Experiment: does AF1q self-associate, and does KMT2A::MLLT11 transform? The two measurements nobody has made on the 10th most common KMT2A fusion

Two facts sit badly together. KMT2A::MLLT11 is the 10th most common direct KMT2A fusion partner, with 28 patients in the largest systematic series — 13 infant AML, 10 paediatric AML, 4 adult AML and one infant ALL 1. And the fusion protein has never been expressed in a cell by anybody 1.

That is not a niche gap. Every mechanistic claim this graph makes about KMT2A leukaemia is a claim about what the partner contributes, and the models were built on partners chosen because they had a module to point at. MLLT11 has none: 90 residues, one nuclear export signal, no domain, no structure, and a protein so poorly behaved that recombinant expression goes entirely into inclusion bodies 23.

The field's two accounts make opposite predictions, which is what makes the experiment worth doing rather than merely overdue:

• If forced oligomerisation is the mechanism, AF1q must self-associate. A 90-residue protein with no annotated module could still do so — short peptides oligomerise — and the measurement settles it in a day.
• If elongation hijack is the mechanism, MLLT11 has nothing to hijack with, and the fusion should either not transform or transform by something else.
• If it transforms and AF1q is a monomer, both models have a genuine counterexample at 28 patients, which is the outcome the dispatch for this beat anticipated and which nobody has tested.

Design. Sedimentation-velocity analytical ultracentrifugation on purified, refolded AF1q at three concentrations spanning at least a tenfold range, to distinguish a monomer from a concentration-dependent self-association. Mass photometry and SEC-MALS are acceptable orthogonal or substitute readouts and are cheaper; at least two independent methods should agree before a monomer is reported, because a weak self-association is exactly what a single method at one concentration misses.

Why it is nearly free. The reagent exists. AF1q was expressed in E. coli as an N-terminal His6 fusion, recovered from inclusion bodies under denaturing conditions, refolded by one-step dialysis and purified to homogeneity by gel filtration at 35 mg/L of culture — and the same study already ran NMR and AUC on it to map the interaction with the TCF7 HMG domain (residues 154-237) 3. The published abstract simply does not report AF1q's own sedimentation behaviour.

Controls and the trap. The protein is refolded from denaturant, so a negative result (monomer) is weaker than a positive one: misfolded protein may fail to self-associate for reasons unrelated to the native state. The TCF7-binding activity should be used as a folding control — protein that binds the TCF7 HMG domain with the reported behaviour is at least partly native. Note also that this is a fusion-relevant question asked on the free partner; the fusion context could create or destroy self-association, which experiment B partly addresses and a pulldown of the expressed fusion would address directly.

Readout and interpretation. Monomeric at all concentrations, with folding controls passing → the forced-oligomerisation model has a counterexample at 28 patients, contingent on B showing transformation. Self-associating → the model absorbs MLLT11 and the partner landscape becomes more unified, not less.

Design. Retroviral transduction of murine bone-marrow haematopoietic progenitors with KMT2A::MLLT11 (the common architecture: KMT2A N-terminus fused in frame to the complete 90-residue MLLT11 ORF 6), followed by serial replating in methylcellulose over 3-4 rounds. This is the standard assay in which MLL-GAS7 and MLL-GPHN were both scored 45.

Controls, which matter more than usual here.

• Positive: MLL-AF9 or MLL-ENL, an elongation-partner fusion with an established replating phenotype.
• Second positive: a synthetic-dimeriser MLL construct, which is the reference for what pure forced oligomerisation produces 4.
• Negative: truncated MLL with no partner, to exclude a phenotype from the KMT2A moiety alone.
• Structure-function arm: a construct in which MLLT11's nuclear export signal (aa 24-32) is mutated 2. The complete ORF is retained in patients, so the fusion carries an NES — a configuration not known for any other KMT2A fusion — and whether it is functional changes where the oncoprotein is.

Readouts. Colony number and morphology per round; immunophenotype (c-Kit, Mac-1, Gr-1, B220) to place the arrest; Hoxa9 and Meis1 expression, which every oligomerisation-driven MLL fusion tested so far upregulates 5; and, if replating is positive, subcellular localisation of the fusion by immunofluorescence — the single most informative cheap add-on, since no KMT2A::MLLT11 localisation exists at all.

Interpretation, both ways. Positive replating in a fusion whose partner is monomeric would be the first real counterexample to both models. Negative replating is equally publishable and arguably more interesting: it would mean a fusion found in 28 patients is not sufficient for transformation on its own, forcing a cooperating-lesion account — for which there is already a hint, since t(1;11) occurs in three-way and complex rearrangements including one with MYO18A as a reciprocal partner 6.

The honest limitation. A negative result in a mouse replating assay does not establish that the fusion does not transform. Infant leukaemia arises in a fetal-liver-derived compartment that adult murine bone marrow does not model, and this project's own material on cell of origin makes that point repeatedly. A negative should trigger a fetal-liver or human-CD34 progenitor arm, not a conclusion.