Oncofusion

start / graph / localization-condensate-evidence-standards

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q4b-localization

The case that the KMT2A-fusion condensate is an artefact — what 1,6-hexanediol, overexpressed tags and in-vitro droplets actually prove

also called hexanediol artefact · LLPS controls · phase separation methodological critique · condensate false positive

Almost every claim that a leukemia protein forms liquid droplets rests on three experiments, and all three have known ways of giving the wrong answer: a chemical that dissolves droplets also poisons the enzyme being measured, glowing tagged proteins clump when overproduced, and purified proteins in a tube form droplets at concentrations cells never reach. This entry states the strongest version of the sceptical case so that the graph's positive condensate claims can be graded honestly rather than repeated.

The same thing, for a scientist

Three standard assays underpin the condensate literature for AFF1/AFF4, ENL and AF9, and each has a documented failure mode. (1) **1,6-hexanediol.** Duster et al. showed that at 1% v/v hexanediol already strongly impairs kinases and phosphatases and partly blocks DNA polymerases, and that at the 5-10% used to dissolve droplets kinases and phosphatases are virtually inactive [E1]. This is not a generic caveat here: the super elongation complex's functional output is CDK9/P-TEFb phosphorylation of the Pol II CTD, so in any SEC-condensate experiment hexanediol independently abolishes the readout used to show the condensate mattered. The one AFF4-adjacent condensate report in the literature — FUS co-localising with AFF4 in nuclear puncta dispersed by aliphatic alcohol — rests precisely on this reagent [E5]. (2) **Overexpressed fluorescent constructs.** McSwiggen, Mir, Darzacq and Tjian surveyed the in-vivo evidence across processes and concluded it is often phenomenological and inadequate to discriminate phase separation from other mechanisms, urging measurement of proteins at endogenous abundance [E2]; Chong et al. then showed by live-cell single-molecule imaging that transcription-factor low-complexity domains form local high-concentration hubs that stabilise DNA binding, recruit Pol II and activate transcription **without detectable phase separation** at physiological levels, and that hexanediol sensitivity differs between LCDs [E3]. (3) **Fixation.** Irgen-Gioro et al. showed paraformaldehyde fixation can both enhance and diminish apparent LLPS, and for some proteins creates droplet-like puncta in cells that have none when live [E4]. The practical consequence for this graph: no KMT2A-fusion condensate claim currently clears this bar, and a claim resting only on hexanediol plus an overexpressed tag plus an in-vitro droplet should be graded weak.

Go deeper
1,6-hexanediol is not a specific probe, and for the SEC it is worse than non-specific

The reagent is used as though it were a selective solvent for weak multivalent interactions. Duster et al. tested it directly against enzymes in vitro: at 1% volume concentration it already strongly impaired kinases and phosphatases and partly blocked DNA polymerases (DNase activity was unaffected); at 5-10%, the range used to dissolve droplets, both kinases and phosphatases were virtually inactive 1.

Now apply that to the super elongation complex. The SEC's mechanism of action is delivery of active P-TEFb (CDK9-cyclin T1) to paused Pol II. The published SEC-droplet work reports that SEC components compartmentalise and concentrate P-TEFb away from the inactive HEXIM1-bound pool, that ENL or its IDR alone initiates droplet formation, that AFF4 fluidises the droplets, and that fusion of ENL with MLL boosts SEC phase separation 6. Any experiment in that series that uses hexanediol to test whether the droplet is needed for transcription is confounded at the level of the enzyme: hexanediol turns off CDK9 whether or not it dissolves anything. A negative transcription readout after hexanediol is therefore uninformative about condensates specifically.

The single AFF4-adjacent condensate observation in the whole literature — FUS co-localising with AFF4 in nuclear puncta in live cells, puncta disrupted by treating cells with aliphatic alcohol — is an HIV-latency study in HEK293 and Jurkat cells, not leukemia, with no KMT2A fusion present, and its condensate claim rests on exactly this reagent 5.

Overexpressed tagged constructs droplet readily; endogenous ones often do not1 deeper

McSwiggen et al. reviewed the in-vivo evidence across gene expression, cell division and stress responses and concluded that the evidence for LLPS in living cells is often phenomenological and inadequate to discriminate between phase separation and other possible mechanisms, with the causal relationship and functional consequences even more elusive; their explicit recommendation is quantitative measurement of proteins in their endogenous state and at physiological abundance 2.

Chong et al. supplied the positive control for the alternative. Live-cell single-molecule imaging showed TF low-complexity domains forming local high-concentration interaction hubs at synthetic and endogenous loci. The hubs stabilise DNA binding, recruit Pol II and activate transcription. The interactions are highly dynamic, selective for partners, differentially sensitive to hexanediols, and — under physiological conditions — occur without detectable phase separation 3. So the observable that condensate papers treat as diagnostic (a bright nuclear focus that concentrates machinery and drives transcription) is fully produced by a mechanism that is not phase separation.

The ENL literature contains its own version of this warning from inside the positive camp: expression of ENL mutants beyond the endogenous level produces non-functional condensates (already recorded in mechanism-condensates-load-bearing-or-correlative). Overexpression does not merely risk a false positive, it can generate structures whose behaviour is opposite to the real ones.

Fixation can manufacture the phenotype

Most leukemia imaging is done on fixed cells, often on cytospins of primary blasts, because that is what is available. Irgen-Gioro et al. compared images of the same cells before and after paraformaldehyde fixation and found PFA can both enhance and diminish apparent LLPS behaviour, and for specific proteins can cause droplet-like puncta to appear artificially in cells with no detectable puncta when live. Fixing in the presence of glycine reversed the direction of the artefact; a kinetic model attributes the effect to the difference between the fixation rate and the protein-protein interaction rate 4. Any claim that a KMT2A fusion forms condensates in fixed patient material inherits this artefact unless a live-cell or glycine-controlled comparison is shown.

In-vitro droplets at non-physiological concentration prove that a protein *can* demix, not that it does

Purified IDR-containing proteins droplet at low micromolar concentrations and low salt. The demonstration establishes a material property. It does not establish the intracellular concentration, the presence of competing partners, the crowding environment, or that the same demixing occurs at the endogenous copy number of a blast nucleus. The graph already records the cleanest published example of the gap: in NUP98-NSD1, interactome binding depends on the FG repeats that mediate condensates, SMARCA5 is a condensate constituent and is functionally essential, yet SMARCA5 inhibition does not affect the condensates — so condensate formation is not sufficient for the transformed phenotype.

And the direction of the effect is not even fixed. Chong et al. 2022 showed that increasing LCD-LCD interactions toward putative LLPS represses transcription of EWS::FLI1 target genes in patient-derived cells, and that deliberately sequestering EWS::FLI1 into the nucleolus — a well-documented LLPS compartment — inhibits its transcription and oncogenic transformation 7. More condensate is not more oncogenic. That finding should be read directly against the existing graph entry reporting MLL-AF4 as predominantly nucleolar: on the EWS::FLI1 precedent, nucleolar partition is a candidate off state, not an active one.

The bar a KMT2A-fusion condensate claim would have to clear

Stated as a checklist the lab can hold any future claim (including its own) against:

  1. Endogenous knock-in tag at the native fusion locus in a t(4;11) line, not a transfected construct, with the tagged allele's expression level measured against untagged.
  2. Live-cell imaging, or fixed imaging with a glycine-modulated fixation control 4.
  3. A separation-of-function mutant that abolishes demixing while preserving chromatin binding and partner recruitment — the design that carried the NUP98 and ENL claims — rather than a chemical.
  4. If a chemical is used at all, an orthogonal one, plus a demonstration that the functional readout survives the chemical in a condensate-independent control 1.
  5. Quantified intracellular concentration versus the in-vitro saturation concentration.
  6. FRAP or single-molecule kinetics distinguishing a dynamic hub from a demixed phase 3.

No published KMT2A-fusion claim meets even points 1 and 3.

Facts and figures
checklist items6
kmt2a claims meeting checklist0
artefact classes
[
  {
    "assay": "1,6-hexanediol dispersal",
    "failure_mode": "directly inactivates kinases and phosphatases including CDK9 at the concentrations used",
    "affects_claims_about": [
      "SEC/AFF4",
      "ENL",
      "FUS-AFF4 puncta"
    ]
  },
  {
    "assay": "overexpressed fluorescent construct",
    "failure_mode": "puncta form above a concentration threshold; supra-endogenous ENL mutant condensates are non-functional",
    "affects_claims_about": [
      "all published KMT2A-fusion imaging"
    ]
  },
  {
    "assay": "paraformaldehyde-fixed imaging",
    "failure_mode": "fixation can create or erase droplet-like puncta relative to the live cell",
    "affects_claims_about": [
      "fixed patient-material imaging"
    ]
  },
  {
    "assay": "in-vitro droplet with purified protein",
    "failure_mode": "demonstrates a material property at non-physiological concentration and ionic strength, not intracellular behaviour",
    "affects_claims_about": [
      "SEC droplets",
      "PRC1/CBX2 droplets"
    ]
  }
]
How this could bite us back
  • If the lab builds a drug-discovery campaign on disrupting a KMT2A-fusion condensate and the condensate is an artefact of the assay, every hit will be a hexanediol-like general kinase or interaction poison with no therapeutic window — the assay itself selects for promiscuity.
  • Conversely, if condensates are real but the direction is inverted (more demixing equals less transcription, as in EWS::FLI1), a condensate-disrupting drug would *activate* the oncogenic program. The sign of the effect has never been established for any KMT2A fusion and should be determined before any screen is designed.
  • Publishing a condensate claim for KMT2A::AFF1 that later fails replication would damage the lab's credibility in a field where the open ground is precisely that nobody else is working here; the reputational downside of a soft positive is larger than usual.
Still open
  1. Does hexanediol at 5% abolish KMT2A::AFF1 target-gene transcription in SEM cells even when the fusion is degron-depleted first? If the transcriptional collapse is the same with and without the fusion present, the reagent is measuring general kinase poisoning and every hexanediol result in this field needs re-reading.
  2. What is the actual intranuclear concentration of KMT2A::AFF1 in a t(4;11) blast? Nobody has measured it, and without it no in-vitro saturation concentration can be interpreted.
  3. Would an AFF1 IDR separation-of-function allele that abolishes demixing while preserving CHD dimerisation and SEC binding still transform? This is the only design that settles the question, and it is the experiment the graph should fund before any further imaging.
  4. Is the reported nucleolar partition of MLL-AF4 an active state or, as the EWS::FLI1 precedent suggests, an inactivating sequestration? These make opposite predictions for whether increasing nucleolar partition would be therapeutic.
Every source behind this entry (7)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    At 1% volume concentration 1,6-hexanediol strongly impairs kinases and phosphatases and partly blocks DNA polymerases while leaving DNase activity unaffected, and at the 5-10% concentrations usually used to dissolve liquid-liquid phase separated droplets both kinases and phosphatases are virtually inactive.

    PMID 33814344DOI
    What this evidence cannot show

    In-vitro enzyme assays on purified enzymes, not measurements inside cells; the paper does not show that any specific published condensate result is wrong, only that the reagent has a large confounding activity. The enzymes tested include CDK7 and CDK9, which is why the confound is acute for super-elongation-complex work, but the authors did not test the SEC itself.

    Evidence that argues against this (1)
    • There is a second, deeper objection to this reagent than its enzyme inhibition: it does not discriminate the mechanism it is used to test. Muzzopappa et al. benchmarked candidate LLPS assays against well-controlled model systems in vitro and in living cells and found that 1,6-hexanediol treatment AND classical FRAP both fail to distinguish liquid-liquid phase separation from the alternative of molecules binding to spatially clustered binding sites without phase-separating; only preferential internal mixing in half-bleach experiments separated the two, which is the basis of their MOCHA-FRAP workflow probing the condensate interfacial barrier (applied to heterochromatin foci, nucleoli, stress granules and nuage granules). So even a hexanediol experiment with perfect enzyme controls would not establish phase separation. This entry should require half-bleach or an equivalent interface measurement, not merely warn about kinases. Muzzopappa F et al. 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    Evidence for liquid-liquid phase separation in living cells is often phenomenological and inadequate to discriminate phase separation from other possible mechanisms, and the causal relationship and functional consequences of LLPS in vivo are even more elusive; quantitative measurement of proteins in their endogenous state and at physiological abundance is required.

    Evaluating phase separation in live cells: diagnosis, caveats, and functional consequences · McSwiggen DT, Mir M, Darzacq X, Tjian R · Genes & Development 2019
    PMID 31594803DOI
    What this evidence cannot show

    A perspective/review, cited as an authoritative statement of methodological standards rather than as primary data, and written by a group that has publicly argued against the LLPS framing, so it is a partisan position in a live dispute. It does not evaluate any leukemia fusion protein.

    Evidence that argues against this (2)
    • The McSwiggen/Tjian verdict that in-cell LLPS evidence is 'phenomenological and inadequate' is a position in a live dispute, not a settled finding, and the opposing evidence is of comparable technical quality. Cho et al., published in the same journal and year as the Chong hub paper, used live-cell superresolution and light-sheet imaging of Mediator and RNA Pol II in mouse embryonic stem cells and found large stable clusters that colocalise, associate with chromatin, show properties of phase-separated condensates, and are sensitive to transcriptional inhibitors. This is live-cell imaging of endogenous transcription machinery, not an overexpressed reporter, so it answers the specific objection E2 raises. Read: abstract only. Cho WK et al. 2018
    • The field's own consensus methods paper takes a middle position that this entry's framing flattens. Alberti, Gladfelter and Mittag state that evidence is mounting that LLPS underlies membraneless compartments, and then set out guidelines for rigorous characterisation in vitro and in cells together with the caveats of common approaches. Their conclusion is that LLPS claims need better controls, not that the in-vivo evidence is generally uninformative. Cited here so the entry is not read as 'the field has rejected LLPS'; it has not. Read: abstract only. Alberti S et al. 2019
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Live-cell single-molecule imaging shows transcription-factor low-complexity domains forming local high-concentration interaction hubs at synthetic and endogenous genomic loci that stabilise DNA binding, recruit RNA Pol II and activate transcription; these LCD-LCD interactions are highly dynamic, partner-selective, differentially sensitive to hexanediols, and under physiological conditions occur without detectable phase separation.

    Imaging dynamic and selective low-complexity domain interactions that control gene transcription · Chong S, Dugast-Darzacq C, Liu Z, Dong P, Dailey GM, Cattoglio C, Heckert A, Banala S, Lavis L, Darzacq X, Tjian R · Science 2018
    PMID 29930090DOI
    What this evidence cannot show

    Model TFs (including FET-family LCDs) in cultured cells with synthetic arrays as well as endogenous loci; 'without detectable phase separation' is bounded by the sensitivity of the imaging, and absence of detection is not proof of absence. Neither KMT2A nor AFF1 was tested.

    Evidence that argues against this (1)
    • The entry's checklist says no fusion-oncoprotein condensate claim meets point 3 (a separation-of-function mutant that abolishes demixing while preserving chromatin binding). For a leukemia fusion that is no longer true. Ahn et al. show the NUP98 FG-repeat IDRs are required both for LLPS puncta of NUP98-HOXA9 and for leukaemic transformation; that LLPS promotes chromatin occupancy and broad super-enhancer-like binding; that replacing the FG repeats with an unrelated LLPS-competent FUS IDR restores genome-wide binding and target activation; and that Hi-C detects CTCF-independent loops at proto-oncogenes only in the phase-separating form. That is an IDR-swap separation-of-function design in a fusion oncoprotein. It remains true that no KMT2A-fusion claim meets the bar, but the bar is clearable and has been cleared next door. Read: abstract only. Ahn JH et al. 2021
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Paraformaldehyde fixation can both enhance and diminish apparent liquid-liquid phase separation behaviour in cells, and for specific proteins can cause droplet-like puncta to appear artificially in cells that show no detectable puncta when live; fixing in the presence of glycine reverses the direction of the artefact.

    Fixation can change the appearance of phase separation in living cells · Irgen-Gioro S, Yoshida S, Walling V, Chong S · eLife 2022
    PMID 36444977DOI
    What this evidence cannot show

    Demonstrated for a panel of intrinsically disordered proteins in cultured cells; the magnitude and even the direction of the artefact is protein-specific, so it cannot be used to argue that any particular published image is wrong, only that the control was not done.

    Evidence that argues against this (1)
    • 'The control was not done' understates the position: a positive, discriminating assay exists, so the standard this entry sets can be raised from a caveat to a requirement. Muzzopappa et al. showed that hexanediol and classical FRAP cannot separate phase separation from binding to clustered sites, that preferential internal mixing in half-bleach experiments robustly can, and used the resulting MOCHA-FRAP workflow to measure interfacial barrier strength across heterochromatin foci, nucleoli, stress granules and nuage granules - finding the barrier increases in that order, i.e. that 'condensate' is a graded property rather than a binary one. The fixation artefact is therefore one of several reasons fixed-cell puncta are weak evidence, and the constructive standard for any KMT2A-adjacent condensate claim is live-cell half-bleach at endogenous expression. Muzzopappa F et al. 2022
  5. E5weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    FUS co-localises with AFF4 within nuclear punctate condensates in live cells and those condensates are disrupted by treating cells with aliphatic alcohol; this is an HIV-latency study in HEK293 and Jurkat cells with no KMT2A fusion present.

    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

    The condensate observation is a single live-imaging panel in a virology paper whose subject is HIV latency; the dispersal control is the aliphatic alcohol that E1 shows inactivates kinases including CDK9, the SEC's own kinase. It is nonetheless the only AFF4-adjacent condensate report in the literature, which is itself the point.

    Evidence that argues against this (1)
    • The aliphatic-alcohol control in that study is compromised twice over, and the second way is specific to a chromatin-associated punctum. Itoh et al. used live-cell single-nucleosome imaging to show that 1,6-hexanediol drastically suppresses chromatin motion and hyper-condenses chromatin in human cells in a dose-dependent way, with chromatin effectively 'frozen' at 5% or above, and showed in vitro that 1,6-HD greatly facilitates cation-dependent chromatin condensation - an action the authors state is distinct from its droplet-melting activity, and which they attribute to removal of water molecules around chromatin. Their explicit conclusion is that droplet results obtained with 1,6-HD should be carefully reinterpreted when the droplets are associated with chromatin. A nuclear FUS/AFF4 punctum is by definition chromatin-associated, so its dispersal by aliphatic alcohol carries essentially no mechanistic information. Itoh Y et al. 2021
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Super elongation complex components compartmentalise and concentrate P-TEFb by liquid-liquid phase separation away from the soluble inactive HEXIM1-containing pool; ENL or its intrinsically disordered region alone initiates SEC droplet formation, AFF4 fluidises the droplets, droplets form rapidly on serum exposure and are required for rapid transcriptional induction, and fusion of ENL with MLL boosts SEC phase separation.

    PMID 32270036DOI
    What this evidence cannot show

    This is the strongest positive SEC condensate paper and is listed here because it is the claim the critique applies to. It uses recombinant proteins in vitro and transfected constructs in cell lines; the MLL-ENL result is 'boosts phase separation' in that system, not a demonstration in a leukemia cell at endogenous level, and MLL::AFF1 was not tested.

    Evidence that argues against this (1)
    • Additional and independent reason the hexanediol arm of the SEC condensate case cannot be rescued by dose-lowering or by orthogonal alcohols. Itoh et al. used live-cell single-nucleosome imaging in human cells and found 1,6-hexanediol drastically suppresses chromatin motion and hyper-condenses chromatin dose-dependently, with chromatin effectively frozen at 5 per cent or above; the effect is distinct from the reagent's droplet-melting activity and is reproduced in vitro as cation-dependent chromatin condensation. For a chromatin-bound complex such as the SEC on paused Pol II, hexanediol therefore has at least three simultaneous actions - droplet melting, kinase inactivation (E1) and chromatin freezing - and no experiment using it can attribute a transcriptional effect to any one of them. Read: abstract only. Itoh Y et al. 2021
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Increasing low-complexity-domain interactions toward putative LLPS represses transcription of EWS::FLI1 target genes in patient-derived Ewing sarcoma cells, and ectopically sequestering EWS::FLI1 into the nucleolus, a well-documented LLPS compartment, inhibits EWS::FLI1-driven transcription and oncogenic transformation.

    Tuning levels of low-complexity domain interactions to modulate endogenous oncogenic transcription · Chong S, Graham TGW, Dugast-Darzacq C, Dailey GM, Darzacq X, Tjian R · Molecular Cell 2022
    PMID 35483357DOI
    What this evidence cannot show

    Ewing sarcoma, not leukemia; the sequestration is engineered rather than physiological, so it shows what forced nucleolar partition does to a fusion TF rather than what any fusion does naturally. Quantitative single-cell and single-molecule imaging in patient-derived cells is nonetheless a higher standard than any KMT2A condensate study.

    Evidence that argues against this (1)
    • The sign of the effect is disputed within the FET-fusion literature itself, so the entry's inference that 'more condensate equals less transcription, therefore nucleolar partition of MLL-AF4 is plausibly an off state' is not safe. Zuo et al. built an in-vitro single-molecule assay to watch phase separation on DNA and found FET fusion proteins phase-separate specifically at their target binding loci, that the resulting condensates recruit RNA Pol II and enhance transcription, and that there is a threshold number of fusion-binding DNA elements above which condensate formation is favoured. That is the opposite sign to Chong et al. 2022 in a system that is more reductionist but directly reads transcription. The honest position is that the sign is unresolved, which if anything strengthens the entry's own bite-back warning that a condensate-disrupting drug could activate rather than silence the oncogenic program. Read: abstract only. Zuo L et al. 2021
Connected to