ENL's disordered region and condensates, restricted to what actually bears on KMT2A::MLLT1 - which is one figure panel

ENL's disordered regions and AFF4's serine-rich IDR (AFF4 residues 760-901) both phase separate as eGFP fusions. Droplets form without crowding agent, are still visible at 200 nM protein, shrink and disappear as NaCl rises, fuse on contact into single spherical droplets, are completely inhibited by 3% 1,6-hexanediol, and are disrupted at 65 C - so both hydrophobic and electrostatic interactions contribute 1.

In cells, ENL knockout reduces heterotypic AFF4/cyclin T1 droplets, rescuable by ENL re-expression 1. So in this system ENL is the initiating component and AFF4 is the modulator.

eGFP-ENL with CDK9-mRFP forms heterotypic droplets that recover slowly by FRAP - the authors call this a static, solid-like state. Adding FLAG-AFF4 speeds recovery. Co-transfecting AFF4, but not cyclin T1, increases the number of ENL/CDK9 droplets while reducing their size, which the authors read as AFF4 preventing excessive coalescence. Knocking AFF4 down by shRNA, or inhibiting it with the small molecule KL-1, produces irregular shape and incomplete coalescence 1.

Read conservatively, this says the ENL-P-TEFb assembly is not intrinsically liquid and that a second component is needed to keep it dynamic. That is a mechanistically specific claim and it is the part of this paper least dependent on the contested assays, because it is a within-system comparison rather than an assertion that any particular structure is a condensate.

Cells overexpressing wild-type ENL or MLL-ENL were co-stained for elongation components. The tested components did not fully overlap, consistent with their existing outside the elongation complex. Transfection of MLL-ENL significantly increased the number of AFF4/ENL, AFF4/CDK9 and AFF4/ELL2 colocalised puncta per nucleus, each with n > 20 nuclei 1. The authors write that this 'may provide a biophysical mechanism for MLL-ENL fusion in leukemogenesis'.

That is the whole fusion-context condensate literature: one panel and its quantification, in HeLa, by transient transfection.

Three limits, in order of severity.

1. The assays do not discriminate. 1,6-hexanediol sensitivity and classical FRAP recovery both fail to separate liquid-liquid phase separation from binding to spatially clustered sites 2; only model-free calibrated half-bleaching does. Worse in a chromatin context, 1,6-hexanediol rapidly immobilises and condenses chromatin in living human cells, an action distinct from melting a droplet 3.
2. Overexpression inflates the phenotype. At physiological expression, low-complexity-domain interactions that control transcription form dynamic hubs with no detectable phase separation 4, and the ENL field's own mutant work found that expression beyond endogenous level produces non-functional condensates 5.
3. The system is wrong. HeLa has no haematopoietic context, no KMT2A rearrangement and none of the chromatin state the fusion acts in.

The fix is inexpensive and specific: a knock-in fluorescent tag at the fusion junction in HB1119 or KOPN-8, imaged at endogenous level, with calibrated half-bleaching rather than FRAP. That would simultaneously produce the first junction-specific reagent for this fusion and the first condensate measurement that could be believed.

Both control arms make the same forward fusion protein, so any condensate property of KMT2A::MLLT1 itself is identical between them by construction. The condensate axis can only distinguish the arms through the reciprocal product, MLLT1::KMT2A, which would carry MLLT1 exon 1 but neither the ENL IDR bulk nor the AHD.

So the testable condensate prediction for this beat is narrow and falsifiable: if reciprocal-positive patients differ, it is not because their forward fusion behaves differently in a droplet. Anyone proposing a condensate explanation for a reciprocal-status effect has to explain how a protein lacking the IDR changes the behaviour of one that has it.

And a bounded negative that constrains the whole area: (ENL OR MLLT1) AND condensate AND phospho* returns 0 PubMed records, re-run 2026-09-20 with the query translation inspected 6. There is no phospho-regulation literature for these condensates to build a switch model on.