Two inhibitors of the same DOT1L active site leave different residual H3K79me2 landscapes, and the difference sits at the KMT2A-fusion target loci - so 'how much H3K79me2 is left' is a compound-specific readout, not a generic measure of DOT1L inhibition
also called residual landscape is compound-specific · the HOXA remainder differs by drug · PD thresholds do not transfer between DOT1L inhibitors
Two drugs that block the same enzyme leave different leftovers behind, and the leftovers are concentrated at the very genes the leukaemia depends on. If that is real, a chromatin test used to decide whether one drug is working cannot be used to decide whether another is.
The same thing, for a scientist
**Statement.** At equal times after treatment, EPZ5676 and compound 10 strip H3K79me2 to similar global depths but leave **different residual distributions**: compound 10 leaves the canonical KMT2A-fusion target loci significantly above a baseline-matched expectation at days 6 and 9 in both RS4;11 and MOLM13 (4 of 4 contrasts, p = 0.0012 to <0.0001), while EPZ5676 does so in none of six contrasts. In MOLM13 at day 9 the difference is large in absolute terms - HOXA9 retains 24.5% under compound 10 and 1.8% under EPZ5676. **Origin: data pattern**, from GSE149183, in a single deposit with no replicates. **Why it is not simply potency.** Compound 10 produces a *larger* global collapse than EPZ5676 in RS4;11 (90.6x vs 52.4x) and a *smaller* one in MOLM13 (36.9x vs 63.5x), yet the target sparing appears in both. It does not track the depth of global inhibition in either direction. **Prior probability: 0.2.** Low, because the null explanation - four single libraries sharing a batch - is strong and untestable in this deposit. Raised above a token value because the pattern replicates across two cell lines, two fusion partners and two timepoints, is absent at day 3 in both lines, and is absent from all six comparator-drug contrasts.
Mechanistic routes that would make it true
- Residence-time differences. Two SAM-competitive inhibitors with different off-rates would leave different amounts of catalytically competent DOT1L at loci with high re-deposition demand, and the fusion's target domains are exactly those.
- Differential access to a complexed pool. DOT1L in the fusion-associated complex (AF4/AF9/ENL/P-TEFb) may be less accessible to one chemotype than free DOT1L, which would spare exactly the fusion-bound loci.
- Pharmacokinetics in culture. Compound 10 at 100 nM may be depleted faster than EPZ5676 at 1 uM, giving intermittent inhibition; loci with the highest deposition rate would recover most between troughs - again the fusion targets.
The third of these is the most likely and is the least interesting; it is also the easiest to exclude, by matching exposure rather than nominal concentration. 12
What it would change
If true, then a single-locus H3K79me2 pharmacodynamic assay - which is what a clinical PD sample realistically permits - reports a different number for two compounds at the same true level of target engagement. A PD threshold established for pinometostat would not transfer, and compound selection on a global H3K79me2 endpoint could pick the compound that spares the loci that matter.
It would also put a concrete mechanism under this project's long-standing observation that near-complete H3K79me2 removal did not translate into clinical benefit: what was measured clinically was a global average, and the residual at the loci that carry the dependency may never have been measured at all. 3
| statement | At matched global depth of DOT1L inhibition, different DOT1L inhibitors leave different residual H3K79me2 at KMT2A-fusion target loci, so residual-mark pharmacodynamic readouts are compound-specific. |
|---|---|
| origin | data-pattern |
| prior probability | 0.2 |
predictions
[ "Two DOT1L inhibitors titrated to equal global H3K79me2 loss (spike-in measured) will differ in residual signal at the HOXA cluster by more than the same-condition replicate floor.", "The difference will be larger at loci with the highest baseline H3K79me2 and will grow with time on drug, being absent at the earliest timepoint.", "Continuous exposure that removes any trough between doses will shrink or abolish the difference if the cause is pharmacokinetic." ]
experiments
[ "exp-two-dot1l-inhibitors-at-matched-global-depth-with-replicates-and-a-spike-in" ]
analyses
[ "ana-compound-10-spares-the-canonical-targets-and-epz5676-does-not" ]
falsifiers
[ "In RS4;11, with three replicates per arm and a dm6 spike-in, two DOT1L inhibitors titrated to within 20% of the same global H3K79me2 loss at day 9 give median residual signal across the ten canonical target loci (HOXA5, HOXA7, HOXA9, HOXA10, MEIS1, PBX3, CDK6, MEF2C, MYB, JMJD1C) differing by LESS than 1.5-fold, against a baseline-matched control set. That kills this hypothesis.", "If the compound-10 sparing in GSE149183 fails to appear when the same test is run on the day-6 arm of an independent deposit with a second DOT1L inhibitor, in any KMT2A-rearranged human line, the pattern is a property of these four libraries.", "If matching exposure rather than nominal concentration abolishes the difference, the hypothesis is reduced to a pharmacokinetic artefact and should be withdrawn rather than narrowed." ]
- If residual mark at target loci is what matters, a potency-driven medicinal chemistry campaign optimises the wrong endpoint and may select compounds that spare the leukaemia while stripping normal cells - the worst possible therapeutic ratio.
- If it is not what matters, then the entire premise of H3K79me2 as a pharmacodynamic biomarker is weaker than assumed, because pinometostat achieved deep global reduction and modest responses.
- Either resolution narrows the case for DOT1L inhibition as a monotherapy, which is the direction the clinical data already points.
- What is compound 10? Without a structure this cannot be connected to any chemical series or to any later clinical candidate.
- Is there a second deposit anywhere with two different DOT1L inhibitors and a spike-in in the same experiment? If not, this hypothesis is currently untestable from public data.
- Does the same question have an answer for menin inhibitors? Revumenib, ziftomenib and bleximenib are chemically distinct and no comparison of their residual chromatin landscapes exists.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In GSE149183, canonical KMT2A-fusion target genes retain significantly more H3K79me2 than baseline-matched genes under compound 10 at days 6 and 9 in both cell lines, and never under EPZ5676.
GSE149183 series supplementary files: RS4_11 and MOLM13 H3K79me2 TSS-ChIPseq signal + dm6 spike-in sitepro tables · Perner F, Gadrey JY, Xiong Y, et al. · NCBI Gene Expression Omnibus 2020What this evidence cannot show
No replicates; dose confounded with compound (1 uM vs 100 nM); ten genes; matching on baseline signal only, so gene length and the fixed 4 kb window are uncontrolled and the HOXA cluster is short and tightly packed.
Evidence that argues against this (1)
- EPZ5676 is the compound that shows no sparing here and it is also the one that reached the clinic with only modest activity. If residual mark at fusion targets were the variable that decided efficacy, EPZ5676 should have looked better on that axis and still did not deliver - so the proposed readout has no demonstrated clinical correlate. Stein EM et al. 2018
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Compound 10 is a novel DOT1L inhibitor reported alongside EPZ5676 in the same study, tested at 100 nM against EPZ5676 at 1 uM in the H3K79me2 time course.
Novel inhibitors of the histone methyltransferase DOT1L show potent antileukemic activity in patient-derived xenografts. · Perner F, Gadrey JY, Xiong Y, Hatton C, Eschle BK, Weiss A, et al. · Blood 2020What this evidence cannot show
Concentrations read from GEO sample titles. Compound 10 has no resolvable identifier in ChEMBL or any public chemistry database reachable by this beat, so nothing about its chemotype, selectivity or residence time can be checked.
Evidence that argues against this (1)
- Without the structure, 'different chemotype leaves different residual' is untestable as a chemical hypothesis and remains a statement about two labelled arms of one experiment. NCBI GEO 2020
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
DOT1L-dependent H3K79 methylation at KMT2A-fusion target loci is the dependency DOT1L inhibitors are designed to break.
MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L. · Bernt KM, Zhu N, Sinha AU, Vempati S, et al. · Cancer Cell 2011What this evidence cannot show
Mouse MLL-AF9 genetic model; the inference to human KMT2A::AFF1 is a transfer this project makes constantly and this beat does not test.
Evidence that argues against this (1)
- Treatment-emergent resistance to pinometostat arises in MLL-rearranged cells without restoration of H3K79me2, which argues the dependency can be bypassed and that residual mark is not the only variable. Campbell CT et al. 2017
the observation this hypothesis generalises
the matched-depth replicated design that scores the falsifier
- related to →Pinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field
the clinical-stage compound is the one that leaves no target-selective residual
two inhibitors of one active site, two residual landscapes
- related to →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it
the locus where the compound difference is largest: 24.5% vs 1.8% retained