Where KMT2A::MLLT1 actually sits: chromatin occupancy is mapped, subnuclear localisation of the endogenous fusion has never been imaged

Chromatin immunoprecipitation in HB1119 places native MLL-ENL, together with AF4 and AF5q31, at the promoter-adjacent regions of HOXA9 and MEIS1, and also at CDKN1B and CDKN2C and within the transcribed bodies of HOXA9 and MEIS1. Occupancy at ACTB and GAPDH was minimal or negligible 1.

In the murine retroviral system, FLAG-ChIP against the fusion's N-terminal tag gives occupancy at the Eya1 promoter regions P1/P3 and the intragenic P7, and at the Meis1 regulatory regions P1/P2. Deleting the ENL YEATS domain reduces that occupancy; the reader-dead point mutants F59A and Y78A do not, retaining occupancy while losing H3K9ac, H3K79me2 and H3K4me3 output 2. PAF1 occupancy tracks the fusion, which the authors read as cooperative assembly 2.

So at the level of 'which genes', this fusion is one of the better-characterised KMT2A chimeras.

Three imaging observations touch this fusion, and none of them is of the endogenous protein.

1. Nuclear dot pattern. ENL and hPc3 (CBX8) tagged with fluorescent proteins co-localised in living cells in a nuclear dot pattern, in the paper that first showed the ENL moiety of MLL-ENL recruits hPc3 3. The recruitment was mapped to the ENL protein-protein interaction domain - the same C-terminal region the fusion retains - but the imaging is of ENL, not of the fusion.
2. Nuclear speckled pattern. ENL and AF4 co-localised in vivo in a nuclear speckled pattern, in the paper that mapped the ENL C-terminus as the AF4/AF5q31/AF10 binding domain and noted it coincides with the region essential for MLL-ENL transformation 4. Again, wild-type proteins.
3. Increased puncta on fusion expression. Transfecting MLL-ENL into HeLa significantly increased the number of nuclear puncta containing AFF4 with ENL, AFF4 with CDK9, and AFF4 with ELL2, relative to transfecting wild-type ENL; the tested components did not fully overlap 5. This is the only fusion-context image in the literature, and it is a transient overexpression in an epithelial line.

Nobody has imaged KMT2A::MLLT1 in a t(11;19) leukaemia cell. The graph's standing constraint explains why: every reagent in this field is raised against retained sequence and cannot distinguish fusion from wild-type parent, and no junction-specific antibody exists for any KMT2A fusion.

First, 'more puncta' is not 'more condensate'. Counting colocalised foci per nucleus in a transfected cell reports on clustering, and clustering has at least two rival explanations - phase separation, and binding to spatially clustered genomic sites. The assays used in that paper (1,6-hexanediol sensitivity, classical FRAP) are exactly the two that cannot tell those apart 6, and 1,6-hexanediol itself rapidly immobilises and condenses chromatin in living human cells, which is close to the opposite of the action the assay assumes. This graph's standing condensate-evidence entry sets the bar.

Second, an overexpression phenotype is weak evidence about the endogenous protein. At physiological expression, low-complexity-domain interactions that control transcription form dynamic hubs without detectable phase separation, and even within the ENL literature, expression of ENL mutants beyond the endogenous level produces non-functional condensates 7. The MLL-ENL puncta experiment is transient transfection into HeLa.

Both cautions cut in the direction of 'not established', not in the direction of 'artefact'. Clustered-site binding is a substantive rival model and for a fusion whose 5' partner carries CXXC and AT-hooks, arguably the more interesting one.

If the reciprocal product MLLT1::KMT2A exists in 101 of 254 YEATS-retaining patients, the first question anyone would ask about it is where it goes. It would carry MLLT1 exon 1 and the 3' half of KMT2A - so no AHD, no YEATS, but a SET domain and PHD fingers - and there is no published image, no ChIP and no antibody for it.

That means the localisation axis of the 153-versus-101 comparison is currently unmeasurable by imaging and can only be approached through the consequences: chromatin occupancy of the forward fusion, transcriptional output, and outcome. A junction-specific reagent or a knock-in tag on the reciprocal allele in an isogenic pair would open it.

No phospho-condensate literature exists for this protein. (ENL OR MLLT1) AND condensate AND phospho* returns 0 records, and the query translation is clean - PubMed expanded condensate to its twelve word forms and ENL/MLLT1 to All Fields with no [UID] term injected, so this is a real zero and not a parser artefact 8. That confirms an earlier beat's finding rather than restating it.

The only primary fusion-context condensate paper is unreachable from the obvious query. (MLL-ENL OR KMT2A-MLLT1 OR KMT2A::MLLT1) AND (condensate OR phase separation OR droplet) returns 1 record, and it is a review 8. The primary paper - which contains the MLL-ENL puncta experiment in its Figure 3 - does not appear, because its abstract phrases the result as 'the fusion of ENL with MLL' and the token MLL-ENL occurs only in the body. Anyone who ran the sensible query and stopped would conclude that no fusion-context condensate data exist at all, and would be wrong by exactly one paper: the only one there is.

This is the same family as the complex-name trap this project already documents. For a fusion, search the partner alone with the mechanism term, not only the hyphenated fusion name.