MAML2 (mastermind-like 2) - the Notch co-activator that is a KMT2A partner in leukaemia and in thymoma

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• If the fusion's MAML2 moiety is a Notch dominant-negative, then a gamma-secretase inhibitor - the obvious Notch-directed drug - would push in the same direction as the oncoprotein rather than against it. Notch-pathway drugs are the intuitive choice for a Notch co-activator fusion and may be exactly wrong here.
• MAML1 and MAML3 redundancy means that any attempt to target MAML function pharmacologically has three paralogues to contend with, and the flat DepMap score for MAML2 alone is consistent with that buffering rather than with irrelevance.
• The MAML2 C-terminus is shared with CRTC1::MAML2 in salivary-gland tumours, so anything that drugs it is not leukaemia-specific - which is an opportunity for reuse and a warning about off-tumour effects in secretory epithelium.

1. Where is KMT2A::MAML2 in the nucleus? Wild-type MAML2 is a speckle protein and the fusion has lost the domain that docks MAML onto DNA-bound Notch complexes, while gaining CXXC-mediated CpG-island targeting. Those two point to different compartments and a single imaging experiment would decide it.
2. Does the fusion suppress Notch targets in a cell that actually has Notch signalling - a T-cell progenitor - rather than in a reporter assay? The disease distribution includes secondary T-ALL, where Notch is usually activated, so the dominant-negative result and the clinical picture currently point opposite ways.
3. MAML1 and MAML3 are paralogues and MAML3 has now been found fused to KMT2A in thymoma. Is the leukaemia literature undercounting MAML-family fusions by searching only for MAML2?
4. Is the retained MAML2 fragment sufficient on its own, fused to anything, to transform? CRTC1::MAML2 and KMT2A::MAML2 share it exactly, which makes the fragment testable independently of either 5' partner.
5. Co-stain MAML2 with SC35/SRSF2 and with PML. One experiment decides whether MAML2's punctate nuclear pattern is a speckle, a PML body, or neither - and whether this graph is right to place it beside the AFF-family speckle proteins.
6. Does KMT2A-MAML2 go where MAML2 goes? The fusion loses the basic N-terminal MamL-1 domain (33-87), which is the region a basic-patch-driven nuclear pattern would depend on, so the wild-type compartment may be the wrong prior for the chimera.
7. MAML2 is widely expressed by bulk tissue and has never been measured in a haematopoietic cell. Is it present at all in the pro-B or myeloid progenitor in which KMT2A-MAML2 arises?

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   MAML2 is UniProt Q8IZL2, 1156 aa, with the MamL-1 domain at residues 33-87 (Pfam PF09596) and a transcriptional activation domain at 609-681 (Pfam PF20804), and carries the UniProt subcellular location term 'Nucleus speckle'. SEE E_q89_maml2_loc1 AND E_q89_maml2_loc2: that term over-specifies its own source, which reports a speckled nuclear pattern in GFP-tagged overexpressed MAML2 in COS7 and U2OS cells and calls it nuclear BODY localisation, with no speckle marker co-stained.

   UniProtKB entry Q8IZL2 (MAML2_HUMAN), Mastermind-like protein 2, with InterPro/Pfam domain boundaries · UniProt Consortium; InterPro (EMBL-EBI) · UniProtKB / InterPro 2026
   ▸What this evidence cannot show

   Database annotation for wild-type MAML2, not for any fusion protein. The Pfam boundaries are model matches rather than experimentally determined limits, and much of the protein between the two domains is annotated only as disordered. The nucleus-speckle assignment is for the full-length protein; the fusion has never been imaged. Accession resolved by fetching the record and reading back the PRIMARY gene name (MAML2, synonym KIAA1819) and length (1156 aa) rather than trusting a gene_exact synonym match.

   ▸Evidence that argues against this (1)

   • A subnuclear localisation for the wild-type protein says little about a fusion that has lost the N-terminal 204 residues including the domain that docks it onto a DNA-bound complex. Applying the speckle annotation to KMT2A::MAML2 would be an unsupported transfer, and no imaging of the fusion exists to check it. Nemoto N et al. 2007
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In MLL-MAML2, MLL exon 7 is fused to MAML2 exon 2, giving a putative protein of 1408 MLL N-terminal residues plus 952 MAML2 C-terminal residues; the N-terminal basic domain of MAML2 containing the NOTCH intracellular-domain binding site is deleted; MLL-MAML2 and MECT1-MAML2 contain the same C-terminal part of MAML2; and MLL-MAML2 suppressed HES1 promoter activation by the NOTCH1 intracellular domain in a luciferase assay.

   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 2007
   PMID 17551948DOI
   ▸What this evidence cannot show

   Two patients (one secondary AML, one MDS), both adults. The luciferase experiment is a reporter assay in a transfected cell line, not a measurement of endogenous Notch target genes in a leukaemia cell, and a single target promoter (HES1) was used. The residue counts are for the predicted protein from the chimeric transcript; no protein was detected. Abstract read in full; figures not re-derived.

   ▸Evidence that argues against this (1)

   • A dominant-negative effect on Notch sits awkwardly with the disease distribution: KMT2A-MAML2 causes secondary T-ALL in several reported children, and T-ALL is the lineage in which Notch signalling is most commonly ACTIVATED, not suppressed. Either the reporter assay does not reflect what the fusion does in a T-cell progenitor, or Notch is not the relevant axis in those patients. Mariani RA et al. 2020
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   MAML2 accounts for 4 patients among the 3401 genomically characterised KMT2A rearrangements in the DCAL recombinome.

   The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, Burmeister T, Gröger D, et al., Marschalek R · Leukemia 2023
   PMID 37019990DOI
   ▸What this evidence cannot show

   Four patients at one referral centre. Because KMT2A-MAML2 is enriched in therapy-related disease, which is often diagnosed and managed outside the paediatric networks that refer to DCAL, four is likely an undercount relative to the published case literature. Counts read from the published article in this session.

   ▸Evidence that argues against this (1)

   • At least eight distinct KMT2A-MAML2 patients are individually described in the case literature (secondary AML and MDS, secondary T-ALL in several children, B-ALL after sarcoma, a clone that regressed, a T-LBL after RUNX1::RUNX1T1 AML), so the recombinome's 4 plainly does not capture the full published experience. Menu E et al. 2017
4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   KMT2A-MAML2 is the first recurrent gene fusion described in type B thymoma, found in 11 patients with fusion breakpoints between KMT2A exon 8, 9, 10 or 11 and MAML2 exon 2, at 4% of all thymomas (10 of 242) and 6% of B2/B3 histology (10 of 169).

   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 2020
   PMID 32923872DOI
   ▸What this evidence cannot show

   A commercial comprehensive-genomic-profiling cohort of advanced cancers (255,008 samples, 242 thymomas), so the thymoma denominator is enriched for advanced or refractory disease and the 4% figure may not apply to resected early-stage thymoma. Only a portion of samples had RNA profiling, so fusion detection was not uniform across the cohort. No functional work. Abstract read in full.

   ▸Evidence that argues against this (1)

   • A KMT2A::MAML3 fusion has since been reported in a type B3 thymoma, with the authors noting MAML3 and MAML2 are paralogues sharing the same functional domains downstream of exon 2. So the thymoma-defining lesion is a KMT2A-MAML-family fusion rather than KMT2A-MAML2 specifically, and any count keyed to MAML2 alone will undercount it. Zhou Z et al. 2025
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

   MAML2 shows no dependency in KMT2A-rearranged leukaemia lines in either the myeloid (z +0.33) or the B-lineage (z +0.29) half of the index arm, in the same DepMap run in which MEN1 scored -2.29 and -2.28.

   DepMap Chronos gene-effect scores for USP2, MAML2, USP22, USP8, MEN1 and KMT2A in KMT2A-rearranged leukaemia lines, lineage-split (analyses/depmap-usp2-maml2-lineage-split) · Broad Institute DepMap; Open Targets Platform; analysis run for this graph · DepMap / Open Targets Platform 2026
   ▸What this evidence cannot show

   Eight myeloid and five B-lineage KMT2A-rearranged lines; a single line moves a mean. Split by lineage LABEL, not expression signature. No KMT2A::MAML2 line exists in DepMap or anywhere else, and no thymoma line carries the fusion either, so this measures wild-type MAML2 in cells carrying other KMT2A fusions. Ensembl ENSG00000184384 was asserted to resolve to approvedSymbol MAML2 by the query, which aborts on mismatch.

   ▸Evidence that argues against this (1)

   • MAML1 and MAML3 are paralogues that can substitute for MAML2 in Notch co-activation, so a single-gene knockout is expected to be buffered. A flat MAML2 score is therefore the predicted result of redundancy rather than evidence that MAML function is dispensable, and only a combined MAML1/2/3 perturbation would test that. Zhou Z et al. 2025
6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   PubMed retrieval for MAML2 is dominated by salivary-gland work: MAML2 alone returns 544 records, MAML2 AND (mucoepidermoid OR salivary) returns 345, CRTC1 AND MAML2 returns 151 and the retired symbol MECT1 AND MAML2 a further 49, against 21 for (MLL OR KMT2A) AND MAML2.

   PubMed E-utilities esearch counts for MAML2 symbol, alias and partner queries, run 2026-09-19 · National Center for Biotechnology Information (NCBI), PubMed · PubMed (NCBI E-utilities) 2026
   ▸What this evidence cannot show

   One day's index snapshot; counts drift. The CRTC1 and MECT1 sets overlap to an unmeasured degree, so 151 and 49 cannot simply be added. [All Fields] queries also match reference lists. The 21-record intersection was individually title-checked; the larger counts were not, and are reported as raw counts for the purpose of showing the noise ratio, not as characterisations of those literatures.

   ▸Evidence that argues against this (1)

   • Restricting to (MLL OR KMT2A) removes the salivary noise but introduces the opposite error: it excludes any paper that discusses MAML2 fusion biology generically without naming KMT2A, including most of the mechanistic work on the shared MAML2 C-terminus that is directly relevant here. Nemoto N et al. 2007
7. E_q89_maml2_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt Q8IZL2 (MAML2_HUMAN, primary gene MAML2, 1,156 aa, read back 2026-09-21) carries exactly one SUBCELLULAR LOCATION value, 'Nucleus speckle', on a single experimental attribution ECO:0000269 PubMed:12370315, with the uncoded note 'Nuclear, in a punctate manner'. Tissue specificity, also uncoded: 'Widely expressed with high levels detected in placenta, salivary gland and skeletal muscle'.

   UniProtKB Q8IZL2 (MAML2_HUMAN) SUBCELLULAR LOCATION comment and note, fetched 2026-09-21 · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Established as a statement about the database record. Note the internal tension inside UniProt's own entry: the controlled-vocabulary LOCATION says 'Nucleus speckle' while the free-text NOTE says only 'Nuclear, in a punctate manner' - the note is the more faithful rendering of the source and the location term is the one that gets machine-read and propagated.

   ▸Evidence that argues against this (1)

   • A reader might defend UniProt here on the grounds that 'Nucleus speckle' is simply the nearest available term in a controlled vocabulary and no deception is intended, which is true and beside the point. The consequence is what matters: this graph imported the term, wrote that MAML2 has 'a genuine subnuclear assignment and one of very few available for any KMT2A partner', and used that to make MAML2 exceptional among KMT2A partners. Once the source is read, MAML2 is NOT exceptional - it has the same class of evidence as most of them, a transfected-construct punctate pattern with no marker. The vocabulary did not merely round the answer, it changed the entry's argument. That is the general hazard rule 91 names, arriving through the location term rather than through the citation. Wu L et al. 2002
8. E_q89_maml2_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   READ IN FULL. The source of MAML2's subnuclear annotation expressed MAML2 and MAML3 cDNAs FUSED TO GFP, TRANSIENTLY, in COS7 and U2OS cells and visualised them by fluorescence microscopy: 'MAML2 and MAML3 displayed a speckled nuclear staining pattern, with considerable heterogeneity in the sizes of the dots'; FLAG-tagged constructs gave a similar pattern ('data not shown'). The paper's section heading is 'Nuclear body localization of MAML2 and MAML3', its abstract states 'All MAML proteins localize to nuclear bodies', and the precedent it builds on is that MAML1 'was previously found to localize in nuclear bodies'. The string 'speckle' occurs ONCE in the whole article, as the adjective 'speckled' describing the pattern; SC35, SRSF2 and PML do not occur at all.

   Identification of a family of mastermind-like transcriptional coactivators for mammalian notch receptors. · Wu L, Sun T, Kobayashi K, Gao P, Griffin JD · Mol Cell Biol 2002
   PMID 12370315DOI
   ▸What this evidence cannot show

   Read in full via the PMC HTML route (PMC135662, 73,198 characters after tag-strip-then-unescape) because Europe PMC fullTextXML returned HTTP 500 - the four-route procedure in the brief, and route 4 was the one that worked. The word counts above are mechanical string counts over that stripped text and would miss a marker named only in a figure image. The experiment itself is transient overexpression of a GFP fusion in two non-haematopoietic lines, with no endogenous protein, no marker co-stain and no quantitation.

   ▸Evidence that argues against this (1)

   • The obvious rebuttal is that 'nuclear bodies' and 'nuclear speckles' are loose terms that authors in 2002 used interchangeably, so the distinction drawn here is pedantry. It is not, for one concrete reason: nuclear speckles are DEFINED by the splicing-factor content that SC35/SRSF2 marks, and the AFF-family entries in this graph - gene-aff3 and gene-aff4 - rest their speckle calls on a study that tied the compartment to G-quadruplex RNA binding and to measurable splicing modulation. If MAML2's 'speckle' is really a nuclear body, then MAML2 does not belong in the same subnuclear class as the AFF scaffolds, and a graph that files them together will draw a false parallel between a Notch co-activator and the SEC. The rebuttal would carry if anyone had run the marker; nobody has, in twenty-four years, and the residual uncertainty is the finding. UniProt Consortium 2026

• ← fuses withKMT2A / MLL1 — the gene that breaks

  inv(11)(q21q23), 4 patients in the recombinome and at least 8 individually reported; also the recurrent fusion of type B2/B3 thymoma

• ← part ofKMT2A::MAML2, inv(11)(q21q23) - a therapy-related fusion that also causes a solid tumour

  MAML2 exon 2 onward, ~residues 205-1156: transactivation domain retained, MamL-1 Notch-binding domain lost

• ← part ofKMT2A::MAML2 keeps MAML2's activation domain and loses its Notch-binding domain, and suppresses Notch output

  MamL-1 (33-87) lost, transactivation domain (609-681) retained

• ← part ofMAML2 exon 2: the strongly preferred 3' acceptor, the same in leukaemia, thymoma and salivary-gland tumours

  exon 2 acceptor across every reported fusion; retained fragment approximately residues 205-1156

• ← related toKMT2A::MAML2 type B2/B3 thymoma - the same fusion, a different organ, a different disease

  MAML2 exon 2 acceptor, identical to the leukaemia junctions

• ← testsEnumeration for the USP2 / MAML2 / second-BCR beat: what was searched, what was counted, what was rejected

  21 raw, 15 on-topic, 6 itemised rejections; 345 of 544 bare MAML2 records are salivary-gland work

• related to →A partner's 'normal' compartment is usually read off a different tissue, and for at least two KMT2A partners that reading is wrong in blood cells

  CORRECTION: MAML2's 'Nucleus speckle' is UniProt vocabulary over-specifying its source, which reports GFP-MAML2 in COS7/U2OS as a speckled pattern and calls it NUCLEAR BODIES, with no SC35 co-stain (PMID 12370315, read in full) (Q89 2026-09-21)

• related to →Nuclear speckles — the compartment everyone assumes KMT2A::AFF1 uses, on evidence that skips AFF1 itself

  MAML2 was this graph's example of a KMT2A partner with a real speckle assignment; after reading the source the assignment is downgraded to 'punctate nuclear, compartment unresolved' (Q89 2026-09-21)