Spliced KMT2A fusions: breaks upstream of the partner's exon 1, and the reciprocal-free natural experiment
also called spliced fusion · read-through KMT2A fusion · promoter-capture fusion · KMT2A::MLLT1 spliced type
Sometimes the break does not land inside the partner gene at all - it lands in front of it. The KMT2A promoter then simply reads straight through into the intact partner gene, and the cell splices the two together. These cases make only one hybrid protein instead of two.
The same thing, for a scientist
In a subset of KMT2A rearrangements the 5' KMT2A fragment is placed upstream of an intact partner gene rather than inside it. Because there is no transcription terminator between them, transcription initiated at the KMT2A promoter reads through into the partner gene, and the most 3' KMT2A exon splices to partner exon 2 - the first exon carrying a bona fide splice acceptor [E1]. Meyer et al. identified 275 such cases among 3401 patients. They are overwhelmingly a KMT2A::MLLT1 phenomenon: 241 of 460 KMT2A::MLLT1 cases (about 50%) break upstream of MLLT1 exon 1, and in these cases only the direct KMT2A::MLLT1 transcript is expressed, with no reciprocal product [E1][E_corr]. Smaller numbers occur with other partners: EPS15 12/51, ELL 10/119, PRPF19 3/3, AFDN 2/138, MLLT3 2/612 and AFF1 1/1262 [E_corr]. The reciprocal-free half of KMT2A::MLLT1 is often described as a natural experiment on whether the reciprocal fusion protein matters, but the useful version of that comparison is WITHIN MLLT1: across partners it is confounded almost perfectly with MLLT1 versus AFF1, and it is not the only reciprocal-free population, because the reciprocal AF4-MLL transcript is absent in about half of t(4;11) patients who carry an intact reciprocal allele.
Why KMT2A::MLLT1 is the spliced-fusion partner1 deeper
About half of all KMT2A::MLLT1 rearrangements - 241 of 460 - place the 5' KMT2A fragment upstream of an intact MLLT1 gene rather than inside it, and the recombinome reports no other partner anywhere near that rate [E_corr]. The mechanism is generic: there is no transcription terminator downstream of a translocated 5'-KMT2A fragment, so transcripts starting at the KMT2A promoter read into whatever gene lies downstream 1. What is partner-specific is how often the rearrangement happens to land in front of the partner gene rather than within it, and the primary text does not explain why MLLT1 is the outlier.
The consequence for the atlas is important: KMT2A::MLLT1 is not one fusion but two molecularly distinct entities - a reciprocal-expressing type and a reciprocal-free spliced type - that are indistinguishable by FISH or by RT-PCR for the direct transcript. Any analysis of KMT2A::MLLT1 that does not separate them averages over two different molecular configurations.
Correction, 2026-09-19. An earlier version of this section asserted that KMT2A::MLLT1 rearrangements are 'unusually often complex rather than simple reciprocal translocations - inversions, insertions and three-way events' and that 'the intron phase of the MLLT10 breakpoint region means KMT2A breakpoints in MLLT10 cases shift even upstream of intron 9'. Neither is supported by the cited paper. The upstream-of-intron-9 statement belongs to SEPTIN9, verbatim: 'breakpoints are shifting to regions even upstream of intron 9 because of the intron phase of the BCR of this partner gene.' The named inversion and insertion examples in that paper are PRPF19, MAML2, USP2 and SEPTIN6, not MLLT1 or MLLT10. Both sentences have been withdrawn.
Why this is the test the field needs
The Marschalek group's central claim is that the reciprocal fusion protein opens chromatin and expands the direct fusion's target-gene repertoire roughly tenfold, establishing the pre-leukaemic program 2. If true, reciprocal-free spliced KMT2A::MLLT1 should differ from reciprocal-expressing KMT2A::MLLT1 in target-gene breadth, HOXA signature, age at diagnosis or outcome. The DCAL database already contains roughly 241 versus 219 patients in the two arms with clinical annotation, so that comparison is a database query rather than a new study, and to our knowledge it has not been published.
Two bounds on how clean the experiment is. First, the comparison has to be run WITHIN MLLT1; across partners it is a comparison of MLLT1 against AFF1 (241/460 versus 1/1262) and cannot separate the missing reciprocal from the partner. Second, spliced fusions are not the only reciprocal-free population and are not even the largest: in t(4;11) the direct MLL-AF4 transcript is always expressed but the reciprocal AF4-MLL transcript is expressed in only about half of patients, who nonetheless carry an intact der(4) [E_corr]. Whatever makes a reciprocal allele silent in those patients is a bigger and less selected natural experiment than promoter capture is.
Mechanics of the spliced junction
- A DNA break in the KMT2A BCR frees a 5' KMT2A fragment carrying the KMT2A promoter, AT-hooks, SNL/MBM region and CXXC.
- The rearrangement places this fragment 5' of an intact partner gene, in the same orientation, with no intervening polyadenylation/termination signal.
- RNA polymerase II initiating at the KMT2A promoter transcribes through the junction into the partner gene body.
- The spliceosome joins the last KMT2A exon to the first partner exon possessing a genuine splice acceptor, which is normally exon 2 (exon 1 has no upstream acceptor) 1.
- Because the partner gene's own 5' end and the KMT2A 3' end are not joined into a second transcription unit, no reciprocal fusion mRNA exists.
Practical implication for diagnostics: the mRNA junction of a spliced fusion can look identical to that of a conventional in-frame fusion, so distinguishing them requires genomic breakpoint mapping (LDI-PCR or targeted NGS), not RNA-seq alone 1.
Correction: the partner is MLLT1, not MLLT10
An earlier version of this entry attributed the 241-of-460 spliced-fusion figure to MLLT10. It belongs to MLLT1. Verified against the primary text: "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." The full list, with denominators:
| Partner | spliced / total | share |
|---|---|---|
| MLLT1 | 241 / 460 | 52% |
| EPS15 | 12 / 51 | 24% |
| ELL | 10 / 119 | 8% |
| PRPF19 | 3 / 3 | 100% |
| AFDN | 2 / 138 | 1.4% |
| MLLT3 | 2 / 612 | 0.3% |
| AFF1 | 1 / 1,262 | 0.08% |
275 cases in total use this mechanism.
The two gene symbols differ by one character and denote different proteins, which is how the error survived. Caught by a later agent cross-checking the primary text rather than the entry. Second pass, 2026-09-19: that correction was applied to the evidence array only, and the summary, the section prose, the data payload, the tags and the open questions were all left saying MLLT10. They have now been corrected too.
| gene | KMT2A |
|---|---|
| cytoband | 11q23.3 |
| genome build | hg38 |
| mechanism class | promoter capture / transcriptional read-through with splicing to partner exon 2 |
| n spliced cases total | 275 |
| population | 3401 KMT2A-rearranged acute leukaemia patients (DCAL 2003-2022) |
| dominant partner | MLLT1 |
| gene3 exon first rule | partner exon 2 (first exon with a bona fide splice acceptor) |
| reciprocal product | none expressed |
| diagnostic note | Cannot be distinguished from a conventional in-frame fusion by RNA-seq or RT-PCR of the direct transcript alone; requires genomic breakpoint mapping (LDI-PCR or targeted NGS) |
| other partners with spliced cases | reported for 9 further partners at 1-12 cases each out of their respective totals |
| reciprocal free note | Spliced fusions are reciprocal-free by construction, but they are not the only reciprocal-free group: the reciprocal AF4-MLL transcript is expressed in only about half of t(4;11) patients (PMID 30679325). |
mllt1 spliced — the numbers behind the chart
{
"spliced": 241,
"total_mllt1": 460,
"pct": 52
}per partner spliced
{
"MLLT1": "241/460",
"EPS15": "12/51",
"ELL": "10/119",
"PRPF19": "3/3",
"AFDN": "2/138",
"MLLT3": "2/612",
"AFF1": "1/1262"
}- Spliced fusions are invisible to RNA-based diagnostics as a distinct class, so any biomarker or therapy-response signal specific to them will be diluted into the general KMT2A::MLLT1 group and missed.
- If reciprocal-free disease has a narrower transcriptome and, per the recombinome authors' own reasoning, therefore fewer druggable targets and greater chemoresistance, spliced KMT2A::MLLT1 could be an intrinsically harder-to-treat subgroup hiding in plain sight.
- This entry frames reciprocal-free spliced fusions as the natural experiment the field needs, but the cross-partner comparison it implies is confounded almost perfectly with MLLT1 versus AFF1. The within-MLLT1 comparison is the one worth arguing for.
- The reciprocal-free argument is also weaker than it looks because about half of t(4;11) patients already express no reciprocal transcript. If reciprocal-free disease behaved distinctly, that difference should already be visible in every t(4;11) cohort ever collected, and no one has reported it.
- Do spliced (reciprocal-free) KMT2A::MLLT1 patients differ from reciprocal-expressing KMT2A::MLLT1 patients in age, lineage, HOXA signature, MRD kinetics or survival? The DCAL database can answer this today with roughly 241 versus 219 patients, and holding the partner constant is what makes it interpretable.
- Because a spliced fusion keeps the partner gene's coding sequence intact from exon 2 onward, is the resulting protein longer and more nearly wild-type on the 3' side than a conventional intragenic fusion? If so, spliced fusions are the best available model of the pure KMT2A-N-terminus-plus-full-partner oncoprotein.
- Why is the reciprocal AF4-MLL transcript silent in about half of t(4;11) patients who carry an intact der(4)? That is a larger reciprocal-free population than the spliced fusions, it is not explained by promoter capture, and nobody has reported what distinguishes the expressing from the non-expressing half.
- Does promoter capture also happen in the opposite direction - a partner promoter reading into 3'-KMT2A - and would that generate a menin-independent KMT2A species analogous to KMT2A*?
- How many KMT2A-r cases classified by RNA-seq in modern diagnostic pipelines are actually spliced fusions that were never recognised as such, and does that matter for MRD assay design?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
275 of 3401 KMT2A-rearranged patients carry spliced fusions in which 5'-KMT2A lies upstream of an intact partner gene and the last KMT2A exon splices to partner exon 2; 241 of 460 KMT2A::MLLT1 cases break upstream of MLLT1 exon 1 and express only the direct fusion with no reciprocal product.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
The absence of a reciprocal transcript in spliced cases is inferred from genomic architecture plus a cited prior study, not measured transcript-by-transcript in this cohort. Single-centre referral cohort. The per-partner spliced-case denominators reported in the text are partly ambiguous in the published prose because partner names are italicised and can be mis-parsed; the MLLT10 figure (241/460) is the one stated unambiguously.
Evidence that argues against this (2)
- The natural experiment this entry proposes is confounded by partner identity, and the confound is visible in these same numbers. Reciprocal-free spliced fusions are not spread across partners: 241 of 460 are KMT2A::MLLT1, against 1 of 1,262 for KMT2A::AFF1. So a comparison of reciprocal-free against reciprocal-bearing cases is very nearly a comparison of MLLT1 against AFF1, and MLLT1 brings its own biology - a YEATS-domain reader engaging SEC and DOT1L directly. Any difference in target-gene breadth could be attributed to the missing reciprocal or to the partner, and this design cannot separate them. Leukemia 2023
- The design is rescuable, and the entry should say how rather than treat the comparison as available. Because KMT2A::MLLT1 splits roughly 50/50 - 241 of 460 spliced, the rest not - a WITHIN-MLLT1 comparison of spliced against reciprocal-bearing cases holds the partner constant and isolates the reciprocal. That is the analysis the falsification argument needs. The cross-partner version, which is what 'reciprocal-free natural experiment' invites a reader to run, is uninterpretable. Leukemia 2023
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Reciprocal KMT2A fusion proteins are proposed to act as chromatin-opening complexes that greatly expand the direct fusion's target-gene set, which is the hypothesis that reciprocal-free spliced fusions can falsify.
Influence of IGF2BP3 on t(4;11) fusion mRNA and protein abundance · Hanewald T, Benz T, Marschalek R · Leukemia Research 2026What this evidence cannot show
Engineered cell-line models from a single laboratory; concerns t(4;11)/AFF1 rather than MLLT1, so applying it to spliced KMT2A::MLLT1 fusions is an extrapolation.
Evidence that argues against this (1)
- Absence of a reciprocal TRANSCRIPT is weaker than absence of the proposed chromatin-opening activity, so a null result would not falsify the hypothesis cleanly. The reciprocal proposal is about a protein complex, and fusion protein abundance in this disease is actively regulated rather than read off transcript presence - IGF2BP1-3 influence t(4;11) fusion mRNA and protein abundance independently. A spliced case that makes no reciprocal transcript is good evidence about transcripts and indirect evidence about the protein activity actually at issue. Leukemia Research 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
KMT2A::MLLT1 is a major AML partner (209/1116 AML cases, 18.7%) and a major T-ALL partner (40/123), so the reciprocal-free spliced subtype is not a rare curiosity but affects a clinically important group.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Referral cohort frequencies, not population incidence.
Evidence that argues against this (1)
- Clinical importance does not transfer across the split. MLLT1's prominence (209/1116 AML, 40/123 T-ALL) is a statement about KMT2A::MLLT1 overall, while the reciprocal-free subtype is about half of it, and the entry's own figures do not establish that the spliced and non-spliced halves are clinically alike. Treating the subtype as inheriting the partner's full clinical weight assumes exactly what the natural experiment sets out to test. Leukemia 2023
- E_corrstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Spliced fusions are rare except for KMT2A::MLLT1, where 241 of 460 cases (about 50%) break upstream of MLLT1 exon 1 and express no reciprocal transcript; for KMT2A::AFF1 the figure is 1 of 1,262 cases.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Referral-based cohort from a single diagnostic centre, so denominators reflect referral patterns. Detection of a spliced fusion depends on the assay reaching upstream of partner exon 1.
Evidence that argues against this (2)
- Reciprocal-free is not the same as spliced, and the larger reciprocal-free population is a different one. In t(4;11) acute lymphoblastic leukaemia the direct MLL-AF4 fusion is expressed in every patient, but the reciprocal AF4-MLL fusion is expressed in only about HALF of them - and those patients carry a structurally intact der(4), so their reciprocal allele is silent for reasons that have nothing to do with promoter capture. A design that treats the 241 spliced KMT2A::MLLT1 cases as the field's only access to reciprocal-free disease therefore picks the smaller, more selected of two available populations. Bueno C et al. 2019
- The per-partner spliced-fusion rates measure the detection pipeline as much as the biology. Spliced fusions are only visible when genomic mapping reaches upstream of partner exon 1, and an independent group found one - a cryptic insertion of a 653 kb 11q23 fragment carrying KMT2A exons 1-9 into the region upstream of CT45A2, splicing to CT45A2 exon 2 - only because FISH, RT-PCR and LDI-PCR were all run on the same case. Partners that are more often referred for full genomic characterisation will therefore show higher spliced rates, and a denominator of 460 MLLT1 cases is a denominator of cases that were mapped, not of cases that occurred. Cerveira N et al. 2010
- related to →KMT2A::MLLT1 (MLL-ENL), t(11;19)(q23;p13.3)
Correction applied 2026-09-19: the 241-of-460 reciprocal-free spliced population is KMT2A::MLLT1, not KMT2A::MLLT10. The summary, sections, data payload, tags, open questions and bite-back of this entry all still said MLLT10 after an earlier evidence-array-only correction.
Counter-evidence bounding the reciprocal-free natural experiment: spliced fusions are not the largest reciprocal-free population. The reciprocal AF4-MLL transcript is expressed in only about half of t(4;11) patients, who carry a structurally intact der(4) (Bueno 2019, PMID 30679325).
- refutes →AFF1 as the subtraction control: the refutation is WITHDRAWN - the degron is lost, not retained
CORRECTION FOR THE OWNER of subtraction-aff1-control: its open question states 'About half of KMT2A::AFF1 cases are spliced fusions with the genomic break upstream of AFF1 exon 1'. The recombinome gives AFF1 as 1 of 1,262 (0.08%); the partner with about half is MLLT1 (241/460). Meyer C et al., Leukemia 2023, PMID 37019990, DOI 10.1038/s41375-023-01877-1, Results, read verbatim from PMC10169636 on 2026-09-19.
CORRECTION FOR THE OWNER of fusion-kmt2a-mllt3: it states 'about half of MLLT3 fusions in the DCAL series are spliced fusions'. The recombinome gives MLLT3 as 2 of 612 (0.3%). Meyer C et al., Leukemia 2023, PMID 37019990, DOI 10.1038/s41375-023-01877-1, read verbatim from PMC10169636 on 2026-09-19.
- ← refutesSpliced KMT2A fusions join to partner exon 2 — with the per-partner counts, and a correction
corrects the partner: the 241/460 spliced-fusion finding is MLLT1, not MLLT10
- ← depends onThe reciprocal fusion (AFF1::KMT2A and friends): a second oncoprotein that opens chromatin for the first
only exists when the break leaves a reciprocal allele
partner-side break determines whether a reciprocal allele exists, which does predict outcome
no reciprocal allele exists in the spliced architecture, so no reciprocal target can be built