The missing measurement: is MHC class I bivalent, does a menin inhibitor raise it, and does raising it cost NK control - in a KMT2A-rearranged cell, for the first time
also called class I bivalency CUT&Tag in KMT2A-r · NK degranulation dose-response against class I · the experiment the beat is built to demand
Nobody has ever checked whether the immune identity-card genes are held in the half-on state in this particular leukaemia, whether a menin inhibitor turns them up, or whether turning them up makes a different kind of immune killer worse at its job. One set of experiments on cells this project already uses would answer all three.
The same thing, for a scientist
Three questions this beat could not answer from the literature, in one design on material this graph already holds entries for (`line-sem-kmt2a-aff1`, `line-rs4-11-kmt2a-aff1-adult`, `line-mv4-11-kmt2a-aff1-myeloid`, `line-kopn-8-kmt2a-mllt1-infant-yeats-negative`, plus primary infant material from `panel-infant-all-pdx-kmt2a-aff1-hopkins`). **(1) Are MHC class I promoters bivalent in a KMT2A-rearranged blast?** The entire menin-inhibitor class I inference rests on bivalency demonstrated by ChIP-reChIP in K-562 [E1]. **(2) Does a menin inhibitor raise class I here, and is the raise KMT2A/KMT2B-dependent when one allele is a SET-less fusion?** [E1] **(3) Does raising class I reduce NK degranulation - is the inhibitory brake saturated?** Every published experiment removes or blocks class I; none raises it [E2][E3]. Question 3 is the decision-relevant one, because it is the only one that can turn the beat's headline from a benefit into a trade.
Design
Arm A - is it bivalent. H3K4me3 and H3K27me3 CUT&Tag (and ChIP-reChIP in at least one line, to match the method the premise rests on) at HLA-A, HLA-B, HLA-C, B2M, TAP1, TAP2, NLRC5 and CIITA, in four KMT2A-rearranged lines spanning both lineages and two partners, plus at least two primary infant KMT2A::AFF1 samples. Comparator: K-562, run in the same batch, because that is the cell the premise comes from and a direct comparison is worth more than a literature comparison.
Arm B - does the drug move it. Revumenib and VTP50469 at published exposures, 0-10 days, timecourse at days 0, 2, 4, 7, 10, reading in parallel: surface MHC-I by W6/32 and by an allele-resolved reagent; HLA-A, HLA-B, HLA-C, B2M, NLRC5 and CIITA transcript; the same CUT&Tag panel at day 4. With and without IFN-gamma, because the published effects are largely cytokine-conditioned. Plus a washout arm at day 7, sampled to day 14, which is the measurement the ordering hypothesis needs and which nobody has reported.
Arm C - does it cost NK control. Degranulation (CD107a) and killing by KIR-genotyped healthy-donor NK cells against the same cells at each timecourse point, so that NK output is measured as a function of achieved class I level rather than against an all-or-nothing knockout. Pan-class-I blockade as the positive control for missing-self, replicating the 2012 design 2.
Arm D - the allele question. In one line, degron-tag the wild-type KMT2A allele and repeat Arm B. If class I induction requires the wild-type allele's SET domain, this is where it shows.
What each outcome would mean
| result | consequence for this graph |
|---|---|
| MHC-I promoters are not bivalent in KMT2A-r blasts | the beat's central inference fails; mechanism-mhc-class-i-bivalency-under-menin-inhibition must be re-scoped to 'not applicable to this disease' |
| bivalent, and the drug raises class I | the inference is confirmed for the first time in the index disease; the immune rationale for menin inhibition becomes a finding rather than an extrapolation |
| bivalent, drug does not raise class I | the KMT2A/KMT2B dosage argument is supported and Arm D becomes the story |
| NK degranulation falls as class I rises | risk-raising-mhc-class-i-trades-t-cell-visibility-for-nk-inhibition becomes a quantified trade-off and the post-transplant proposal needs rethinking |
| NK degranulation is already at floor | the trade-off is notional, the risk entry downgrades, and the menin-inhibitor immune rationale is clean |
| class I induction decays within days of washout | hyp-menin-inhibitor-immunotherapy-ordering-is-antigen-dependent gains its key parameter and continuous dosing becomes mandatory |
Every row changes an existing entry. That is the test of whether an experiment is worth proposing in this project.
Controls and known failure modes
- Differentiation confound. Menin inhibitors drive myelomonocytic differentiation and differentiated myeloid cells raise MHC on their own. Read CD11b, CD14 and CD117 alongside, and analyse MHC within differentiation-matched gates. No published menin-inhibitor MHC experiment reports this control and it is the most likely alternative explanation for the whole phenotype.
- Lineage confound. MV4-11 and THP-1 are myelomonocytic and constitutively class II-high; SEM and RS4;11 are B-lymphoid. Both lineages must be present or the result will be a lineage result.
- B2M as the ceiling. Read B2M transcript and, if possible, protein. A heavy-chain rise with a flat B2M is a ceiling, and no published study has looked.
- Antibody specificity, per rule 69. W6/32 reads assembled heavy-chain/B2M complexes; the Bw6 clone REA143 reads a public serological epitope shared across HLA-B alleles and present on some HLA-A alleles. Record the clone, and HLA-type the lines, or the allele-resolved arm is uninterpretable.
- NK donor variation. KIR genotype and HLA of the donor determine the inhibitory repertoire; use at least four genotyped donors and report per-donor.
- This is a cell-line experiment. The primary-sample arm is the one that matters and is the hardest;
panel-infant-all-pdx-kmt2a-aff1-hopkinsandpanel-kocl-japanese-infant-kmt2a-aff1-linesare the routes to it.
| design | Four-arm timecourse in KMT2A-rearranged lines and primary infant material: (A) H3K4me3/H3K27me3 CUT&Tag and ChIP-reChIP at class I and antigen-processing loci to test bivalency, with K-562 as an in-batch comparator; (B) menin-inhibitor timecourse 0-10 days with and without IFN-gamma plus a day-7 washout arm, reading surface class I, allele-resolved transcript, B2M, NLRC5, CIITA and the CUT&Tag panel; (C) CD107a degranulation and killing by KIR-genotyped donor NK cells at each achieved class I level, with pan-class-I blockade as the missing-self positive control; (D) degron on the wild-type KMT2A allele to test whether class I induction needs the SET domain the fusion has lost. |
|---|---|
| readout | bivalency call per locus; class I dose-response curve under drug; NK degranulation as a function of achieved class I level; washout decay constant for induced antigen |
| controls | differentiation-matched gating on CD11b/CD14/CD117 (the control no published menin-inhibitor MHC experiment reports); both lineages represented; B2M read as a possible ceiling; antibody clone recorded and lines HLA-typed because W6/32 and the Bw6 clone REA143 read different things; at least four KIR- and HLA-genotyped NK donors reported individually |
| cost tier | med |
| timeline | 6-9 months for arms A-C in lines; add 6 months for primary material and arm D |
entries this would change
[ "mechanism-mhc-class-i-bivalency-under-menin-inhibition", "risk-raising-mhc-class-i-trades-t-cell-visibility-for-nk-inhibition", "mechanism-mhc-class-i-is-an-nk-brake-in-kmt2a-fusion-leukaemia", "hyp-menin-inhibitor-immunotherapy-ordering-is-antigen-dependent", "analogy-ezh2-driven-mhc-silencing-in-fus-erg-and-cbfa2t3-glis2", "gene-b2m", "gene-hla-a", "gene-hla-b", "gene-nlrc5" ]
material
[ "line-sem-kmt2a-aff1", "line-rs4-11-kmt2a-aff1-adult", "line-mv4-11-kmt2a-aff1-myeloid", "line-kopn-8-kmt2a-mllt1-infant-yeats-negative", "panel-infant-all-pdx-kmt2a-aff1-hopkins" ]
- A negative bivalency result would invalidate the premise of several entries written this cycle, which is the correct behaviour for an experiment and should be stated before it is run rather than after.
- The differentiation confound could account for the entire published MHC phenotype, and if it does, the immune rationale for menin inhibition reduces to 'differentiated cells present antigen better', which is true of any differentiating agent and is not a reason to prefer this one.
- An NK degranulation assay against cell lines with donor NK cells is a weak proxy for in vivo NK control in an infant, and a clean negative in Arm C should not be over-read as safety.
- If the answer to Arm C is that the trade-off is real, the finding argues against the post-transplant maintenance setting that is currently the most advanced clinical proposal in this space - which makes it a result with a constituency against it.
- Should the primary-sample arm come first? The cell-line arms are cheap and the primary arm is the one that would change practice; a two-sample pilot on banked infant material would de-risk the whole design.
- Is a degron on the wild-type KMT2A allele feasible in a KMT2A-rearranged line, given that both alleles carry KMT2A sequence? This graph's `exp-endogenous-tag-and-degron-knock-in-kmt2a-aff1` addresses the same problem for the fusion and its solution should transfer.
- Would an in vivo arm in an NK-competent host be worth the cost? It is the only design that measures the trade-off as a survival outcome rather than as degranulation.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Bivalent modification of MHC class I gene promoters was confirmed by ChIP-reChIP with H3K27me3- and H3K4me3-specific antibodies in K-562; a Menin-KMT2A/B inhibitor activated MHC-I expression dose-dependently and enhanced IFN-gamma-induced surface MHC-I; and combined KMT2A and KMT2B knockout both impaired IFN-gamma-induced MHC-I activation and abolished the increase induced by menin inhibition.
Targeting Menin disrupts the KMT2A/B and polycomb balance to paradoxically activate bivalent genes. · Sparbier CE, Gillespie A, Gomez J, Kumari N, Motazedian A, Chan KL, et al., Dawson MA · Nat Cell Biol 2023What this evidence cannot show
K-562 is the entirety of the bivalency demonstration, and it is a BCR::ABL1 CML line. The KMT2A/KMT2B requirement is established by double knockout, which models neither a heterozygous rearrangement nor a SET-domain-less fusion product. These are the two gaps Arms A and D are designed to fill. Read: full text via Europe PMC fullTextXML (PMC7614190), 2026-09-20.
Evidence that argues against this (1)
- The same paper shows the KMT2A/B requirement can be bypassed entirely once H3K27me3 is removed, with the NLRC5-RFX5 enhanceosome and SETD1A/B taking over, so a negative result in Arm D would not settle the question without an EZH2-inhibitor co-treatment arm. Sparbier CE et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Three KMT2A-rearranged leukaemia lines - RS4;11, SEM and MV4-11 - and primary blasts were resistant to NK-cell lysis with inhibitory KIR ligands, protease inhibitor-9, and low NKG2D activating ligands and adhesion molecules; adding a pan-MHC class I antibody to block inhibitory KIR-HLA interaction further augmented NK degranulation in all KIR2DL1, KIR2DL2/3 and KIR3DL1 subsets.
Antibody-dependent cell-mediated cytotoxicity overcomes NK cell resistance in MLL-rearranged leukemia expressing inhibitory KIR ligands but not activating ligands. · Chan WK, Kung Sutherland M, Li Y, Zalevsky J, Schell S, Leung W · Clin Cancer Res 2012What this evidence cannot show
This is the design Arm C replicates and extends: the 2012 experiment blocks class I and measures the increase in degranulation, which establishes that the brake is engaged at baseline but not whether it can be engaged further. The reagents are fourteen years old and the lines are exactly the three this graph already carries. Read: abstract only; identifiers fetched 2026-09-20.
Evidence that argues against this (1)
- If baseline class I already saturates inhibitory KIR signalling, blockade would still increase degranulation (removing a saturated signal has a large effect) while raising class I would have none. The 2012 result is therefore fully compatible with both outcomes of Arm C, which is precisely why Arm C is needed. Chan WK et al. 2012
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A classical MHC class I molecule, H2-K1, is a top in vivo dependency of MLL-AF9 leukaemia-initiating cells; its disruption depletes leukaemia in vivo and extends survival, and the effect is completely rescued by NK-cell depletion.
H2-K1 protects murine MLL-AF9 leukemia stem cells from natural killer cell-mediated immune surveillance. · Ghosh S, Rodriguez-Zabala M, Dushime GT, Reinbach K, Ramakrishnan R, Sitnicka E, Jaras M · Haematologica 2025What this evidence cannot show
Murine and myeloid; a loss-of-function experiment, which is the opposite manipulation from the one the therapeutic question requires. Read: full text via the PMC HTML page (PMC12130766), 2026-09-20, the PubMed abstract field being empty.
Evidence that argues against this (1)
- A murine in vivo dependency cannot be reproduced in the human cell-line assay proposed here; Arm C measures degranulation against donor NK cells in vitro, which is a proxy. The definitive version of this experiment is a KMT2A-rearranged model in an NK-competent host with class I titrated pharmacologically, and that is a larger undertaking. Ghosh S et al. 2025
the washout arm supplies the decay constant the ordering hypothesis needs
tests bivalency and drug response in a KMT2A-rearranged cell for the first time
arm C measures NK degranulation as a function of achieved class I level
index material for the bivalency and NK arms
companion design: class I plus NK ligands here, class II headroom there; the two panels should be run on the same samples