SEPTIN9 (septin-9 / MSF) — cytoskeletal GTPase, KMT2A fusion partner in t(11;17)(q23;q25)
also called SEPT9 · MSF · MSF-A · AF17q25 · KIAA0991 · septin-9
Septin-9 is a scaffolding GTPase of the cell skeleton whose official protein name still records how it was found: 'MLL septin-like fusion protein'. It was discovered because it turned up fused to KMT2A in leukaemia, and only later studied as a normal protein.
The same thing, for a scientist
SEPTIN9 (UniProt Q9UHD8, SEPT9_HUMAN, 586 aa in the canonical isoform) is a group-III septin whose UniProt protein names still include 'MLL septin-like fusion protein MSF-A' [E1]. Its architecture differs from the other KMT2A-partner septins in two ways that matter mechanistically: a very long disordered N-terminal extension (disordered regions at 1-49, 62-108 and 134-268) preceding the septin-type G domain at 295-567, and **no annotated coiled coil** on the canonical isoform, whereas SEPTIN6, SEPTIN5 and SEPTIN11 all carry one [E1]. Since the coiled coil is one of the two elements individually required for transformation by MLL-SEPT6 [E2], KMT2A::SEPTIN9 either transforms by a different route or the coiled-coil requirement is not a class property. Nobody has tested it. SEPTIN9 is the one KMT2A-partner septin with a substantial independent cancer literature of its own, as a differentially expressed and isoform-switched gene in ovarian and breast neoplasia [E3] and as a driver of migration and ECM degradation in mammary epithelial cells [E4] — but that literature concerns epithelial tumours and over-expression, not the leukaemic fusion. In DepMap SEPTIN9 is not selectively required in KMT2A-rearranged lines (rank 13,148 of 17,728; mean effect -0.17 across 1090 lines) [E5].
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.
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.
The name is a fossil of the discovery
SEPTIN9 entered the literature twice in 1999 under two names from two groups analysing t(11;17)(q23;q25) AML: MSF, for 'MLL septin-like fusion' 6, and AF17q25 7. The UniProt record still carries MSF-A among its protein names and MSF among its gene synonyms 1. The practical consequence is a literature-search trap. A PubMed [Title/Abstract] search intersecting MLL/KMT2A with 'SEPT9' returns 20 records, with 'SEPTIN9' returns 1, and with 'MSF' returns 14, and the three sets are not nested. Any systematic enumeration that uses only the current HGNC symbol will miss almost the entire literature.
Architecture: the septin without a coiled coil
The canonical Q9UHD8 isoform is 586 aa, of which roughly the first 270 are disordered extension and the G domain occupies 295-567 with G1 at 305-312, G3 at 362-365 and G4 at 444-447 1. UniProt annotates no coiled coil. Three high-resolution structures of the isolated SEPTIN9 G domain exist -- with GDP, with GTP-gamma-S and apo 8 -- so the nucleotide-binding module is structurally well characterised while the fusion's self-association interface is entirely undefined.
Against the native-assembly data this is less anomalous than it looks: septin filaments polymerise through the G interface and the N/C extensions, not through coiled coils 9. On that reading a coiled-coil-free SEPTIN9 should oligomerise perfectly well. The anomaly is on the other side -- why MLL-SEPT6 needs its coiled coil at all.
A separate cancer literature that is not about this fusion
SEPTIN9 is altered in epithelial neoplasia independently of any KMT2A involvement: altered expression and isoform usage in ovarian neoplasia 3, and specific isoforms promoting tumorigenesis in mammary epithelial cells by increasing migration and ECM degradation through metalloproteinase secretion at focal adhesions 4. This body of work is a reason SEPTIN9 is comparatively well reagented, and a reason to be careful: none of it is in haematopoietic cells and none of it involves the fusion. It should not be imported as mechanism for KMT2A::SEPTIN9.
| uniprot | Q9UHD8 |
|---|---|
| uniprot verified | Q9UHD8 = SEPT9_HUMAN, PRIMARY gene name SEPTIN9, 586 aa, Homo sapiens. Fetched from the UniProt REST API on 2026-09-19; primary gene name read back rather than trusting a gene_exact synonym match, which for septins is unusually dangerous because MSF, SEPT9 and SEPTIN9 all appear on the record. |
| length AA | 586 |
| gene symbol note | Renamed SEPT9 -> SEPTIN9 by HGNC, and previously published under two independent names, MSF and AF17q25. PubMed [Title/Abstract] intersections with MLL/KMT2A: SEPT9 = 20 records, MSF = 14, SEPTIN9 = 1. All three must be searched. |
| PDB note | 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. All verified against the RCSB entry API on 2026-09-19. |
| alphafold ID | AF-Q9UHD8-F1 |
| domains absent note | No coiled coil is annotated on the canonical isoform, in contrast to SEPTIN6 (321-409), SEPTIN5 (338-369) and SEPTIN11 (320-415). This is a real architectural difference between KMT2A septin partners and is the single most testable prediction in this entry. |
| localization | cytoplasm; focal adhesions in epithelial models |
| family | septin, group III (SEPT3/9/12 group) |
PDB IDS
[ "5CYO", "5CYP", "4YQF" ]
domains — the numbers behind the chart
[
{
"name": "Septin-type G domain",
"start": 295,
"end": 567
},
{
"name": "G1/P-loop motif",
"start": 305,
"end": 312
},
{
"name": "G3 motif",
"start": 362,
"end": 365
},
{
"name": "G4 motif",
"start": 444,
"end": 447
}
]essentiality
{
"depmap_common_essential": false,
"chronos_mean": -0.172,
"chronos_n_lines": 1090,
"kmt2a_selectivity_rank": 13148,
"kmt2a_selectivity_of": 17728,
"mouse_ko": "not established in this entry"
}depmap kmt2a selectivity
{
"chronos_mean_kmt2a_rearranged": -0.0922,
"chronos_mean_other_leukaemia": -0.1249,
"difference": 0.0326,
"pan_essentiality_all_lines": -0.1724,
"selectivity_rank": 13148,
"rank_denominator": 17728,
"provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
"how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}- SEPTIN9 has a real oncogenic literature in epithelial cancer, which makes it tempting to import mechanism. That mechanism is cytoplasmic and adhesion-based and cannot operate on a nuclear fusion protein; importing it would be a category error.
- The gene has been called MSF, AF17q25, SEPT9 and SEPTIN9 in the primary literature. Any automated enumeration of KMT2A partners keyed to the current symbol will silently under-count this partner, which is how a recurrent fusion comes to look like a curiosity.
- Does a coiled-coil-free septin still supply a transforming self-association interface to KMT2A? KMT2A::SEPTIN9 versus KMT2A::SEPTIN6 is a natural experiment that nobody has run.
- Which SEPTIN9 isoform is present in the fusion protein in patients? Three protein types were described in the founding paper and the transcripts are splice-variable.
- Does the long disordered N-terminal extension of SEPTIN9, mostly retained in the fusion, contribute anything — for example condensate behaviour — independent of the G domain?
- E1established 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, with protein names including 'MLL septin-like fusion protein MSF-A' and gene synonyms KIAA0991, MSF and SEPT9; it has a septin-type G domain at aa 295-567 with G1 at 305-312, G3 at 362-365 and G4 at 444-447, disordered regions at 1-49, 62-108 and 134-268, GTP-binding sites at 305-312, 339, 365, 445-453, 501 and 516, and no annotated coiled coil.
UniProtKB Q9UHD8 (SEPT9_HUMAN) Septin-9 / MLL septin-like fusion protein MSF-A, Homo sapiens, 586 aa · UniProt Consortium · UniProtKB 2026What this evidence cannot show
Canonical isoform only; SEPTIN9 has multiple N-terminally distinct isoforms, and the 1999 founding paper already described three protein types of 568, 594 and 574 aa. The absence of a coiled-coil annotation reflects prediction and curation, not an experiment.
Evidence that argues against this (1)
- Absence of a coiled coil need not imply absence of oligomerisation: native septin complexes assemble through the nucleotide-binding G interface and the N/C extensions, and the predicted coiled coils of SEPT2/6/7 were shown not to be involved in complex or filament formation at all. Sirajuddin M et al. 2007
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Myeloid immortalisation by MLL-SEPT6 requires the coiled coil as well as the GTP-binding domain, and a synthetic 4-OHT-inducible dimerisation module substituted for the coiled coil did not rescue it.
Dimerization of MLL fusion proteins and FLT3 activation synergize to induce multiple-lineage leukemogenesis. · Ono R, Nakajima H, Ozaki K, et al. · J Clin Invest 2005What this evidence cannot show
Established for SEPTIN6 only; the transposition to SEPTIN9 made in this entry is an inference and is flagged as an open question, not a finding.
Evidence that argues against this (1)
- The coiled-coil requirement conflicts with the way native septins assemble, where the predicted coiled coils are dispensable, so it may be specific to the MLL-SEPT6 construct rather than a property of septin fusions as a class. Sirajuddin M et al. 2007
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
SEPT9 expression is altered in ovarian neoplasia.
Altered expression of the septin gene, SEPT9, in ovarian neoplasia. · Burrows JF, Chanduloy S, McIlhatton MA, et al. · J Pathol 2003What this evidence cannot show
Epithelial tumours, expression-level observation. Read as abstract-level evidence that SEPTIN9 has a cancer biology outside leukaemia, not as mechanism.
Evidence that argues against this (1)
- This literature does not transfer to the fusion: in haematopoietic lines SEPTIN9 shows no selective dependency in KMT2A-rearranged leukaemia (DepMap rank 13,148 of 17,728), so an epithelial oncogenic role gives no reason to expect the wild-type gene to matter here. Broad Institute DepMap; Cellosaurus; Open Targets Platform; scan run by this project 2026
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Septin 9 isoforms promote tumorigenesis in mammary epithelial cells by increasing migration and ECM degradation through metalloproteinase secretion at focal adhesions.
Septin 9 isoforms promote tumorigenesis in mammary epithelial cells by increasing migration and ECM degradation through metalloproteinase secretion at focal adhesions. · Marcus J, Bejerano-Sagie M, Patterson N, et al. · Oncogene 2019What this evidence cannot show
Mammary epithelial cell models and over-expression of specific isoforms. No haematopoietic data and no fusion protein.
Evidence that argues against this (1)
- The mechanism described is cytoplasmic and adhesion-based (focal adhesions, metalloproteinase secretion) and is therefore exactly the septin function the KMT2A fusion cannot use, because the fusion protein is relocalised to nuclear speckles by the KMT2A N-terminus. Ono R et al. 2005
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In a genome-wide DepMap CRISPR (Chronos) scan, SEPTIN9 had a mean gene effect of -0.172 across 1090 lines, a KMT2A-rearranged group mean of -0.092 versus -0.125 in other leukaemia lines, and ranked 13,148 of 17,728 for KMT2A selectivity (74.2nd percentile) — no selective requirement in KMT2A-rearranged leukaemia.
DepMap CRISPR (Chronos) genome-wide KMT2A-selectivity scan of 17,728 protein-coding genes, 11 Cellosaurus-annotated KMT2A-rearranged leukaemia lines vs other leukaemia lines (analyses/depmap-kmt2a-selective-scan-full/results.json, all_results; septin/ABI1/MLLT11 rows looked up for this beat) · Broad Institute DepMap; Cellosaurus; Open Targets Platform; scan run by this project · Database and derived analysis 2026What this evidence cannot show
No KMT2A-SEPTIN9 cell line exists in DepMap, so this is wild-type SEPTIN9 in leukaemias driven by other partners. Eleven KMT2A-rearranged lines is a small group.
Evidence that argues against this (1)
- A pooled CRISPR score for one septin cannot detect a requirement buffered by the other twelve septins, the same redundancy that made the Sept6 knockout mouse phenotypically normal. Ono R et al. 2005
- E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The MLL partner in a t(11;17)(q23;q25) therapy-related AML was cloned and named MSF (MLL septin-like fusion); it is a septin family member with a conserved GTP-binding domain, cytoplasmic localisation and interaction with cytoskeletal filaments, and is highly homologous to hCDCrel-1.
MSF (MLL septin-like fusion), a fusion partner gene of MLL, in a therapy-related acute myeloid leukemia with a t(11;17)(q23;q25). · Osaka M, Rowley JD, Zeleznik-Le NJ · Proc Natl Acad Sci U S A 1999What this evidence cannot show
Single index case; the paper defines the gene, not the mechanism.
Evidence that argues against this (1)
- The same gene was named AF17q25 in an independent contemporaneous report, so the 'MSF' name is one of two competing primary designations and neither is the current HGNC symbol; treating MSF as the gene's identity produces incomplete literature retrieval. Taki T et al. 1999
- E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
AF17q25, a putative septin family gene on 17q25 encoding at least three protein types (568, 594 and 574 aa), was identified fused to MLL in one de novo and one therapy-related AML with t(11;17)(q23;q25).
AF17q25, a putative septin family gene, fuses the MLL gene in acute myeloid leukemia with t(11;17)(q23;q25). · Taki T, Ohnishi H, Shinohara K, et al. · Cancer Res 1999What this evidence cannot show
Two patients. No DOI is registered for this 1999 Cancer Research paper; the PMID is the only stable identifier and was fetched from NCBI for this entry.
Evidence that argues against this (1)
- The authors' proposal that AF17q25 and hCDCrel might define a septin subfamily specifically involved in 11q23 leukaemia is not supported by later data: KMT2A partners now span at least five septin genes from different septin groups. Cerveira N et al. 2011
- E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
High-resolution crystal structures of the isolated human septin-9 GTPase domain exist in three nucleotide states: with GDP at 2.04 A (PDB 5CYO), with GTP-gamma-S at 2.89 A (5CYP) and apo at 2.73 A (4YQF).
RCSB PDB entries 5CYO, 5CYP and 4YQF — human septin 9 GTPase domain structures · RCSB Protein Data Bank · RCSB PDB entry API 2026What this evidence cannot show
Titles and resolutions read back from the RCSB entry API on 2026-09-19. These are isolated G domains; no structure contains the SEPTIN9 N-terminal extension, and none contains any part of KMT2A.
Evidence that argues against this (1)
- A well-characterised G domain does not make the fusion tractable: the transforming element of a KMT2A-septin fusion is a self-association interface that has never been mapped for SEPTIN9, and no structure of any KMT2A fusion protein exists. Ono R et al. 2005
- E9established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Mammalian septin filaments assemble through an extended G domain forming a bipolar building block, via conserved interactions between adjacent nucleotide-binding sites and via the N- and C-terminal extensions, with the predicted coiled coils shown not to be involved in or required for complex or filament formation.
Structural insight into filament formation by mammalian septins. · Sirajuddin M, Farkasovsky M, Hauer F, et al. · Nature 2007What this evidence cannot show
Structural and EM work on SEPT2/6/7, not on SEPTIN9-containing complexes.
Evidence that argues against this (1)
- The leukaemic fusion does not obey this rule: deleting the SEPT6 coiled coil abolished both self-association of MLL-SEPT6 and myeloid immortalisation, so fusion assembly and native filament assembly are not the same process. Ono R et al. 2005
SEPTIN9 G domain retained; no coiled coil on the canonical isoform
- ← fuses withKMT2A / MLL1 — the gene that breaks
t(11;17)(q23;q25)
- fuses with →KMT2A / MLL1 — the gene that breaks
KMT2A::SEPTIN9, t(11;17)(q23;q25)
both KMT2A-partner septins; SEPTIN9 has no annotated coiled coil, SEPTIN6 does
G domain retained in the fusion