Oncofusion

start / graph / resistance-adaptive-nongenetic-menin

Resistancemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q6b-therapeutics2

Adaptive, non-genetic resistance — the leukemia keeps the fusion and rewires how it is maintained

also called epigenetic escape · non-mutational menin resistance · transcriptional reprogramming resistance · MEN1-wild-type resistance

Some leukemias stop responding to a drug without changing their DNA at all. They simply find another way to keep the same cancer genes switched on, which means sequencing the tumour will show nothing wrong and the drug will still fail.

The same thing, for a scientist

Acquired resistance to menin inhibition occurs in KMT2A-rearranged ALL both through MEN1 mutation and independently of it. The most directly informative model is the DOT1L precedent: KMT2A::AFF1 B-ALL cells driven to pinometostat resistance became largely independent of H3K79 methylation while still requiring the physical presence of DOT1L, HOXA9 and the fusion, selectively abandoning some fusion target genes and drifting myeloid. Analogous non-genetic escape from menin inhibition has been documented in patient-derived xenografts and in primary samples. This class of resistance is invisible to mutation panels, is likely to dominate in infant leukemia because of its exceptionally low mutation burden, and is the mechanism most likely to be shared between every chromatin-directed drug in this disease.

Go deeper
Three independent demonstrations of the same phenomenon1 deeper

1. Under DOT1L pressure. A paediatric KMT2A::AFF1 B-ALL line driven to acquired pinometostat resistance became mostly independent of DOT1L-mediated H3K79 methylation but still relied on the physical presence of DOT1L, HOXA9 and the KMT2A::AFF1 fusion. It selectively lost epigenetic regulation and expression of some fusion target genes (PROM1/CD133) while leaving others (HOXA9, CDK6) unaffected, and it upregulated myeloid genes including CD33 and LILRB4 1. The circuit survived; the dependency on the drugged step did not.

2. Under menin pressure. In KMT2A-rearranged ALL, acquired revumenib resistance arises either through MEN1 mutations or independently of MEN1 mutation 2. That is the same phenomenon in the drug class that matters.

3. Before any pressure at all. In patient-derived xenografts of highly pretreated KMT2A::AFF1 ALL, samples from later in the same patient's disease course were much less sensitive to menin inhibition than samples from earlier in that patient's course, despite none of the patients ever having received a menin inhibitor. Transcriptomic analysis confirmed sustained on-target efficacy against the canonical menin-inhibitor targets in the resistant cells — the drug was still doing its job, and the cells no longer cared 3.

That third result is the one that should worry the lab most. It means resistance to menin inhibition can be pre-existing, generated by conventional chemotherapy exposure rather than by the targeted drug, and it means a relapsed, multiply-treated infant may be the patient least likely to respond — which is precisely the population in which the drugs were approved.

Named molecular routes

The escape is not formless. At least three concrete mechanisms have been characterised:

  • PRC1.1 loss. Deletion or loss of PCGF1 or BCOR causes menin-inhibitor resistance by epigenetically reactivating non-canonical menin targets, including MYC. These cells show reduced monocytic signatures and become BCL-2-dependent, and venetoclax resensitises them 4.
  • KMT2C/KMT2D inactivation. Downregulation of KMT2D may account for upfront resistance in at least one patient; inactivation of KMT2C/D has been reported to cause menin-inhibitor resistance, and KMT2C-edited cells from that patient were selected for under menin-inhibitor pressure 3.
  • KAT6/KAT7-dependent chromatin maintenance. KAT7 interacts with menin and the MLL complex and co-regulates the MLL-fusion program; catalytic inhibition of KAT6/KAT7 with PF-9363 evicts the fusion from chromatin by an orthogonal route and overcomes both primary and acquired, genetic and non-genetic, resistance to menin inhibition 5.

The third of these is the most actionable: an orthogonal chromatin lever that works specifically where menin inhibition fails.

Why infants are the worst case

KMT2A-rearranged infant leukemia has one of the lowest somatic mutation burdens of any cancer. The graph covers this separately. The consequence for resistance is direct and under-appreciated:

  • In a high-mutation-burden adult AML, genetic escape routes (MEN1 mutation, RAS pathway, FLT3) are abundant and are what you see.
  • In an infant leukemia with almost no mutations, the leukemia has fewer genetic options and correspondingly more reliance on plasticity. The cell of origin is a fetal progenitor with intrinsically high lineage and state plasticity, which is the substrate non-genetic resistance runs on.

The testable prediction: menin-inhibitor resistance in infants should be disproportionately MEN1-wild-type and disproportionately state-based, compared with adults. Nobody has the sample series to check this yet, and the trials that will generate it (NCT05761171, NCT05848687, NCT06376162) should be banking serial samples with transcriptomic, not just genomic, readouts.

Facts and figures
kindresistance
mechanism classnon-genetic / adaptive / epigenetic
seen in patientspartially — documented in patient-derived xenografts from treated patients and inferred from MEN1-wild-type clinical failures; not yet characterised prospectively in patients on menin inhibitors
seen in modelstrue
detectabilityMOSTLY invisible to DNA mutation panels, but NOT uniformly - and the difference is actionable. Corrected 2026-09-19: this field previously read 'invisible to DNA mutation panels' while named_routes includes KMT2C/KMT2D inactivation, which is loss-of-function in two genes carried on comprehensive sequencing panels. A clinician reading the old wording could stop looking for something their existing assay already reports. Split by route: KMT2C/KMT2D inactivation is DETECTABLE on a comprehensive DNA panel; PRC1.1/BCOR loss is detectable where BCOR is on the panel; H3K79me-independent maintenance, pre-existing resistance in pretreated disease and myeloid transcriptional drift are genuinely invisible to DNA and need transcriptomic or chromatin profiling of serial samples.
severityhigh, and probably higher in infants than adults because of the very low somatic mutation burden
detectability previous wordinginvisible to DNA mutation panels; requires transcriptomic or chromatin profiling of serial samples
cross reference 2026 09 19The SWI-SNF/BET-dependent state in MEN1-wild-type resistant cells is a further route and is written up separately as resistance-swi-snf-bet-dependent-state rather than duplicated here.
affects
[
  "drug-revumenib",
  "drug-ziftomenib",
  "drug-bleximenib",
  "drug-enzomenib",
  "drug-icovamenib",
  "drug-pinometostat"
]
named routes
[
  "H3K79me-independent maintenance of the fusion program with retained DOT1L/HOXA9/fusion requirement",
  "PRC1.1 (PCGF1/BCOR) loss reactivating non-canonical menin targets including MYC",
  "KMT2C/KMT2D inactivation or downregulation",
  "pre-existing resistance in multiply pretreated disease with sustained on-target drug effect",
  "myeloid transcriptional drift"
]
pre emption strategies
[
  "venetoclax, which specifically resensitises the PRC1.1-deficient/MYC-reactivated state",
  "KAT6/KAT7 catalytic inhibition as an orthogonal route to evict the fusion from chromatin (preclinical, preprint)",
  "CDK9 inhibition to remove the MYC and MCL1 output regardless of how it is being maintained",
  "use in frontline rather than multiply-relapsed disease, where pre-existing resistance is already present",
  "T-cell engagers and cellular therapy, which are indifferent to transcriptional state"
]
How this could bite us back
  • Invisibility: this resistance class produces no mutation, so a relapsed child will be sequenced, told nothing was found, and treated as if the drug should still work.
  • Pre-existing rather than acquired: multiply pretreated disease may already be resistant on day one, which inverts the usual assumption that targeted agents should be saved for later lines.
  • Lineage drift: the adaptive state trends myeloid, which is the same direction as immunotherapy-driven lineage switch, so chromatin therapy may prime immunotherapy failure.
  • Generalisation: because the escape is to keep the circuit and change its maintenance, it threatens every obligate-partner interface strategy equally, not just menin.
  • Combination does not automatically help: if the adaptive state is entered before treatment, adding a second chromatin agent to a resistant cell buys nothing, and only mechanistically orthogonal agents (BCL-2, immune, CDK9) will.
Still open
  1. Is menin-inhibitor resistance in infants disproportionately MEN1-wild-type? The low mutation burden of infant KMT2A-rearranged leukemia predicts yes, and the paediatric menin trials now open could answer it if they bank serial samples with transcriptomic readouts.
  2. Does conventional chemotherapy itself select for menin-inhibitor resistance? The observation that late samples from menin-naive, multiply-treated patients are resistant implies yes, and would argue strongly for moving menin inhibitors frontline rather than reserving them.
  3. Can a transcriptional signature predict adaptive resistance before it is clinically apparent? A signature scored on diagnostic samples would be the single most useful biomarker in this programme, and the analysis node for it already exists in this graph.
  4. Do menin inhibitors and DOT1L inhibitors select for the SAME adaptive state? If the H3K79me-independent state and the menin-independent state are one state, the field has one problem rather than two, and one orthogonal agent could cover both.
Every source behind this entry (5)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Paediatric KMT2A::AFF1 B-ALL cells with acquired pinometostat resistance became largely independent of DOT1L-mediated H3K79 methylation while still requiring the physical presence of DOT1L, HOXA9 and the KMT2A::AFF1 fusion, selectively losing regulation and expression of some fusion target genes while upregulating myeloid genes including CD33 and LILRB4.

    Modelling acquired resistance to DOT1L inhibition exhibits the adaptive potential of KMT2A-rearranged acute lymphoblastic leukemia · Schneider P, Crump NT, Arentsen-Peters STCJM, et al. · Experimental Hematology & Oncology 2023
    PMID 37740239DOI
    What this evidence cannot show

    A single cell-line model of acquired resistance generated by prolonged in vitro drug exposure; cell-line adaptation is not proof of the route taken in patients, and no patient with acquired pinometostat resistance exists to check against because the drug was never used long enough clinically.

    Evidence that argues against this (1)
    • This adaptive-resistance model was built under a selection pressure that was never clinically effective, which limits how far it can be generalised to menin inhibitors. In the adult phase 1 of pinometostat (n=51, continuous intravenous infusion, six dose cohorts), the drug reduced H3K79 methylation and reached apparent steady-state exposure without a maximum tolerated dose, yet produced complete remission in only 2 patients, both with t(11;19); the authors describe single-agent efficacy as modest. Cells selected in vitro against a drug that achieves target engagement but almost no clinical response may be modelling adaptation to a weak pressure rather than the adaptation that matters under an effective one. Stein EM et al. 2018
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Acquired resistance to revumenib in KMT2A-rearranged ALL cells occurs either through acquisition of MEN1 mutations or independently of mutations in MEN1.

    Distinct Responses to Menin Inhibition and Synergy with DOT1L Inhibition in KMT2A-Rearranged Acute Lymphoblastic and Myeloid Leukemia · Adriaanse FRS, Schneider P, Arentsen-Peters STCJM, et al. · International Journal of Molecular Sciences 2024
    PMID 38892207DOI
    What this evidence cannot show

    In vitro acquired-resistance models in KMT2A-rearranged ALL cells, not patient samples. The paper does not establish the relative frequency of the two routes.

    Evidence that argues against this (1)
    • The in vitro dichotomy of MEN1-mutant versus MEN1-independent resistance has no prospective clinical correlate so far. In the registration-enabling phase 2 of ziftomenib (KOMET-001, 92 patients with R/R NPM1-mutant AML), PRESPECIFIED subgroup analyses found CR/CRh comparable regardless of TYPE OF COMUTATION and regardless of previous therapy including venetoclax. Genomic stratification did not separate responders from non-responders in the largest prospective menin-inhibitor dataset, so a resistance taxonomy built on the presence or absence of a MEN1 mutation in cell lines should not yet be used to predict or explain clinical failure. Wang ES et al. 2025
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In patient-derived xenografts, highly pretreated KMT2A::AFF1 ALL samples were much less sensitive to menin inhibition than samples taken earlier in the same patients' disease course despite no prior menin-inhibitor exposure; transcriptomic analysis documented sustained on-target efficacy against canonical menin-inhibitor targets in the resistant cells, with emergence of RAS pathway and TP53 co-mutations that were individually insufficient to induce resistance in vitro, and possible contribution from KMT2D downregulation and KMT2C inactivation.

    Upfront menin-inhibitor resistance in multiply pretreated leukemias · See PubMed record PMID 41027490 for the full author list (not transcribed here) · Experimental Hematology 2025
    PMID 41027490DOI
    What this evidence cannot show

    Serial samples from only two patients with multiply relapsed ALL, plus three high-risk AML and two infant ALL samples that were sensitive. Very small n. The authors themselves state that which genomic or epigenomic alterations drive upfront resistance needs broader clarification. Xenograft drug sensitivity is an imperfect proxy for clinical response.

    Evidence that argues against this (1)
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Loss of PRC1.1 subunits PCGF1 or BCOR causes menin-inhibitor resistance through epigenetic reactivation of non-canonical menin targets including MYC; genetic or pharmacological MYC inhibition resensitises these cells, and PRC1.1-deficient cells show reduced monocytic signatures, are susceptible to BCL-2 inhibition, and are resensitised to menin inhibition by venetoclax.

    Epigenetic regulation of noncanonical menin targets modulates menin inhibitor response in acute myeloid leukemia · See PubMed record PMID 39158067 for the full author list (not transcribed here) · Blood 2024
    PMID 39158067DOI
    What this evidence cannot show

    Chromatin-focused CRISPR screens in human and mouse KMT2A-rearranged AML models. This is a defined genetic lesion producing a non-genetic (epigenetic reactivation) resistance phenotype, so it sits at the boundary of this node's category. No patient data show that PRC1.1-deficient relapses respond to venetoclax.

    Evidence that argues against this (1)
    • 'Pharmacological MYC inhibition resensitises these cells' names a clinical option that does not currently exist at an effective dose. The nearest available approach, CDK9 inhibition, has been taken into patients with matched pharmacodynamics: voruciclib downregulated MYC and NF-kB transcriptional gene sets, decreased MCL1 messenger RNA and reduced RNA polymerase II phosphorylation, at an intermittent 200 mg schedule with no dose-limiting toxicities — and among 21 patients with AML produced one morphologic leukaemia-free state, two stable diseases and no complete remissions. Achieving the transcriptional effect this entry relies on is therefore feasible and, on its own, inactive. Davids MS et al. 2025
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KAT7 interacts with menin and the MLL complex and co-regulates the MLL-fusion transcriptional program; combined catalytic inhibition of KAT6A/B and KAT7 with PF-9363 rapidly evicts the MLL fusion protein from chromatin, represses oncogenic transcription, overcomes primary resistance to menin inhibitors, and overcomes acquired genetic and non-genetic resistance to menin inhibition.

    PMID 40536420DOI
    What this evidence cannot show

    Preclinical only. Claims of overcoming 'primary and acquired, genetic and non-genetic' resistance are broad and should be treated as provisional until published. PF-9363 has no paediatric data. KAT7 is required for normal haematopoiesis, so the therapeutic index is a real question. REGRADED weak->strong 2026-09-20: this evidence item restates what PMID 40536420 (Gordon et al., Cancer Discov 2025, 'Catalytic Inhibition of KAT6/KAT7 Enhances the Efficacy and Overcomes Primary and Acquired Resistance to Menin Inhibitors in MLL Leukemia') reports in its own models, and two other entries carry the same claim at strong. Per concept-the-confidence-grading-rule the grade attaches to the CLAIM, so identical claims must carry identical grades; a weaker grade here was expressing how load-bearing the paper is for THIS entry's argument, which belongs in the caveat rather than in the confidence field. The hedging this entry needs is that the work is preclinical and from one group.

    Evidence that argues against this (1)
    • Catalytic KAT7 inhibition removes a factor that normal haematopoietic stem cells require, so the therapeutic index — not the resistance rescue — is the open question. Conditional deletion of Hbo1 (KAT7) causes rapid loss of haematopoietic progenitors with reduced erythroid, myeloid, B- and T-lineage-restricted progenitors, drives quiescent HSCs abnormally into cycle so that they produce progenitors at the expense of self-renewal, and exhausts the HSC pool; HBO1 is required for genome-wide H3K14 acetylation and specifically at loci controlling HSC quiescence and self-renewal, including Mpl, Tek/Tie-2, Gfi1b, Egr1, Tal1, Gata2, Erg, PBX1, MEIS1 and Hox9. Note that the last three are the same transcription-factor network the KMT2A fusion program exploits, so 'evicting the fusion from chromatin' by KAT6/KAT7 inhibition and 'collapsing normal HSC self-renewal' are not separable effects in the way a cancer-cell viability assay implies. Yang Y et al. 2022
Connected to