H3K4me2 - the LSD1 substrate, and the only histone mark ever compared head-to-head between two KMT2A fusion partners in human cells
also called H3K4me2 · H3K4Me2 · H3K4 dimethylation · histone H3 lysine 4 dimethylation · dimethyl-H3K4 · H3K4di
H3K4me2 is a half-methylated tag near active genes, and it is the tag the enzyme LSD1 rubs off. LSD1 is one of the few drug targets in KMT2A-fusion leukaemia whose whole rationale is a histone mark - and in the one study that compared two different KMT2A fusions directly, the amount of this mark at a differentiation gene was the thing that differed between them.
The same thing, for a scientist
H3K4me2 is the dimethyl state of histone H3 lysine 4, written by the KMT2/COMPASS family and removed by the FAD-dependent demethylase LSD1/KDM1A (O60341, 852 aa), for which H3K4me1 and H3K4me2 - not H3K4me3 - are the substrates. **In KMT2A-rearranged leukaemia this mark is the mechanistic basis of a real therapeutic programme.** KDM1A is an essential regulator of leukaemia stem-cell potential in murine MLL-AF9, acting at genomic loci bound by MLL-AF9 to sustain the oncogenic programme, and tranylcypromine analogues phenocopy Kdm1a knockdown in murine and primary human MLL-translocated AML while sparing normal HSPCs [E3]. A selective LSD1 inhibitor raises bulk H3K4me1 and H3K4me2 - and explicitly **not** H3K4me3 - in MV4-11 (KMT2A::AFF1) and MOLM-13 (KMT2A::MLLT3) cells while suppressing HOXA9 and MEIS1 [E2]. **And the mark carries this graph's only published head-to-head comparison of a histone modification between two KMT2A fusion partners in human cells**: H3K4me2 at the RARalpha promoter, the PU.1 upstream regulatory region and the RUNX1 +24/+25 intronic enhancer is higher in MLL-AF9-positive cells (THP-1, MOLM-13) than in the MLL-AF4-positive line KOCL-48, and LSD1 inhibition with tranylcypromine raises it in KOCL-48 and restores ATRA sensitivity [E1]. **The measurements are at named differentiation loci and in bulk - not at fusion-bound sites**, and no H3K4me2 ChIP-seq anchored on KMT2A-fusion occupancy has ever been published [E6].
Why this mark and not H3K4me3
This graph already carries mark-h3k4me3, and the two are routinely conflated. They should not be.
- Different enzymes remove them. LSD1/KDM1A cannot demethylate a trimethyl-lysine at all; its substrates are me1 and me2. KDM5-family enzymes remove me3. An LSD1 inhibitor therefore raises H3K4me1 and H3K4me2 and leaves H3K4me3 essentially untouched, which 2 demonstrates directly in two KMT2A-rearranged lines.
- Different distributions. H3K4me3 is a sharp start-site mark; H3K4me2 extends further into gene bodies and enhancer regions, which is why the loci where it has been measured in this disease are a promoter, an upstream regulatory region and an intronic enhancer 1.
- Different relevance to the fusion. The graph's KMT2A H3K4 story is an H3K4me3 story: broad H3K4me3 domains coincide with H3K79me2 mistargeting at MLL-AF4 targets in SEM cells 4. H3K4me2 was not measured in that work, and the abstract-level count of nine records for this mark in this disease 6 contains no fusion-occupancy-anchored map at all.
And a third qualifier applies to both: MLL1's own SET domain, the enzyme that would write either state, is dispensable for HSC maintenance and for MLL-AF9 leukaemogenesis 5. The KMT2A locus's functionally decisive histone output at its targets is recruited H4K16 acetylation, not its own H3K4 methylation.
The partner comparison, which is what makes this entry worth having
PMID 24769646 asked why MLL-AF9-positive AML responds to all-trans retinoic acid and MLL-AF4-positive disease does not, and answered it with a histone mark 1.
| system | fusion | H3K4me2 at RARalpha promoter / PU.1 URE / RUNX1 +24/+25 | ATRA response |
|---|---|---|---|
| THP-1, MOLM-13 | MLL-AF9 (KMT2A::MLLT3) | higher | differentiation, RARalpha/CEBPA/CEBPE/PU.1 up, synergy with cytarabine |
| KOCL-48 | MLL-AF4 (KMT2A::AFF1) | lower | no differentiation |
| murine immortalised lines | MLL-AF9 vs MLL-AF5q31 | same direction | same direction |
| KOCL-48 + tranylcypromine (LSD1 inhibitor) | MLL-AF4 | raised, dose-dependently | ATRA sensitivity restored |
This is the only place in this graph where a histone modification is measured comparatively across KMT2A fusion partners in human cells, and the difference is not at a fusion-bound locus - it is at the retinoic-acid receptor's own promoter and at two myeloid transcription-factor regulatory elements. The mark is behaving as a property of the differentiation programme the fusion permits, not of the fusion's direct targets.
The caveats are real: three cell lines, ChIP-qPCR at four amplicons, no genome-wide data, and KOCL-48 is a single MLL-AF4 line, so 'MLL-AF4 cells have less H3K4me2 here' rests on one line against two. But the therapeutic prediction it makes - that an LSD1 inhibitor should convert an ATRA-resistant KMT2A::AFF1 cell into an ATRA-responsive one - is specific, cheap and testable, and it has not been taken into the infant B-lineage disease.
The eraser is the drug target, and the target is not the mark
KDM1A is an essential regulator of leukaemia stem-cell potential in MLL-AF9 leukaemia, acting at genomic loci bound by MLL-AF9 to sustain the oncogenic programme and prevent differentiation and apoptosis; nanomolar tranylcypromine analogues phenocopy Kdm1a knockdown in murine and primary human MLL-translocated AML while sparing normal HSPC clonogenic and repopulating potential 3. That is a locus-anchored statement about the enzyme.
It is not a locus-anchored statement about the mark. The abstract of 3 reports KDM1A occupancy at MLL-AF9-bound loci and a phenotype; it does not report an H3K4me2 map, and the paper is not in PMC so the full text was not opened for this entry - the caveat records that. Everything downstream of it that does measure the mark measures it in bulk: 2 is a western blot in two cell lines. So the field has a drug target defined by a mark, a phenotype defined by loci, and no published dataset joining the two.
LSD1 also has substrates that are not histones - p53, DNMT1, E2F1 - and 2 itself reports that higenamine changes p53 levels in an LSD1-dependent manner in MV4-11 cells, which is exactly the confound. An LSD1 inhibitor's effect in this disease has never been shown to run through H3K4me2 rather than around it.
The bounded absence
(KMT2A term set)[tiab] AND (H3K4me2 OR "H3K4 dimethylation")[tiab] returns 9 records on 2026-09-20, against 926 for H3K4me2[tiab] alone and 3,209 for the KMT2A set alone 6. Of the nine, one (1) is a primary measurement of this mark in KMT2A-rearranged cells; the others are a ribosomal-RNA study, a case report, a prostate-cancer methylation paper, a WDR5 PROTAC, a nucleus-pulposus degeneration study, a mouse-embryo survey, an unrelated review, and the LSD1 inhibitor paper 2. Nine records is one measurement plus one pharmacodynamic blot.
Fourteen core KMT2A-r chromatin full texts were then read for this beat. H3K4me2 appears in none of them - Guenther 2008, Kerry 2017, Prange 2017, Godfrey 2017/2019/2021, Harman 2021, Crump 2023, Chen 2015, Ropa 2018/2020, Chu 2020 and Valerio 2017 all measure H3K4me3 or H3K4me1 where they measure H3K4 at all. So unlike H3K27ac, H3K9me3 and H4K16ac, rule 80 did not rescue this mark - the full-text sweep confirmed the abstract-level thinness rather than overturning it. That is the correct use of the rule and it is worth recording that it cuts both ways.
| mark | H3K4me2 |
|---|---|
| histone | H3 |
| residue | K4, N-terminal tail |
| chemistry | dimethylation; the demethylase LSD1/KDM1A is FAD-dependent and can act on me1 and me2 but NOT on me3, which is the chemical reason the two states separate pharmacologically |
| writer note | the MLL1-MOF complex carries H3K4 mono/di/trimethyltransferase activity (see gene-kat8); MLL1's SET domain is nonetheless dispensable for HSC maintenance and MLL-AF9 leukaemogenesis (PMID 24813891) |
| eraser uniprot | O60341 |
| eraser length AA | 852 |
| measured in kmt2a r | true |
| measured in kmt2a aff1 | true |
| measured in infant material | false |
| measurement scope | NEITHER at fusion targets nor genome-wide by sequencing - the only locus-resolved data are ChIP-qPCR at four differentiation-pathway amplicons (PMID 24769646) and the only KMT2A::AFF1 data are bulk immunoblot-level (PMID 33618250). No H3K4me2 ChIP-seq anchored on KMT2A-fusion occupancy has been published. |
| menin inhibitor effect | never measured |
| dot1l inhibitor effect | never measured |
| rule 80 changed the answer | false |
| rule 80 note | fourteen core KMT2A-r chromatin full texts were read and H3K4me2 appears in none of them; for this mark the full-text sweep CONFIRMED the abstract-level absence rather than overturning it |
| named in entries when written | 35 |
| mentions when written | 93 |
| count method | regex /\bH3K4me2/i over all 1,701 node JSON files, run 2026-09-20 |
| abstract query count 2026 09 20 | 9 |
| only cross partner histone comparison in this graph | true |
writers
[ "gene-kmt2a", "gene-kmt2c", "gene-kmt2d", "gene-setd1a", "gene-setd1b" ]
erasers
[ "gene-kdm1a" ]
named loci with data
[ "RARalpha promoter and 5'UTR", "SPI1/PU.1 upstream regulatory region (URE)", "RUNX1 +24/+25 intronic enhancer" ]
cell systems with data
[ "THP-1 (human, KMT2A::MLLT3, AML)", "MOLM-13 (human, KMT2A::MLLT3, AML)", "KOCL-48 (human, KMT2A::AFF1)", "MV4-11 (human, KMT2A::AFF1, AML)", "murine MLL-AF9 and MLL-AF5q31 immortalised lines" ]
chemical tools
{
"tranylcypromine (TCP)": "non-selective MAO/LSD1 inhibitor, 10-25 uM in PMID 24769646",
"higenamine": "selective natural-product LSD1 inhibitor, PMID 33618250",
"TCP analogues": "nanomolar, PMID 22464800"
}- LSD1 inhibitors raise H3K4me2 genome-wide, and LSD1 is required for normal erythroid and granulocytic differentiation; the published sparing of normal HSPC clonogenicity was measured in colony assays, not in an infant marrow over months, and thrombocytopenia has been the dose-limiting toxicity in LSD1-inhibitor trials.
- Tranylcypromine, the scaffold every LSD1 inhibitor in this disease derives from, is a non-selective monoamine oxidase inhibitor with hypertensive-crisis and dietary-tyramine interactions - a real constraint in a paediatric inpatient on multiple co-medications.
- The ATRA-combination rationale rests on raising H3K4me2 at the RARalpha promoter in one MLL-AF4 cell line; if the difference between fusion partners is a line property rather than a fusion property, an LSD1-plus-ATRA trial stratified by fusion would be stratifying on the wrong variable.
- Because LSD1 cannot touch H3K4me3, an LSD1 inhibitor cannot reverse the broad H3K4me3 domains that coincide with fusion occupancy - so the drug is aimed at a different methylation state from the one the fusion's chromatin signature is written in, and nobody has shown the two are connected.
- Where is H3K4me2 in a KMT2A-rearranged genome? No ChIP-seq or CUT&RUN for this mark anchored on fusion occupancy has been published in any KMT2A-rearranged system, so the mark underpinning an entire drug class is spatially unmapped in the disease that class is aimed at.
- Does an LSD1 inhibitor restore ATRA sensitivity in infant KMT2A::AFF1 B-ALL, as it does in the MLL-AF4-positive line KOCL-48? The prediction is explicit, the compounds are clinical-stage, and the experiment has not been reported in B-lineage or infant material.
- Is the MLL-AF9 versus MLL-AF4 difference in H3K4me2 at the RARalpha promoter a fusion property or a cell-line property? One MLL-AF4 line against two MLL-AF9 lines cannot separate them, and the murine isogenic arm uses MLL-AF5q31 rather than MLL-AF4.
- Does LSD1 inhibition in this disease act through H3K4me2 at all? Every published effect is compatible with the LSD1-GFI1 scaffolding route or with p53, and no experiment separates catalysis from scaffolding here - a catalytically dead LSD1 knock-in would.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In human MLL-rearranged AML cell lines, the level of H3K4me2 by ChIP-qPCR at the RARalpha gene promoter and 5'UTR, the PU.1 upstream regulatory region and the RUNX1 +24/+25 intronic enhancer is higher in MLL-AF9-positive cells (THP-1, MOLM-13) than in the MLL-AF4-positive line KOCL-48, MLL-AF9-positive cells differentiate in response to all-trans retinoic acid while MLL-AF4-positive cells do not, and inhibiting LSD1 with tranylcypromine raises H3K4me2 at those regions in KOCL-48 and restores ATRA sensitivity; murine immortalised lines expressing MLL-AF9 versus MLL-AF5q31 behave the same way.
Sensitivity of MLL-rearranged AML cells to all-trans retinoic acid is associated with the level of H3K4me2 in the RARα promoter region. · Sakamoto K; Imamura T; Yano M; Yoshida H; Fujiki A; Hirashima Y; et al. · Blood Cancer J 2014What this evidence cannot show
Read in full text (PMC4003419). **Three human cell lines - two MLL-AF9 against one MLL-AF4 (KOCL-48) - so the partner comparison rests on n=1 on the MLL-AF4 side and cannot separate fusion identity from line identity.** ChIP-qPCR at four amplicons, no genome-wide data, antibody given only as 'anti-H3K4me2 (Abcam)' without a catalogue number. The loci are differentiation-pathway regulatory elements, **not** KMT2A-fusion-bound sites, so this is not a measurement of the mark at fusion targets. Tranylcypromine is a non-selective monoamine oxidase inhibitor used here at 10-25 micromolar, concentrations at which off-target activity is expected. Myeloid disease throughout; KOCL-48 carries MLL-AF4 but the comparison is framed as AML.
Evidence that argues against this (1)
- A selective LSD1 inhibitor raises bulk H3K4me1 and H3K4me2 in MV4-11, which is also an MLL-AF4/KMT2A::AFF1 line, and suppresses HOXA9 and MEIS1 there - so MLL-AF4 cells are not intrinsically unable to accumulate the mark, and the difference reported between MLL-AF9 and MLL-AF4 lines is a difference in steady-state occupancy at specific loci rather than in the capacity to acquire it. Fang Y; Yang C; Teng D; Su S; Luo X; Liu Z; et al. 2021
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In the KMT2A-rearranged human leukaemia lines MV4-11 (KMT2A::AFF1) and MOLM-13 (KMT2A::MLLT3), the selective natural-product LSD1 inhibitor higenamine significantly increases the LSD1 substrates H3K4me1 and H3K4me2 while having nearly no effect on LSD1 protein or on H3K4me3, dose-dependently suppresses HOXA9 and MEIS1, induces CD11b/CD14/CD86 differentiation markers and apoptosis, and alters p53 levels in an LSD1-dependent manner.
Discovery of higenamine as a potent, selective and cellular active natural LSD1 inhibitor for MLL-rearranged leukemia therapy. · Fang Y; Yang C; Teng D; Su S; Luo X; Liu Z; et al. · Bioorg Chem 2021What this evidence cannot show
Read at abstract level only; no PMCID. **The H3K4me1/me2 readout is a bulk cellular measurement (immunoblot-level), not locus-resolved**, so it says the drug engages its target and nothing about where in the genome the mark matters. Higenamine has 'acceptable' rather than high potency by the authors' own description, and the paper reports it does not significantly inhibit proliferation - so the differentiation and apoptosis phenotypes sit alongside a weak growth effect. The p53 observation is the standing non-histone confound for every LSD1 inhibitor.
Evidence that argues against this (1)
- If the therapeutic effect ran through H3K4me2, a compound that raises it substantially should inhibit proliferation, and this one does not significantly. The authors describe the mechanism as 'non-classical' and route it partly through p53, which is a non-histone LSD1 substrate. Harris WJ; Huang X; Lynch JT; Spencer GJ; Hitchin JR; Li Y; et al. 2012
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In a mouse model of human MLL-AF9 leukaemia, the H3K4me1/me2 demethylase KDM1A (LSD1) is an essential regulator of leukaemia stem-cell potential, acting at genomic loci bound by MLL-AF9 to sustain the associated oncogenic programme and prevent differentiation and apoptosis; nanomolar tranylcypromine analogues phenocopy Kdm1a knockdown in murine and primary human AML cells with MLL translocations while sparing the clonogenic and repopulating potential of normal haematopoietic stem and progenitor cells.
The histone demethylase KDM1A sustains the oncogenic potential of MLL-AF9 leukemia stem cells. · Harris WJ; Huang X; Lynch JT; Spencer GJ; Hitchin JR; Li Y; et al. · Cancer Cell 2012What this evidence cannot show
**Read at abstract level only.** The paper has no PMCID - checked directly against efetch db=pubmed for an ArticleId of type pmc, per this project's route-1 rule - and no full text was opened for this entry, so whether it contains an H3K4me2 ChIP is unverified and this entry does not assert that it does. Murine MLL-AF9 plus primary human MLL-translocated AML; no B-lineage and no KMT2A::AFF1 data. The claim as written is about the enzyme's occupancy and phenotype, not about the mark's distribution.
Evidence that argues against this (1)
- KDM1A sits in CoREST and NuRD complexes with HDAC1/2 and has abundant non-histone substrates, and the therapeutic effect of tranylcypromine analogues in this disease has been attributed by other groups to disrupting the LSD1-GFI1 interaction rather than to demethylase activity as such. An enzyme dependency is not a mark dependency. Fang Y; Yang C; Teng D; Su S; Luo X; Liu Z; et al. 2021
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In human SEM cells (KMT2A::AFF1 / MLL-AF4 B-ALL), H3K4me3 occurs normally at the start sites of about 90% of active genes despite monoallelic loss of the KMT2A SET domain, and at many regions of H3K79me2 mistargeting H3K4me3 extends across broad domains of similar size that overlap MLL-AF4 occupancy at HOXA7, HOXA9, PROM1 and HMGA2.
Aberrant chromatin at genes encoding stem cell regulators in human mixed-lineage leukemia. · Guenther MG; Lawton LN; Rozovskaia T; Frampton GM; Levine SS; Volkert TL; et al. · Genes Dev 2008What this evidence cannot show
Read in full text (PMC2607073). **This is H3K4me3, not H3K4me2** - it is cited here to establish what the KMT2A-r H3K4 literature actually measured, and to bound the claim that no fusion-anchored H3K4me2 map exists. Single cell line for the fusion arm, 2008-era ChIP-seq depth, and the '90% of active genes' figure is drawn from a supplemental table.
Evidence that argues against this (1)
- Because this study measured only the trimethyl state, it cannot say whether H3K4me2 is also broadened at MLL-AF4 targets, spared, or inversely distributed - and since LSD1 acts on me1/me2 at exactly these loci in the myeloid counterpart disease, the missing measurement is the one that would matter therapeutically. Harris WJ; Huang X; Lynch JT; Spencer GJ; Hitchin JR; Li Y; et al. 2012
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The SET domain, and therefore the H3K4 methyltransferase activity, of MLL1 is dispensable for maintaining haematopoietic stem cells and for supporting leukaemogenesis driven by the MLL-AF9 fusion oncoprotein; a SET-deleted mutant shows unchanged H4K16 acetylation at both gene-specific and global level, and SIRT1 inhibition restores H4K16Ac and MLL1 target-gene expression without affecting H3K4me3.
The histone methyltransferase activity of MLL1 is dispensable for hematopoiesis and leukemogenesis. · Mishra BP; Zaffuto KM; Artinger EL; Org T; Mikkola HK; Cheng C; et al. · Cell Rep 2014What this evidence cannot show
Read in full text (PMC4120120). Mouse genetics; the leukaemia arm is MLL-AF9 myeloid. The experiment deletes the SET domain of wild-type Mll1 and does not test H3K4me2 specifically - the histone readouts reported are H4K16Ac, H3K4me3 and H3K27me3. So 'H3K4 methyltransferase activity is dispensable' is supported at the level of the domain and of H3K4me3, and the dimethyl state was not separately assayed.
Evidence that argues against this (1)
- Redundancy among the six human KMT2 enzymes means deleting one SET domain does not remove H3K4 methylation from the genome; KMT2B, KMT2C, KMT2D, SETD1A and SETD1B all remain. 'Dispensable' here means dispensable in the presence of the rest of the family, not that the mark is dispensable. Guenther MG; Lawton LN; Rozovskaia T; Frampton GM; Levine SS; Volkert TL; et al. 2008
- E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
As of 2026-09-20 PubMed returns 9 records for the KMT2A term set intersected with (H3K4me2 OR 'H3K4 dimethylation')[tiab], against 926 for H3K4me2[tiab] alone and 3,209 for the KMT2A term set alone; only one of the nine is a primary measurement of this mark in KMT2A-rearranged cells, and H3K4me2 appears in none of the fourteen core KMT2A-rearranged chromatin full texts read for this beat.
NCBI E-utilities esearch over PubMed, run 2026-09-20 by the Q59 marks2 beat, with every querytranslation read back and recorded · US National Library of Medicine 2026What this evidence cannot show
**Nine in titles and abstracts; supporting figures are not searched by the query** - but for this mark, unlike H3K27ac, H3K9me3 and H4K16ac, the full-text sweep of fourteen core papers recovered nothing additional, so the abstract count and the document-level picture agree. Fourteen full texts is not the whole literature, and the name list was H3K4me2 and 'H3K4 dimethylation' only; 'H3K4me1/2', 'dimethylated H3K4' and 'H3K4 me2' were not run and would widen it. Querytranslations were read back and none contained a [UID] term.
Evidence that argues against this (1)
- The thinness is of the mark's measurement, not of its biology in this disease: LSD1, whose substrates are H3K4me1 and H3K4me2, is one of the best-validated non-menin, non-DOT1L targets in KMT2A-rearranged leukaemia, with clinical-stage inhibitors. The mark is load-bearing for a drug programme and almost unmeasured. Harris WJ; Huang X; Lynch JT; Spencer GJ; Hitchin JR; Li Y; et al. 2012
LSD1 demethylates H3K4me1/me2 and not me3; inhibition raises both in KMT2A::AFF1 and KMT2A::MLLT3 lines
same residue, different state, different erasers - an LSD1 inhibitor moves me2 and leaves me3
- related to →KMT2A / MLL1 — the gene that breaks
written by the KMT2/COMPASS family, whose MLL1 SET domain is dispensable for MLL-AF9 leukaemogenesis
- related to →HOXA9 — required for survival of established MLL-rearranged leukemia, and dispensable for starting it
LSD1 inhibition raises the mark and suppresses HOXA9 in MV4-11 and MOLM-13
LSD1 inhibition raises the mark and suppresses MEIS1 in MV4-11 and MOLM-13
- related to →The differentiation block — what it actually is, operationally, in KMT2A::AFF1 leukaemia
low H3K4me2 at RARalpha, PU.1 URE and the RUNX1 enhancer tracks with failure to differentiate under ATRA
- part of →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here
layman explainer for the mark layer
- ← related toKMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
KOCL-48 and MV4-11 (both KMT2A::AFF1) are the only index-fusion systems with any H3K4me2 data
the SET-domain-dispensability result puts H4K16ac, not H3K4 methylation, as the decisive histone output at MLL1 targets
supplies the counted bound; the full-text sweep confirmed the absence
mutual exclusivity is annotated for H3K4me2 as well as H3K4me3