DOT1L and H3K79 in KMT2A::MLLT1: a direct partner contact, mutually exclusive with the elongation complex, and the only fusion with a DOT1L-inhibitor complete remission

Immunopurification of ENL-associated proteins (EAP) from human cells recovered DOT1L alongside P-TEFb, other frequent KMT2A partners (AF4, AF5q31, LAF4) and polycomb group members RING1, CBX8 and BCoR. The purified material had H3K79-specific methylase activity and robust RNAPII CTD kinase activity, and counteracted the P-TEFb inhibitor DRB. ENL knockdown reduced genome-wide and gene-specific H3K79 dimethylation, global run-on elongation and reporter activity, and structure-function data put DOT1L recruitment as important for transformation by the MLL-ENL derivative 1.

This is not a property KMT2A::AFF1 shares. AF4-family proteins do not directly associate with DOT1L, and the artificial MLL-AF5q31 fusion transformed myeloid progenitors while being unable to bind DOT1L 2. So when this graph says 'KMT2A-rearranged leukaemia is DOT1L-dependent', the route by which DOT1L arrives is partner-specific even though the dependency is shared.

Co-immunoprecipitation from cells expressing ENL, AF5q31 and DOT1L together showed ENL pulling down both partners while AF5q31 and DOT1L did not pull each other down - no trimeric complex. A GAL4-AF5 construct co-precipitated ENL but not DOT1L under conditions where GAL4-ENL pulled DOT1L cleanly 2. The ENL/DOT1L species is therefore a separate biochemical entity from the AF4/ENL/P-TEFb complex, not a subcomplex of it.

The functional test followed: an artificial MLL-DOT1L fusion, which cannot bind AF4 but retains the methyltransferase domain and so mimics the MLL-ENL/DOT1L species, failed to activate Hoxa9 sufficiently to immortalise myeloid progenitors, while MLL-AF5q31 transformed in the same experiment without touching DOT1L 2. The authors state plainly that this contrasts with the earlier account in which DOT1L recruitment was sufficient.

The consequence for this beat is that KMT2A::MLLT1 has two routes to its targets - elongation-complex tethering and direct DOT1L tethering - and the evidence says the first is the load-bearing one while the second shapes the chromatin output.

In HB1119 (t(11;19), KMT2A::MLLT1), DOT1L-dependent H3K79 methylation co-localised with MLL-ENL at every target locus tested. In MV4-11 (KMT2A::AFF1) H3K79me marks also co-localised at most occupied loci, but the H3K79 dimethylation signal intensities were relatively low at MLL-AF4 target loci compared with MLL-ENL target loci when normalised to the same beta-actin and GAPDH internal standards, and were minimal at the CDKN2C promoter despite abundant elongation-complex occupancy there 2.

That is the closest thing in the literature to a direct, same-experiment, partner-stratified measurement of H3K79 load, and it points the same way as the clinical observation below. It is a ChIP-qPCR comparison across two different cell lines rather than an isogenic one, so it is suggestive rather than decisive.

Independently, in the murine MLL-ENL system, H3K79me2 at the Eya1 and Meis1 promoters falls in proportion to transcriptional output when the ENL YEATS domain is deleted or reader-dead 4 - so H3K79 deposition downstream of this fusion is coupled to the partner's acetyl-reading module as well as to its DOT1L contact.

Pinometostat (EPZ-5676) is a first-in-class DOT1L inhibitor. Its adult phase 1 (NCT01684150) enrolled 51 patients across six dose-escalation and two expansion cohorts at 54 and 90 mg/m2/day by continuous IV infusion. Two patients achieved complete remission, both with t(11;19), both at 54 mg/m2/day. Central NGS identified the partner in every t(11;19) case, and the two responders differed: one KMT2A::ELL and one KMT2A::MLLT1 3.

The MLLT1 responder is identifiable from the figure: an 81-year-old woman with KMT2A-rearranged CMML who enrolled after abandoning azacitidine, reached CR at 113 days with a duration of response of 85 days, had resolution of leukaemia cutis, and had a FISH cytogenetic remission in which translocation-positive cells fell from 90% to 0.2%, with the MLLT1 signal returning to normal segregation. She developed AML within three weeks of stopping the drug 3.

The authors themselves raise partner-specific sensitivity as a hypothesis for the pattern and immediately qualify it: not all patients with such fusions derived detectable benefit 3.

Very little on its own, and it is worth being explicit about why, because this is the single most quotable partner-specific fact in the DOT1L literature and it is a two-patient observation.

• n=2 responders in 51 patients, in a dose-escalation study with no prospective partner stratification and no comparator.
• The two responders do not share a partner gene - one is ELL - so the common factor is the cytoband, not the protein. t(11;19)(q23;p13.3) is MLLT1 and t(11;19)(q23;p13.1) is ELL, and this graph already holds an entry on how often that distinction is lost.
• The MLLT1 responder had CMML, an adult myelodysplastic/myeloproliferative disease, at age 81. The modal KMT2A::MLLT1 patient is an infant with B-ALL. Nothing about this case transfers to that population without argument.
• The response was not durable: DOR 85 days, and transformation to AML within three weeks of stopping.

The honest reading is that DOT1L inhibition can produce a real, cytogenetically documented remission in a KMT2A::MLLT1 patient, that this is more than any other partner group achieved in that trial, and that nobody has since tested it prospectively in t(11;19).

The pinometostat paper's discussion describes the two responding fusions as '1 DOT1L-ELL and 1 DOT1L-ENL' 3. Neither fusion exists; the fusions are MLL-ELL and MLL-ENL, and DOT1L is the drug target rather than a fusion partner. The rest of the paper is consistent with MLL-ELL and MLL-ENL, including the FISH figure showing MLL and MLLT1 signals. Anyone searching for 'DOT1L-ENL' as a fusion will find nothing, and anyone quoting that sentence verbatim will propagate a non-existent fusion name.