Oncofusion

start / graph / gene-ezh2

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q41-prc2-arm

EZH2 - the Polycomb enzyme that collaborates with the fusion instead of opposing it, and that no single-agent inhibitor can switch off

also called EZH2 · KMT6 · ENX-1 · Enhancer of zeste homolog 2 · histone-lysine N-methyltransferase EZH2

EZH2 is the enzyme that puts the main 'switch this gene off' chemical tag on DNA packaging proteins, and textbooks say it is the opposite of the KMT2A machinery this leukaemia runs on. In this leukaemia it turns out to work with the fusion rather than against it - but blocking EZH2 by itself does almost nothing, because a near-identical back-up enzyme called EZH1 takes over.

The same thing, for a scientist

EZH2 is the catalytic subunit of Polycomb repressive complex 2 (PRC2, UniProt Q15910, 746 aa) and writes H3K27me1/2/3. KMT2A is a trithorax-group protein and Polycomb is its canonical antagonist, so the expectation is opposition. **In KMT2A-rearranged leukaemia the published evidence points the other way.** Menin binds the *Ezh2* promoter and drives its expression, and EZH2 then represses C/EBPalpha target genes to hold MLL-AF9 cells undifferentiated - menin depletion lowers EZH2 and de-represses those targets [E1]. In the index disease, EZH2 acts together with KMT2A::AFF1 to silence *HDAC7* in t(4;11) infant pro-B ALL, and a menin inhibitor restores HDAC7 [E2]. In both lineages the axis reads menin -> EZH2 -> repression of a differentiation gene, which makes PRC2 a **downstream effector of the menin interface**, not its opponent. **But EZH2 alone is not the dependency.** *Ezh2*-null MLL-AF9 AML still self-renews to a third transplant with locus-specific H3K27me3 persisting via EZH1, while inactivating *Eed* ablates PRC2 and is incompatible with leukaemia growth [E3]. DepMap agrees and is decisive on the pharmacology: across 11 KMT2A-rearranged lines EZH2 gene effect is -0.201 against -0.204 in 43 other leukaemia lines (rank 9,672 of 17,728), and EZH1 is -0.100 against -0.102 - **neither catalytic paralogue is required alone** - while EED is -0.458 and SUZ12 -0.482 [E4]. EZH1 is transcribed in 7/7 panel lines (median 13.7 nTPM), so the redundancy is not an artefact of a silent paralogue [E4]. The therapeutic reading is specific: an EZH2-selective inhibitor should fail and a dual EZH1/EZH2 inhibitor should work, which is what is observed [E5][E6]. **The direction is contested outside this disease and that contest is real**: EZH2 carries inactivating, tumour-suppressor mutations in MDS/MPN and myelofibrosis [E7] and cooperates with RUNX1 loss to initiate ETP leukaemia [E8]. Nothing in this graph should state 'EZH2 is an oncogene' without a lineage and a genotype attached.

Go deeper
Dependency or tumour suppressor? The answer is lineage-and-lesion specific, and this graph should carry all three readings

Three well-evidenced positions coexist and they are not in conflict once the population is stated:

settingdirectionevidence
KMT2A-rearranged leukaemiaPRC2 is required; EZH2 alone is not (EZH1 covers it)345
KMT2A-rearranged leukaemia, mechanismEZH2 collaborates with menin and the fusion to repress differentiation genes12
MDS/MPN, myelofibrosis, ETP-ALLEZH2 is a tumour suppressor; loss-of-function mutations are recurrent78

The brief that dispatched this beat asked whether EZH2 is a dependency or a tumour suppressor here. The honest answer is that in KMT2A-rearranged leukaemia it is neither on its own: the complex is required, the catalytic subunit is redundant, and the mechanism by which the complex helps the leukaemia is repression of differentiation genes downstream of menin. The tumour-suppressor literature is real but belongs to a different genotype - myeloid neoplasms with 7q loss and acquired uniparental disomy 7, and early thymic progenitors carrying RUNX1 loss 8. Neither population is infant KMT2A::AFF1 B-ALL.

What this does not license is treating the tumour-suppressor result as irrelevant. It is the mechanistic basis of a real bite-back: partial, chronic EZH2 inhibition in a haematopoietic compartment is the pharmacological analogue of the loss-of-function state that predisposes to myeloid neoplasia, and the patients in question are infants with decades of life ahead.

The redundancy is measurable, and it decides the drug class

This is a rare case where a DepMap number, an expression control and a drug result all say the same thing, and the project's own DepMap rules are satisfied rather than violated.

From analyses/prc2-dependency-lineage-split/ (run 2026-09-20 for this beat) 4; index arm = 11 Cellosaurus-annotated KMT2A-rearranged DepMap lines, comparator = 43 other leukaemia lines, raw Chronos effects only, no z:

geneindex meancomparator meandifferencerank / 17,728pan-essentiality
EED-0.458-0.340-0.117824-0.277
SUZ12-0.482-0.452-0.0305,332-0.405
RBBP4-1.486-1.462-0.0246,113-1.347 (floor)
EZH2-0.201-0.204+0.0049,672-0.049
EZH1-0.100-0.102+0.0029,427-0.134
RBBP7-0.013-0.016+0.0039,520-0.155

Neither catalytic paralogue is required alone. That is the CRISPR signature of paralogue redundancy, and it is the same statement as Ezh2-null leukaemia surviving with EZH1-dependent H3K27me3 3. The expression control rules out the trivial explanation: EZH1 is detected at median 13.7 nTPM in all seven panel lines Human Protein Atlas carries, so a cell treated with an EZH2-selective inhibitor still contains a transcribed H3K27 methyltransferase 4.

The pharmacology matches exactly. UNC1999, which inhibits both enzymes, suppresses global H3K27me3/me2, de-represses Cdkn2a, and prolongs survival of MLL-AF9 mice, while its inactive analogue UNC2400 does not 5. Valemetostat (DS-3201), a dual EZH1/EZH2 inhibitor, inhibited B-ALL line growth more strongly than the EZH2-selective inhibitor EPZ-6438 (tazemetostat) and worked in a PDX carrying MLL-AF4 6. A separate dual-inhibitor programme reported activity against leukaemia lines with MLL-AF9, MLL-AF4 and AML1-ETO fusions 9.

So the drug-class question in this disease is already answered, and the answer is not the approved drug.

Lineage: the literature is myeloid, but the DepMap lean is not

Almost every KMT2A epigenetics result this graph holds has turned out to be MLL-AF9 and myeloid, and the PRC2 literature is mostly no exception: 3, 5 and 9 are MLL-AF9; 1 is mouse MLL-AF9 and its effector is C/EBPalpha, a myeloid differentiation factor a pro-B blast does not run.

Two things break the pattern and both matter.

  1. The index-disease result exists. 2 is human t(4;11) infant pro-B ALL, and reports EZH2 acting with KMT2A::AFF1 to silence HDAC7 - the same shape as the myeloid menin -> EZH2 -> differentiation-gene axis, with a B-lineage effector instead of C/EBPalpha. 6 adds a human MLL-AF4 B-ALL PDX responding to dual EZH1/EZH2 inhibition.
  2. The DepMap signal leans lymphoid, against expectation. Splitting the index arm 7 myeloid : 4 B-lineage and comparing each half to the matching half of the comparator 4, EED's difference is -0.058 in the myeloid half and -0.224 in the lymphoid half; EZH2's index-arm mean is -0.112 myeloid against -0.355 lymphoid. The lymphoid half is four lines and no infant line exists in DepMap, so this is weak. But it is the first epigenetic panel in this project where a lineage split did not reveal a myeloid artefact, and it points the same way as 2 and 6.

The honest summary: the mechanistic literature is myeloid, the two index-disease results are lymphoid and agree with it, and the dependency data lean lymphoid. No claim in this entry needs to be imported across lineage.

Accession resolution (Rules 62, 63, 65)

gene_exact:EZH2 AND organism_id:9606 AND reviewed:true returns exactly one reviewed human record, Q15910, primary gene name EZH2, 746 aa, protein name Histone-lysine N-methyltransferase EZH2 10. Resolved 2026-09-20 with gene_primary and length requested in fields.

The family hazard is real. EZH1 is Q92800, 747 aa - one residue longer - so a length check separates EZH2 from EZH1 by a single amino acid and must not be leaned on. gene-rbbp4 records the worse case in the same complex, where the two paralogues are exactly the same length and only the accession can detect a swap.

Facts and figures
symbolEZH2
uniprotQ15910
length AA746
uniprot primary gene nameEZH2
uniprot resolution notegene_exact:EZH2 AND organism_id:9606 AND reviewed:true returned exactly one row; primary gene name EZH2 matched; length 746 asserted so verify-accessions.py can catch an EZH1 (Q92800, 747 aa) swap. Resolved 2026-09-20.
paraloguegene-ezh1
complexprotein-prc2-complex
mark writtenmark-h3k27me3
catalytictrue
expression hpa nTPM median kmt2a r panel47.2
expression notehighest in the two KMT2A::AFF1 B-ALL lines of the panel (SEM 83.0, RS4;11 136.4) against a myeloid median of 40.0; NOMO-1 at 13.4 is below EZH1 in the same line
direction in this diseasecollaborator with menin and the fusion in repressing differentiation genes, NOT an antagonist - but catalytically redundant with EZH1
direction elsewheregain-of-function oncogene in follicular lymphoma and DLBCL; loss-of-function tumour suppressor in MDS/MPN, myelofibrosis and ETP-ALL
subnuclear localizationpolycomb bodies, inferred from the compartment's dependence on PRC2/EZH2 activity rather than measured as endogenous EZH2 protein in a body (PMID 22275876, PMID 16314526). This is a weaker record than it looks and is written here as an inference, not an observation.
shown for fusionNO. Neither the polycomb-body assembly work nor the extranuclear talin work touches a KMT2A-rearranged cell. The graph's own localization-polycomb-bodies-kmt2a-r entry holds what is known about that compartment in this disease; what is added here is that EZH2's placement in it is inferential.
essentiality
{
  "depmap_kmt2a_r_mean": -0.2005,
  "depmap_other_leukaemia_mean": -0.2045,
  "depmap_difference": 0.004,
  "depmap_rank_of_17728": 9672,
  "depmap_pan_essentiality_all_lines": -0.049,
  "depmap_index_myeloid_mean": -0.1121,
  "depmap_index_lymphoid_mean": -0.3551,
  "n_index_lines": 11,
  "n_comparator_lines": 43,
  "source": "analyses/prc2-dependency-lineage-split/results.json",
  "interpretation": "not a dependency and not selective; the informative experiment (EZH1/EZH2 double knockout) is absent from DepMap"
}
drugs
[
  "drug-tazemetostat",
  "drug-valemetostat"
]
localization
{
  "subcellular": "nucleus, chromatin-associated, as the catalytic subunit of PRC2 - with a documented and functionally distinct EXTRANUCLEAR pool that this graph should not omit: cytosolic Ezh2 methylates talin to control integrin-dependent adhesion and migration in neutrophils and dendritic cells, and an extranuclear VAV-dependent mechanism contributes to neoplastic transformation.",
  "subnuclear": "POLYCOMB BODIES, as a component of the compartment rather than as an imaged protein in its own right. Polycomb bodies form by progressive assembly of Polycomb complexes on H3K27me3 compartments - the mark EZH2 writes - and their motion is developmentally regulated; separately, BMI1 association with polycomb bodies is dynamic and REQUIRES PRC2/EZH2. Both are therefore statements that EZH2 activity builds the compartment, and neither is a direct image of endogenous EZH2 in it.",
  "how_shown": "live and fixed imaging of polycomb body assembly on H3K27me3 compartments (PMID 22275876); FRAP and dependency analysis of BMI1 in polycomb bodies under PRC2/EZH2 and DNMT1 perturbation (PMID 16314526); the extranuclear pool by biochemistry and imaging in primary innate immune cells (PMID 25751747)",
  "systems": "Drosophila and mammalian cells for the polycomb-body work; mouse neutrophils and dendritic cells for the extranuclear pool. No KMT2A-rearranged system in any of them.",
  "node": "localization-polycomb-bodies-kmt2a-r"
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.2005,
  "chronos_mean_other_leukaemia": -0.2045,
  "difference": 0.004,
  "pan_essentiality_all_lines": -0.049,
  "selectivity_rank": 9672,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • If PRC2 restrains the fusion programme in any compartment, an EZH2 or EZH1/EZH2 inhibitor makes the disease worse there. The published direction in KMT2A-rearranged leukaemia is collaboration, not restraint, but EZH2 is a bona fide tumour suppressor in MDS/MPN and in early thymic progenitors, and the marrow being treated contains those compartments.
  • EZH2 inhibition depletes H3K27me3 and raises H3K27ac at the same residue, increasing transcription and, in B cells with high AID activity or DNA-repair deficiency, increasing chromosomal translocation frequency (PMID 40825195). Giving a translocation-promoting agent in a translocation-defined B-lineage leukaemia in infants, whose remaining lifespan is measured in decades, requires a specific argument that the study's monotherapy-safe result covers the intended combination.
  • EZH1 makes EZH2-selective inhibition pharmacologically close to inert in this disease - so a negative result from a tazemetostat-class agent must NOT be recorded as 'PRC2 is not a target here'. The failure mode would be the drug, not the biology.
  • EZH2 is required for normal early B-cell development and Igh rearrangement (PMID 12496962). Inhibiting it in an infant B-lineage leukaemia means inhibiting it in the compartment that has to rebuild a B-cell repertoire afterwards.
  • The main published anti-leukaemic mechanism for PRC2 inhibition is de-repression of CDKN2A, which is deleted in a substantial fraction of KMT2A-rearranged leukaemias - so the biomarker most likely to be used to select patients may also be the lesion that abolishes the mechanism.
  • EZH2 inhibitors are not nuclear-selective by virtue of their target being a histone methyltransferase. The extranuclear Ezh2-talin axis controls integrin-dependent adhesion and transendothelial migration of neutrophils and dendritic cells (PMID 25751747), so an EZH2 inhibitor given to a patient is also an innate-immune-trafficking drug. In a neutropenic infant receiving intensive chemotherapy that is a specific and testable toxicity hypothesis, and no EZH2 trial in this graph reports a migration or infection endpoint framed that way.
Still open
  1. Is H3K27me3 actually present at HDAC7 in t(4;11) infant blasts, and does an EZH1/EZH2 dual inhibitor reproduce the MI-538 effect on HDAC7? PMID 40722046 names EZH2 as the silencer but tests menin inhibition, not PRC2 inhibition, so the causal step from EZH2 to HDAC7 has not been perturbed directly.
  2. Does the menin -> EZH2 -> differentiation-gene axis established for C/EBPalpha in MLL-AF9 (PMID 23349306) run through a B-lineage effector in KMT2A::AFF1, and is HDAC7 that effector? If so the two lineages share one mechanism with different output genes, and menin inhibition is a PRC2-directed therapy nobody has described as one.
  3. Is there any KMT2A-rearranged system in which the EZH1/EZH2 DOUBLE knockout has been performed? Every dependency statement here rests on single-gene screens, which are the one design guaranteed to miss a redundant pair, or on inhibitors, which cannot separate catalysis from scaffolding.
  4. Does menin inhibition lower EZH2 protein in patients? If PMID 23349306 transfers, EZH2 level is a free pharmacodynamic readout of menin inhibition that no revumenib or ziftomenib trial has reported.
Every source behind this entry (11)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Menin binds the Ezh2 promoter and promotes Ezh2 expression; EZH2 interacts with C/EBPalpha and represses its target genes, and menin depletion reduces MLL binding at the Ezh2 locus, EZH2 expression, and EZH2 binding and repressive H3K27 methylation at C/EBPalpha targets, inducing pro-differentiation C/EBPalpha target expression in MLL-AF9 leukaemia.

    PMID 23349306DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. Mouse MLL-AF9 primary leukaemia cells with acute menin depletion - a myeloid fusion and a myeloid effector (C/EBPalpha), not the index KMT2A::AFF1 pro-B disease. 'Menin binds the Ezh2 promoter' is a ChIP-level statement; the microarray design identifies menin-regulated pathways rather than proving the EZH2 arm is the sole route to the differentiation block. The paper does not report whether EZH1 substitutes when EZH2 falls.

    Evidence that argues against this (1)
    • If EZH2 protein level were the mediator, removing EZH2 should phenocopy menin depletion - and it does not. Ezh2-null MLL-AF9 leukaemia still self-renews through three rounds of transplantation because EZH1 sustains locus-specific H3K27me3, whereas Eed inactivation is incompatible with growth. Menin-driven EZH2 expression is therefore at most one of two redundant routes to the same mark. Neff T et al. 2012
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In t(4;11) infant B-ALL, HDAC7 is epigenetically silenced by EZH2 together with the KMT2A::AFF1 fusion protein; the menin inhibitor MI-538 restores HDAC7 expression, an effect enhanced by the class I HDAC inhibitor chidamide, driving leukaemic pro-B cells toward differentiation and increasing glucocorticoid sensitivity of PDX cells.

    HDAC7 induction combined with standard-of-care chemotherapy provides a therapeutic advantage in t(4;11) infant B-cell acute lymphoblastic leukemia. · de Barrios O, Ocón-Gabarró I, Gusi-Vives M, Collazo O, Meler A, Romecín PA, et al. · Biomark Res 2025
    PMID 40722046DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. This is the only entry point from EZH2 into the index disease and it is a single 2025 report. 'Silenced by EZH2 and KMT2A::AFF1' is the stated conclusion of a multiomics plus cell-line analysis; the abstract does not say whether EZH2 occupancy at HDAC7 was measured by ChIP, whether H3K27me3 was measured there, or whether an EZH2 inhibitor reproduces the MI-538 effect. The therapeutic arm tested is menin inhibition plus an HDAC inhibitor, NOT an EZH2 inhibitor, so the paper does not test PRC2 pharmacology in this disease at all.

    Evidence that argues against this (1)
    • HDAC7 and HDAC9 are the trap Rule 65 was written for: HDAC9 carries HDAC7 as a synonym and its protein name reads 'Histone deacetylase 7B'. In this disease the two point in opposite therapeutic directions - HDAC7 low and to be induced, HDAC9 high and to be repressed - so any downstream use of this result must assert the accession (HDAC7 = Q8WUI4, 952 aa) rather than the symbol. UniProt 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Ezh2-null MLL-AF9 AML failed to accelerate on secondary transplantation but maintained self-renewal to the third round, indicating Ezh2 is not strictly required and acts in progression, with H3K27me3 persisting locus-specifically through suggested partial compensation by Ezh1; inactivation of the essential PRC2 gene Eed completely ablated PRC2 function and was incompatible with leukaemia growth, with down-regulated Myc target genes and up-regulated PRC2 targets.

    Polycomb repressive complex 2 is required for MLL-AF9 leukemia. · Neff T, Sinha AU, Kluk MJ, Zhu N, Khattab MH, Stein L, et al. · Proc Natl Acad Sci U S A 2012
    PMID 22396593DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. Conditional mouse alleles in an MLL-AF9 model - a myeloid fusion, not KMT2A::AFF1, and no infant or B-lineage arm. 'Partial compensation by Ezh1' is inferred from persistence of locus-specific H3K27me3 after Ezh2 inactivation, not from an Ezh1 deletion. Eed inactivation removes the whole complex and therefore cannot separate loss of H3K27 methylation from loss of PRC2 scaffolding or of PRC2's non-canonical functions.

    Evidence that argues against this (1)
    • The Eed requirement is not solely about de-repressing Cdkn2a, which is how the result is usually summarised. Combined genetic inactivation of Cdkn2a and Eed only PARTIALLY rescues growth of Eed-null MLL-AF9 cells in vitro and in vivo; Gata2, Egr1 and Cdkn2b are also de-repressed and MLL-fusion direct targets are negatively enriched. An intervention aimed at phenocopying Eed loss therefore cannot be evaluated on a Cdkn2a readout alone. Danis E et al. 2015
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Across 11 Cellosaurus-annotated KMT2A-rearranged DepMap leukaemia lines against 43 other leukaemia lines, neither EZH2 (index -0.201, comparator -0.204, difference +0.004, rank 9,672 of 17,728) nor EZH1 (-0.100 vs -0.102, +0.002) is a dependency or selective, while EED (-0.458 vs -0.340, difference -0.117, rank 824) and SUZ12 (-0.482) are required; both EZH1 (median 13.7 nTPM) and EZH2 (median 47.2 nTPM) are transcribed in 7/7 panel lines in Human Protein Atlas cell-line RNA, so the absence of single-paralogue dependency reflects redundancy and not silence.

    What this evidence cannot show

    A CRISPR knockout is not a drug, and a single-gene knockout is exactly the design that cannot see a redundant pair - the informative measurement is the EZH1/EZH2 DOUBLE knockout, which DepMap does not contain. 11 index lines, none infant-derived and none a true infant KMT2A::AFF1 model; the lymphoid half is 4 lines. Cellosaurus fusion annotation may be incomplete, biasing toward the null. DepMap release version is not stated by the Open Targets API. nTPM is bulk RNA from immortalised cell lines and is not protein. Human Protein Atlas does not carry SHI-1 or HB11;19, so the expression control covers 7 of the 11 index lines. Raw effect sizes only; no z and no permutation null, so nothing here is a hit.

    Evidence that argues against this (1)
    • Reading 'EZH2 is not a dependency' off a single-gene screen is the standard failure of paralogue biology, and the published genetics say the double perturbation gives a different answer: shRNA suppression of Eed, Suz12 OR Ezh1/Ezh2 together caused proliferation arrest and differentiation in MLL-AF9;Nras-G12D AML with minimal effect on non-transformed haematopoietic lines. The DepMap null for EZH2 is evidence about EZH2 ALONE and says nothing about PRC2 catalysis. Shi J et al. 2013
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    UNC1999, a selective inhibitor of both EZH2 and EZH1, suppresses global H3K27me3/me2 and inhibits growth of MLL-rearranged leukaemia cells; its transcriptome overlaps EED knockdown, it preferentially affects distal regulatory elements such as enhancers, it de-represses Polycomb targets including Cdkn2a with a concurrent gain of H3K27 acetylation, and oral administration prolongs survival of MLL-AF9 mice, while the inactive analogue UNC2400 does not.

    Selective inhibition of EZH2 and EZH1 enzymatic activity by a small molecule suppresses MLL-rearranged leukemia. · Xu B, On DM, Ma A, Parton T, Konze KD, Pattenden SG, et al. · Blood 2015
    PMID 25395428DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. UNC1999 is a chemical probe, not a clinical agent, and 'selective for EZH2 and EZH1' means selective against other methyltransferases rather than between the two paralogues. The in vivo arm is a murine MLL-AF9 model - myeloid, not the index fusion. The inactive-analogue control (UNC2400) is the right control and is why this is graded strong rather than moderate.

    Evidence that argues against this (1)
    • A large part of the anti-leukaemic effect is attributed to Cdkn2a de-repression, and CDKN2A is deleted in a substantial fraction of human KMT2A-rearranged leukaemias and of the lines used to model them. Where CDKN2A is absent the predicted mechanism cannot operate; Eed-null Cdkn2a-null MLL-AF9 cells are only partially rescued, so the residual effect runs through genes a trial design would not be measuring. Danis E et al. 2015
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The EZH1/EZH2 dual inhibitor DS-3201 (valemetostat) inhibited growth of B-cell ALL cell lines more significantly and strongly than the EZH2-specific inhibitor EPZ-6438 (tazemetostat), induced cell-cycle arrest and apoptosis, up-regulated CDKN1C and TP53INP1 as direct EZH1/2 targets confirmed by CRISPR/Cas9 knockout, and significantly inhibited growth of a B-ALL patient-derived xenograft harbouring MLL-AF4.

    Dual inhibition of EZH1/2 induces cell cycle arrest of B cell acute lymphoblastic leukemia cells through upregulation of CDKN1C and TP53INP1. · Ito J, Yamagata K, Shinohara H, Shima Y, Katsumoto T, Aikawa Y, et al. · Int J Hematol 2023
    PMID 36280659DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. This is the single most index-relevant pharmacology result in the beat and it rests on one PDX; the abstract gives no n, no dose, and no statement of whether the PDX was paediatric or infant. The head-to-head with EPZ-6438 is a cell-line growth comparison, not a matched-exposure or matched-target-engagement comparison, so 'dual beats selective' is confounded with potency unless the full text shows H3K27me3 depletion matched between arms. The mechanism (CDKN1C, TP53INP1) is established in B-ALL lines generally, not specifically in the MLL-AF4 PDX.

    Evidence that argues against this (1)
    • Dual EZH1/EZH2 inhibition is not free of the tumour-suppressor direction. EZH2 loss of function is recurrent and causative in myeloid neoplasia, so a dual inhibitor achieves pharmacologically what those mutations achieve genetically, in a marrow that must be repopulated after treatment. A response measured over weeks in a xenograft cannot see a clonal-selection risk that in patients would play out over years. Ernst T et al. 2010
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Homozygous EZH2 mutations were found in 9 of 12 individuals with 7q acquired uniparental disomy, and screening 614 individuals with myeloid disorders revealed 49 monoallelic or biallelic EZH2 mutations in 42 individuals - most commonly in MDS/MPN (27 of 219, 12%) and myelofibrosis (4 of 30, 13%) - with the mutations causing premature chain termination or direct abrogation of histone methyltransferase activity, indicating EZH2 acts as a tumour suppressor in myeloid malignancy.

    Inactivating mutations of the histone methyltransferase gene EZH2 in myeloid disorders. · Ernst T, Chase AJ, Score J, Hidalgo-Curtis CE, Bryant C, Jones AV, et al. · Nat Genet 2010
    PMID 20601953DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. The population is adult myeloid neoplasia - MDS/MPN and myelofibrosis - and contains no KMT2A-rearranged, paediatric or infant cohort. A lifetime constitutive loss-of-function state is not the same exposure as partial reversible pharmacological inhibition, so this is a mechanism-level warning about direction, not a quantified drug risk.

    Evidence that argues against this (1)
    • The opposite direction is equally well established in the same organ system: activating EZH2 mutations are present in roughly 20% of follicular lymphomas and predict response to an EZH2 inhibitor, with a 69% objective response rate in the EZH2-mutant cohort against 35% in wild-type. EZH2's direction is set by the disease, not by the gene. Morschhauser F et al. 2020
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Targeting Ezh2 and Runx1 inactivation to early lymphoid progenitors in mice causes marked expansion of pre-leukaemic early thymic progenitors with transcriptional signatures characteristic of ETP leukaemia, and adding Flt3-ITD produces an aggressive leukaemia co-expressing myeloid and lymphoid genes that can be established and propagated in vivo by the expanded ETPs.

    Ezh2 and Runx1 Mutations Collaborate to Initiate Lympho-Myeloid Leukemia in Early Thymic Progenitors. · Booth CAG, Barkas N, Neo WH, Boukarabila H, Soilleux EJ, Giotopoulos G, et al. · Cancer Cell 2018
    PMID 29438697DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. Mouse model of ETP-ALL, a T-lineage disease with RUNX1 and signalling lesions; there is no KMT2A rearrangement in the model and no B-lineage arm. Cited only to establish that EZH2 loss is leukaemogenic in a lymphoid progenitor compartment, which is the population most relevant to a bite-back argument for a B-lineage disease.

    Evidence that argues against this (1)
    • In this model Ezh2 loss alone is not sufficient - it requires Runx1 co-inactivation and a signalling lesion - so the transformation risk it demonstrates is conditional on a second and third hit rather than being a property of EZH2 inhibition per se. The same paper reports that both mouse and human ETP leukaemias are sensitive to BET inhibition, which reverses the gene-expression changes induced by Ezh2 inactivation, so the state is pharmacologically reversible. Booth CAG et al. 2018
  9. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Orally bioavailable EZH1/2 dual inhibitors suppressed cellular H3K27 trimethylation more than EZH2-selective inhibitors, showed greater antitumour efficacy in vitro and in vivo, and in a haematological cancer panel assay were effective against leukaemia carrying fusion genes including MLL-AF9, MLL-AF4 and AML1-ETO, with no severe toxicity seen in rats treated for 14 days at levels above those used in the antitumour study.

    Novel orally bioavailable EZH1/2 dual inhibitors with greater antitumor efficacy than an EZH2 selective inhibitor. · Honma D, Kanno O, Watanabe J, Kinoshita J, Hirasawa M, Nosaka E, et al. · Cancer Sci 2017
    PMID 28741798DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. A cell-line panel assay from the originating pharmaceutical programme (the valemetostat sponsor), so publication and selection bias apply. 'Effective against leukaemia with MLL-AF9, MLL-AF4' is a panel-level statement with no n, no potency values and no fusion-versus-non-fusion comparison in the abstract; the primary efficacy demonstration is in EZH2-mutant DLBCL. The 14-day rat study is a short general-toxicology statement and carries no information about haematopoietic effects over the months a paediatric leukaemia protocol would require.

    Evidence that argues against this (1)
    • EZH2 inhibition depletes H3K27me3 and raises H3K27ac at the same residue, increasing transcriptional activity. In B cells with high AID activity or DNA-repair deficiency this significantly increases the frequency of chromosomal translocations - a specific hazard in a disease defined by a translocation and treated in a B-lineage compartment. The same study found monotherapy did NOT change translocation frequency, so the risk is conditional on combination or on a repair defect rather than intrinsic. Tao J et al. 2025
  10. E10established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    EZH2 resolves to a single reviewed human UniProt record, Q15910 (EZH2_HUMAN), primary gene name EZH2, 746 amino acids, protein name Histone-lysine N-methyltransferase EZH2; its paralogue EZH1 is Q92800, 747 amino acids.

    What this evidence cannot show

    Resolved 2026-09-20. EZH2 and EZH1 differ by one residue in length (746 vs 747), so a length check alone is not a safe discriminator between these two paralogues and the accession must be asserted. UniProt length is for the canonical isoform.

    Evidence that argues against this (1)
    • Length is an even weaker discriminator elsewhere in this complex: RBBP4 (Q09028) and RBBP7 (Q16576) are BOTH 425 aa, so for that pair the length check Rule 65 prescribes cannot detect a swap at all and only the accession can. UniProt 2026
  11. E11moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Polycomb bodies are built by progressive assembly of Polycomb complexes on H3K27me3 compartments - the mark EZH2 deposits - with developmentally regulated motion, and BMI1's association with polycomb bodies is dynamic and requires PRC2/EZH2; EZH2's subnuclear placement in this compartment is therefore an inference from the compartment's dependence on its catalytic product rather than a direct observation of the protein.

    PMID 22275876DOI
    What this evidence cannot show

    This is deliberately recorded as weaker than the usual textbook statement. Neither cited study images endogenous human EZH2 in a polycomb body; one images the compartment forming on the mark, the other images BMI1 and perturbs PRC2. Neither is a KMT2A-rearranged or haematopoietic system. This graph's condensate standards apply to any claim that polycomb bodies are phase-separated as opposed to clustered binding sites.

    Evidence that argues against this (1)
    • A substantial fraction of EZH2's biology is not in the nucleus at all. Ezh2 directly methylates talin, an extranuclear regulatory protein, disrupting talin-F-actin binding and controlling adhesion turnover in neutrophils and dendritic cells, with the effect abolished by disrupting the Ezh2-Vav1 interaction. An EZH2 inhibitor is therefore not a nuclear-only agent, and a localisation record that writes EZH2 as a chromatin protein will not predict its effects on leukocyte migration. Gunawan M et al. 2015
Connected to