Hypothesis: a defined subset of fusion binding sites survives saturating menin inhibition on menin-independent tethering, and that residual set is where bypass resistance comes from
The drug works by prising menin off the DNA and taking the fusion protein with it. But the fusion has other ways of holding on, so some sites should stay occupied no matter how much drug you give. We think those leftover sites are exactly the ones that let the leukemia escape.
The same thing, for a scientist
Apparent authority survives revocation. Menin inhibitors act by evicting menin and MLL1 from chromatin, and resistance mutants work by preventing that eviction [E1]. But the fusion holds chromatin by at least three routes that do not run through menin: CXXC-domain binding to unmethylated CpG, LEDGF/PSIP1 tethering via the PWWP-H3K36me nucleosome interaction, and the AFF1-side SEC contacts. We hypothesise a reproducible, mappable residual occupancy set that persists at saturating drug, that this set is enriched for the non-canonical menin targets bivalently marked by menin and H2AK119ub whose de-repression through PRC1.1 loss reactivates MYC [E2], and that the size of this set at baseline predicts depth and duration of clinical response. The recently reported redistribution of JUND onto chromatin under menin inhibition [E3] shows the drug does not simply subtract occupancy — it reshuffles it — which is a direct observation in the same spirit and increases our prior.
Why this is not the same as the cell-state-plasticity hypothesis
hyp-cell-state-plasticity-limits-interface-disruption predicts that the fusion IS evicted and the program is maintained anyway, through a changed cell state. This hypothesis predicts the opposite mechanism at the same clinical endpoint: the fusion is NOT fully evicted, at a specific and mappable subset of sites, and that residual occupancy is the substrate for bypass. The two are discriminable by one experiment — CUT&RUN for the fusion under saturating drug. Whichever way the result falls, one of the two hypotheses loses, which is the main reason this one is worth writing down.
The three menin-independent tethers, ranked by how likely they are to matter
- LEDGF/PSIP1. The IBD binds a composite surface formed by menin and KMT2A jointly, but the PWWP domain reads H3K36me on nucleosomes independently. Whether the IBD contact survives menin displacement is the crux and is structurally unresolved.
- CXXC-DNA. Retained in every leukemogenic fusion and binds unmethylated CpG directly. Sequence-specific and menin-independent by construction; the open question is whether it is sufficient for stable occupancy or only for local search.
- AFF1-side SEC contacts. The 3' partner brings its own affinity for the elongation machinery, which could hold the complex at actively transcribed loci through a route the drug never touches.
Honest prior and the strongest objection
Prior 0.5, which is deliberately near the middle because the existing evidence points both ways. In favour: the drug is known to work by eviction, eviction is reported as a property of sensitive cells rather than as a complete phenomenon, and there are at least three tethers it does not engage. Against: if substantial residual occupancy existed at important sites, the drug would probably not produce complete remissions at all, and it does produce them in about 22.8% of heavily pretreated patients 4. The reconciliation the hypothesis requires is that the residual set is small, site-specific, and matters only for bypass rather than for the canonical HOXA program — which is a real prediction and not a hedge, because it says exactly where to look.
| statement | At saturating menin-inhibitor concentration a reproducible subset of KMT2A::AFF1 chromatin binding sites retains fusion occupancy through menin-independent tethering (CXXC-DNA, LEDGF/PSIP1 PWWP, AFF1-SEC contacts); this residual set is significantly enriched for non-canonical, bivalently marked menin target loci including MYC regulatory elements; and residual-set size at baseline correlates inversely with depth and duration of response. |
|---|---|
| origin | analogy |
| source analogy | analogy-law-agency-authority-revocation |
| prior probability | 0.3 |
| discriminating against | hyp-cell-state-plasticity-limits-interface-disruption, which predicts complete eviction with maintained output; a single CUT&RUN experiment separates the two. |
| decision value | If a residual set exists and is mappable, it names the specific loci and the specific tether that a second agent must break, which converts a vague call for combination therapy into a target. |
predictions
[ "CUT&RUN for the KMT2A N-terminus under saturating revumenib retains a defined peak set rather than showing uniform loss, reproducible across at least two KMT2A::AFF1 models.", "The residual set is enriched for loci carrying H2AK119ub together with menin, the non-canonical class identified in the PRC1.1 screens.", "Deleting the CXXC DNA-binding function, or disrupting the LEDGF IBD contact, shrinks the residual set — identifying which tether holds it.", "Cells with a larger baseline residual set are intrinsically less sensitive, and PRC1.1-deficient cells show an expanded residual set.", "Adding an agent that evicts the fusion by an independent route (KAT6/KAT7 catalytic inhibition) collapses the residual set where a second menin inhibitor does not." ]
experiments
[ "exp-residual-occupancy-cutrun-saturating-menin" ]
analyses
[ "ana-menin-resistance-cell-state-signature" ]
falsifiers
[ "CUT&RUN under saturating drug shows essentially complete, uniform loss of occupancy BY ANY OF the fusion, ENL or SEC members, with no reproducible residual peak set. (Revised 2026-09-19: the original wording named only the fusion, and was arguably met by PMID 39472576 while the underlying idea survived - a falsifier that fires on the wrong molecule is the wrong falsifier.)", "A residual set exists but is random across replicates and models, meaning it is technical background rather than biology.", "The residual set exists but shows no enrichment for non-canonical or bivalent loci and no relationship to bypass, making it real but irrelevant.", "Cells with larger residual sets are no less drug-sensitive than cells with smaller ones.", "A residual set exists and is occupied by ENL, but ENL occupancy there is present before drug and unchanged by it, making it a constitutive feature rather than a response to inhibition." ]
prior revision 2026 09 19
{
"from": 0.5,
"to": 0.3,
"reason": "Falsifier [0] partially met by PMID 39472576 - fusion and menin undetectable, ENL retained - in the only human sample under revumenib. Not collapsed to near-zero because that sample is lineage-switched and switching independently lowers oncoprotein levels, so the observation is confounded for the question asked."
}predicted occupant set
{
"was": "the fusion protein itself",
"now": "the fusion, ENL, or another SEC/DOT1L-complex member. The locus set is the claim; the occupant is now an open variable and ENL is the leading candidate.",
"source_pmid": "39472576"
}- If residual occupancy is real, it is a hard ceiling on every menin-interface compound regardless of potency, and improving affinity will not help.
- The same logic applies to every obligate-partner interface the lab might pursue: a protein held to chromatin by three tethers cannot be removed by blocking one.
- Menin inhibition increasing JUND chromatin occupancy means the drug has an unmeasured gain-of-function arm in patients, which could contribute to both resistance and toxicity and is not monitored anywhere.
- Does the LEDGF IBD contact survive menin displacement? The IBD binds a composite surface formed by both menin and KMT2A, and whether half of that surface is sufficient has never been measured.
- Is the residual set shared between KMT2A::AFF1 and NPM1-mutant leukemia, which depend on the same adaptor through different principals?
- Now that menin inhibition is known to redistribute JUND onto chromatin, how much of the transcriptional response attributed to fusion eviction is actually JUND gain of function?
LEDGF/PSIP1 cannot be a menin-independent tether: menin is the adaptor that physically links MLL to LEDGF (Yokoyama & Cleary 2008), and the LEDGF surface is formed jointly by menin and MLL1 (Huang 2012)
- ← related toKAT7 / HBO1 — an enzyme, a leukaemia-stem-cell dependency, and an inhibitor already in a crystal
competing mechanism for residual fusion retention under menin inhibition: KAT6/KAT7 catalytic inhibition rapidly evicts MLL-FP from chromatin and overcomes primary and acquired menin-inhibitor resistance
generated by apparent authority surviving revocation
predicts a ceiling on what blocking this one interface can achieve
LEDGF tethering is one of the three candidate menin-independent tethers
- depends on →Step 1 — How the fusion finds its genes: AT-hooks and the CXXC domain reading unmethylated CpG
CXXC-DNA binding is the second candidate menin-independent tether
predicts the residual occupancy set is enriched for the non-canonical bivalent targets that drive MYC bypass
- refutes →Hypothesis: the ceiling on menin-interface disruption is cell-state plasticity, not target mutation
predicts incomplete eviction where the plasticity hypothesis predicts complete eviction with maintained output
- ← testsExperiment: map what stays on chromatin when the menin inhibitor has done everything it can do
saturation-ladder CUT&RUN with genetic dissection of the three tethers
predicts a residual-occupancy ceiling that better potency cannot fix
redistribution to lower-affinity sites is one of five unresolved fates
the residual set that hypothesis predicts is already reported, and it is the majority rather than a subset
its opening sentence transfers eviction to the fusion; correction owed, with PMID 36922589 and 31821784 as the sources
a stronger form of the same intuition: the residual is the majority, not a subset
direct evidence that menin inhibition redistributes chromatin occupancy rather than only subtracting it
- ← refutesENL stays after the fusion leaves — an epigenetic lesion that outlives the protein that made it
the only human sample profiled under revumenib shows the fusion undetectable and a cofactor retained, not retained fusion occupancy