Hypothesis: the KMT2A breakpoint map is shaped by product viability, not only by DNA fragility, and the signature is that breaks respect inter-domain linkers
Chromosome breaks in this gene are not scattered randomly; they cluster in one small region. The usual explanation is that the DNA there is fragile, but there may be a second reason: those are the only places you can cut the protein and still get something that works.
The same thing, for a scientist
The observed KMT2A breakpoint cluster regions are conventionally explained by DNA fragility (topoisomerase II cleavage sites, Alu density, scaffold attachment regions, apoptotic nuclease cleavage). Imported from the blue-pencil rule in contract law, an orthogonal explanation is that the map is partly a *survivorship* map: breaks are observed where the resulting protein is still competent. These two explanations make different predictions about non-selected breaks. We predict that, after conditioning on fragility covariates, breakpoint density in leukemic cases is still elevated in the CXXC-to-PHD1 inter-domain linker relative to positions inside folded modules, and that this excess is absent or much smaller in KMT2A breaks detected without leukemic selection — in healthy cord blood, in etoposide-treated cells in vitro, and in non-recurrent single-case rearrangements. This is a speculative inference of ours drawn from an analogy, not a reported finding.
Why the two explanations are separable
Fragility acts on DNA before selection; viability acts on protein after it. The only place they can be told apart is a population of KMT2A breaks that has not passed through the leukemic filter. Three such populations exist or are obtainable: (i) KMT2A rearrangements detectable at low frequency in healthy cord blood and in normal tissue, (ii) etoposide- or apoptosis-induced KMT2A breaks mapped in vitro, and (iii) the tail of non-recurrent, single-case breakpoints in the recombinome itself, which by definition have not been repeatedly selected 1.
If the linker excess is present in all three, fragility explains everything and this hypothesis is dead. If it appears only in the recurrent leukemic set, selection on product viability is real and measurable.
Why anyone should care
Three consequences, in ascending order of practical value.
- It reframes the exon-11 cut point. The recombinome identifies exon 11 as the point past which PHD1-3 is destroyed 1. If viability selection is detectable, exon 11 is not a passive coordinate but a boundary between two structurally different fusion classes, which strengthens the case for stratifying trials on break position — the question
ana-interfant-breakpoint-prognostic-reanalysisalready exists to answer. - It predicts where a new partner is allowed to break. Every mapping rule generated here applies equally to the 100-plus documented KMT2A partners, giving a prior over which novel partner-side breaks should be viable.
- It bears on prevention. If the breakpoint map is mostly fragility, interventions at the level of DNA damage (topoisomerase-II-poison exposure in utero, replication stress) are aimed at the right target. If a large part of the map is selection, then reducing fragility shifts where breaks occur without reducing how many leukemias result, and prevention arguments based on the breakpoint map are weaker than they look.
Honest statement of the prior
Prior 0.35. Both mechanisms almost certainly contribute; the risky part of the claim is not that viability matters but that its signature is detectable above fragility in existing data. Intron length alone explains a great deal of the subregion shares (the 5,026 bp exon 11-13 subregion carries 39.5% of breaks and the 679 bp exon 10-intron 10 subregion carries 19.8%, which is a 5-fold density difference in the opposite direction to length — already suggestive that length is not the whole story) 1. The chief risk is that fragility covariates and domain boundaries are collinear because intron/exon architecture and domain architecture co-evolved, in which case the regression cannot separate them and the analysis returns an honest null.
| statement | After conditioning on DNA-fragility covariates (intron length, Alu and other repeat density, topoisomerase II consensus cleavage sites, scaffold/matrix attachment regions, apoptotic nuclease sensitivity), KMT2A breakpoint density in recurrent leukemic cases remains significantly elevated at positions encoding inter-domain linkers relative to positions encoding folded modules, and this excess is absent or substantially smaller in KMT2A breaks that have not passed through leukemic selection. |
|---|---|
| origin | analogy |
| source analogy | analogy-law-severability-domain-retention |
| prior probability | 0.35 |
| decision value | Determines whether break position should be treated as a stratification variable with mechanistic meaning or as a passive coordinate, and bears directly on how much weight prevention arguments built on the breakpoint map can carry. |
predictions
[ "In a regression of breakpoint counts on genomic position across the 3,401-case recombinome, a linker-versus-domain indicator derived from the KMT2A protein domain model retains a positive coefficient after fragility covariates are included.", "Etoposide-induced or apoptosis-induced KMT2A breaks mapped in unselected cells show fragility-driven clustering but no additional linker enrichment.", "Non-recurrent single-case breakpoints in the recombinome tail are less linker-enriched than recurrent ones.", "The same linker constraint appears on the partner side: AFF1 breakpoints should cluster upstream of the ALF and pSER modules rather than inside them.", "Breaks at positions predicted to sever a folded module should be enriched among cases with atypical or complex rearrangements requiring more than a simple reciprocal exchange." ]
experiments
[]
analyses
[ "ana-breakpoint-linker-vs-fragility-decomposition", "ana-interfant-breakpoint-prognostic-reanalysis" ]
falsifiers
[ "The linker indicator loses significance once intron length and repeat density are in the model — fragility explains the map and the analogy adds nothing.", "Unselected KMT2A breaks (cord blood, etoposide-treated cells in vitro) show the same linker enrichment as leukemic breaks, proving the pattern is pre-selective.", "Mapping the recombinome onto a current KMT2A domain model shows the major cluster region straddles rather than avoids a folded module, contradicting the premise.", "Isogenic fusions differing only in whether the break severs a folded module transform equally well, showing product viability is insensitive to the distinction." ]
- A confirmed result would say the observed fusion repertoire is a survivorship sample, which means the rare structural classes now invisible to the field may be exactly the ones a menin inhibitor covers least well.
- A null result is genuinely informative but would be tempting to report as a positive by adjusting the domain model until significance appears. The domain model must be pre-registered before the regression is run.
- Is a mapped set of KMT2A breaks from unselected cells (healthy cord blood, in vitro etoposide) publicly available at the resolution this test needs, or would it have to be generated?
- Can intron/exon architecture and protein domain architecture be statistically decoupled in this gene at all, or are they too collinear for any regression to separate them?
- Does a similar analysis in a fusion where the breakpoint region is long and domain-poor, such as BCR::ABL1, act as a negative control?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 3,401 patients, 98% of KMT2A breaks fall in a major cluster region from intron 7 to exon 13 and 1.4% in a minor region from intron 20 to exon 24; subregion shares are 37.2% over 1,761 bp, 19.8% over 679 bp and 39.5% over 5,026 bp; exon 11 is identified as the functional cut point because breaks at or past it destroy the PHD1-3 module.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Referral-based series with changing detection methods; the breakpoint coordinates are diagnostic-grade rather than research-grade for a subset of cases.
Evidence that argues against this (1)
- A sufficient alternative explanation is already the field's default and this entry has to beat it, not merely condition on it. Cowell and Austin review the DNA-fragility case for the KMT2A breakpoint cluster region: translocation break sites fall in an 8 kb BCR, therapy-related breaks concentrate in a hotspot near the intron 11/exon 12 boundary, and genome-wide analyses demonstrate etoposide-induced TOP2 DNA cleavage within the BCR, with TOP2 cleavage-site sequence preferences and DNA secondary-structure features as additional determinants. Notably, they also point out that the therapy-related hotspot and the centre of preferential cleavage are offset by more than 250 nucleotides — which shows the fragility account is itself incomplete, but the residual it leaves is a *sequence/structure* residual, not obviously a protein-linker residual. The hypothesis proposes to condition on fragility covariates and look for a linker excess; this review is the statement of how much of the signal those covariates are expected to absorb, and it is most of it. Cowell IG et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Breakpoint position in KMT2A is associated with age and lineage in 2,345 patients — intron 11 breaks with lymphoblastic leukemia and younger age, intron 9 with myeloid leukemia and older patients — and the pattern reverses for MLLT10 and for adult MLLT3, showing that position is entangled with partner identity.
The MLL recombinome of acute leukemias in 2017 · Meyer C, Burmeister T, Groger D, et al. · Leukemia 2017What this evidence cannot show
Univariate associations only; partner identity, age, geography and detection method are all confounded with each other, which is precisely the confounding the proposed analysis must model rather than ignore.
Evidence that argues against this (1)
- The hypothesis's discriminating prediction is that the linker excess should be absent or much smaller in KMT2A breaks that never passed through leukaemic selection. A published counterexample in outline already exists: exposing normal cord blood and normal infant and child peripheral blood mononuclear cells to etoposide in vitro produced MLL double-strand breaks detected by ligation-mediated PCR, mapping to the telomeric side of intron 8 near exon 9 — inside the breakpoint cluster region, in cells that underwent no leukaemic selection whatsoever, and with no difference in incidence between cord blood and other samples. Breaks land in the same part of the gene without selection. Caveat: an older, low-resolution study with a small number of sequenced junctions, and it does not resolve breaks at the codon-level precision the hypothesis needs — so it weakens rather than kills the prediction. But it means the null the hypothesis must beat already has data behind it. Ishii E et al. 2002
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Severance in contract law is subtractive only: under the blue-pencil rule a court may strike offending words but may not rewrite them to save the agreement.
Severability clause (Wex legal encyclopedia) · Legal Information Institute, Cornell Law School · Cornell Law School Legal Information Institute 2024What this evidence cannot show
Source-domain reference. It is the origin of the idea and is not evidence for any biological claim.
Evidence that argues against this (1)
- WHERE THE ANALOGY BREAKS. The entry treats severance as necessarily subtractive — the blue-pencil rule permits striking words but not rewriting. That is one of at least four approaches actually in use, and it is not the majority one. Practitioner analysis distinguishes the red pencil (refuse to rescue an over-reaching agreement at all, invalidating it entirely), the blue pencil (delete only, enforce what remains if reasonable), reformation (courts revise language, including adding or changing words, to make the agreement enforceable — with some states legislatively mandating reformation), and the purple pencil (reformation conditioned on good faith in drafting). So the source-domain premise that 'severance is subtractive only' is jurisdiction-specific. The transfer to protein structure is worse: severance presupposes a text of grammatically independent clauses whose surviving portions retain their meaning unchanged, whereas the domains flanking a KMT2A breakpoint are not independent — the truncated N-terminus acquires an entirely new C-terminal partner and a new function, which in contract terms is reformation, not severance. Reformation is the doctrine the biology actually matches, and reformation puts no premium on cutting at clause boundaries, which is the hypothesis's whole prediction. Beck R 2025
the fragility null is strong: TOP2 cleavage inside the BCR is demonstrated genome-wide, and etoposide produces MLL breaks in the same region in normal cord and peripheral blood cells with no leukaemic selection
- proposed by →Severability (contract law) — what survives when a clause is struck, and what survives a KMT2A break
generated by the blue-pencil rule
offers a selection-based account of why the cluster regions sit where they do
predicts the major cluster region coincides with the CXXC-to-PHD1 linker
claims fragility alone is an incomplete account of the breakpoint map
positional regression with fragility covariates and a pre-registered domain model