Oncofusion

start / graph / exp-isomer-resolved-arginine-marks-at-kmt2a-fusion-targets

Experimentmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q103-arginine-marks

Measure H3R2 and H4R3 methylation states at KMT2A-fusion targets by isomer-resolving mass spectrometry, and test the gate by writer perturbation read out on the enzyme's product - the experiment that has never been done in any fusion

also called the arginine-mark measurement in KMT2A-rearranged leukaemia

Nobody has ever measured these arginine tags in a cell carrying a KMT2A fusion, in any fusion, in any species. This is the cheapest way to find out whether they are there, and a second step that tests what they do without relying on antibodies that cannot tell the two versions apart.

The same thing, for a scientist

Across all four H3R2/H4R3 dimethyl states, exactly **two** records in the literature co-occur an arginine mark with KMT2A in a title or abstract, and neither is locus-resolved at a fusion target: one is wild-type KMT2A in murine neural differentiation [E1], the other a bulk H4R3me2s reduction in murine MLL-AF9 downstream of *Cdc73* excision and *Prmt5* transcription [E2]. All four marks return **zero** against menin and zero against MLL-AF4 [E3]. **Arm 1** is therefore purely descriptive and needs no new reagent: quantify all H3R2 and H4R3 methylation states, both dimethyl isomers resolved, by mass spectrometry on acid-extracted histones and on fractionated chromatin from human KMT2A::AFF1 lines (SEM, RS4;11) and a KMT2A::MLLT3 line (MOLM-13 or THP-1) against non-rearranged B-ALL and AML comparators. **Arm 2 tests the gate while sidestepping the antibody problem entirely**, by copying the design that produced the KMT2D trans-tail result [E4]: perturb the writers - PRMT6, PRMT5, PRMT7 - and read the **downstream output**, KMT2-complex occupancy and H3K4me3 at HOXA9/MEIS1, rather than the arginine mark. **Arm 3** is the allele-asymmetry test the mechanism uniquely predicts: because the fusion lacks the SET domain, the WIN motif and the PHD1-3 cassette, a substrate-side gate should restrain the **wild-type** allele and leave the fusion's menin/CXXC-driven occupancy untouched. **Arm 1 alone can close the axis**: if the marks are absent or invariant, no further work is warranted.

Go deeper
Why arm 1 comes first, and why it is cheap

Everything on this axis currently rests on transfer from other systems - neural differentiation, embryonal carcinoma, erythroid globin, bladder cancer, HeLa. This graph has measured that 0 of 24 sampled load-bearing experiments in the corpus rest on a dedicated human KMT2A::AFF1 experiment, and the arginine axis is a pure case: not one measurement, anywhere, in any fusion.

Arm 1 requires no antibody, no new cell line, no perturbation and no genomics. Acid-extract histones from existing lines, propionylate, digest, and run LC-MS/MS with a chromatographic separation that resolves the dimethylarginine isomers. The output is a table of H3R2me0/me1/me2a/me2s and H4R3me0/me1/me2a/me2s stoichiometries per line.

Three outcomes and what each licenses:

resultinterpretationnext step
both residues essentially unmethylated in all linesthe marks are not a feature of this chromatinCLOSE THE AXIS. One table retires four node entries' worth of speculation
present and similar across rearranged and non-rearranged linesthe marks exist but carry no fusion-associated signal in bulkproceed to arm 2 only if a locus-resolved hypothesis is specified; bulk is uninformative
present and differing by fusion statusa fusion-associated arginine-methylation state existsproceed to arms 2 and 3

Note the asymmetry deliberately built in: the cheapest arm is the one most likely to kill the hypothesis. That is the point of running it first, and it is the opposite of the usual order in which the confirmatory experiment goes first.

Arm 2: test the gate without reading the mark

The measurement constraint on this axis is absolute and is recorded as its own entry: me2a and me2s are isobaric with opposite consequences, so an antibody that cannot separate them returns no information, and a perturbation that converts one state into the other reads as no change.

The design that evades it is already published. PMID 23249737 established the H4R3 gate on KMT2D by perturbing the writers (PRMT7 versus PRMT5 knockdown) and reading the enzyme's own catalytic product, H3K4me3 4. Applied here:

  • Perturbations, one at a time, in a t(4;11) line: PRMT6 knockdown or knockout (writer of the H3R2 block); PRMT5 (writer of H3R2me2s and of H4R3me2s at some loci); PRMT7 (writer of the H4R3 block); plus the PRMT7 chemical probe SGC3027 as a pharmacological arm with an independent time course 5.
  • Readout, and this is the part that matters: KMT2A and WDR5 occupancy plus H3K4me3 at HOXA9, MEIS1 and a menin-inhibitor-responsive gene set, with expression-matched control loci. Not the arginine mark.
  • Directional predictions, each falsifiable: PRMT6 loss removes the H3R2 block and should raise WDR5 occupancy and H3K4me3 at loci where the block was operating; PRMT7 loss removes the H4R3 block and should raise H3K4me3 deposited by a PHD-competent KMT2 enzyme; PRMT5 loss should do neither at KMT2D targets, which is the published negative control built into the original design 4.
  • Kill criterion: no change in H3K4me3 or WDR5 occupancy at any tested locus under any of the three writer perturbations, with confirmed protein loss. That would mean the gate does not operate in this chromatin regardless of whether the marks are present.
Arm 3: the allele asymmetry, which only this mechanism predicts

This is the arm that distinguishes the arginine gate from every other mechanism in this graph's WRAD arm, and it is a within-cell control, so it needs no cross-line normalisation.

The logic. The gate acts on the histone tail, not on the KMT2A polypeptide. KMT2A::AFF1 retains the AT-hooks, CXXC and menin-binding motifs and has lost PHD1-3 (Q03164 aa 1431-1627), the bromodomain (1635-1765), FYRN, the TAD and SET 6. So:

  • The wild-type allele is SET-competent, WIN-competent and PHD-competent - every module the gate could act through is present.
  • The fusion has none of them, and is targeted by menin plus CXXC.

Prediction: a PRMT6 perturbation should change wild-type KMT2A occupancy and H3K4me3 while leaving fusion occupancy unchanged. PMID 30232013 reports the wild-type half of exactly this in a non-leukaemic model - loss of promoter H3R2me2a enhances KMT2A binding and H3K4me3 1.

Feasibility. Allele discrimination is the hard part and is solved in this graph's own literature: the antibody used in the murine MLL-AF9 WDR5 study detects only the 180 kDa MLL1 C-terminal fragment and therefore cannot see the fusion, which makes it a wild-type-specific reagent; and a C-terminal-tagged endogenous allele, or a fusion-specific reagent against the partner moiety, gives the other half. Running both in the same chromatin under one perturbation is the experiment.

What a positive result would and would not mean. It would establish a substrate-side control point on the wild-type allele in a fusion-driven cell - which matters, because this graph already records that a menin-independent function of wild-type KMT2A/B maintains H3K4me3 at bivalent genes and opposes polycomb, and that menin inhibition redistributes KMT2A. It would not be evidence against menin dependence of the fusion, and any entry citing it that way would be misreading it.

Facts and figures
designTHREE ARMS, run in order, with arm 1 gating the rest. ARM 1 (descriptive, no antibody, no perturbation): acid-extract histones and separately prepare fractionated chromatin from human KMT2A::AFF1 lines (SEM, RS4;11), a KMT2A::MLLT3 line (MOLM-13 or THP-1), and non-rearranged B-ALL and AML comparators; propionylate, digest, and quantify by LC-MS/MS with a chromatographic separation that RESOLVES the dimethylarginine isomers, reporting stoichiometry for H3R2me0/me1/me2a/me2s and H4R3me0/me1/me2a/me2s per line. ARM 2 (functional, evades the antibody problem by design, copying PMID 23249737): single-gene knockdown or knockout of PRMT6, PRMT5 and PRMT7 in one t(4;11) line, plus a pharmacological PRMT7 arm with SGC3027 on an independent time course; read the DOWNSTREAM OUTPUT, not the arginine mark. ARM 3 (allele asymmetry, within-cell): under a PRMT6 perturbation, measure wild-type KMT2A and fusion occupancy separately - the wild-type-specific route is a C-terminal KMT2A reagent that cannot detect the fusion, the fusion-specific route is a reagent against the partner moiety or an epitope-tagged endogenous fusion allele.
readoutARM 1: per-line stoichiometry table of all four methylation states at H3R2 and at H4R3, both dimethyl isomers resolved, in bulk histone and in chromatin fractions. ARM 2: KMT2A and WDR5 occupancy plus H3K4me3 by CUT&RUN or calibrated ChIP at HOXA9, MEIS1 and a menin-inhibitor-responsive gene set, with promoter and enhancer classes reported SEPARATELY (because PMID 30232013 finds opposite signs in the two compartments), against expression-matched control loci; plus HOXA9/MEIS1 transcript levels. ARM 3: allele-resolved KMT2A occupancy at the same loci under the same perturbation.
controlsConfirmed protein loss for every writer perturbation by western. PRMT5 arm retained as the PUBLISHED NEGATIVE for the H4R3/KMT2D gate (PMID 23249737 reports PRMT7 but not PRMT5). Expression-matched control loci for all occupancy measurements. Promoter versus enhancer stratification, not averaged. For the pharmacological arm, an independent target-engagement control, because SGC3027's validated pharmacodynamic is HSP70 monomethylation and NOT a histone mark, plus a timepoint earlier than any viability effect because PRMT7 inhibition sensitises to proteotoxic stress. Non-rearranged comparator lines in arm 1. Breakpoint of each fusion line stated, because a fusion retaining PHD1-3 would abolish arm 3's predicted asymmetry. NO ARGININE-MARK ANTIBODY is used as a primary readout anywhere in this design; if one is used at all, its clone and lot must first be shown to discriminate the isomers by peptide microarray or isomer-defined nucleosome standards.
cost tierlow
timelineARM 1: 4-8 weeks in a lab with histone-PTM mass spectrometry, using cell lines already held in most KMT2A laboratories, and it can kill the axis on its own. ARM 2: 3-4 months including knockdown line generation. ARM 3: 4-6 months, dependent on establishing allele-resolved detection, and worth starting only if arms 1 and 2 are positive.
decidesmechanism-arginine-methylation-gates-kmt2-access-independently-of-menin-and-the-win-motif
kill criterionARM 1 kills the axis if H3R2 and H4R3 are essentially unmethylated across all lines, or methylated indistinguishably between KMT2A-rearranged and non-rearranged lines with no chromatin-fraction difference. ARM 2 kills the gate if none of the three writer perturbations changes H3K4me3 or WDR5 occupancy at any tested locus with confirmed protein loss. ARM 3 kills the substrate-side model specifically if a PRMT6 perturbation moves FUSION occupancy as much as wild-type occupancy, since the fusion has no SET domain, no WIN motif and no PHD1-3 for a substrate-side gate to act through.
why nobody has done itAcross all four dimethyl states, two records in the whole literature co-occur an arginine mark with KMT2A in a title or abstract, and neither is locus-resolved at a fusion target; all four return zero against menin and zero against MLL-AF4 (2026-09-22, translations read back, zeros re-run). The arginine-methylation field and the KMT2A-fusion field have not intersected. NOTE THE BOUND: PubMed titles and abstracts only - a histone-MS supplementary table in an existing fusion paper would make arm 1 redundant, and that should be checked by full-text search before the experiment is started.
measurement constraintBINDING. me2a and me2s are isobaric with OPPOSITE consequences for KMT2-complex access, so a non-discriminating antibody returns the sum of an inhibitory and a permissive mark, and a state conversion reads as no change. This design uses mass spectrometry for the descriptive arm and downstream catalytic output for the functional arm precisely so that no conclusion depends on an arginine-mark antibody. See concept-an-assay-that-cannot-separate-me2a-from-me2s-cannot-test-the-arginine-switch.
what a positive result would NOT meanIt would not be evidence against menin dependence of the fusion. The gate acts on the histone substrate and is downstream of, and orthogonal to, the menin/CXXC targeting the drug class exploits. A positive arm 3 in particular would establish a substrate-side control point on the WILD-TYPE allele in a fusion-driven cell, which is interesting on its own terms - this graph already records a menin-independent function of wild-type KMT2A/B maintaining H3K4me3 at bivalent genes - and is not a statement about the fusion.
How this could bite us back
  • A positive result points at writer inhibition, and both blocking states are written by enzymes that are NOT DepMap dependencies - PRMT6 (mean Chronos -0.0666) and PRMT7 (mean -0.1513). If the blocks are brakes on a HOXA program, inhibiting their writers is the pro-leukaemic direction, so a mechanistically successful experiment could produce a therapeutically inverted conclusion. The experiment should be designed to detect that sign, not just the existence of an effect.
  • The pharmacological arm carries an on-target confound that is easy to misread: PRMT7 inhibition reduces tolerance for proteotoxic stress, so an apparent anti-leukaemic effect of SGC3027 in a stressed culture may be proteostasis rather than chromatin. Reading the arm on chromatin endpoints before any viability change is not optional.
  • Arm 1 is the arm most likely to close the axis, which means the honest version of this plan risks retiring four node entries written in the same cycle as the plan. That is the correct order and it is recorded here so a later reader can see the axis was not protected from its own cheapest test.
Still open
  1. Does a histone mass-spectrometry dataset from a KMT2A-rearranged line already exist with the dimethylarginine isomers resolved? If so arm 1 is a reanalysis rather than an experiment, and the whole plan starts at arm 2. A Europe PMC full-text and supplementary-table search is the cheap first move and was not run in this beat.
  2. Which chromatographic method resolves H3R2me2a from H3R2me2s at the peptide level with usable sensitivity on a limited chromatin fraction? The design assumes this is solved; if it is not, arm 1's cost tier moves from low to high and the plan should invert to lead with arm 2, which needs no isomer resolution at all.
  3. Can wild-type KMT2A and the fusion be distinguished at chromatin in the same experiment reliably enough for arm 3? A C-terminal reagent that cannot see the fusion gives one half, and this graph records such a reagent being used in murine MLL-AF9; the fusion-specific half is harder and may require an engineered tagged allele, which changes arm 3's timeline substantially.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a murine neural-differentiation model with PRMT6 knockout, loss of promoter H3R2me2a leads to enhanced KMT2A binding and increased H3K4me3 deposition with increased target-gene transcription, while loss at enhancers reduces KMT2D binding and H3K4me1/H3K27ac - and this is the ONLY record returned by '"H3R2me2a"[tiab] AND KMT2A[tiab]'.

    PMID 30232013DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Murine neural differentiation, wild-type alleles, no fusion protein present - so it supplies the prediction for arm 3's wild-type half and nothing about a fusion. The ChIP is antibody-based and the PRMT6 knockout removes the writer rather than the mark, so indirect effects through PRMT6's non-histone substrates are not excluded.

    Evidence that argues against this (1)
    • The same paper finds the opposite sign at enhancers, so an arm-3 readout restricted to promoters could report a direction that does not hold genome-wide. The experiment must therefore report promoter and enhancer classes separately or it will produce a compartment-averaged answer that means nothing. Guccione E et al. 2007
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In murine MLL-AF9 leukaemia, Prmt5 is a direct PAF-complex target positively regulated by a transcriptional unit including the PAFc, MLL1, HOXA9 and STAT5, and PRMT5-mediated H4R3me2s was reduced following conditional excision of Cdc73 - the only arginine-mark measurement reported in any KMT2A-fusion system.

    PMID 28945229DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. Murine MLL-AF9, myeloid, not the index fusion. The mark is reported as a bulk reduction, NOT at named fusion target loci, and the causal chain runs PAFc to Prmt5 transcription to PRMT5 protein to the mark - a writer-abundance result rather than a demonstration that the fusion controls the mark on chromatin. The antibody and its isomer specificity are not stated in the abstract.

    Evidence that argues against this (1)
    • Because the reduction follows the writer's gene being downregulated, the result is equally consistent with the mark being an inert bystander of PRMT5 abundance - which is precisely why this experiment's arm 2 reads the downstream KMT2 output under writer perturbation rather than the mark itself. Dhar SS et al. 2012
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed esearch 2026-09-22, translations read back and every zero re-run identical: across all four H3R2/H4R3 dimethyl states, only two records co-occur with KMT2A in a title or abstract ('"H3R2me2a" AND KMT2A' = 1, PMID 30232013; '"H4R3me2s" AND KMT2A' = 1, PMID 28945229); '"H3R2me2s" AND KMT2A' = 0, '"H4R3me2a" AND KMT2A' = 0 and AND MLL = 0; all four marks AND menin = 0; '"H3R2me2a" AND revumenib' = 0; '"H3R2me2a" AND "MLL-AF4"' = 0. Operand control KMT2A[tiab] = 1652.

    What this evidence cannot show

    Bounds TITLES AND ABSTRACTS only; no Europe PMC full-text sweep was run, and this project has a measured precedent of a histone-mark measurement hiding in a supporting figure. Two symbol artefacts were measured and must travel with these numbers: H4R3[tiab] AND KMT2A[tiab] = 0 while H4R3[tiab] AND MLL[tiab] = 3, and '"H4R3me2s" AND "MLL-AF9"' = 0 although the one KMT2A record IS an MLL-AF9 study. Before this experiment is funded, the founding KMT2A-fusion chromatin papers' supplementary tables should be searched for arginine marks.

    Evidence that argues against this (1)
    • A quantitative histone mass-spectrometry supplementary table in a KMT2A-fusion paper would report both dimethyl isomers and would be invisible to every query here. If such a table exists, arm 1 is already done and the experiment should start at arm 2 - which is a cheaper outcome and a reason to check before starting. Van Vlierberghe P et al. 2008
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The KMT2D trans-tail gate was established by perturbing the arginine writers and reading the enzyme's own catalytic product: PRMT7 but not PRMT5 represses MLL4/KMT2D target genes by upregulating H4R3me2s, PRMT7 knockdown increases MLL4-catalysed H3K4me3, PHD(4-6) binding is inhibited by H4R3me2s while unmethylated or asymmetrically dimethylated H4R3 is recognised, and PHD(4-6) is required for MLL4's nucleosomal methyltransferase activity.

    PMID 23249737DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. NOMENCLATURE: 'MLL4' is KMT2D, settled by the MLL2/ALR aliases and the Kabuki-syndrome attribution in the abstract. NT2/D1 embryonal carcinoma differentiation, not haematopoietic and not leukaemic; the enzyme is KMT2D not KMT2A. This item is cited for the DESIGN PRINCIPLE - read the catalytic output under writer perturbation - as much as for the biology.

    Evidence that argues against this (1)
    • The design has a built-in ambiguity: a writer knockdown changes that enzyme's non-histone substrates too, so a change in H3K4me3 could be indirect. Mitigating it requires the PRMT5 negative arm (which the original study had and which must be kept) plus, ideally, a catalytically dead rescue - and that rescue has not been done for any of these enzymes in a leukaemic cell. Szewczyk MM et al. 2020
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    SGC3027 is a potent, selective, cell-active chemical probe for PRMT7 - a cell-permeable prodrug converted in cells to the SAM-competitive inhibitor SGC8158 - whose validated cellular pharmacodynamic is loss of arginine monomethylation of HSP70-family stress proteins, and whose use decreases tolerance for proteostasis perturbations including heat shock and proteasome inhibitors.

    Pharmacological inhibition of PRMT7 links arginine monomethylation to the cellular stress response. · Szewczyk MM, Ishikawa Y, Organ S, Sakai N, et al. · Nat Commun 2020
    PMID 32409666DOI
    What this evidence cannot show

    Abstract read from the eutils record 2026-09-22; full text not read. A chemical probe, not a clinical agent. IMPORTANT FOR THIS DESIGN: the probe's validated target-engagement readout is HSP70 MONOmethylation, NOT any histone mark, so the pharmacological arm must carry its own independent engagement control and cannot assume a chromatin effect. The proteostasis sensitisation is an on-target confound in any cell under stress, which includes leukaemic lines under drug.

    Evidence that argues against this (1)
    • Because PRMT7 inhibition sensitises cells to proteotoxic stress, a proliferation or viability change in the pharmacological arm may reflect proteostasis rather than chromatin - so the arm must be read on chromatin endpoints at a timepoint before any viability effect, and a genetic arm is required alongside it. Liu C et al. 2022
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt Q03164 (KMT2A_HUMAN, 3969 aa) places the CXXC-type zinc finger at 1147-1195, PHD-type zinc fingers 1-3 at 1431-1482, 1479-1533 and 1566-1627, the bromodomain at 1635-1765, FYR N-terminal at 2018-2074, FYR C-terminal at 3666-3747 and the SET domain at 3829-3945 - so the PHD1-3 and bromodomain cassette and everything C-terminal to it lie beyond the breakpoint cluster region and are absent from a direct 5'-KMT2A fusion, while the CXXC module is retained.

    What this evidence cannot show

    Feature-table coordinates fetched live rather than recalled. The inference that these modules are lost from the fusion follows from their position relative to the breakpoint cluster region as this graph records it; no breakpoint coordinate is asserted here and the exact retained boundary is breakpoint-specific. Reciprocal der products and NUP98::KMT2A retain the C-terminal half and are the exception.

    Evidence that argues against this (1)
    • Which modules a given patient's fusion retains depends on the breakpoint, and this graph records that reciprocal products and NUP98::KMT2A carry the C-terminal half. An allele-asymmetry experiment must therefore state the breakpoint of the line used, because a fusion retaining PHD1-3 would abolish the predicted asymmetry. Guccione E et al. 2007
Connected to