At the KMT2A anchor itself, KDM3B loss does not raise H3K9me1 AND H3K9me2 - it raises me1 3.9-fold while me2 FALLS to 0.44x, an inversion of the mono-to-di ratio that the paper's abstract describes as a joint increase
also called me1 up, me2 down, at the same 1 kb · the mono-to-di inversion at exon 11 · 'increases H3K9me1/2' is not what the element does
The paper says switching off one enzyme raises two closely related tags on the KMT2A gene. At the exact spot that matters, only one goes up - the other goes down. The tags are not moving together, they are swapping.
The same thing, for a scientist
PMID 37788669's abstract states that *"KDM3B depletion increases H3K9me1/2 levels ... at the MLL/KMT2A locus"* [E2]. Read from the paper's own deposited tracks at Ensembl canonical exon 11 and the 1,048 bp CTCF element, KDM3B depletion raises **H3K9me1 to 3.85x and 3.74x** of control while lowering **H3K9me2 to 0.44x**, with both siKDM3B replicates below both controls for me2 (0.383 and 0.153 against 0.608 and 0.610) [E1]. The same inversion holds over the whole 7,466 bp BCR core (me1 2.67x, me2 0.69x) and at the DCAL cut point (2.37x, 0.62x). **H3K9me2 rises only on the flanks** - 1.26x on the 20 kb centromeric flank, 1.06x telomeric [E1] - which is exactly what this graph's mechanism entry records from the paper's *figures* (*"raises H3K9me1 uniformly across the BCR and H3K9me2 on its flank"*). **So the deposited data agree with the figures and disagree with the abstract's compression of them**, and the compression matters: a joint me1/me2 increase and a mono-for-di substitution are different chemical events with different enzymology. G9a/EHMT2 writes me1 and me2 processively; KDM3B removes me2 and me1. Losing the demethylase should raise both unless something is limiting the second methylation step at this element specifically. **Two readings, both testable, neither tested:** either the element's chromatin restricts G9a to a single methylation round (nucleosome accessibility, H3 tail occlusion by the CTCF-cohesin footprint), or a me2-specific demethylase or a me2-to-me3 conversion is active there. **RPE cells; 1 kb resolution; n = 2.**
The inversion, interval by interval
siKDM3B / control ratios of fold-over-chr11-mean, both replicates shown, from the four-arm main series 1:
| interval | H3K9me1 | H3K9me1 per replicate | H3K9me2 | H3K9me2 per replicate | direction |
|---|---|---|---|---|---|
| canonical exon 11 | 3.85 | ctrl [0.131, 0.472] -> [1.226, 1.097] | 0.44 | ctrl [0.608, 0.610] -> [0.383, 0.153] | opposite |
| CTCF element, 1,048 bp | 3.74 | ctrl [0.148, 0.474] -> [1.240, 1.085] | 0.44 | ctrl [0.584, 0.626] -> [0.368, 0.159] | opposite |
| DCAL cut point +/-250 | 2.37 | ctrl [0.532, 0.822] -> [1.859, 1.348] | 0.62 | ctrl [0.575, 0.645] -> [0.476, 0.280] | opposite |
| BCR core, 7,466 bp | 2.67 | ctrl [0.417, 0.661] -> [1.617, 1.267] | 0.69 | ctrl [0.532, 0.622] -> [0.515, 0.281] | opposite |
| 20 kb centromeric flank | 1.01 | - | 1.26 | ctrl [1.011, 0.751] -> [1.145, 1.075] | me2 up, me1 flat |
| 20 kb telomeric flank | 1.85 | ctrl [0.948, 1.051] -> [1.818, 1.872] | 1.06 | - | me1 up, me2 flat |
The replicate structure is the part that matters. For me1 at the element both knockdown replicates exceed both controls with no overlap. For me2 both knockdown replicates fall below both controls with no overlap. These are the two cleanest directional results in the whole dataset and they point opposite ways at the same 1 kb of DNA in the same cells.
And the geometry is complementary, not random. me1 rises at the BCR and its telomeric flank and not its centromeric flank. me2 rises at the centromeric flank and not the BCR. Those two patterns tile the locus rather than overlapping it - which is the signature of a substitution moving along the locus, not of two marks responding independently. This graph records that de novo KMT2A breakpoints skew centromeric within the BCR and therapy-related breakpoints skew telomeric (P = 0.02), so a mark change confined to the telomeric side is the shape the therapy-related aetiology would predict. No entry in this graph currently records either the inversion or the telomeric asymmetry.
Why this is not just a normalisation artefact - and the one way it could be
The arguments that it is real. (a) Both marks were measured in the same cells, the same transfections and the same pipeline, and normalised the same way; a scaling artefact would push them the same direction, not opposite ways. (b) The me2 fall is replicate-consistent with no overlap. (c) The me2 chr11 means move only 7-8% between arms (1.019 and 1.047 in control, 1.101 and 1.115 in siKDM3B) and move UP, so the shared denominator cannot manufacture a 0.44x ratio - if anything it slightly understates the fall. (d) The pattern is spatially structured - BCR versus flank - rather than uniform, which a global artefact would not produce.
The one way it could be an artefact, stated plainly. These are ChIP-over-input ratio tracks, and the H3K9me2 antibody is Abcam ab1220, a different reagent from the H3K9me1 ab8896-100. If the me2 antibody cross-reacts with me1 - a well-known hazard for H3K9 methyl-state antibodies, and the reason the clone is the identifier and not the label - then a large me1 increase could compete the me2 antibody's apparent signal at a site where me1 becomes dominant, producing an apparent me2 fall from a real me1 rise. That mechanism would produce exactly this result and cannot be excluded from these data. It is the first thing a peptide-competition or mass-spectrometry check would settle, and it is why this entry's confidence is moderate rather than strong. Note that it would not explain the centromeric-flank me2 rise, which is in the opposite direction at a site where me1 does not change.
What would settle it. Quantitative mass spectrometry of H3K9 methyl states at the locus is not feasible; the tractable test is an me2 ChIP with a second, independent antibody clone, or a spike-in-normalised me1/me2/me3 ChIP with recombinant nucleosome standards of defined methyl state - a standard reagent set now, and the control this experiment lacked.
| mechanism class | methyl-state redistribution at a single element, as opposed to a joint increase in two marks |
|---|---|
| stage of disease | GENESIS - before any fusion exists |
| the observation | at the KMT2A CTCF element and across the BCR core, KDM3B depletion raises H3K9me1 (3.7-3.9x at the element) and LOWERS H3K9me2 (0.44x) in the same cells and the same 1 kb bins; H3K9me2 rises only on the 20 kb centromeric flank (1.26x) where H3K9me1 does not change, and H3K9me1 rises on the telomeric flank (1.85x) where H3K9me2 does not |
| me1 over me2 ratio change at the element | 8.5-fold (me1 x3.74 while me2 x0.44) |
| the alternative that cannot be excluded | antibody competition. H3K9me1 (Abcam ab8896-100) and H3K9me2 (ab1220) are different reagents and cross-reactivity between adjacent H3K9 methyl states is a known hazard; a four-fold me1 rise could depress apparent me2 signal by competition with no change in true H3K9me2. This would reproduce the element result exactly. It would NOT explain the centromeric-flank me2 rise where me1 is flat. |
| what would settle it | an H3K9me2 ChIP with a second independent antibody clone, or spike-in normalisation with recombinant nucleosomes of defined H3K9 methyl state - a standard reagent set now and the control this experiment lacked |
| geometry | the me1 and me2 responses TILE the locus rather than overlapping it: me1 at the BCR and its telomeric flank, me2 at the centromeric flank. This graph records de novo KMT2A breakpoints skewed centromeric and therapy-related breakpoints skewed telomeric within the BCR (P = 0.02), so a telomeric-confined me1 change is the shape the therapy-related aetiology predicts - and no entry here records it. |
| localization | chromatin at chr11q23.3, the KMT2A breakpoint cluster region and its 20 kb flanks |
| subnuclear localization | not measured. A CTCF-cohesin footprint occluding the H3 tail is one of the two proposed explanations, which would be a nucleosome-scale rather than a subnuclear property; no subnuclear assay bears on it |
| cell type | hTERT-immortalised human retinal pigment epithelial (RPE) cells - epithelial, non-haematopoietic, no KMT2A rearrangement |
| cell state | asynchronous proliferating culture 72 h after siRNA transfection with two independent KDM3B siRNAs |
| epigenetic readout | H3K9me1 versus H3K9me2 ChIP-over-input coverage compared in the same samples at 1 kb resolution; the mono-to-di ratio at the element inverts 8.5-fold on KDM3B depletion |
| shown for fusion | Not shown for any fusion. Measured in unrearranged RPE cells at the intact KMT2A locus before any rearrangement exists; no KMT2A::AFF1 or other fusion-bearing cell was assayed. |
| absence class | measured-not-analysed |
two mechanistic readings neither tested
[ "the element's chromatin restricts G9a/EHMT2 to a single methylation round - nucleosome accessibility, or H3 tail occlusion by the CTCF-cohesin footprint, limiting the processive me1-to-me2 step specifically here", "a me2-selective demethylase, or a me2-to-me3 conversion, is active at this element; note that H3K9me3 does NOT rise at the element (1.01x), which argues against conversion to me3 as the sink" ]
- If the inversion is real, a drug chosen to lower H3K9 methylation is acting on a locus where the two relevant methyl states move in opposite directions - so 'lowering H3K9 methylation' is not a single intervention at this element, and a G9a inhibitor that blocks the me1-writing step may have a different effect from one that blocks the me1-to-me2 step. No available probe distinguishes them.
- This entry sets a published abstract against a re-analysis of the authors' own deposited files, with n = 2 and a named antibody confound. That is a position to hold carefully: the most likely resolution is that the abstract is a compression of figures this entry actually AGREES with, and the genuinely new part is the element-level inversion, not any error by the authors.
- Is the H3K9me2 fall at the element real, or antibody competition from the H3K9me1 rise? A second me2 clone, or spike-in normalisation with defined-methyl-state recombinant nucleosomes, settles it. Until then the inversion is an observation with a named confound and should be quoted that way.
- If it is real, what limits the me1-to-me2 step at this element specifically? G9a/EHMT2 is processive, so a site where mono accumulates without di is a site where the second round is blocked - and the CTCF-cohesin footprint over the same 147-550 bp is the obvious candidate occluder. That is testable in vitro with reconstituted nucleosomes plus or minus CTCF.
- Does the telomeric-only H3K9me1 rise line up with the therapy-related breakpoint skew at base-pair resolution? This graph records the skew (de novo centromeric, t-AML telomeric, P = 0.02) and now records a telomerically confined mark change, and nobody has put the two coordinate sets side by side.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Read from Gray 2023's deposited GSE210480 tracks in RPE cells, KDM3B depletion raises H3K9me1 to 3.85x of control at Ensembl canonical exon 11 and 3.74x at the 1,048 bp CTCF element while lowering H3K9me2 to 0.44x at both, with both replicates separated from both controls in each direction; the same inversion holds over the 7,466 bp BCR core (2.67x, 0.69x) and at the DCAL cut point (2.37x, 0.62x); H3K9me2 rises only on the flanks (1.26x centromeric, 1.06x telomeric) while H3K9me1 rises on the telomeric flank (1.85x) and not the centromeric one (1.01x); and the H3K9me2 chr11 means move only 7-8% between arms and move UPWARD (1.019/1.047 control, 1.101/1.115 siKDM3B), so the shared denominator cannot manufacture the fall.
GEO GSE210480 H3K9me1 and H3K9me2 bigWigs read at interval resolution with pybigtools 0.3.0, 2026-09-22; scripts, per-replicate values and full output at analyses/q108-gse210480-h3k9me3-kmt2a-bcr/ (results.json, arm_summary and main blocks) · Gray ZH, Chakraborty D, Duttweiler RR, Alekbaeva GD, Murphy SE, Chetal K, Ji F, Ferman BI, Honer MA, Wang Z, Myers C, Sun R, Kaniskan HU, Toma MM, Bondarenko EA, Santoro JN, Miranda C, Dillingham ME, Tang R, Gozani O, Jin J, Skorski T, Duy C, Lee H, Sadreyev RI, Whetstine JR · NCBI GEO (data accompanying Cell 186:4528-4545.e18) 2023What this evidence cannot show
n = 2 per arm, no spike-in, no recombinant-nucleosome methyl-state standards, and the histone tracks are 1 kb binned so 'at canonical exon 11' means 'in the 1 kb bin containing it'. The two marks were measured with different antibodies (H3K9me1 ab8896-100, H3K9me2 ab1220 as listed in the paper's STAR Methods), and antibody cross-reactivity between adjacent H3K9 methyl states is a known hazard that could produce an apparent me2 fall from a real me1 rise by competition - this cannot be excluded from these data and is the reason the entry is moderate overall. RPE cells, epithelial, no KMT2A rearrangement.
Evidence that argues against this (1)
- The competition artefact is the live alternative and it would reproduce this result exactly: if ab1220 has appreciable me1 cross-reactivity, a site where me1 rises four-fold could show a falling apparent me2 signal with no change in true H3K9me2. What that alternative does NOT explain is the centromeric-flank me2 rise (1.26x) at a site where me1 is flat (1.01x) - an antibody artefact driven by me1 abundance should produce no me2 change where me1 does not change. Gray ZH et al. 2023
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
PMID 37788669's abstract states that KDM3B depletion increases H3K9me1/2 levels and reduces CTCF occupancy at the MLL/KMT2A locus, describing the two methyl states as moving together; its figures, as recorded in this graph, instead show H3K9me1 rising across the BCR and H3K9me2 rising on its flank.
Epigenetic balance ensures mechanistic control of MLL amplification and rearrangement. · Gray ZH, Chakraborty D, Duttweiler RR, Alekbaeva GD, Murphy SE, Chetal K, Ji F, Ferman BI, Honer MA, Wang Z, Myers C, Sun R, Kaniskan HU, Toma MM, Bondarenko EA, Santoro JN, Miranda C, Dillingham ME, Tang R, Gozani O, Jin J, Skorski T, Duy C, Lee H, Sadreyev RI, Whetstine JR · Cell 2023What this evidence cannot show
The abstract was read in full from PubMed; the figure-level description is taken from this graph's `mechanism-h3k9me1-2-ctcf-balance-gates-kmt2a-rearrangement`, which tabulates it from the paper, rather than from a fresh reading of the panels by this beat. An abstract is a compression and a mismatch between an abstract and a figure is ordinary scientific shorthand, not an error by the authors - the point here is that the compression is the version this graph and any casual reader would carry, and it is the version the deposited data contradict.
Evidence that argues against this (1)
- 'At the MLL/KMT2A locus' in the abstract could legitimately mean the locus and its surroundings taken together, in which case a me1 rise over the BCR plus a me2 rise on the flank IS an increase in H3K9me1/2 at the locus and there is no discrepancy at all - only a resolution difference between an abstract and a 1 kb track. On that reading this entry's contribution is the element-level inversion, which is new, rather than any correction to the paper. Gray ZH et al. 2023
me1 rises 3.7-3.9x at the element while me2 falls to 0.44x in the same 1 kb
H3K9me2 FALLS at the KMT2A CTCF element on KDM3B depletion; the rise is flank-confined
G9a writes me1 and me2 processively, so mono accumulating without di implies a blocked second round at this element
same measurement, same normalisation
the me1 rise is telomerically confined, the shape the therapy-related breakpoint skew would predict
H3K9me2 at the BCR is the median chr11 bin - no change at all at BCR-width resolution
the resting-state H3K9me2 baseline that mechanism starts from, now measured in a CD34+ progenitor at domain resolution