If the fusion's address requires unmethylated CpG, a hypermethylating co-mutation should be selected AGAINST - and IDH2 is selected FOR, which means the address is not the limiting resource
also called the methylome antagonism hypothesis · IDH2 and the uCpG address
The KMT2A fusion can only land on DNA that has no methyl marks. A mutation that adds methyl marks should therefore make life harder for it, and should be rarer in those leukaemias - but the data show it is commoner. Something about the simple picture is wrong, and this entry says which part.
The same thing, for a scientist
**The prediction and the observation point opposite ways, and the resolution is quantitative rather than rhetorical.** The prediction: the KMT2A CXXC domain binds unmethylated CpG and is blocked by 5-methyl-CpG on either strand [E1]; mutant IDH2 produces 2-hydroxyglutarate, inhibits the TET 5mC hydroxylases and yields a DNA hypermethylator phenotype [E2]; therefore IDH2 mutation should shrink the fusion's address space and be **selected against** in KMT2A-rearranged leukaemia. The observation: **IDH2 is significantly ENRICHED** in KMT2A-altered AML, 16.0% versus 10.4%, odds ratio 1.630, p = 0.0004 [E3]. **The hypothesis this entry proposes is that the prediction fails because unmethylated CpG is not a scarce resource for this fusion.** In SEM cells - the index fusion, human, B-lineage - every MLL-AF4 binding site lies on open unmethylated CpG, **and the fusion occupies only 20% of the available uCpG sites, against CXXC1's 50% and KDM2B's 89%**; what selects the fusion's subset is menin occupancy, correlating with fusion occupancy at r2 = 0.96 while CXXC1 correlates with menin at only 0.27 [E4]. **A fusion with a fivefold surplus of address is not address-limited, and moderate hypermethylation removes sites it was not using.** That is a direct, quantitative prediction of the menin-interface hypothesis and it explains the observation without special pleading. **Prior probability 0.25 that a hypermethylating co-mutation measurably reduces KMT2A-fusion chromatin occupancy**, and the entry is written to be killed: if 2-HG-driven hypermethylation preferentially hits menin-bound promoters rather than falling at random over the uCpG pool, the antagonism is real and the headroom is irrelevant.
The three readings, and how to tell them apart
| reading | what it says | what it predicts in an isogenic IDH2-R140Q KMT2A-fusion cell |
|---|---|---|
| A. Antagonism (the naive prediction) | hypermethylation blocks the CXXC contact, fusion occupancy falls | Bio-CAP-seq uCpG pool shrinks; fusion ChIP peaks fall in number and height; growth disadvantage |
| B. Headroom (this entry's hypothesis) | the fusion uses 20% of uCpG and is selected by menin, so losing uCpG it never used changes nothing | uCpG pool shrinks; fusion peaks essentially unchanged; no growth effect |
| C. Facilitation | 2-HG also inhibits KDM2A and KDM2B, two CXXC rivals occupying 89% and a large share of the same sites, so the crowd thins | KDM2B occupancy falls; fusion occupancy at shared sites rises; possible growth advantage |
All three are testable in one experiment - exp-idh2-r140q-versus-kmt2a-fusion-occupancy-in-sem - and the three readouts (uCpG pool, fusion peaks, competitor peaks) separate them cleanly. Nobody has run it in any system.
And note that reading C would make an approved drug look dangerous. Enasidenib lowers 2-HG, restores TET activity, restores unmethylated CpG and relieves KDM inhibition. Under C that is a mixed intervention; under A it is straightforwardly pro-leukaemic for the fusion. On no reading is an IDH2 inhibitor obviously the right thing to give an IDH2-mutant KMT2A-rearranged patient on this graph's own model, and that deserves to be said out loud.
Four ways this hypothesis is already weaker than it looks
- The enrichment may not be about fusions at all. 43.1% of the KMT2A-altered arm in the source series are KMT2A duplications, and the table is not stratified 3.
- The IDH2 and TET2 rows are not independent. IDH and TET2 mutations are mutually exclusive, and the summed hypermethylator burden is 26.1% in KMT2A-altered against 26.8% in wild-type - essentially identical 5. On that arithmetic there is no enrichment of hypermethylating lesions at all, only a redistribution, and the hypothesis has nothing to explain.
- IDH1 is flat. Same oncometabolite, same phenotype, odds ratio 0.876, p = 0.53 3. A 2-HG-mediated story predicts both paralogues and gets one.
- The observation is adult AML and the disease is infant B-ALL. No infant or paediatric KMT2A-rearranged cohort reporting IDH status was found in this beat, so the index disease has no estimate in either direction.
Taken together these four say the hypothesis may be answering a question that does not exist. That possibility is written into the falsifiers rather than left for a later cycle to discover.
| statement | A hypermethylating co-mutation does NOT measurably antagonise the KMT2A fusion's chromatin occupancy, because the fusion uses only ~20% of available unmethylated CpG and its site selection is dominated by menin rather than by uCpG availability. The naive antagonism prediction fails because unmethylated CpG is not the limiting resource. |
|---|---|
| origin | data-pattern |
| prior probability | 0.25 |
| prior note | 0.25 is the prior that a hypermethylating co-mutation measurably REDUCES fusion occupancy, i.e. that reading A is right. Reading B (headroom, no effect) is this entry's favoured reading at ~0.55 and reading C (facilitation via KDM inhibition) at ~0.20. These are stated so a later cycle can score them. |
predictions
[ "In an isogenic IDH2-R140Q versus wild-type KMT2A-fusion cell, the Bio-CAP-seq unmethylated-CpG pool shrinks while the number and height of fusion ChIP peaks are essentially unchanged.", "The 5mC gained under IDH2-R140Q falls at random over the uCpG pool with respect to menin occupancy - i.e. menin-bound promoters are not preferentially methylated.", "IDH2-mutant KMT2A-rearranged patients do not differ in menin-inhibitor response from IDH2-wild-type KMT2A-rearranged patients.", "The IDH2 enrichment does not survive restriction to KMT2A fusions (as opposed to duplications)." ]
experiments
[ "exp-idh2-r140q-versus-kmt2a-fusion-occupancy-in-sem" ]
analyses
[ "ana-the-methylome-comutation-table-recomputed-and-the-p-value-that-does-not-match-its-own-paper", "ana-the-fusion-sits-on-one-fifth-of-the-unmethylated-cpg-sites-and-menin-picks-which" ]
falsifiers
[ "5mC gain under mutant IDH2 is enriched at menin-bound promoters rather than random over the uCpG pool - the headroom argument then fails and antagonism is real.", "Fusion ChIP peak number or height falls significantly in an isogenic IDH2-R140Q background - reading A wins directly.", "The IDH2 enrichment disappears in a fusion-only cohort - the hypothesis has nothing to explain and should be withdrawn rather than rescued.", "The summed hypermethylator-lesion burden turns out to differ significantly between KMT2A-altered and wild-type arms in patient-level data - the redistribution explanation fails and a real enrichment needs explaining.", "KDM2B occupancy falls and fusion occupancy at shared sites rises under mutant IDH2 - reading C wins, and an IDH2 inhibitor becomes a candidate KMT2A therapy for a reason nobody has proposed." ]
cxxc dna kd conditions
{
"_why": "Added by king sweep 2026-09-21 (cycle 46). This entry quotes the KMT2A CXXC-DNA 44 uM. That number is a property of five choices, not of the molecule, and quoting it bare is how this graph manufactured a 1,300-fold 'discrepancy' with the menin contact.",
"value": "44 uM",
"construct": "isolated CXXC domain, KMT2A 1147-1203, untagged, 15N-labelled",
"ligand": "DNA pal oligonucleotide, one CpG element",
"method": "NMR titration",
"buffer": "50 mM phosphate, pH 7.1, 300 mM NaCl, 2 mM DTT, 25 C",
"nacl_mM": 300,
"source_pmid": "20010842",
"source_quote": "\"Determination of the dissociation constants by NMR for the CXXC domain (MLL 1147-1203) and mutants was performed by titration of 15N-labeled protein with concentrated solutions of DNA pal oligonucleotide. Experiments were carried out in 50 mM phosphate buffer (pH 7.1), 300 mM NaCl, 2 mM DTT at 25 C.\" -- Cierpicki 2010 NSMB, Methods, fetched and read from PMC2908503 on 2026-09-21. The result sentence reads \"The wild type protein yielded a Kd of 44 uM\".",
"this_is_the_high_salt_end": "300 mM NaCl is roughly twice physiological ionic strength and SIX TIMES the 50 mM the menin-MBM1 53 nM was measured in. The same domain measured tagless at 150 mM NaCl gives 5.7 uM -- 7.7-fold tighter from salt alone.",
"do_not_compare_naively": {
"against": "menin-MBM1 53 nM (PMID 20961854), measured at 50 mM NaCl by FP",
"best_matched_pair": "108-fold, not ~1,000-fold",
"see": "ana-the-menin-affinity-comparison-was-made-at-one-sixth-the-salt"
},
"added": "2026-09-21"
}- The favoured reading of this hypothesis is that the methylome co-mutation does not matter. That conclusion is convenient for the project's central hypothesis, and convenient conclusions from single retrospective series are exactly what this graph's own rules warn against. The prior is stated numerically (0.55) so that a later cycle can mark it wrong.
- Under reading A or reading C, an IDH2 inhibitor given to an IDH2-mutant KMT2A-rearranged patient restores unmethylated CpG and could expand the fusion's address space. An approved drug that plausibly helps the fusion, given for an orthogonal indication, is the most actionable hazard in this beat and nobody has looked for it.
- Does 5mC gain under mutant IDH2 fall at random over the unmethylated-CpG pool, or preferentially at menin-bound promoters? This single question decides between the three readings.
- Is there any infant or paediatric KMT2A-rearranged cohort with IDH status? Without one, the entire hypothesis is being argued in adult AML about an infant disease.
- Does menin-inhibitor response differ by IDH2 status in the KMT2A-rearranged arms of AUGMENT-101 or SAVE? IDH status is collected on those panels and has not been reported as a covariate.
- If the fusion has fivefold headroom in uCpG, why does it not use more of it? The answer is presumably menin and LEDGF availability, and nobody has measured the stoichiometry.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The KMT2A CXXC (MT) domain binds DNA containing unmethylated CpG, and 5-methyl-CpG on one or both strands precludes efficient interaction - so the fusion's DNA contact is directly opposed by CpG methylation.
The MT domain of the proto-oncoprotein MLL binds to CpG-containing DNA and discriminates against methylation. · Birke M, Schreiner S, Garcia-Cuellar MP, et al. · Nucleic Acids Res 2002What this evidence cannot show
Recombinant GST-MT in vitro with defined oligonucleotides; no fusion protein, no cell, no methylation manipulation. The reported SPR KD (~3.3e-8 M) is for a dimerising GST fusion and is ~1,300-fold tighter than the ~44 uM NMR figure this graph quotes elsewhere for the isolated domain, a discrepancy neither entry currently reconciles. This hypothesis uses only the qualitative methylation discrimination. RESOLVED 2026-09-21 by Q82. This caveat was right to flag the discrepancy and its GST-dimerisation attribution is now bounded rather than confirmed. Ten published measurements of this contact form a continuous series from 44 uM to 33 nM. A same-laboratory, same-construct (1147-1203), same-oligonucleotide pair puts GST plus a 300-to-150 mM salt drop at 5.5-fold, and a tagless-against-tagless pair (44 uM at 300 mM NaCl against 5.7 uM at 150 mM NaCl) puts about 8-fold on the salt alone - so GST dimerisation explains very little of the gap. The remaining 19-fold, between GST-1147-1203 at 0.64 uM and GST-1147-1244 at 33 nM, confounds the 41-residue post-CXXC basic patch, six CpGs against two, a 52-bp duplex against a 12-mer, and an immobilised kinetic geometry, and is unresolved. See ana-cxxc-dna-affinity-ladder. This entry's own use of the paper - qualitative methylation discrimination - is unaffected.
Evidence that argues against this (2)
- In a cell the fusion is also held by AT-hooks, menin and LEDGF, so methylation discrimination by the isolated domain does not establish that a physiological rise in 5mC displaces the fusion - and the genome-wide data say site selection tracks menin, not uCpG availability. Kerry J et al. 2017
- The reconciliation this caveat called for was published in 2008, by the same laboratory that produced the 44 uM figure, and this graph did not cite it. Erfurth et al. state the SPR value is about two orders of magnitude tighter than their ITC and attribute it to GST dimerisation and bivalent binding, and the same paper gives the MT domain boundaries as MLL 1147-1244 and an ITC ladder of 3.5 to 32 uM across five Hoxa9-derived oligonucleotides at fixed construct, salt and method. Erfurth FE et al. 2008
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Mutant IDH1 and IDH2 produce 2-hydroxyglutarate, which competitively inhibits alpha-ketoglutarate-dependent dioxygenases including the TET 5mC hydroxylases and the histone demethylases, and IDH-mutant AML shows global DNA hypermethylation with a specific hypermethylation signature and impaired TET2 catalytic function.
Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of alpha-ketoglutarate-dependent dioxygenases. · Xu W, Yang H, Liu Y, et al. · Cancer Cell 2011What this evidence cannot show
Biochemistry and ectopic expression; cellular correlation in glioma. Read from abstract only. No inhibition constants for individual dioxygenases and no ranking of TET against KDM inhibition, so the relative size of the two arms of this hypothesis is unknown. Nothing in a KMT2A-rearranged system.
Evidence that argues against this (1)
- Because 2-HG inhibits the JmjC histone demethylases as well as TET, and KDM2A and KDM2B are CXXC-family competitors of the fusion, the same lesion plausibly helps the fusion at the reader-competition level while hindering it at the DNA level - so the hypothesis's own premise does not fix a direction. Kerry J et al. 2017
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In adult AML, IDH2 alteration is significantly ENRICHED rather than depleted in KMT2A-altered cases (16.0% of 521 vs 10.4% of 3,342, recomputed odds ratio 1.630, p = 4.1e-4, matching the paper's Table 2 value of 0.0004), while IDH1 is flat (6.7% vs 7.6%, p = 0.53) - so the naive antagonism prediction fails and fails with the sign reversed.
Characterizing KMT2A Rearrangement in Acute Myeloid Leukemia: A Comprehensive Genomic Study. · Batayneh O, Moein M, Naji NS, et al. · Cancers (Basel) 2026What this evidence cannot show
Adult AML, median age 62, retrospective commercial panel, no outcome data, ~17 unadjusted comparisons, and 43.1% of the KMT2A-altered arm are duplications rather than fusions with no stratification. The paper's abstract states p < 0.0001 and its Table 2 states 0.0004; the recomputation reproduces the table. Read in full text from PMC12784994 on 2026-09-21. A co-occurrence frequency is not a selection coefficient: mutation order, clonal architecture and detection thresholds all intervene between 'selected for' and 'found together'.
Evidence that argues against this (1)
- IDH and TET2 mutations are mutually exclusive, and summing the two hypermethylator lesions gives 26.1% in KMT2A-altered against 26.8% in wild-type - so there may be no enrichment of hypermethylating lesions at all, only a redistribution, in which case this hypothesis has nothing to explain. Figueroa ME et al. 2010
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In SEM cells (human t(4;11) KMT2A::AFF1 B-ALL) all MLL-AF4 binding occurs at open unmethylated-CpG regions but the fusion occupies only 20% of uCpG sites against CFP1's 50% and KDM2B's 89%, and fusion occupancy correlates with menin occupancy at r2 = 0.96 while menin and CFP1 correlate at only 0.27 - so unmethylated CpG is abundant relative to the fusion's use of it and menin, not uCpG availability, selects the address.
MLL-AF4 Spreading Identifies Binding Sites that Are Distinct from Super-Enhancers and that Govern Sensitivity to DOT1L Inhibition in Leukemia. · Kerry J, Godfrey L, Repapi E, et al. · Cell Rep 2017What this evidence cannot show
Read in full text from PMC5263239 on 2026-09-21; the percentages are in the running text and Figure S1F, not the abstract. One cell line. The uCpG denominator is defined by Bio-CAP, itself a recombinant CXXC affinity reagent, so the 20/50/89 split is measured against a CXXC-defined reference; the ranking is more robust than the absolute values. The two r2 values are computed over different peak sets and are not a like-for-like comparison. Steady-state occupancy only - no reader was depleted and no methylation was perturbed.
Evidence that argues against this (1)
- The same paper shows the menin interaction is sufficient but NOT necessary for recruitment in a tethering system, so 'menin selects the address' overstates what an r2 of 0.96 licenses - and headroom in the uCpG pool says nothing about whether a hypermethylating insult falls at random across it. Kerry J et al. 2017
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Because IDH and TET2 mutations are mutually exclusive in AML, the IDH2 enrichment and the TET2 depletion measured in the same KMT2A comparison are not independent, and the summed hypermethylator-lesion frequency is 26.1% in KMT2A-altered against 26.8% in KMT2A-wild-type cases.
Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. · Figueroa ME, Abdel-Wahab O, Lu C, et al. · Cancer Cell 2010What this evidence cannot show
The summation is valid only under strict exclusivity, is computed from published marginals, has no interval and is NOT a statistical test - it is a hypothesis about what the table may mean. Exclusivity itself is reported in a 2010 adult AML cohort without a test statistic in the abstract, and co-mutated cases are reported in later series. Read from abstract only.
Evidence that argues against this (1)
- If exclusivity is imperfect, or if double-mutant cases concentrate in one arm, the near-identity of the two sums is an artefact of the assumption and not a property of the data - and it cannot be checked from the published marginals. Batayneh O et al. 2026
THE HYPOTHESIS CANNOT CURRENTLY BE TESTED IN THE INDEX DISEASE, and this entry says why in one sentence: the co-mutation frequency it turns on has never been reported in an infant KMT2A-rearranged cohort. If the infant frequency is near zero, the hypermethylating route to challenging the CXXC address closes for this disease whatever the adult numbers say. The test costs an afternoon with an open-access supplementary table.
The one experiment that discriminates antagonism, headroom and facilitation in a single design.
The enrichment that falsifies the naive antagonism prediction with the sign reversed.
Asks directly whether unmethylated CpG is a limiting resource for the fusion; the measured answer is that the fusion uses a fifth of it.
The headroom reading rests entirely on the 20/50/89 measurement and the r2 = 0.96 menin correlation.
- related to →Is the CpG address a contested resource? On current evidence, no - and the beat that went looking says so
The methylome version of the same question: last cycle asked whether rival READERS contest the address, this asks whether the SUBSTRATE can be taken away.
The mechanism that makes the direction ambiguous: 2-HG raises 5mC and impairs the fusion's CXXC rivals at the same time.
The deflating possibility written into the falsifiers: the summed hypermethylator burden is 26.1% vs 26.8% and there may be nothing to explain.
The observation the hypothesis has to account for, and the sign is the wrong way round for the naive prediction.
Same resolution for the same caveat carried by this hypothesis entry.