Oncofusion

start / graph / hyp-the-cxxc-address-is-a-contested-resource

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q79-cxxc-readers

Is the CpG address a contested resource? On current evidence, no - and the beat that went looking says so

also called contested CXXC address hypothesis · do CXXC readers compete with the KMT2A fusion

The idea tested here is that the dozen human proteins reading the same unmethylated CG addresses get in each other's way, so the KMT2A fusion's address depends on who else is in the room. After checking every family member, the evidence does not support it - and the one direct test showed the fusion staying put while a rival pushed the normal protein off.

The same thing, for a scientist

**REASONING SUPERSEDED 2026-09-21 (cycle 45).** This hypothesis' prior was lowered partly because 0 of 11 CXXC readers had been co-measured with a KMT2A fusion. That premise is FALSE - Kerry 2017 (PMID 28076791) measured CFP1/CXXC1 and KDM2B by ChIP-seq in SEM. **The conclusion does not change and the support for it improves**: MLL-AF4 sits on 20% of unmethylated CpG sites against CFP1's 50% and KDM2B's 89%, so the fusion leaves four fifths of the address unused and is placed by menin (MLL(N)-menin r² = 0.96) rather than by uCpG availability. The address is not contested because the fusion is not competing for it - a stronger reason than the absence of evidence originally cited. **Statement.** The unmethylated-CpG-island address that the KMT2A fusion reads through its CXXC domain is a limiting resource contested by the other eleven human CXXC-domain proteins, so the fusion's occupancy - and therefore its leukaemogenicity - depends on the expression and activity of its competitors rather than on its own affinity alone. **Prior probability assigned at 0.15, and this entry exists mainly to record that the hypothesis was tested and is currently unsupported**, because in this project an entry that establishes an absence precisely is worth more than one that asserts a plausible mechanism. **The null** (rule 84): of the order of 25,000 human CpG islands, at most eleven nuclear CXXC readers with different partner complexes, different secondary reader modules and different cell-type expression, binding different subsets at different loci and never interfering. Pairwise promoter-class overlap is expected by construction. **Where the evidence landed.** *Against the hypothesis, and this is the strongest datum:* the one published experiment that displaced a CXXC reader's chromatin neighbours at a KMT2A fusion's loci - KDM2A overexpression - **moved wild-type MLL and LEDGF and did not move the fusion** [E1]. *Also against:* TET2 mutation, the obvious methylome co-lesion, is **depleted** in KMT2A-rearranged AML (10.2% versus 16.4%, p = 0.0002) and absent from the infant index disease [E2][E3]; and TET3's CXXC prefers 5-carboxylcytosine, so at least one family member is not even reading the same ligand [E4]. *For the hypothesis, weakly:* CXXC5 is transcriptionally down-regulated in MLL-rearranged AML, though also in t(8;21) and GATA2-mutant AML, so not fusion-specific [E5]; and MBD1 and TET1 do have a documented reciprocal occupancy cycle, in heterochromatin rather than at CpG islands [E6]. **The reformulation that survives**, and which the beat recommends pursuing instead: *the address is set by the chemistry of the DNA, not by the census of proteins reading it.* The CXXC-swap experiment's own conclusion is that what distinguished functional from non-functional substitutions was **which CpG residues were protected from methylation** [E7], and the index cohort's authors independently conclude that the infant target cell may already have a permissive chromatin state [E3]. That is a methylome hypothesis, not a competition hypothesis, and nothing in this beat contradicts it.

Go deeper
Predictions, and how each fared
prediction of the contested-address hypothesisoutcome
Depleting a CXXC reader raises fusion occupancy at shared islandsuntested - 0 of 11 family members has an occupancy measurement in the same cells as a fusion 8
Overexpressing a CXXC reader lowers fusion occupancytested once, failed. KDM2A overexpression moved wild-type MLL and LEDGF, not the fusion 1
A methylome-altering co-mutation should be enriched with the fusionfailed, with the sign reversed. TET2 mutation is depleted in KMT2A-rearranged AML 2 and absent from 22 infant whole genomes 3
Competitors should be lost or silenced in the diseaseone weak hit. CXXC5 is down in MLL-rearranged AML - and also in t(8;21) and GATA2-mutant AML, so not fusion-attributable 5
Family members should read the same ligandfailed. TET3's CXXC prefers 5caC; MLL1's tolerates no modification at all 4
Family members should all be nuclearfailed for at least two. CXXC4's only UniProt location is cytoplasm by similarity; FBXL19's experimental record is three cytoplasmic supports to one nuclear

Five of six predictions fail or are unsupported, one is weakly and non-specifically supported, and the single direct test came out against. The honest prior after this beat is lower than the 0.15 it started with.

Why the hypothesis is still worth a node rather than a deletion

Three reasons.

1. The decisive experiment has never been run. Every failure above is either an indirect test or an overexpression artefact-prone one. The clean test - degrade one reader, ChIP the fusion - costs one degron line and two ChIPs and is specified in exp-cxxc-reader-depletion-versus-kmt2a-fusion-occupancy. Until that runs, 'unsupported' is the right word and 'refuted' is not.

2. The failure is informative about the menin-interface hypothesis, in its favour. If the fusion holds its address against an overexpressed rival CXXC reader 1, the fusion's occupancy is not set by CXXC affinity alone - it is buttressed by the AT-hooks and by the menin-LEDGF tether. A fusion whose address is robust to competition is a fusion whose address depends more on the protein-protein contacts this project's central hypothesis names. That is a positive result for the lab's position, arrived at by a beat dispatched to look for a challenge to it.

3. The reformulation is testable and cheaper. 'The address is the methylome' makes a prediction the competition version does not: the fusion's target set should be predictable from the cell of origin's methylation landscape, and should differ between infant and older-child KMT2A-rearranged disease in the direction the mutation data already hints at. That is a computational analysis over existing methylation data, and it is the follow-up this beat would prioritise.

What would change the verdict

Falsifiers, in the strict sense - results that would show the hypothesis right after this entry says it is unsupported:

  • Fusion occupancy at CpG islands rises measurably on acute degradation of CXXC1, KDM2A or KDM2B in a KMT2A-rearranged line.
  • Absolute copy-number measurement shows the nuclear CXXC readers in large molar excess over available unmethylated CpG islands, making the resource genuinely limiting.
  • A CXXC reader is found somatically mutated or deleted recurrently in KMT2A-rearranged disease in a modern cohort - currently only TET3, once, in the older-child arm 3.
  • The fusion's target set is shown to shift when a single competitor's level is changed, without any change in the methylome.

And the results that would strengthen the current verdict:

  • Fusion occupancy is unchanged by degradation of any reader.
  • Absolute copy numbers show islands in excess of readers.
  • An independent cohort reproduces TET2 depletion in KMT2A-rearranged AML.
Facts and figures
statementThe unmethylated-CpG-island address the KMT2A fusion reads through its CXXC domain is a limiting resource contested by the other eleven human CXXC-domain proteins, so fusion occupancy and leukaemogenicity depend on competitor expression and activity rather than on the fusion's own affinity alone.
originliterature-gap
prior probability0.15
posterior after this beat0.08
null hypothesisOf the order of 25,000 human CpG islands and at most eleven nuclear CXXC readers with different partner complexes, secondary reader modules and cell-type expression; they bind different subsets at different loci and never interfere. Pairwise overlap of a promoter class is expected by construction, so co-occupancy is not evidence of competition.
verdictUNSUPPORTED on current evidence. Five of six predictions fail or are untested; the one direct test showed the fusion NOT displaced. Not refuted, because the decisive experiment has never been run.
reformulation recommendedThe address is set by the chemical state of the DNA - which CpGs are protected from methylation - rather than by the census of proteins reading it. Supported by the CXXC-swap experiment's own conclusion and by the index cohort's permissive-chromatin interpretation, and testable computationally against existing methylation data.
prior reasoning supersededThe 0/11 co-measurement premise was false; the low prior is retained on the positive Kerry 2017 evidence instead.
predictions
[
  "Acute degradation of CXXC1, KDM2A or KDM2B raises KMT2A-fusion occupancy at shared CpG islands.",
  "Overexpression of a CXXC reader lowers fusion occupancy - TESTED ONCE FOR KDM2A AND FAILED.",
  "A methylome-altering co-mutation is enriched with the fusion - FAILED for TET2, with the sign reversed.",
  "CXXC readers are lost or silenced in KMT2A-rearranged disease - one weak, non-specific hit (CXXC5).",
  "The nuclear CXXC readers are in molar excess over available unmethylated CpG islands - never measured."
]
experiments
[
  "exp-cxxc-reader-depletion-versus-kmt2a-fusion-occupancy"
]
analyses
[
  "ana-cxxc-domain-family-census-human"
]
falsifiers
[
  "KMT2A-fusion occupancy at CpG islands is unchanged by acute degradation of any CXXC-family reader in a KMT2A-rearranged line - this would settle the null.",
  "Absolute copy-number measurement shows unmethylated CpG islands in molar excess over the nuclear CXXC readers, so the resource is not limiting.",
  "An independent cohort reproduces the depletion of TET2 mutation in KMT2A-rearranged leukaemia, confirming the methylome-co-lesion prediction runs the wrong way.",
  "A CXXC reader's overexpression at physiological range leaves fusion occupancy unchanged, as KDM2A overexpression already did."
]
How this could bite us back
  • If the fusion's address is robust to competition, then a strategy of raising a rival CXXC reader to evict the fusion will not work - and the one experiment resembling it confirms that, having moved wild-type MLL instead.
  • The reformulation is worse news for drug discovery than the original hypothesis. A protein competitor could be modulated; a methylation landscape set by the cell of origin before birth cannot be, and hypomethylating agents move it in the direction that OPENS more address for the fusion.
Still open
  1. Does acute degradation of a CXXC reader change KMT2A-fusion occupancy? Nobody has run it, and it is the only measurement that settles this.
  2. What is the molar ratio of nuclear CXXC readers to unmethylated CpG islands in a leukaemic blast? Without it, 'limiting resource' is not a well-posed claim.
  3. IDH2 is enriched in KMT2A-rearranged AML and inhibits TET2 catalytically. Is the methylome-co-lesion hypothesis correct with IDH2 as the gene rather than TET2, and why does this graph hold no IDH1 or IDH2 entry?
  4. Is the fusion's target set predictable from the methylation landscape of the cell of origin, and does it differ between infant and older-child KMT2A-rearranged disease?
Every source behind this entry (8)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    KDM2A overexpression reduces H3K36me2 and reduces chromatin occupancy of LEDGF and of wild-type MLL, but not of the MLL fusion, while antagonising MLL-associated leukaemogenesis.

    ASH1L Links Histone H3 Lysine 36 Dimethylation to MLL Leukemia. · Zhu L, Li Q, et al. · Cancer Discov 2016
    PMID 27154821DOI
    What this evidence cannot show

    Abstract read this session; the occupancy detail is held in this graph's existing mark-h3k36me2-3 and ledgf-axis entries, read before writing. Overexpression rather than physiological perturbation, H3K36me2 rather than CpG occupancy as the manipulated variable, murine MLL leukaemia. It is the only direct test available and it is not the ideal one.

    Evidence that argues against this (1)
    • A failure to displace under overexpression is the strongest form of this negative, since overexpression should saturate an address more thoroughly than any physiological change - which makes the null reading more secure, not less. Zhu L et al. 2016
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    TET2 alteration is significantly less frequent in KMT2A-rearranged AML than in KMT2A-wild-type AML in a 3,863-case single-assay series: 10.1% of 521 versus 16.4% of the remainder, p = 0.0002.

    PMID 41514668DOI
    What this evidence cannot show

    Targeted panel on adult AML with median age 62; 43.1% of the KMT2A-altered arm are duplications rather than fusions. Recomputed here as 53/521 versus 548/3,342, Fisher odds ratio 0.577, p = 1.6e-4. Says nothing about infant B-lineage disease.

    Evidence that argues against this (1)
    • The same assay finds IDH2 ENRICHED in KMT2A-rearranged AML (16.0% versus 10.4%, p < 0.0001), and IDH2 mutation inhibits TET2 catalytically - so a methylome-altering co-lesion IS enriched, and the hypothesis may have named the wrong gene rather than being wrong. Batayneh O et al. 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In 22 infant KMT2A-rearranged ALL whole genomes the string TET2 does not occur, and the only CXXC-family gene named anywhere in the study is TET3, among eleven epigenetic regulators mutated in 9/20 non-infant versus 3/22 infant cases (P = 0.04), with the authors concluding the infant target cell may already have a more permissive chromatin state.

    PMID 25730765DOI
    What this evidence cannot show

    Full-text search of one paper, 71,489 characters via efetch db=pmc after fullTextXML returned HTTP 500; supplementary tables not parsed. 0/22 bounds the TET2 rate below 13.6% by the rule of three and no lower. The permissive-chromatin conclusion is the authors' interpretation of a mutation-burden difference, not a chromatin measurement.

    Evidence that argues against this (1)
    • A lower epigenetic-mutation burden in infants is equally explained by less time for mutations to accumulate before a potent fusion transforms the cell, with no permissive-chromatin hypothesis needed. Andersson AK et al. 2015
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    TET3's CXXC domain binds unmethylated CpG but with highest affinity toward 5-carboxylcytosine, with a co-crystal of the CXXC-5caC-DNA complex, so at least one family member's preferred ligand differs from the KMT2A CXXC's.

    PMID 26774490DOI
    What this evidence cannot show

    Isolated domain binding and a co-crystal; genomic mapping in mouse neuronal cells. Whether 5caC is abundant at CpG islands in a haematopoietic cell is not addressed.

    Evidence that argues against this (1)
    • The systematic family comparison still places all twelve domains as CpG readers with graded selectivity rather than as readers of different marks, so 'different ligand' may overstate a difference in preference. Xu C et al. 2018
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CXXC5 mRNA is down-regulated in AML with MLL rearrangements, and also in t(8;21) and GATA2-mutant AML, with evidence for promoter-methylation silencing.

    PMID 25805812DOI
    What this evidence cannot show

    Adult AML expression cohort, mRNA only, no protein and no occupancy, and the change is shared with two non-KMT2A subtypes so it cannot be attributed to the fusion. It is the single best supporting datum for the hypothesis and it is weak.

    Evidence that argues against this (1)
    • If CXXC5 is a tumour suppressor as this study argues, its loss benefits the leukaemia for reasons unrelated to CpG-address competition, and the competition reading is superfluous. Kühnl A et al. 2015
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MBD1's CXXC3 domain enhances TET1 heterochromatin localisation and 5mC oxidation, and TET1 activity then displaces MBD1 from methylated DNA - the family's one documented reciprocal occupancy cycle.

    PMID 28449087DOI
    What this evidence cannot show

    Cell lines, fluorescent fusions, live imaging; pericentromeric heterochromatin rather than CpG islands; displacement inferred from imaging dynamics. No leukaemia and no KMT2A fusion.

    Evidence that argues against this (1)
    • This is recruitment then hand-off, which is co-operation followed by chemical succession rather than competition for a limiting resource. Zhang P et al. 2017
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the MLL-AF9 CXXC-swap experiment all substituted fusions still targeted Hoxa9, and what distinguished the leukaemogenic from the non-leukaemogenic substitutions was which CpG residues were protected from methylation, preventing H3K9me3 and enabling Hoxa9 expression.

    Functional specificity of CpG DNA-binding CXXC domains in mixed lineage leukemia. · Risner LE, Kuntimaddi A, et al. · J Biol Chem 2013
    PMID 23990460DOI
    What this evidence cannot show

    Murine MLL-AF9 myeloid system; three heterologous domains; mechanism read out at one locus. This is the strongest support in the literature for a methylome-centred reformulation and it is still a single-locus murine result.

    Evidence that argues against this (1)
    • The DNMT1 CXXC substitution worked while the CGBP one did not, although both bind unmethylated DNA - so an unmeasured property of the domains, not only the resulting methylation state, may be doing the work. Risner LE et al. 2013
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Of the eleven human CXXC-domain proteins other than KMT2A, none has a published measurement of chromatin occupancy in the same cells as a KMT2A fusion protein.

    PubMed E-utilities esearch record counts with querytranslation inspected and every returned title read · Oncofusion knowledge-graph project · NCBI E-utilities, queries run 2026-09-21; exact query strings given in the caveat of each item that uses this source 2026
    What this evidence cannot show

    Five bounded queries with every title read; exact strings are tabulated in absence-no-experiment-tests-cxxc-reader-competition-at-a-kmt2a-fusion-target. Titles and abstracts only, so a supporting ChIP panel would be invisible. 0/11 gives Wilson 95% 0-25.9% and rule of three <27.3%.

    Evidence that argues against this (1)
    • The ASH1L/KDM2A study measured fusion occupancy under KDM2A overexpression, one reagent short of the missing experiment, so the field has come closer to this than the counted absence suggests. Zhu L et al. 2016
Connected to