Oncofusion

start / graph / fusion-kmt2a-septin9

Fusionmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q-nonelongation-partners

KMT2A::SEPTIN9 (MLL-SEPT9 / MLL-MSF), t(11;17)(q23;q25)

also called MLL-SEPT9 · MLL-MSF · MLL-AF17q25 · KMT2A-SEPT9 · MLL::SEPTIN9 · t(11;17)(q23;q25)

The second most common septin partner: KMT2A is joined to septin-9, a scaffolding protein of the cell's skeleton. Unlike most KMT2A fusions it shows up in adults and in leukaemias caused by earlier chemotherapy, and it is easy to confuse with a completely different t(11;17) leukaemia that is treated in a completely different way.

The same thing, for a scientist

KMT2A::SEPTIN9 arises from t(11;17)(q23;q25) and was discovered twice in 1999, independently, as MSF ('MLL septin-like fusion') [E1] and as AF17q25 [E2]. It is the 13th most frequent KMT2A partner in the DCAL recombinome, n=14 of 3401, distributed 2 infant AML / 6 paediatric AML / 6 adult AML with zero ALL cases in any age class -- the flattest age distribution of any septin partner and the only one with as many adult as paediatric cases [E3]. Reported cases are enriched for therapy-related disease and for myeloid neoplasms outside classical AML, including de novo MDS [E5] and acute megakaryoblastic leukaemia [E6]. The partner gene is SEPTIN9 (UniProt Q9UHD8, 586 aa), whose UniProt record still carries 'MLL septin-like fusion protein MSF-A' as a protein alias [E7]. Two facts make this fusion a diagnostic trap: t(11;17) at the cytogenetic level covers at least six molecularly distinct fusions, of which one is ZBTB16::RARA, and the MLL-SEPT9 transcripts themselves are highly variable in splice form, so a single primer pair will miss cases [E4].

What it looks like
loading 5CYO…
PDB 5CYO ↗
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
Two independent discoveries, one gene

Osaka, Rowley and Zeleznik-Le cloned the partner from a 10-year-old who developed therapy-related AML 17 months after treatment for Hodgkin's disease, named it MSF, and noted it was a septin family member with the conserved GTP-binding domain, cytoplasmic localisation and cytoskeletal association -- and that it was highly homologous to hCDCrel-1, already an MLL partner in t(11;22) 1. Taki and colleagues independently identified AF17q25 in one de novo and one therapy-related AML with the same t(11;17)(q23;q25), and concluded that AF17q25 and hCDCrel 'might define a new septin family particularly involved in the pathogenesis of 11q23-associated leukemia' 2. Both groups arrived at the same inference from the same observation: septins keep turning up as KMT2A partners, and nobody yet knows why.

Age and disease distribution: the adult septin

In 3401 genomically characterised patients SEPTIN9 accounts for 14 cases: 2 infant AML, 6 paediatric AML, 6 adult AML, and zero ALL at any age 3. That adult share is unusual for the septin class (SEPTIN6 has 2 of 19 adult; ABI1 has none) and it sits alongside a literature that is visibly enriched for therapy-related myeloid neoplasia 12. SEPTIN9 fusions also carry a distinctive breakpoint signature on the KMT2A side: their breakpoints shift significantly towards KMT2A introns 7-9, i.e. upstream, because of the intron phase of the partner's breakpoint cluster region 3.

The phenotype is not confined to classical AML. Reported cases include de novo MDS with multilineage dysplasia that responded to decitabine 5 and a 2-year-old with acute megakaryoblastic leukaemia 6.

The t(11;17) diagnostic trap

Translocations between 11q and 17q in AML look identical cytogenetically and are molecularly diverse. In 14 t(11;17) cases analysed by FISH and molecular methods, Strehl and colleagues found six different gene fusions: seven MLL-MLLT6/AF17, three MLL-SEPT9/MSF, one each of MLL-LASP1, MLL-ACACA and ZBTB16/PLZF-RARA, and one MLL-SEPT6/Xq24 indicating a complex rearrangement 4. ZBTB16::RARA is an APL variant with entirely different therapy. Separately, MLL-SEPT9 transcripts showed a high degree of splice variability, so primer selection determines detection 4. Case reports independently describe AML with t(11;17) being clinically mistaken for APL 8.

Mechanism: assumed, not measured

There is no published structure-function dissection of KMT2A::SEPTIN9. Everything mechanistic said about it is transposed from MLL-SEPT6 [see fusion-kmt2a-septin6] and the general dimerisation literature. Two specifics make the transposition non-trivial. First, the canonical SEPTIN9 isoform (Q9UHD8) carries the septin-type G domain at aa 295-567 but no annotated coiled coil at all, unlike SEPTIN6, SEPTIN5 and SEPTIN11 7 -- and the coiled coil is one of the two elements shown to be individually required in MLL-SEPT6. Second, the SEPTIN9 transcripts in patients are heterogeneous 4, so which septin isoform is actually in the fusion protein varies between cases. Anyone modelling this fusion has to choose an isoform, and the literature has not.

Facts and figures
gene5KMT2A
gene3SEPTIN9
cytobandt(11;17)(q23;q25)
uniprot5Q03164
uniprot3Q9UHD8
uniprot3 verifiedQ9UHD8 = SEPT9_HUMAN, PRIMARY gene name SEPTIN9, 586 aa, Homo sapiens (UniProt REST, checked 2026-09-19). Synonyms carried on the record are KIAA0991, MSF and the retired symbol SEPT9; resolved by reading back the primary gene name, not by a synonym match.
prognosisNo survival estimate exists. One reported MDS case responded to decitabine and one AML case achieved persistent complete remission on decitabine-based treatment; these are isolated reports, not an outcome signal.
exon numbering noteKMT2A exon numbers follow the Meyer/Marschalek DCAL consensus. SEPTIN9 exon numbering is isoform-dependent and deliberately not asserted.
PDB noteNo structure of the fusion. 5CYO = high-resolution septin 9 GTPase domain with GDP, 2.04 A; 5CYP = septin 9 GTPase domain with GTP-gamma-S, 2.89 A; 4YQF = GTPase domain of human septin 9, 2.73 A. Titles and resolutions verified against the RCSB entry API on 2026-09-19. These cover the G domain that the fusion retains; nothing covers a KMT2A-SEPTIN9 self-association interface, which has never been defined.
mechanism class noteAssigned by analogy to MLL-SEPT6 and to the general dimerisation literature, NOT by any experiment on this fusion. No transformation assay, no interactome, no chromatin readout has been published for KMT2A::SEPTIN9.
druggability classunexplored
druggability noteNo targeted agent has been reported in a KMT2A::SEPTIN9 patient except anecdotal decitabine responses. Menin dependence untested.
stage of hematopoiesismyeloid progenitor; one reported case was megakaryoblastic and one was MDS with multilineage dysplasia, so the cell of origin is not pinned down
age peakdistributed across paediatric and adult AML, with the flattest age profile of any septin partner
frequency
all_cases_any_age_n0.000all_cases_any_age_n: 0kmt2a_r_aml_pct1.300kmt2a_r_aml_pct: 1.3infant_aml_n2.000infant_aml_n: 2paediatric_aml_n6.000paediatric_aml_n: 6adult_aml_n6.000adult_aml_n: 6dcal_recombinome_n14.0dcal_recombinome_n: 14dcal_recombinome_total3,401dcal_recombinome_total: 3401
diseases
[
  "paediatric AML",
  "adult AML",
  "therapy-related AML",
  "de novo MDS",
  "acute megakaryoblastic leukaemia"
]
frequency — the numbers behind the chart
{
  "dcal_recombinome_n": 14,
  "dcal_recombinome_total": 3401,
  "kmt2a_r_aml_pct": 1.3,
  "infant_aml_n": 2,
  "paediatric_aml_n": 6,
  "adult_aml_n": 6,
  "all_cases_any_age_n": 0
}
breakpoints
[
  {
    "id": "bp-kmt2a-intron-7-9-septin9",
    "gene5_exon_last": 9,
    "gene3_exon_first": 2,
    "population": "SEPTIN9 fusions shift significantly towards KMT2A introns 7-9, upstream of the cohort mean, because of the intron phase of the SEPTIN9 breakpoint cluster region (n=14). gene3_exon_first is indicative: the SEPTIN9 transcripts are highly splice-variable and no single consensus junction is published.",
    "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 295-567)",
      "G1 motif (aa 305-312)",
      "G3 motif (aa 362-365)",
      "G4 motif (aa 444-447)"
    ],
    "domains_lost_3": [
      "a variable amount of the disordered SEPTIN9 N-terminal extension, isoform-dependent"
    ],
    "phenotype_notes": "CXXC is omitted from both arrays because the breakpoint interval spans the CXXC-encoding region and no primary source resolves its fate for this partner. The canonical SEPTIN9 isoform has no annotated coiled coil, so the domain content of this fusion differs from MLL-SEPT6 in exactly the element shown to be required there."
  }
]
PDB IDS
[
  "5CYO",
  "5CYP",
  "4YQF"
]
mechanism class
[
  "forced-oligomerisation"
]
drugs
[]
key interactors
[
  "gene-men1"
]
trials
[]
How this could bite us back
  • The commonest practical failure here is diagnostic, not therapeutic: a cytogenetic t(11;17) reported without molecular typing can be an APL variant (ZBTB16::RARA), an MLLT6 fusion or a SEPTIN9 fusion, and each implies different therapy. An assay built on one primer pair will miss splice variants of this fusion.
  • Any drug against a septin self-association interface inherits the essential-function problem: SEPTIN9 is required in normal cells and is separately implicated as a pro-tumorigenic isoform switch in epithelial cancer, so the therapeutic window is unknown in both directions.
  • If a septin-partner patient is enrolled in a menin-inhibitor programme on 'KMT2A-rearranged' status and fails, the failure will be recorded as primary menin-inhibitor resistance rather than as a partner-class question, which is how a testable hypothesis becomes invisible.
Still open
  1. Does the canonical SEPTIN9 isoform, which has no annotated coiled coil, oligomerise in the fusion context at all? If it does, the coiled-coil requirement demonstrated for MLL-SEPT6 is not a class property, and the two fusions work differently.
  2. Which SEPTIN9 isoform is present in patient fusion proteins? Transcript variability is documented but no one has determined the protein product.
  3. Is the adult and therapy-related skew real, or an ascertainment artefact of who gets t(11;17) molecularly typed?
  4. Is KMT2A::SEPTIN9 menin-dependent? Never tested.
Every source behind this entry (8)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The partner of MLL in a t(11;17)(q23;q25) therapy-related AML arising 17 months after treatment for Hodgkin's disease was cloned as a new septin family member and named MSF (MLL septin-like fusion); septins conserve the GTP-binding domain, localise in the cytoplasm and interact with cytoskeletal filaments, and MSF is highly homologous to hCDCrel-1, the MLL partner in t(11;22)(q23;q11.2).

    PMID 10339604DOI
    What this evidence cannot show

    A single index patient; the paper establishes the fusion's existence and the partner's identity, not its mechanism. No functional transformation assay.

    Evidence that argues against this (1)
    • The therapy-related framing of the index case does not define the entity: in the recombinome series KMT2A::SEPTIN9 appears across infant, paediatric and adult AML, and de novo (non-therapy-related) presentations including MDS are documented. Kreuziger LM et al. 2007
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In one de novo and one therapy-related AML with t(11;17)(q23;q25), the gene AF17q25 on 17q25 -- a putative septin family gene encoding at least three protein types (568, 594 and 574 aa) -- was found fused to MLL, and the fusion transcript used the type I AF17q25 transcript.

    PMID 10485469
    What this evidence cannot show

    Two patients. AF17q25 is the same gene as MSF/SEPTIN9; the duplicate naming persisted for years and is still a source of literature-search error. No DOI is registered for this 1999 Cancer Research article; the PMID was fetched from NCBI and is the only stable identifier.

    Evidence that argues against this (1)
    • The authors' framing that AF17q25 and hCDCrel 'might define a new septin family' particularly involved in 11q23 leukaemia was too narrow: five different septin genes across three septin subgroups (SEPTIN2, SEPTIN5, SEPTIN6, SEPTIN9, SEPTIN11) are now documented KMT2A partners, so the property being selected for is not specific to one septin subfamily. Cerveira N et al. 2011
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In 3401 genomically characterised KMT2A-rearranged patients SEPTIN9 was the 13th most frequent partner with n=14 (2 infant AML, 6 paediatric AML, 6 adult AML, 0 ALL in any age class), representing 1.3% of the KMT2A-rearranged AML cohort, and SEPTIN9 fusions shift significantly towards KMT2A introns 7-9 because of the intron phase of the partner's breakpoint cluster region.

    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

    Referral cohort, not population incidence. The age distribution rests on 14 patients, so the apparently adult-weighted profile has wide confidence intervals.

    Evidence that argues against this (1)
    • A case series specifically dissecting t(11;17) found 3 MLL-SEPT9 among 14 t(11;17) cases, a much higher local share, which indicates ascertainment depends heavily on whether molecular typing of t(11;17) is performed at all; the recombinome count is not a stable denominator-free frequency. Strehl S et al. 2006
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Fourteen AML cases with a cytogenetic t(11;17) resolved into six distinct gene fusions -- seven MLL-MLLT6/AF17, three MLL-SEPT9/MSF, one MLL-LASP1, one MLL-ACACA, one ZBTB16/PLZF-RARA and one MLL-SEPT6/Xq24 complex rearrangement -- and the MLL-SEPT9 transcripts showed a high degree of splice and transcript variability, so primer choice determines detection.

    PMID 16897742DOI
    What this evidence cannot show

    Fourteen cases from a referral series; the proportions are not generalisable, and the point being cited is the qualitative heterogeneity, not the counts.

    Evidence that argues against this (1)
    • The heterogeneity cuts the other way too: the recombinome's genomic-breakpoint approach resolves partner identity without depending on transcript primer design, and it finds MLLT6 and SEPTIN9 at comparable frequencies (27 and 14 of 3401), so transcript variability limits RT-PCR assays specifically rather than partner assignment in general. Meyer C et al. 2023
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A patient with multilineage dysplasia and t(11;17)(q23;q25) who responded to decitabine had MLL rearrangement by FISH and an MLL-SEPT9 transcript fusion by RT-PCR and sequencing; MLL-SEPT9 had at that point been described in only seven cases of AML and not in MDS.

    An MLL-SEPT9 fusion and t(11;17)(q23;q25) associated with de novo myelodysplastic syndrome. · Kreuziger LM, Porcher JC, Ketterling RP, et al. · Leuk Res 2007
    PMID 17250889DOI
    What this evidence cannot show

    Single case. A decitabine response in one patient is not evidence of hypomethylating-agent sensitivity for the fusion class.

    Evidence that argues against this (1)
    • The decitabine signal is anecdotal and uncontrolled: it rests on isolated reports rather than any partner-stratified trial, and no KMT2A-septin patient has been reported in a prospective study of any targeted agent, including the registrational menin-inhibitor trial. Issa GC et al. 2025
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A 2-year-old girl with acute megakaryoblastic leukaemia and t(11;17)(q23;25) was found by targeted RNA sequencing to carry a KMT2A-SEPT9 fusion, which the authors state had not previously been reported in that subset.

    A case of KMT2A-SEPT9 fusion-associated acute megakaryoblastic leukemia. · Forlenza CJ, Zhang Y, Yao J, et al. · Cold Spring Harb Mol Case Stud 2018
    PMID 30455225DOI
    What this evidence cannot show

    Single case report; establishes phenotypic range, nothing about mechanism or outcome for the class.

    Evidence that argues against this (1)
    • Megakaryoblastic presentation is a one-case observation against a recombinome distribution that classifies all 14 SEPTIN9 patients simply as AML, so it should not be read as a characteristic phenotype of the fusion. Meyer C et al. 2023
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Human septin-9 is UniProt Q9UHD8 (SEPT9_HUMAN), primary gene name SEPTIN9, 586 aa, carrying a septin-type G domain at aa 295-567 with G1 motif 305-312, G3 362-365 and G4 444-447, extensive disordered N-terminal regions (1-49, 62-108, 134-268) and NO annotated coiled coil; the entry's protein aliases include 'MLL septin-like fusion protein MSF-A'.

    What this evidence cannot show

    Canonical isoform only. SEPTIN9 has many isoforms with different N-terminal extensions, and the absence of a coiled-coil annotation on the canonical record does not prove the protein lacks a self-association helix -- the annotation is prediction-driven.

    Evidence that argues against this (1)
    • Absence of a coiled-coil annotation does not mean absence of oligomerisation capacity: native septin filaments assemble through the nucleotide-binding G interface and the N/C extensions, and the predicted coiled coils were shown by crystallography and EM NOT to be involved in complex or filament formation at all. A SEPTIN9 G domain could oligomerise perfectly well without one. Sirajuddin M et al. 2007
  8. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Monocytic AML with KMT2A translocations to chromosome 17q can clinically mimic acute promyelocytic leukaemia.

    PMID 32984885DOI
    What this evidence cannot show

    Small case series; the title as published contains a typographical error in the gene symbol ('KM2TA'), reproduced here verbatim from the NCBI record. The article covers KMT2A-17q translocations generally, not SEPTIN9 specifically.

    Evidence that argues against this (1)
    • The confusion runs in both directions and is molecular, not just clinical: within cytogenetic t(11;17) the actual APL-variant fusion ZBTB16::RARA and the KMT2A fusions coexist in the same case series, so neither morphology nor karyotype settles it and only molecular typing does. Strehl S et al. 2006
Connected to