The H3K9me1/2-CTCF brake: a KDM3B-G9a balance keeps the KMT2A breakpoint cluster region methyl-poor and CTCF-bound, and losing either end makes the locus amplify and rearrange - in epithelium, in primary cord-blood progenitors and in mice

[
  "KDM3B binds the KMT2A breakpoint cluster region and keeps H3K9me1 low there",
  "G9a/EHMT2 writes H3K9me1/2 at the same region; the two set a balance",
  "loss of KDM3B (or gain of G9a) raises H3K9me1 across the BCR and H3K9me2 on its flank",
  "CTCF occupancy at the conserved KMT2A exon-11 CTCF site falls",
  "the locus undergoes site-specific copy gain and break-apart, transient at 12 h, inherited by 72 h",
  "doxorubicin or etoposide enters at step 3 by lowering KDM3B and CTCF protein, partly via the proteasome"
]

{
  "gene": "KMT2A",
  "element": "the 8.3 kb breakpoint cluster region; the CTCF site is in EXON 11",
  "exon_span_as_stated_by_the_source": "exons 8-14",
  "numbering_conventions_in_this_graph_for_the_SAME_8.3_kb": {
    "breakpoint-kmt2a-major-bcr": "intron 7 - exon 13 (DCAL numbering)",
    "breakpoint-kmt2a-bcr-architecture": "exons 5-11 (historic numbering)",
    "PMID 37788669": "exons 8-14",
    "warning": "three different exon spans for one 8.3 kb interval. These are numbering-convention differences, not errors, and they must NOT be silently reconciled. Any coordinate work on this element must start from hg38 base positions, not exon numbers.",
    "RESOLVED_cycle_57": "The DCAL/recombinome frame is Ensembl canonical + 1, established by three exact bp identities rather than by assertion. The warning in this block was correct and this entry violated it three paragraphs later; the violation is now the reason the rule is stated with a measurement attached. A fourth frame exists and is NOT reconciled: Canela 2017's 'near exon 10'/'at exon 12' are MOUSE Kmt2a exons."
  },
  "coordinate_coincidence_not_yet_checked": "ANSWERED, cycle 57 - and the answer is YES, they are the same element. Atkin's break-prone window chr11:118,488,500-800 (PMID 34405474) CONTAINS all 147 bp of canonical exon 11 (chr11:118,488,614-118,488,760) and sits wholly inside the CTCF/RAD21/SMC3 cluster at chr11:118,488,159-118,489,206. The break-prone window and the protective CTCF site are 0 bp apart, so the fragile-versus-guarded dispute cannot be dissolved by geography. It was not a one-afternoon check by the time it was done properly: it needed the numbering frame settled first, which is the step that had been silently skipped.",
  "functional_cut_point_overlap": "RETRACTED AND CORRECTED, cycle 57. This field read 'The CTCF site sits at that boundary.' It does not. breakpoint-kmt2a-major-bcr's 'intron-10/exon-11' boundary is DCAL numbering = Ensembl canonical + 1 = canonical intron 9/exon 10 = chr11:118,484,861|118,484,862. The CTCF site is CANONICAL exon 11 = chr11:118,488,614-118,488,760, which is 3,752 bp telomeric of the cut point (3,297 bp to the CTCF anchor's near edge). The subregion-C and telomeric-half statements are unaffected and still hold. See evidence E_k57_coordinates."
}

{
  "responsive": [
    "TCF3/E2A - copy gains INHERITED",
    "AFF3/LAF4 - copy gains transient, NOT inherited (gene-aff3 exists in this graph and AFF3 is a KMT2A fusion partner)"
  ],
  "not_responsive": [
    "MLLT1/ENL",
    "MLLT3/AF9"
  ],
  "reading": "the specificity is real but not KMT2A-exclusive, and all three responsive loci are lymphoid-rearrangement genes. That MLLT1 and MLLT3 do NOT respond is the informative negative: the 3' partner loci of the index fusions are not regulated this way, which argues against a two-sided version of any shared-set-point model."
}

{
  "causative": [
    "drug-doxorubicin - lowers KDM3B and CTCF transcript and protein",
    "drug-etoposide - same effect, and etoposide IS in Interfant-06 while doxorubicin is not",
    "drug-jdi-12 - the KDM3-family inhibitor ('KDM3i'), 25 nM to 1 microM",
    "G9a/EHMT2 over-expression, 24 h, sufficient alone"
  ],
  "protective": [
    "drug-unc0642 - dual G9a/GLP inhibitor ('EHMTi'), 1.5-2.5 microM in cells, 5 mg/kg i.p. daily in mice",
    "siG9a - rescues the transient form, does NOT rescue the inherited form",
    "KDM3B over-expression - blocks Dox-induced gains and rearrangements",
    "CTCF over-expression - prevents Dox-induced copy gain (break-apart unassessable because CTCF over-expression itself separated the probes)"
  ],
  "not_deliverable": "KDM3B and CTCF over-expression are the best-evidenced protective interventions and are not clinical modalities"
}

{
  "dox_only": "mixed C57BL/6-129Sv, 3-5 months, 4 males + 4 females, doxorubicin 1.5 mg/kg i.v. in 100 ul daily x3 or saline; spleen isolated 24 h after last dose; Kmt2a copy gain by DNA FISH with a Control-9 probe unchanged",
  "combination": "B6129SF1/J (JAX 101043), 7 males + 8 females in 4 arms - vehicle n=4, EHMTi 5 mg/kg i.p. daily x4 n=4, doxorubicin 1.5 mg/kg i.v. daily x3 n=4, both n=3; spleen scored double-blinded",
  "endpoint": "cytogenetic (Kmt2a copy gain by FISH). NOT leukaemia incidence, NOT survival, no latency arm.",
  "n_warning": "3-4 mice per arm in the combination study. This is the weakest quantitative link in the chain and it is the link the therapeutic proposal rests on."
}

{
  "new": "GSE210480 - 60 human RPE samples: H3K9me1, H3K9me2, H3K9me3, CTCF and Input across siCTRL, siKDM3B (2 independent siRNAs), siG9A (2 siRNAs) and double knockdown, 2 replicates each",
  "reused": "GSE71885 - published KDM3B ChIP-seq in colorectal cancer stem cells, the source of all KDM3B occupancy in the paper",
  "unreported_arm": "8 H3K9me3 samples are deposited and no H3K9me3 result appears anywhere in the paper"
}

[
  "a catalytically dead KDM3B that rescues locus stability - the mark would be a bystander",
  "RUN IN CYCLE 59, ANSWER IS SCALE-DEPENDENT AND THE FALSIFIER NAMED NO SCALE. At the 1,048 bp CTCF element H3K9me3 is 1.014x against H3K9me1's 3.738x, and 1.042 against 3.852 at canonical exon 11 and the summit - it does NOT fire there. Over the 7,466 bp BCR core me3 rises 1.344x against me1's 2.675x, ranking 3,877 of 18,082 chr11 bins against me1's 72 - it PARTLY fires there. Mark identity was controlled (ZNF274 body me3 3.76 vs GAPDH promoter 0.309). See data.FALSIFIER_ADJUDICATION_cycle_61 for the full reading. Original text: H3K9me3 moving with me1/me2 at the BCR in the deposited GSE210480 data - the me1/me2-specific framing would collapse",
  "CTCF occupancy at the exon-11 site NOT falling in a primary haematopoietic cell, as Lehman et al. found for CTCF binding genome-wide under stress",
  "JDI-12 still producing KMT2A copy gains in a KDM3B-null cell - the pharmacological arm would belong to JMJD1C",
  "a normal 11q23 in the Diets-Jongmans AML case, or no excess of KMT2A-rearranged leukaemia in larger DIJOS cohorts",
  "a properly powered mouse study with a leukaemia-incidence endpoint showing the copy gains are epiphenomenal - which the authors themselves half-expect, writing that KDM3B loss alone confers no major fitness advantage and that t-AML latency runs to 15 years"
]

{
  "entries": [
    "mechanism-loop-anchor-dsb-hotspots",
    "gene-ctcf (E2, E4)",
    "hyp-kmt2a-aff1-shared-loop-anchor-hub",
    "genome-aff1-breakpoint-region-architecture"
  ],
  "the_contradiction": "those entries make CTCF/cohesin-bound loop anchors the PLACES WHERE KMT2A BREAKS; this mechanism makes CTCF occupancy the thing that PREVENTS rearrangement. Both are published. PMID 37788669 cites both Gothe 2019 and Atkin 2021 and addresses neither conflict.",
  "do_not_amend": "none of those entries should be amended on the strength of this one. The conflict is recorded here and in their direction only as a dispute to be resolved by the coordinate overlap check and by a CTCF ChIP in the right cell type."
}

{
  "drug-doxorubicin E3": "doxorubicin induces site-specific cleavage in the MLL BCR at positions colocalising with etoposide's, in primary progenitor haematopoietic cells from healthy newborns and adults, with religation demonstrated - the DNA-cleavage half, which this mechanism sits upstream of",
  "breakpoint-kmt2a-bcr-architecture": "de novo and therapy-related breakpoints are non-randomly distributed within the BCR (de novo centromeric, t-AML telomeric, P=0.02), implying different initiating mechanisms - this mechanism is a candidate for the therapy-related half specifically",
  "gene-ehmt2": "G9a is a selective regulator of fast-proliferating myeloid progenitors with no discernible HSC function, and its inhibition delays established AML - so G9a inhibition is pointed the same way at both disease stages, which is unusual and is the one clean part of the therapeutic picture"
}

{
  "the_refuter": "H3K9me3 moving with H3K9me1/2 at the BCR would collapse the mark-identity distinction this mechanism depends on",
  "answer_in_RPE_at_the_element": "it does NOT move. All four me3 arms lie inside the control replicate spread at the CTCF element while me1 rises 3.74x; the me3 log response is 1.1% of me1's. The refuter does not fire at the element.",
  "but_it_partly_fires_over_the_BCR_core": "me3 rises 1.34x over the 7,466 bp core and 1.55x at the cut point - and equally under siG9A, which is the WRONG SHAPE for the balance this mechanism models. A quarter of that is a chromosome-wide shift (chr11 median me3 log2FC +0.114 vs +0.004 for me1).",
  "and_a_harder_problem_for_the_first_arrow": "H3K9me1 rises broadly - 1.38-1.85x at MLLT1, TCF3, GAPDH and ACTB - while CTCF falls to 0.30x at canonical exon 11 and is unchanged at eleven other intervals including the cut point 3.8 kb away. So the LOCUS ADDRESS sits on the CTCF arrow, not the mark arrow, and the locus label on this mechanism's first arrow is not supported by the deposited data.",
  "and_the_mark_is_neither_necessary_nor_sufficient": "at the paper's own four comparison loci: MLLT1 (non-responsive) gains me1 +1.38x and me2 +1.43x - mark without event; AFF3 (responsive) shows me1 0.86x - event without mark. The sufficiency failure is robust; the necessity failure is fragile because it averages 570 kb.",
  "scope": "RPE epithelium, 1 kb resolution, 72 h siRNA. Nothing haematopoietic."
}

{
  "why_this_block_exists": "analyses/the-unrun-falsifier-queue/run.py swept all 2,125 entries for stated falsifiers: 161 entries state one, 1,964 state none, and of 624 falsifier items only 8 carry any sign of having been run. Ranked by edge exposure this entry is the TOP of that queue - 38 edges, confidence moderate, status verified, six falsifiers, no outcomes recorded - and it is the entry carrying this graph's prophylactic-intervention proposal. Cycle 60's headline was bought by cashing exactly one such field, so these were read.",
  "[0] a catalytically dead KDM3B that rescues locus stability": {
    "verdict": "NOT RUN",
    "note": "No catalytically dead KDM3B rescue experiment is cited anywhere in this graph. This remains the cleanest single test of whether the mark is causal or a bystander, and it is unattempted."
  },
  "[1] H3K9me3 moving with me1/me2 at the BCR in GSE210480": {
    "verdict": "RUN IN CYCLE 59, AND THE ANSWER IS SPLIT BY SCALE. THIS ENTRY NEVER RECORDED IT.",
    "where": "analyses/q108-gse210480-h3k9me3-kmt2a-bcr/run.py; entries ana-q108-h3k9me3-at-the-kmt2a-bcr-measured and ana-q108-chr11-rank-of-the-kmt2a-bcr-h3k9-response",
    "at the element, exon 11 and the summit - NOT MET": "H3K9me3 siKDM3B/control is 1.014 at the 1,048 bp CTCF element and 1.042 at canonical exon 11 and the summit, while H3K9me1 is 3.738 and 3.852 at the same intervals. me3 does not move where me1 moves nearly fourfold, so the me1/me2-specific framing SURVIVES at the element. Mark identity was controlled: ZNF274 body me3 3.76 against GAPDH promoter 0.309, ratio 12.168.",
    "over the wider BCR core - PARTIALLY MET": "H3K9me3 rises 1.344 over the 7,466 bp BCR core while H3K9me1 rises 2.675, and on the chr11 null the BCR bin ranks 3,877 of 18,082 for me3 (top 21.4%) against 72 of 18,082 for me1. So me3 DOES move with me1 over the wider locus, just less - and about a quarter of that is a chromosome-wide shift, since chr11's median siKDM3B/control log2 is +0.1138 for me3 against +0.0044 for me1. Median-corrected the locus-specific me3 change is 1.22x against me1's 2.23x.",
    "so the honest verdict": "The falsifier as written does not specify a scale, and the answer depends entirely on scale. At the element it fails to fire; over the BCR it partially fires. A falsifier that can be met or not met depending on an interval the falsifier does not name is underspecified, and that is a defect in this entry rather than in the measurement.",
    "AND THE MARK THAT TRACKS me1 LEAST IS me2, WHICH IS IN THIS ENTRY'S OWN TITLE": "Over the BCR core H3K9me2's log2 is -0.0326, rank 10,276 of 18,082 - the middle of chromosome 11 - so me2 does not rise at all. At the element itself, mechanism-kdm3b-loss-shifts-the-kmt2a-anchor-from-h3k9me2-to-h3k9me1 records me2 FALLING to 0.44x while me1 rises 3.9x. This entry is named for an 'H3K9me1/2 balance', and the measured behaviour is me1 up, me2 down, me3 slightly up. The pairing in the title is the part under strain, and no falsifier here tests it."
  },
  "[2] CTCF occupancy at the exon-11 site NOT falling in a primary haematopoietic cell": {
    "verdict": "NOT RUN, and it is the experiment this graph has already designed and costed",
    "note": "exp-h3k9me3-and-ctcf-at-the-kmt2a-anchor-in-a-cd34-progenitor specifies it - cord-blood CD34+ HSPCs, control and KDM3B knockdown, ChIP for H3K9me1/2/3 and CTCF with ZNF274 and GAPDH as mark-identity controls - and it is still a proposal. Every CTCF measurement behind this entry is RPE.",
    "and something adjacent DID happen in cycle 60": "analyses/q111-ctcf-anchor-matched-ranking/ showed the CTCF fall is not locus-specific in RPE: 84.6% of 12,443 anchors fall to <=0.9x, and against 1,625 strength-matched anchors the KMT2A anchor is the 3.57th percentile with a matched median of 0.633x. That does NOT test this falsifier, which is about a primary haematopoietic cell - but it weakens the same mechanism by a different route, and the entry should not be read as untouched merely because its own falsifier is unrun."
  },
  "[3] JDI-12 still producing KMT2A copy gains in a KDM3B-null cell": {
    "verdict": "NOT RUN",
    "note": "No KDM3B-null JDI-12 experiment is cited here. This is the falsifier that would separate the pharmacological arm from KDM3B and assign it to JMJD1C - and note that JMJD1C is the HGNC-approved symbol for the gene this graph writes as gene-kdm3c, so the two arms of that question are one gene under two names."
  },
  "[4] a normal 11q23 in the Diets-Jongmans AML case, or no excess in larger DIJOS cohorts": {
    "verdict": "NOT RUN - and it is a cohort question, not a bench question",
    "note": "Nothing in this graph reports a larger DIJOS cohort. This falsifier cannot be cashed by any analysis this project can run; it waits on a published cohort."
  },
  "[5] a powered mouse study with a leukaemia-incidence endpoint": {
    "verdict": "NOT RUN - and the entry already records that the authors half-expect it to fire",
    "note": "The falsifier's own text notes the authors write that KDM3B loss alone confers no major fitness advantage and that t-AML latency runs to 15 years. A falsifier the original authors expect to fire, left unrun, is the strongest reason this entry's confidence should not rise."
  },
  "SCORE": "1 of 6 run - and its answer is scale-dependent, not met at the element and partly met over the BCR. 5 of 6 unattempted, of which 2 (the dead-enzyme rescue and the CD34+ CTCF measurement) are bench experiments this project could specify today and one is already written up as exp-h3k9me3-and-ctcf-at-the-kmt2a-anchor-in-a-cd34-progenitor.",
  "WHAT THIS DOES NOT DO": "It does not lower this entry's confidence. Nothing here refuted the mechanism: the one falsifier that was run came back against the me1/me2-specific framing only at the wider BCR scale and in favour of it at the element. What it does is remove the entry's claim to having been tested, which it never made explicitly and which a reader would reasonably infer from six confidently-worded falsifiers sitting under a status of 'verified'.",
  "adjudicated_by": "king, cycle 61, 2026-09-22"
}