Oncofusion

start / graph / interface-ell-eaf1-phospho-gated-condensate

Interfaceweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q25-partner-phospho

ELL-EAF1: a phosphorylation-gated interface reported to drive phase separation - and the assays used cannot distinguish that from clustered binding

also called ELL phase separation · ELL EAF1 condensate · ATM ELL condensate · ELL DNA-PKcs Ku

ELL and its partner EAF1 are reported to form liquid-like droplets that a damage kinase makes form more readily, which if true is a kinase-controlled droplet on a piece of protein the leukaemia fusion keeps. But the tests used to call something a droplet cannot tell a droplet apart from a lot of protein bound to nearby sites, which is a different thing.

The same thing, for a scientist

ELL is reported to phase separate intrinsically in vitro and in cells; association with EAF1 enhances that phase separation and changes the material properties of ELL-EAF1 condensates. Physiologically, genotoxic stress is reported to cause ATM-mediated phosphorylation-dependent increased EAF1 binding, enhancing phase separation and increasing ELL interaction with DNA-PKcs and the Ku complex for their recruitment to chromatin, repair, transcriptional restart and cell survival; an EAF1 knockdown or an ELL mutant that cannot show enhanced EAF1 interaction during damage fails at all of these. The interface region and the ATM sites are both retained in MLL-ELL. **This entry is written at weak-to-moderate confidence for two independent reasons**: the primary report could not be read in full by any of this project's four routes, and the standing condensate caveat applies with full force - 1,6-hexanediol sensitivity and classical FRAP recovery cannot distinguish liquid-liquid phase separation from binding to spatially clustered sites, and hexanediol itself immobilises and condenses chromatin in living cells.

Go deeper
What the accessible record asserts

From the abstract, which is all that could be obtained 1:

  • ELL has intrinsic phase-separation ability in vitro and in mammalian cells.
  • EAF1 association enhances it and changes the material property of ELL-EAF1 condensates.
  • Under genotoxic stress, ATM-mediated phosphorylation-dependent increased EAF1 binding enhances phase separation and changes ELL's material property.
  • This increases ELL interaction with DNA-PKcs and Ku complex components, important for their chromatin recruitment, DNA repair and transcriptional restart.
  • EAF1 knockdown, or an ELL mutant that fails to show enhanced EAF1 interaction during damage, fails at repair, restart and survival.

Nothing beyond this is asserted here. The specific ELL mutant, the assays used to establish phase separation, the concentrations, the cell types and the effect sizes are all unknown to this entry.

Why the condensate claim is held at weak confidence regardless of readability

This graph carries a standing rule, and it applies here directly:

  • 1,6-hexanediol sensitivity and classical FRAP recovery both fail to distinguish LLPS from binding to spatially clustered sites 2. Only model-free calibrated half-bleaching, which measures internal mixing, discriminates them.
  • 1,6-hexanediol rapidly immobilises and condenses chromatin in living human cells 3, an action the authors state is distinct from melting a droplet and which they explicitly instruct readers to use in reinterpreting chromatin-associated droplet results.
  • At physiological expression, low-complexity-domain interactions that control transcription form dynamic hubs without detectable phase separation 4.

This is not a claim that ELL does not condense. 'Binding to spatially clustered sites' is a substantive rival model and in a chromatin context the more interesting one, because clustered binding predicts sequence-specific targeting where phase separation predicts bulk enrichment. The point is that the record available here does not say which assay was used, so the model cannot be assigned.

The ENL condensate literature has already been through this correction cycle: the Wan laboratory's own work shows ENL mutant puncta are expression-level dependent, with near-100% FRAP recovery on overexpression and only partial recovery at endogenous levels, and the authors state that expression beyond endogenous levels gives non-functional condensates 5. Any condensate claim about an elongation factor should be read with that in front of it.

Retained, and the odd consequence

MLL-ELL contains ELL 46-621, so both the EAF1-interaction machinery in the OCEL region and the ATM sites T551/S561/S589 are retained. If phospho-gated ELL-EAF1 condensation is real, the fusion should be able to do it.

The consequence is strange and worth flagging: this mechanism says phospho-ELL recruits DNA-PKcs - and this beat's other headline says DNA-PKcs phosphorylates AF9 S395 to expel AF9 from the SEC, which contains ELL. The two partners of KMT2A are wired to the same kinase in opposite directions. Nobody has put those two results side by side, and this graph now does.

Facts and figures
partner agene-ell
partner bgene-eaf1
kinase gateATM, via phosphorylation of ELL (T551/S561/S589 per the companion paper)
gated contactELL-EAF1 binding, phase separation, and downstream ELL interaction with DNA-PKcs and Ku
retained in fusiontrue
retained basisMLL-ELL contains ELL aa 46-621; both the OCEL EAF1 module and the ATM sites lie within it
mapped in fusionfalse
primary paper readablefalse
assay discrimination statusunknown - the paper's assays could not be read, so LLPS cannot be distinguished from clustered binding
druggabilityNo ligand is known for the ELL-EAF1 interface. It is a protein-protein interface on an intrinsically disordered-adjacent module, and the condensate framing does not by itself make it tractable.
unreadable routes tried
[
  "EPMC search EXT_ID core -> no pmcid",
  "NCBI PMC ID converter -> Identifier not found in PMC",
  "fullTextXML / efetch db=pmc / PMC HTML unreachable without a PMCID"
]
How this could bite us back
  • Condensate-disrupting strategies are a live therapeutic idea in this disease, and this entry is a warning about the evidence they would rest on: an interface called a condensate on assays that cannot distinguish it from clustered binding is not a validated target.
  • If ELL-EAF1 condensation is really a DNA-repair mechanism, disrupting it pharmacologically would impair repair in every proliferating cell, which is a genotoxic liability rather than a selective one.
  • The ELL-EAF1 interaction is the same contact this graph records as critical for MLL-ELL immortalisation. A molecule that blocked it would be blocking a normal damage-response interaction at the same time, and there is no reported structural basis on which to try to separate the two.
Still open
  1. Which assays did the ELL phase-separation study use? If it used only hexanediol and bulk FRAP, the conclusion is compatible with clustered binding and the entry should be rewritten.
  2. Does MLL-ELL form condensates, and are they phospho-gated? No one has looked at the fusion.
  3. DNA-PKcs phosphorylates AF9 to expel it from the SEC, and phospho-ELL is reported to recruit DNA-PKcs. Is DNA-PKcs a coordinator of SEC disassembly during damage, and is that why two different KMT2A partners both route through it?
  4. Can calibrated half-bleaching be applied to ELL-EAF1 to settle the LLPS-versus-clustered-binding question directly? The method exists and is published.
Every source behind this entry (5)
  1. E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    ELL phase separates intrinsically; EAF1 association enhances it and alters condensate material properties; under genotoxic stress ATM-mediated phosphorylation-dependent increased EAF1 binding enhances phase separation and increases ELL interaction with DNA-PKcs and Ku complex components, which is required for repair, transcriptional restart and cell survival.

    PMID 41424126DOI
    What this evidence cannot show

    ABSTRACT ONLY. Full text could not be obtained on 2026-09-19 by any of this project's four routes: the Europe PMC core record for EXT_ID:41424126 returns no pmcid; the NCBI PMC ID converter returns 'Identifier not found in PMC'; and without a PMCID the fullTextXML, efetch db=pmc and PMC HTML routes cannot be attempted. Confidence is weak for that reason AND because the assays used to support the phase-separation claim are unknown and the commonest such assays cannot distinguish LLPS from clustered binding. From the same laboratory as the two other ELL entries in this beat, so not independent of them.

    Evidence that argues against this (1)
    • The two commonest assays for phase separation, 1,6-hexanediol sensitivity and classical FRAP recovery, cannot distinguish liquid-liquid phase separation from binding to spatially clustered sites; only model-free calibrated half-bleaching measuring internal mixing discriminates the two models. Unless this study used half-bleaching, its condensate claim is compatible with clustered binding. Muzzopappa F et al. 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    1,6-hexanediol sensitivity and classical FRAP recovery both fail to distinguish liquid-liquid phase separation from binding to spatially clustered sites; model-free calibrated half-bleaching, which measures internal mixing, is required to discriminate them.

    PMID 36526633DOI
    What this evidence cannot show

    A methods paper. It establishes what the assays cannot do, not that any particular published condensate is misassigned.

    Evidence that argues against this (1)
    • The methodological critique does not make condensates artefactual. Work on cancer-associated ENL condensates at endogenous expression reports submicron chromatin-bound condensates with partial FRAP recovery correlating with target-gene hyperactivation, and extensive mutagenesis establishing condensate formation as a driver of oncogenic gene activation - a case built on more than hexanediol and bulk FRAP. Song L et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    1,6-hexanediol rapidly immobilises and condenses chromatin in living human cells, an action distinct from dissolving a liquid droplet, and the authors explicitly instruct readers to use this in reinterpreting chromatin-associated droplet results.

    PMID 33536240DOI
    What this evidence cannot show

    Established in human cell lines. It is an argument about the reagent, not about any specific protein.

    Evidence that argues against this (1)
    • The reagent critique does not reach studies that did not use it, and it is not known which assays the ELL phase-separation report used, because that report could not be read. Pal S et al. 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    At physiological expression levels, low-complexity-domain interactions that control gene transcription form dynamic hubs without detectable phase separation.

    PMID 29930090DOI
    What this evidence cannot show

    A single-molecule imaging study of specific transcription factors; generalising it to elongation factors is an extrapolation.

    Evidence that argues against this (1)
    • The opposing case for this protein family is made at endogenous expression with a chromatin-bound readout, so the overexpression objection does not straightforwardly apply to it. Song L et al. 2022
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ENL mutant puncta are strongly expression-level dependent: overexpressed ENL mutants form large droplet-like condensates showing approximately 100% FRAP recovery within 80 seconds, whereas at endogenous levels ENL mutants form submicron chromatin-bound condensates with only partial recovery, and expression beyond the endogenous level produces non-functional condensates.

    PMID 36272410DOI
    What this evidence cannot show

    Full text read via efetch db=pmc (PMC10071517) after fullTextXML returned HTTP 500. This is about ENL YEATS hotspot mutants in a Wilms-tumour/AML context, not about ELL or about a KMT2A fusion; it is cited here for the expression-level lesson, not for ELL.

    Evidence that argues against this (1)
    • The same study establishes condensate formation as a driver of oncogenic gene activation through extensive mutagenesis, so its expression-level caveat is a refinement of a positive result rather than a retraction of it. Reading it as evidence that condensates are artefacts inverts the paper. Kabra A et al. 2022
Connected to