In plain terms

The claim being assessed: Checkpoint-driven, epigenetically fixed CD8+ T-cell exhaustion is the dominant mechanism by which human solid tumours evade immune control.

Where the evidence stands: Γ — Plausible mechanism. There is a believable biological story and partial support, but the causal step has not been demonstrated.

What this mechanism does: This makes the process worse or keeps it going, but something else starts it.

Strongest evidence for it: The preregistered scope-matched gate SC5 yields Γ. Defined experimental lineage and tumour-control sensitivities reach Ζ; clinical checkpoint efficacy alone can reach Η in selected indications, but neither establishes terminal-state mediation or cross-solid-tumour dominance. The weakest-link bundle would be Β because of the observational existence subclaim; it is not the registered gate.

What decisive evidence would look like

In plain terms, researchers would change the proposed cause, confirm it changed as intended, and measure a meaningful health outcome while checking other explanations. The result should be able to support or count against the claim.

Relative explanatory fit: Alternative ranked ahead

What this review means for Immune Checkpoint Exhaustion

Claim being assessed: Checkpoint-driven, epigenetically fixed CD8+ T-cell exhaustion is the dominant mechanism by which human solid tumours evade immune control.. This review asks whether the proposed factor causes the health outcome—not just whether the two appear together. For the wording assessed here, the evidence is rated Plausible mechanism (HSEF Γ; Moderate confidence). There is a believable biological story and partial support, but the causal step has not been demonstrated. That rating applies to this claim and scope; it does not mean the biological process is irrelevant in every setting.

Implications for health and treatment: This is an assessment of a causal explanation, not proof that a particular supplement, diet, test, or treatment improves health. A treatment claim needs studies that test the intervention in people and measure meaningful benefits and harms. These findings can help frame research and informed clinical questions, but they do not identify what is right for one person.

Limits for individual care: Evidence from cells, animals, associations, or a plausible mechanism may not predict what happens in people. The grade is limited to the exact claim, outcomes, and sources reviewed, and it is not a diagnosis or an individual forecast. Do not start, stop, or change care based on this page; discuss personal decisions with a qualified clinician.

Medical information notice: This review provides general health information, not medical advice. It cannot diagnose a condition or replace an individual discussion with a qualified clinician.

This assessment rests on 69 sources and 75 individually checked claims. 75 of those claims were verified against the original paper.

A low grade is not an accusation. It means the causal work has not been done yet, which is the ordinary state of most actively studied mechanisms in biology.

At a glance

Sub-claims assessed separately

ID	Sub-claim	Type	Level	Pivotal card
SC1	Tumour-infiltrating T cells enter a reproducible dysfunctional state with inhibitory receptors and lost polyfunctionality.	existence	Β	—
SC2	Exhaustion is a distinct TOX-associated lineage maintained by defined epigenetic programmes.	mechanism	Ζ	—
SC3	The exhaustion programme causally constrains tumour control.	necessity	Ζ	Card ↗
SC4	PD-1 blockade restores tumour control by reversing terminally exhausted T-cell states.	mechanism	Γ	Card ↗
SC5	Exhaustion is the dominant evasion mechanism across human solid tumours.	primacy	Γ	—

Visual Evidence Cards · Pivotal Subclaims

Pivotal subclaim illustrations

Dark-theme simulated-micrograph renders and concise causal evidence summaries for the pivotal subclaims that govern this review's grade. Click any card to inspect the master 2400×2400 render.

Master 2400×2400 render ↗

Master 2400×2400 render ↗

Figures

Explanatory figures summarizing this review. They carry no evidential weight of their own; where a figure and the review differ, the review governs.

Full assessment

From here the review is written for a technical reader: caps and thresholds, competing models, the independence basis, and the citation ledger.

Sources: 69 primary reports; claims: 75; verified: 75; unverified: 0. Verification is bounded attribution. Access: 6 partial-body and 63 complete abstract/status examinations. Cutoff: 2026-10-05. No complete-paper audit or blinded calibration claimed. Adversarial review: Generated by Sol 6.1; adversarially reviewed by Claude Opus 5.5 (Anthropic) on 2026-10-06. 1 corrections applied (see ADVERSARIAL REVIEW AND APPROVAL). Status: Approved as revised.

AT A GLANCE

Exhaustion programs constrain immunity in defined tumour contexts, but terminal-state reversal and dominance across human solid tumours are not established.

Deciding experiment: A prospective multicancer study enrolls prespecified inflamed antigen-presenting, excluded and presentation-defective strata; randomizes matched checkpoint, lineage-reprogramming and alternative-barrier interventions; tracks pretreatment terminal and progenitor tumour-reactive clones separately with state, function and durable tumour outcomes.

Belief versus evidence: The preregistered scope-matched gate SC5 yields Γ. Defined experimental lineage and tumour-control sensitivities reach Ζ; clinical checkpoint efficacy alone can reach Η in selected indications, but neither establishes terminal-state mediation or cross-solid-tumour dominance. The weakest-link bundle would be Β because of the observational existence subclaim; it is not the registered gate.

Best-supported rival: Progenitor-supported contextual immune control. Prospective pretreatment stratification and factorial barrier-specific interventions with direct clone and state tracking distinguish a progenitor-supported network from a universal terminal-exhaustion bottleneck.

COMPETING MECHANISTIC MODELS

Rank 1: Progenitor-supported contextual immune control (Ζ). Responsive precursors, niches and ongoing differentiation sustain antitumour immunity. Subset ablation and model tracing converge with distinct human clonal trajectories. Human clone persistence or appearance does not directly label cell fate.

Rank 2: Antigen-presentation loss and immunoediting (Ε). Tumour recognition fails through HLA, B2M or IFNgamma pathway defects. Paired acquired-resistance lesions and functional assays support presentation and signaling escape. Small samples and evolutionary inference do not establish all-tumour prevalence or dominance.

Rank 3: Stromal exclusion and niche failure (Ζ). Fibroblast programs and absent APC niches limit access or support. Cell-resolved CAF depletion and TGFbeta combinations improve experimental control with human spatial concordance. CAF removal changes composition and function; human depletion efficacy is not established.

Rank 4: Myeloid and regulatory-cell suppression (Ζ). TAM, Treg and glucocorticoid circuits restrain effector immunity and priming. Conditional perturbations and complementary functional assays support several circuits. These compatible mechanisms do not supply one universal clinical ranking.

Rank 5: Dominant fixed checkpoint exhaustion (Γ). A checkpoint-induced irreversible CD8 program is the central escape mechanism across human solid tumours. Clinical checkpoint benefit and stable dysfunctional chromatin make the mechanism plausible. Program initiation, terminal reversal and cross-tumour dominance cannot be inferred from efficacy or marker abundance.

The tournament does not alter the nominal level. Compatible pathways and flexible alternatives are not automatically proved by explaining observations.

FULL REVIEW

Registered claim and scope

Checkpoint-driven, epigenetically fixed CD8+ T-cell exhaustion is the dominant mechanism by which human solid tumours evade immune control.

Use index scope-matched assessment. SC5 is the gate matching the broad human solid-tumour dominance claim; mechanism and efficacy inside prespecified inflamed, antigen-presenting tumour contexts are reported separately. Terminal-state reversal and progenitor expansion are distinct claims. No responder-defined scope.

Selection: public-interest rank 15 from the editorial ranking compiled 2026-09-23; categorical prior verdicts were visible. Private fresh review; earlier documents are preserved.

Evidence and causal interpretation

Clinical checkpoint efficacy is robust in specified settings. Separate randomized melanoma, lung and renal studies show survival benefits; prospectively PD-L1-high lung and MSI-high colorectal groups illustrate valid restricted enrollment. Initial and ten-year CheckMate067 reports are one cohort. CheckMate816 interim OS and KEYNOTE177 interim OS are not mature proof of all survival endpoints. Strong efficacy does not identify a terminal CD8 lineage mediator [33–41,58,62].

The exhaustion program is real but heterogeneous. Tumour antigen, chromatin remodeling, TOX/NR4A and cell-specific perturbations converge; the AST liver-antigen and engineered CAR lines are partly shared platforms. Original Scott Results show Tox deletion removes receptor signatures without rescuing cytokines or killing, followed by cell loss; the Methods state no randomization, blinding or sample-size predetermination. Receptor abundance, lineage survival and function must be separately measured [7,13,18,50,51,63–69].

Checkpoint-responsive sources often precede the terminal compartment. Progenitor ablation, nodal targeting and niche experiments favor precursor-supported replenishment. Original Yost Results identify 11 BCC and four SCC paired patients with bulk and serial-biopsy controls, but explicitly cannot distinguish peripheral sources from rare unexpanded tumour cells. Renal maintenance and lung mixed new/pre-existing clone studies constrain a universal replacement narrative [1,6,8,20,22–27,52–54,65].

Fixed describes resistance under tested conditions. Original Philip Results distinguish early plastic and later IL15-resistant mouse states, while human chromatin and marker patterns are comparisons. CAR rest, IL10-Fc, FcIL4, STAT5 and inducible TOX removal demonstrate partial context-dependent plasticity; apoptosis and persistence costs remain. Newer IRF8 original Results show matched co-transfer and two tumour-line growth contrasts, with IFN-dependent virus/tumour differences [3,14,21,28–30,51,68,69].

Alternative barriers have causal support. HLA, B2M and JAK defects impair recognition or cytokine response; fibroblast, Treg and macrophage interventions alter access, priming or function. Original LRRC15 depletion Results include genotype/toxin and CD8-depletion controls but alter total fibroblast abundance, and their progression-to-size or ulceration endpoint is not human all-cause survival. These can interact with exhaustion rather than replace it everywhere [9,17,29,31,42–49,55–57,66].

Negative clinical contexts remain load bearing. Original CheckMate143 Methods and Results confirm a powered recurrent-GBM primary OS null after 301 deaths; steroid/MGMT subgroup hypotheses do not rescue it. CheckMate026 first-line lung PFS is null; KEYNOTE361 misses multiplicity-adjusted combination thresholds and its later monotherapy comparisons are exploratory. These failures do not erase benefit elsewhere, but defeat unqualified pan-solid-tumour claims [32,59–61].

HYPOTHESIS STATE EVALUATION FRAMEWORK (HSEF v5.3)

Subclaim matrix and aggregation

Preregistered aggregation: scope-matched. The preregistered scope-matched gate SC5 yields Γ. Defined experimental lineage and tumour-control sensitivities reach Ζ; clinical checkpoint efficacy alone can reach Η in selected indications, but neither establishes terminal-state mediation or cross-solid-tumour dominance. The weakest-link bundle would be Β because of the observational existence subclaim; it is not the registered gate. Sensitivities: weakest-link Β; primary-subclaim Γ; scope-matched Γ.

Decisive-perturbation inventory and causal tally

[1] Progenitor exhausted T cells sustain tumour control and respond to checkpoint therapy; terminal cells have different responsiveness. Subset dependency in models does not establish terminal-state reversal or pan-tumour dominance; linked erratum details unexamined.

[3] Transient CAR signaling rest restores function and remodels exhaustion-associated chromatin in engineered-cell experimental systems. CAR rest is a different intervention from PD1 blockade and does not prove terminal human tumour T-cell reversal.

[4] MYB-dependent progenitor maintenance enables checkpoint-associated response and protection during chronic viral infection. Chronic-virus lineage dependency is a mechanism comparator rather than human solid-tumour causality.

[5] Continuous antigen plus hypoxia promotes metabolic dysfunction; reducing hypoxic stress improves experimental immunotherapy. Metabolic and environmental effects are not isolated proof that checkpoints initiate the epigenetic program.

[7] Original Methods and Results show tumour-specific Tox deletion removes inhibitory receptor signatures but retains impaired cytokines and killing, followed by loss of persistence; bystander and acute-infection controls separate antigen context. TOX marker loss is not functional rescue. Experiments were not randomized or blinded and had no sample-size predetermination; human samples are observational, including infiltrate-selected breast cancers.

[8] Tumour-draining-node memory-like cells supply checkpoint-responsive progenitors with epigenetic states distinct from exhausted tumour cells. A nodal source mechanism does not establish reversal of terminal intratumoral cells or clinical dominance.

[9] Macrophage depletion and antigen-specific interactions implicate a reciprocal macrophage-exhausted T-cell circuit. Changing macrophages affects additional functions and does not establish a single universal exhaustion bottleneck.

[10] In-vivo CRISPR and Fli1 perturbation improve effector responses and tumour or infection control while preserving precursor potential. Different disease systems and transcription-factor perturbation do not isolate ordinary PD1-mediated terminal reversal.

[11] Combined LAG3 and PD1 genetic interventions improve experimental clearance and survival with autocrine IFNgamma dependency. Dual conditional pathway effects do not prove that all human resistant tumours are limited by fixed CD8 exhaustion.

[13] During chronic infection TOX loss transiently increases function but removes progenitor support and persistence with immunopathology. The outcome argues against a simple uniformly harmful lineage; it is not a human tumour treatment trial.

[14] STAT5 and orthogonal cytokine signaling partially reprogram exhausted cells and improve experimental antitumour function. Targeted cytokine rewiring is not equivalent to PD1 blockade or complete durable terminal-state reversal in patients.

[15] Tumour-antigen exposure induces early functional and molecular dysfunction before cell division. Rapid onset in a defined model does not identify a checkpoint as the initiating cause across solid tumours.

[16] Human melanoma nivolumab and relatlimab immune profiling shows enhanced cytotoxicity with retained exhaustion signatures and clonal changes. Molecular response is compatible with continuing exhaustion programs; it is not a terminal-fate-mediated survival trial.

[17] Cell-resolved glucocorticoid receptor and myeloid steroidogenesis interventions improve antitumour immunity. Parallel environmental suppression and checkpoint effects need mediation separation.

[18] TOX/TOX2 perturbation in an engineered CAR setting implicates a feedback program with NR4A in hyporesponsiveness. Shared CAR platform with NR4A experiments limits independence; linked erratum details unexamined.

[19] IL10R and STAT3 support progenitor-like T cells and disease control in CLL models; human breast observations are complementary. CLL is not a solid tumour and protective IL10 effects are context dependent; no cross-solid-tumour clinical proof.

[21] IL10-Fc boosts metabolic and antitumour function of terminal exhausted cells with progenitor-independent effects in models. A distinct intervention shows some plasticity but does not imply PD1 blockade normally reverses all terminal states.

[22] TCF1-associated cells supply the proliferative burst after blockade in chronic LCMV. Chronic viral infection is a mechanism comparator rather than clinical solid-tumour efficacy or human fate proof.

[23] Tumour-draining-node targeted PDL1 intervention supports progenitor seeding and tumour control; human recurrence association is complementary. Human nodal association does not establish clinical source necessity or pan-cancer mechanism dominance.

[25] PD1 signaling supports high-affinity stem-like T-cell maintenance; blockade can drive differentiation and death. The protective axis is context dependent and does not negate clinical benefits or prove human cross-tumour dominance.

[28] In-vivo CRISPR identifies RBPJ, IRF1, ETS1 and IKAROS-dependent transitions with intervention-induced reprogramming. Gene-specific cell rewiring is distinct from routine checkpoint treatment and may alter several lineage compartments.

[29] ATF4-associated hypoxic stress can promote terminal-like dysfunction independently of TCR signaling; pathway intervention improves experimental function. Stress pathways provide an alternative input and do not establish exhaustion dominance in patients.

[30] FcIL4 improves terminal exhausted T-cell function and checkpoint synergy in experimental tumour contexts. Functional plasticity with another cytokine treatment does not establish ordinary PD1-mediated complete lineage reversal.

[31] Conditional CD39 perturbation identifies suppressive exhausted T-cell activity under hypoxic tumour conditions and improves tumour control. A suppressor mechanism in specified models does not determine prevalence or dominance across human solid cancers.

[33] Randomized CheckMate649 shows nivolumab plus chemotherapy improves OS and PFS in its primary gastric-cancer CPS-defined population. Combination efficacy does not isolate CD8 exhaustion mediation or imply identical benefits in every biomarker context.

[34] Ten-year CheckMate067 follow-up supports durable melanoma survival benefit with nivolumab-containing regimens. This is follow-up of the same randomized cohort as the initial report, not an independent replication or terminal-state fate assay.

[35] Randomized CheckMate816 reports neoadjuvant lung-cancer event-free survival and pathological response benefit. The interim OS result in this report did not cross its prespecified threshold; chemotherapy combination does not isolate terminal exhaustion reversal.

[36] Initial CheckMate067 randomization demonstrates melanoma PFS benefit from nivolumab-containing regimens with higher combination toxicity. PFS and the subsequent ten-year follow-up share one cohort; neither directly proves terminal CD8 epigenetic reversal. Linked erratum details unexamined.

[37] Randomized ipilimumab treatment improves metastatic melanoma survival in the enrolled HLA-selected population. CTLA4 acts in broader immune compartments and the trial does not identify terminal CD8 exhaustion as the dominant mediator; linked erratum details unexamined.

[38] CheckMate214 improves OS and response in the prespecified intermediate or poor-risk renal-cancer group. PFS did not meet the prespecified significance threshold and favorable-risk or mechanistic generalization is separate.

[39] KEYNOTE006 randomization demonstrates melanoma survival and PFS benefit of pembrolizumab versus ipilimumab. A comparative clinical endpoint does not identify which CD8 lineage mediated benefit.

[40] CheckMate057 improves previously treated nonsquamous NSCLC survival with nivolumab. Median PFS was not uniformly favorable and PD-L1-dependent heterogeneity does not establish cross-tumour dominance.

[41] CheckMate017 improves previously treated squamous NSCLC survival regardless of measured PD-L1 groups. Clinical efficacy is not an assay of exhaustion causality or terminal-state reversal.

[42] Original Results show LRRC15 CAF depletion, toxin and genotype controls, CD8-depletion reversal and improved checkpoint-associated tumour control in subcutaneous and orthotopic pancreatic models. Ablation changes total fibroblast abundance and immune state, so its full mechanism is not a single physical barrier; progression-to-size or ulceration survival is not all-cause human survival.

[43] Human bladder cytotoxic CD4 cells kill autologous tumour in an MHCII-dependent assay and are restrained by Tregs. Ex-vivo killing and response prediction do not establish in-vivo clinical necessity; CD8 exhaustion is not the only candidate compartment.

[44] PD1-positive Treg biology and conditional experiments supply a potential immunosuppressive consequence of blockade alongside gastric hyperprogression associations. Human observational hyperprogression is not randomized proof that checkpoint treatment causes accelerated disease.

[45] CD25-directed Treg depletion preserving IL2 signaling improves selected preclinical immune responses. Cross-system preclinical function is not established human solid-tumour efficacy or a universal CD8-exhaustion bottleneck.

[46] Conditional CXCR3 Treg manipulation releases DC1 cross-presentation and antitumour immunity with checkpoint combinations. Changing upstream priming provides another constraint and does not establish exhausted CD8 states as autonomous initiators.

[48] PI3Kgamma targeting alters macrophage programs and improves experimental checkpoint responses across tumour models. The intervention is not fully cell-specific and preclinical benefit is not human clinical efficacy.

[49] MHC-I-deficient pancreatic models reveal TAP1-associated presentation and CD4 or Treg immune constraints; CTLA4 intervention improves control and metastatic endpoints. Experimental loss and engineered rescue do not determine frequency or dominance in unselected human pancreatic cancer.

[50] PDL1 blockade in chronic infection yields limited chromatin remodeling and transient response followed by renewed exhaustion; antigen removal does not produce ordinary memory. Chronic-virus stability is not absolute irreversibility under every intervention or proof of terminal human tumour fate.

[51] Original Results distinguish early plastic and later fixed mouse tumour chromatin states by IL15 reprogramming, with B16-OVA confirmation and matching features in human PD1-high TILs. The human marker and chromatin comparison does not directly test patient T-cell reversibility; fixed means resistant to the tested conditions, not forever irreprogrammable.

[55] Paired melanoma relapse biopsies and functional tests identify JAK1, JAK2 and B2M resistance-associated defects affecting IFNgamma signaling or MHC-I presentation. Four patients do not estimate population prevalence or prove every escape event independently; acquired resistance is a distinct stage.

[56] Human melanoma and mismatch-repair-deficient colon nonresponse plus cell-line assays implicate JAK1/2 loss and absent IFNgamma signaling. Small selected samples and in-vitro mechanism do not establish universal primary resistance or ranking versus exhaustion.

[58] KEYNOTE024 demonstrates PFS and OS benefit in prospectively selected PD-L1-high NSCLC lacking specified oncogenic drivers. The selection was before outcome but defines a narrower context; clinical benefit is not terminal-state reversal.

[59] CheckMate026 first-line NSCLC randomization does not improve PFS in its prespecified PD-L1-at-least-five-percent population; OS is similar. This null pertains to its treatment, comparator and population, not absence of all exhausted cells or failure of every checkpoint regimen.

[60] Original Methods and Results confirm powered CheckMate143 OS null in recurrent glioblastoma after 301 deaths; nivolumab PFS and response are inferior to bevacizumab, and steroid or MGMT subsets are hypothesis generating. Archival biomarker tissue, unstandardized MGMT and small subgroups limit mechanisms; subgroup trends do not rescue the null primary outcome.

[61] KEYNOTE361 combination treatment misses both multiplicity-adjusted primary thresholds; subsequent monotherapy OS comparisons are exploratory. Nominal P values or exploratory subsets cannot override the testing hierarchy; the trial does not directly measure exhaustion mediation.

[62] KEYNOTE177 improves first-line colorectal-cancer PFS in the prospectively MSI-high or dMMR population. OS is not mature in this examined interim report and results cannot be generalized to microsatellite-stable colorectal cancer.

[63] Inducible autochthonous liver tumour and defined antigen show early antigen-driven dysfunction transitioning from reversible to less reversible states. Model antigen and restricted TCR context do not establish human checkpoint-driven initiation; this shares the liver platform with later chromatin and TOX work.

[64] TOX loss and molecular profiling show a defined feed-forward regulator of exhaustion differentiation under persistent stimulation. Lineage commitment mainly from experimental chronic stimulation does not alone prove that removing TOX restores human tumour control.

[65] Ablation of TCF1-positive PD1-positive tumour cells restricts checkpoint response and identifies stem-like subset expansion as its source. The source of benefit is distinct from reversal of terminal exhaustion and human observations remain complementary rather than equivalent causal tests.

[66] Human urothelial fibroblast TGFbeta signatures associate with exclusion and nonresponse; combined TGFbeta and PDL1 blockade releases infiltration and regression in a matching mouse model. Human signatures are observational and combination effects do not establish TGFbeta alone as a universal human dominant bottleneck.

[67] NR4A triple-knockout hCD19-reactive CAR T cells improve regression and survival in engineered-antigen mouse solid tumours with chromatin changes. Artificial antigen and CAR system limit transfer to endogenous human cancers; shared TOX/TOX2 CAR platform limits independence.

[68] Induced TOX ablation after exhaustion establishment changes transcription and chromatin, causes apoptotic cell loss and permits more effector-like fate flexibility. Continuous maintenance and plasticity coexist with survival costs; the abstract does not specify the whole model scope or prove clinical terminal human tumour reversal.

[69] IRF8 gain and loss perturbation alters TOX, effector molecules and experimental tumour control; chronic IFN-I repression differs from tumour expression. Tumour-versus-virus regulation supports context specificity rather than a universal common pathway or direct human dominance.

Tally: 56 discriminating contrasts. Null and negative contrasts are discriminators, not positive threshold crossings. Reports and replications are not independent causal lines. Shared cohorts, experimental platforms, transgenic lineages and repeated outcomes do not constitute fully independent causal lines.

Cap audit, precedence and Β/Γ/Δ boundary

Immunology Γ caps ex-vivo-only or correlational human-plus-in-vitro causal claims. Cell-resolved in-vivo perturbations cross that boundary for SC2 and SC3; clinical targeted-therapy effects can attain upper levels for efficacy. These narrower successes do not lift SC4 or SC5. Cancer-rubric progression and relapse effects remain distinct from incidence/initiation or universal escape necessity.

Gamma hard-rule audit: Human receptor or atlas associations do not cross an in-vivo causal boundary. Broad interventions, developmental immune effects and artificial CAR antigens cannot automatically generalize to the full clinical scope.

SC1 cap: “Inflammation association without pathway-specific causality caps at Β.” Reproducible human cellular states and functional patterns establish association; neither receptor expression nor an atlas alone demonstrates pathway-specific clinical causality.

SC1 field-valid definition: “ - Cell-type-specific conditional knockout or depletion - Pathway-specific blockade (not broad immunosuppression) changing disease course - Human targeted therapy with mechanistic alignment showing disease-modifying effect” Positive full-scope result: no. Reproducible human cellular states and functional patterns establish association; neither receptor expression nor an atlas alone demonstrates pathway-specific clinical causality.

SC2 field-valid definition: “ - Cell-type-specific conditional knockout or depletion - Pathway-specific blockade (not broad immunosuppression) changing disease course - Human targeted therapy with mechanistic alignment showing disease-modifying effect” Positive full-scope result: yes. Defined antigen, chromatin trajectories, TOX or NR4A and other lineage interventions converge on a distinct maintained program in specified experimental contexts; tumour and virus programs are not identical.

SC3 field-valid definition: “ - Cell-type-specific conditional knockout or depletion - Pathway-specific blockade (not broad immunosuppression) changing disease course - Human targeted therapy with mechanistic alignment showing disease-modifying effect” Positive full-scope result: yes. Complementary cell-resolved genetic and targeted pathway interventions change tumour control in prespecified inflamed antigen-presenting models, with human efficacy providing compatible clinical evidence. No claim that terminal exhaustion is the exclusive mediator is made.

SC4 field-valid definition: “ - Cell-type-specific conditional knockout or depletion - Pathway-specific blockade (not broad immunosuppression) changing disease course - Human targeted therapy with mechanistic alignment showing disease-modifying effect” Positive full-scope result: no. Ordinary PD1 benefit often arises from progenitor expansion, persistence or newly detected clones; direct reversal of pre-existing terminal human cells is not established. Distinct rest or cytokine interventions demonstrate partial plasticity rather than validating PD1 reversal.

SC5 field-valid definition: “ - Cell-type-specific conditional knockout or depletion - Pathway-specific blockade (not broad immunosuppression) changing disease course - Human targeted therapy with mechanistic alignment showing disease-modifying effect” Positive full-scope result: no. Human clinical efficacy and mechanistic mouse studies do not compare causal dominance across presentation-defective, excluded, myeloid-suppressed and inflamed solid tumours. Negative randomized trials and alternative escape mechanisms block generalization.

Upward threshold ladder

SC1 Δ: “Requires pathway- or cell-type-resolved causal perturbation affecting disease or phenotype.” Satisfied: no. Broader causal interpretation requires cell-resolved intervention with relevant tumour control, rather than greater receptor abundance.

SC2 Δ: “Requires pathway- or cell-type-resolved causal perturbation affecting disease or phenotype.” Satisfied: yes.

SC2 Ε: “Requires context-specific causality with explicit scope restriction.” Satisfied: yes.

SC2 Ζ: “Requires multiple convergent causal manipulations and reduced plausibility of downstream-only interpretations.” Satisfied: yes.

SC2 Η: “Requires robust upstream causal evidence across systems and interventions.” Satisfied: no. Wider human lineage generalization requires timed perturbation of established states with surviving clone controls, chromatin and durable function.

SC3 Δ: “Requires pathway- or cell-type-resolved causal perturbation affecting disease or phenotype.” Satisfied: yes.

SC3 Ε: “Requires context-specific causality with explicit scope restriction.” Satisfied: yes.

SC3 Ζ: “Requires multiple convergent causal manipulations and reduced plausibility of downstream-only interpretations.” Satisfied: yes.

SC3 Η: “Requires robust upstream causal evidence across systems and interventions.” Satisfied: no. Human mediator-specific control requires a prospective intervention separating exhaustion programs from priming, recruitment and other immune compartments.

SC4 Δ: “Requires pathway- or cell-type-resolved causal perturbation affecting disease or phenotype.” Satisfied: no. Label established terminal tumour-reactive human clones before blockade, then demonstrate durable chromatin and functional reversal, not selective survival or replacement.

SC5 Δ: “Requires pathway- or cell-type-resolved causal perturbation affecting disease or phenotype.” Satisfied: no. Prespecify representative inflamed, excluded and presentation-defective human tumours, compare barrier-specific causal interventions and show exhaustion has dominant attributable effects with appropriate rescue.

Upper ladder: Targeted checkpoint interventions against hard clinical endpoints are live upper-ladder equivalents. Multiple separate randomized melanoma, lung, renal and biomarker-selected colorectal trials justify Η for scoped clinical efficacy as a sensitivity, while null contexts remain explicit. Η, Θ and Ω for terminal reversal or pan-solid-tumour dominance require lineage-specific mediation, comparative bottleneck tests and predictive consistency; they cannot borrow the efficacy score.

Independence: Different instruments exclude different errors for constrained contribution. Chronic infection is a comparator, not independent human-tumour replication. CheckMate067 early and late reports are one cohort. AST, CAR and TOX follow-ups share biological platforms; atlas reuse, shared authors and common antigens limit independence. Human state prediction, clonal tracking and randomized clinical efficacy measure different questions.

Chromatin dynamics and antigen controls [63, 51, 7]: Distinguish antigen-specific differentiation from solely bulk infiltration, while sharing a liver platform.

Cell-resolved program perturbations [67, 28, 69]: Reduce marker-only and purely downstream explanations; CAR/artificial-antigen boundaries persist.

Precursor depletion and nodal support [65, 8, 23]: Test source dependency distinct from terminal receptor abundance.

Independent randomized clinical settings [39, 41, 38, 58, 60]: Exclude a solely observational efficacy account and expose clinical heterogeneity; they do not directly trace exhaustion mediation.

Alternative barrier interventions [42, 66, 46, 9]: Challenge a sole checkpoint bottleneck using different stromal, priming and myeloid instruments.

Falsification and pitfalls

Phenotype signature removed without claimed functional rescue. Scott Tox deletion uncouples receptor expression from cytokines and killing, then loses persistence. Additional downgrade: False.

Intervention efficacy without terminal-state reversal. TCF1 subset ablation, clonal appearance and persistent exhausted phenotypes separate clinical response from the nominal mediation claim. Additional downgrade: False.

Specific intervention has context-restricted clinical efficacy. Negative CheckMate026 and143 and multiplicity-corrected KEYNOTE361 contrast with positive disease-specific trials. Additional downgrade: False.

Checkpoint marker treated as causal exhaustion. TOX marker/function uncoupling and observational state cards retain the distinction.

Therapeutic efficacy substituted for terminal reversal. Clinical endpoints, clonal dynamics and chromatin mechanisms are separate sensitivities.

Chronic virus substituted for human tumour proof. Viral lineage experiments are explicitly tagged; IRF8 contrasts expose mechanistic differences.

Newly detected clone treated as proven newly recruited cell. Original Yost sampling caveat and alternative sources are retained.

Fixed treated as absolutely irreversible. CAR rest, IL10-Fc, FcIL4, STAT5 and inducible TOX loss support intervention-dependent plasticity.

Provider paper identity trusted without reconciliation. Consensus Pro substituted Khan for Scott; no substituted Scott details were promoted.

Adversarial steelman: above, then below

Δ for pan-solid-tumour dominance would require a comparative human bottleneck perturbation, not merely positive treatment trials in selected diseases. Η for SC2 or SC3 would require broader endogenous context and human mediation convergence beyond model antigens and partly shared platforms.

Β for the scope-matched causal hypothesis would omit the extensive in-vivo perturbational and mechanistic basis. Conversely, the strongest Δ proposal based on clinical efficacy tests a weaker claim than universal fixed-exhaustion dominance.

Retain Γ for the registered broad gate and Ζ for defined experimental contribution; retain the separate Η clinical-efficacy sensitivity. The real exhaustion phenomenon and clinical benefits are not labelled refuted.

Final score, causal direction and M/S/D

The preregistered scope-matched gate SC5 yields Γ. Defined experimental lineage and tumour-control sensitivities reach Ζ; clinical checkpoint efficacy alone can reach Η in selected indications, but neither establishes terminal-state mediation or cross-solid-tumour dominance. The weakest-link bundle would be Β because of the observational existence subclaim; it is not the registered gate.

Persistent antigen, hypoxia and stromal or myeloid inputs often precede dysfunction, but specific immune-program and precursor perturbations alter control, preventing a purely noncausal or downstream-only classification at steps1–2. Pan-tumour initiation, prevention and rate limitation or universal necessity are not established at steps3–4. The program can stabilize immune escape and preserve some persistent T-cell activity, meeting step5 within mapped contexts.

Momentum: Decisive work underway. Named recent primary perturbation studies in this corpus test mechanistic or functional boundaries; no claim of a presently recruiting trial is made.

Stability: Stable. No named pending result is asserted to be likely to move the registered score imminently.

Divergence: Contested; no single implied-level arithmetic. These dimensions do not change level or confidence.

PROPOSED DECISIVE EXPERIMENTS

A prospective multicancer study enrolls prespecified inflamed antigen-presenting, excluded and presentation-defective strata; randomizes matched checkpoint, lineage-reprogramming and alternative-barrier interventions; tracks pretreatment terminal and progenitor tumour-reactive clones separately with state, function and durable tumour outcomes.

Readout: Clonal survival versus recruitment, durable chromatin and cytotoxic changes, objective tumour control and progression or survival, analyzed by baseline stratum. Controls: Antigen-specific and bystander controls, receptor-signature versus killing assays, equivalent target engagement, terminal versus progenitor tracing, and matched stromal or presentation rescue. Supporting: In most representative strata direct exhaustion-program reversal accounts for the largest causal control effect, traced terminal clones recover durably, and alternative barrier removal adds little with verified engagement. Falsifying: Benefit depends on progenitors, new clones or another baseline barrier, or terminal signatures change without function and persistence.

RESEARCH TIMELINE

KNOWLEDGE GAPS

Which established human terminal clones can be durably reprogrammed without depletion or replacement, and which interventions accomplish this?

Can prospective barrier-stratified clinical comparisons estimate causal dominance rather than predict response from correlated signatures?

How do TOX and checkpoint functions balance survival, function and immunopathology across early and terminal states?

How do tumour-specific IRF8 and chronic-virus IFN programs alter the transportability of lineage results?

What explains target-engaged negative clinical settings and which alternative barriers must be removed before CD8-directed treatment?

Can shared clone specificity, spatial niches and longitudinal blood or node sampling distinguish recruitment from rare pre-existing cells with known detection limits?

METHODS AND CONSENSUS CONTRIBUTION

Consensus was used through the signed-in browser for discovery, counterevidence, research-gap and timeline suggestions, and attempted checks of specific indexed full-text passages. Returned Methods/Results passages were reconciled with original-paper passages where available; failed retrievals, source substitutions and unverified aspects are recorded in the case reconciliation. Successful passage checks are distinguished from provider summaries. Provider screening counts, citation counts and meters were not used as primary counts or HSEF scores. This source log records access depth; it does not claim complete-paper or supplement review. The monthly API allowance was conserved: zero Consensus API calls in this batch.

REVIEWER NOTES

69 primary reports examined with six selected original-body examinations. The cohort includes explicitly relevant chronic-virus and CLL mechanism comparators, which do not supply human solid-tumour proof. The newer TOX maintenance and IRF8 papers were Consensus leads independently verified. Several linked errata remain unexamined; none is counted as primary evidence. Consensus Pro misidentified the requested Scott paper as Khan, while portions of its Yost sampling caveat matched original Results. No complete-paper, supplement, trial-protocol or independent blinded scientific audit is claimed.

Search leads, reviews, protocols and correction notices do not pad the primary count. Downloaded XML does not earn examination credit. Linked notices remain attached; unexamined correction details limit inference. This is a coordinator source/claim consistency assessment, not independent blinded calibration.

ADVERSARIAL REVIEW AND APPROVAL

Authoring model: Sol 6.1. Adversarial reviewer: Claude Opus 5.5 (Anthropic). Date: 2026-10-06.

Scope of checks. All 69 reference PMIDs verified against PubMed metadata (title, first author, year, journal, DOI) — 69/69 resolve to the stated record; none carries a retraction or expression of concern. Twelve load-bearing claims were spot-checked against the cited abstracts or the retained original passages: the Tox-deletion result (inhibitory-receptor signature lost, function not rescued, persistence lost); the clonal-replacement study's 11 basal-cell and 4 squamous-cell paired patients and its sampling caveat; CheckMate 143 as a powered null analysed after 301 deaths; CheckMate 816's interim overall-survival result not meeting its significance criterion; CheckMate 214's progression-free survival missing the prespecified threshold; KEYNOTE-177's immature overall survival; CheckMate 026's null first-line result; KEYNOTE-361's multiplicity-adjusted misses; the HLA-A*0201-selected ipilimumab trial; CheckMate 057's unfavourable median progression-free survival; the TCF1+PD-1+ ablation result; and the induced-TOX-ablation maintenance study. Scoring logic was re-derived from references/hsef-framework.md and SKILL.md Steps 3, 8–12: the index-prescribed scope-matched rule is applied to the scope-matching gate SC5 (Γ) rather than to the narrower Ζ subclaims, and the reported alternatives (weakest-link Β, primary-subclaim Γ) are arithmetically correct for this matrix.

Findings.

No evidence card content, number, direction of effect, subclaim level, confidence or aggregation result changed.

Rejected findings (checked, no change). (a) Six references carry a print/issue year one later than PubMed's e-publication year (refs 25, 27, 31, 34, 47, 56); the convention is consistent and is not an error. (b) SC5 was tested against Δ: positive randomised trials in selected diseases test efficacy, not comparative causal dominance across presentation-defective, excluded and myeloid-suppressed tumours, and the glioblastoma and first-line lung nulls remain unexplained by the nominal — Γ stands. (c) SC1 at Β under the immunology Γ cap was tested against the argument that single-cell atlases plus functional assays should cross Γ; cap 2 (pure association cannot exceed Β) binds, since the atlas and receptor evidence carries no pathway-specific causal manipulation of its own. (d) The separate Η sensitivity for scoped clinical efficacy is correctly kept out of the registered letter, consistent with the framework's rule that a cap governs the inference from a system to the claimed scope. (e) The erratum notices recorded for several references are linked notices only; none is a retraction, and the review already states that their contents were not examined.

Residual limitations. Verification depth is abstract/metadata plus the six retained partial bodies; no complete-paper or supplement audit was performed by either model. Several linked errata remain unexamined upstream.

Approval: Following adversarial review and the corrections listed above, the revised document is approved. — Claude Opus 5.5 (Anthropic), 2026-10-06.

REFERENCES

1. Miller BC, Sen DR, Al Abosy R et al. Subsets of exhausted CD8+ T cells differentially mediate tumor control and respond to checkpoint blockade. Nature immunology. 2019. PMID 30778252. DOI: 10.1038/s41590-019-0312-6. Access: abstract_metadata. Study: PMID30778252.

2. Guo X, Zhang Y, Zheng L et al. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. Nature medicine. 2018. PMID 29942094. DOI: 10.1038/s41591-018-0045-3. Access: abstract_metadata. Study: PMID29942094.

3. Weber EW, Parker KR, Sotillo E et al. Transient rest restores functionality in exhausted CAR-T cells through epigenetic remodeling. Science (New York, N.Y.). 2021. PMID 33795428. DOI: 10.1126/science.aba1786. Access: abstract_metadata. Study: PMID33795428.

4. Tsui C, Kretschmer L, Rapelius S et al. MYB orchestrates T cell exhaustion and response to checkpoint inhibition. Nature. 2022. PMID 35978192. DOI: 10.1038/s41586-022-05105-1. Access: abstract_metadata. Study: PMID35978192.

5. Scharping NE, Rivadeneira DB, Menk AV et al. Mitochondrial stress induced by continuous stimulation under hypoxia rapidly drives T cell exhaustion. Nature immunology. 2021. PMID 33398183. DOI: 10.1038/s41590-020-00834-9. Access: abstract_metadata. Study: PMID33398183.

6. Pai JA, Hellmann MD, Sauter JL et al. Lineage tracing reveals clonal progenitors and long-term persistence of tumor-specific T cells during immune checkpoint blockade. Cancer cell. 2023. PMID 37001526. DOI: 10.1016/j.ccell.2023.03.009. Access: abstract_metadata. Study: PMID37001526.

7. Scott AC, Dündar F, Zumbo P et al. TOX is a critical regulator of tumour-specific T cell differentiation. Nature. 2019. PMID 31207604. DOI: 10.1038/s41586-019-1324-y. Access: full_text_partial. Study: AST liver-antigen platform; TOX perturbation and human samples.

8. Huang Q, Wu X, Wang Z et al. The primordial differentiation of tumor-specific memory CD8+ T cells as bona fide responders to PD-1/PD-L1 blockade in draining lymph nodes. Cell. 2022. PMID 36208623. DOI: 10.1016/j.cell.2022.09.020. Access: abstract_metadata. Study: PMID36208623.

9. Kersten K, Hu KH, Combes AJ et al. Spatiotemporal co-dependency between macrophages and exhausted CD8+ T cells in cancer. Cancer cell. 2022. PMID 35623342. DOI: 10.1016/j.ccell.2022.05.004. Access: abstract_metadata. Study: PMID35623342.

10. Chen Z, Arai E, Khan O et al. In vivo CD8+ T cell CRISPR screening reveals control by Fli1 in infection and cancer. Cell. 2021. PMID 33636129. DOI: 10.1016/j.cell.2021.02.019. Access: abstract_metadata. Study: PMID33636129.

11. Andrews LP, Butler SC, Cui J et al. LAG-3 and PD-1 synergize on CD8+ T cells to drive T cell exhaustion and hinder autocrine IFN-γ-dependent anti-tumor immunity. Cell. 2024. PMID 39121848. DOI: 10.1016/j.cell.2024.07.016. Access: abstract_metadata. Study: PMID39121848.

12. Pritykin Y, van der Veeken J, Pine AR et al. A unified atlas of CD8 T cell dysfunctional states in cancer and infection. Molecular cell. 2021. PMID 33891860. DOI: 10.1016/j.molcel.2021.03.045. Access: abstract_metadata. Study: Integrated public exhaustion datasets plus new viral experiments.

13. Alfei F, Kanev K, Hofmann M et al. TOX reinforces the phenotype and longevity of exhausted T cells in chronic viral infection. Nature. 2019. PMID 31207605. DOI: 10.1038/s41586-019-1326-9. Access: abstract_metadata. Study: PMID31207605.

14. Beltra JC, Abdel-Hakeem MS, Manne S et al. Stat5 opposes the transcription factor Tox and rewires exhausted CD8+ T cells toward durable effector-like states during chronic antigen exposure. Immunity. 2023. PMID 38091951. DOI: 10.1016/j.immuni.2023.11.005. Access: abstract_metadata. Study: PMID38091951.

15. Rudloff MW, Zumbo P, Favret NR et al. Hallmarks of CD8+ T cell dysfunction are established within hours of tumor antigen encounter before cell division. Nature immunology. 2023. PMID 37537361. DOI: 10.1038/s41590-023-01578-y. Access: abstract_metadata. Study: PMID37537361.

16. Cillo AR, Cardello C, Shan F et al. Blockade of LAG-3 and PD-1 leads to co-expression of cytotoxic and exhaustion gene modules in CD8+ T cells to promote antitumor immunity. Cell. 2024. PMID 39121849. DOI: 10.1016/j.cell.2024.06.036. Access: abstract_metadata. Study: PMID39121849.

17. Acharya N, Madi A, Zhang H et al. Endogenous Glucocorticoid Signaling Regulates CD8+ T Cell Differentiation and Development of Dysfunction in the Tumor Microenvironment. Immunity. 2020. PMID 32937153. DOI: 10.1016/j.immuni.2020.08.005. Access: abstract_metadata. Study: PMID32937153.

18. Seo H, Chen J, González-Avalos E et al. TOX and TOX2 transcription factors cooperate with NR4A transcription factors to impose CD8+ T cell exhaustion. Proceedings of the National Academy of Sciences of the United States of America. 2019. PMID 31152140. DOI: 10.1073/pnas.1905675116. Access: abstract_metadata. Study: Related hCD19 CAR TOX/TOX2 platform.

19. Hanna BS, Llaó-Cid L, Iskar M et al. Interleukin-10 receptor signaling promotes the maintenance of a PD-1int TCF-1+ CD8+ T cell population that sustains anti-tumor immunity. Immunity. 2021. PMID 34879221. DOI: 10.1016/j.immuni.2021.11.004. Access: abstract_metadata. Study: PMID34879221.

20. Yost KE, Satpathy AT, Wells DK et al. Clonal replacement of tumor-specific T cells following PD-1 blockade. Nature medicine. 2019. PMID 31359002. DOI: 10.1038/s41591-019-0522-3. Access: full_text_partial. Study: PMID31359002.

21. Guo Y, Xie YQ, Gao M et al. Metabolic reprogramming of terminally exhausted CD8+ T cells by IL-10 enhances anti-tumor immunity. Nature immunology. 2021. PMID 34031618. DOI: 10.1038/s41590-021-00940-2. Access: abstract_metadata. Study: PMID34031618.

22. Im SJ, Hashimoto M, Gerner MY et al. Defining CD8+ T cells that provide the proliferative burst after PD-1 therapy. Nature. 2016. PMID 27501248. DOI: 10.1038/nature19330. Access: abstract_metadata. Study: PMID27501248.

23. Dammeijer F, van Gulijk M, Mulder EE et al. The PD-1/PD-L1-Checkpoint Restrains T cell Immunity in Tumor-Draining Lymph Nodes. Cancer cell. 2020. PMID 33007259. DOI: 10.1016/j.ccell.2020.09.001. Access: abstract_metadata. Study: PMID33007259.

24. Au L, Hatipoglu E, Robert de Massy M et al. Determinants of anti-PD-1 response and resistance in clear cell renal cell carcinoma. Cancer cell. 2021. PMID 34715028. DOI: 10.1016/j.ccell.2021.10.001. Access: abstract_metadata. Study: PMID34715028.

25. Hor JL, Schrom EC, Wong-Rolle A et al. Inhibitory PD-1 axis maintains high-avidity stem-like CD8+ T cells. Nature. 2026. PMID 41299179. DOI: 10.1038/s41586-025-09440-x. Access: abstract_metadata. Study: PMID41299179.

26. Wang K, Coutifaris P, Brocks D et al. Combination anti-PD-1 and anti-CTLA-4 therapy generates waves of clonal responses that include progenitor-exhausted CD8+ T cells. Cancer cell. 2024. PMID 39214097. DOI: 10.1016/j.ccell.2024.08.007. Access: abstract_metadata. Study: PMID39214097.

27. Liu B, Hu X, Feng K et al. Temporal single-cell tracing reveals clonal revival and expansion of precursor exhausted T cells during anti-PD-1 therapy in lung cancer. Nature cancer. 2022. PMID 35121991. DOI: 10.1038/s43018-021-00292-8. Access: abstract_metadata. Study: PMID35121991.

28. Zhou P, Shi H, Huang H et al. Single-cell CRISPR screens in vivo map T cell fate regulomes in cancer. Nature. 2023. PMID 37968405. DOI: 10.1038/s41586-023-06733-x. Access: abstract_metadata. Study: PMID37968405.

29. Alicea Pauneto CDM, Riesenberg BP, Gandy EJ et al. Intra-tumoral hypoxia promotes CD8+ T cell dysfunction via chronic activation of integrated stress response transcription factor ATF4. Immunity. 2025. PMID 41005293. DOI: 10.1016/j.immuni.2025.09.003. Access: abstract_metadata. Study: PMID41005293.

30. Feng B, Bai Z, Zhou X et al. The type 2 cytokine Fc-IL-4 revitalizes exhausted CD8+ T cells against cancer. Nature. 2024. PMID 39322665. DOI: 10.1038/s41586-024-07962-4. Access: abstract_metadata. Study: PMID39322665.

31. Vignali PDA, DePeaux K, Watson MJ et al. Hypoxia drives CD39-dependent suppressor function in exhausted T cells to limit antitumor immunity. Nature immunology. 2023. PMID 36543958. DOI: 10.1038/s41590-022-01379-9. Access: abstract_metadata. Study: PMID36543958.

32. Lee AH, Sun L, Mochizuki AY et al. Neoadjuvant PD-1 blockade induces T cell and cDC1 activation but fails to overcome the immunosuppressive tumor associated macrophages in recurrent glioblastoma. Nature communications. 2021. PMID 34836966. DOI: 10.1038/s41467-021-26940-2. Access: abstract_metadata. Study: PMID34836966.

33. Janjigian YY, Shitara K, Moehler M et al. First-line nivolumab plus chemotherapy versus chemotherapy alone for advanced gastric, gastro-oesophageal junction, and oesophageal adenocarcinoma (CheckMate 649): a randomised, open-label, phase 3 trial. Lancet (London, England). 2021. PMID 34102137. DOI: 10.1016/S0140-6736(21)00797-2. Access: abstract_metadata. Study: PMID34102137.

34. Wolchok JD, Chiarion-Sileni V, Rutkowski P et al. Final, 10-Year Outcomes with Nivolumab plus Ipilimumab in Advanced Melanoma. The New England journal of medicine. 2025. PMID 39282897. DOI: 10.1056/NEJMoa2407417. Access: abstract_metadata. Study: CheckMate067 NCT01844505.

35. Forde PM, Spicer J, Lu S et al. Neoadjuvant Nivolumab plus Chemotherapy in Resectable Lung Cancer. The New England journal of medicine. 2022. PMID 35403841. DOI: 10.1056/NEJMoa2202170. Access: abstract_metadata. Study: PMID35403841.

36. Larkin J, Chiarion-Sileni V, Gonzalez R et al. Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. The New England journal of medicine. 2015. PMID 26027431. DOI: 10.1056/NEJMoa1504030. Access: abstract_metadata. Study: CheckMate067 NCT01844505.

37. Hodi FS, O'Day SJ, McDermott DF et al. Improved survival with ipilimumab in patients with metastatic melanoma. The New England journal of medicine. 2010. PMID 20525992. DOI: 10.1056/NEJMoa1003466. Access: abstract_metadata. Study: PMID20525992.

38. Motzer RJ, Tannir NM, McDermott DF et al. Nivolumab plus Ipilimumab versus Sunitinib in Advanced Renal-Cell Carcinoma. The New England journal of medicine. 2018. PMID 29562145. DOI: 10.1056/NEJMoa1712126. Access: abstract_metadata. Study: PMID29562145.

39. Robert C, Schachter J, Long GV et al. Pembrolizumab versus Ipilimumab in Advanced Melanoma. The New England journal of medicine. 2015. PMID 25891173. DOI: 10.1056/NEJMoa1503093. Access: abstract_metadata. Study: PMID25891173.

40. Borghaei H, Paz-Ares L, Horn L et al. Nivolumab versus Docetaxel in Advanced Nonsquamous Non-Small-Cell Lung Cancer. The New England journal of medicine. 2015. PMID 26412456. DOI: 10.1056/NEJMoa1507643. Access: abstract_metadata. Study: PMID26412456.

41. Brahmer J, Reckamp KL, Baas P et al. Nivolumab versus Docetaxel in Advanced Squamous-Cell Non-Small-Cell Lung Cancer. The New England journal of medicine. 2015. PMID 26028407. DOI: 10.1056/NEJMoa1504627. Access: abstract_metadata. Study: PMID26028407.

42. Krishnamurty AT, Shyer JA, Thai M et al. LRRC15+ myofibroblasts dictate the stromal setpoint to suppress tumour immunity. Nature. 2022. PMID 36171287. DOI: 10.1038/s41586-022-05272-1. Access: full_text_partial. Study: Related LRRC15 fibroblast lineage, distinct experimental depletion.

43. Oh DY, Kwek SS, Raju SS et al. Intratumoral CD4+ T Cells Mediate Anti-tumor Cytotoxicity in Human Bladder Cancer. Cell. 2020. PMID 32497499. DOI: 10.1016/j.cell.2020.05.017. Access: abstract_metadata. Study: PMID32497499.

44. Kamada T, Togashi Y, Tay C et al. PD-1+ regulatory T cells amplified by PD-1 blockade promote hyperprogression of cancer. Proceedings of the National Academy of Sciences of the United States of America. 2019. PMID 31028147. DOI: 10.1073/pnas.1822001116. Access: abstract_metadata. Study: PMID31028147.

45. Solomon I, Amann M, Goubier A et al. CD25-Treg-depleting antibodies preserving IL-2 signaling on effector T cells enhance effector activation and antitumor immunity. Nature cancer. 2020. PMID 33644766. DOI: 10.1038/s43018-020-00133-0. Access: abstract_metadata. Study: PMID33644766.

46. Moreno Ayala MA, Campbell TF, Zhang C et al. CXCR3 expression in regulatory T cells drives interactions with type I dendritic cells in tumors to restrict CD8+ T cell antitumor immunity. Immunity. 2023. PMID 37392735. DOI: 10.1016/j.immuni.2023.06.003. Access: abstract_metadata. Study: PMID37392735.

47. Dominguez CX, Müller S, Keerthivasan S et al. Single-Cell RNA Sequencing Reveals Stromal Evolution into LRRC15+ Myofibroblasts as a Determinant of Patient Response to Cancer Immunotherapy. Cancer discovery. 2020. PMID 31699795. DOI: 10.1158/2159-8290.CD-19-0644. Access: abstract_metadata. Study: Human LRRC15 fibroblast program.

48. De Henau O, Rausch M, Winkler D et al. Overcoming resistance to checkpoint blockade therapy by targeting PI3Kγ in myeloid cells. Nature. 2016. PMID 27828943. DOI: 10.1038/nature20554. Access: abstract_metadata. Study: PMID27828943.

49. Schmiechen ZC, Cruz-Hinojoza E, Hilk AL et al. Treg cells promote immunotherapy-induced immune evasion by restraining CD4 T cell control of MHC-I-deficient metastatic pancreatic cancer. Science immunology. 2026. PMID 42361199. DOI: 10.1126/sciimmunol.adz4302. Access: abstract_metadata. Study: PMID42361199.

50. Pauken KE, Sammons MA, Odorizzi PM et al. Epigenetic stability of exhausted T cells limits durability of reinvigoration by PD-1 blockade. Science (New York, N.Y.). 2016. PMID 27789795. DOI: 10.1126/science.aaf2807. Access: abstract_metadata. Study: PMID27789795.

51. Philip M, Fairchild L, Sun L et al. Chromatin states define tumour-specific T cell dysfunction and reprogramming. Nature. 2017. PMID 28514453. DOI: 10.1038/nature22367. Access: full_text_partial. Study: AST liver-antigen platform plus B16-OVA and human chromatin.

52. Jansen CS, Prokhnevska N, Master VA et al. An intra-tumoral niche maintains and differentiates stem-like CD8 T cells. Nature. 2019. PMID 31827286. DOI: 10.1038/s41586-019-1836-5. Access: abstract_metadata. Study: PMID31827286.

53. Tumeh PC, Harview CL, Yearley JH et al. PD-1 blockade induces responses by inhibiting adaptive immune resistance. Nature. 2014. PMID 25428505. DOI: 10.1038/nature13954. Access: abstract_metadata. Study: PMID25428505.

54. Sade-Feldman M, Yizhak K, Bjorgaard SL et al. Defining T Cell States Associated with Response to Checkpoint Immunotherapy in Melanoma. Cell. 2018. PMID 30388456. DOI: 10.1016/j.cell.2018.10.038. Access: abstract_metadata. Study: PMID30388456.

55. Zaretsky JM, Garcia-Diaz A, Shin DS et al. Mutations Associated with Acquired Resistance to PD-1 Blockade in Melanoma. The New England journal of medicine. 2016. PMID 27433843. DOI: 10.1056/NEJMoa1604958. Access: abstract_metadata. Study: PMID27433843.

56. Shin DS, Zaretsky JM, Escuin-Ordinas H et al. Primary Resistance to PD-1 Blockade Mediated by JAK1/2 Mutations. Cancer discovery. 2017. PMID 27903500. DOI: 10.1158/2159-8290.CD-16-1223. Access: abstract_metadata. Study: PMID27903500.

57. McGranahan N, Rosenthal R, Hiley CT et al. Allele-Specific HLA Loss and Immune Escape in Lung Cancer Evolution. Cell. 2017. PMID 29107330. DOI: 10.1016/j.cell.2017.10.001. Access: abstract_metadata. Study: PMID29107330.

58. Reck M, Rodríguez-Abreu D, Robinson AG et al. Pembrolizumab versus Chemotherapy for PD-L1-Positive Non-Small-Cell Lung Cancer. The New England journal of medicine. 2016. PMID 27718847. DOI: 10.1056/NEJMoa1606774. Access: abstract_metadata. Study: PMID27718847.

59. Carbone DP, Reck M, Paz-Ares L et al. First-Line Nivolumab in Stage IV or Recurrent Non-Small-Cell Lung Cancer. The New England journal of medicine. 2017. PMID 28636851. DOI: 10.1056/NEJMoa1613493. Access: abstract_metadata. Study: PMID28636851.

60. Reardon DA, Brandes AA, Omuro A et al. Effect of Nivolumab vs Bevacizumab in Patients With Recurrent Glioblastoma: The CheckMate 143 Phase 3 Randomized Clinical Trial. JAMA oncology. 2020. PMID 32437507. DOI: 10.1001/jamaoncol.2020.1024. Access: full_text_partial. Study: PMID32437507.

61. Powles T, Csőszi T, Özgüroğlu M et al. Pembrolizumab alone or combined with chemotherapy versus chemotherapy as first-line therapy for advanced urothelial carcinoma (KEYNOTE-361): a randomised, open-label, phase 3 trial. The Lancet. Oncology. 2021. PMID 34051178. DOI: 10.1016/S1470-2045(21)00152-2. Access: abstract_metadata. Study: PMID34051178.

62. André T, Shiu KK, Kim TW et al. Pembrolizumab in Microsatellite-Instability-High Advanced Colorectal Cancer. The New England journal of medicine. 2020. PMID 33264544. DOI: 10.1056/NEJMoa2017699. Access: abstract_metadata. Study: PMID33264544.

63. Schietinger A, Philip M, Krisnawan VE et al. Tumor-Specific T Cell Dysfunction Is a Dynamic Antigen-Driven Differentiation Program Initiated Early during Tumorigenesis. Immunity. 2016. PMID 27521269. DOI: 10.1016/j.immuni.2016.07.011. Access: abstract_metadata. Study: Inducible AST SV40 liver-antigen platform.

64. Khan O, Giles JR, McDonald S et al. TOX transcriptionally and epigenetically programs CD8+ T cell exhaustion. Nature. 2019. PMID 31207603. DOI: 10.1038/s41586-019-1325-x. Access: abstract_metadata. Study: Chronic-stimulation TOX lineage program.

65. Siddiqui I, Schaeuble K, Chennupati V et al. Intratumoral Tcf1+PD-1+CD8+ T Cells with Stem-like Properties Promote Tumor Control in Response to Vaccination and Checkpoint Blockade Immunotherapy. Immunity. 2019. PMID 30635237. DOI: 10.1016/j.immuni.2018.12.021. Access: abstract_metadata. Study: PMID30635237.

66. Mariathasan S, Turley SJ, Nickles D et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature. 2018. PMID 29443960. DOI: 10.1038/nature25501. Access: abstract_metadata. Study: PMID29443960.

67. Chen J, López-Moyado IF, Seo H et al. NR4A transcription factors limit CAR T cell function in solid tumours. Nature. 2019. PMID 30814732. DOI: 10.1038/s41586-019-0985-x. Access: abstract_metadata. Study: Engineered hCD19 CAR and NR4A platform.

68. Huang YJ, Ngiow SF, Baxter AE et al. Continuous expression of TOX safeguards exhausted CD8 T cell epigenetic fate. Science immunology. 2025. PMID 40053604. DOI: 10.1126/sciimmunol.ado3032. Access: abstract_metadata. Study: Related TOX program, inducible maintenance intervention.

69. Ongaro M, Thouenon R, Crespo I et al. The transcription factor IRF8 drives tumor-specific exhaustion in CD8+ T cells. The Journal of experimental medicine. 2026. PMID 42329236. DOI: 10.1084/jem.20252115. Access: full_text_partial. Study: PMID42329236.

Source-specific evidence cards

[1] Subsets of exhausted CD8+ T cells differentially mediate tumor control and respond to checkpoint blockade.

Shows: Progenitor exhausted T cells sustain tumour control and respond to checkpoint therapy; terminal cells have different responsiveness.

Does not show: Subset dependency in models does not establish terminal-state reversal or pan-tumour dominance; linked erratum details unexamined.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. ErratumIn: PMID31582823. Subclaims: SC2, SC3, SC4.

[2] Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing.

Shows: Single-cell NSCLC profiling resolves heterogeneous immune states and suppressive populations.

Does not show: Atlas association does not rank causal bottlenecks; linked erratum details unexamined.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. ErratumIn: PMID30093729. Subclaims: SC1, SC5.

[3] Transient rest restores functionality in exhausted CAR-T cells through epigenetic remodeling.

Shows: Transient CAR signaling rest restores function and remodels exhaustion-associated chromatin in engineered-cell experimental systems.

Does not show: CAR rest is a different intervention from PD1 blockade and does not prove terminal human tumour T-cell reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[4] MYB orchestrates T cell exhaustion and response to checkpoint inhibition.

Shows: MYB-dependent progenitor maintenance enables checkpoint-associated response and protection during chronic viral infection.

Does not show: Chronic-virus lineage dependency is a mechanism comparator rather than human solid-tumour causality.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC4.

[5] Mitochondrial stress induced by continuous stimulation under hypoxia rapidly drives T cell exhaustion.

Shows: Continuous antigen plus hypoxia promotes metabolic dysfunction; reducing hypoxic stress improves experimental immunotherapy.

Does not show: Metabolic and environmental effects are not isolated proof that checkpoints initiate the epigenetic program.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC5.

[6] Lineage tracing reveals clonal progenitors and long-term persistence of tumor-specific T cells during immune checkpoint blockade.

Shows: Multiregion human lung-cancer profiling links persistent progenitor clones in lymph nodes with tumour descendants.

Does not show: Clonal and computational relationships are not direct in-vivo human fate labeling or universal causal mediation.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2, SC4.

[7] TOX is a critical regulator of tumour-specific T cell differentiation.

Shows: Original Methods and Results show tumour-specific Tox deletion removes inhibitory receptor signatures but retains impaired cytokines and killing, followed by loss of persistence; bystander and acute-infection controls separate antigen context.

Does not show: TOX marker loss is not functional rescue. Experiments were not randomized or blinded and had no sample-size predetermination; human samples are observational, including infiltrate-selected breast cancers.

Access/status: Complete primary abstract and linked-status metadata examined. Selected body sections: 0:METHODS:paragraph0, 0:METHODS:paragraph2, 0:METHODS:paragraph6, 0:METHODS:paragraph11, 0:METHODS:paragraph13, 0:METHODS:paragraph14, 0:METHODS:paragraph17, 0:METHODS:paragraph18, 0:METHODS:paragraph23, 0:METHODS:paragraph31, 0:METHODS:paragraph37, 0:METHODS:paragraph38, 3:Main text (unsectioned body paragraphs):paragraph0, 3:Main text (unsectioned body paragraphs):paragraph1, 3:Main text (unsectioned body paragraphs):paragraph2, 3:Main text (unsectioned body paragraphs):paragraph3, 3:Main text (unsectioned body paragraphs):paragraph4, 3:Main text (unsectioned body paragraphs):paragraph5, 3:Main text (unsectioned body paragraphs):paragraph6, 3:Main text (unsectioned body paragraphs):paragraph7, 3:Main text (unsectioned body paragraphs):paragraph8, 3:Main text (unsectioned body paragraphs):paragraph9, 3:Main text (unsectioned body paragraphs):paragraph10, 3:Main text (unsectioned body paragraphs):paragraph11. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2, SC3, SC4.

[8] The primordial differentiation of tumor-specific memory CD8+ T cells as bona fide responders to PD-1/PD-L1 blockade in draining lymph nodes.

Shows: Tumour-draining-node memory-like cells supply checkpoint-responsive progenitors with epigenetic states distinct from exhausted tumour cells.

Does not show: A nodal source mechanism does not establish reversal of terminal intratumoral cells or clinical dominance.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[9] Spatiotemporal co-dependency between macrophages and exhausted CD8+ T cells in cancer.

Shows: Macrophage depletion and antigen-specific interactions implicate a reciprocal macrophage-exhausted T-cell circuit.

Does not show: Changing macrophages affects additional functions and does not establish a single universal exhaustion bottleneck.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC5.

[10] In vivo CD8+ T cell CRISPR screening reveals control by Fli1 in infection and cancer.

Shows: In-vivo CRISPR and Fli1 perturbation improve effector responses and tumour or infection control while preserving precursor potential.

Does not show: Different disease systems and transcription-factor perturbation do not isolate ordinary PD1-mediated terminal reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3.

[11] LAG-3 and PD-1 synergize on CD8+ T cells to drive T cell exhaustion and hinder autocrine IFN-γ-dependent anti-tumor immunity.

Shows: Combined LAG3 and PD1 genetic interventions improve experimental clearance and survival with autocrine IFNgamma dependency.

Does not show: Dual conditional pathway effects do not prove that all human resistant tumours are limited by fixed CD8 exhaustion.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC4.

[12] A unified atlas of CD8 T cell dysfunctional states in cancer and infection.

Shows: Integrated chromatin and transcriptional datasets plus viral experiments resolve exhaustion-associated trajectories.

Does not show: Reused atlas datasets share failure modes and viral mechanisms need separate tumour validation.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2.

[13] TOX reinforces the phenotype and longevity of exhausted T cells in chronic viral infection.

Shows: During chronic infection TOX loss transiently increases function but removes progenitor support and persistence with immunopathology.

Does not show: The outcome argues against a simple uniformly harmful lineage; it is not a human tumour treatment trial.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[14] Stat5 opposes the transcription factor Tox and rewires exhausted CD8+ T cells toward durable effector-like states during chronic antigen exposure.

Shows: STAT5 and orthogonal cytokine signaling partially reprogram exhausted cells and improve experimental antitumour function.

Does not show: Targeted cytokine rewiring is not equivalent to PD1 blockade or complete durable terminal-state reversal in patients.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[15] Hallmarks of CD8+ T cell dysfunction are established within hours of tumor antigen encounter before cell division.

Shows: Tumour-antigen exposure induces early functional and molecular dysfunction before cell division.

Does not show: Rapid onset in a defined model does not identify a checkpoint as the initiating cause across solid tumours.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2.

[16] Blockade of LAG-3 and PD-1 leads to co-expression of cytotoxic and exhaustion gene modules in CD8+ T cells to promote antitumor immunity.

Shows: Human melanoma nivolumab and relatlimab immune profiling shows enhanced cytotoxicity with retained exhaustion signatures and clonal changes.

Does not show: Molecular response is compatible with continuing exhaustion programs; it is not a terminal-fate-mediated survival trial.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC3, SC4.

[17] Endogenous Glucocorticoid Signaling Regulates CD8+ T Cell Differentiation and Development of Dysfunction in the Tumor Microenvironment.

Shows: Cell-resolved glucocorticoid receptor and myeloid steroidogenesis interventions improve antitumour immunity.

Does not show: Parallel environmental suppression and checkpoint effects need mediation separation.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[18] TOX and TOX2 transcription factors cooperate with NR4A transcription factors to impose CD8+ T cell exhaustion.

Shows: TOX/TOX2 perturbation in an engineered CAR setting implicates a feedback program with NR4A in hyporesponsiveness.

Does not show: Shared CAR platform with NR4A experiments limits independence; linked erratum details unexamined.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. ErratumIn: PMID31527257. Subclaims: SC2, SC3.

[19] Interleukin-10 receptor signaling promotes the maintenance of a PD-1int TCF-1+ CD8+ T cell population that sustains anti-tumor immunity.

Shows: IL10R and STAT3 support progenitor-like T cells and disease control in CLL models; human breast observations are complementary.

Does not show: CLL is not a solid tumour and protective IL10 effects are context dependent; no cross-solid-tumour clinical proof.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC5.

[20] Clonal replacement of tumor-specific T cells following PD-1 blockade.

Shows: Original Results profile 11 paired BCC patients and four SCC patients; bulk TCR sequencing in eight BCC patients and serial same-site biopsies address sampling. Expanded exhausted clones often are newly detected and retain exhausted phenotypes.

Does not show: The source of new clones was not identified and rare tumour-resident cells remain possible. Blood detection supports but does not prove recruitment; this is not terminal-state lineage perturbation in humans.

Access/status: Complete primary abstract and linked-status metadata examined. Selected body sections: 0:Main. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC4.

[21] Metabolic reprogramming of terminally exhausted CD8+ T cells by IL-10 enhances anti-tumor immunity.

Shows: IL10-Fc boosts metabolic and antitumour function of terminal exhausted cells with progenitor-independent effects in models.

Does not show: A distinct intervention shows some plasticity but does not imply PD1 blockade normally reverses all terminal states.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[22] Defining CD8+ T cells that provide the proliferative burst after PD-1 therapy.

Shows: TCF1-associated cells supply the proliferative burst after blockade in chronic LCMV.

Does not show: Chronic viral infection is a mechanism comparator rather than clinical solid-tumour efficacy or human fate proof.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC4.

[23] The PD-1/PD-L1-Checkpoint Restrains T cell Immunity in Tumor-Draining Lymph Nodes.

Shows: Tumour-draining-node targeted PDL1 intervention supports progenitor seeding and tumour control; human recurrence association is complementary.

Does not show: Human nodal association does not establish clinical source necessity or pan-cancer mechanism dominance.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC4.

[24] Determinants of anti-PD-1 response and resistance in clear cell renal cell carcinoma.

Shows: Longitudinal multiregion renal-cancer samples associate maintained expanded T-cell clones with nivolumab response.

Does not show: Maintenance correlations constrain a universal replacement account but do not prove causally reversed terminal cells.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC4.

[25] Inhibitory PD-1 axis maintains high-avidity stem-like CD8+ T cells.

Shows: PD1 signaling supports high-affinity stem-like T-cell maintenance; blockade can drive differentiation and death.

Does not show: The protective axis is context dependent and does not negate clinical benefits or prove human cross-tumour dominance.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[26] Combination anti-PD-1 and anti-CTLA-4 therapy generates waves of clonal responses that include progenitor-exhausted CD8+ T cells.

Shows: Serial melanoma profiling identifies distinct waves associated with CTLA4 and PD1 treatment and progenitor or effector responses.

Does not show: Temporal molecular waves do not uniquely establish reversal of pre-existing terminal cells or mediation of survival.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC4.

[27] Temporal single-cell tracing reveals clonal revival and expansion of precursor exhausted T cells during anti-PD-1 therapy in lung cancer.

Shows: Human lung-cancer paired samples support revival of pre-existing and new clones rather than a single universal clonal route.

Does not show: Clonal sampling and state inference do not directly trace terminal human T-cell reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC4.

[28] Single-cell CRISPR screens in vivo map T cell fate regulomes in cancer.

Shows: In-vivo CRISPR identifies RBPJ, IRF1, ETS1 and IKAROS-dependent transitions with intervention-induced reprogramming.

Does not show: Gene-specific cell rewiring is distinct from routine checkpoint treatment and may alter several lineage compartments.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[29] Intra-tumoral hypoxia promotes CD8+ T cell dysfunction via chronic activation of integrated stress response transcription factor ATF4.

Shows: ATF4-associated hypoxic stress can promote terminal-like dysfunction independently of TCR signaling; pathway intervention improves experimental function.

Does not show: Stress pathways provide an alternative input and do not establish exhaustion dominance in patients.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC5.

[30] The type 2 cytokine Fc-IL-4 revitalizes exhausted CD8+ T cells against cancer.

Shows: FcIL4 improves terminal exhausted T-cell function and checkpoint synergy in experimental tumour contexts.

Does not show: Functional plasticity with another cytokine treatment does not establish ordinary PD1-mediated complete lineage reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[31] Hypoxia drives CD39-dependent suppressor function in exhausted T cells to limit antitumor immunity.

Shows: Conditional CD39 perturbation identifies suppressive exhausted T-cell activity under hypoxic tumour conditions and improves tumour control.

Does not show: A suppressor mechanism in specified models does not determine prevalence or dominance across human solid cancers.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[32] Neoadjuvant PD-1 blockade induces T cell and cDC1 activation but fails to overcome the immunosuppressive tumor associated macrophages in recurrent glioblastoma.

Shows: Neoadjuvant glioblastoma immune profiling shows T-cell and progenitor engagement alongside persistent myeloid barriers.

Does not show: Target engagement does not guarantee clinical efficacy or removal of non-T-cell resistance.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC3, SC5.

[33] First-line nivolumab plus chemotherapy versus chemotherapy alone for advanced gastric, gastro-oesophageal junction, and oesophageal adenocarcinoma (CheckMate 649): a randomised, open-label, phase 3 trial.

Shows: Randomized CheckMate649 shows nivolumab plus chemotherapy improves OS and PFS in its primary gastric-cancer CPS-defined population.

Does not show: Combination efficacy does not isolate CD8 exhaustion mediation or imply identical benefits in every biomarker context.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[34] Final, 10-Year Outcomes with Nivolumab plus Ipilimumab in Advanced Melanoma.

Shows: Ten-year CheckMate067 follow-up supports durable melanoma survival benefit with nivolumab-containing regimens.

Does not show: This is follow-up of the same randomized cohort as the initial report, not an independent replication or terminal-state fate assay.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[35] Neoadjuvant Nivolumab plus Chemotherapy in Resectable Lung Cancer.

Shows: Randomized CheckMate816 reports neoadjuvant lung-cancer event-free survival and pathological response benefit.

Does not show: The interim OS result in this report did not cross its prespecified threshold; chemotherapy combination does not isolate terminal exhaustion reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[36] Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma.

Shows: Initial CheckMate067 randomization demonstrates melanoma PFS benefit from nivolumab-containing regimens with higher combination toxicity.

Does not show: PFS and the subsequent ten-year follow-up share one cohort; neither directly proves terminal CD8 epigenetic reversal. Linked erratum details unexamined.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. ErratumIn: PMID31442371. Subclaims: SC3, SC5.

[37] Improved survival with ipilimumab in patients with metastatic melanoma.

Shows: Randomized ipilimumab treatment improves metastatic melanoma survival in the enrolled HLA-selected population.

Does not show: CTLA4 acts in broader immune compartments and the trial does not identify terminal CD8 exhaustion as the dominant mediator; linked erratum details unexamined.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. ErratumIn: PMIDnot supplied. Subclaims: SC3, SC5.

[38] Nivolumab plus Ipilimumab versus Sunitinib in Advanced Renal-Cell Carcinoma.

Shows: CheckMate214 improves OS and response in the prespecified intermediate or poor-risk renal-cancer group.

Does not show: PFS did not meet the prespecified significance threshold and favorable-risk or mechanistic generalization is separate.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[39] Pembrolizumab versus Ipilimumab in Advanced Melanoma.

Shows: KEYNOTE006 randomization demonstrates melanoma survival and PFS benefit of pembrolizumab versus ipilimumab.

Does not show: A comparative clinical endpoint does not identify which CD8 lineage mediated benefit.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[40] Nivolumab versus Docetaxel in Advanced Nonsquamous Non-Small-Cell Lung Cancer.

Shows: CheckMate057 improves previously treated nonsquamous NSCLC survival with nivolumab.

Does not show: Median PFS was not uniformly favorable and PD-L1-dependent heterogeneity does not establish cross-tumour dominance.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[41] Nivolumab versus Docetaxel in Advanced Squamous-Cell Non-Small-Cell Lung Cancer.

Shows: CheckMate017 improves previously treated squamous NSCLC survival regardless of measured PD-L1 groups.

Does not show: Clinical efficacy is not an assay of exhaustion causality or terminal-state reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[42] LRRC15+ myofibroblasts dictate the stromal setpoint to suppress tumour immunity.

Shows: Original Results show LRRC15 CAF depletion, toxin and genotype controls, CD8-depletion reversal and improved checkpoint-associated tumour control in subcutaneous and orthotopic pancreatic models.

Does not show: Ablation changes total fibroblast abundance and immune state, so its full mechanism is not a single physical barrier; progression-to-size or ulceration survival is not all-cause human survival.

Access/status: Complete primary abstract and linked-status metadata examined. Selected body sections: 4:LRRC15+ CAF depletion slows tumour growth, 6:LRRC15+ CAFs impede CD8+ T cell function, 7:LRRC15+ CAF ablation boosts ICB responses, 8:Discussion. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[43] Intratumoral CD4+ T Cells Mediate Anti-tumor Cytotoxicity in Human Bladder Cancer.

Shows: Human bladder cytotoxic CD4 cells kill autologous tumour in an MHCII-dependent assay and are restrained by Tregs.

Does not show: Ex-vivo killing and response prediction do not establish in-vivo clinical necessity; CD8 exhaustion is not the only candidate compartment.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[44] PD-1+ regulatory T cells amplified by PD-1 blockade promote hyperprogression of cancer.

Shows: PD1-positive Treg biology and conditional experiments supply a potential immunosuppressive consequence of blockade alongside gastric hyperprogression associations.

Does not show: Human observational hyperprogression is not randomized proof that checkpoint treatment causes accelerated disease.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[45] CD25-Treg-depleting antibodies preserving IL-2 signaling on effector T cells enhance effector activation and antitumor immunity.

Shows: CD25-directed Treg depletion preserving IL2 signaling improves selected preclinical immune responses.

Does not show: Cross-system preclinical function is not established human solid-tumour efficacy or a universal CD8-exhaustion bottleneck.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[46] CXCR3 expression in regulatory T cells drives interactions with type I dendritic cells in tumors to restrict CD8+ T cell antitumor immunity.

Shows: Conditional CXCR3 Treg manipulation releases DC1 cross-presentation and antitumour immunity with checkpoint combinations.

Does not show: Changing upstream priming provides another constraint and does not establish exhausted CD8 states as autonomous initiators.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[47] Single-Cell RNA Sequencing Reveals Stromal Evolution into LRRC15+ Myofibroblasts as a Determinant of Patient Response to Cancer Immunotherapy.

Shows: Pancreatic and multicancer profiling associates LRRC15 fibroblast programs with checkpoint nonresponse.

Does not show: Clinical correlations are not causal depletion results; this shares a fibroblast biological program with later intervention work.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC5.

[48] Overcoming resistance to checkpoint blockade therapy by targeting PI3Kγ in myeloid cells.

Shows: PI3Kgamma targeting alters macrophage programs and improves experimental checkpoint responses across tumour models.

Does not show: The intervention is not fully cell-specific and preclinical benefit is not human clinical efficacy.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[49] Treg cells promote immunotherapy-induced immune evasion by restraining CD4 T cell control of MHC-I-deficient metastatic pancreatic cancer.

Shows: MHC-I-deficient pancreatic models reveal TAP1-associated presentation and CD4 or Treg immune constraints; CTLA4 intervention improves control and metastatic endpoints.

Does not show: Experimental loss and engineered rescue do not determine frequency or dominance in unselected human pancreatic cancer.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[50] Epigenetic stability of exhausted T cells limits durability of reinvigoration by PD-1 blockade.

Shows: PDL1 blockade in chronic infection yields limited chromatin remodeling and transient response followed by renewed exhaustion; antigen removal does not produce ordinary memory.

Does not show: Chronic-virus stability is not absolute irreversibility under every intervention or proof of terminal human tumour fate.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC4.

[51] Chromatin states define tumour-specific T cell dysfunction and reprogramming.

Shows: Original Results distinguish early plastic and later fixed mouse tumour chromatin states by IL15 reprogramming, with B16-OVA confirmation and matching features in human PD1-high TILs.

Does not show: The human marker and chromatin comparison does not directly test patient T-cell reversibility; fixed means resistant to the tested conditions, not forever irreprogrammable.

Access/status: Complete primary abstract and linked-status metadata examined. Selected body sections: 1:CD8 T cell chromatin changes during infection, 2:Chromatin state dynamics of TST dysfunction, 3:Chromatin states are associated with reprogrammability, 4:Surface proteins associated with chromatin states, 5:Memory T cells enter state 2 dysfunction in tumors, 6:Chromatin accessibility in human TIL, 7:Discussion. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2, SC4.

[52] An intra-tumoral niche maintains and differentiates stem-like CD8 T cells.

Shows: Human tumour stem-like T cells occupy APC-rich niches and correlate with infiltration and nonprogressive disease.

Does not show: Spatial association does not prove that niche absence causes clinical escape or that terminal exhaustion dominates it.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2, SC5.

[53] PD-1 blockade induces responses by inhibiting adaptive immune resistance.

Shows: Serial melanoma biopsies associate pre-existing margin CD8 cells and PD1 or PDL1 proximity with pembrolizumab response, with independent predictive validation.

Does not show: Predictive association and proliferation are not a necessity perturbation or proof of terminal-state mediation.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC3.

[54] Defining T Cell States Associated with Response to Checkpoint Immunotherapy in Melanoma.

Shows: Melanoma single-cell profiling identifies TCF7-associated states predicting checkpoint outcome and complementary experimental targets.

Does not show: Predictive human states do not establish fate reversal; linked erratum details unexamined.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. ErratumIn: PMID30633907. Subclaims: SC1, SC2, SC4.

[55] Mutations Associated with Acquired Resistance to PD-1 Blockade in Melanoma.

Shows: Paired melanoma relapse biopsies and functional tests identify JAK1, JAK2 and B2M resistance-associated defects affecting IFNgamma signaling or MHC-I presentation.

Does not show: Four patients do not estimate population prevalence or prove every escape event independently; acquired resistance is a distinct stage.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[56] Primary Resistance to PD-1 Blockade Mediated by JAK1/2 Mutations.

Shows: Human melanoma and mismatch-repair-deficient colon nonresponse plus cell-line assays implicate JAK1/2 loss and absent IFNgamma signaling.

Does not show: Small selected samples and in-vitro mechanism do not establish universal primary resistance or ranking versus exhaustion.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[57] Allele-Specific HLA Loss and Immune Escape in Lung Cancer Evolution.

Shows: Allele-specific HLA loss and tumour phylogenetic patterns support immune selection and presentation escape in NSCLC.

Does not show: Computational evolutionary evidence does not directly restore HLA or quantify causal superiority over exhaustion in patients.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC5.

[58] Pembrolizumab versus Chemotherapy for PD-L1-Positive Non-Small-Cell Lung Cancer.

Shows: KEYNOTE024 demonstrates PFS and OS benefit in prospectively selected PD-L1-high NSCLC lacking specified oncogenic drivers.

Does not show: The selection was before outcome but defines a narrower context; clinical benefit is not terminal-state reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[59] First-Line Nivolumab in Stage IV or Recurrent Non-Small-Cell Lung Cancer.

Shows: CheckMate026 first-line NSCLC randomization does not improve PFS in its prespecified PD-L1-at-least-five-percent population; OS is similar.

Does not show: This null pertains to its treatment, comparator and population, not absence of all exhausted cells or failure of every checkpoint regimen.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[60] Effect of Nivolumab vs Bevacizumab in Patients With Recurrent Glioblastoma: The CheckMate 143 Phase 3 Randomized Clinical Trial.

Shows: Original Methods and Results confirm powered CheckMate143 OS null in recurrent glioblastoma after 301 deaths; nivolumab PFS and response are inferior to bevacizumab, and steroid or MGMT subsets are hypothesis generating.

Does not show: Archival biomarker tissue, unstandardized MGMT and small subgroups limit mechanisms; subgroup trends do not rescue the null primary outcome.

Access/status: Complete primary abstract and linked-status metadata examined. Selected body sections: 1:Methods, 2:Results, 3:Discussion. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[61] Pembrolizumab alone or combined with chemotherapy versus chemotherapy as first-line therapy for advanced urothelial carcinoma (KEYNOTE-361): a randomised, open-label, phase 3 trial.

Shows: KEYNOTE361 combination treatment misses both multiplicity-adjusted primary thresholds; subsequent monotherapy OS comparisons are exploratory.

Does not show: Nominal P values or exploratory subsets cannot override the testing hierarchy; the trial does not directly measure exhaustion mediation.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[62] Pembrolizumab in Microsatellite-Instability-High Advanced Colorectal Cancer.

Shows: KEYNOTE177 improves first-line colorectal-cancer PFS in the prospectively MSI-high or dMMR population.

Does not show: OS is not mature in this examined interim report and results cannot be generalized to microsatellite-stable colorectal cancer.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[63] Tumor-Specific T Cell Dysfunction Is a Dynamic Antigen-Driven Differentiation Program Initiated Early during Tumorigenesis.

Shows: Inducible autochthonous liver tumour and defined antigen show early antigen-driven dysfunction transitioning from reversible to less reversible states.

Does not show: Model antigen and restricted TCR context do not establish human checkpoint-driven initiation; this shares the liver platform with later chromatin and TOX work.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2, SC4.

[64] TOX transcriptionally and epigenetically programs CD8+ T cell exhaustion.

Shows: TOX loss and molecular profiling show a defined feed-forward regulator of exhaustion differentiation under persistent stimulation.

Does not show: Lineage commitment mainly from experimental chronic stimulation does not alone prove that removing TOX restores human tumour control.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC1, SC2.

[65] Intratumoral Tcf1+PD-1+CD8+ T Cells with Stem-like Properties Promote Tumor Control in Response to Vaccination and Checkpoint Blockade Immunotherapy.

Shows: Ablation of TCF1-positive PD1-positive tumour cells restricts checkpoint response and identifies stem-like subset expansion as its source.

Does not show: The source of benefit is distinct from reversal of terminal exhaustion and human observations remain complementary rather than equivalent causal tests.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[66] TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells.

Shows: Human urothelial fibroblast TGFbeta signatures associate with exclusion and nonresponse; combined TGFbeta and PDL1 blockade releases infiltration and regression in a matching mouse model.

Does not show: Human signatures are observational and combination effects do not establish TGFbeta alone as a universal human dominant bottleneck.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC3, SC5.

[67] NR4A transcription factors limit CAR T cell function in solid tumours.

Shows: NR4A triple-knockout hCD19-reactive CAR T cells improve regression and survival in engineered-antigen mouse solid tumours with chromatin changes.

Does not show: Artificial antigen and CAR system limit transfer to endogenous human cancers; shared TOX/TOX2 CAR platform limits independence.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3.

[68] Continuous expression of TOX safeguards exhausted CD8 T cell epigenetic fate.

Shows: Induced TOX ablation after exhaustion establishment changes transcription and chromatin, causes apoptotic cell loss and permits more effector-like fate flexibility.

Does not show: Continuous maintenance and plasticity coexist with survival costs; the abstract does not specify the whole model scope or prove clinical terminal human tumour reversal.

Access/status: Complete primary abstract and linked-status metadata examined. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC4.

[69] The transcription factor IRF8 drives tumor-specific exhaustion in CD8+ T cells.

Shows: IRF8 gain and loss perturbation alters TOX, effector molecules and experimental tumour control; chronic IFN-I repression differs from tumour expression.

Does not show: Tumour-versus-virus regulation supports context specificity rather than a universal common pathway or direct human dominance.

Access/status: Complete primary abstract and linked-status metadata examined. Selected body sections: 1:Results:paragraph19, 1:Results:paragraph21, 1:Results:paragraph24, 1:Results:paragraph25, 1:Results:paragraph26, 1:Results:paragraph27, 1:Results:paragraph30. No complete-paper or supplement audit claimed. none_found. Subclaims: SC2, SC3, SC5.

Documentary record

Every grade on this site is reproducible from its record. The files below are the posted edition exactly as assessed.

• SHA256.txt
• checkpoint-exhaustion-tumor-evasion.review.md
• checkpoint-exhaustion-tumor-evasion.scorecard.json
• codex-correction-provenance.json
• consensus-addendum.md
• consensus-validation.json
• edition-approval-audit.json
• validation-report.json
• validation-round2-2026-10-14.json
• validation.txt

Sources: 69 · load-bearing claims: 75 · verified: 75 · unverified: 0 · contradicted: 0 · schema 5.3