Oncofusion

start / graph / risk-dot1l-inhibition-deepens-an-existing-h3k79-deficit

Riskmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q55-calm-af10

Bite-back: a DOT1L inhibitor in CALM-AF10 deepens a genome-wide H3K79 deficit that already causes chromosome breaks

also called DOT1L inhibition risk in CALM-AF10

CALM-AF10 cells already have far too little of the DOT1L chemical tag across most of their genome, and that shortage makes their chromosomes break. A drug that blocks DOT1L removes the small amount that is in the wrong place, and also removes what little is left everywhere else.

The same thing, for a scientist

The mechanistically obvious drug for this fusion is a DOT1L inhibitor, and the preclinical case is strong [E1]. The bite-back is that **CALM-AF10 cells already carry global H3K79 hypomethylation, are more sensitive to gamma-irradiation, show increased chromosomal instability, and the patients carry more secondary chromosomal aberrations than patients with other fusions** [E2]. A DOT1L inhibitor removes the focal excess *and* deepens the global deficit. **Nobody has measured what that does**: no study in this beat's reading reports chromosome breaks, micronuclei or mutational burden under DOT1L inhibition in a CALM-AF10 system. Two further edges: DOT1L is required for normal adult haematopoiesis, and Dot1L-deficient mouse embryonic stem cells show aneuploidy and telomere elongation alongside H3K79 hypomethylation - the same phenotype the fusion produces [E3]. And the clinical experience that would have surfaced this was never obtained here, because both pinometostat phase 1 trials required an **11q23** rearrangement, which excludes t(10;11)(p13;q14) [E4].

Go deeper
The specific measurement that is missing

Any DOT1L-inhibitor experiment in a CALM-AF10 model should carry a genome-stability readout alongside the efficacy readout: micronucleus frequency, metaphase breaks and gaps, or gamma-H2AX. That is one extra panel on an experiment already being run, and its absence is the reason this risk is currently a hypothesis rather than a number. A second measurement worth having is whether the global mark recovers on washout faster or slower than the focal mark - if the focal mark returns first, the therapeutic index of intermittent dosing is worse than it looks.

Facts and figures
kindessential-function
severityunquantified - the hypothesis is specific and no measurement exists
mechanismDOT1L inhibition deepens a pre-existing genome-wide H3K79 deficit in a cell type already showing chromosomal instability and irradiation sensitivity
mitigationCarry a genome-stability readout (micronuclei, metaphase breaks/gaps, gamma-H2AX) alongside every efficacy readout in a CALM-AF10 DOT1L-inhibitor experiment; measure washout kinetics of the global versus the focal mark before choosing an intermittent schedule.
affects
[
  "drug-pinometostat"
]
compounding factors
[
  "DOT1L is required for normal adult haematopoiesis",
  "the disease already accumulates secondary chromosomal aberrations, which is also a resistance engine",
  "no clinical DOT1L-inhibitor experience exists in this fusion because the trials excluded it by cytogenetic band"
]
How this could bite us back
  • The drug with the best mechanistic case in this disease may be the one that accelerates its worst feature.
  • The clinical failure of DOT1L inhibition in KMT2A-rearranged disease is being carried into this fusion as though it were evidence, when the trials structurally excluded these patients. An absence of testing is not a negative result - and the corollary is that the safety experience does not transfer either.
Still open
  1. Does DOT1L inhibition increase chromosome breaks in CALM-AF10 cells? One extra panel on an experiment people are already running.
  2. Do the global and focal H3K79 pools recover at the same rate after washout? If the focal pool returns first, intermittent dosing is worse than continuous by this logic.
  3. Is the elevated secondary-aberration rate in CALM-AF10 patients enough to make acquired resistance to any targeted agent faster than in other fusions?
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Genetic Dot1l inactivation and pharmacological Dot1l inhibition with EPZ004777 both abrogate CALM-AF10-mediated transformation in mouse models, suppress Hoxa and Meis1, and reduce in vivo spleen-colony-forming ability.

    PMID 23138183DOI
    What this evidence cannot show

    Murine; tool compound. No genome-stability readout is reported alongside the efficacy readout.

    Evidence that argues against this (1)
    • The counterweight is clinical tolerability: the paediatric pinometostat phase 1 study completed, and no excess genotoxicity signal in KMT2A-rearranged children has been reported from it. That does not address CALM-AF10, where the baseline deficit is the concern, but it shows DOT1L inhibition is not overtly genome-destabilising in patients who do not start from a deficit. ClinicalTrials.gov 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CALM-AF10-positive human and murine leukaemic cells have globally reduced H3K79 methylation, are more sensitive to gamma-irradiation, display increased chromosomal instability, and CALM-AF10 patients carry more secondary chromosomal aberrations than patients with other fusion genes such as KMT2A::MLLT3 and PML::RARA.

    PMID 19443658DOI
    What this evidence cannot show

    The patient-level comparison is observational with few CALM-AF10 cases. The causal link from H3K79 hypomethylation to instability is supported by the irradiation experiments but is not proven in patients.

    Evidence that argues against this (1)
    • The link from global H3K79 hypomethylation to chromosomal instability is supported by irradiation sensitivity in cells and by an observational count of secondary aberrations in a small number of patients; it is not a demonstrated causal chain, and CALM-AF10 patients could accumulate aberrations for reasons unrelated to the mark. Lin YH et al. 2009
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Endogenous AF10 directs progressive H3K79 methylation and is required for normal HOX regulation across diverse AML subtypes, and DOT1L/AF10 inhibition is proposed as broadly applicable - which is the same axis normal cells use.

    PMID 25464900DOI
    What this evidence cannot show

    Murine and human AML systems. Used here to establish that the target is not tumour-specific; the normal-haematopoiesis requirement for DOT1L is carried in this graph's existing mechanism-dot1l-normal-haematopoiesis entry.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT against the size of this risk. The on-target concern is that a global H3K79 requirement makes DOT1L inhibition indiscriminately toxic. In the one place it has been tested in humans, that did not dominate: pinometostat given by continuous intravenous infusion to 51 adults did not establish a maximum tolerated dose during dose escalation, and the commonest adverse events of any cause were fatigue (39%), nausea (39%), constipation (35%) and febrile neutropenia (35%) - the profile of a heavily pretreated leukaemia population rather than a specific H3K79-withdrawal syndrome. The risk this entry describes is mechanistically real and was not dose-limiting, and the entry should carry the clinical measurement alongside the mechanism. Stein EM et al. 2018
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

    Both pinometostat phase 1 studies required rearrangement involving the MLL gene including 11q23 translocations or MLL partial tandem duplication, so t(10;11)(p13;q14) PICALM::MLLT10 patients were not eligible.

    NCT02141828
    What this evidence cannot show

    Raw ClinicalTrials.gov v2 API, 2026-09-20. The exclusion is an inference from the cytogenetic band (PICALM at 11q14, KMT2A at 11q23) rather than an explicit exclusion criterion.

    Evidence that argues against this (1)
    • PUBLISHED CONFLICTING RESULT about whether the excluded patients were excluded from the relevant pathway. This risk is framed on deepening an H3K79 deficit, which assumes the drug acts through the canonical H3K79-to-HOXA axis. At achievable, sub-saturating exposures it does not: low-dose pinometostat is cytotoxic far below the concentrations that downregulate HOXA9 and MEIS1, the major route is transcriptional downregulation of FLT3, and overexpression of constitutively active STAT5A largely rescues both the cytotoxicity and the expression changes - a pathway that also depends on MLL1. If that is the operative route, then the harm and the benefit of a DOT1L inhibitor in a CALM-AF10 patient both run through a pathway this entry does not model, and eligibility by cytogenetic band is not what determines whether the drug would help or hurt. Richter WF et al. 2021
Connected to