Giving a translocation-promoting epigenetic drug in a translocation-defined B-lineage leukaemia, to infants
also called EZH2 inhibitor genomic instability · H3K27me3-to-H3K27ac switch and AID · PRC2 inhibitor second-malignancy risk
The drugs that block Polycomb work by stripping off a 'switch off' tag, which gets replaced by an 'switch on' tag and makes nearby genes more active. In B cells that are actively rearranging their antibody genes, that extra activity makes chromosome breaks join up wrongly more often - and this leukaemia is caused by exactly that kind of wrong join.
The same thing, for a scientist
The mechanism by which a PRC2 inhibitor acts is the same mechanism by which it can create chromosomal translocations in a B cell. EZH2 inhibition depletes H3K27me3 while concurrently **enhancing H3K27ac**, selectively increasing transcriptional activity and **facilitating chromosomal translocation formation** where AID activity is high or DNA ligase IV is deficient [E1]. **The authors' own headline is reassuring and must be stated first: monotherapy did not significantly change AID expression or translocation frequency** in either mouse CH12F3 or human MEC-1 B cells [E1]. The hazard appeared only when EZH2 inhibition was combined with PI3K-delta inhibition, which raises AID, or in a *Ligase4*-deficient background. So this is a **conditional** risk, not a contraindication. **Three features of the intended setting make the condition worth taking seriously anyway.** (i) The disease is *defined* by a translocation, and the graph's own break-machinery arm exists because the mechanism of that break is a live question. (ii) The patients are **infants**, whose remaining lifespan is measured in decades, so a second-malignancy risk that a 25-patient adult R/R trial at median age 69 could never detect is the relevant one [E2]. (iii) The class is specifically B-lineage-directed here, and the same UNC1999 and valemetostat pharmacology shows the H3K27me3-to-H3K27ac exchange is the standard on-target signature [E3][E4] - so the pharmacodynamic that would be used to confirm target engagement **is** the hazard signature. Independently, **EZH2 loss of function is the causative lesion of a class of myeloid neoplasia** [E5], which is a second and distinct second-malignancy route.
What is and is not claimed
Claimed:
- The on-target pharmacodynamic of a PRC2 inhibitor in a B cell is an H3K27me3-to-H3K27ac exchange that raises transcription at affected loci 13.
- In B cells with high AID activity or DNA-repair deficiency, that exchange increases translocation frequency 1.
- The intended combination partners in this disease have not been tested for this interaction.
- EZH2 loss of function independently predisposes to myeloid neoplasia 5 and, with RUNX1 loss, to ETP leukaemia in mice.
Not claimed:
- That PRC2 inhibitor monotherapy is genotoxic. The paper explicitly concludes the opposite and this entry would be misrepresenting it to say otherwise 1.
- That AID activity is high in infant t(4;11) pro-B blasts. It is not established either way in the cited work, and pro-B cells are generally pre-AID; the relevant compartment for an AID-dependent hazard would be the patient's normal germinal-centre B cells, not the leukaemia.
- That any translocation has been observed in a patient on this drug class.
The honest position is: a specific, published, mechanistically plausible route by which this drug class could cause a second B-lineage malignancy exists, it is conditional on factors nobody has measured in this population, and it has never been assessed in a paediatric protocol.
What would retire this risk, and what would confirm it
Retire it if: AID expression is undetectable in infant t(4;11) blasts and in the normal B-cell compartment during and after therapy, and the intended combination (glucocorticoid plus menin inhibitor plus a PRC2 inhibitor) does not raise AID, and a translocation-capture assay in treated primary B cells is negative. The first is a single RNA measurement on samples that already exist.
Confirm it if: the intended combination raises AID, or a KMT2A breakpoint-cluster-region translocation-capture assay in PRC2-inhibitor-treated primary human B cells shows increased junction frequency. This graph's break-machinery arm already holds the KMT2A BCR biology needed to design that readout.
Either way the measurement is cheap and nobody has made it. The reason this node exists is the project's standing rule that a bite-back must be on record before an intervention is proposed, not after it fails.
| kind | toxicity |
|---|---|
| also affects | any EED-directed PRC2 inhibitor, since the H3K27me3-to-H3K27ac exchange is subunit-agnostic |
| severity | unquantified; conditional on AID activity or DNA-repair status, and on treatment duration |
| mechanism | PRC2 inhibition depletes H3K27me3 and raises H3K27ac at the same residue, increasing transcriptional activity and facilitating chromosomal translocation formation where AID is high or Ligase4 is deficient (PMID 40825195) |
| second independent route | EZH2 loss of function is the causative lesion of a class of myeloid neoplasia (PMID 20601953) and, with RUNX1 loss and a signalling mutation, initiates ETP leukaemia in mice (PMID 29438697) |
| monotherapy status | NOT demonstrated - the same study found EZH2 inhibitor monotherapy did not change AID expression or translocation frequency |
| population at risk | infants with a translocation-defined B-lineage leukaemia and decades of expected survival; the normal germinal-centre B-cell compartment rather than the pro-B blast is the likely AID-positive tissue |
| status of assessment | never assessed in any paediatric protocol or in any KMT2A-rearranged system |
affects
[ "drug-tazemetostat", "drug-valemetostat" ]
mitigation
[ "measure AID (AICDA) expression in infant t(4;11) blasts and in the normal B-cell compartment during therapy - a single RNA measurement on existing samples", "test the intended combination (glucocorticoid + menin inhibitor + PRC2 inhibitor) for AID induction before proposing it", "run a translocation-capture assay across the KMT2A breakpoint cluster region in PRC2-inhibitor-treated primary human B cells", "limit exposure duration; the hazard is cumulative and the mechanism is inseparable from the therapeutic effect" ]
- The pharmacodynamic that confirms this drug class is working - H3K27me3 down, H3K27ac up - is the same event that constitutes the hazard. It cannot be engineered out of the molecule, only restricted by setting, duration or combination.
- The hazard is conditional and the conditions are unmeasured in this population, which is the worst state for a safety question: it will read as 'no evidence of risk' in a protocol document while actually being 'no one looked'.
- Two independent second-malignancy routes exist and they point at different lineages - AID-dependent B-lineage translocations, and EZH2-loss-driven myeloid neoplasia. A monitoring plan designed around one would miss the other.
- The population is infants. A risk that materialises at a rate too low for a 25-patient adult trial to see, over a latency longer than that trial's follow-up, is exactly the risk profile that paediatric oncology has historically discovered too late.
- Is AICDA expressed in infant t(4;11) blasts, or in the normal B-cell compartment of treated infants? This is the single measurement that decides whether the AID-dependent half of the hazard applies, it can be made on samples that already exist, and nobody has reported it.
- Do the intended combination partners raise AID? PI3K-delta inhibition does, and that is what converted a safe monotherapy into a hazardous combination in PMID 40825195. Glucocorticoids and menin inhibitors have not been tested for this.
- Would a translocation-capture assay across the KMT2A breakpoint cluster region detect increased junction frequency under a PRC2 inhibitor? This graph's break-machinery arm holds the BCR biology to design it and the experiment has never been proposed.
- Is the myeloid second-malignancy route separable from the therapeutic effect? The ETP model shows the Ezh2-loss state is BET-inhibitor reversible, which raises the unusual possibility that a specific agent could be held in reserve against a drug-induced pre-leukaemic state.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
EZH2 inhibition in B cells depletes H3K27me3 while concurrently enhancing H3K27ac, selectively increasing transcriptional activity and facilitating chromosomal translocation formation in the presence of high AID activity or Ligase4 deficiency; treatment with EZH2 inhibitors as monotherapy did not significantly change AID expression or AID-dependent chromosomal translocation frequency in either mouse CH12F3 or human MEC-1 B cells, but combining EZH2 inhibition with PI3K-delta inhibition - which enhances AID expression - significantly increased translocation frequency in both, and EZH2 inhibition further enhanced translocation formation in DNA ligase IV-deficient mouse B cells.
Epigenetic changes by EZH2 inhibition increase translocations in B cells with high AID activity or DNA repair deficiency. · Tao J, Alessandri L, Gasparetto A, Zhao L, Zhang X, Alt FW, et al. · Blood 2025What this evidence cannot show
Abstract read from the eutils record; full text not read. The systems are a mouse B-cell lymphoma line (CH12F3) and a human CLL line (MEC-1), neither of which is a pro-B cell, an infant cell or a KMT2A-rearranged blast. The sensitising conditions - high AID activity from PI3K-delta inhibition, or Ligase4 deficiency - are not established features of infant t(4;11) ALL or of the compartment treated. **The authors' stated conclusion is that EZH2 inhibitors are genetically SAFE as monotherapy**, and this entry must be read as recording a conditional hazard, not a contraindication.
Evidence that argues against this (1)
- The authors' own framing is the counter: monotherapy changed neither AID expression nor translocation frequency in either system, so the demonstrated hazard requires a second perturbation. Presenting this as a general genotoxicity warning would misrepresent the paper, and this graph has twice had to correct entries that left a body asserting something its counter-evidence refuted. Tao J et al. 2025
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The registrational experience for the dual EZH1/EZH2 inhibitor valemetostat is 25 patients with relapsed or refractory adult T-cell leukaemia/lymphoma at median age 69 with a median of 3 prior lines, in which the grade 3 or above adverse events were thrombocytopenia, anaemia, lymphopenia, leukopenia and neutropenia - a population and a follow-up in which a late second malignancy could not be detected.
An open-label, single-arm phase 2 trial of valemetostat for relapsed or refractory adult T-cell leukemia/lymphoma. · Izutsu K, Makita S, Nosaka K, Yoshimitsu M, Utsunomiya A, Kusumoto S, et al. · Blood 2023What this evidence cannot show
Abstract read from the eutils record; full text not read. This citation is used as an argument about what the trial could NOT observe, which is an absence-of-testing argument rather than a finding: 25 heavily pretreated patients at median age 69 followed for a period the abstract does not state cannot bound a decades-long second-malignancy risk in infants. That is a statement about statistical and temporal power, not a criticism of the trial.
Evidence that argues against this (1)
- An absence of observed second malignancies in an adult R/R population is not evidence of risk either. The correct reading is that the question is unaddressed in both directions, and that a paediatric development programme would have to address it prospectively - which is the standard position for any epigenetic agent in children and not unique to this class. Keam SJ 2022
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The dual EZH1/EZH2 inhibitor UNC1999 suppresses global H3K27me3/me2 in MLL-rearranged leukaemia cells and de-represses Polycomb targets, with gene de-repression correlating with a decrease in H3K27me3 and a concurrent gain in H3K27 acetylation, preferentially at distal regulatory elements such as enhancers.
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 2015What this evidence cannot show
Abstract read from the eutils record; full text not read. Establishes that the methyl-to-acetyl exchange is the on-target signature in MLL-rearranged cells, which is what links the therapeutic mechanism to the hazard mechanism. UNC1999 is a chemical probe, not a clinical agent, and the system is a leukaemia line rather than a normal B cell - so this is the pharmacodynamic half of the argument, not the genotoxicity half.
Evidence that argues against this (1)
- The exchange being the on-target signature also means it cannot be engineered away: a PRC2 inhibitor that did NOT raise H3K27ac would not be working. Unlike most bite-backs, this one cannot be separated from the mechanism by a better molecule - only by restricting the setting, the duration or the combination. Tao J et al. 2025
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
EZH2 is required for normal early B-cell development and for rearrangement of the immunoglobulin heavy chain gene, and is a key regulator of histone H3 methylation in early B-cell progenitors, established by Cre-mediated conditional mutagenesis in mice.
Ezh2 controls B cell development through histone H3 methylation and Igh rearrangement. · Su IH, Basavaraj A, Krutchinsky AN, Hobert O, Ullrich A, Chait BT, et al. · Nat Immunol 2003What this evidence cannot show
Abstract read from the eutils record; full text not read. Mouse conditional knockout, 2003, no pharmacological arm and no EZH1 analysis - so it does not tell you what an EZH2-selective inhibitor would do to a human infant's B-cell development, and EZH1 might buffer some of it. Cited to establish that the compartment being treated is the compartment that performs V(D)J recombination, which is the mechanistic link between this developmental requirement and the translocation hazard.
Evidence that argues against this (1)
- A constitutive genetic requirement in development is not a safety signal for partial reversible pharmacological inhibition in a child - the same scope objection this project applies to gnomAD constraint. Ezh2 conditional knockout removes the protein for the life of the lineage; a drug occupies the active site for hours. Su IH et al. 2003
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Screening 614 individuals with myeloid disorders identified 49 monoallelic or biallelic EZH2 mutations in 42 individuals - most commonly in MDS/MPN (12%) and myelofibrosis (13%) - causing premature chain termination or direct abrogation of methyltransferase activity, establishing EZH2 as a tumour suppressor for myeloid malignancies; separately, Ezh2 and Runx1 inactivation targeted to early lymphoid progenitors in mice expands pre-leukaemic early thymic progenitors and, with Flt3-ITD, produces an aggressive lympho-myeloid leukaemia.
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 2010What this evidence cannot show
Abstract read from the eutils record; full text not read. This citation covers the first half of the claim only; the ETP-leukaemia half is PMID 29438697 and is cited in full on `gene-ezh2`. Adult myeloid neoplasia with 7q abnormalities - no paediatric, infant or KMT2A-rearranged cohort. A germline-to-lifetime loss-of-function state is a different exposure from reversible pharmacological inhibition.
Evidence that argues against this (1)
- In the ETP model, Ezh2 loss alone is insufficient - it requires Runx1 co-inactivation and a signalling lesion - so the transformation risk is conditional on additional hits. And in the same model the resulting leukaemia is sensitive to BET inhibition, which reverses the Ezh2-inactivation gene-expression changes, so the state is pharmacologically reversible rather than a fixed commitment. Booth CAG et al. 2018
- ← confers resistance toAnother fusion family did the experiment this beat wants: in FUS::ERG and CBFA2T3::GLIS2 AML, MHC class I and II are absent at diagnosis, EZH2 drives it, and an EZH2 inhibitor restores presentation
importing the EZH2-inhibitor programme into B-lineage disease is not neutral
bite-back: the on-target H3K27me3-to-H3K27ac exchange is also the reported translocation-promoting event in B cells
same hazard, more completely, because dual inhibition depletes the mark further
the hazard is a property of the mark's removal, not of any one molecule
EZH2 is required for early B-cell development and Igh rearrangement; the treated compartment is the one that recombines