The MLL3/4-UTX switch fails to throw — menin inhibition kills by turning something ON, and cells that cannot turn it on survive

The received account of menin inhibition is subtractive: evict menin, evict the fusion, HOXA9 and MEIS1 fall, the differentiation block lifts. 1 shows that account is incomplete in a way that matters for resistance. MLL1-menin does not only activate the oncogenic program — it also impedes MLL3/4-UTX from reaching a subset of promoters. When the drug displaces menin, MLL3/4-UTX moves in and activates a tumour-suppressive program, and it is that activation which dictates therapeutic response in murine and human leukaemia 1.

So the drug has two arms, one subtractive and one additive, and only the subtractive one is routinely measured. A cell can therefore show a textbook on-target response — HOXA9 down, MEIS1 down, differentiation markers up — and still be resistant, because the arm that actually kills it never fired.

This explains an otherwise puzzling observation that already sits in this graph: in resistant, multiply pretreated KMT2A::AFF1 patient-derived xenografts, transcriptomic analysis documented sustained on-target efficacy against the canonical menin-inhibitor targets 2. On the subtractive account that is a contradiction. On the switch account it is what resistance looks like.

If the additive arm runs through MLL3/4-UTX, then loss of KMT2C (MLL3), KMT2D (MLL4) or KDM6A (UTX) should confer resistance. Two independent lines of evidence support this:

• 2 states that inactivation of KMT2C/D has been reported to result in menin-inhibitor resistance, that downregulation of KMT2D may account for resistance in one of their two patients, and — the strongest observation in that paper for this mechanism — that KMT2C-edited cells from that patient were selected for in menin-inhibitor-containing growth conditions. That is a selection experiment, not an association.
• In a non-haematopoietic system, Kmt2c/d knockout caused redistribution of KMT2A-menin away from KMT2D-localised enhancers onto CpG-high and bivalent promoters, with derepression of signal-induced immediate early genes 3. This is urothelium, not leukaemia, and the phenotypic endpoint is tumorigenesis rather than drug response — but it is direct, orthogonal evidence that the MLL3/4 arm and the KMT2A-menin arm compete for the same chromatin, which is the physical premise of the switch.

KMT2C and KMT2D are among the most frequently mutated genes in human cancer and are recurrently altered in relapsed leukaemia, so the lesion is not exotic.

1 reports that therapeutic reactivation of the tumour-suppressive program using CDK4/6 inhibitors mitigates treatment resistance in leukaemia cells insensitive to menin inhibitors. This is worth dwelling on, because CDK4/6 inhibitors (palbociclib, ribociclib, abemaciclib) are licensed, orally available, and have paediatric experience in other indications — and they are not on anybody's list of menin-inhibitor partners. The trials now running pair menin inhibitors with venetoclax, hypomethylating agents, FLT3 inhibitors and chemotherapy. A menin-plus-CDK4/6 combination follows from a published mechanism, uses two approved drug classes, and as far as this beat could determine is not being tested in KMT2A-rearranged leukaemia.

resistance-adaptive-nongenetic-menin already lists 'KMT2C/KMT2D inactivation or downregulation' as one of five named routes inside a broader non-genetic entry. That listing is correct and this entry does not overturn it; it does the thing the granularity rule asks for, which is to give the route its own entry with its own mechanism, its own primary source, and its own countermeasure. The important difference is that KMT2C/D loss is genetic, and the entry that currently houses it is about escape that leaves the genome unchanged. Filing a recurrent loss-of-function lesion under 'invisible to mutation panels' is exactly the kind of mismatch that causes a reader to stop looking for it on a panel where it would in fact show up.

This entry reads the KMT2C/KMT2D and UTX findings as selection under menin-inhibitor pressure. PMID 41027490 - the entry's own source - says the opposite, in its abstract, and I fetched and read it back on 2026-09-20:

"Because none of the patients had been treated with a menin inhibitor, resistance in these highly pretreated samples was acquired in the absence of menin-inhibitor exposure."

So what the paper describes is upfront resistance in multiply pretreated disease, not resistance selected by the drug. The entry's title and framing survive - the lesions are associated with reduced menin-inhibitor sensitivity - but the causal direction does not, and it was the load-bearing part.

There is a published alternative that fits the observation better. PMID 33596429 (Chen, Cell Reports 2021, fetched and read back) shows that haploid Kmt2c deletions convey a selective advantage only when HSCs are driven into cycle by a strong proliferative stimulus, such as chemotherapy, and offers that as the reason Kmt2c deletions are commoner in therapy-related MDS/AML than in de novo disease. In multiply pretreated patients a KMT2C-deficient clone is expected from the chemotherapy alone, which confounds any reading that attributes it to the menin axis.

And the same source records a second thing this entry should carry: the resistant samples showed "the emergence of multiple comutations, including RAS pathway and TP53 mutations, although neither was sufficient to induce menin-inhibitor resistance in vitro" - a clean negative on two obvious candidate mechanisms.

Editorial note on the source text. The abstract of PMID 41027490 says "KMT3D", which is not a gene. It is almost certainly KMT2D, and this graph reads it as such - but the substitution is made explicitly here rather than silently, because a graph that silently repairs its sources cannot be audited against them.