Oncofusion

start / graph / mechanism-kmt2a-subnuclear-localization

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q4-cells

Where KMT2A and its fusions sit inside the nucleus — SNL1/SNL2, nuclear speckles and the condensate question

also called MLL subnuclear localization · SNL1 SNL2 · ALL-1 nuclear punctate distribution · MLL nuclear speckles · SEC condensates · KMT2A::AFF1 subnuclear localization

KMT2A does not float evenly through the nucleus — it collects into bright dots, and two short stretches of the protein (amino acids 388-432 and 1,051-1,089) are what put it there. Every leukemia fusion keeps both stretches, so the fusion protein inherits the same dotted address, and whether those dots are true liquid droplets is still an open question for KMT2A::AFF1 specifically.

The same thing, for a scientist

Immunofluorescence of native ALL-1/MLL shows a punctate nuclear distribution, and deletion mapping localised this to two short trithorax-conserved elements named SNL-1 (aa 388-432) and SNL-2 (aa 1,051-1,089), plus three nuclear targeting sequences including a classical RKRKRK motif at aa 734-739 [E1]. Both SNLs lie inside the N-terminal ~1,400 residues that every MLL fusion retains, so MLL-AF4, MLL-AF9 and MLL-AF17 all reproduce the speckled pattern [E1][E2]. Mechanistically, punctate positioning is necessary but not sufficient: high-resolution mutagenesis identified three domains (including CXXC/Basic) that are required for myeloid transformation by mechanisms that leave subnuclear localization intact [E2], and chromatin targeting itself runs through the menin-LEDGF axis at the extreme N-terminus [E7]. On the 3' side, the super elongation complex shared by essentially all MLL chimeras [E5] is itself a phase-separating assembly nucleated by the ENL intrinsically disordered region and fluidised by AFF4, and MLL-ENL fusion boosts SEC phase separation [E3]. Direct condensate data for KMT2A::AFF1 do not exist; the condensate framing for this specific fusion is an inference from ENL/SEC [E3][E4] and from NUP98-HOXA9 [E6].

Go deeper
The two subnuclear localization signals, with the residue numbers as published2 deeper

Yano and colleagues raised antibodies against ALL-1 (now KMT2A/MLL) and found that the endogenous protein distributes in cultured cells in a nuclear punctate pattern rather than diffusely. Dissecting the protein, they mapped this to two elements within the N-terminal ~1,100 residues and named them subnuclear localization signals:

  • SNL-1: amino acids 388-432
  • SNL-2: amino acids 1,051-1,089

Both correspond to short sequences conserved with Drosophila TRITHORAX. Separately they mapped three nuclear targeting sequences (NTS-1, NTS-2 co-mapping with SNL-1 in aa 322-480, and NTS-3, a classical RKRKRK motif at aa 734-739) 1.

These numbers matter because they are frequently mis-stated in secondary sources. The often-quoted "SNL1 in the 1100s, SNL2 in the 1300s" is wrong: SNL-1 is at aa 388-432 and SNL-2 at aa 1,051-1,089, and both sit upstream of the CXXC domain, in the interval between the AT-hooks and CXXC. This node asserts only the ranges printed in the primary paper.

Why the fusion inherits the address

MLL fusion oncoproteins invariably retain the KMT2A N-terminus fused to one of over thirty partners, conventionally described as the amino-terminal ~1,400 residues 2. PRECISION NOTE 2026-09-19: the retention is invariable; the extent is not. Breakpoints fall across introns 9 to 11 with an age gradient, so ~1,400 is a convention rather than a constant. Both SNLs and all three NTSs fall inside that retained block. Yano et al. expressed several chimeric ALL-1 proteins encoded by translocation products — ALL-1/AF-4 (t(4;11)), ALL-1/AF-9 and ALL-1/AF-17 — and all showed the same nuclear speckles as wild-type ALL-1 1.

So subnuclear positioning is not something the fusion acquires; it is something the fusion cannot lose. Any therapeutic idea built on "relocalise the fusion" has to contend with two independent, redundant, conserved elements plus three NLSs, all of them inside the obligatory 5' block.

The complementary result is equally important: enforced nuclear expression of other MLL domains — the PHD fingers, the SET motif — produced uniform, non-punctate patterns 1. Dotted distribution is a property of the N-terminus specifically, which is exactly the part the translocation keeps.

What the dots are (and what the 1997 paper did not test)

Yano et al. proposed that the speckles represent protein complexes containing multiple copies of ALL-1, positioned at ALL-1 target sites on chromatin — i.e. the N-terminus's job is to deliver the protein to its target genes 1.

Two honest limitations. First, this was largely transfection/overexpression work in COS and HeLa cells with epitope-tagged fragments; overexpressed fragments aggregate readily and "speckles" in 1997 microscopy is a low bar. Second, the paper did not report co-staining against canonical nuclear-speckle markers (SC35, snRNP), so the phrase "MLL localises to nuclear speckles" in the sense of interchromatin granule clusters is not established by this paper. What is established is a reproducible punctate, non-uniform nuclear distribution, retained by fusions, driven by two mappable elements. Anyone writing "MLL is in nuclear speckles" should say which definition of speckle they mean.

Punctate localization is necessary but not sufficient — the CXXC dissociation experiment1 deeper

The single most useful piece of evidence here is a negative one. Ayton, Chen and Cleary performed high-resolution mutagenesis across the MLL N-terminus and identified three domains essential for in vitro myeloid transformation by mechanisms that do not compromise subnuclear localization — the CXXC/Basic domain plus two then-uncharacterised domains. Point mutations in CXXC that abolished myeloid transformation also abolished recognition of non-methylated CpG DNA in vitro and transactivation in vivo, while the mutant protein still went to the right subnuclear compartment 2.

That cleanly separates two things people conflate:

  1. Compartment — getting into the right nuclear neighbourhood (SNL1/SNL2, NTSs).
  2. Engagement — actually gripping unmethylated CpG at target promoters (CXXC) and being held there by menin-LEDGF 7.

A fusion protein can be in the right dots and still be transformation-dead. So "the fusion is in nuclear puncta" is not, on its own, a mechanism, and it is not on its own a drug target either.

The chromatin-tethering layer underneath

Menin binds the extreme N-terminus of MLL and physically bridges it to LEDGF/PSIP1, a chromatin-associated protein; LEDGF is required both for MLL-dependent transcription and for leukemic transformation. A subset of MEN1 patient mutations abrogate the LEDGF interaction while preserving MLL binding, and still compromise MLL/menin-dependent function 7. So the practical chromatin address of the fusion is set by a three-body module (MLL-N / menin / LEDGF) that is upstream of, and separable from, the SNL-driven compartment.

Condensates: what is demonstrated, what is analogy1 deeper

This is where the literature and the popular framing diverge, and it is worth being blunt.

Demonstrated, on the 3' partner side. The super elongation complex is a phase-separating assembly. ENL — or its intrinsically disordered region alone — is sufficient to initiate liquid droplet formation of SEC, AFF4 acts together with ENL to fluidise those droplets, SEC droplets compartmentalise and concentrate P-TEFb away from the soluble inactive HEXIM1-P-TEFb pool, they form within minutes of serum stimulation, and they are required for rapid transcriptional induction. Critically for us, the authors report that fusion of ENL with MLL boosts SEC phase separation 3.

Demonstrated, in vivo, for a condensate-promoting ENL mutant. A conditional knock-in mouse carrying the gain-of-function ENL mutation found in AML and Wilms tumour shows that mutant ENL forms condensates in haematopoietic stem/progenitor cells at leukemogenic loci, that mutagenesis which disrupts condensate formation impairs its chromatin and oncogenic function, and that an acetyl-binding inhibitor displaces those condensates from target loci and delays AML in vivo 4. This is the strongest existing proof-of-principle that a leukemic condensate is a druggable object.

Demonstrated, for a different fusion family. NUP98-HOXA9 forms LLPS puncta via the nucleoporin FG-repeat IDR; the IDR is required for leukemic transformation; LLPS promotes chromatin occupancy of the chimera, produces a broad super-enhancer-like binding pattern at leukaemogenic genes, and induces CTCF-independent chromatin loops enriched at proto-oncogenes. Swapping the FG repeats for an unrelated LLPS-forming FUS IDR reproduced the effect 6.

Not demonstrated: KMT2A::AFF1 itself. Targeted searches of the primary literature return no paper showing that KMT2A::AFF1 (MLL-AF4) forms phase-separated condensates, no FRAP or 1,6-hexanediol sensitivity data for the fusion, and no droplet reconstitution with the AFF1 pSER region. The condensate story for this fusion is an inference by analogy, resting on (i) AFF1 being an AFF-family paralogue of AFF4, which is a demonstrated SEC droplet component 35; (ii) the fusion's obligate SEC association 5; and (iii) the general rule from NUP98-HOXA9 that IDR-bearing fusion transcription factors use LLPS 6. It is a good hypothesis. It is not a finding. Confidence for the KMT2A::AFF1-specific condensate claim is speculative, and this node marks it so.

What an honest experiment would look like

The gap is small and cheap to close. In a t(4;11) line (SEM, MV4-11, RS4;11) with an endogenously tagged fusion: (i) live-cell imaging for discrete foci, (ii) FRAP to test internal mobility, (iii) fusion-to-fusion vs fusion-to-AFF1/ENL/CDK9 co-localisation, (iv) an IDR-deletion or pSER-deletion allele scored for both foci and transformation. Nobody appears to have published this for KMT2A::AFF1.

A methodological warning for whoever does it: a widely used condensate test in the AFF4 literature is disruption by aliphatic alcohol. FUS was shown to co-localise with AFF4 in nuclear punctate condensates that dissolve on aliphatic-alcohol treatment 8. 1,6-hexanediol is a notoriously blunt instrument — it perturbs chromatin, kinase activity and general nuclear organisation — so alcohol sensitivity alone should not be reported as evidence of liquid-liquid phase separation.

Where AFF1 and the SEC actually sit — on chromatin, at paused polymerase

The functional address of the 3' half is better defined than its physical one. Purification of several commonly occurring MLL chimeras recovered the super elongation complex — ELL, P-TEFb (CDK9/cyclin T), AFF4 and others — associated with every chimera tested. AFF4 is required for SEC stability and for proper transcription by poised RNA polymerase II in metazoans, and AFF4 knockdown in leukemic cells reduces MLL-chimera target gene expression 5.

So the SEC's job is pause release: it sits at promoter-proximal paused Pol II and converts it to productive elongation. The phase-separation work adds a physical account of how that is regulated — P-TEFb is held inactive in a soluble HEXIM1 complex and becomes active by being concentrated into SEC droplets 3.

On genomic position, MLL-AF4 is not only a promoter factor. Capture-C at the BCL2 locus showed MLL-AF4 directly activating BCL2 and MCL1, with a BCL2-specific enhancer made of two H3K27ac clusters at the 3' end of the gene; loss of MLL-AF4 reduced both H3K79me3 in the gene body and H3K27ac at that 3' enhancer. The same study found MLL-AF4 repressing BIM via the PRC1 component CBX8 9. That is promoter and enhancer occupancy, with a documented physical contact between them — which is the kind of architecture the NUP98-HOXA9 loop result 6 predicts a condensate would help build, and a concrete place to test it.

What has not been shown for the SEC is co-localisation with canonical interchromatin-granule nuclear speckles, or with classical "transcription factories". Treat both phrases as unsupported for this system.

Is subnuclear localization itself a drug target?1 deeper

Three distinguishable ideas, in decreasing order of credibility.

1. Disrupt chromatin tethering, not the compartment. This is already the clinical strategy and it works: menin inhibitors break the MLL-N/menin interface and evict the fusion from its target genes. Note what that means for this node — the successful drugs target engagement (menin-LEDGF-CXXC), not compartment (SNL). The CXXC dissociation experiment 2 is why: you can be in the right dots and still be dead.

2. Modulate the condensate. There is now genuine in vivo proof of concept, but for a mutant ENL rather than an MLL fusion: an acetyl-binding inhibitor displaced mutant-ENL condensates from target loci, blocked the chromatin changes and delayed AML initiation and progression 4. Note the mechanism — the inhibitor worked by blocking a specific reader interaction, and condensate dispersal was the consequence. That is the tractable version. Generic "condensate-dissolving chemistry" is not.

3. Disrupt SNL1/SNL2 directly. Weakest of the three. SNL-1 and SNL-2 are two short, redundant, non-enzymatic, non-pocket-bearing peptide stretches with no known binding partner and no structure. There is no published ligand, no reported binding partner to compete with, and, from 2, no evidence that losing punctate localization is sufficient to block transformation. This is a concept, not a programme.

Bite-back specific to localization-directed strategies
  • Redundancy. Two independent SNLs plus three NTSs. Knocking out one element is unlikely to move the protein.
  • Wrong node in the chain. Localization may be an epiphenomenon of chromatin binding rather than its cause; 2 shows mutants can keep the compartment and lose the biology, and the converse (lose compartment, keep biology) has never been tested for a fusion.
  • Condensates are everywhere. SEC droplets form within minutes on serum stimulation and serve normal rapid gene induction 3. Anything that broadly modulates them hits an acutely needed general mechanism, not a leukemia-specific one — a therapeutic-index problem by construction. CDK9, ELL and BRD4 are all DepMap common-essential for the same reason.
  • Fluidity cuts both ways. AFF4 fluidises SEC droplets 3; a drug that makes condensates more solid could plausibly stabilise the oncogenic assembly rather than disperse it. Direction of effect is not guaranteed.
  • Assay artefact risk. If the field builds a KMT2A::AFF1 condensate programme on hexanediol sensitivity and overexpressed GFP constructs 8, it may be chasing an aggregation artefact. This is the main reason this node keeps the KMT2A::AFF1 condensate confidence at speculative.
Facts and figures
proteingene-kmt2a
uniprotQ03164
localization classnucleoplasmic, non-uniform punctate / speckled
residue ranges verifiedtrue
residue range source pmid9207083
fusion block retained AAN-terminal ~1400 residues of MLL (source E2)
elements
NTS-2 (co-maps with SNL-1)322.0NTS-2 (co-maps with SNL-1): 322SNL-1 (subnuclear localiz…388.0SNL-1 (subnuclear localization signal 1): 388NTS-3 (nuclear targeting …734.0NTS-3 (nuclear targeting sequence 3): 734SNL-2 (subnuclear localiz…1,051SNL-2 (subnuclear localization signal 2): 1051
elements — the numbers behind the chart
[
  {
    "name": "SNL-1 (subnuclear localization signal 1)",
    "start": 388,
    "end": 432,
    "function": "directs punctate (dotted) subnuclear distribution; conserved with Drosophila trithorax",
    "retained_in_fusions": true,
    "source": "E1"
  },
  {
    "name": "SNL-2 (subnuclear localization signal 2)",
    "start": 1051,
    "end": 1089,
    "function": "directs punctate (dotted) subnuclear distribution; conserved with Drosophila trithorax; lies immediately upstream of the CXXC domain",
    "retained_in_fusions": true,
    "source": "E1"
  },
  {
    "name": "NTS-3 (nuclear targeting sequence 3)",
    "start": 734,
    "end": 739,
    "motif": "RKRKRK",
    "function": "classical basic nuclear localization signal",
    "retained_in_fusions": true,
    "source": "E1"
  },
  {
    "name": "NTS-2 (co-maps with SNL-1)",
    "start": 322,
    "end": 480,
    "function": "nuclear targeting; region co-maps with SNL-1",
    "retained_in_fusions": true,
    "source": "E1"
  }
]
fusions shown punctate
[
  "ALL-1/AF-4 (KMT2A::AFF1)",
  "ALL-1/AF-9 (KMT2A::MLLT3)",
  "ALL-1/AF-17 (KMT2A::MLLT6)"
]
domains giving uniform distribution
[
  "PHD/zinc fingers",
  "SET"
]
condensate status
[
  {
    "entity": "ENL (MLLT1) IDR",
    "claim": "sufficient to initiate SEC liquid droplet formation",
    "confidence": "strong",
    "source": "E3"
  },
  {
    "entity": "AFF4",
    "claim": "fluidises SEC droplets; component of SEC droplets",
    "confidence": "strong",
    "source": "E3"
  },
  {
    "entity": "MLL-ENL (KMT2A::MLLT1)",
    "claim": "boosts SEC phase separation",
    "confidence": "moderate",
    "source": "E3"
  },
  {
    "entity": "ENL gain-of-function mutant",
    "claim": "forms condensates in HSPCs at leukemogenic genes in vivo; condensate-disrupting mutagenesis impairs oncogenic function; inhibitor displaces condensates",
    "confidence": "strong",
    "source": "E4"
  },
  {
    "entity": "NUP98-HOXA9",
    "claim": "IDR-driven LLPS puncta required for transformation; drives super-enhancer-like binding and CTCF-independent loops",
    "confidence": "strong",
    "source": "E6"
  },
  {
    "entity": "KMT2A::AFF1 (MLL-AF4)",
    "claim": "no primary evidence of condensate formation found; condensate framing rests on analogy to ENL/SEC and NUP98-HOXA9",
    "confidence": "speculative",
    "source": "inference from E3, E5, E6"
  },
  {
    "entity": "AFF1 pSER domain",
    "claim": "no published droplet reconstitution or IDR-deletion condensate experiment located",
    "confidence": "speculative",
    "source": "literature gap"
  }
]
chromatin address
{
  "tethering_module": [
    "gene-kmt2a N-terminus",
    "gene-men1",
    "gene-psip1"
  ],
  "dna_recognition": "CXXC domain binds non-methylated CpG; point mutants lose CpG binding, transactivation and transformation while retaining subnuclear localization (E2)",
  "elongation_function": "SEC-mediated RNA Pol II promoter-proximal pause release (E5)",
  "genomic_features_occupied": [
    "promoters",
    "enhancers (e.g. BCL2 3' H3K27ac enhancer)"
  ],
  "speckle_marker_colocalisation": "not tested in any source located; claims of SC35/snRNP nuclear-speckle co-localisation are unsupported",
  "transcription_factory_colocalisation": "not tested in any source located; unsupported"
}
druggability
{
  "engagement_layer": "validated clinically (menin inhibitors evict the fusion from target genes)",
  "condensate_layer": "proof of concept in vivo for mutant ENL via a reader-pocket inhibitor (E4); no KMT2A::AFF1 data",
  "snl_layer": "no ligand, no known binding partner, no structure, redundancy across two elements; concept only"
}
How this could bite us back
  • Redundancy: two independent subnuclear localization signals (aa 388-432 and 1,051-1,089) plus three nuclear targeting sequences, all inside the obligatory N-terminal ~1,400-residue fusion block. Any single-element disruption is likely to be absorbed.
  • Wrong layer: the CXXC dissociation experiment shows a fusion can hold its subnuclear position and still be transformation-dead, so compartment may be a consequence of chromatin binding rather than its cause — making localization a biomarker, not a target.
  • Therapeutic index: SEC droplets are a general mechanism for rapid, serum-responsive gene induction, and the core components (CDK9, ELL, BRD4) are broadly essential. Condensate-modulating chemistry that is not anchored to a fusion-specific pocket is a toxicity programme.
  • Direction of effect is unknown: AFF4 fluidises SEC droplets, so a compound that rigidifies condensates could entrench the oncogenic assembly instead of dissolving it.
  • Assay artefact: the condensate literature in this space leans on overexpressed fluorescent constructs and 1,6-hexanediol sensitivity, both of which can manufacture the phenotype. A KMT2A::AFF1 condensate programme built on those readouts could be chasing an aggregate.
  • If the fusion's foci are phase-separated, drug partitioning becomes a real problem: a small molecule may be excluded from, or sequestered by, the condensate, decoupling biochemical potency from cellular effect.
Still open
  1. Does KMT2A::AFF1 form condensates at all? Nobody has published live-cell imaging plus FRAP on an endogenously tagged fusion in a t(4;11) line. This is the single cheapest high-value experiment in this node.
  2. Is losing punctate localization sufficient to block transformation? Ayton et al. showed the converse (transformation lost, localization kept) [E2], but no one has built an SNL1/SNL2 double-mutant fusion and asked whether it still transforms. Without that, 'localization matters' is untested in the direction that would make it a target.
  3. Does the AFF1 pSER region behave as a phase-separating IDR in isolation, and is it required for transformation independently of its role in SEC assembly and in SIAH-mediated turnover?
  4. Do MLL fusion foci co-localise with canonical nuclear speckles (SC35/SRRM2) or are they a distinct compartment? The literature uses 'speckle' loosely and the experiment appears never to have been done with modern markers.
  5. Does menin inhibition disperse the fusion's nuclear foci, or evict it from chromatin while leaving the foci intact? The answer discriminates whether the foci are chromatin-templated or self-assembling, and it can be read out in existing revumenib-treated cell lines.
  6. Do KMT2A::AFF1 target loci show CTCF-independent chromatin loops of the kind NUP98-HOXA9 produces [E6]? Deeply sequenced Hi-C on a t(4;11) line with and without fusion degradation would settle it.
Every source behind this entry (9)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Native ALL-1/MLL shows a nuclear punctate distribution conferred by two trithorax-conserved subnuclear localization signals, SNL-1 at amino acids 388-432 and SNL-2 at amino acids 1,051-1,089, together with three nuclear targeting sequences including a classical RKRKRK motif at residues 734-739; chimeric ALL-1/AF-4, ALL-1/AF-9 and ALL-1/AF-17 proteins show the same speckles, while the PHD fingers and SET motif give uniform non-punctate patterns.

    Nuclear punctate distribution of ALL-1 is conferred by distinct elements at the N terminus of the protein · Yano T, Nakamura T, Blechman J, Sorio C, Dang CV, Geiger B, Canaani E · Proceedings of the National Academy of Sciences of the USA 1997
    PMID 9207083DOI
    What this evidence cannot show

    Residue ranges were read directly from the PMC full text of this paper and are reported as published. The work relies heavily on transfected epitope-tagged fragments overexpressed in COS and HeLa cells, where aggregation can mimic specific foci; 1997-era immunofluorescence cannot distinguish a condensate from an aggregate from a clustered chromatin-bound pool. The paper does not report co-staining with canonical nuclear-speckle markers (SC35, snRNP), so it does not license the claim that MLL resides in interchromatin granule clusters. Numbering is for the human ALL-1/MLL isoform used in that paper and should be re-checked against the current KMT2A reference (UniProt Q03164) before being used to design a mutant.

    Evidence that argues against this (1)
    • Punctate subnuclear localization is not sufficient for oncogenic function: MLL mutants that lose myeloid transformation can retain normal subnuclear localization, so the speckled pattern is not by itself the mechanism. Ayton PM et al. 2004
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL fusion oncoproteins invariably retain the KMT2A N-terminus, conventionally given as the amino-terminal ~1,400 residues although the exact extent varies with breakpoint; high-resolution mutagenesis identified three domains essential for in vitro myeloid transformation by mechanisms that do not compromise subnuclear localization, and CXXC point mutations that block transformation also abolish non-methylated CpG binding and transactivation.

    PMID 15542854DOI
    What this evidence cannot show

    In vitro myeloid colony-forming transformation with a single MLL fusion partner in mouse bone marrow, not an in vivo leukemia model; 'does not compromise subnuclear localization' was scored by fluorescence microscopy of transfected constructs, which has limited sensitivity to partial or quantitative relocalisation. The residue boundaries of the three essential domains were not read from the full text for this node (the PMC page was not retrievable), so only the qualitative claim is asserted here.

    Evidence that argues against this (1)
    • The retention of the KMT2A N-terminus is genuinely without exception; what is not constant is its EXTENT, and the entry previously stated the two together as though both were fixed. Breakpoints fall across introns 9 to 11 with an age gradient, and spliced fusions retain a different complement again, so the amino-terminal ~1,400 residues is a convention over a distribution. The entry body was made precise on 2026-09-19 rather than merely annotated. Separately, the three domains identified as essential were mapped in ONE chimera in an in vitro myeloid transformation assay, and a partner contributing nothing but oligomerisation - gephyrin's 15-residue tubulin-binding domain - can still transform, so necessity in that assay does not establish that every fusion uses those same three domains. Molecular and Cellular Biology 2004
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The super elongation complex forms liquid droplets by multivalent phase separation: ENL or its intrinsically disordered region alone is sufficient to initiate SEC droplet formation, AFF4 acts with ENL to fluidise the droplets, droplets concentrate P-TEFb away from the soluble inactive HEXIM1-P-TEFb pool and are required for rapid transcriptional induction, and fusion of ENL with MLL boosts SEC phase separation.

    PMID 32270036DOI
    What this evidence cannot show

    Droplet assays rest substantially on ectopic expression of fluorescent fusions and in vitro reconstitution, both of which favour condensate formation relative to endogenous stoichiometry. The MLL-ENL result is that the fusion 'boosts' SEC phase separation, which is not the same as showing the fusion itself is the condensate scaffold, and it says nothing directly about KMT2A::AFF1. Serum-response kinetics are from cell lines, not primary haematopoietic cells.

    Evidence that argues against this (1)
    • SEC droplet formation is demonstrated with purified components and in overexpression, where concentration is set by the experimenter and is the single variable that determines whether a multivalent protein demixes. Puncta, FRAP recovery and alcohol sensitivity are all produced by DNA-mediated clustering without phase separation, and only half-bleach internal mixing separates the two. eLife 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A condensate-promoting ENL mutation drives AML in a conditional knock-in mouse; mutant ENL forms condensates in haematopoietic stem/progenitor cells at key leukemogenic genes, mutagenesis that disrupts condensate formation impairs its chromatin and oncogenic function, and an acetyl-binding inhibitor displaces the condensates from target loci and delays AML initiation and progression in vivo.

    Condensate-Promoting ENL Mutation Drives Tumorigenesis In Vivo Through Dynamic Regulation of Histone Modifications and Gene Expression · Liu Y, Li Q, Song L, Gong C, Tang S, Budinich KA, Vanderbeck A, Mathias KM, Wertheim GB, Nguyen SC, Outen R, Joyce EF, Maillard I, Wan L · Cancer Discovery 2024
    PMID 38655899DOI
    What this evidence cannot show

    This is a point-mutant ENL, not an MLL fusion. Transfer of the conclusion to KMT2A::AFF1 is an assumption, not a result. The inhibitor blocks a specific acetyl-lysine reader interaction, so condensate displacement is downstream of a conventional pocket-binding event; the study does not show that condensates can be drugged as condensates.

    Evidence that argues against this (1)
    • A condensate-promoting ENL mutation driving AML in mouse is the strongest condensate evidence in this beat and it remains correlative on the key point: mutagenesis that disrupts condensate formation also disrupts the multivalent interactions that mediate chromatin binding, so an allele cannot separate the two properties it is being used to separate. eLife 2019
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Purification of several commonly occurring MLL chimeras recovered the super elongation complex (ELL, P-TEFb, AFF4 and other factors) associated with every chimera tested; AFF4 is required for SEC stability and for proper transcription by poised RNA polymerase II, and AFF4 knockdown reduces MLL-chimera target gene expression in leukemic cells.

    AFF4, a component of the ELL/P-TEFb elongation complex and a shared subunit of MLL chimeras, can link transcription elongation to leukemia · Lin C, Smith ER, Takahashi H, Lai KC, Martin-Brown S, Florens L, Washburn MP, Conaway JW, Conaway RC, Shilatifard A · Molecular Cell 2010
    PMID 20159561DOI
    What this evidence cannot show

    Biochemical purification from cell lines with tagged constructs; composition of a purified complex does not establish stoichiometry or physical organisation in the nucleus, and says nothing about spatial distribution. AFF4 knockdown phenotypes are confounded by AFF4's requirement in normal poised-Pol II transcription.

    Evidence that argues against this (3)
    • 'Every chimera tested' is the load-bearing phrase and the tested set is narrow. The purified chimeras are elongation-class partners, which is the class that recovers SEC by construction. This project's non-elongation-partner beat shows the generalisation does not hold across partner classes: septins, ABI1 and gephyrin have no elongation machinery, and a 15-residue gephyrin tubulin-binding domain from a protein with no transcriptional activity is sufficient to transform. SEC recruitment is what one partner class does, not what every chimera does. SOURCING CORRECTION 2026-09-20 (Q40 counter-evidence quality sweep): the gephyrin sentence above is NOT in the paper cited on this note. It comes from Eguchi et al. 2004, which this graph already cites in ten other entries. The note carrying the claim's own PMID made an independent finding look like a same-source caveat. The primary source is now given as the note below. Molecular Cell 2010
    • INDEPENDENT SOURCE for the gephyrin sentence in the preceding note, added 2026-09-20 by the Q40 sweep after fetching the record. MLL-GPHN transforms haematopoietic progenitors DESPITE gephyrin having no transcriptional activity, and a 15-amino-acid tubulin-binding domain of GPHN is necessary and sufficient for that activity in vitro and in vivo, conferring oligomerisation on MLL; transduced progenitors gave both myeloid and lymphoid leukaemias in mice. A partner contributing nothing but self-association is therefore sufficient, which is the counter-example to 'every chimera recovers SEC' and it does not come from the SEC purification paper. Eguchi M et al. 2004
    • The same beat bounds the substitution logic that would otherwise rescue the generalisation. Forced self-association is the better class-wide description, but 'any dimeriser will do' is falsified twice - MLL-SEPT6 lacking its coiled coil failed to immortalise, and heterologous RA domains did not substitute for AFDN RA1. So neither SEC recruitment nor generic oligomerisation is the universal mechanism this item implies. Leukemia 2023
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    NUP98-HOXA9 requires its nucleoporin intrinsically disordered FG-repeat region to form liquid-liquid phase separated puncta and to induce leukemic transformation; phase separation promotes chromatin occupancy, generates a broad super-enhancer-like binding pattern at leukaemogenic genes, and induces CTCF-independent chromatin loops enriched at proto-oncogenes, and the effect is reproduced when the FG repeats are replaced by an unrelated LLPS-forming FUS IDR.

    Phase separation drives aberrant chromatin looping and cancer development · Ahn JH, Davis ES, Daugird TA, Zhao S, Quiroga IY, Uryu H, et al. · Nature 2021
    PMID 34163069DOI
    What this evidence cannot show

    This is a NUP98 fusion, a structurally different oncoprotein class from KMT2A fusions: its IDR is a tandem FG repeat contributed by the 3' partner's 5' component, and the chimera is a homeodomain transcription factor rather than a CXXC/menin-tethered chromatin adaptor. Using it to argue that KMT2A::AFF1 forms condensates is explicit reasoning by analogy across fusion families, and this node treats it as such.

    Evidence that argues against this (1)
    • The FG-repeat requirement is real and the IDR-swap experiment that accompanies it in this literature shows an UNRELATED IDR restoring function, which means the contribution is generic rather than sequence-specific. A generic requirement is consistent with clustered binding as well as with phase separation. eLife 2019
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Menin acts as a molecular adaptor that physically links MLL to LEDGF/PSIP1 on chromatin; LEDGF is required both for MLL-dependent transcription and for leukemic transformation, and a subset of MEN1 patient mutations abrogate the LEDGF interaction while preserving MLL binding yet still compromise MLL/menin-dependent function.

    PMID 18598942DOI
    What this evidence cannot show

    Establishes a biochemical and functional link, not a subnuclear imaging result: this paper does not show that menin or LEDGF form or co-localise in discrete nuclear puncta with the fusion. Claims about 'menin/LEDGF puncta' are therefore not supported by this citation and were not asserted in this node.

    Evidence that argues against this (1)
    • The menin-LEDGF adaptor result is about chromatin tethering and is used here for subnuclear localisation, which is a different scale: tethering explains which nucleosomes are bound, not why a protein appears concentrated in a nuclear region. The separation-of-function MEN1 alleles are also endocrine patient mutations read into a leukaemia assay. Cancer Research 2014
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    FUS co-localises with AFF4 within nuclear punctate condensates in live cells, and those condensates are disrupted by treating cells with aliphatic alcohol.

    Fused in sarcoma silences HIV gene transcription and maintains viral latency through suppressing AFF4 gene activation · Krasnopolsky S, Marom L, Victor RA, Kuzmina A, Schwartz JC, Fujinaga K, Taube R · Retrovirology 2019
    PMID 31238957DOI
    What this evidence cannot show

    Live-cell imaging of ectopically expressed fluorescent proteins in HIV latency models (HEK293, Jurkat), not haematopoietic progenitors, and not an MLL fusion context. Aliphatic-alcohol (1,6-hexanediol) sensitivity is a weak and much-criticised criterion for liquid-liquid phase separation: hexanediol also perturbs chromatin compaction and kinase signalling, so dissolution on treatment does not establish a liquid phase. Included here mainly as a methodological warning for anyone designing the missing KMT2A::AFF1 experiment.

    Evidence that argues against this (1)
    • This is the weakest item in the entry and it is doing condensate work. FUS co-localising with AFF4 in puncta disrupted by aliphatic alcohol is exactly the pair of observations that cannot distinguish phase separation from clustered binding — hexanediol condenses chromatin, and co-localisation in live cells is limited by diffraction. eLife 2019
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF4 occupies both promoters and enhancers: it directly activates BCL2 and MCL1 and binds a BCL2-specific enhancer composed of two H3K27ac clusters at the 3' end of the gene, with loss of MLL-AF4 reducing both gene-body H3K79me3 and 3' enhancer H3K27ac, while repressing BIM via the PRC1 component CBX8.

    MLL-AF4 binds directly to a BCL-2 specific enhancer and modulates H3K27 acetylation · Godfrey L, Kerry J, Thorne R, Repapi E, Davies JOJ, Tapia M, Ballabio E, Hughes JR, Geng H, Konopleva M, Milne TA · Experimental Hematology 2016
    PMID 27856324DOI
    What this evidence cannot show

    Performed in t(4;11) cell lines with shRNA knockdown of the fusion; knockdown kinetics confound direct from indirect effects, and Capture-C reports contact frequency rather than a functional enhancer-promoter requirement. Genomic occupancy is not subnuclear localization — this evidences where on the genome the fusion sits, not where in the nucleus.

    Evidence that argues against this (1)
    • Promoter and enhancer occupancy with a 3' H3K27ac-marked element at BCL2 is a genomic-location result and this entry is about SUBNUCLEAR location. Where in the genome a fusion binds does not determine where in the nucleus it concentrates, and the two are conflated wherever a beat runs both arguments. eLife 2021
Connected to