Oncofusion

start / graph / fusion-kmt2a-septin6

Fusionstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q-nonelongation-partners

KMT2A::SEPTIN6 (MLL-SEPT6), t(X;11)(q24;q23) and its cryptic insertion variants

also called MLL-SEPT6 · MLL::SEPT6 · MLL-SEPTIN6 · KMT2A-SEPT6 · t(X;11)(q24;q23)

Here the broken KMT2A gene is joined to septin-6, a cytoskeleton protein that normally helps a cell divide and has nothing to do with reading genes. It is the best-studied of the septin fusions, and in mice it causes only a slow blood disorder unless a second mutation is added.

The same thing, for a scientist

KMT2A::SEPTIN6 is the most frequent KMT2A-septin fusion (n=19 of 3401 genomically characterised patients; 1.7% of the KMT2A-rearranged AML cohort) and is confined to myeloid disease: 7 infant AML, 10 paediatric AML, 2 adult AML and zero ALL in any age class [E3]. KMT2A (11q23) and SEPTIN6 (Xq24) lie in opposite transcriptional orientations, so the productive 5'-KMT2A/SEPTIN6-3' transcript sits on the der(X) rather than the der(11); the rearrangement is correspondingly cryptic or complex and is missed by karyotype alone [E6]. The fusion retains almost the entire septin ORF, including the septin-type G domain (aa 39-305 of Q14141) with its G1/P-loop, G3 and G4 motifs, and the C-terminal coiled coil (aa 321-409) [E1][E7]. This is the only KMT2A-septin fusion with a real structure-function dissection, and the result is more restrictive than the field's general dimerisation model: homo-oligomerisation is necessary, but so are the GTP-binding domain, the P-loop and the coiled coil individually, and replacing the coiled coil with a 4-OHT-inducible synthetic oestrogen-receptor dimerisation module failed to restore immortalisation even in the presence of 4-OHT [E4]. Alone, MLL-SEPT6 produced long-latency myeloproliferative disease in mice, not acute leukaemia; FLT3-ITD was required for short-latency acute myeloid or biphenotypic leukaemia [E4].

What it looks like
loading 6UPA…
PDB 6UPA ↗
shape

This is real experimental data — atoms placed by crystallography or cryo-EM, deposited by the people who did the experiment.

I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
What is kept and what is lost

The KMT2A side is a standard 5' retention: menin-binding motifs, AT-hooks, SNL1/2 and CXXC kept; PHD fingers, bromodomain, FYRN, TAD, FYRC, WIN and SET lost. The two founding infant AML cases broke at KMT2A exon 7 and exon 8, joined in frame to SEPTIN6 exon 2 2. Across the 19-patient recombinome series, however, SEPTIN6 fusions preferentially break at KMT2A intron 11, a statistically significant shift away from the cohort mean in both paediatric and adult patients 3. The exon-7/8 junctions of the founding cases are therefore not representative, and the mouse construct the whole mechanistic literature rests on was built from the atypical architecture.

The SEPTIN6 side keeps almost the whole open reading frame 1: septin-type G domain (39-305), G1/P-loop (49-56), G3 (101-104), G4 (184-187) and coiled coil (321-409) 7.

Mechanism: oligomerisation is necessary, but a generic dimeriser is not sufficient

Ono and colleagues built the mutant series this whole beat turns on 4. Deleting the GTP-binding domain, deleting the P-loop, the single P-loop substitution S56N that reduces GTP binding, and deleting the coiled coil each abolished serial myeloid immortalisation. Co-immunoprecipitation showed that only intact MLL-SEPT6 self-associated; no mutant did.

Then the decisive negative. The coiled coil was replaced with the 4-hydroxytamoxifen-inducible mutant oestrogen-receptor ligand-binding domain -- the exact synthetic dimerisation trick that converts MLL into an oncoprotein when supplied in place of the AF1p/EPS15 or GAS7 coiled coils 8. MLL-SEPT6 Δcoil-ER did not restore immortalisation, with or without 4-OHT. The authors' own wording is that 'homo-oligomerization with proper whole structure is essential'.

That is a real constraint on the unifying story. Forced dimerisation is sufficient for some partners and is demonstrably not sufficient for this one.

How a cytoskeletal GTPase ends up on chromatin

By confocal immunofluorescence of transiently transfected 293T cells, MLL-SEPT6 distributed as small nuclear speckles or dots, whereas overexpressed SEPT6 alone did not shift from its physiological cytoplasmic localisation 4. The KMT2A N-terminus supplies the nuclear localisation signals and the chromatin-targeting apparatus; the septin is a passenger dragged into a compartment it never normally occupies, where it contributes self-association. MLL-SEPT6-immortalised cells upregulated Hoxa7, Hoxa9 and Meis1, so the canonical KMT2A-fusion output is engaged 4.

The wild-type allele is not the story

Sept6-deficient mice showed no gross abnormality, no cytokinesis defect and no spontaneous malignancy; Sept6 loss did not change the phenotype of MLL-SEPT6-induced myeloproliferative disease; and depleting the close homologue Sept11 on the Sept6-null background did nothing either 5. This is the only KMT2A partner for which disrupting the partner gene has been formally shown not to contribute to leukemogenesis by the fusion, so the loss-of-function/subtraction reading that applies to MLLT10 and AFDN does not apply here.

Diagnosis: why this fusion is missed

KMT2A and SEPTIN6 are transcribed towards each other, so the functional chimera lands on the der(X) and reported cases are dominated by complex insertions such as inv ins(X;11)(q24;q23q13) rather than a simple reciprocal t(X;11) 6. FISH break-apart plus RT-PCR is required. One founding case was originally karyotyped as t(X;3)(q22;p21)ins(X;11)(q22;q13q25) 2. A simple reciprocal t(X;11)(q22;q23) does also occur 2, so the cryptic architecture is the usual case and not a universal one.

Facts and figures
gene5KMT2A
gene3SEPTIN6
cytobandt(X;11)(q24;q23); frequently cryptic inv ins(X;11) or three-way rearrangements
uniprot5Q03164
uniprot3Q14141
uniprot3 verifiedQ14141 = SEPT6_HUMAN, PRIMARY gene name SEPTIN6, 434 aa, Homo sapiens (UniProt REST, checked 2026-09-19). Resolved by reading back the primary gene name, not a synonym match: the old symbol SEPT6 is carried only as a synonym after the HGNC septin renaming.
prognosisNot separately estimated. No cohort large enough for a survival estimate exists for any KMT2A-septin fusion; the recombinome series reports frequency and breakpoints only.
exon numbering noteKMT2A exon numbers follow the Meyer/Marschalek DCAL consensus. Domain arrays for the index-case breakpoint are provisional for the reason stated in its phenotype_notes.
PDB noteNo structure of KMT2A::SEPTIN6 exists, and none of the fusion's self-association interface. 6UPA = crystal structure of the GTPase domain of the human SEPTIN2/SEPTIN6 heterocomplex, 2.51 A; 2QAG = human septin 2/6/7 trimer, 4.0 A; 2QNR = human septin 2 with GDP, 2.6 A. All three titles and resolutions verified against the RCSB entry API on 2026-09-19.
mechanism class noteNOT an elongation hijack in any demonstrated sense. Neither SEC recruitment nor DOT1L recruitment nor H3K79 methylation has ever been measured for this fusion; see ana-kmt2a-septin-fusion-mechanism-absences.
druggability classunexplored
druggability noteMenin dependence has never been tested for this or any other KMT2A-septin fusion. The reagent exists: HF6, an IL-3-dependent mouse line immortalised by MLL/SEPT6, is published and was used as a screening background for cooperating lesions.
stage of hematopoiesismyeloid progenitor; serial replating of mouse bone marrow progenitors, and infant/paediatric AML in patients
age peakinfancy to childhood, with a minority of adult AML
frequency
all_cases_any_age_n0.000all_cases_any_age_n: 0kmt2a_r_aml_pct1.700kmt2a_r_aml_pct: 1.7adult_aml_n2.000adult_aml_n: 2infant_aml_n7.000infant_aml_n: 7paediatric_aml_n10.0paediatric_aml_n: 10dcal_recombinome_n19.0dcal_recombinome_n: 19dcal_recombinome_total3,401dcal_recombinome_total: 3401
diseases
[
  "infant AML",
  "paediatric AML",
  "adult AML"
]
frequency — the numbers behind the chart
{
  "dcal_recombinome_n": 19,
  "dcal_recombinome_total": 3401,
  "kmt2a_r_aml_pct": 1.7,
  "infant_aml_n": 7,
  "paediatric_aml_n": 10,
  "adult_aml_n": 2,
  "all_cases_any_age_n": 0
}
breakpoints
[
  {
    "id": "bp-kmt2a-intron-11-septin6",
    "gene5_exon_last": 11,
    "gene3_exon_first": 2,
    "population": "the preferential KMT2A breakpoint location for SEPTIN6 fusions in the 3401-patient recombinome (significant shift towards intron 11 in paediatric and adult patients). No per-case percentage is published, so frequency_pct is deliberately omitted rather than guessed.",
    "domains_retained_5": [
      "MBM1",
      "MBM2",
      "AT-hook 1",
      "AT-hook 2",
      "AT-hook 3",
      "SNL1/2",
      "CXXC"
    ],
    "domains_lost_5": [
      "PHD1",
      "PHD2",
      "PHD3",
      "Bromodomain",
      "PHD4",
      "FYRN",
      "TAD",
      "FYRC",
      "WIN",
      "SET"
    ],
    "domains_retained_3": [
      "Septin-type G domain (aa 39-305)",
      "G1/P-loop motif (aa 49-56)",
      "G3 motif (aa 101-104)",
      "G4 motif (aa 184-187)",
      "coiled coil (aa 321-409)"
    ],
    "domains_lost_3": [
      "SEPTIN6 exon 1 / extreme N-terminal segment upstream of the junction"
    ],
    "phenotype_notes": "Myeloid, HOXA-high. Both the G domain and the coiled coil are individually required for transformation and a synthetic dimerisation module does not substitute for the coiled coil."
  },
  {
    "id": "bp-kmt2a-intron-7-septin6-index",
    "gene5_exon_last": 7,
    "gene3_exon_first": 2,
    "population": "the two founding infant AML cases (MLL exon 7 and MLL exon 8 to SEPTIN6 exon 2); atypical relative to the 19-patient series but the architecture on which the mouse mechanistic work was modelled",
    "domains_retained_5": [
      "MBM1",
      "MBM2",
      "AT-hook 1",
      "AT-hook 2",
      "AT-hook 3",
      "SNL1/2"
    ],
    "domains_lost_5": [
      "PHD1",
      "PHD2",
      "PHD3",
      "Bromodomain",
      "PHD4",
      "FYRN",
      "TAD",
      "FYRC",
      "WIN",
      "SET"
    ],
    "domains_retained_3": [
      "Septin-type G domain (aa 39-305)",
      "coiled coil (aa 321-409)"
    ],
    "domains_lost_3": [
      "SEPTIN6 exon 1"
    ],
    "phenotype_notes": "The KMT2A intron 7 breakpoint was a functional topoisomerase II cleavage site in vitro. CXXC is deliberately listed in neither retained nor lost array for this junction: the founding papers report exon junctions, not domain fate, and the CXXC boundary relative to exon 7/8 was not verified from a primary source for this entry."
  }
]
PDB IDS
[
  "6UPA",
  "2QAG",
  "2QNR"
]
mechanism class
[
  "forced-oligomerisation"
]
drugs
[]
key interactors
[
  "gene-men1",
  "gene-hoxa9",
  "gene-meis1",
  "gene-flt3"
]
trials
[]
How this could bite us back
  • If menin inhibitors are given on the basis of 'KMT2A-rearranged' status alone, septin-partner patients are being treated on an untested mechanistic assumption; a partner-class-specific non-response would look like ordinary primary resistance and be misattributed.
  • The fusion alone gave only long-latency myeloproliferative disease in mice. If human KMT2A::SEPTIN6 AML likewise needs a cooperating lesion such as FLT3-ITD, targeting the fusion axis alone may be insufficient and the cooperating lesion may be the more tractable target.
  • Septins are broadly required for cytokinesis, so a drug aimed at a septin self-association interface risks on-target toxicity in every dividing tissue -- while the Sept6-null mouse shows the family is redundant enough that partial inhibition may simply be compensated, giving the worst of both.
Still open
  1. Is KMT2A::SEPTIN6 menin-dependent? Nobody has asked: a PubMed intersection of menin / menin-inhibitor terms with septin terms returns zero records.
  2. Why does a synthetic ER dimerisation module rescue AF1p/EPS15 and GAS7 fusions but not MLL-SEPT6? Is the difference stoichiometric (filament-order oligomer versus dimer), geometric, or is the negative construct an artefact?
  3. Does the intron-11 breakpoint that dominates real SEPTIN6 patients retain CXXC and part of PHD1, and does that change the mechanism relative to the exon-7 index architecture the mouse work was built on?
  4. Does MLL-SEPT6 recruit the super elongation complex or DOT1L at all, and does H3K79me2 spread over HOXA in these cells?
Every source behind this entry (8)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    SEPTIN6 on Xq24 is fused to MLL in three de novo infant AML with complex abnormalities involving 11q23 and Xq22-24, and the MLL-SEPTIN6 fusion protein contains almost the entire septin protein, as do MLL-CDCREL1 and MLL-AF17q25/MSF.

    PMID 11809673
    What this evidence cannot show

    Three index patients. The authors' further inference that MLL-SEPT6 AML is a subset restricted to FAB M1/M2 rests on those three plus the literature of the time, and later cases contradict it.

    Evidence that argues against this (1)
    • The FAB restriction does not hold: Slater et al. report MLL-SEPTIN6 in infant acute myeloMONOcytic leukaemia (M4), and Kim et al. in acute monocytic leukaemia, so the founding paper's lineage claim is refuted by subsequent cases. Slater DJ et al. 2002
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In two infant AML cases MLL exon 7 and MLL exon 8 were joined in frame to SEPTIN6 exon 2, the MLL genomic breakpoint lay in intron 7 and was a functional DNA topoisomerase II cleavage site in an in vitro assay, and FISH placed the MLL-SEPTIN6 junction on the der(X).

    PMID 12096348DOI
    What this evidence cannot show

    Two patients. The topoisomerase II cleavage assay is in vitro on naked DNA and does not establish that etoposide or any dietary topo II poison caused these particular breaks.

    Evidence that argues against this (1)
    • The intron-7 breakpoint of these index cases is atypical: across 19 SEPTIN6 patients the KMT2A breakpoint distribution shifts significantly towards intron 11 in both paediatric and adult patients, the opposite end of the major breakpoint cluster region. Meyer C et al. 2023
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In 3401 genomically characterised KMT2A-rearranged acute leukaemia patients, SEPTIN6 was the 12th most frequent translocation partner with n=19 (7 infant AML, 10 paediatric AML, 2 adult AML, 0 ALL in any age class), representing 1.7% of the KMT2A-rearranged AML cohort, and SEPTIN6 fusions preferentially carry KMT2A intron 11 breaks.

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

    A referral cohort accrued 2003-2022 from 20 countries, not a population incidence estimate. Cryptic der(X) rearrangements are precisely what a karyotype-led referral stream under-ascertains, so 19 is a floor.

    Evidence that argues against this (1)
    • Because KMT2A::SEPTIN6 is usually cryptic or complex and sits on the der(X), the founding series required panhandle PCR and spectral karyotyping to detect it at all; the recombinome count therefore cannot be read as an unbiased frequency and the true share of AML is probably higher. Slater DJ et al. 2002
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Fusion-partner-mediated homo-oligomerisation of MLL-SEPT6 is essential for myeloid immortalisation and requires both the GTP-binding domain and the coiled coil: the GTP-domain deletion, the P-loop deletion, the S56N P-loop substitution and the coiled-coil deletion all failed to immortalise, only intact MLL-SEPT6 self-associated by co-immunoprecipitation, replacing the coiled coil with a 4-OHT-inducible oestrogen-receptor dimerisation module failed to rescue immortalisation, MLL-SEPT6 localised to nuclear speckles while SEPT6 alone remained cytoplasmic, Hoxa7/Hoxa9/Meis1 were upregulated, and MLL-SEPT6 alone gave long-latency myeloproliferative disease rather than acute leukaemia unless FLT3-ITD was co-expressed.

    PMID 15761502DOI
    What this evidence cannot show

    Retroviral overexpression in mouse bone marrow progenitors. The synthetic-dimeriser experiment is a single negative construct, and a failure to rescue can always be an expression, folding or geometry artefact rather than a biological requirement. Localisation was assayed in transiently transfected 293T cells, not leukaemic blasts. Full text read at jci.org (open access), not abstract only.

    Evidence that argues against this (1)
    • The same synthetic-dimerisation strategy that fails here succeeds for other cytoplasmic KMT2A partners: So et al. showed the coiled-coil oligomerisation domains of AF1p/EPS15 and GAS7 are necessary AND sufficient, and that homodimerisation of MLL by synthetic dimerisation modules mimics bona fide MLL fusion proteins with Hox activation and enhanced progenitor self-renewal. Either the septin requirement is qualitatively different or the negative construct is an artefact; the two results have never been reconciled experimentally. So CW et al. 2003
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Sept6-deficient mice showed no gross abnormality, no change in cytokinesis and no spontaneous malignancy; loss of Sept6 did not alter the myeloproliferative disease induced by MLL-SEPT6; and depletion of the close homologue Sept11 did not affect Sept6-null cells, indicating SEPT6 does not act as a tumour suppressor in this fusion.

    PMID 16314519DOI
    What this evidence cannot show

    Mouse knockout scoring gross phenotype and disease latency. The authors themselves attribute the absence of phenotype to a high degree of redundancy in the septin system, so a requirement masked by paralogues cannot be excluded.

    Evidence that argues against this (1)
    • Wild-type partner dosage IS measurably reduced in patients: SEPT6 expression is significantly down-regulated 1.7-fold (p=0.002) in MLL-SEPT6 leukaemia alongside a 6.4-fold MLL down-regulation (p=0.008), so a dosage contribution in humans is not excluded by a mouse knockout that scores only gross phenotype and latency. Santos J et al. 2010
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KMT2A and SEPTIN6 lie in reverse orientation on their respective chromosomes, so the 5'-MLL/SEPTIN6-3' transcript resides on the derivative X chromosome rather than the derivative 11, and MLL/SEPTIN6 rearrangements are correspondingly cryptic or complex; in a 29-month-old with acute monocytic leukaemia and inv ins(X;11)(q24;q23q13), FISH showed the break-apart 5'-MLL segment translocated to the der(X).

    PMID 12874781DOI
    What this evidence cannot show

    One case plus a review of all reported 11q23/Xq22-24 rearrangements up to 2003; only six molecularly confirmed SEPTIN6 cases existed at that time.

    Evidence that argues against this (1)
    • Simple reciprocal t(X;11)(q22;q23) also occurs -- Slater et al.'s second case had that karyotype with two MLL rearrangements by Southern blot -- so a normal-looking t(X;11) must still be typed molecularly, and the cryptic architecture is the usual case rather than a universal one. Slater DJ et al. 2002
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Human septin-6 (UniProt Q14141, SEPT6_HUMAN, primary gene name SEPTIN6, 434 aa) carries a septin-type G domain at aa 39-305 with G1/P-loop at 49-56, G3 at 101-104 and G4 at 184-187, plus a coiled coil at aa 321-409.

    What this evidence cannot show

    Database annotation for the canonical isoform. SEPTIN6 is predicted to encode at least two protein types by alternative splicing, so residue numbers are isoform-dependent.

    Evidence that argues against this (1)
    • The coiled coil UniProt annotates is not what drives native septin polymerisation: X-ray crystallography and electron microscopy of the human SEPT2-SEPT6-SEPT7 complex showed the predicted coiled coils are NOT involved in, or required for, complex or filament formation, which proceed through the nucleotide-binding G interface and the N/C extensions. The fusion's absolute requirement for the coiled coil is therefore not explained by normal septin assembly. Sirajuddin M et al. 2007
  8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The coiled-coil oligomerisation domains of AF1p/EPS15 and GAS7 are necessary and sufficient for leukemogenic transformation by their MLL fusions, and homodimerisation of MLL by synthetic dimerisation modules mimics bona fide MLL fusion proteins, producing Hox gene activation and enhanced self-renewal of haematopoietic progenitors.

    PMID 12957285DOI
    What this evidence cannot show

    Mouse progenitor assays with retroviral constructs; the synthetic modules were tested on the KMT2A N-terminus, not on a septin backbone.

    Evidence that argues against this (1)
    • The synthetic module does not generalise to this partner: substituting a 4-OHT-inducible ER dimerisation domain for the SEPT6 coiled coil failed to immortalise myeloid progenitors with or without 4-OHT. Ono R et al. 2005
Connected to