Spliced KMT2A fusions join to partner exon 2 — with the per-partner counts, and a correction
also called read-through KMT2A fusions · partner exon 2 rule · promoter-capture KMT2A fusions
In some cases the break lands in front of the partner gene rather than inside it, and the cell then splices KMT2A's last surviving exon straight onto the partner's second exon. This entry lists how often that happens for each partner, and corrects an existing entry in this graph that attributed the phenomenon to the wrong gene.
The same thing, for a scientist
Where the 5' KMT2A fragment lands upstream of an intact partner gene, transcription from the KMT2A promoter reads through and the most 3' KMT2A exon splices to partner **exon 2**, the first exon carrying a bona fide splice acceptor. Meyer and colleagues identified 275 such cases among 3,401 patients. The dominant partner is **MLLT1**: 241 of 460 MLLT1 fusions (~52%) break upstream of MLLT1 exon 1, and those cases express only the direct KMT2A::MLLT1 transcript with no reciprocal product. Per-partner counts are EPS15 12/51, ELL 10/119, PRPF19 3/3, AFDN 2/138, MLLT3 2/612, AFF1 1/1262, plus DCPS, MYO1F, SEPTIN5 and CT45A2 at 1 case each.
A correction to an existing entry in this graph
The existing node breakpoint-kmt2a-spliced-fusion (in the atlas/ folder, written by another beat) states that spliced fusions are "overwhelmingly a KMT2A::MLLT10 phenomenon: 241 of 460 KMT2A::MLLT10 cases (~50%) break upstream of MLLT10 exon 1".
The source says MLLT1, not MLLT10 1. Quoting the Results section verbatim: "Spliced fusions are rare events except for MLLT1 gene fusions (n = 460), where about 50% of the breakpoints (n = 241) localize upstream of MLLT1 exon 1. In these cases, only KMT2A::MLLT1 fusion transcripts but no reciprocal one are expressed."
The arithmetic corroborates the correction independently of the gene label. Meyer's cohort contains 404 KMT2A::MLLT1 ALL cases plus 2 infant, 28 pediatric and 16 adult AML cases — consistent with a partner total near 460. MLLT10 totals 80 in ALL and roughly 200 across AML strata, nowhere near 460 2.
This matters beyond bookkeeping. The spliced-fusion class is the natural experiment for asking whether the reciprocal fusion protein is needed, because these patients are constitutively reciprocal-free. If you believe that experiment is available in half of all KMT2A::MLLT10 cases, you will design it in the wrong disease: MLLT10 is predominantly myeloid, MLLT1 predominantly lymphoid. We have not edited the atlas/ node, which belongs to another beat. This entry records the correction and the evidence for it.
The mechanism and the exon-2 rule
The 5' KMT2A fragment is recombined upstream of an intact partner gene rather than into it. There is no transcription terminator between the two, so transcripts initiating at the KMT2A promoter read straight through into the partner. The spliceosome then joins the most 3' KMT2A exon to partner exon 2 — because exon 2 is the first partner exon presenting a genuine splice acceptor; partner exon 1 has none available, being normally the transcription start 1.
So the partner-side junction coordinate for this whole class is known exactly, and it is the same for every partner: exon 2. That is the one partner-side exon assignment this beat can state with confidence, and it is why this entry exists as a breakpoint node rather than a footnote.
The diagnostic consequence is awkward. A spliced fusion produces a direct fusion transcript that looks ordinary by RT-PCR or RNA-seq. Distinguishing it requires genomic breakpoint mapping. So a series that classified junctions from RNA alone will have silently mixed spliced and conventional cases together — including, potentially, in any analysis correlating junction type with outcome.
Per-partner counts, and what the denominators reveal
From the same cohort 1:
| Partner | Spliced cases / total for that partner | Rate |
|---|---|---|
| MLLT1 | 241 / 460 | ~52% |
| EPS15 | 12 / 51 | 24% |
| ELL | 10 / 119 | 8.4% |
| PRPF19 | 3 / 3 | 100% |
| AFDN | 2 / 138 | 1.4% |
| MLLT3 | 2 / 612 | 0.3% |
| AFF1 | 1 / 1262 | 0.08% |
| DCPS, MYO1F, SEPTIN5, CT45A2 | 1 each | — |
| Total | 275 / 3401 | 8.1% |
The spread is enormous and demands an explanation nobody has given. AFF1 is 1 in 1,262. MLLT1 is 1 in 1.9. These are the two commonest lymphoid partners, both recruiting the same super elongation complex, and they differ by a factor of about 650 in how often the break lands in front of the gene rather than inside it.
Two readings, both testable. Either the genomic architecture upstream of MLLT1 is unusually fragile or unusually often juxtaposed to KMT2A in nuclear space — a 3D-proximity or fragility explanation — or the product is selected: a reciprocal-free KMT2A::MLLT1 is viable and a reciprocal-free KMT2A::AFF1 is not. The second reading would directly contradict the argument that reciprocal fusion proteins are dispensable, and it cuts against the observation reported in the same paper that t(4;11) patients expressing both direct and reciprocal alleles have better outcome 3.
Note also the denominators themselves: 1,262 AFF1, 612 MLLT3, 460 MLLT1, 138 AFDN, 119 ELL, 51 EPS15. These are the partner totals across the whole cohort and they are a cleaner frequency source than the age-stratified figures, because they do not depend on age annotation being complete.
| gene5 | KMT2A |
|---|---|
| cytoband | 11q23.3 |
| genome build | hg38 |
| mechanism class | promoter capture / transcriptional read-through with splicing to partner exon 2 |
| gene3 exon first rule | partner exon 2, invariant across partners — the first partner exon carrying a bona fide splice acceptor |
| n spliced cases total | 275 |
| n cohort | 3401 |
| spliced share of cohort PCT | 8.1 |
| population | 3,401 genomically characterised KMT2A-rearranged acute leukemia patients, DCAL Frankfurt, 2003-2022 |
| reciprocal product | absent in spliced cases |
| diagnostic note | Indistinguishable from a conventional in-frame fusion by RT-PCR or RNA-seq of the direct transcript alone; requires genomic breakpoint mapping by LDI-PCR, LD multiplex PCR or targeted NGS. |
per partner — the numbers behind the chart
[
{
"partner": "MLLT1",
"spliced": 241,
"partner_total": 460,
"pct": 52.4,
"note": "dominant; reciprocal product absent in these cases"
},
{
"partner": "EPS15",
"spliced": 12,
"partner_total": 51,
"pct": 23.5
},
{
"partner": "ELL",
"spliced": 10,
"partner_total": 119,
"pct": 8.4
},
{
"partner": "PRPF19",
"spliced": 3,
"partner_total": 3,
"pct": 100,
"note": "every case; n too small to interpret"
},
{
"partner": "AFDN",
"spliced": 2,
"partner_total": 138,
"pct": 1.4
},
{
"partner": "MLLT3",
"spliced": 2,
"partner_total": 612,
"pct": 0.3
},
{
"partner": "AFF1",
"spliced": 1,
"partner_total": 1262,
"pct": 0.08
},
{
"partner": "DCPS",
"spliced": 1,
"partner_total": 1,
"pct": 100
},
{
"partner": "MYO1F",
"spliced": 1,
"partner_total": 5,
"pct": 20
},
{
"partner": "SEPTIN5",
"spliced": 1,
"partner_total": 1,
"pct": 100
},
{
"partner": "CT45A2",
"spliced": 1,
"partner_total": 1,
"pct": 100
}
]correction recorded
{
"target_node": "breakpoint-kmt2a-spliced-fusion",
"target_folder": "data/graph/nodes/atlas/",
"error": "attributes the 241/460 spliced-fusion finding to KMT2A::MLLT10; the source says MLLT1",
"evidence": "Meyer et al. 2023 Results, PMC10169636, verbatim: 'Spliced fusions are rare events except for MLLT1 gene fusions (n = 460), where about 50% of the breakpoints (n = 241) localize upstream of MLLT1 exon 1.'",
"corroboration": "MLLT1 cohort totals (404 ALL + ~46 AML) are consistent with ~460; MLLT10 totals (80 ALL) are not.",
"action_taken": "not edited — the node belongs to another beat; correction recorded here and reported to the King"
}- If reciprocal expression really does improve chemotherapy response by broadening the druggable transcriptome, then a therapy that suppresses the reciprocal allele — or a clone that loses it under pressure — would produce a narrower, more resistant transcriptome. That is a resistance route created by the treatment.
- Spliced fusions are invisible to RNA-based diagnostics, so a stratification scheme that assigns risk by junction type cannot be implemented in centres that do not do genomic breakpoint mapping.
- Why is the spliced-fusion rate 52% for MLLT1 and 0.08% for AFF1, two partners that recruit the same complex in the same lineage? A ~650-fold difference is either a genomic-architecture fact about the MLLT1 locus or a selection fact about reciprocal-free products, and distinguishing them would say whether reciprocal fusion proteins are required in general or only for some partners.
- Reciprocal-free KMT2A::MLLT1 leukemias are common, fully leukemogenic and constitutively available. Has anyone compared their transcriptome and drug sensitivity against reciprocal-expressing KMT2A::MLLT1? That is a ready-made, zero-cost natural experiment on the reciprocal-product question sitting in existing biobanks.
- How many series that classified KMT2A junctions from RNA alone have silently mixed spliced and conventional cases? If spliced fusions differ in outcome, that misclassification would attenuate any junction-type effect towards the null.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Spliced KMT2A fusions arise when the 5' KMT2A fragment lands upstream of an intact partner gene; transcription reads through from the KMT2A promoter and the most 3' KMT2A exon splices to partner exon 2, the first exon with a bona fide splice acceptor. 275 such cases were identified among 3,401 patients: MLLT1 241/460, EPS15 12/51, ELL 10/119, PRPF19 3/3, AFDN 2/138, MLLT3 2/612, AFF1 1/1262, and DCPS, MYO1F, SEPTIN5 and CT45A2 at one case each. In the MLLT1 spliced cases only the direct fusion transcript is expressed, with no reciprocal product.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, Burmeister T, Gröger D, Sutton R, Venn NC, Cazzaniga G, et al., Marschalek R · Leukemia 37:988-1005 2023What this evidence cannot show
Retrieved as PMC full text (PMC10169636) on 2026-09-18 and the Results passage read verbatim, because a machine-readable version of the same text had the italicised gene symbols stripped — which is precisely how the MLLT1/MLLT10 confusion this entry corrects can arise. Referral-based single-network cohort. The per-partner denominators are cohort totals and include out-of-frame and complex rearrangements, so the rates are approximate. Spliced fusions cannot be identified from RNA data alone, so their frequency in series that used RNA-based diagnosis is unknown and probably underestimated.
Evidence that argues against this (2)
- The per-partner spliced rates measure how often each partner's cases were genomically mapped, not only how often the mechanism occurs. An independent group found the same mechanism in a partner that appears nowhere in the frequent list - a cryptic insertion of a 653 kb 11q23 fragment carrying KMT2A exons 1-9 upstream of CT45A2, splicing in frame to CT45A2 exon 2 - and found it only because cytogenetics, FISH, RT-PCR and LDI-PCR were all run on one patient. Partners more often referred for full genomic characterisation will show higher spliced rates, so 241 of 460 is a rate among MLLT1 cases that were mapped. Cerveira N et al. 2010
- Reciprocal-free is a larger category than spliced, and the bigger group is not made by promoter capture. In t(4;11) acute lymphoblastic leukaemia the direct MLL-AF4 transcript is expressed in every patient while the reciprocal AF4-MLL transcript is expressed in only about half, and those patients carry a structurally intact der(4). Treating the 275 spliced cases as the graph's access to reciprocal-free disease therefore selects the smaller and more unusual of two populations, and any conclusion drawn from them should be checked against the reciprocal-silent t(4;11) half. Bueno C et al. 2019
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In the same cohort, KMT2A partner counts by lineage are: ALL (n=2182) AFF1 1233 (56.5%), MLLT1 404 (18.5%), MLLT3 258 (11.8%), MLLT10 80 (3.7%), AFDN 46 (2.1%), EPS15 36 (1.6%); the seven most frequent partners plus KMT2A-PTD account for more than 90% of the 3,401 patients.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al., Marschalek R · Leukemia 37:988-1005 2023What this evidence cannot show
We use these counts here only to corroborate, by arithmetic, that a partner total near 460 is consistent with MLLT1 and not with MLLT10. They are referral-based and not population frequencies.
Evidence that argues against this (1)
- The membership of the dominant-partner set is not stable across editions of this same series, so 'the seven most frequent partners plus KMT2A-PTD account for more than 90%' is an edition-specific statement. At n=1590 in 2013 the seven were AFF1/AF4, MLLT3/AF9, MLLT1/ENL, MLLT10/AF10, ELL, KMT2A-PTD and MLLT4/AF6, covering about 90%. EPS15, which appears in the lineage table quoted here, was not in that set, and ELL and MLLT4/AF6 were. The 90% figure is robust; which seven genes deliver it is not. Meyer C et al. 2013
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
Two recent publications report better outcome in t(4;11)/pro-B ALL patients when both the direct and the reciprocal fusion alleles are expressed, and reciprocal fusion proteins AFF1::KMT2A and AFDN::KMT2A act as chromatin-opening complexes that let the direct fusion activate roughly ten-fold more target genes.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al., Marschalek R · Leukemia 37:988-1005 2023What this evidence cannot show
This is a summary in a Discussion section citing other primary papers whose identity we could not resolve from the stripped reference list, so we cite the review rather than the primary work and mark confidence as moderate. The 'roughly ten-fold' figure is given as ~10-fold in the Discussion and as ~12-fold in the companion 2024 consensus Letter by the same group — an internal inconsistency we record rather than average. Better outcome with reciprocal expression is an association in retrospective series, not a demonstrated causal effect, and the proposed explanation (a broader transcriptome exposes more druggable targets to chemotherapy) is speculative.
Evidence that argues against this (1)
- If reciprocal expression were required, reciprocal-free spliced KMT2A::MLLT1 leukemias should not exist — yet they are about half of all MLLT1 cases and are fully leukemogenic. The two observations are only compatible if the reciprocal requirement is partner-specific. Meyer C et al. 2023
corrects the partner: the 241/460 spliced-fusion finding is MLLT1, not MLLT10
- related to →KMT2A::MLLT1 (MLL-ENL), t(11;19)(q23;p13.3)
about 52% of KMT2A::MLLT1 cases are spliced fusions with no reciprocal product
reciprocal-free KMT2A::MLLT1 leukemias are a ready-made natural experiment on whether the reciprocal product is required
- related to →The reciprocal fusion (AFF1::KMT2A and friends): a second oncoprotein that opens chromatin for the first
the 650-fold MLLT1-vs-AFF1 spliced-fusion rate difference bears directly on whether the reciprocal product is partner-specifically required
- part of →KMT2A / MLL1 — the gene that breaks
partner-side junction rule for the KMT2A recombinome: the most 3' KMT2A exon splices to partner exon 2