Tumour-infiltrating clonal haematopoiesis

The authors proposed this themselves: the prognostic effect of tumor-infiltrating clonal hematopoiesis (TI-CH) was measured almost entirely in white patients at two centres, and they want it checked in bigger cohorts whose patients are not mostly of European ancestry. Three later cohorts have now checked the size of the effect, but none has reported it by ancestry, so the design below is an extension of those, not a first test.

1. Has anyone done this since?

Status

partly addressed

Works since

• Tumor-Infiltrating Clonal Hematopoiesis (letter) (Spetzler, Antonarakis and Sledge 2025, N Engl J Med, doi:10.1056/NEJMc2507106)
  Caris Life Sciences reported, in a letter to the journal, an independent validation in 3,255 matched tumor-blood sequencing cases with TI-CH most common in non-small cell lung cancer (NSCLC) and about 30% higher risk of death; the letter is not open access, and these numbers come from the company's press summary, which does not report ancestry.
• Tumor-Infiltrating Clonal Hematopoiesis and Pan-Cancer Prognosis in Patients With Solid Tumors (Yun, Chen, Qin, Wang and Song 2026, JAMA Oncology, doi:10.1001/jamaoncol.2026.1036)
  In 10,571 Genomics England 100,000 Genomes Project patients, TI-CH defined from tumor-only whole-genome sequencing (variant allele frequency 2% to 30%) was present in 18.38% and associated with worse pan-cancer overall survival (hazard ratio 1.13, 95% CI 1.02-1.25), strongest in breast cancer (1.95); a United Kingdom cohort, with no ancestry breakdown in the abstract.
• The interplay of clonal hematopoiesis and cancer: Clinical implications in oncology (Tanguay et al. 2026, Cancer Treatment Reviews, doi:10.1016/j.ctrv.2026.103109)
  A review that states TI-CH has been associated with inferior survival 'in some cancer cohorts', which is the state of the replication record, and names no diverse-ancestry cohort.

Searched

All 49 titles in the citing-works file, with Europe PMC abstracts for the eight that mention clonal hematopoiesis outcomes; Europe PMC searches for 'tumor-infiltrating clonal hematopoiesis' (33 hits), for clonal hematopoiesis with ancestry, ethnicity, African American, Hispanic or East Asian in the title, and for clonal hematopoiesis cancer cohorts in Japanese, Chinese, Korean, Asian, African or Hispanic populations; one web search for a 2026 TI-CH validation cohort.

Still open

The size of the TI-CH effect has been replicated three times in European-ancestry-dominated cohorts, but no cohort has reported TI-CH prevalence or its hazard ratio within patients of African, East Asian, South Asian or Hispanic ancestry, so whether the association transfers across ancestries is untested.

Registry query

4 of 49 records in the open node mention any of: tumor-infiltrating clonal hematopoiesis, chip, ancestry, all-cause mortality, solid tumor, replication. arcadia-alcalase-path: Alcalase pretreatment for CRISPR-Cas9 HDR knock-in in cell-walled Chlamydomonas reinhardtii (iced project); robin-damd-screen-r1-arpe19: Round 1 phagocytosis screen of Robin's candidates in ARPE-19 cells; robin-damd-screen-r2-arpe19: Round 2 phagocytosis screen of Robin's candidates in ARPE-19 cells

2. A proposed experiment

Question

Does TI-CH carry the same prevalence and the same adjusted hazard for all-cause mortality in patients of non-European ancestry as in European-ancestry patients?

Hypothesis

Within each self-reported race or genetic-ancestry stratum, TI-CH (tumor VAF >= 2%) raises adjusted all-cause mortality relative to blood-only CHIP with a hazard ratio whose 95% CI includes 1.17.

Attempt type

observational-study

Design

A pre-registered retrospective cohort analysis pooling matched tumor-blood panel sequencing from holders whose patients are not mostly of European ancestry, run with the paper's own definitions: clonal hematopoiesis of indeterminate potential (CHIP) at blood VAF >= 2% across the 77-gene list, TI-CH at tumor VAF >= 2% in at least one region. Each patient is assigned both self-reported race or ethnicity and genetic ancestry inferred from the matched normal. The three exposure categories are no CHIP, blood-only CHIP and TI-CH. The outcome is all-cause mortality in a Cox model adjusted for age, sex, smoking, prior treatment, stage, histology and tumor purity, exactly as in the record, fitted within each ancestry stratum and then with an ancestry-by-TI-CH interaction term. The pre-registered test is whether the stratum-specific hazard ratio for TI-CH versus blood-only CHIP differs from the pooled 1.17 (95% CI 1.06-1.29). Secondary analyses repeat the gene-level prevalence (TET2 highest, PPM1D lowest) by stratum, because Bick 2020 found a germline TET2 enhancer variant specific to African ancestry that raises CHIP risk.

Controls

• reference-case The European-ancestry stratum of the pooled data, analysed first with the same code and covariates.
  Rules out: That a null in another stratum is caused by the pooled pipeline, panel or covariate set rather than by ancestry: if the reference stratum does not reproduce a TI-CH hazard ratio near 1.36 versus no CHIP, nothing else in the analysis can be read.
• background A pre-specified outcome that TI-CH has no plausible route to, such as death from external causes, modelled with the same covariates.
  Rules out: Residual confounding by frailty or access to care that differs between ancestry groups and would also inflate the hazard for an unrelated outcome.
  Vocabulary note: This is a negative-control outcome in the epidemiological sense (an outcome TI-CH should not move); no control kind in the vocabulary describes it.

Sensitivity

The target effect is the record's TI-CH versus blood-only CHIP hazard ratio of 1.17 (95% CI 1.06-1.29), estimated in 22,141 patients of whom 7,450 had CHIP and 1,974 TI-CH; a stratum would need a power calculation from its own CHIP count and death count, which the record does not give for any ancestry group, so the detectable hazard ratio per stratum cannot be set here. MSK-IMPACT's Table 1 shows 41% of NSCLC patients with race 'Other/Missing', which is the first number a data holder would have to resolve.

Would count as

Supported → positive · opposed → negative-not-replicated
Inconclusive if: Any non-European stratum has fewer deaths than the power calculation requires, or the reference European stratum fails to reproduce a hazard ratio near the record's, or more than about a third of patients lack usable ancestry data, as in MSK-IMPACT.

Known before

• CHIP prevalence and its driver-gene mix differ by ancestry: in 97,691 TOPMed genomes a germline locus at TET2 that raises CHIP risk was specific to individuals of African ancestry. (Bick et al. 2020, Nature, doi:10.1038/s41586-020-2819-2)
  So: The gene-level TI-CH prevalence (TET2 highest) is expected to shift by ancestry, so the design must report gene-level prevalence per stratum rather than assume it.
• Clonal hematopoiesis was detected in 12% of 495 multiple myeloma patients and was associated with inferior overall survival specifically in non-Hispanic Black patients (hazard ratio 4.36, 95% CI 1.36-14.00). (Peres et al. 2022, Blood Advances, doi:10.1182/bloodadvances.2021006652)
  So: The only published signal that the prognostic weight of clonal hematopoiesis may differ by race, in a hematologic cancer, with a wide interval; it motivates the interaction term rather than a pooled estimate.
• In 1,550 treatment-naive CPTAC patients, 18.3% had CH and one third of CH mutations were detectable in tumor DNA (CH-Tum), which was associated with worse survival and an inflammatory tumor microenvironment. (Buttigieg, Vlasschaert, Bick, Vanner and Rauh 2025, Cell Reports Medicine, doi:10.1016/j.xcrm.2025.101989)
  So: An independent, contemporaneous cohort with the same construct and the same direction of effect; CPTAC also lacks an ancestry breakdown, so it does not close this issue but could be pooled.
• In 415 Korean NSCLC patients resected and given adjuvant therapy, pre-operative CH (21%) carried a multivariable overall survival hazard ratio of 1.58 (95% CI 1.08-2.32). (Yun et al. 2023, Genome Medicine, doi:10.1186/s13073-023-01266-4)
  So: An East Asian NSCLC cohort with blood CHIP and outcome data already exists; adding tumor genotyping of the CHIP mutations to it would be the fastest route to an East Asian TI-CH estimate.

Searched: Europe PMC searches for clonal hematopoiesis with ancestry, ethnicity or named populations in the title (254 hits, top 12 read), for clonal hematopoiesis cancer cohorts in Asian, African or Hispanic populations (37 hits), and for clonal hematopoiesis in solid tumors with survival since 2021 (15 hits); abstracts read for Bick 2020, Peres 2022, Buttigieg 2025, Yun 2023; I looked for and did not find any TI-CH or tumor-genotyped CHIP study in a Chinese, Japanese, Korean, African or Latin American cohort.

Feasibility

Holders of matched tumor-normal clinical sequencing with outcomes and ancestry: MSK-IMPACT, Caris Life Sciences, the Genomics England cohort, CPTAC, and East Asian hospital cohorts such as the Korean NSCLC cohort of Yun 2023; or an AACR Project GENIE-style consortium if matched normals are available. · 6 to 12 months for data access and a pre-registered analysis, longer if tumor genotyping must be added to an existing blood-CHIP cohort · low, because the sequencing already exists and the cost is analyst time and data-access agreements
Blockers: Ancestry is missing for 41% of MSK-IMPACT NSCLC patients and is unreported in the Caris and Genomics England analyses; Most non-European cohorts sequence tumor only, so blood-only CHIP cannot be separated from TI-CH without adding a blood panel; Data-access agreements across institutions

Strongest objection

An epidemiologist would say that ancestry strata in clinical sequencing cohorts are confounded by institution, stage at diagnosis, treatment access and panel version, so a different hazard ratio in one stratum will be read as a cohort artefact rather than biology, and the interaction test will be underpowered because non-European patients are a minority in every holder's data; the honest version of this study is a prevalence estimate, not a hazard ratio.

Not derivable from what is published

• The number of deaths per ancestry stratum in any existing cohort, needed for the power calculation
• Whether the Caris 3,255-case analysis recorded race or ancestry at all
• Whether the Yun 2023 Korean cohort has archived tumor DNA for genotyping

3. Send it to a lab

Send to a lab

This design is a reanalysis of existing data. There is nothing for a bench to run; it needs the data holder.
No bench job exists: this is a reanalysis of existing matched tumor-blood sequencing with survival and ancestry data, and the only laboratory step, if any, is genotyping known CHIP mutations in archived tumor DNA from an existing blood-CHIP cohort.

The authors could show that TET2-mutant clones infiltrate tumors most often and PPM1D-mutant clones least often, but not whether a TET2-driven TI-CH is more lethal than a DNMT3A-driven one; they asked for larger cohorts. Since publication, gene-level outcome estimates have appeared in breast cancer and pan-cancer, and two independent groups found TET2-mutant clonal hematopoiesis improves response to immune checkpoint therapy, so the design below is an extension that must separate treatment context, not a first look.

1. Has anyone done this since?

Status

partly addressed

Works since

• Tumor-Infiltrating Clonal Hematopoiesis and Pan-Cancer Prognosis in Patients With Solid Tumors (Yun, Chen, Qin, Wang and Song 2026, JAMA Oncology, doi:10.1001/jamaoncol.2026.1036)
  Gene-level TI-CH outcomes in 10,571 Genomics England patients: GATA2 variants carried worse pan-cancer overall survival (hazard ratio 3.00, 95% CI 1.61-5.59) and TET2 variants worse breast cancer survival (2.92, 95% CI 1.59-5.37); tumor-only sequencing, so blood-only CHIP could not be separated and germline or tumor-derived variants in these genes could contaminate the calls.
• Clonal Hematopoiesis of Indeterminate Potential Influences Breast Cancer Outcomes in a Genotype-Specific Manner (Reed et al. 2025, Clinical Cancer Research, doi:10.1158/1078-0432.CCR-25-2009)
  In 125 breast cancer patients, non-DNMT3A CHIP and high-burden CHIP shortened distant metastasis-free survival, and in chimeric mice Tet2-CHIP produced larger primary tumors and more lung metastasis while Dnmt3a-CHIP did not differ from controls; blood CHIP, not TI-CH, and a small human cohort.
• TET2-mutant clonal hematopoiesis enhances macrophage antigen presentation and improves immune checkpoint therapy in solid tumors (Herbrich et al. 2025, Cancer Cell, doi:10.1016/j.ccell.2025.09.011)
  In 35,971 NSCLC and 25,064 colorectal cancer patients, TET2-mutant clonal hematopoiesis was associated with improved outcome specifically with immune checkpoint therapy, and Tet2-mutant tumor macrophages cross-primed CD8 T cells; this reverses the sign of the TET2 effect in one treatment context.
• Response to Immune Checkpoint Blockade Is Enhanced in the Presence of Hematopoietic TET2 Inactivation (Rondeau et al. 2025, Cancer Research, doi:10.1158/0008-5472.CAN-24-3329)
  Independent mouse and patient evidence in the same direction: TET2-mutant clonal hematopoiesis gave six-fold greater odds of clinical benefit from checkpoint blockade in melanoma, and the effect was lost at 20% Tet2-mutant hematopoiesis in mice, so clone size matters.
• Clonal hematopoiesis is associated with reduced response to atezolizumab in non-small cell lung cancer (Chat et al. 2025, Journal for ImmunoTherapy of Cancer, doi:10.1136/jitc-2025-011565)
  In 1,281 NSCLC patients from five anti-PD-L1 trials, any CHIP at VAF >= 2% raised the odds of non-response (odds ratio 1.69, 95% CI 1.08-2.63) with no effect in comparator arms; not gene-resolved, and pointing the opposite way to the two TET2 studies.

Searched

All 49 citing titles, with abstracts for Yun 2026, Reed 2025, Herbrich 2025, Rondeau 2025, Chat 2025 and Yang 2026; Europe PMC searches for clonal hematopoiesis in solid tumors with gene-specific survival (335 hits, top 12 read) and for clonal hematopoiesis solid-tumor or lung cancer prognosis papers since 2021 (15 hits, all titles read).

Still open

No study has estimated gene-specific hazard ratios for TI-CH, as opposed to blood CHIP, in treatment-naive patients with matched blood to exclude blood-only CHIP, and the sign of the TET2 effect appears to flip between untreated or chemotherapy-treated patients (worse) and checkpoint-treated patients (better), which no single dataset has tested.

Registry query

0 of 49 records in the open node mention any of: tet2, dnmt3a, ppm1d, tumor-infiltrating clonal hematopoiesis, immune checkpoint, all-cause mortality. The open node holds biology and chemistry so far, so this is the expected answer; a private node of a clinical group would be asked the same question.

2. A proposed experiment

Question

Among patients with TI-CH, does the driver gene (TET2, DNMT3A, ASXL1, PPM1D) change the adjusted hazard for all-cause mortality, and does the TET2 effect reverse under immune checkpoint therapy?

Hypothesis

TET2-driven TI-CH carries a higher all-cause mortality hazard than DNMT3A-driven TI-CH in patients not given checkpoint inhibitors, and a lower hazard than DNMT3A-driven TI-CH in patients given them.

Attempt type

secondary-analysis

Design

A pre-registered individual-patient reanalysis of the MSK-IMPACT pan-cancer cohort (7,450 CHIP patients, 1,974 with TI-CH; 1,191 TET2, 3,753 DNMT3A, 392 ASXL1, 622 PPM1D CHIP carriers), extended with any matched tumor-blood cohort that will share data (Caris, CPTAC). Exposure is a six-level factor: no CHIP; blood-only CHIP; TI-CH driven by TET2, DNMT3A, ASXL1 or PPM1D (patients with two driver genes assigned to the gene with the higher tumor VAF, with a sensitivity analysis dropping them). Outcome is all-cause mortality in the record's Cox model, adjusted for age, sex, smoking, stage, histology, prior treatment, tumor purity and blood VAF, with the tumor-over-blood VAF ratio as a continuous covariate so that clone size is not mistaken for gene effect. The pre-registered contrast is TET2 TI-CH versus DNMT3A TI-CH. A second model adds receipt of immune checkpoint therapy as a time-varying covariate and tests the gene-by-checkpoint interaction, which is the only way to reconcile the record with Herbrich 2025 and Rondeau 2025.

Controls

• reference-case The pooled TI-CH versus blood-only CHIP contrast in the same data, which must return a hazard ratio near 1.17 (95% CI 1.06-1.29) before gene-level contrasts are read.
  Rules out: A reanalysis pipeline or covariate set that no longer reproduces the record's main effect.
• permutation-null Driver-gene labels shuffled among TI-CH patients 1,000 times to give the distribution of the TET2 versus DNMT3A hazard-ratio difference under no gene effect.
  Rules out: A gene-level difference that is the expected spread of estimates across groups of these sizes rather than a real difference.
• within-subject For each gene, blood-only CHIP carriers of that gene as the comparator for TI-CH carriers of that gene.
  Rules out: A gene effect that is really a blood-clone effect (for example, PPM1D clones being therapy-selected and marking prior chemotherapy) rather than an infiltration effect.
  Vocabulary note: Used for a within-gene comparison: blood-only CHIP of the same gene versus TI-CH of the same gene.

Sensitivity

The record gives the group sizes (1,191 TET2 and 3,753 DNMT3A CHIP carriers, of whom 33% and 25% had TI-CH, so roughly 381 and 926 TI-CH patients) but not their death counts, so the minimum detectable hazard-ratio difference cannot be computed here; what is needed is the number of deaths in each gene-level TI-CH group from the MSK-IMPACT survival file. With 381 versus 926 patients the comparison will not detect a hazard-ratio difference much smaller than the overall TI-CH effect (1.17), so the pre-registration should state that a null here is uninformative below that size.

Would count as

Supported → positive · opposed → negative-no-association
Inconclusive if: Fewer deaths than the power calculation requires in the TET2 TI-CH group, or the checkpoint-treated subgroup is too small to fit the interaction, or the tumor-over-blood VAF ratio absorbs the gene effect so that gene and clone size cannot be separated.

Known before

• Cancer therapy selects clones with DNA damage response mutations (PPM1D, TP53), so PPM1D CHIP marks prior cytotoxic exposure. (Bolton et al. 2020, Nature Genetics, doi:10.1038/s41588-020-00710-0)
  So: PPM1D-driven TI-CH is confounded by treatment history; the within-gene blood-only comparator and the prior-treatment covariate are there for this.
• In CANTOS, canakinumab's cardiovascular benefit was concentrated in TET2-mutant CHIP carriers, and TET2 carriers on canakinumab had the lowest incidence of non-hematological malignancy. (Svensson et al. 2022, JAMA Cardiology; Woo et al. 2024, Cancer Prevention Research, doi:10.1001/jamacardio.2022.0386; doi:10.1158/1940-6207.CAPR-23-0342)
  So: TET2 is already the gene with a treatment-specific signal, which supports testing a gene-by-treatment interaction rather than a gene main effect only.
• A validated clonal hematopoiesis risk score (CHRS) predicts myeloid malignancy; in the record, TI-CH was more frequent at intermediate or high CHRS (45% vs 32%). (Weeks et al. 2023, NEJM Evidence, doi:10.1056/EVIDoa2200310)
  So: Gene-level TI-CH risk must be adjusted for, or stratified by, CHRS so that a TET2 effect is not a high-risk-clone effect.
• In 97,691 TOPMed genomes, blood-cell, lipid and inflammatory trait associations were specific to different CHIP driver genes. (Bick et al. 2020, Nature, doi:10.1038/s41586-020-2819-2)
  So: Gene-specific biology is established for CHIP in blood, which makes a gene-specific TI-CH effect plausible and worth a pre-registered test.

Searched: Europe PMC searches for clonal hematopoiesis with solid tumor, TET2, DNMT3A or PPM1D and gene-specific survival (335 hits, top 12 read) and for clonal hematopoiesis solid-tumor prognosis since 2021 (15 hits); abstracts for Bolton 2020, Svensson 2022, Woo 2024, Bick 2020; I looked for and did not find any published gene-level TI-CH hazard ratio in a matched tumor-blood cohort.

Feasibility

The MSK-IMPACT computational oncology group that holds the 49,351-patient dataset, or any group granted access through cBioPortal or AACR GENIE with matched normals; Caris and CPTAC could add patients. · 3 to 6 months · low, because no new sequencing is needed and the analysis is a re-specification of an existing Cox model
Blockers: Death counts per gene-level TI-CH group are not public; Checkpoint-therapy exposure must be extracted from treatment records; Patients with multiple driver genes need a pre-specified assignment rule

Strongest objection

A clonal hematopoiesis specialist would object that gene is inseparable from clone size and from therapy history in observational data: TET2 clones are larger and infiltrate more, PPM1D clones mark prior chemotherapy, and adjusting for blood VAF and the tumor-over-blood ratio removes most of what makes TET2 different, so the model will report no gene effect because it has adjusted the gene effect away, not because there is none.

Not derivable from what is published

• Death counts per gene-level TI-CH group in MSK-IMPACT
• The share of MSK-IMPACT patients who received immune checkpoint therapy and when
• A pre-specified rule for patients carrying mutations in two driver genes (the record reports a median of 1, range 1-4)

3. Send it to a lab

Send to a lab

This design is a reanalysis of existing data. There is nothing for a bench to run; it needs the data holder.
No bench job exists: the design re-specifies the record's Cox model on existing matched tumor-blood sequencing and survival data with driver gene and checkpoint exposure as factors.

The paper showed that human TET2-mutant myeloid cells make lung tumor organoids grow larger and more numerous, but not through which signal; the authors list candidate cytokines (IL-1 beta, IL-6, PF4, CXCL1, CXCL2, OSM) and note that blocking IL-1 beta prevents lung cancer but does not treat it. The question is which of these signals the organoid effect depends on. Two later papers found TET2-mutant macrophages present antigen better and help checkpoint therapy, so the mechanism may be context-dependent.

1. Has anyone done this since?

Status

partly addressed

Works since

• TET2-mutant clonal hematopoiesis enhances macrophage antigen presentation and improves immune checkpoint therapy in solid tumors (Herbrich et al. 2025, Cancer Cell, doi:10.1016/j.ccell.2025.09.011)
  In a Tet2 bone marrow chimera model, Tet2-mutant tumor macrophages act as immunogenic antigen-presenting cells that cross-prime CD8 T cells in response to interferon gamma; this is a T-cell-dependent mechanism that the record's organoid co-culture, which has no T cells, cannot have been measuring.
• Response to Immune Checkpoint Blockade Is Enhanced in the Presence of Hematopoietic TET2 Inactivation (Rondeau et al. 2025, Cancer Research, doi:10.1158/0008-5472.CAN-24-3329)
  Tet2-mutant monocytes in tumors activated costimulatory programs, and the checkpoint benefit required phagocytes plus CD4 and CD8 T cells; again a mechanism outside the record's co-culture, and it does not test what Tet2-mutant myeloid cells do to tumor cells directly.

Searched

All 49 citing titles, abstracts for Herbrich 2025 and Rondeau 2025; Europe PMC searches for Tet2 with macrophage or myeloid in mouse lung cancer with IL-1, IL-6, cytokine or angiogenesis (1,218 hits, dominated by reviews, top 12 read), and for canakinumab in NSCLC with CANOPY (118 hits, CANOPY-A and CANOPY-1 abstracts read).

Still open

Nothing has tested which secreted or contact signal from TET2-mutant myeloid cells drives the direct growth effect on lung tumor cells shown in the record's organoid co-culture, and no one has blocked IL-1, IL-6 or the S100A8/A9 axis in that system.

Registry query

1 of 49 records in the open node mention any of: tet2, myeloid, lung adenocarcinoma organoid, il-1, il-6, co-culture. robin-damd-assay-selection: Robin selects an in vitro assay for dAMD: RPE phagocytosis (sample trajectory of 2025-05-28)

2. A proposed experiment

Question

Which signal from human TET2-mutant myeloid cells is required for the increase in lung adenocarcinoma organoid number and size in 14-day co-culture?

Hypothesis

Blocking IL-1 receptor signalling abolishes the excess organoid number and size caused by TET2-mutant over wild-type myeloid cells, whereas blocking IL-6 receptor does not.

Attempt type

wet-lab

Design

Repeat the record's co-culture: TRACERx patient-derived lung adenocarcinoma organoid cells cultured 14 days with human TET2-mutant or wild-type myeloid cells isolated from lungs of immunodeficient mice engrafted with TET2-edited or control human hematopoietic stem and progenitor cells (the Huerga Encabo 2023 system). Arms, each with both myeloid genotypes: no blocker; anti-IL-1 receptor 1 antibody (blocks IL-1 alpha and beta); anti-IL-1 beta; anti-IL-6 receptor; anti-S100A8/A9 or anti-EMMPRIN after Nguyen 2021; isotype control; organoids alone. Two format arms separate contact from secretion: direct co-culture and transwell with the myeloid cells above a membrane. Readouts are organoid count and size at day 14 by brightfield imaging, as in Figure 4F, plus a multiplex cytokine panel on day-3 and day-7 supernatants covering IL-1 alpha, IL-1 beta, IL-6, PF4, CXCL1, CXCL2, OSM and S100A8/A9. Five to six biological replicates per arm across three independent experiments, matching the record. Analysis is a two-way model of genotype by blocker on log organoid count, with the pre-registered test being the genotype-by-anti-IL-1R1 interaction.

Controls

• internal-positive The unblocked TET2-mutant versus wild-type arm, which must reproduce the record's larger and more numerous organoids (p<0.05 in the record) in this run.
  Rules out: A null for every blocker that is really a failure of the myeloid cells, the organoid line or the 14-day culture to show the effect at all.
• isotype Isotype-matched antibody at the same concentration as each blocker.
  Rules out: Fc-mediated or non-specific antibody effects on myeloid cells or organoids.
• vehicle Wild-type myeloid cells with each blocker, and organoids alone with each blocker.
  Rules out: A blocker that shrinks organoids by acting on tumor cells directly rather than on the TET2-dependent signal.
  Vocabulary note: Used for the wild-type-plus-blocker arms, which show the blocker's effect on baseline growth.
• known-positive-task Recombinant IL-1 beta added to organoids alone at a dose that increases organoid number, run alongside the anti-IL-1R1 arm.
  Rules out: An anti-IL-1R1 antibody that does not neutralise in this culture medium, which would make a null for IL-1 blockade uninterpretable.

Sensitivity

Cannot be set from the record: Figure 4F reports only p<0.05 with n=5-6 biological replicates and no means, variances or fold change for organoid number or size. The raw values behind Figure 4F, or a two-arm pilot of the unblocked co-culture, are needed to compute how large a reduction of the TET2 effect this replicate number can detect; until then the design should be read as powered to reproduce the record's effect, not to detect a partial block.

Would count as

Supported → positive · opposed → negative-no-effect
Inconclusive if: The unblocked TET2-mutant arm does not out-grow the wild-type arm in this run, or recombinant IL-1 beta does not increase organoid number (so IL-1 blockade cannot be read), or the TET2-mutant myeloid cells arrive at fewer than the replicates needed because the humanized-mouse yield is too low.

Known before

• Myeloid-specific Tet2 deficiency enhanced tumor growth in a mouse lung cancer model through S100a8/S100a9-driven VEGFA production and angiogenesis, and an antibody against EMMPRIN, the S100a8/a9 receptor, suppressed growth. (Nguyen et al. 2021, Cancer Science, doi:10.1111/cas.15165)
  So: A tested mechanism other than IL-1 already exists in mouse lung cancer, so the S100A8/A9 or EMMPRIN arm is required, not optional.
• Hematopoietic aging drives lung cancer progression through IL-1 alpha from myeloid progenitor-like cells, and IL-1 receptor 1 blockade early in tumor development slowed lung, colon and pancreatic tumor growth. (Park et al. 2024, Science, doi:10.1126/science.adn0327)
  So: IL-1 alpha, not only IL-1 beta, may carry the signal, which is why the design blocks the receptor rather than IL-1 beta alone.
• Adjuvant canakinumab after complete resection of stage II-IIIA NSCLC did not improve disease-free survival in 1,382 patients (hazard ratio 0.94, 95% CI 0.78-1.14), and canakinumab added to first-line pembrolizumab and chemotherapy did not improve survival in 643 patients. (Garon et al. 2024 (CANOPY-A) and Tan et al. 2024 (CANOPY-1), Journal of Clinical Oncology, doi:10.1200/JCO.23.00910; doi:10.1200/JCO.23.00980)
  So: Two powered negative trials of IL-1 beta blockade in established NSCLC mean a positive in vitro result for anti-IL-1 beta would still not predict clinical benefit; the question is mechanism, not therapy.
• Tet2-deficient macrophages secrete more IL-1 beta, and Tet2 clonal hematopoiesis accelerates atherosclerosis through the IL-1 beta/NLRP3 inflammasome. (Fuster et al. 2017, Science; Sano et al. 2018, JACC, doi:10.1126/science.aag1381; doi:10.1016/j.jacc.2017.12.037)
  So: The IL-1 beta hypothesis comes from the vascular literature, which is why it is the primary arm, but the lung organoid system has not tested it.
• Myeloid cells from metastatic NSCLC patients with high CHIP burden show enrichment of NF-kappa-B-mediated inflammatory pathways in single-cell RNA sequencing. (Sim et al. 2024, eLife, doi:10.7554/elife.96951)
  So: Human myeloid cells in CHIP carriers do carry an inflammatory programme, which supports profiling the co-culture supernatant for the NF-kappa-B targets listed.

Searched: Europe PMC searches for Tet2 and macrophage or myeloid in mouse lung cancer with cytokine terms (1,218 hits, top 12 read, mostly reviews), for canakinumab in NSCLC (118 hits), and direct abstract retrieval for Nguyen 2021, Park 2024, Garon 2024, Tan 2024, Fuster 2017 and Sim 2024; I looked for and did not find any blocking-antibody experiment in a TET2-mutant myeloid and lung tumor organoid co-culture.

Feasibility

A lab with access to the Huerga Encabo humanized TET2-mutant hematopoiesis mice and to patient-derived lung adenocarcinoma organoids, in practice the Crick Haematopoietic Stem Cell Laboratory and TRACERx organoid collaborators, or a lab that substitutes CRISPR-edited TET2 knockout human CD34+ cells differentiated to macrophages in vitro. · 6 to 9 months, dominated by humanized-mouse engraftment time · moderate, because the humanized mice and organoid lines are expensive to maintain but the assay itself is an antibody-blocking co-culture
Blockers: TRACERx organoid lines are consented for the study and not freely distributable; Yield of human lung myeloid cells per humanized mouse is not stated in the record; Neutralising antibody doses validated in organoid medium are not published for this system

Strongest objection

A tumor immunologist would say the organoid co-culture strips out the step the record actually showed matters, which is monocyte migration into the tumor and conversion to CD11b+ macrophages in vivo, and that a cytokine the myeloid cells secrete in a dish may be irrelevant to growth in a lung; they would also point out that two independent groups found TET2-mutant macrophages are better antigen presenters, so the pro-tumor effect may be a T-cell-absent artefact of the co-culture itself.

Not derivable from what is published

• Raw organoid counts and sizes behind Figure 4F, for the power calculation
• Neutralising antibody concentrations effective in the organoid medium
• Yield and purity of human lung myeloid cells per humanized mouse
• Whether the TET2-edited human cells carry a heterozygous or homozygous loss, which the record does not state

3. Send it to a lab

The authors observed that CHIP clones infiltrate squamous lung tumors more often than adenocarcinomas and guessed that the squamous immune microenvironment (which their own earlier work showed is neutrophil-rich) pulls more CHIP-derived cells in; they did not test it. A rival explanation, that squamous patients simply carry larger blood clones, has support from an independent cohort and has to be excluded first.

1. Has anyone done this since?

Status

nothing found

Works since

None found that bears on this issue.

Searched

All 49 citing titles (none concerns histology-specific TI-CH); Europe PMC search for tumor-infiltrating clonal hematopoiesis (33 hits, none on histology), and for squamous or NSCLC titles with neutrophil, adenocarcinoma and immune or microenvironment (41 hits, top 12 read).

Still open

No study has asked whether the squamous excess of TI-CH survives adjustment for blood clone size and for myeloid or neutrophil density, so the authors' explanation remains an untested guess.

Registry query

0 of 49 records in the open node mention any of: lung squamous carcinoma, lung adenocarcinoma, tumor-infiltrating clonal hematopoiesis, neutrophil, myeloid density, mediation. The open node holds biology and chemistry so far, so this is the expected answer; a private node of a clinical group would be asked the same question.

2. A proposed experiment

Question

Is the higher odds of TI-CH in lung squamous carcinoma explained by its myeloid-rich microenvironment, or by larger blood CHIP clones in squamous patients?

Hypothesis

After adjusting for blood VAF and tumor purity, the squamous-versus-adenocarcinoma odds ratio for TI-CH stays above 1, and is attenuated toward 1 when tumor myeloid or neutrophil density is added to the model.

Attempt type

secondary-analysis

Design

A pre-registered reanalysis in two existing datasets. First, the MSK-IMPACT stage I-III NSCLC validation cohort (917 CHIP patients: 685 lung adenocarcinoma, 132 lung squamous carcinoma), fitting logistic models for TI-CH with histology as the exposure: model A unadjusted (must reproduce odds ratio 1.53, 95% CI 1.03-2.27); model B adding age, sex, smoking, stage, blood VAF and tumor purity; model C adding the CHIP driver gene. Second, the TRACERx imaging mass cytometry subset (163 regions from 70 patients, of whom 21 had CHIP) and its parent set of 198 samples with the four tumor microenvironment archetypes from Enfield 2024, fitting model D with per-region myeloid, monocyte and neutrophil density and the archetype label added to model B, and reporting the proportion of the histology effect mediated by myeloid density. The pre-registered readout is the change in the histology odds ratio from model B to model D. A positive control for the mediation approach is the known association of tumor myeloid density with CHIP (odds ratio 1.20 in the record). The design is extended if Caris's 3,255 matched cases can be added, which would roughly quadruple the squamous CHIP count.

Controls

• reference-case Model A must return the record's unadjusted histology odds ratio of 1.53 (95% CI 1.03-2.27) in the MSK-IMPACT NSCLC cohort.
  Rules out: A reanalysis that has lost or re-defined patients, so that attenuation in later models is a data artefact.
• permutation-null Histology labels shuffled among CHIP patients 1,000 times, re-fitting models B and D each time.
  Rules out: An apparent attenuation of the histology odds ratio that is the expected behaviour of adding correlated covariates to a model of this size.
• internal-positive Within the TRACERx imaging subset, the record's association between CHIP and myeloid density (age-adjusted odds ratio 1.20, 95% CI 1.01-1.43) must be recovered.
  Rules out: An imaging dataset in which myeloid density is too noisy to mediate anything, which would make a null mediation result uninformative.

Sensitivity

The record's own estimate is the only input: 132 squamous versus 685 adenocarcinoma CHIP patients gave an odds ratio of 1.53 with a lower bound of 1.03, so the current sample barely detects the effect at all and cannot detect a partial attenuation; the number of squamous CHIP patients needed to detect attenuation of the odds ratio from 1.53 to, say, 1.2 is not derivable here and requires a power calculation using the covariance of histology with blood VAF and myeloid density, which the record does not give. The imaging subset (21 CHIP patients) is too small for mediation and should be treated as hypothesis-generating unless Caris or additional TRACERx imaging is added.

Would count as

Supported → positive · opposed → negative-no-association
Inconclusive if: Model A does not reproduce an odds ratio near 1.53, or the squamous CHIP count after exclusions falls below what the power calculation requires, or blood VAF and histology are so correlated that model B's interval spans both 1 and 1.53.

Known before

• In 100 metastatic NSCLC patients, squamous carcinoma patients had a higher burden of large blood CHIP clones than adenocarcinoma patients (8/43 vs 2/50, p=0.04). (Sim et al. 2024, eLife, doi:10.7554/elife.96951)
  So: This is the rival explanation: if squamous patients carry larger blood clones, the TI-CH excess may be a clone-size effect, so blood VAF must be adjusted before the microenvironment is invoked.
• Immune cell composition is fundamentally different between lung adenocarcinoma and lung squamous carcinoma, and neutrophils are the most prevalent immune cell type in NSCLC. (Kargl et al. 2017, Nature Communications, doi:10.1038/ncomms14381)
  So: Establishes that histology predicts myeloid composition, which is the premise of the authors' explanation and the reason neutrophil density is a candidate mediator.
• In TRACERx (81 patients, 198 samples), a tumor-associated-neutrophil-high archetype with sparse lymphocytes was frequent in squamous carcinoma, carried PIK3CA mutations, and had shorter disease-free survival independent of stage. (Enfield et al. 2024, Cancer Discovery, doi:10.1158/2159-8290.CD-23-1380)
  So: The archetype labels already exist for the same TRACERx samples used in the record, so model D can be run without new imaging.
• In the record itself, TI-CH was independent of tumor purity, tumor-infiltrating neutrophils and the proportion of myeloid cells as predictors of survival (Fig S14B-D). (Pich et al. 2025, NEJM (the record), doi:10.1056/NEJMoa2413361)
  So: The authors adjusted the survival model for myeloid content but not the histology-prevalence model, which is the gap this design fills.

Searched: Europe PMC search for squamous or NSCLC titles with neutrophil, adenocarcinoma and immune or microenvironment (41 hits, top 12 read), direct abstract retrieval for Kargl 2017, Enfield 2024 and Sim 2024; I looked for and did not find any study of CHIP clone infiltration by lung histology other than the record.

Feasibility

The TRACERx consortium (which holds both the CHIP genotyping and the imaging mass cytometry archetypes) with the MSK-IMPACT computational oncology group; Caris could contribute matched cases. · 3 to 6 months · low, because all data exist and the work is model re-specification plus a mediation analysis
Blockers: Only 21 CHIP patients in the imaging subset; 132 squamous CHIP patients in MSK-IMPACT is barely enough for the unadjusted effect; Per-region myeloid density must be matched to per-region CHIP VAF across the two TRACERx datasets

Strongest objection

A lung pathologist would say that squamous and adenocarcinoma differ in purity, necrosis and stromal content in ways that change the denominator of a tumor VAF, so a 2% tumor VAF threshold captures a different fraction of infiltrating cells in the two histologies and the odds ratio of 1.53 may be a measurement artefact of the threshold, not of infiltration; the record's alternative metric (fraction of non-tumor cells carrying the mutation, Fig S14E) should be the exposure instead.

Not derivable from what is published

• Covariance of histology with blood VAF and myeloid density, needed for the attenuation power calculation
• Whether per-region imaging mass cytometry data can be matched to per-region CHIP VAF for the 21 imaged CHIP patients
• Squamous versus adenocarcinoma counts in the Caris 3,255-case analysis

3. Send it to a lab

Send to a lab

This design is a reanalysis of existing data. There is nothing for a bench to run; it needs the data holder.
No bench job exists: the design re-fits logistic models on existing MSK-IMPACT and TRACERx data, with imaging mass cytometry densities that have already been measured.