KMT2A::MLLT6 (MLL-AF17), t(11;17)(q23;q12-q21) — the bridge case: a leucine zipper on a DOT1L-complex subunit

Prasad and colleagues cloned the ALL-1 partner at 17q21 and reported that AF17 encodes a 1093-residue protein containing a leucine-zipper dimerisation motif located 3' of the fusion point and an N-terminal cysteine-rich domain that can be arranged as three zinc fingers with homology to Br140 (peregrin); AF17 also contains stretches previously associated with transcriptional repression or activation 2.

Two things about that paper are worth preserving because the field has quietly dropped both.

First, "3' of the fusion point" is the whole argument. A dimerisation motif retained in the chimera, in a partner whose chromatin-reading module is lost, is precisely the forced-oligomerisation architecture — stated in 1994, nine years before the EPS15/GAS7 coiled-coil experiments that the graph treats as founding it. But Prasad et al. never tested it functionally. There is no transformation assay, no deletion series and no self-association measurement for the AF17 leucine zipper in any source read for this entry.

Second, the model the authors actually proposed was the opposite of an oncogenic gain of function: they argued that ALL-1 rearrangements result in loss of function, with the partner polypeptide playing an accessory role by repressing the truncated ALL-1 protein or blocking the normal protein still present in the leukaemic cell 2. That model did not survive, but it is worth recording that the paper the field cites for "the AF17 leucine zipper dimerises MLL" argued for something else entirely.

The graph sorts KMT2A partners into elongation/chromatin partners (recruit SEC, DOT1L, H3K79 methylation) and non-elongation partners (supply an oligomerisation surface and nothing else). MLLT6 is in both sets, and not by a technicality.

It is a DotCom subunit. Affinity purification and MudPIT identified a multisubunit Dot1 complex containing ENL, AF9/MLLT3, AF17/MLLT6 and AF10/MLLT10 together with TRRAP, SKP1 and beta-catenin; the human complex trimethylates H3K79, and Dot1 knockdown in Drosophila reduces expression of a subset of Wingless targets, as does loss of the Drosophila homologues of the Dot1-associated proteins 3. Three of the four MLL partners inside DotCom — ENL, AF9, AF10 — are the canonical elongation/H3K79 partners of this disease. MLLT6 is the fourth.

It also carries the dimerisation motif. Retained in the fusion, 3' of the breakpoint 2.

So the natural experiment is available and has not been done: delete the leucine zipper from an MLL-AF17 construct and ask whether transformation survives. If it does, MLLT6 works through DotCom and the oligomerisation reading is wrong for this partner. If it does not, then a DotCom subunit still requires an oligomerisation module, which would argue that forced oligomerisation is the deeper mechanism and elongation recruitment is one way of achieving it — the position hyp-elongation-hijack-is-a-special-case-of-forced-oligomerisation takes. MLLT6 is the single fusion on which that hypothesis could be tested directly, and that is the reason this entry exists.

One caution against over-reading DotCom membership: the DotCom composition was determined for the wild-type proteins, not for any MLL fusion, and nothing establishes that KMT2A::MLLT6 enters DotCom or brings DOT1L to chromatin. That inference is exactly the kind this graph is supposed to refuse to make silently.

From Table 1 of the 2023 recombinome 1:

27 patients, rank 11, 1.6% of the AML cohort. The adult-AML weighting (12 of 27) is the strongest of any partner in this beat and is unusual for a KMT2A fusion; the infant cases are all lymphoid.

The cytoband is not settled in the literature. Prasad et al. cloned the partner at 17q21 2; later case reports describe t(11;17)(q23;q12-q21) or (q23;q21) 4; the recombinome's historical partner table records MLLT6 at 17q12 with t(11;17)(q23.3;q12), citing Prasad et al. (1994) with disease annotation AML, ALL, T-ALL 1. This entry records the discrepancy rather than choosing, because a reader searching cytogenetic reports needs both.

The region is also crowded. Molecular dissection of t(11;17) in AML revealed a variety of gene fusions with heterogeneous fusion transcripts and multiple splice variants 4, and KMT2A::LASP1 arises from the same t(11;17)(q23;q21) 5. A karyotypic t(11;17)(q23;q21) is therefore not a KMT2A::MLLT6 diagnosis, which matters because it is the third distinct impersonation route this project has documented for a KMT2A partner.

Cell type. Patient-level only: adult AML predominantly, infant ALL in a minority, and one reported case of acute megakaryoblastic leukaemia with t(11;17)(q23;q21) and KMT2A::MLLT6 6. No experimental cell-of-origin work exists for this fusion.

Cell state. Not characterised. No differentiation-block phenotype, no serial replating, no mouse model was found.

Subcellular localization. Not reported for the fusion. Wild-type AF17 is a nuclear chromatin-complex subunit by virtue of DotCom membership 3; unlike EPS15, GAS7, GPHN and the septins, MLLT6 is not a cytoplasmic protein, so the nuclear-relocalisation question that defines the cytoplasmic-partner class does not arise here.

Subnuclear localization. Not reported for the fusion.

Epigenetics. The only epigenetic readout attaching to MLLT6 is indirect and is about the wild-type protein: DotCom, of which MLLT6 is a subunit, trimethylates H3K79 3. No H3K79 measurement has been made on KMT2A::MLLT6 itself, nor any H3K4 methylation, HOXA ChIP or transcriptome. Given that this is the partner where an H3K79 link is most plausible, the absence is the most conspicuous in this beat.