Oncofusion

start / graph / ana-q108-mark-response-versus-cytogenetic-response-at-four-loci

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q108-h3k9me3-at-the-bcr

The paper's four comparison loci, marks read against FISH calls: the H3K9 mark response agrees with the cytogenetic response at two loci and contradicts it at two - the mark is neither necessary nor sufficient

also called MLLT1's marks move and its locus does not · AFF3 responds without a mark change · the four-locus mark-versus-FISH table

The paper tested four other genes to show its mechanism is specific: two broke and two did not. Reading its own raw data shows the chemical tag it blames moved at one of the genes that did not break, and did not move at one of the genes that did.

The same thing, for a scientist

PMID 37788669 uses four loci to argue specificity: *TCF3* and *AFF3* undergo copy gain on KDM3B depletion (*TCF3* inherited, *AFF3* transient) while *MLLT1* and *MLLT3* do not [E2]. The paper never shows the chromatin at any of the four. Its own deposited H3K9me1/2/3 tracks were read over all four gene bodies in the same samples [E1]. **The mark response matches the cytogenetic call in two of four and contradicts it in two.** *TCF3*, responsive: H3K9me1 +1.46x - agrees. *MLLT3*, non-responsive: +0.90x, no change - agrees. ***MLLT1*, non-responsive: H3K9me1 +1.38x and H3K9me2 +1.43x** - a mark change with no cytogenetic consequence. ***AFF3*, responsive: H3K9me1 0.86x**, i.e. down - a cytogenetic consequence with no mark change [E1]. This is the direct test of `ana-gse210480-h3k9me3-at-the-kmt2a-bcr`'s third open question, which asked exactly this and predicted it would be informative. It is: **at these four loci an H3K9me1/2 rise is neither necessary nor sufficient for the cytogenetic event**, which means the mark cannot be the whole address. RPE cells; gene-body averages, which is the design's main weakness.

Go deeper
The table, and how much weight it can carry

siKDM3B / control ratios of fold-over-chr11-mean, two replicates per arm 1:

locuschromosomeFISH call in the paperH3K9me1H3K9me2H3K9me3CTCFverdict
TCF319responsive, inherited copy gain1.461.231.030.98mark up, locus responds - agrees
AFF32responsive, transient copy gain0.861.061.101.03locus responds, mark does not move - contradicts
MLLT119non-responsive1.381.431.250.93mark up, locus does not respond - contradicts
MLLT39non-responsive0.900.971.091.01no mark change, no response - agrees
KMT2A BCR core (reference)11responsive, inherited2.670.691.340.62-

Two things this does establish. (a) MLLT3 is a genuinely clean negative on every channel - no mark change, no CTCF change, no cytogenetic event - and it is the best internal negative control in the whole series. (b) MLLT1 and AFF3 each break the mark-to-event mapping, in opposite directions.

Three things it does not establish, stated because they bound the claim. (a) These are gene bodies, 15-570 kb wide, averaged. A localised mark change at a single 1 kb element inside AFF3 - which spans 570 kb - would be invisible in a gene-body mean, and AFF3's 'no mark change' is therefore much weaker evidence than MLLT1's 'mark change' (a rise cannot be created by averaging). (b) The paper's FISH calls concern copy gain and probe separation, scored at particular timepoints; 'non-responsive' means not detected in that assay, not impossible. (c) No sequenced junction exists anywhere in this paper, at any locus.

Where the asymmetry leaves the argument. The MLLT1 result is the robust half: a 1.38x me1 and 1.43x me2 rise over a whole gene body, in both replicates, at a locus the paper scores as not responding. Averaging can only dilute a rise, never manufacture one, so this is a mark change without a cytogenetic consequence and the mark is not sufficient. The AFF3 result is the fragile half and 'the mark is not necessary' should be held more loosely, at 1 kb resolution over 570 kb, until someone looks at AFF3 element by element.

Facts and figures
scriptanalyses/q108-gse210480-h3k9me3-kmt2a-bcr/run.py
statusrun
run date2026-09-22
result summarythe H3K9 mark response agrees with the paper's cytogenetic call at TCF3 and MLLT3 and contradicts it at MLLT1 (mark up, no event) and AFF3 (event, no mark change); MLLT3 is a clean negative on every channel
readingan H3K9me1/2 rise is neither necessary nor sufficient for the cytogenetic event at these four loci. The sufficiency failure (MLLT1) is robust because a gene-body average can only dilute a rise. The necessity failure (AFF3) is fragile because a 570 kb average can hide a local change.
asymmetry of evidencea POSITIVE gene-body result is strong (averaging cannot manufacture a rise); a NEGATIVE gene-body result is weak (averaging can erase one). This asymmetry is why the two contradictions carry different weight and it is stated rather than averaged away.
answers open questionana-gse210480-h3k9me3-at-the-kmt2a-bcr open question 3: 'Do MLLT1 and MLLT3 show no H3K9 methylation change at all, or a change without a cytogenetic consequence?' Answer: MLLT3 shows no change; MLLT1 shows a change without a consequence.
localizationchromatin at four gene bodies on chr2, chr9 and chr19 (two of them, MLLT1 and TCF3, both on chr19)
subnuclear localizationnot measured - a ChIP coverage comparison has no subnuclear coordinate
cell typehTERT-immortalised human retinal pigment epithelial (RPE) cells - epithelial, non-haematopoietic, no KMT2A rearrangement
cell stateasynchronous proliferating culture 72 h after siRNA transfection
epigenetic readoutH3K9me1, H3K9me2 and H3K9me3 ChIP-over-input coverage at 1 kb resolution, averaged over whole gene bodies, compared with the founding paper's FISH copy-gain and break-apart calls
shown for fusionNot shown for any fusion, but two of the four loci are KMT2A fusion partners - MLLT1 (KMT2A::MLLT1) and MLLT3 (KMT2A::MLLT3). They were assayed as unrearranged loci in RPE cells; nothing here was measured in a fusion-bearing cell.
absence classmeasured-not-analysed
four locus table
AFF30.860AFF3: 0.86MLLT30.900MLLT3: 0.9MLLT11.380MLLT1: 1.38TCF31.460TCF3: 1.46
inputs
[
  "GSE210480 H3K9me1/2/3 and CTCF bigWigs, four arms, RPE, hg19",
  "MLLT1, MLLT3, TCF3, AFF3 gene bodies looked up directly in the Ensembl GRCh37 archive"
]
four locus table — the numbers behind the chart
{
  "TCF3": {
    "fish": "responsive, inherited",
    "h3k9me1": 1.46,
    "h3k9me2": 1.23,
    "h3k9me3": 1.03,
    "ctcf": 0.98,
    "verdict": "agrees"
  },
  "AFF3": {
    "fish": "responsive, transient",
    "h3k9me1": 0.86,
    "h3k9me2": 1.06,
    "h3k9me3": 1.1,
    "ctcf": 1.03,
    "verdict": "contradicts - event without a mark change"
  },
  "MLLT1": {
    "fish": "non-responsive",
    "h3k9me1": 1.38,
    "h3k9me2": 1.43,
    "h3k9me3": 1.25,
    "ctcf": 0.93,
    "verdict": "contradicts - mark change without an event"
  },
  "MLLT3": {
    "fish": "non-responsive",
    "h3k9me1": 0.9,
    "h3k9me2": 0.97,
    "h3k9me3": 1.09,
    "ctcf": 1.01,
    "verdict": "agrees - clean negative on every channel"
  }
}
How this could bite us back
  • This entry weakens the mark's claim to be the mechanism's address, and a reader who wants the mechanism to be true will reach for the gene-body-averaging objection - which is a genuine objection to the AFF3 half and no objection at all to the MLLT1 half. The two halves must not be dismissed together.
  • If an H3K9me1 rise is not sufficient for the cytogenetic event, then a therapy aimed at lowering H3K9 methylation to prevent rearrangement is acting on a marker rather than a cause at some loci. That does not make the KMT2A result wrong, but it removes the inference that the mark is the lever everywhere.
Still open
  1. Is AFF3's apparent non-response an averaging artefact? AFF3 spans 570 kb and was read as one mean. Splitting it into 1 kb windows - the native resolution of these tracks - and asking whether any window rises the way the KMT2A BCR does would settle whether the necessity failure is real. Same four files, minutes of compute.
  2. Why does MLLT1's chromatin respond when its locus does not? If the mark rise is real and the locus is genuinely stable, MLLT1 is the best available system for asking what protects a locus from an H3K9me1 rise - and the obvious candidate, given this beat's CTCF result, is that MLLT1 has no CTCF anchor whose occupancy the mark can displace.
  3. Do the two loci on chromosome 19 behave differently from each other for a reason other than sequence? MLLT1 and TCF3 are both chr19 and both show a ~1.4x me1 rise, yet one is scored responsive and the other not. A chromosome-level confound in the FISH scoring, or in the mark tracks, would be visible as a chr19-wide skew and was not checked.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Read from Gray 2023's deposited GSE210480 bigWigs over whole GRCh37 gene bodies in the same RPE samples: on KDM3B depletion H3K9me1 rises 1.46x over TCF3 and 1.38x over MLLT1 (with H3K9me2 also 1.43x over MLLT1), does not rise over MLLT3 (0.90x) and falls slightly over AFF3 (0.86x), while CTCF is unchanged at all four (0.93x-1.03x) - against 2.67x H3K9me1 and 0.62x CTCF over the KMT2A BCR core in the same samples.

    GEO GSE210480 H3K9me1/2/3 and CTCF bigWigs read over GRCh37 gene-body intervals for MLLT1, MLLT3, TCF3 and AFF3 (coordinates looked up directly in the Ensembl GRCh37 archive, no lift) with pybigtools 0.3.0, 2026-09-22; scripts and full output at analyses/q108-gse210480-h3k9me3-kmt2a-bcr/ · 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) 2023
    PMID 37788669DOI
    What this evidence cannot show

    Whole-gene-body averages, 15 kb (MLLT1) to 570 kb (AFF3). A gene-body mean can dilute a localised change to nothing, so a NEGATIVE result here is weak evidence and a POSITIVE result is strong - which makes the MLLT1 finding much more load-bearing than the AFF3 one. n = 2 per arm, no spike-in, histone tracks 1 kb binned. Gene coordinates were taken from the Ensembl GRCh37 archive directly rather than lifted, so no lift error is possible in these four intervals; the strand-aware promoter windows used elsewhere in this beat were not used here. RPE cells, epithelial. Confidence is moderate rather than strong because the design is a gene-body average, not because the numbers are uncertain.

    Evidence that argues against this (1)
    • The cytogenetic calls this table is compared against come from FISH probe separation and copy-number scoring at specific timepoints, and the paper's own AFF3 result is explicitly TRANSIENT (not inherited at 72 h) while TCF3's is inherited. A transient event may be produced by a chromatin change too small or too brief to appear in a bulk ChIP at the single timepoint sequenced, which would dissolve the AFF3 contradiction without touching the MLLT1 one. Gray ZH et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PMID 37788669 reports that TCF3/E2A and AFF3/LAF4 undergo KDM3B-depletion-dependent copy gain - TCF3's inherited, AFF3's transient - while MLLT1/ENL and MLLT3/AF9 do not respond, and presents no chromatin measurement at any of the four.

    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 2023
    PMID 37788669DOI
    What this evidence cannot show

    The four-locus panel and the inherited-versus-transient distinction are recorded in this graph's `mechanism-h3k9me1-2-ctcf-balance-gates-kmt2a-rearrangement` from the paper's figures; this beat did not re-read the FISH panels themselves. The claim that the paper shows no chromatin at the four loci is a negative about the paper's figure set and rests on that entry's reading plus the full-text search recorded in `absence-h3k9me3-at-the-kmt2a-locus-was-assayed-and-never-reported`, not on an independent figure-by-figure audit by this beat.

    Evidence that argues against this (1)
    • If the paper does contain a chromatin panel at one of these four loci that this graph's reading missed, the whole framing of this entry as 'the paper never shows it' is wrong even though the measurements stand. That is a live possibility: the absence at these loci was established by the same text-search method whose stated bound is that a result present only as an unlabelled figure panel would not be seen. Iqbal Lab oncofusion knowledge graph 2026
Connected to