a feasibility and effect-size-prior pilot. It cannot itself confirm the age claim at two-sided alpha=0.05.

• The study tests ex-vivo macrophage state-transition hysteresis; it does not establish a thermodynamic energy barrier or catastrophe-theory mechanism.
• It does not determine therapeutic IL-4 dosing, senolytic efficacy, or clinical benefit.
• It does not identify a causal locus within receptor signaling, IKKbeta/NF-kB, chromatin, metabolism, extracellular matrix, or tissue niche.
• It does not test tissue heterogeneity, systemic bile-acid effects, CHIP, CMV status, or other donor-level causes of age-associated variation, although selected covariates may be exploratory.
• It does not claim that simplified M1/M2 surface-marker states represent all in-vivo macrophage diversity.
• Cross-ligand LPS-versus-IL-4 EC50 differences alone do not establish hysteresis and must be distinguished from unequal receptor potency or pathway-specific pharmacology.

Trained immunity provides a framework for durable, history-dependent innate-cell reprogramming, but does not establish a bistable or hysteretic architecture in primary human macrophage polarization [Defining trained immunity and its role in health and disease; https://doi.org/10.1038/s41577-020-0285-6]. Macrophage activation is better treated as a dynamic and heterogeneous state space than as a complete M1/M2 dichotomy. Regulatory motifs involving signaling, transcriptional, and metabolic feedback may support switch-like behavior, but do not prove it in aged donor macrophages [Multistability in Macrophage Activation Pathways and Metabolic Implications; https://doi.org/10.3390/cells11030404].

NF-kB is a biologically motivated orthogonal inflammatory readout, but p65 localization alone cannot identify a hysteretic state [NF-κB in biology and targeted therapy: new insights and translational implications; https://doi.org/10.1038/s41392-024-01757-9]. Macrophage subsets are context dependent, reinforcing the need not to equate a constrained marker panel with in-vivo macrophage diversity [The significance of macrophage polarization subtypes for animal models of tissue fibrosis and human fibrotic diseases; https://doi.org/10.1186/s40169-015-0047-4]. Metabolic and epigenetic reprogramming are plausible persistence mechanisms, but their causal role in aged macrophages is unresolved [Metabolic reprogramming and epigenetic modifications on the path to cancer; https://doi.org/10.1007/s13238-021-00846-7].

Key unresolved gaps are:

• No supplied decision-grade evidence demonstrates wider bidirectional hysteresis in aged versus young primary human MDMs.
• Cross-ligand LPS/IL-4 EC50 separation cannot establish hysteresis because their receptors, units, efficacies, and signaling differ.
• The molecular basis of any retained state—NF-kB, chromatin, metabolism, or another pathway—remains unresolved.
• The reduced CD86/CD206/HLA-DR/CD163/cytokine/p65 framework is operational rather than exhaustive.

This is an ex-vivo pilot using CD14+ M-CSF monocyte-derived macrophages from 12 younger adults (20–35 years; 6 female/6 male) and 12 older adults (65–80 years; 6 female/6 male). Fresh PBMCs are processed within 6 hours; monocytes differentiate for 7 days with M-CSF before each donor is split across randomized matched plates.

The study includes:

• Young donor bidirectional-axis pilot: 12 donors, paired directions, duplicate wells.
• Older donor bidirectional-axis pilot: 12 donors, paired directions, duplicate wells.
• Duration- and cumulative-exposure-matched order control.
• Repeated common-R probe control.
• Withdrawal-rest-rechallenge persistence arm.
• Fresh-history-matched reference arm.
• No-ligand differentiated-MDM controls.
• Every-batch cryopreserved healthy-donor PBMC bridge reference.
• Independent pooled-MDM classifier-reference controls.

For each calibration-valid donor, define:

R=log10[(IL-4/EC80_IL-4)/(LPS/EC80_LPS)]

At each dwell:

LPS=EC80_LPS×10^(-R/2)

IL-4=EC80_IL-4×10^(R/2)

Forward cultures traverse R=-2,-1.5,-1,-0.5,0,0.5,1,1.5,2; reverse cultures traverse the same positions in reverse order. Each dwell lasts 24 hours.

• Young and older donor arms use paired, donor-normalized forward/reverse sweeps; the pilot estimates age-stratified variance and effect-size priors.
• The PRIMARY node is the locked composite threshold measurement that produces donor-level hysteresis width W.
• Matched order and repeated common-R controls test history effects at identical terminal R=0 with duration and cumulative ligand exposure matched.
• Washout, 72-hour rest, and R=0 rechallenge test persistence against fresh-history references.
• Passing all pilot gates supports confirmatory-study planning; this pilot does not itself confirm the central age claim.

The primary functional readout is the pre-enrollment locked composite regulatory-state probability:

P_reg=logit^-1(β0+β1z(CD206)+β2z(CD163)−β3z(CD86)−β4z(HLA-DR)+β5z(log10(IL-10+LLOQ))−β6z(log10(TNF-alpha+LLOQ))−β7z(log10(IL-12p70+LLOQ))−β8z(p65 nuclear:cytoplasmic ratio))

Reference-derived means, SDs, coefficients, signs, winsorization bounds, LLOQ handling, missing-data rules, and the 0.50 threshold rule must be frozen before study-donor enrollment.

Readouts that test function:

• Same-axis paired forward/reverse thresholds, yielding W_d=|T_d,forward−T_d,reverse|.
• Matched-duration, cumulative-exposure-matched opposite-history differences at identical R=0.
• Fresh-reference-adjusted ΔP after washout/rest and after standardized R=0 rechallenge.
• Secreted TNF-alpha, IL-12p70, and IL-10, as orthogonal effector-output confirmation.

Readouts that support signaling, anatomy, morphology, or plausibility but cannot prove hysteresis alone:

• CD86, CD206, HLA-DR, and CD163 surface states.
• p65 nuclear-to-cytoplasmic localization and p65-high fraction.
• Fixed-cell morphology, cell counts, and confluence.
• Exploratory transcriptomics, if later added.

All six functional readouts have evidence_status='needs_source'. Full-text-verified primary-method evidence remains required before activation.

High-content p65 immunofluorescence is repeated at every R-axis dwell, matched common-R probe, transition-region endpoint, after 72-hour washout/rest, and 24 hours after R=0 rechallenge. It provides an orthogonal single-cell inflammatory-signaling measurement, but p65 localization alone cannot prove persistent hysteresis or assign a causal mechanism. Full-text-verified method support for this aged-primary-MDM application is required.

Histology is not a required functional readout in this ex-vivo MDM pilot. Fixed-cell immunofluorescence provides localization and morphology context; viable-cell flow establishes analyzed-cell identity and quality. Neither morphology nor fixed-cell staining alone can demonstrate threshold separation, matched-history discrimination, persistence, or age-associated hysteresis.

Spatial transcriptomics is not applicable. Single-cell transcriptomics is exploratory and, if added after the pilot, should profile donors with reproducibly wide versus narrow W, and opposite-history cultures after rest/rechallenge. It may identify candidate programs but cannot replace same-axis functional threshold, matched-history, and persistence testing or establish causality.

Organism/model: Human ex-vivo primary-cell pilot using CD14+ primary monocyte-derived macrophages from fresh leukapheresis-derived PBMCs.

Obtain IRB/ethics approval and written informed consent before recruitment, screening, collection, or laboratory use. Consent and the IRB protocol must authorize age-stratified PBMC collection, coded laboratory use, medication/infection screening, and CMV serostatus collection where consented.

Screen candidates to exclude acute infection, systemic immunosuppression, and recent vaccination. Record sex, BMI, medication class, CBC differential, CMV serostatus where consented, and monocyte-subset proportions as exploratory covariates.

Before launch, confirm 12 scheduled eligible donors plus 3 reserves in each age stratum. Replace donors only before unblinded outcome review; record all screened, excluded, scheduled, replaced, and processed donors.

Pre-enrollment assay lock, donor processing, and MDM differentiation

1. Before study-donor enrollment, use independent non-study pooled-MDM samples to establish no-ligand, inflammatory-reference, and regulatory-reference conditions.
2. Lock staining, acquisition, compensation, cytokine range, p65 segmentation, scaling, feature weights, winsorization bounds, lower-limit handling, QC scripts, and the composite classifier.
3. Do not alter locked elements using study-donor age, direction, history, or outcome data.
4. Confirm: reference technical-replicate ICC≥0.75; required analytes within validated range; ≥5,000 live singlet events/well; inflammatory-versus-regulatory composite separation≥0.30 probability units with 95% CI excluding 0; and blinded held-out reference classification accuracy≥85%.
5. If any criterion fails, revise and revalidate only on new independent technical references; do not process study donors for outcome analysis.
6. Recruit 24 enrolled adults: 12 younger (20–35 years; 6 female/6 male) and 12 older (65–80 years; 6 female/6 male). Recruit 30 eligible donors in total.
7. Process fresh leukapheresis-derived PBMCs within 6 hours; do not mix fresh and cryopreserved study samples.
8. Record coded donor ID, collection time, anticoagulant, monocyte yield, isolation recovery, processing operator, batch, and protocol deviations.
9. Isolate CD14+ monocytes using Miltenyi Biotec human CD14 MicroBeads, catalog 130-050-201.
10. Seed 1.5×10^5 viable CD14+ cells/cm2 in RPMI-1640 with 10% heat-inactivated pooled human AB serum, 1% penicillin-streptomycin, and recombinant human M-CSF 50 ng/mL (PeproTech 300-25).
11. Maintain cultures at 37°C, 5% CO2, humidified incubator.
12. Renew differentiation medium on day 3 with the same formulation and M-CSF concentration.
13. Use primary passage-0 cells only.
14. On differentiation day 7, verify 70–90% adherent confluence; do not permit overgrowth beyond 90%.
15. Record viability, cell count/confluence, contamination status, and plate identity before allocation.
16. Differentiate a fixed cryopreserved healthy-donor PBMC bridge reference to MDMs on every processing batch.

M-CSF-driven differentiation has been reported to express a substantial part of an M2-associated transcriptome, so these MDMs are a constrained operational primary-human macrophage model rather than an exhaustive representation of in-vivo macrophage diversity (https://doi.org/10.4049/jimmunol.177.10.7303).

Pilot-confirm: Use one prevalidated tissue-culture-treated plate format and working volume sufficient for all required matched flow, supernatant, and imaging wells; maintain the same surface-area-to-volume ratio across conditions.

Fresh matched-aliquot LPS and IL-4 EC80 calibration

1. On Day 0, allocate fresh matched differentiated-MDM aliquots from each donor to separate LPS and IL-4 concentration-response calibrations.
2. Use one validated lot of ultrapure E. coli LPS (InvivoGen tlrl-3pelps) and one validated lot of recombinant human IL-4 (PeproTech 200-04) for the entire pilot.
3. Prepare aliquots once, avoid repeated freeze-thaw, and use serum-matched exposure medium.
4. Apply the pre-specified ligand concentration range and maximum ligand ceiling to calibration wells.
5. At the locked calibration endpoint, acquire locked component measurements and viability data required for the calibration-response definition.
6. Fit a pre-specified four-parameter Hill model separately for each donor and ligand.
7. Accept EC80_LPS and EC80_IL-4 only when fit convergence occurs, adjusted R²≥0.80, observed range spans ≥70% of fitted amplitude, EC80 lies within tested range and below the locked ceiling, and calibration viability is ≥70%.
8. If either ligand fails, classify the donor as technically excluded from axis-based analyses; retain the reason, curve diagnostics, tested range, and viability results.
9. Do not substitute cohort-average EC80 values or manually adjust a donor EC80.
10. For calibration-valid donors calculate R=log10[(IL-4/EC80_IL-4)/(LPS/EC80_LPS)].
11. At every subsequent dwell calculate and record LPS=EC80_LPS×10^(-R/2) and IL-4=EC80_IL-4×10^(R/2).
12. Independently verify the concentration worksheet against donor ID, R value, ligand stock, dilution factors, and batch record before each exposure.

Pilot-confirm: Lock the calibration concentration range, maximum ligand ceiling, and Hill-fit response feature during independent reference-assay development before enrollment.

Paired bidirectional nine-dwell normalized-axis sweeps

1. Allocate matched MDM aliquots to paired forward and reverse sweep cultures, with duplicate wells per direction/timepoint.
2. Execute nine sequential 24-hour dwells at R=-2,-1.5,-1,-0.5,0,0.5,1,1.5,2 for the forward direction.
3. Execute the exact reverse sequence R=2,1.5,1,0.5,0,-0.5,-1,-1.5,-2 for the reverse direction.
4. At every dwell administer both ligands using donor-specific concentrations from the locked R-axis formulation; do not expose sweep wells to a single ligand alone.
5. At the end of every dwell collect from designated parallel wells: viable-cell flow for CD86, CD206, HLA-DR, and CD163; TNF-alpha, IL-12p70, and IL-10 supernatants; high-content fixed-cell p65 nuclear-to-cytoplasmic imaging; viability; live-singlet event yield; and cell count/confluence.
6. Analyze only live singlet CD45+ CD14-derived macrophage events.
7. Require baseline viability≥70%, required-dwell viability≥60%, and ≥5,000 live singlet macrophage events at every required measurement.
8. At every medium exchange retain donor code, concealed age stratum, direction, R, calculated ligand concentrations, exact dosing and sampling time, viable cell count/confluence, well position, operator, reagent lot, and deviations.
9. Apply locked P_reg unchanged to every study sample.
10. Call a directional threshold only when adjacent positions bracket P_reg=0.50, local monotonicity occurs over the surrounding three-point neighborhood, interpolation lies in the tested R range, duplicate-derived threshold ICC≥0.75, and bootstrap CI width≤1.0 log10 R unit.
11. Retain one valid technical replicate through the locked missing-data rule; failure of both duplicates at a required measurement makes the dwell missing.
12. Do not exclude donors for discordant marker patterns or small estimated hysteresis widths.

Duration- and cumulative-exposure-matched history controls and fresh references

1. Construct two order-control histories per donor, both terminating at an identical 24-hour R=0 probe after nine 24-hour dwells.
2. Construct one order-preserving, history-compressed schedule enriched toward the initial inflammatory side followed by balanced compensatory exposures; construct the opposite order as its reverse.
3. Calculate cumulative administered LPS and IL-4 using donor-specific dwell concentrations.
4. During pre-enrollment sequence design, adjust only pre-terminal compensatory exposures so cumulative exposure for each ligand is matched within ±5%, while retaining nine dwells, equal total duration, and the identical 24-hour terminal R=0 probe.
5. In parallel, impose the same 24-hour R=0 probe after early inflammatory-biased, early regulatory-biased, and balanced alternating histories.
6. Match total duration and cumulative exposure for each ligand within ±5%.
7. Include balanced-history comparators and process them identically.
8. Prepare freshly differentiated MDM references receiving a single 24-hour exposure at each donor transition-region R and at R=0, without prior sweep history.
9. Apply the same handling, medium changes, and sampling schedule to fresh references.
10. At terminal probes and fresh references collect the full locked flow, cytokine, p65, viability, event-yield, and confluence set.
11. Archive every R sequence, concentration, total duration, cumulative LPS, cumulative IL-4, terminal probe identity, and matching calculation.
12. A result is non-evaluable if either cumulative-exposure matching or identical-duration requirements are not met.

Pilot-confirm: Lock exact nonterminal R schedules during assay development, with numerical proof of nine dwells, matched duration, identical terminal R=0, and ±5% matching for cumulative LPS and IL-4.

Transition-region washout, ligand-removal verification, rest, and R=0 rechallenge

1. Identify each donor’s direction-specific putative transition region from the locked sweep schedule and reserve duplicate experimental and matched fresh-reference wells.
2. Do not select the region using unblinded age-group comparisons.
3. At the selected transition region collect the pre-washout full orthogonal measurement set from designated parallel wells.
4. Wash experimental cultures three times with prewarmed ligand-free macrophage medium.
5. Maintain identical wash handling, aspiration timing, medium temperature, and rest medium across forward, reverse, control, and fresh-reference wells.
6. Retain the complete third wash, labeled by coded donor, direction, well, and time.
7. Submit third washes for residual-LPS bioactivity testing and quantify residual IL-4 where sensitivity permits.
8. Record residual-ligand LOD/LLOQ before testing.
9. Mark persistence non-evaluable if final-wash LPS bioactivity is not below the prevalidated LOD or residual IL-4 is not below the available LLOQ.
10. Rest cultures for 72 hours in ligand-free macrophage medium at 37°C, 5% CO2.
11. At Day 12 collect the complete orthogonal measurement set.
12. Apply a standardized 24-hour R=0 rechallenge to rested experimental and matched fresh-reference cultures using donor-normalized EC80 concentrations.
13. At Day 13 collect full flow, cytokine, imaging, viability, event-yield, confluence, and residual-ligand records.

Pilot-confirm: Select and validate the LPS-bioactivity assay, IL-4 assay where sensitivity permits, storage conditions, and LOD/LLOQ during independent assay development.

Randomization, blinding, plate layout, and handling equivalence

1. Before each batch, generate a coded allocation list randomizing donor, age group, sweep direction, history-control condition, and well position.
2. Maintain allocation keys separately from flow, cytokine, imaging, and gating personnel.
3. Balance each plate across direction and history-control conditions, distributing matched donor aliquots across nonadjacent positions where compatible with the locked assay.
4. Include the cryopreserved healthy-donor PBMC bridge reference in every batch.
5. Ensure experimental, no-ligand, order-control, common-probe, persistence, and fresh-reference wells receive matched medium changes, dwell times, sampling times, imaging times, and operator contact whenever their assigned intervention permits.
6. Maintain no-ligand differentiated-MDM controls for each donor/batch handling context.
7. Mask age group, donor code, direction, history assignment, and outcomes from flow acquisition/gating, cytokine acquisition, p65 image acquisition/segmentation, and initial QC.
8. Use prelocked acquisition templates, gating hierarchy, segmentation, classifier coefficients, and exclusion rules.
9. Record deviations contemporaneously; do not retrospectively reclassify wells to improve threshold identifiability.
10. Release the allocation key only after raw-data completeness checks, locked technical-failure adjudication, and version-controlled locking of raw files and analysis inputs.

The primary assay measures donor-level paired direction-conditioned threshold separation on the same normalized two-ligand axis. It establishes feasibility, variance, and effect-size priors; it does not prove the full age claim.

For each valid donor:

• T_d,forward and T_d,reverse are interpolated R values at which locked P_reg crosses 0.50.
• W_d=|T_d,forward−T_d,reverse|.
• Thresholds require bracketing, local monotonicity, in-range interpolation, duplicate-derived ICC≥0.75, and bootstrap CI width≤1.0 log10 R unit.
• Do not extrapolate beyond R=-2 to R=2 or call an unbracketed plateau a threshold.
• Technical duplicates assess precision and are not independent biological observations.

At terminal R=0, calculate:

ΔP=(P_reg,history A,R=0−P_reg,history B,R=0)−(P_reg,fresh matched reference A,R=0−P_reg,fresh matched reference B,R=0)

A retained matched-history effect requires donor-level median |ΔP| above both 0.15 probability units and two times the locked duplicate-well error bound; paired donor-level 95% CI must exclude zero in the hypothesized direction; and ≥75% of assay-valid donors must show sweep/probe directional concordance.

The following confounds are specifically addressed:

• Same donor-normalized axis and two ligands at every dwell: cross-ligand potency/pharmacology.
• Matched nine-dwell duration and ±5% ligand-exposure controls: duration and integrated ligand dose.
• Fresh no-history references: static dose response and handling artifacts.
• Three washes, residual-LPS/IL-4 testing, 72-hour rest, and identical R=0 rechallenge: ligand carryover, receptor occupancy, and acute ligand effects.
• Live-singlet gating, viability, event yield, confluence, contamination criteria: selective death, debris, doublets, and compositional effects.
• Randomization, blinding, bridge controls, and single ligand lots: plate, batch, and operator bias.

Relevant cited context: https://doi.org/10.3389/fbioe.2020.00666; https://doi.org/10.1016/j.heliyon.2024.e28332; https://doi.org/10.1038/s41593-020-00789-y.

No omics platform is a required primary endpoint in this pilot.

If exploratory single-cell transcriptomics is added after the pilot, select donors with reproducibly wide versus narrow W and opposite-history cultures after rest/rechallenge. Integrate donor- and cell-state transcriptional programs with donor-level thresholds, common-R probe effects, cytokines, and p65 distributions. This is hypothesis-generating only and cannot substitute for the same-axis threshold assay, matched-history controls, or persistence tests.

Primary endpoint: assay-valid donor-level W_d=|T_d,forward−T_d,reverse| in log10 R units.

Pilot estimands: mean(W_older)−mean(W_younger), mean(W_older)/mean(W_younger), and the fresh-adjusted opposite-history contrast at matched terminal R=0.

Threshold model:

T~age_group+direction+age_group:direction+sex+batch+(1|donor)

Common-R probe model:

P_reg~age_group+history_order+age_group:history_order+sex+batch+(1|donor)

Use REML. Derive directional contrasts and W-related contrasts from the pre-specified approach, but calculate and report unsigned donor-level W directly. Donor-level W summaries use donor-stratified nonparametric bootstrap 95% CIs, resampling donors independently within age stratum while retaining paired directions, histories, and duplicates together.

Use at least 2,000 bootstrap resamples. Pilot-confirm: lock the exact count before enrollment.

Enroll 12 donors per age group and target ≥9 assay-valid donors per stratum. If donor-level SD is about 1.0 log10 R unit, the intended approximate 95% CI half-width for group mean W is ≤0.75 log10 R units. Technical duplicates are not independent biological n.

One fixed composite endpoint supports the pilot feasibility decision. Matched-probe and persistence gates are co-required validity criteria, not independent efficacy claims. Individual markers, cytokines, p65, calibration, and exploratory covariates are descriptive with 95% CIs. No Bayesian alternative, post-hoc classifier refit, age-stratified refit, direction-specific refit, or batch-specific refit is permitted.

Technical exclusions are limited to baseline viability<70%; required-dwell viability<60%; <5,000 live singlet CD45+ CD14-derived macrophage events; contamination; documented protocol deviation; donor-specific EC80 calibration failure; or failure of both duplicate wells at a required measurement. Do not exclude for age, sex, small effects, biologically discordant markers, unfavorable outcomes, or missed go thresholds.

Before enrollment, timestamp and archive the analysis plan, exclusions, randomization, blinding, assay/classifier version, scaling/coefficient values, threshold rules, missing-data rules, REML formulas, bootstrap method, gates, and confirmatory sizing algorithm.

Go/no-go gates:

• Pre-enrollment assay qualification: ICC≥0.75; all analytes validated; ≥5,000 live singlet events/well; reference separation≥0.30 with 95% CI excluding 0; held-out accuracy≥85%. Miss: no-go.
• Cohort readiness: ≥12 scheduled eligible plus 3 reserves/stratum; planned 6 female/6 male. Miss: pause recruitment.
• Calibration: Hill fit adjusted R²≥0.80; ≥70% amplitude; EC80 in range and below ceiling; viability≥70%. Miss: technical exclusion from axis analyses.
• Threshold feasibility: ≥75% valid paired thresholds/stratum; W ICC≥0.75; median threshold CI≤1.0 log10 R. Miss: no-go for age interpretation.
• Matched-history discrimination: ≥75% directional agreement; median |ΔP|>0.15 and >2× error; paired 95% CI excludes 0. Miss: no-go for history interpretation.
• Residual-ligand clearance: LPS below LOD and IL-4 below LLOQ in ≥90% evaluable cultures. Miss: no-go for persistence claims.
• Persistence: ≥75% retain specified ΔP after rest and rechallenge; ≥2/3 orthogonal classes agree in ≥75% of persistence-positive donors. Miss: no-go for persistent hysteresis.
• Pilot advancement: all validity gates pass; older mean W≥1.0 log10 R; older/younger W≥1.5; 95% CI not opposite the hypothesized direction. Miss: stop or redesign.
• Confirmatory claim: a separately powered preregistered cohort demonstrates aged mean W≥1.5-fold young W at two-sided alpha=0.05 while all validity gates pass.

Deliverables include the locked classifier and validation dataset; paired donor-level normalized-axis dataset; matched-history R=0 dataset; time-resolved flow/cytokine/p65/quality dataset; residual-ligand and persistence evidence; and a pilot report supporting advancement, redesign, or discontinuation.