Two classes of KMT2A rearrangement keep the SET domain, and the graph says otherwise
also called not every KMT2A fusion loses SET · 3-prime KMT2A fusions · SET-retaining rearrangements
It is repeated everywhere, including in this graph, that every KMT2A fusion deletes the enzyme end of the protein. Two kinds of rearrangement do not: the mirror-image product of a balanced swap, and the rare inversions where KMT2A is the second partner rather than the first.
The same thing, for a scientist
The statement 'the SET domain is lost in every KMT2A fusion' is true only of 5'-KMT2A direct fusion products. Two classes falsify it as a universal. (1) Reciprocal der products of balanced translocations - AFF1::KMT2A, and by the same logic MLLT3::KMT2A, MLLT1::KMT2A and the rest - fuse the partner N-terminus to KMT2A exons 12 onwards and therefore retain the WIN motif and SET domain [E1][E4]. (2) 3'-KMT2A rearrangements, of which the molecularly confirmed human example is NUP98::KMT2A from inv(11)(p15q23), fuse NUP98 FG repeats to almost the entire KMT2A open reading frame, deleting only exon 1 [E2][E3]. That fusion is the clean natural experiment: it keeps WRAD and loses menin, and it is menin-inhibitor resistant with REDUCED rather than elevated HOXA output [E3]. The universal claim should be restated as a claim about the direct product, and the exceptions are informative rather than pedantic, because they dissociate the two halves of KMT2A that the direct fusion always separates.
The two classes
Reciprocal der products. A balanced translocation makes two fusion genes. The one nobody assays keeps the KMT2A C-terminus. Reciprocal X-MLL proteins are described as exhibiting a PHD H3K4me3-reader domain, sequestering CREBBP and MOF, and carrying the SET methyltransferase domain at their C-terminus 4. Expression is not universal - roughly half of t(4;11) cases and, by the recombinome's count, about half of KMT2A::MLLT1 cases break upstream of the partner's first exon and make no reciprocal transcript at all - but where it exists, the WRAD interface exists with it.
3'-KMT2A rearrangements. NUP98::KMT2A arises from inv(11)(p15q23) - an inversion, both genes being on chromosome 11 - and fuses NUP98 FG repeats to nearly the whole KMT2A ORF while deleting exon 1, which encodes the menin- and LEDGF-binding region 23. Everything downstream is retained, including the WIN motif and SET domain. In an inducible transgenic mouse the resulting leukaemia expresses very low Hoxa5/a7/a9/a10 and Meis1 and is resistant to the menin inhibitor MI-2-2 3.
Why these two cases are the control arms the WRAD question needs
The direct der(11) fusion confounds two variables: it gains a partner and it loses the C-terminus. The two exceptions break the confound in opposite directions.
- NUP98::KMT2A gains a partner and keeps the C-terminus. If WRAD assembly on a KMT2A fusion has any oncogenic consequence, this is where it should show, and the phenotype is the least HOXA-driven of the family.
- AFF1::KMT2A keeps the C-terminus and gains an elongation-factor partner, in a cell that simultaneously contains the direct fusion.
Together they say something the direct-fusion literature cannot: retaining WRAD is not sufficient to produce the canonical KMT2A-rearranged HOXA program. In the one case where a KMT2A fusion keeps the whole methyltransferase module and loses menin, HOXA output goes DOWN. That is a substantive argument for menin primacy, and it is a better one than the 2004 co-IP because it is a functional measurement in the right direction.
What should be corrected, and where
The universal phrasing appears in this beat's own dispatch brief and in several entries. The correct phrasing is 'lost from the direct 5'-KMT2A fusion product'. The entries this affects are listed in ana-wrad-fusion-evidence-enumeration with the exact sentences. Notably this graph already holds both exceptions - mechanism-reciprocal-kmt2a-fusion-chromatin-opening records that reciprocal products carry the SET domain, and fusion-nup98-kmt2a records the retained WIN motif and SET domain - so the correction is a linking failure rather than a knowledge gap.
| corrected phrasing | the SET domain and its WRAD contacts are lost from the direct 5'-KMT2A fusion product |
|---|
set retaining classes
[
{
"class": "reciprocal der product of a balanced translocation",
"example": "AFF1::KMT2A (AF4-MLL)",
"menin_motif": "lost",
"win_motif": "retained",
"set_domain": "retained",
"expressed_in": "about 45% of t(4;11) infant B-ALL"
},
{
"class": "3-prime KMT2A rearrangement",
"example": "NUP98::KMT2A, inv(11)(p15q23)",
"menin_motif": "lost (exon 1 deleted)",
"win_motif": "retained",
"set_domain": "retained",
"expressed_in": "2 molecularly confirmed adult AML patients"
}
]existing graph entries holding the exceptions
[ "mechanism-reciprocal-kmt2a-fusion-chromatin-opening", "fusion-nup98-kmt2a", "subtraction-kmt2a-set-domain" ]
- Any therapeutic claim built on 'the fusion has no methyltransferase activity' is wrong for the reciprocal-positive half of t(4;11) and for NUP98::KMT2A, and the eligibility criteria currently in use do not distinguish those patients.
- NUP98::KMT2A is already documented as menin-inhibitor resistant, so this class is a known blind spot of the entire approved modality rather than a hypothetical one.
- A SET-domain or WIN-site inhibitor aimed at the exceptions would also engage the wild-type KMT2A allele and all five other SET1-family enzymes, so the therapeutic index would have to come from somewhere other than target selectivity.
- Do reciprocal products of other KMT2A translocations - MLLT3::KMT2A, MLLT1::KMT2A, MLLT10::KMT2A - exist at the protein level, and at what frequency? The recombinome answers the transcript question partner by partner; the protein question is unanswered for every partner except AFF1.
- Is the NUP98::KMT2A low-HOXA phenotype caused by retained WRAD or merely by lost menin? A NUP98::KMT2A construct with the SET domain deleted would separate them and has not been made.
- Does any 3'-KMT2A rearrangement other than NUP98::KMT2A exist in the recombinome, and was it counted as a KMT2A fusion or discarded as an artefact?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The reciprocal t(4;11) product AF4-MLL comprises AF4 exons 1-3 fused to MLL exons 12-37 and therefore retains the KMT2A C-terminus through the final exon.
The role of reciprocal fusions in MLL-r acute leukemia: studying the chromosomal translocation t(4;11). · Wilhelm A, Marschalek R · Oncogene 2021What this evidence cannot show
Describes the constructs used in that study, designed to represent the infant-typical intron 11 breakpoint. Patient-level exon composition varies with breakpoint, and breaks 3' of KMT2A exon 12 would shorten the retained fragment - though not below the SET domain, which is encoded near the 3' end.
Evidence that argues against this (1)
- About half of KMT2A::MLLT1 cases break upstream of MLLT1 exon 1 and express no reciprocal transcript at all, so reciprocal expression is partner- and breakpoint-dependent and cannot be assumed for any given patient. Meyer C et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
NUP98-MLL is a fusion of NUP98 to MLL described in human acute myeloblastic leukaemia in which MLL fusion proteins associated with leukaemia normally contain the menin interaction peptide.
NUP98-MLL fusion in human acute myeloblastic leukemia. · Kaltenbach S, Soler G, Barin C, et al. · Blood 2010What this evidence cannot show
Two reported patients. The abstract was read; the full text was not retrieved for this entry, so exon-level breakpoint detail here is taken from the second citation and from this graph's existing fusion-nup98-kmt2a entry rather than from this record.
Evidence that argues against this (1)
- inv(11)(p15q23) has been reported cytogenetically in other malignancies without fusion confirmation, and the fusion is absent from the 160-patient paediatric NUP98 landscape cohort, so the two-patient basis is the whole human evidence. Fisher JN et al. 2020
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Inducible iNUP98-KMT2A transgenic mice develop disease after a median latency of 80 weeks with anaemia and reduced marrow cellularity, establishing the transforming activity of a KMT2A fusion that retains the SET domain and lacks the menin-binding exon 1.
Transforming activities of the NUP98-KMT2A fusion gene associated with myelodysplasia and acute myeloid leukemia. · Fisher JN, Thanasopoulou A, Juge S, et al. · Haematologica 2020What this evidence cannot show
A mouse transgenic model with very long latency, which itself argues the fusion is a weak oncogene. The menin-inhibitor resistance and the low-HOXA phenotype are recorded in this graph's fusion-nup98-kmt2a entry from this source; only the abstract was read for the present entry.
Evidence that argues against this (1)
- An 80-week median latency in a transgenic model is long enough that cooperating lesions are likely required, so the phenotype attributed to the fusion may be the phenotype of the fusion plus whatever it selected for over 80 weeks. Kaltenbach S et al. 2010
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
Reciprocal X-MLL fusion proteins exhibit a PHD H3K4me3-reader domain, sequester the histone acetyltransferases CREBBP and MOF, and carry the SET histone methyltransferase domain at their C-terminus.
MLL leukemia and future treatment strategies. · Marschalek R · Arch Pharm (Weinheim) 2015What this evidence cannot show
A review by the group that has argued longest for the importance of reciprocal fusions, so the framing is partisan. It is used here for the domain composition, which is a structural consequence of the breakpoint and not in dispute, not for the functional claims.
Evidence that argues against this (1)
- The SET domain the reciprocal product retains is genetically dispensable for MLL-AF9-driven leukemogenesis in mouse, so retaining it does not by itself imply a leukaemogenic contribution. Mishra BP et al. 2014
- ← supportsAbout one KMT2A-rearranged patient in twenty has no direct fusion at all - only the reciprocal one
Adds a third and larger class of menin-blind KMT2A oncoprotein: patients who have no direct fusion at all.
A third SET-retaining, menin-blind KMT2A species beyond the in-frame reciprocal product and NUP98::KMT2A.
NUP98::KMT2A is the second SET-retaining class: it deletes KMT2A exon 1 and the menin-binding motif while keeping the WIN motif and SET domain, and it is menin-inhibitor resistant with reduced HOXA output.
Correction to 'deleted from every leukemia fusion': the SET domain is deleted from the direct 5'-KMT2A product and retained in reciprocal der products and in NUP98::KMT2A.
The reciprocal der product is the commoner of the two SET-retaining classes.
The lost-module ledger is a ledger of the direct fusion; the two SET-retaining classes need their own rows or an explicit scope statement.
- ← depends onThe graph's universal 'every KMT2A fusion loses the SET domain' is a silent, propagating error class
The error class is the universal quantifier that the two SET-retaining fusion classes falsify.