Oncofusion

start / graph / gene-kdm6b

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q44-kmt2-family

KDM6B (JMJD3) - UTX's paralogue, which this graph has argued about for cycles without writing

also called JMJD3 · KIAA0346 · Kdm6b (mouse)

KDM6B removes the repressive mark that keeps leukaemia genes switched off, and it is the twin of UTX, the protein this graph's menin-resistance story is built on. In mice it is needed for leukaemia that starts before birth and not for leukaemia that starts in an adult - which is the same developmental split this disease shows.

The same thing, for a scientist

KDM6B (UniProt O15054, 1,643 aa) is the H3K27me3 demethylase paralogous to KDM6A/UTX; both sit in the MLL3/MLL4 COMPASS-like complex [E5]. This graph writes `gene-kdm6a` and `resistance-mll34-utx-tumour-suppressor-failure` and has had no entry for the paralogue. Three things make it load-bearing. **(1) Developmental context-dependence.** Kdm6b is required for effective fetal-derived T-ALL and adult-derived M5 AML, **but not vice versa** - a developmental-origin dependency that is exactly the axis this project's infant disease sits on [E1]. **(2) A HOX-directed pharmacology.** GSK-J4 raised global H3K27me3, reduced proliferation and colony formation of primary AML cells and lines, synergised with cytarabine, attenuated disease in a human AML xenograft, and **abrogated expression of HOX genes with increased H3K27me3 at their transcription start sites** [E2] - the same program a menin inhibitor is supposed to shut down, reached by the opposite chromatin move. **(3) It is coupled to KDM6A in DNA repair.** Deficiency of either KDM6A or KDM6B impairs DNA-damage-repair transcriptional activation, and KDM6A loss sensitises AML to PARP and BCL2 inhibition [E3]. **The tension this creates for this graph is explicit and unresolved**: the menin-resistance entry says the drug works by letting UTX *activate* genes, while the GSK-J4 literature says blocking KDM6 demethylation *kills* AML by repressing HOX. Both are published; neither cites the other.

Go deeper
The developmental split, and why it belongs in an infant-leukaemia graph

1 is a conditional-deletion study in mouse HSCs with two leukaemia arms. Loss of Kdm6b depletes phenotypic and functional HSCs in adult mice and is required for HSC self-renewal under inflammatory and proliferative stress. Mechanistically, Kdm6b loss puts HSCs into a pro-differentiation poised state through increased AP-1 (Fos, Jun) and immediate-early-response gene expression - and the authors state these expression changes occurred independently of chromatin modifications, which is a striking result for a demethylase and is the kind of thing this graph's catalytic-versus-scaffold arguments should absorb. Targeting AP-1 restored function in Kdm6b-deficient HSCs.

The leukaemia arms give the split: Kdm6b is required for effective fetal-derived T-ALL and adult-derived AML, but not vice versa. So the same gene is required or dispensable depending on the developmental origin of the leukaemia, not on its lineage. This project's entire premise is that infant KMT2A-rearranged leukaemia arises in a fetal cell and behaves differently from the adult disease with the same fusion. A published, gene-level demonstration that a chromatin dependency flips with developmental origin is directly relevant and no KMT2A-rearranged arm exists in it.

The contradiction with this graph's own menin-resistance story

resistance-mll34-utx-tumour-suppressor-failure holds that menin inhibition works partly by allowing MLL3/4-UTX to activate a tumour-suppressive program, so that cells which cannot do so are resistant. UTX activates by removing H3K27me3.

2 reports the opposite therapeutic direction with the same enzyme family. GSK-J4, an H3K27 demethylase inhibitor, increases global H3K27me3, and kills AML cells while abrogating HOX gene expression with increased H3K27me3 at HOX transcription start sites. KDM6B mRNA is upregulated in AML and correlates with poor survival.

So one literature says KDM6-mediated demethylation is the therapeutic effector and its loss confers resistance; the other says KDM6-mediated demethylation sustains the oncogenic HOX program and inhibiting it is therapeutic. Possible reconciliations - different paralogues, different loci, different disease contexts, KDM6A versus KDM6B non-equivalence, or GSK-J4 off-target activity - are all plausible and none has been tested. This is recorded here because a graph that holds both claims without noticing is exactly the failure mode the contradiction checker exists for.

Reagent caveat that bears on both. GSK-J4 is a KDM6 inhibitor, not a KDM6B-selective one: 2's own methods call it 'a KDM6B-specific inhibitor' while the compound's family activity is against KDM6 demethylases generally, and 3 treats KDM6A and KDM6B as jointly regulating DNA repair. Attributing a GSK-J4 result to KDM6B rather than KDM6A is naming and not measurement - the same problem this project logged for CK2 and CX-4945.

What is NOT here

No study connects KDM6B to a KMT2A fusion protein. (KDM6B OR JMJD3) AND (KMT2A OR MLL-rearranged OR "MLL fusion") AND leukemia returns 6 records, title-checked; the only one in a KMT2A-rearranged system is 4, an AURKA study in THP-1 (KMT2A::MLLT3) in which alisertib de-represses KDM6B and drives monocytic differentiation, and GSK-J4 blocks PMA-driven differentiation of the same cells. That is a differentiation result in one line, not a fusion interaction. The complex forms ASCOM, "MLL3/4 complex" and "COMPASS-like complex" were run as required by rule 64; the COMPASS-like query returns 37 records dominated by plant biology (Arabidopsis, rice, sorghum, peony), which is the ordinary-word contamination that rule warns about, and none of them is a KDM6B-plus-KMT2A-fusion paper 6.

So the counted absence is: no co-immunoprecipitation, proximity-labelling, structure or degron experiment places KDM6B with any KMT2A fusion protein. Same shape as the WDR5 verdict already in this graph.

Facts and figures
uniprotO15054
length AA1643
gene primary verifiedKDM6B
accession resolutiongene_exact:KDM6B AND organism_id:9606 AND reviewed:true returned exactly one row; gene_primary equals KDM6B; length 1643 asserted
paralogueKDM6A/UTX (gene-kdm6a in this graph)
catalytic activityH3K27me3/me2 demethylase
developmental context dependencerequired for fetal-derived T-ALL and adult-derived M5 AML but not vice versa (PMID 30936419)
chemical matterGSK-J4 - a KDM6-family inhibitor, NOT KDM6B-selective; prodrug ester with documented cross-reactivity on other JmjC enzymes
fusion association evidencenone published in either direction
unresolved tensionthe menin-resistance entry in this graph makes KDM6 demethylation the therapeutic effector; the GSK-J4 literature makes inhibiting KDM6 demethylation the therapy. Both published, neither cites the other.
pubmed counts 2026 09 20
ASCOM AND (leukemia OR le…3.000ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2): 3(KDM6B OR JMJD3) AND (KMT…6.000(KDM6B OR JMJD3) AND (KMT2A OR MLL-rearranged OR MLL fusion) AND leukemia: 6KDM6B AND (acute myeloid …17.0KDM6B AND (acute myeloid leukemia OR myelodysplastic): 17
pubmed counts 2026 09 20 — the numbers behind the chart
{
  "(KDM6B OR JMJD3) AND (KMT2A OR MLL-rearranged OR MLL fusion) AND leukemia": 6,
  "KDM6B AND (acute myeloid leukemia OR myelodysplastic)": 17,
  "ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2)": 3
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.0632,
  "chronos_mean_other_leukaemia": -0.0865,
  "difference": 0.0233,
  "pan_essentiality_all_lines": -0.0386,
  "selectivity_rank": 12113,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • The two literatures predict opposite treatments. A clinician acting on this graph could reasonably conclude either that KDM6 activity should be inhibited or that it must be preserved, and the graph currently contains both without saying so.
  • GSK-J4 is not selective, so every KDM6B attribution in this entry is a family attribution. A result credited to KDM6B may be KDM6A, and in this disease KDM6A is the gene with the resistance story.
  • KDM6B loss depletes HSCs and blocks stress self-renewal. Inhibiting it in a child whose marrow is already replaced by leukaemia targets the same stress-response machinery the normal graft needs to recover.
  • A non-catalytic mechanism would make the whole inhibitor class inert where it matters. The genetic phenotype in HSCs is reported as chromatin-modification-independent; if that generalises to leukaemia, a demethylase inhibitor cannot reproduce it and an efficacy signal from GSK-J4 must be coming from somewhere else.
Still open
  1. Does GSK-J4 help or hurt in a KMT2A-rearranged cell? The HOX-suppressing result would predict help, the menin-resistance switch model would predict harm, and no KMT2A-rearranged line appears in either literature. A five-line panel would settle the direction in a fortnight.
  2. Is the fetal-versus-adult dependency split reproduced with a KMT2A fusion? Kdm6b is required for fetal-derived T-ALL and adult-derived AML but not the converse. Nobody has asked the question with MLL::ENL or MLL-AF9 in a temporally controlled model, although exactly such a model now exists and was used for the MLL3 fetal-barrier work.
  3. Do KDM6A and KDM6B do the same thing here? The DNA-repair result treats them together; the PARP/venetoclax result is KDM6A-only; GSK-J4 cannot separate them. A KDM6B-selective degrader does not exist and would resolve most of this entry's uncertainty.
  4. Why do the Kdm6b HSC expression changes occur independently of chromatin modification? If the leukaemia-relevant function is non-catalytic, every catalytic inhibitor in this area is testing the wrong hypothesis.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Loss of Kdm6b depletes phenotypic and functional HSCs in adult mice and is necessary for HSC self-renewal in response to inflammatory and proliferative stress, producing a pro-differentiation poised state through increased AP-1 (Fos, Jun) and immediate-early-response gene expression that occurred independently of chromatin modifications; Kdm6b is required for effective fetal-derived T-ALL and adult-derived M5 AML but not vice versa.

    Kdm6b regulates context-dependent hematopoietic stem cell self-renewal and leukemogenesis. · Mallaney C, Ostrander EL, Celik H, et al. · Leukemia 2019
    PMID 30936419DOI
    What this evidence cannot show

    Abstract fetched and read 2026-09-20; full text (PMC6773521) not obtained in this beat, so the leukaemia drivers used in the T-ALL and M5 AML arms are not recorded here. **Neither arm is KMT2A-rearranged as far as the abstract states.** Mouse, conditional deletion, which permits compensation an acute degron would not. The claim that the expression changes occur 'independently of chromatin modifications' is a negative result about the assays performed and is the kind of statement that a more sensitive method could overturn.

    Evidence that argues against this (1)
    • A demethylase whose principal reported phenotype occurs independently of chromatin modification is at least as likely to be acting as a scaffold, in which case a catalytic inhibitor such as GSK-J4 would not reproduce the genetic phenotype - and the GSK-J4 literature nevertheless reports strong anti-leukaemic activity, which the two accounts do not reconcile. Li Y et al. 2018
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    KDM6B mRNA is upregulated in AML and positively correlated with poor survival; GSK-J4 increased global H3K27me3, reduced proliferation and colony formation of primary AML cells and AML lines, induced apoptosis and cell-cycle arrest in Kasumi-1, synergised with cytosine arabinoside, attenuated disease progression in a human AML xenograft, and abrogated expression of critical cancer-promoting HOX genes with ChIP-qPCR-validated increased H3K27me3 enrichment at their transcription start sites.

    PMID 29594337DOI
    What this evidence cannot show

    Abstract fetched and read 2026-09-20; full text (PMC5948279) not obtained. **GSK-J4 is called 'KDM6B-specific' by these authors and it is not** - it inhibits KDM6 demethylases including KDM6A, and it is a prodrug ester whose cell-permeable form has documented off-target activity on other JmjC enzymes. Kasumi-1 is RUNX1::RUNX1T1, not KMT2A-rearranged; no KMT2A-rearranged line is named in the abstract. The HOX result is the most interesting element for this project and it is ChIP-qPCR at selected loci, not genome-wide.

    Evidence that argues against this (1)
    • This graph's own menin-resistance entry holds the opposite therapeutic polarity: that KDM6-family demethylation is the ACTIVATING arm through which menin inhibition kills, and that losing it confers resistance. Blocking KDM6 demethylation cannot be both the therapy and the resistance mechanism in the same cell, and nobody has run the experiment that would separate them. Soto-Feliciano YM et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KDM6 demethylase function regulates DNA-damage-repair gene expression in AML: KDM6 expression is regulated by genotoxic stress, and deficiency of KDM6A (UTX) or KDM6B (JMJD3) impairs DDR transcriptional activation and compromises repair potential; olaparib reduced engraftment of KDM6A-mutant AML patient-derived xenografts, higher KDM6A expression correlated with venetoclax tolerance, and loss of KDM6A increased mitochondrial activity and BCL2 expression and sensitised AML cells to venetoclax, with dual PARP and BCL2 targeting superior to either monotherapy.

    PMID 36720973DOI
    What this evidence cannot show

    Abstract fetched and read 2026-09-20; the fetched record carries no PMCID so full text was not obtained through routes 1-4. The therapeutic arm is about **KDM6A**, not KDM6B - KDM6B appears in the mechanistic DDR claim and not in the PARP/BCL2 synthetic-lethality result. Adult AML throughout; no KMT2A-rearranged subgroup and no paediatric material identified in the abstract. The venetoclax correlation is expression-based and correlative.

    Evidence that argues against this (1)
    • The paper notes that although acquired KDM6A loss-of-function is implicated in chemoresistance, a significant percentage of relapsed AML has UPREGULATED KDM6A - so KDM6 status at relapse is bidirectional and a therapeutic strategy keyed to loss will be wrong in a substantial fraction of exactly the patients it targets. Mallaney C et al. 2019
  4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    In THP-1, a KMT2A::MLLT3 (MLL-AF9) AML line, PMA-induced monocytic differentiation involves downregulation of AURKA; the AURKA inhibitor alisertib accelerates KDM6B expression by dissociating AURKA and YY1 from the KDM6B promoter and induces monocytic differentiation, while the KDM6B inhibitor GSK-J4 perturbs PMA-mediated differentiation of the same cells.

    PMID 29477140DOI
    What this evidence cannot show

    Abstract fetched and read 2026-09-20; full text (PMC5974621) not obtained. **A single cell line, THP-1, with PMA-forced differentiation** - a pharmacological differentiation model that does not represent leukaemic biology in a patient. The KMT2A fusion is never manipulated and never assayed; THP-1's fusion status is why this entry cites the paper, not something the paper tests. GSK-J4 is not KDM6B-selective. Graded weak deliberately.

    Evidence that argues against this (1)
    • Within this same paper the two pharmacological arms point in opposite directions - raising KDM6B drives differentiation, inhibiting KDM6B blocks differentiation - which is the opposite polarity from the AML studies where GSK-J4 is anti-leukaemic. A cell line plus a phorbol ester is not a basis for choosing between them. Li Y et al. 2018
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt reviewed human record fetched 2026-09-20 gives KDM6B as O15054, 1,643 aa, gene_primary KDM6B, synonyms JMJD3 and KIAA0346, protein name 'Lysine-specific demethylase 6B ... ([histone H3]-trimethyl-L-lysine(27) demethylase 6B)'; the KDM6 family demethylases KDM6A and KDM6B are the H3K27me3-removing partners of the MLL3/MLL4 COMPASS-like complexes.

    What this evidence cannot show

    gene_exact:KDM6B AND organism_id:9606 AND reviewed:true returned exactly one row whose gene_primary equals KDM6B; the length 1,643 is asserted here because a value one might reach from recall is wrong. Whether KDM6B, as opposed to KDM6A, is a stable component of the MLL3/4 complex is not established by the database record and the published complex compositions vary; the family relationship is what is asserted.

    Evidence that argues against this (1)
    • The quantitative survey of the SET1/MLL complexes that assigned PTIP and PA1 to MLL3/4 did not resolve KDM6B as a stoichiometric subunit, so 'KDM6B is in the MLL3/4 complex' is a statement inherited from the UTX literature and applied to the paralogue. van Nuland R et al. 2013
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    PubMed run 2026-09-20: `(KDM6B OR JMJD3) AND (KMT2A OR MLL-rearranged OR MLL fusion) AND leukemia` returns 6 records, title-checked, of which only one is in a KMT2A-rearranged system and none reports a physical association between KDM6B and a KMT2A fusion protein; the complex forms required by rule 64 were also run - `ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2)` returns 3 and `"MLL3/MLL4 complex" OR "MLL3/4 complex" OR "COMPASS-like complex"` returns 37 dominated by plant biology.

    What this evidence cannot show

    A counted absence bounded by the symbol set KDM6B and JMJD3 only; the retired-symbol risk is low for this gene but not zero. Titles were checked; abstracts were read only for records judged possibly relevant. A co-IP in a supplementary figure of a paper whose title and abstract do not name KDM6B would be invisible. The repository was grepped before recording this: no entry in this graph cites a KDM6B-plus-KMT2A-fusion experiment.

    Evidence that argues against this (1)
    • This project has previously declared a literature empty while citing the disproving paper elsewhere in the graph, and the complex-form route is exactly how such a paper hides. The COMPASS-like query's plant contamination means its 37 records give weak assurance, and a paper framing a result at 'the MLL3/4-UTX complex' without naming KDM6B would not be retrieved by either query. Soto-Feliciano YM et al. 2023
Connected to