Put mutant IDH2 into SEM and ask three questions at once: does the unmethylated-CpG pool shrink, does the fusion move, and does its competitor move
also called IDH2-R140Q in SEM · the methylome antagonism experiment
Take the standard human infant-leukaemia cell line, switch on the mutant IDH2 enzyme that drives DNA over-methylation, and then map three things: where methyl-free DNA is, where the fusion sits, and where its main rival sits. Three readouts, one experiment, and it distinguishes three competing explanations that argument alone cannot.
The same thing, for a scientist
**Design.** Isogenic SEM (human t(4;11) KMT2A::AFF1 B-ALL) with doxycycline-inducible IDH2-R140Q versus catalytically dead control, plus a wild-type-IDH2 arm; confirm 2-HG accumulation by mass spectrometry as the pharmacodynamic gate before any sequencing is done. After a defined exposure (2 and 6 weeks, because methylation drift is slow and a two-day experiment would return a false null), run in the same cells: **(i) Bio-CAP-seq** for the unmethylated-CpG pool - the same assay used to establish the 20/50/89 baseline, so the comparison is like-for-like [E1]; **(ii) ChIP-seq or CUT&RUN for MLL(N) and AF4(C)** for fusion occupancy; **(iii) ChIP-seq for KDM2B and menin**, the competitor and the selector; **(iv) whole-genome bisulfite sequencing** as an orthogonal methylation readout that does not depend on a CXXC reagent - this matters because the baseline denominator is CXXC-defined and a CXXC-defined denominator cannot fully test a hypothesis about CXXC readers. **Readout and the three-way discrimination.** If the uCpG pool shrinks and fusion peaks fall, the antagonism reading wins. If the pool shrinks and fusion peaks are unchanged, the headroom reading wins and the menin-interface hypothesis gains a direct quantitative confirmation. If KDM2B occupancy falls and fusion occupancy at shared sites rises, the facilitation reading wins and an approved IDH2 inhibitor becomes a potential hazard rather than a candidate therapy [E2]. **The decisive secondary analysis** is whether 5mC gain is random over the uCpG pool with respect to menin occupancy, or concentrated at menin-bound promoters; that single contrast decides whether the fusion's fivefold headroom protects it [E1][E3]. **Cost tier: medium.** Inducible-allele SEM lines, four sequencing assays, two timepoints, three genotypes. Nothing here needs a mouse. **Controls that must not be dropped:** the catalytically dead IDH2 arm (2-HG is the variable, not the protein), a decitabine arm as a positive control for moving the uCpG pool [E4], and spike-in normalisation on every ChIP because a global methylation change can shift background genome-wide and make an unchanged fusion look reduced.
Why SEM and why not a mouse
SEM is where the baseline is. The 20% / 50% / 89% uCpG occupancy figures and the r2 = 0.96 menin correlation were all measured in SEM 1. Running the perturbation anywhere else forfeits the comparison and creates a new baseline problem.
And it is the index fusion in the index lineage. This project's commonest documented synthesis failure is generalising from murine MLL-AF9 myeloid leukaemia to infant B-lineage KMT2A::AFF1; two thirds of its load-bearing evidence carries no lineage marker at all. An experiment that can be done in human t(4;11) B-lineage cells should be.
The honest limitation of that choice: SEM is an established line from a relapsed 5-year-old, not an infant, and it has accumulated its own genome. An isogenic perturbation controls for that far better than a cross-line comparison would, but it does not make SEM an infant.
What this experiment cannot answer, said in advance
Recorded so that a null is not later oversold and a positive is not overread:
- It cannot tell you whether IDH2 mutation is selected for in patients. That is an epidemiological question and needs a fusion-only cohort, not a cell line.
- It cannot model mutation order. In a patient the IDH2 mutation may precede or follow the translocation, and an inducible allele in an established fusion line models only one order.
- It cannot reproduce the infant methylome. Infant KMT2A-rearranged blasts are globally hypermethylated at baseline relative to normal marrow; SEM's baseline is its own.
- A negative result is bounded by the induction time. If six weeks is not long enough for maintenance methylation to reshape CpG islands, the null is about the schedule.
The pre-registered interpretation of a null: under these conditions and this exposure, a hypermethylating co-mutation did not measurably reduce fusion occupancy - not hypermethylation does not affect the fusion.
| design | Isogenic SEM (human t(4;11) KMT2A::AFF1) with doxycycline-inducible IDH2-R140Q, catalytically dead IDH2-R140Q and wild-type IDH2 arms; 2-HG accumulation confirmed by mass spectrometry as a pharmacodynamic gate before sequencing; readouts at 2 and 6 weeks of induction. A decitabine arm serves as a positive control for moving the methylation readouts in the opposite direction. |
|---|---|
| readout | (i) Bio-CAP-seq for the unmethylated-CpG pool, matching the published baseline assay; (ii) MLL(N) and AF4(C) ChIP-seq or CUT&RUN for fusion occupancy; (iii) KDM2B and menin ChIP-seq for the competitor and the selector; (iv) whole-genome bisulfite sequencing as an orthogonal, non-CXXC methylation readout; (v) the decisive secondary analysis - is 5mC gain random over the uCpG pool with respect to menin occupancy, or concentrated at menin-bound promoters? |
| controls | Catalytically dead IDH2-R140Q (isolates 2-HG from the protein); wild-type IDH2 (isolates the neomorphic activity from overexpression); spike-in normalisation on every ChIP, because a global methylation change shifts background genome-wide and an unchanged fusion can look reduced without it; decitabine arm as a positive control for the methylation readouts only, since it also lowers KMT2A protein and therefore cannot control the fusion ChIP. |
| cost tier | med |
| timeline | 9-12 months including line construction and two induction timepoints |
| tests hypothesis | hyp-a-hypermethylating-comutation-should-antagonise-the-unmethylated-cpg-address |
| pre registered null interpretation | Under this exposure and these conditions, a hypermethylating co-mutation did not measurably reduce fusion occupancy - NOT 'hypermethylation does not affect the fusion'. |
| system | Human t(4;11) KMT2A::AFF1 B-lineage cell line - index fusion, index lineage, human. Not infant-derived. |
discriminates
{
"antagonism": "uCpG pool shrinks AND fusion peaks fall in number or height",
"headroom": "uCpG pool shrinks AND fusion peaks essentially unchanged",
"facilitation": "KDM2B occupancy falls AND fusion occupancy at shared sites rises"
}what it cannot answer
[ "whether IDH2 mutation is selected for in patients (needs a fusion-only cohort)", "mutation order, since an inducible allele in an established fusion line models one order only", "the infant methylome, since SEM's baseline is its own", "anything beyond the induction window, so a null is bounded by the schedule" ]
- This experiment could return the result that an IDH2 inhibitor expands the fusion's address space. If it does, an approved drug given to IDH2-mutant patients becomes a candidate hazard in the KMT2A-rearranged subgroup, and the finding would need to reach clinicians rather than only this graph.
- A null here is the outcome the project's central hypothesis predicts, and a null that confirms the sponsor's prior is the least trustworthy kind. The pre-registered interpretation, the catalytically dead arm and the spike-in normalisation are all there to make the null hard to manufacture, and none of them should be dropped for cost.
- Are the published SEM Bio-CAP-seq and ChIP-seq datasets deposited and re-analysable? If so the baseline arm is free and the cost tier drops; this beat did not check.
- How long does IDH2-R140Q induction need to run before CpG-island methylation reaches steady state in a rapidly dividing leukaemia line? A too-short exposure returns a false null and there is no published answer for this line.
- Should the same design be run in an MLL-AF9 myeloid line as well, given that the co-mutation epidemiology is myeloid and the index lineage is lymphoid? Doing both would test whether the answer is lineage-dependent.
- Is there an alternative that needs no new cells - for example re-analysing existing IDH2-mutant AML methylomes for depletion of unmethylated CpG at KMT2A-fusion-bound loci? That is a desk analysis and would come first if the data exist.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The baseline this experiment perturbs already exists in the same cell line: in SEM, Bio-CAP-seq plus ChIP-seq show MLL-AF4 on 20% of unmethylated-CpG sites, CFP1 on 50% and KDM2B on 89%, with fusion occupancy correlating with menin at r2 = 0.96 and menin with CFP1 at only 0.27.
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; percentages are in the running text and Figure S1F and not in the abstract. Whether the published Bio-CAP-seq and ChIP-seq data are deposited and re-analysable was NOT checked in this beat - if they are not, the proposed experiment has to regenerate its own baseline arm, which raises the cost tier. The uCpG denominator is CXXC-defined, which is why this design adds whole-genome bisulfite sequencing as an orthogonal readout.
Evidence that argues against this (1)
- Steady-state occupancy in one line is a weak baseline for a perturbation experiment if the line's own methylome is atypical; SEM's global methylation state relative to primary infant KMT2A-rearranged blasts has not been established here, so the starting point of the perturbation is not characterised. Schafer E et al. 2010
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Mutant IDH2 produces 2-hydroxyglutarate, which competitively inhibits both the TET 5mC hydroxylases and the JmjC histone demethylases - so a single IDH2-R140Q perturbation simultaneously raises CpG methylation and impairs KDM2A and KDM2B, which is why the experiment must read out the competitor as well as the fusion.
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; read from abstract only. No inhibition constants for KDM2A or KDM2B specifically, so the size of the competitor arm is unknown in advance and the experiment may be underpowered for it. R140Q is the commonest IDH2 allele in AML but was not itself the allele characterised in this paper's biochemistry.
Evidence that argues against this (1)
- KDM2B's CpG-island binding is CXXC-mediated and may be independent of its demethylase activity, so inhibiting catalysis with 2-HG could leave its 89% occupancy untouched - which would make the competitor arm of this experiment return a null for a reason unrelated to the hypothesis being tested. Kerry J et al. 2017
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The clinical question this experiment serves is live: IDH2 alteration is enriched in KMT2A-altered AML at 16.0% versus 10.4% (recomputed odds ratio 1.630, p = 4.1e-4), so a non-trivial fraction of patients carry both lesions and nothing is known about how they interact.
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 panel, no outcome data, 43.1% duplications in the KMT2A-altered arm with no stratification, and the paper's abstract disagrees with its own Table 2 on this p-value. SEM is a B-lineage ALL line, so the experiment tests a mechanism in a lineage where the co-mutation frequency has never been measured.
Evidence that argues against this (1)
- If the enrichment belongs to KMT2A-PTD rather than to fusions, an experiment in a fusion cell line is testing an interaction that does not occur in patients, and the whole design should be deferred until a fusion-only cohort has been analysed. Batayneh O et al. 2026
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A decitabine arm is available as a positive control for moving the unmethylated-CpG pool in the opposite direction, because decitabine is directly cytotoxic to MLL-rearranged but not MLL-wild-type cell lines and produces methylation-specific-PCR-confirmed loss of promoter methylation with re-expression of silenced genes.
Promoter hypermethylation in MLL-r infant acute lymphoblastic leukemia: biology and therapeutic targeting. · Schafer E, Irizarry R, Negi S, et al. · Blood 2010What this evidence cannot show
n = 5 MLL-r infant samples for the methylome arm; HELP rather than bisulfite; the cytotoxicity is in cell lines and SEM specifically was not named in the abstract read here. Decitabine is cytotoxic in MLL-r lines, so a decitabine control arm will also perturb viability and cell cycle - it is a control for the methylation readout, not a clean control for occupancy.
Evidence that argues against this (1)
- Because decitabine also lowers KMT2A protein in ALL cells, a decitabine control arm changes the very quantity the experiment is measuring, and cannot be used as a simple positive control for fusion ChIP - only for the Bio-CAP-seq and bisulfite readouts. Brock L et al. 2025
The one experiment that discriminates antagonism, headroom and facilitation in a single design.
Run in SEM, the index fusion in the index lineage, in human cells, with the published Bio-CAP-seq baseline in the same line.
Reads out the competitor as well as the fusion, because 2-HG inhibits JmjC demethylases and KDM2B is on 89% of unmethylated CpG.
Menin ChIP is the decisive secondary readout: is 5mC gain random over the uCpG pool, or concentrated at menin-bound promoters?
Decitabine arm as a positive control for moving the methylation readouts in the opposite direction - valid for Bio-CAP and bisulfite only, since decitabine also lowers KMT2A protein.