Mechanism class 1 — chromatin and elongation hijack (KMT2A-r, NUP98-r)
also called elongation hijack · SEC hijack · chromatin-adaptor class
Some fusion proteins keep the part that finds genes but swap the part that decides whether to switch them on for a machine that switches them on permanently. The cell is then stuck reading a set of stem-cell genes it should have finished with, so it never grows up.
The same thing, for a scientist
In this class the 5-prime partner contributes a chromatin-targeting module (KMT2A AT-hooks/CXXC plus the menin and LEDGF docking motifs; NUP98 FG repeats plus partner reader domains) and the 3-prime partner contributes a transcription-elongation engine (the AFF1/AFF4-nucleated super elongation complex with ELL and P-TEFb, or DOT1L recruited through the AF9/ENL YEATS proteins). The result is not loss of KMT2A function but a neomorphic, constitutively-on elongation platform parked on the HOXA cluster and MEIS1, producing a self-renewing progenitor that cannot differentiate. Because the targeting module, not the effector, is the invariant part, the class is drugged at the adaptor interface (menin-KMT2A) rather than at the fusion itself.
The architecture: a targeting half plus a stolen engine1 deeper
Purification of several different KMT2A chimeras showed that they all pull down the same complex — the super elongation complex (SEC), containing ELL, P-TEFb (CDK9/cyclin T) and the AFF1/AFF4 scaffold — and that knocking down AFF4 reduces expression of KMT2A-chimera target genes 1. That is why partners as dissimilar as AFF1, MLLT1, MLLT3 and ELL converge on one disease program: they are alternative entry points to the same elongation machine.
The second, partly independent effector arm is DOT1L. MLL-AF9-transformed cells require DOT1L catalytic activity and the aberrant H3K79 methylation it deposits; genetic loss of Dot1l collapses the Hoxa9/Meis1 program and releases differentiation 2. So the class has two exploitable effector nodes (SEC/P-TEFb and DOT1L) and one exploitable adaptor node (menin).
Why the adaptor, not the effector, became the drug target
Menin binds the extreme N-terminus of KMT2A, which every fusion retains. VTP50469, a potent and selective menin-KMT2A inhibitor, displaced menin from chromatin, collapsed the fusion gene program and produced long disease-free survival in mice carrying KMT2A-rearranged patient-derived xenografts 3. The clinical version, revumenib, produced a 30% CR/CRh rate in heavily pre-treated relapsed or refractory KMT2A-rearranged and NPM1-mutant acute leukemia, with differentiation of blasts into normal-looking myeloid cells 4. That is the phase 1 figure and it is the optimistic one: the registration-enabling phase 2 of the same trial, in 57 patients with centrally confirmed KMT2A rearrangement, gave CR+CRh of 22.8%, and in the NPM1-mutant cohort the median duration of CR+CRh was 4.7 months (see the counter-evidence on 4).
By contrast the DOT1L inhibitor pinometostat and CDK9/BET inhibitors have not produced a comparable single-agent clinical signal. The effector arms are more redundant, and more broadly essential to normal cells, than the adaptor.
NUP98 fusions belong to the same class but arrive by a different route
NUP98 fusions retain the nucleoporin FG-repeat intrinsically disordered region and add a chromatin reader or writer (the NSD1 methyltransferase, the KDM5A PHD finger). The FG repeats drive liquid-liquid phase separation, and that phase behaviour is required both for chromatin occupancy and for leukemic transformation; replacing the FG repeats with an unrelated condensate-forming IDR from FUS reconstitutes the activity 5. The downstream output is the same HOXA/MEIS1 lock.
So NUP98-r sits at the intersection of this class and the condensate class (mechanism-condensate-transcriptional-rewiring). The two are mechanistically nested, not alternatives.
Therapeutic implication — stated plainly
This class has yielded a drug. Menin inhibition is the first targeted therapy for a chromatin-adaptor fusion class, and revumenib reached FDA approval for relapsed or refractory KMT2A-rearranged acute leukemia. The reason it worked is close to a structural accident: the fusion retains a short, high-affinity peptide-in-groove interaction with a partner protein that happens to present a real, enclosed, small-molecule-sized pocket. That is a partner pocket, not a fusion pocket.
What it has not yielded is cure. Responses are partial and finite, and escape mutations at the drug-menin interface — not at the menin-MLL interface — arose in patients within months 6. Those interface mutations are not the whole of escape: menin-inhibitor-resistant AML also arises with no MEN1 mutation at all, through an altered epigenome and proteome that installs BRD4/SMARCA4/CREBBP dependence, and heavily pretreated KMT2A::AFF1 samples are already resistant before any menin inhibitor has been given, with on-target repression of the canonical targets fully preserved (counter-evidence on 3 and 6). The class lesson is that adaptor-interface drugging buys differentiation and remission, not eradication, unless it is combined with something that removes the cell or the protein - and that the genetic escape route is the one we can see, not necessarily the one that matters most.
| mechanism class | chromatin-elongation-hijack |
|---|---|
| invariant module | 5-prime chromatin-targeting module (KMT2A AT-hooks + CXXC + menin/LEDGF docking motifs; NUP98 FG repeats) |
| variable module | 3-prime elongation or chromatin-reader effector (SEC via AFF1/AFF4, DOT1L via YEATS partners, NSD1 SET, KDM5A PHD) |
| druggable node | partner adaptor interface (menin-KMT2A), not the fusion surface |
| druggability class | clinically-validated |
| why tractable | A short retained peptide motif binds a partner protein that presents a genuine enclosed small-molecule pocket, and the motif is invariant across more than 100 partners so one drug covers the whole class. |
| why still not curative | Adaptor blockade forces differentiation but does not destroy the fusion protein; interface escape mutations and residual disease follow. |
member fusions
[ "fusion-kmt2a-aff1", "fusion-kmt2a-mllt3", "fusion-kmt2a-mllt1", "fusion-kmt2a-mllt10", "fusion-kmt2a-ell", "fusion-kmt2a-afdn", "fusion-nup98-nsd1", "fusion-nup98-kdm5a" ]
drugs yielded
[ "drug-revumenib", "drug-ziftomenib", "drug-bleximenib", "drug-enzomenib" ]
drugs failed or stalled
[ "pinometostat (DOT1L)", "CDK9 inhibitors", "BET inhibitors" ]
- The drug target is a wild-type partner protein (menin) with tumour-suppressor function in endocrine tissue and a role in normal haematopoietic gene regulation, so long-term class-wide inhibition risks on-target endocrine and marrow toxicity.
- Resistance to this class arises at the drug-target interface as well as away from it. MEN1 interface mutants are fully functional leukemia that has simply stopped seeing the drug; but menin-inhibitor-resistant cells also arise with no MEN1 mutation, carrying an altered epigenome and proteome and a new dependence on BRD4, SMARCA4 and CREBBP. A next-generation inhibitor designed around M327I addresses only the visible half. (Corrected 2026-09-19; this item previously read 'Resistance to this class arises at the drug-target interface rather than at the biological interface'.)
- Forcing differentiation in a high-burden blast population causes differentiation syndrome, a potentially fatal on-mechanism toxicity shared with the APL paradigm.
- Prior therapy installs menin-inhibitor resistance before the drug is ever given: multiply relapsed KMT2A::AFF1 samples are markedly less sensitive than earlier samples from the same patient, with on-target transcriptional repression intact. Sequencing this class late in therapy may be testing it in the population least able to respond.
- Response and durability come apart. In the registrational cohorts the CR+CRh rate is roughly a quarter of patients and the median duration of CR+CRh is under five months, so on current evidence the class is a bridge to transplant rather than disease control - and only about one in six responders actually reached transplant.
- Is there a second invariant adaptor in this class (LEDGF/PSIP1, PAF1c, or the AFF1 ALF domain) whose interface is as pocket-like as menin's but whose loss is better tolerated by normal HSC?
- Do the two effector arms (SEC and DOT1L) substitute for each other after menin inhibition, and would blocking adaptor plus one effector simultaneously prevent interface escape mutants from ever being selected?
- Does partner identity change which effector arm dominates enough that a partner-stratified combination would beat a single class-wide drug?
- Are the genetic (MEN1 interface) and non-genetic (BRD4/SMARCA4/CREBBP-dependent) escape routes from menin inhibition mutually exclusive in the same patient, and does pre-treatment chromatin state predict which one a given leukemia will take?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Multiple different KMT2A chimeras all associate with the AFF4-containing super elongation complex (ELL, P-TEFb), and AFF4 knockdown reduces KMT2A-chimera target gene expression.
AFF4, a component of the ELL/P-TEFb elongation complex and a shared subunit of MLL chimeras, can link transcription elongation to leukemia · Lin C, Smith ER, Takahashi H, et al. · Molecular Cell 2010What this evidence cannot show
Biochemical purification from cell lines plus knockdown; it does not establish that SEC association is sufficient for transformation in vivo, and the stoichiometry in primary patient blasts is unknown.
Evidence that argues against this (1)
- The generalisation was tested directly by substitution and it failed. If SEC recruitment via the partner were the mechanism, an artificial dimerisation module could not replace a partner - yet synthetic dimerisation of the MLL N-terminus reproduces transformation, and the coiled coils of the cytoplasmic partners EPS15/AF1p and GAS7 are necessary AND sufficient. Forced self-association is the better class-wide description, and AFF4 knockdown reducing target-gene expression is consistent with SEC being downstream in the partner class that has it, not with SEC being the common cause. Cancer Cell 2003
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL-AF9-driven leukemia requires DOT1L and the aberrant H3K79 methylation it deposits; Dot1l loss collapses the Hoxa9/Meis1 program and permits differentiation.
MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L · Bernt KM, Zhu N, Sinha AU, et al. · Cancer Cell 2011What this evidence cannot show
Mouse models and conditional knockout, largely MLL-AF9; dependency strength differs between partners and between lymphoid and myeloid contexts, and the clinical DOT1L inhibitor pinometostat did not reproduce this magnitude of effect in patients.
Evidence that argues against this (1)
- The clinical DOT1L inhibitor pinometostat produced only limited single-agent activity, which argues that H3K79me-dependence measured genetically overstates the pharmacological window. Issa GC et al. 2023
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A selective menin-KMT2A inhibitor (VTP50469) displaces menin from chromatin, reverses the fusion gene program and produces long disease-free survival in KMT2A-rearranged patient-derived xenografts.
A Menin-MLL inhibitor induces specific chromatin changes and eradicates disease in models of MLL-rearranged leukemia · Krivtsov AV, Evans K, Gadrey JY, et al. · Cancer Cell 2019What this evidence cannot show
Disease-free outcomes were model- and partner-dependent, and immunodeficient mice do not model normal haematopoietic toxicity or immune clearance.
Evidence that argues against this (1)
- The PDX result is conditional on where in the disease course the sample was taken, and the entry states it without that condition. Serial samples from the same patients show that heavily pretreated KMT2A::AFF1 ALL is much less sensitive to menin inhibition than earlier samples from the same individuals -- and that this resistance was acquired with NO prior menin-inhibitor exposure. Transcriptomic analysis showed sustained on-target repression of the canonical menin-inhibitor target genes in the resistant cells, so the drug was still doing to chromatin exactly what VTP50469 was reported to do, and the cells no longer died. 'Displaces menin, reverses the fusion program' and 'produces long disease-free survival' are therefore separable, and the PDX panel that gave the long DFS was not drawn from the multiply-relapsed population the drug is licensed for. Mahdavi L et al. 2025
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the first-in-human trial, the menin inhibitor revumenib produced a 30% CR/CRh rate in relapsed or refractory KMT2A-rearranged or NPM1-mutant acute leukemia, with evidence of blast differentiation.
The menin inhibitor revumenib in KMT2A-rearranged or NPM1-mutant leukaemia · Issa GC, Aldoss I, DiPersio J, et al. · Nature 2023What this evidence cannot show
Single-arm phase 1 with a heterogeneous, heavily pre-treated population; the efficacy-analysis denominator is a subset of enrolled patients, many responders proceeded to transplant, and asymptomatic QTc prolongation was dose-limiting.
Evidence that argues against this (2)
- The 30% figure is the phase 1 dose-escalation number and the registration-enabling phase 2 did not reproduce it. In AUGMENT-101 phase 2 (n=94 treated, 57 efficacy-evaluable with centrally confirmed KMT2A rearrangement) the CR+CRh rate was 22.8% (95% CI 12.7-35.8), i.e. roughly a quarter lower than the figure this entry quotes, in the larger and centrally adjudicated cohort. The same report gives the cost side the entry omits: grade >=3 febrile neutropenia in 37.2%, differentiation syndrome in 16.0% and QTc prolongation in 13.8%. Quoting the phase 1 number for a class-level claim about what adaptor-interface drugging achieves overstates it. Issa GC et al. 2024
- Durability is the missing axis. In the NPM1-mutant phase 2 cohort of the same trial the CR+CRh rate was comparable (23.4%) but the median duration of CR+CRh was 4.7 months, and only 5 of 30 responders (16.7%) reached transplant. A response rate quoted without a duration reads as though the class has produced disease control; under five months of remission is a bridge at best. This also bears on the entry's framing that menin inhibition is a KMT2A-fusion-class drug: the comparable response rate in NPM1-mutant AML, which has no fusion at all, shows the drug targets a menin-dependent HOXA/MEIS1 transcriptional state rather than the fusion class as such. Arellano ML et al. 2025
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The NUP98 FG-repeat intrinsically disordered region drives liquid-liquid phase separation that is required for chromatin occupancy, super-enhancer-like binding and leukemic transformation by NUP98-HOXA9.
Phase separation drives aberrant chromatin looping and cancer development · Ahn JH, Davis ES, Daugird TA, et al. · Nature 2021What this evidence cannot show
The work centres on NUP98-HOXA9 with engineered IDR swaps; whether NSD1 and KDM5A fusions behave identically, and whether puncta exist at physiological expression in primary patient blasts, is less firmly established.
Evidence that argues against this (2)
- Condensate formation and transformation were dissociated in the very fusion this entry generalises to. In NUP98-NSD1/FLT3-ITD immortalised primary murine HSPC, SMARCA5 (ISWI/NURF) is an essential mediator of transformation, yet SMARCA5 inhibition did NOT affect NUP98-NSD1 condensates -- the puncta persist while the transformed phenotype is lost. The authors state plainly that formation of NUP98-NSD1 phase-separated nuclear condensates is not sufficient for maintenance of the transformed phenotype. So 'the FG IDR drives LLPS that is required for transformation' can be true while the condensate is not the thing doing the transforming; what matters is which constituents the condensate holds. Jevtic Z et al. 2022
- The chromatin-occupancy half of the claim does not survive being tested against a NUP98 fusion with a reader domain. In NUP98-PHF23, mutagenesis plus genomics assign the two functions to different halves: the PHD finger targets the condensate to H3K4me3/2-marked developmental genes, while the FG repeats determine condensate composition and gene activation. Occupancy is partner-driven; the FG repeats select which coactivators (KMT2/MLL methyltransferases, histone acetyltransferases, BRD4) partition in. Since NUP98-NSD1 and NUP98-KDM5A likewise carry chromatin reader/writer modules, a class claim that the FG IDR is what determines chromatin occupancy is the wrong attribution for exactly the members this entry names. Ahn JH et al. 2025
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Patients treated with revumenib acquired MEN1 mutations at the drug-menin interface that block inhibitor binding while preserving the menin-MLL1 interaction, restoring chromatin occupancy.
MEN1 mutations mediate clinical resistance to menin inhibition · Perner F, Stein EM, Wenge DV, et al. · Nature 2023What this evidence cannot show
A small number of patients with paired samples; the frequency of interface mutations as the dominant resistance route, versus transcriptional or lineage-switch escape, is not yet defined.
Evidence that argues against this (1)
- MEN1 interface mutation is one escape route, not the route, and treating it as the mechanism of clinical resistance makes the problem look more tractable than it is. Menin-inhibitor-resistant AML cells generated by repeated drug shock carry an altered epigenome, transcriptome and proteome with NO menin mutation, and the resistant state is a druggable dependency on BRD4, SMARCA4 and CREBBP rather than on a restored menin-MLL1 contact. Resistant patient-derived MLL1-r and NPM1-mutant cells behaved the same way. A next-generation inhibitor that defeats M327I does not touch this route, so the entry's conclusion that escape happens 'at the drug-menin interface' covers only the genetic subset. Fiskus W et al. 2026
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
Bound on the PDX result (Mahdavi 2025, PMID 41027490): heavily pretreated KMT2A::AFF1 samples from the same patients resist menin inhibition with no prior exposure, while on-target repression of canonical menin targets is sustained. Efficacy and on-target chromatin action are separable.
Cohort correction (Issa 2024, PMID 39121437): the registrational AUGMENT-101 phase 2 CR+CRh rate in centrally confirmed KMT2A-r was 22.8%, not the 30% phase 1 figure, with grade >=3 differentiation syndrome 16.0% and QTc prolongation 13.8%.
- related to →NUP98::NSD1, t(5;11)(q35;p15.5)
Condensate-transformation dissociation (Jevtic 2022, PMID 35073946): SMARCA5 inhibition abolishes NUP98-NSD1 transformation while leaving the condensates intact; condensate formation is not sufficient for the transformed phenotype.
- related to →NUP98::KDM5A (NUP98-JARID1A), t(11;12)(p15;p13)
Attribution correction (Ahn 2025, PMID 39922194): in NUP98-PHF23 the partner PHD finger targets chromatin and the FG repeats select which coactivators partition in, so chromatin occupancy is partner-driven, not FG-driven, in exactly the reader-domain NUP98 fusions this class names.
Resistance-route breadth (Fiskus 2026, PMID 41734382): menin-inhibitor-resistant AML arises with no MEN1 mutation, via an altered epigenome/proteome creating BRD4, SMARCA4 and CREBBP dependencies. MEN1 interface mutation is a subset of clinical escape, not the mechanism.
Durability bound (Arellano 2025, PMID 40332046): median duration of CR+CRh 4.7 months in the NPM1-mutant AUGMENT-101 cohort, 16.7% of responders reaching transplant; a comparable response rate without any fusion argues the target is a menin-dependent HOXA/MEIS1 state, not the fusion class.
the chaperone that reassembles nucleosomes behind the polymerase the fusion forces forward
Counter-evidence (Chen 2015, PMID 26279188): reducing PAF1 increases release of paused Pol II at thousands of genes by recruiting the super elongation complex - the very complex KMT2A::AFF1 recruits - so PAF1C loss de-represses elongation at the same time as it removes the ubiquitination route
Counter-evidence (Chen 2015, PMID 26279188): a positive MSL/PAF1/RNF20-to-P-TEFb feedback loop cannot be the general rule alongside the genome-wide human finding that PAF1 depletion increases pause release and Ser2P by recruiting SEC
Design objection (Chen 2015, PMID 26279188): PAF1 depletion increases paused-Pol-II release via SEC recruitment, and SEC is what KMT2A::AFF1 recruits, so the PRO-seq/TT-seq arm meant to separate elongation from ubiquitination measures both at once unless an SEC-occupancy arm is added
- ← co-occurs withH2BK120ub (H2Bub1) — the ubiquitin tag on histone H2B that licenses H3K79 methylation, and the mark this graph was missing
deposited co-transcriptionally by PAF1C-associated machinery on elongating Pol II
WW domain binds Pol II, coupling H2B ubiquitination to elongation
- ← part ofDPY30 — the smallest COMPASS subunit, and the only one that behaves like a selective dependency
shared subunit of the H3K4 machinery the fusion works through
the kinase that releases paused polymerase; deepest dependency in the complex
general elongation activation module carried into the fusion complex
the textbook scaffold, which these cells need less of than expected
the elongation machinery only acts once the fusion is anchored
- ← depends onKMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
same mechanism class, different partner
- ← depends onKMT2A::MLLT1 (MLL-ENL), t(11;19)(q23;p13.3)
- ← depends onKMT2A::MLLT10 (MLL-AF10), t(10;11)(p12;q23)
- ← depends onKMT2A::ELL (MLL-ELL), t(11;19)(q23;p13.1)
- ← depends onKMT2A::AFDN (MLL-AF6), t(6;11)(q27;q23)
- ← related toAQR (Aquarius) - a spliceosomal RNA helicase whose loss impairs MLL-rearranged leukaemia stem cells
Acts on productive gene-body elongation WITHOUT altering pause release, unlike the pause-release lever this mechanism is built around, and preferentially affects long genes.
loss of restraint is not sufficient; a partner-supplied activating gain is required