Oncofusion

start / graph / hyp-kmt2a-address-is-dna-structure-not-at-richness

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q4b-localization

Hypothesis: KMT2A's AT-hooks are a DNA-topology reader, not an AT-sequence reader — which would link the fusion's address to the scaffold-attachment architecture that also makes its own gene fragile

also called cruciform hypothesis · structure-selective AT-hook · SAR tether hypothesis

The hooks on KMT2A are usually described as gripping AT-rich DNA, but the only experiment on KMT2A's own hooks found they recognise a DNA shape instead — the four-way cross that forms where DNA is twisted. If that is right, the protein is drawn to the same kind of structurally strained DNA that sits around the point where its own gene breaks in this disease.

The same thing, for a scientist

The generic AT-hook motif is defined as a minor-groove binder of A.T-rich DNA [E1] and is described across DNA-binding proteins as a versatile minor-groove tether [E2]. But the only direct assay of MLL's own AT-hook region reported binding to **cruciform DNA**, explicitly recognising *structure rather than sequence* [E3]. The hypothesis takes that result seriously rather than as an artefact of a 1994 probe set, and makes it load-bearing: **KMT2A's three AT-hooks (aa 169-180, 217-227, 301-309) are a topology reader with preference for cruciform-prone, base-unpairing, negatively supercoiled DNA, i.e. for scaffold/matrix-attachment-region architecture.** Three independent observations become one explanation if this is right. (a) The MLL N-terminus targets the protein to the interphase nuclear matrix and to mitotic chromosome scaffolds alongside topoisomerase II [E4] — matrix attachment is operationally SAR DNA. (b) The KMT2A breakpoint cluster region is flanked by scaffold attachment regions and topoisomerase II consensus sites [E5]. (c) KMT2A is absent from AT-rich pericentromeric satellite, where genuine multi-AT-hook sequence readers such as HMGA1 concentrate. Prior probability 0.35: the direct evidence is a single old experiment, but it is the only direct evidence and it explains three otherwise separate facts. The corollary that makes this more than a curiosity is self-referential: a protein that binds SAR-architecture DNA would be enriched at its own gene's fragile region, which is a candidate mechanism for why KMT2A of all genes is the recurrent translocation partner in infant leukemia.

Go deeper
The self-referential corollary, stated as the risky prediction

If KMT2A's hooks prefer SAR-architecture DNA, and the KMT2A bcr is SAR-architecture DNA, then wild-type KMT2A protein should be enriched at its own breakpoint cluster region, and that enrichment should be independent of transcription. This is a strong, falsifiable, and as far as this search found untested prediction. It would mean the protein whose gene breaks is sitting on the DNA that breaks — and topoisomerase II, whose cleavage is the accepted initiating event, is the protein KMT2A colocalises with on mitotic scaffolds 4.

The prediction generalises to the partners. Most recurrent KMT2A partners (AFF1, MLLT3, MLLT1, MLLT10, ELL, AFDN) have their own breakpoint regions; if SAR architecture is a shared feature of recurrent partner bcrs, that is a sequence-level explanation of partner choice that does not depend on nuclear proximity — and the nuclear-proximity explanation has already failed its direct test, since 3D-FISH gives the wrong ordering for t(4;11) versus t(11;19).

This is the part of the hypothesis most likely to be wrong and most worth testing, because it would reframe partner selection as a DNA-architecture question rather than a spatial one.

Why the alternative is still live

The motif class is unambiguous and large. Reeves & Nissen's synthetic consensus peptide binds AT-rich minor groove and is competed by netropsin, distamycin and Hoechst 33258 1; Aravind & Landsman classify thousands of instances and explicitly name HRX/ALL-1 as carrying them 2. A 1994 gel shift with a handful of probes could easily have found cruciform binding because cruciform probes were in the panel and long AT tracts were not, or because a bacterially expressed fragment behaves differently from the hooks in context.

So the hypothesis is genuinely uncertain, and the first experiment is a binding survey rather than a cell experiment. That ordering matters: running the cell biology first would produce an occupancy map that could be rationalised either way.

Facts and figures
statementKMT2A's three AT-hooks read DNA topology (cruciform-prone, base-unpairing, supercoiled, scaffold-attachment architecture) rather than AT base composition, which explains nuclear-matrix targeting, absence from AT-rich pericentromeric satellite, and — as a risky corollary — predicts wild-type KMT2A enrichment at its own SAR-flanked breakpoint cluster region.
originliterature-gap
prior probability0.35
predictions
[
  "In a modern binding survey, the MLL AT-hook region prefers cruciform and supercoiled inverted-repeat substrates over relaxed long AT tracts of matched length",
  "Minor-groove ligands (netropsin, distamycin) compete HMGA1 binding far more effectively than MLL AT-hook binding on the same AT-rich substrate",
  "Published KMT2A::AFF1 occupancy is enriched at annotated scaffold/matrix attachment regions and at computationally predicted cruciform-forming inverted repeats, beyond what CpG-island overlap explains",
  "Wild-type KMT2A protein is enriched at the KMT2A breakpoint cluster region itself, independent of transcription",
  "A triple AT-hook alanine mutant in an endogenously tagged fusion loses nuclear-matrix retention while retaining CpG-island occupancy",
  "KMT2A remains absent from DAPI-dense chromocentres even when HMGA1 is depleted"
]
experiments
[
  "exp-at-hook-topology-binding-and-triple-mutant"
]
analyses
[
  "ana-sar-cruciform-overlap-kmt2a-occupancy"
]
falsifiers
[
  "A modern binding survey shows the MLL AT-hook region prefers relaxed AT-rich sequence over cruciform substrates, i.e. behaves as a conventional minor-groove tether",
  "KMT2A::AFF1 occupancy shows no enrichment at SAR or predicted cruciform features once CpG-island and accessibility covariates are controlled",
  "Wild-type KMT2A shows no enrichment at its own breakpoint cluster region above a matched-background expectation",
  "HMGA1 depletion reveals KMT2A at chromocentres, showing the hooks are AT-competent and merely outcompeted"
]
How this could bite us back
  • If confirmed, the therapeutic implication is uncomfortable: the fusion's address would be a generic DNA-structural feature present throughout the genome and in every cell, which is far less druggable than a protein-protein interface and would argue against AT-hook-directed chemistry entirely.
  • The self-referential corollary, if true, implies that anything increasing KMT2A expression or chromatin activity increases occupancy at its own fragile region — so a differentiation therapy that de-represses the locus could in principle raise the risk of a second rearrangement in surviving normal progenitors.
Still open
  1. If the hooks are topology readers, do they respond to transcription- or replication-generated supercoiling, making KMT2A's own address dynamic and activity-dependent rather than fixed?
  2. Do the recurrent partner genes' breakpoint regions share the SAR architecture of the KMT2A bcr? A sequence-level explanation of partner choice would displace the failed nuclear-proximity model.
  3. Does the cruciform preference explain why etoposide-induced t-AML breakpoints cluster where they do, given that topoisomerase II itself prefers particular DNA geometries?
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The A.T-hook motif, consensus T-P-K-R-P-R-G-R-P-K-K, binds the minor groove of A.T-rich DNA and is competed by the minor-groove ligands netropsin, distamycin and Hoechst 33258.

    PMID 1692833
    What this evidence cannot show

    Synthetic peptide and purified HMG-I; the motif definition comes from a different protein, and even this founding paper frames the motif as recognising DNA structure in its title.

    Evidence that argues against this (1)
    • The founding definition of this motif is more specific than 'binds A.T-rich DNA', and the specifics work against a clean sequence-versus-structure dichotomy. Huth et al.'s solution structure of HMG-I(Y) bound to the PRDII site shows a central Arg-Gly-Arg core plus two accessory modules making extensive hydrophobic and polar contacts, a one-protein-to-two-DNA stoichiometry, and an architectural role in which the motif STABILISES B-DNA and thereby facilitates binding of other transcription factors in the opposing major groove. That is a structural function performed on ordinary B-form duplex - neither pure sequence readout nor cruciform recognition - so a hypothesis that partitions AT-hook behaviour into 'sequence' versus 'topology' is using a distinction the structure does not support. Huth JR et al. 1997
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AT-hook motifs occur in three sequence types across a wide variety of DNA-binding proteins including HRX/ALL-1 and are described as a versatile minor-groove tether.

    AT-hook motifs identified in a wide variety of DNA-binding proteins · Aravind L, Landsman D · Nucleic Acids Research 1998
    PMID 9742243DOI
    What this evidence cannot show

    Computational only; the functional description is inferred from sequence conservation, not measured for any of the proteins named.

    Evidence that argues against this (1)
    • A sequence-pattern census cannot tell you which copies of a motif bind, and the structural work shows the difference between copies is large. Huth et al. found that the third DNA-binding domain of HMG-I(Y) has roughly 100-fold lower affinity than the second, attributable to the absence of one of the contact modules present in the higher-affinity copy. Aravind and Landsman's three sequence types are therefore consistent with a wide range of actual affinities, and 'versatile minor-groove tether' is a description of the class, not a prediction for KMT2A's three copies. The hypothesis needs the per-hook affinities measured, which is exactly what its companion experiment proposes and what no published work provides. Huth JR et al. 1997
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The MLL AT-hook region binds cruciform DNA, recognising the structure rather than the sequence of the target DNA.

    PMID 7938000DOI
    What this evidence cannot show

    A single 1994 gel-shift study on a bacterially expressed fragment with a limited probe panel, never repeated with modern methods and with no structure of an MLL AT-hook bound to DNA in the PDB. This is the whole direct evidence base for the hypothesis.

    Evidence that argues against this (1)
    • The motif class this domain belongs to is defined as an AT-rich minor-groove binder, so the structural preference could be a property of the probes used. Aravind L et al. 1998
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The MLL amino terminus targets the protein to the interphase nuclear matrix and to mitotic chromosomal scaffolds where it colocalises with topoisomerase II, as well as to punctate nucleoplasm and the nucleolar periphery.

    PMID 11069025DOI
    What this evidence cannot show

    Overexpressed conditional constructs in U937 and HeLa; nuclear-matrix preparations are extraction-dependent and the matrix itself remains a contested operational entity.

    Evidence that argues against this (1)
    • The compartment this observation places MLL in may not exist as an entity. Pederson's review traces the nuclear matrix from its 1948 description and 1974 promotion to a fundamental organising principle, and concludes that convincing evidence for that functional role has been elusive and recently further challenged: the fraction is produced by harsh extraction and precipitation-prone conditions, the biophysical properties of nucleoplasm in vivo are hard to reconcile with an extensive anastomosing filament network, and considerable scepticism continues to surround the matrix fraction as a representation of the in vivo situation. Recovery of MLL in a matrix preparation is therefore a statement about what resists a particular extraction, and the hypothesis's step from 'matrix-associated' to 'reads scaffold/matrix-attachment architecture' inherits that uncertainty whole. Pederson T 2000
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    11q23 breakpoints within the MLL breakpoint cluster region correlate with scaffold attachment regions and topoisomerase II consensus binding sites.

    PMID 8634439
    What this evidence cannot show

    Correlation in 39 patient samples with era-specific SAR mapping methods; SAR assignment is sensitive to extraction conditions and has not been redone with modern approaches.

    Evidence that argues against this (1)
    • Both terms of this correlation are defined by the same contested operation. Scaffold attachment regions are identified by what remains bound after nuclear extraction, and Pederson's review documents the long-standing objection that such preparations use harsh extraction and precipitation-prone conditions, that the in vivo biophysics of the nucleoplasm sit awkwardly with the inferred structure, and that scepticism about the matrix fraction as an accurate representation persists. A correlation between breakpoint positions and SAR assignments made by that operation can therefore reflect a shared dependence on local base composition and extraction chemistry rather than a shared dependence on nuclear architecture - and since AT-rich sequence is both what SAR assays enrich and what the AT-hooks are proposed to read, the correlation is close to circular. Pederson T 2000
Connected to