Oncofusion

start / graph / localization-nuclear-matrix-mitotic-scaffold-kmt2a

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q4b-localization

Nuclear matrix and mitotic chromosome scaffold — KMT2A sits with topoisomerase II, in the same place the gene breaks

also called MLL nuclear matrix · matrix attachment region MLL · SAR MAR AT-hook · mitotic chromosomal scaffold MLL · TOP2 colocalisation

KMT2A stays attached to the nucleus's internal skeleton, and during cell division it rides on the protein scaffold of the condensing chromosomes right next to topoisomerase II. That is striking, because topoisomerase II is the enzyme whose cutting at the KMT2A gene is thought to start infant leukemia in the first place.

The same thing, for a scientist

Caslini et al. recovered both endogenous MLL and a conditionally expressed MLL N-terminal AT-hook fragment in interphase nuclear matrix preparations and showed colocalisation with topoisomerase II on mitotic chromosomal scaffolds by immunofluorescence, confocal microscopy and immuno-EM [E1]. The biochemical basis is plausible: the MLL AT-hooks bind cruciform DNA, recognising structure rather than sequence [E2], and were described from the outset as HMG-like minor-groove binders [E3] — the binding mode associated with scaffold/matrix attachment regions. This gives a second, non-CpG targeting logic for the fusion, and it places the protein in the same physical compartment where TOP2-mediated cleavage of the KMT2A breakpoint cluster region is proposed to generate the translocation. Whether the fusion protein retains matrix association has never been tested directly, although every fusion keeps the AT-hooks.

Go deeper
The observation1 deeper

Three separate readouts in one 2000 study: (i) endogenous MLL and MLL3AT are present in interphase nuclear matrices after extraction; (ii) both colocalise with topoisomerase II on mitotic chromosomal scaffolds; (iii) by immuno-EM the interphase signal includes regions attached to the nuclear envelope and the nucleolar periphery 1. The authors' own interpretation was that the MLL amino terminus has a role in modulating chromatin structure with epigenetic consequences for gene expression maintenance.

'Nuclear matrix' is a contested operational category — it is defined by what survives extraction, and extraction artefacts are a real worry. But mitotic scaffold colocalisation with TOP2 is a separate and more robust observation, because the mitotic chromosome scaffold is a structure with independent biochemical definition.

Why AT-hooks put it there

The AT-hook was the very first thing noticed about this protein: Tkachuk, Kohler and Cleary described HRX as a 431 kDa protein with non-conserved amino-terminal AT-hook motifs related to the DNA-binding motifs of HMG proteins, and proposed DNA binding within the minor groove at AT-rich sites 3. Zeleznik-Le then expressed the MLL AT-hook domain in bacteria and showed by gel shift that it binds cruciform DNA, recognising structure rather than sequence of the target 2.

Structure-not-sequence, minor-groove, AT-rich: that is the recognition signature of scaffold/matrix attachment regions. It is a targeting determinant that operates on DNA shape, orthogonal to the CXXC domain's chemical read of unmethylated CpG, and it is retained by every fusion.

The uncomfortable coincidence with the breakpoint

This graph already holds an entry on TOP2-mediated cleavage at KMT2A as the initiating event for infant t(4;11). Caslini's result says the KMT2A protein itself sits with TOP2 on the mitotic scaffold 1. Two readings are available and they differ in what they predict:

  1. Coincidence of chromatin class. AT-rich matrix-attached chromatin is where TOP2 works anyway; MLL is there for the same reason TOP2 is, and neither causes the other.
  2. Self-targeting. The protein's own localisation biases where TOP2-mediated double-strand breaks accumulate, including at its own locus — which would make KMT2A partly responsible for its own fragility.

Reading 2 is testable and nobody has tested it. If it were true, it would predict that KMT2A protein level or AT-hook competence modulates breakage frequency at the KMT2A BCR under etoposide, which is a straightforward experiment.

What is missing

No published experiment asks whether KMT2A::AFF1 is matrix-associated. Every fusion retains the AT-hooks, so the prior is that it is, but the 3' partner brings a large disordered pSer region that could change partitioning entirely. A matrix/chromatin fractionation of a t(4;11) line with an antibody against the AFF1 C-terminus (which detects the fusion and wild-type AFF1 but not wild-type KMT2A) is a one-week experiment.

Facts and figures
compartmentnuclear matrix / mitotic chromosome scaffold
evidence class for kmt2adirect for wild-type KMT2A and an N-terminal fragment; none for the fusion protein
endogenous tag datafalse
patient material datafalse
targeting determinantAT-hooks, minor-groove, structure-not-sequence recognition of cruciform/AT-rich DNA
at hook count3
marker proteins
[
  "topoisomerase II alpha (TOP2A)",
  "SAF-A/HNRNPU",
  "lamin B"
]
methods used
[
  "nuclear matrix fractionation",
  "indirect immunofluorescence",
  "confocal microscopy",
  "immuno-electron microscopy",
  "EMSA with bacterially expressed domain"
]
How this could bite us back
  • If AT-hook/matrix targeting is a parallel route to chromatin that does not run through menin, then evicting the fusion from the menin-LEDGF axis leaves a second tether intact, and residual occupancy after menin inhibition would be expected rather than surprising.
  • AT-hooks are shared with HMGA proteins and other abundant chromatin factors, so any small molecule aimed at this interface has an intrinsic selectivity problem.
Still open
  1. Is KMT2A::AFF1 matrix-associated? Never tested. Chromatin/matrix fractionation of SEM or RS4;11 blotted with an AFF1 C-terminal antibody answers it directly.
  2. Does KMT2A protein occupancy bias TOP2-mediated cleavage at its own breakpoint cluster region? If yes, the protein contributes to the fragility of its own gene, which would be a genuinely novel loop.
  3. Do the three AT-hooks act redundantly? No AT-hook-null fusion has been built and tested for transformation and for localisation simultaneously, so the field cannot say whether AT-hook targeting is load-bearing or vestigial in the fusion.
  4. Is there a structure of any KMT2A AT-hook bound to DNA? As far as this graph can establish, no — which means a 34-year-old claim about minor-groove binding rests entirely on sequence homology.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Endogenous MLL and a conditionally expressed MLL N-terminal AT-hook fragment are present in interphase nuclear matrices and colocalise with topoisomerase II on mitotic chromosomal scaffolds, with immuno-EM showing interphase signal including regions attached to the nuclear envelope and the nucleolar periphery.

    PMID 11069025DOI
    What this evidence cannot show

    U937 and HeLa, conditional overexpression of the fragment (endogenous MLL was also examined, which strengthens it). 'Nuclear matrix' is an operational category defined by resistance to extraction and is vulnerable to redistribution artefacts during extraction. Fixed cells only; no quantification of the matrix-bound fraction. Nothing here was done on a KMT2A-rearranged cell or on the fusion protein at endogenous level.

    Evidence that argues against this (2)
    • The compartment this entry is named after is contested at the level of whether it exists. Pederson's review notes that the fraction now called the nuclear matrix was first described in 1948 and promoted from 1974 as an organising principle of gene expression, that convincing evidence for that functional role has been elusive and recently further challenged, that objections to the harsh extraction and precipitation-prone conditions used to prepare it are objective rather than merely dismissive, and that considerable scepticism continues to surround the fraction as an accurate representation of the in-vivo nucleus. Caslini's observation that MLL and its N-terminal fragment partition into the nuclear-matrix and mitotic-scaffold fractions is a real biochemical result; what it licenses is 'MLL is retained after nuclease digestion and high-salt extraction', not 'MLL occupies a nuclear matrix'. Read: abstract only. Pederson T 2000
    • An orthogonal reason to read the matrix/scaffold assay as a sequence property rather than an in-vivo attachment. Stanulla et al. showed that the MLL breakpoint cluster region behaves as a SAR, and is cleaved at the same specific site, when it is present as a randomly integrated exogenous fragment or on an extrachromosomal episome - outside its normal chromosomal context altogether - and that a murine MLL bcr episome functions as a SAR in human cells. Scaffold attachment is therefore carried by the naked sequence and reproduced across species and contexts, which is what one expects from an in-vitro reassociation property of AT-rich, easily unwound DNA, and not what one expects from a position in a specific nuclear structure. Read: abstract only. Stanulla M et al. 2001
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The bacterially expressed MLL AT-hook domain binds cruciform DNA in a gel mobility shift assay, recognising the structure rather than the sequence of the target DNA; MLL also contains a strong activation domain telomeric to the breakpoint region and a repression domain centromeric to it, so translocation retains the DNA-binding and repression functions on der(11) while removing activation.

    PMID 7938000DOI
    What this evidence cannot show

    In-vitro gel shift with an isolated bacterially expressed domain and synthetic cruciform substrate; it demonstrates a binding preference, not occupancy of matrix attachment regions in a cell. The transactivation mapping used GAL4 fusions and CAT reporters in transient transfection, a system that overstates domain autonomy.

    Evidence that argues against this (1)
    • Searched and empty, recorded explicitly. PubMed searches on 2026-09-18 for MLL or KMT2A with cruciform DNA, four-way junction, Holliday junction and non-B DNA found no independent replication of the 1994 cruciform gel-shift and no structural or in-cell follow-up. The structure-recognition premise of this entry rests on a single unreplicated experiment; that is stated here so a reader does not mistake age for corroboration. No citation is attached because there is nothing to cite: this note is a bounded absence plus a scope argument about a single unreplicated 1994 result, not a sourced fact. Uncited and unciteable look identical in this field, so it is said here explicitly (Q42 citation-integrity repair, 2026-09-20).
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    HRX/MLL was identified as a predicted 431 kDa protein carrying non-conserved amino-terminal AT-hook motifs related to the DNA-binding motifs of HMG proteins, with the proposal that it acts by DNA binding within the minor groove at AT-rich sites; 11q23 translocations disrupt the gene between the AT-hooks and the zinc fingers.

    PMID 1423624DOI
    What this evidence cannot show

    The minor-groove binding mode was inferred from sequence homology to HMG-I(Y) AT-hooks in 1992, not demonstrated structurally for KMT2A; no experimental structure of a KMT2A AT-hook bound to DNA has been reported. The predicted molecular weight (431 kDa) predates the current annotation.

    Evidence that argues against this (1)
    • The 1992 inference from HMG homology to minor-groove binding 'at AT-rich sites' has a structure behind it, and the structure describes something other than site-specific targeting. Huth et al. solved HMG-I(Y) bound to the PRDII element and found a one-protein-to-two-DNA stoichiometry, a central Arg-Gly-Arg motif plus two accessory contact modules of unequal contribution (the weaker copy binding roughly 100-fold less tightly), and an architectural function in which the protein stabilises B-DNA so that other transcription factors can engage the opposing major groove. Transferring that to KMT2A predicts a DNA-shaping cofactor that can bridge two duplexes, not an AT-address - and a bridging protein is precisely what a preparation enriched for cross-linked, extraction-resistant material would retain. Huth JR et al. 1997
Connected to