MAML2 exon 2: the strongly preferred 3' acceptor, the same in leukaemia, thymoma and salivary-gland tumours
also called MAML2 exon 2 acceptor · MAML2 intron 1 breakpoint · shared MAML2 C-terminus · MAML2 preferred acceptor exon
When MAML2 is joined to another gene in a tumour, the join is almost always at the same place - the start of its second exon - no matter which gene it is joined to or which organ the tumour is in. One published patient is the exception, so this is a strong preference rather than a rule, and that kind of near-sameness usually means the piece that survives is the piece that matters.
The same thing, for a scientist
Across every report that gives a junction, the MAML2 side of a KMT2A fusion is exon 2, while the KMT2A donor varies across exons 7, 8, 9, 10 and 11 [E1][E2][E3]. The same exon 2 acceptor defines CRTC1(MECT1)::MAML2 in mucoepidermoid carcinoma, Warthin tumour and clear-cell hidradenoma, and the founding KMT2A report notes explicitly that MLL-MAML2 and MECT1-MAML2 contain **the same C-terminal part of MAML2** [E1]. The paralogous KMT2A::MAML3 thymoma fusion also joins at MAML3 exon 2, and the authors compared the two exon 2 fusion sites and found they share similar functional protein domains [E4]. So one ~952-residue MAML-family fragment - transactivation domain retained, Notch-binding MamL-1 domain lost - is the invariant unit, recurring under at least two unrelated 5' partners, in at least four tissues, in leukaemia and in epithelial neoplasia. **The variation is overwhelmingly on the 5' side.** One published exception exists - a single B-ALL patient in whom one of three detected products joined KMT2A exon 9 to MAML2 intron 1-2 rather than to exon 2 - so the acceptor is strongly preferred rather than strictly invariant, and this entry was retitled from 'invariant' to 'preferred' on that evidence. The asymmetry is still the finding, and it is the strongest available argument that the MAML2 fragment is a selected functional module rather than incidental C-terminal bulk.
The observation, laid out1 deeper
| 5' partner | tumour | 5' donor exon(s) | MAML acceptor | source |
|---|---|---|---|---|
| KMT2A | secondary AML, MDS | exon 7 | MAML2 exon 2 | 1 |
| KMT2A | B-ALL (one patient, three products) | exon 7, exon 8, exon 9 | MAML2 exon 2; one to intron 1-2 | 3 |
| KMT2A | type B2/B3 thymoma, n = 11 | exon 8, 9, 10 or 11 | MAML2 exon 2 | 2 |
| KMT2A | type B3 thymoma, n = 1 | not stated | MAML3 exon 2 | 4 |
| CRTC1 (MECT1) | mucoepidermoid carcinoma, Warthin tumour, clear-cell hidradenoma | - | MAML2, same C-terminal fragment | 1 |
Five 5' donor exons on the KMT2A side against essentially one acceptor exon on the MAML2 side, plus a completely different 5' partner arriving at the same MAML2 fragment. The one exception is in the table: the B-ALL patient's third product joined KMT2A exon 9 to MAML2 intron 1-2. It occurred alongside two ordinary exon 2 products in the same patient, so it is an alternative product rather than a different fusion, but it is enough to make 'strongly preferred' the correct word and 'invariant' the wrong one 3.
Why exon 2 and not exon 3 or exon 5
Two explanations, and they are not exclusive.
Functional selection. MAML2's MamL-1 domain occupies residues 33-87, encoded in the 5'-most coding sequence. Any junction 3' of it removes Notch-ICD binding while sparing the transactivation domain at 609-681. Exon 2 is the first junction point that achieves that cleanly, so selection for a targeting-free effector would land there 5.
Genomic structure. MAML2's intron 1 is large in many MAML-family genes, and a large intron is a large target: the probability of a break falling in it is proportional to its length. The observed invariance could be partly a consequence of intron 1 dominating the gene's breakable sequence.
These make different predictions and nobody has separated them. If it is pure target size, breaks should be distributed across intron 1 roughly uniformly and occasional out-of-frame or exon-3 products should appear in sequencing data. If it is selection, junctions should cluster at positions that preserve frame with exon 2's acceptor and nothing else should survive. The Felix study's TOP2A cleave-ome mapping in the MAML2 genomic region is the sort of data that could resolve it and was not analysed for this question 6.
What the invariance is worth
Three practical consequences.
- Diagnostics. A single RT-PCR or capture probe anchored on MAML2 exon 2 detects every KMT2A::MAML2 case regardless of KMT2A donor exon, and would also catch CRTC1::MAML2. Assays anchored on the 5' side need to cover five exons.
- Therapeutic reuse. Anything that acts on the shared MAML2 C-terminal fragment is, in principle, active in both a rare leukaemia and a common salivary-gland carcinoma. CRTC1::MAML2 mucoepidermoid carcinoma is a far larger indication than KMT2A::MAML2 leukaemia (151 PubMed records for CRTC1-MAML2 against 21 for the whole KMT2A-MAML2 intersection), so the economics of drugging the shared module are better than the leukaemia case alone suggests.
- Mechanistic leverage. The fragment can be studied independently of either 5' partner. Fusing it to an inert DNA-binding domain isolates what the MAML2 module contributes, which is exactly the decomposition this graph wants for every partner and rarely gets.
Relation to KMT2A's own breakpoint atlas
The KMT2A donor exons reported for MAML2 fusions - 7, 8, 9, 10, 11 - all lie within the major BCR (intron 7 to exon 13). So KMT2A::MAML2 is an ordinary major-BCR fusion in every respect on the KMT2A side, and it loses the PHD fingers. It is not a minor-BCR fusion and must not be filed with KMT2A::USP2 on that basis, even though both are intrachromosomal inversions on chromosome 11. The two fusions differ precisely where it matters: USP2 breaks in BCR2 and keeps PHD1-3 and the bromodomain; MAML2 breaks in BCR1 and loses them.
Correction: what was withdrawn and why
This entry was first written under the title "MAML2 exon 2: the invariant 3' acceptor" and with the id breakpoint-maml2-exon-2-preferred-acceptor. That overstated the evidence and has been withdrawn. A single B-ALL patient had three detected products, one of which joined KMT2A exon 9 to MAML2 intron 1-2 rather than to exon 2 3. The title, id, summaries and section prose now say strongly preferred rather than invariant.
What survives the correction: across seven independent reports, two unrelated 5' partners and four tissues, every other described junction uses the MAML2 exon 2 acceptor, and the resulting ~952-residue fragment is the same in all of them. A near-invariant acceptor is still a strong asymmetry against five variable KMT2A donor exons, and the functional argument - that the junction removes MamL-1 and spares the transactivation domain - is unaffected. What does not survive is any claim that exon 2 is the only acceptor observed.
| acceptor exon | 2 |
|---|---|
| acceptor gene | MAML2 (and MAML3 in the one paralogous thymoma case) |
| retained maml2 fragment | approximately residues 205-1156 (952 residues): TAD at 609-681 retained, MamL-1 at 33-87 lost |
| kmt2a side is major bcr | true |
| kmt2a side note | All reported KMT2A donor exons (7-11) lie within the major BCR (intron 7 - exon 13). KMT2A::MAML2 is NOT a minor-BCR fusion and loses the PHD fingers, unlike KMT2A::USP2. |
| acceptor status | strongly preferred, not invariant - one published exception in one patient, who also carried two ordinary exon 2 products |
| known exception | One B-ALL patient had a KMT2A exon 9 to MAML2 intron 1-2 product alongside two exon 2 products. This is why the entry says 'preferred' rather than 'invariant'. |
donor exons observed kmt2a
[ 7, 8, 9, 10, 11 ]
donor partners observed
[ "KMT2A", "CRTC1 (MECT1)" ]
tissues observed
[ "bone marrow (AML, MDS, T-ALL, B-ALL)", "thymus (type B2/B3 thymoma)", "salivary gland (mucoepidermoid carcinoma, Warthin tumour)", "skin (clear-cell hidradenoma)" ]
competing explanations
[ "functional selection for a targeting-free MAML effector module", "genomic target size: MAML2 intron 1 is large and absorbs breaks in proportion to its length", "TOP2A cleavage-site distribution within the MAML2 region, which the Felix study mapped but did not analyse for junction position" ]
pubmed scale comparison — the numbers behind the chart
{
"CRTC1_AND_MAML2": 151,
"MECT1_AND_MAML2": 49,
"MLL_or_KMT2A_AND_MAML2": 21
}- If a drug were developed against the shared MAML2 C-terminal fragment, it would by construction hit CRTC1::MAML2 salivary tumours as well - which is an opportunity commercially and a caution about specificity, since the module derives from a normally expressed human protein.
- The invariance makes a MAML2 exon 2 assay attractive, but an assay anchored only there would call CRTC1::MAML2 and KMT2A::MAML2 alike unless the 5' side is also sequenced. A laboratory optimising for sensitivity here could lose the ability to tell a leukaemia fusion from a salivary-gland one.
- The one exception in the literature - a KMT2A exon 9 to MAML2 intron 1-2 product - is exactly the kind of case an exon-2-anchored assay would miss, and it occurred alongside two ordinary products in the same patient.
- Where in MAML2 intron 1 do the genomic breaks actually fall? Every published junction is transcript-level, so nobody knows whether the genomic breaks cluster or are scattered - and that single fact distinguishes selection from splicing canalisation.
- Is the retained 952-residue MAML2 fragment sufficient to transform when fused to an unrelated DNA-binding domain? That isolates the module from both KMT2A and CRTC1 and is the clean decomposition experiment.
- Why does the junction remove 117 more residues than the minimum needed to delete MamL-1? Either those residues are dispensable, or they are deleterious, or exon structure simply does not offer a closer option.
- Does CRTC1::MAML2 mucoepidermoid carcinoma share any therapeutic vulnerability with KMT2A::MAML2 leukaemia? They share the effector module exactly, and one is a far larger indication than the other.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL exon 7 was fused to MAML2 exon 2 in secondary AML and MDS, and MLL-MAML2 and MECT1-MAML2 consist of the same C-terminal part of MAML2.
Identification of a novel fusion gene MLL-MAML2 in secondary acute myelogenous leukemia and myelodysplastic syndrome with inv(11)(q21q23). · Nemoto N, Suzukawa K, Shimizu S, Shinagawa A, Takei N, Taki T, Hayashi Y, Kojima H, Kawakami Y, Nagasawa T · Genes, Chromosomes & Cancer 2007What this evidence cannot show
Two patients, and the junction was determined by RT-PCR and sequencing of the transcript, so it reports the spliced product rather than the genomic breakpoint - the genomic break could be anywhere in intron 1. The comparison with MECT1-MAML2 is a statement about shared C-terminal content, not a side-by-side sequence alignment presented in the abstract. Abstract read in full.
Evidence that argues against this (1)
- A transcript-level junction cannot distinguish selection at the protein level from RNA splicing that canalises many different genomic breaks in a large intron 1 onto the same exon 2 acceptor. The invariance may therefore be a property of splicing rather than of selection. Menu E et al. 2017
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Across 11 thymomas with KMT2A-MAML2, fusion breakpoints were identified between KMT2A exon 8, 9, 10 or 11 and MAML2 exon 2.
Pan-Cancer Landscape Analysis Reveals Recurrent KMT2A-MAML2 Gene Fusion in Aggressive Histologic Subtypes of Thymoma. · Massoth LR, Hung YP, Dias-Santagata D, Onozato M, Shah N, Severson E, Duncan D, Gillespie BJ, Williams NF, Ross JS, Vergilio JA, Harkins SK, Glomski K, Nardi V, Zukerberg LR, Hasserjian RP, Louissaint A, Williams EA · JCO Precision Oncology 2020What this evidence cannot show
Detection was by hybrid-capture DNA sequencing with RNA profiling on only a portion of samples, so the junction calls are not uniformly derived. Eleven patients. The exon numbering convention used is not stated and may not be the DCAL convention this graph uses elsewhere, which is a real hazard for cross-reading these numbers against the BCR coordinates. Abstract read in full.
Evidence that argues against this (1)
- If the thymoma series uses a different KMT2A exon numbering from the Meyer/Marschalek convention, the statement that exons 8-11 lie inside the major BCR could be off by a small offset. The conclusion survives because the major BCR spans intron 7 to exon 13 in that convention and any plausible offset stays inside it, but the numbering is not verified. Meyer C et al. 2023
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In a single B-ALL patient, RNA sequencing detected junctions of KMT2A exon 7 to MAML2 exon 2 and KMT2A exon 9 to MAML2 intron 1-2, and Sanger sequencing a junction of KMT2A exon 8 to MAML2 exon 2.
First case of B ALL with KMT2A-MAML2 rearrangement: a case report. · Menu E, Beaufils N, Usseglio F, Balducci E, Lafage Pochitaloff M, Costello R, Gabert J · BMC Cancer 2017What this evidence cannot show
One patient with three distinct products detected by two methods; the paper does not determine whether these represent alternative splicing from one genomic break, subclonal heterogeneity, or technical artefact. The exon-9-to-intron-1-2 product is notable because it is the one junction in the whole literature that does not use the exon 2 acceptor. Abstract read in full.
Evidence that argues against this (1)
- One of this patient's three products joins KMT2A exon 9 to MAML2 intron 1-2, not to exon 2, so the exon 2 acceptor is not universal. This observation is what caused the entry's original title, which called the acceptor invariant, to be withdrawn (see the correction section). It also shows a single patient can carry more than one junction, so 'the junction' is not always a single object. Nemoto N et al. 2007
- E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper
A KMT2A::MAML3 thymoma fusion joins at MAML3 exon 2; the authors compared the MAML2 and MAML3 exon 2 fusion sites and found they share similar functional protein domains, with the same domain appearing downstream of the KMT2A::MAML2 fusion protein.
Type B3 Thymoma With a Novel KMT2A::MAML3 Fusion: Expanding the Spectrum of Gene Fusions Beyond the MAML2 Gene. · Zhou Z, Yu Y, Chen C, Lin J · Genes, Chromosomes & Cancer 2025What this evidence cannot show
One patient, detected by RNA sequencing only after DNA-based NGS found nothing. The domain comparison is a bioinformatic alignment, not a functional demonstration, and the paralogue argument means the 'invariance' claim is really about a MAML-family exon 2 acceptor rather than MAML2 specifically. Abstract read in full.
Evidence that argues against this (1)
- Extending an invariance claim from MAML2 to the MAML family on one case risks manufacturing a pattern. MAML1 has one record in intersection with KMT2A and MAML3 three; neither is enough to say the family behaves alike. Massoth LR et al. 2020
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
MAML2's MamL-1 Notch-binding domain occupies residues 33-87 and its transcriptional activation domain 609-681, so an exon 2 acceptor is the 5'-most junction that removes Notch-ICD binding while sparing the activation domain.
UniProtKB entry Q8IZL2 (MAML2_HUMAN) and InterPro Pfam domain boundaries for MAML2 · UniProt Consortium; InterPro (EMBL-EBI) · UniProtKB / InterPro 2026What this evidence cannot show
Domain boundaries are model matches. The claim that exon 2 is the 5'-most junction achieving this depends on the exon structure of MAML2, which was not independently verified against a genome annotation in this session - it follows from the founding report's statement that the retained fragment is 952 residues, i.e. beginning around residue 205, which is well past residue 87.
Evidence that argues against this (1)
- The retained fragment begins near residue 205, not near residue 88, so exon 2 is not tightly optimal for sparing the activation domain while removing MamL-1 - it removes 117 residues more than necessary. A functional-selection explanation therefore has to account for why the extra 117 residues are also dispensable or deleterious, which nobody has examined. Nemoto N et al. 2007
- E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Etoposide-induced TOP2A cleavage is enriched in the relevant MAML2 genomic region, supporting a TOP2A DNA-damage mechanism for the formation of this fusion.
KMT2A-MAML2 rearrangement emerged and regressed during neuroblastoma therapy without leukemia after 12.8-year follow-up. · Felix CA, Slater DJ, Davenport JW, Yu X, Gregory BD, Li MM, Rappaport EF, Cheung NV · Pediatric Blood & Cancer 2021What this evidence cannot show
The cleave-ome mapping is done in a model system rather than in the patient's own cells, and 'the relevant MAML2 genomic region' is not resolved to intron 1 versus elsewhere in the abstract. The study's primary finding concerns one patient's clone, and the TOP2A data are supporting rather than central. Abstract read in full.
Evidence that argues against this (1)
- A TOP2A-cleavage explanation for where the break falls is a fragility argument, and it competes with rather than supports the functional-selection reading of the exon 2 invariance. If TOP2A cleavage sites in MAML2 intron 1 dictate the break position, the invariance is chemistry, not selection. Nemoto N et al. 2007
exon 2 acceptor across every reported fusion; retained fragment approximately residues 205-1156
KMT2A donor varies across exons 7-11; the MAML2 acceptor does not vary
every KMT2A donor exon reported for MAML2 fusions lies inside the major BCR, so this is a BCR1 fusion that loses the PHD fingers
- related to →Each partner prefers a different KMT2A exon: partner-specific break position in 3,401 patients
a partner-side exon preference so strong it is nearly invariant, and it holds across two unrelated 5' partners and four tissues
the exon 2 acceptor is what removes MamL-1 while sparing the transactivation domain, i.e. it is the junction that creates the dominant-negative composition