Oncofusion

start / graph / gene-sf3a3

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q84-pan-fusion-core

SF3A3 / SF3a60 (Q12874) - a U2 snRNP splicing subunit in the pan-MLL-fusion core, and the paper's own named example of a NON-selective hit

also called SF3A3 · SF3a60 · SAP61 · splicing factor 3A subunit 3

SF3A3 is one of three subunits of SF3a, a part of the machine that recognises introns and splices them out of messenger RNA. It co-purifies strongly with every MLL fusion tested, and the same paper used it as an example of a hit that kills leukaemia cells whether or not they carry a fusion - so it is a real interactor and an explicit non-target.

The same thing, for a scientist

SF3A3 (Q12874, 501 aa; SF3a60) is one of the three subunits of the SF3a heterotrimer within the 17S U2 snRNP, essential for prespliceosome assembly. In Skucha 2018 it co-purified with all seven MLL fusions with one of the better index-fusion calls in the whole beat - MLL-AF4 (= KMT2A::AFF1) p = 5.95e-4 with 15|12 spectra - against a negative control of 0|3. What makes this entry unusual is that the paper itself carries it forward and then rules it out: in the secondary screen of the 40 highest-confidence candidates, SF3A3 is named, verbatim, among 'MLL-interactors with known essential functions' whose knockdown 'led to growth inhibition in both MLL-fusion cells and MLL-wild-type leukemia cells', in contrast to MEN1, which spared the MLL-wild-type lines. **This is a measured negative for fusion selectivity, published in the source paper, and it is the only one of the twelve proteins in this beat for which such a result exists.** A strong interactor that is not a selective dependency is exactly the kind of entry a saturation map needs, and exactly the kind a target list must not inherit.

Go deeper
The Skucha row

7 of 7 fusions. MLL-AF1p 1.34e-4 (16|13), MLL-AF4 5.95e-4 (15|12), MLL-AF9 0.00105 (22|24), MLL-CBP 0.0378 (7|31), MLL-EEN 0.00214 (27|27), MLL-ENL 6.18e-4 (21|24), MLL-GAS7 0.0157 (20|18); negative control 0|3 1. The index-fusion call is the second strongest of the twelve proteins in this beat after MCM2, on 15 and 12 spectra - a real signal by the standards of this table. The negative control is not clean.

The paper's own negative result, quoted

From the Results of Skucha 2018, on the re-screen of the 40 highest-confidence candidates in MLL-AF9 MOLM-13 against MLL-wild-type K562 and HL-60: "As expected, knockdown of MLL-interactors with known essential functions, such as RFC1, SF3A3, or CCT3, led to growth inhibition in both MLL-fusion cells and MLL-wild-type leukemia cells. In contrast, depletion of the known MLL-interaction partner MEN1 selectively inhibited growth of MLL-fusion cells while sparing MLL-wild-type cells, proving the validity of the chosen strategy." 2 So the authors used SF3A3 as a control for non-selectivity. Any downstream use of this interactome as a target list has to carry that forward, and this graph's target-list audit is the place it belongs.

What SF3A3 does, and where it does it

RNAi depletion of each of the three human SF3a subunits individually is lethal to human cells and produces a general block in splicing, showing SF3a is a constitutive splicing factor in vivo; the 17S U2 snRNP cannot be detected in SF3a-depleted extracts, and U2 snRNA and U2-B'' are significantly reduced in nuclear speckles and the nucleoplasm while remaining present in Cajal bodies 3. The crystal structure of the SF3a core (solved in yeast, with subunits Prp9, Prp11 and Prp21) shows a bifurcated three-subunit assembly and places the major U2 snRNA stem-loop IIa binding site on the SF3A3 orthologue Prp9 4. The speckle localisation is the useful part for this graph: it is a named subnuclear compartment, and this graph already carries localization-nuclear-speckles-kmt2a-r and a speckle/paraspeckle hypothesis.

Facts and figures
uniprotQ12874
length AA501
accession resolutionResolved live 2026-09-21: gene_exact:SF3A3 AND organism_id:9606 AND reviewed:true returns exactly one row, gene_primary SF3A3, Q12874, SF3A3_HUMAN, 501 aa, synonym SAP61. Matches Skucha's table.
localizationNucleus
subnuclear localizationNUCLEAR SPECKLES (interchromatin granule clusters), with functional coupling to Cajal bodies - U2 snRNP components are lost from speckles but retained in Cajal bodies when SF3a is depleted.
localization evidenceUniProt Q12874, fetched 2026-09-21: 'Nucleus speckle' with ECO:0000269 experimental evidence, plus 'Nucleus' with four further ECO:0000269 citations. One of only two proteins in this beat (with PML and NOLC1) carrying an experimentally coded NAMED subnuclear compartment.
shown for fusionKMT2A::AFF1 (MLL-AF4) by co-purification at p = 5.95e-4, plus all six other Skucha baits. Functionally tested only against MLL-AF9 MOLM-13, where the effect was NOT fusion-selective. NO independent experiment ties SF3A3 to a KMT2A fusion.
selectivityNEGATIVE, measured in the source paper. SF3A3 should not appear on any target list derived from this interactome.
graph contextSF3A3 appeared in exactly one node file in this graph before this entry (grep -rl, 2026-09-21). Related existing entries: localization-nuclear-speckles-kmt2a-r and hyp-nuclear-speckle-paraspeckle-kmt2a-r.
skucha 2018
{
  "bait": "MLL-AF4 (= KMT2A::AFF1)",
  "p_value": 0.000595,
  "spectral_counts": "15|12",
  "n_fusions_of_seven": 7,
  "negative_control_counts": "0|3",
  "per_bait": {
    "MLL-AF1p": "0.000134 (16|13)",
    "MLL-AF4": "0.000595 (15|12)",
    "MLL-AF9": "0.00105 (22|24)",
    "MLL-CBP": "0.0378 (7|31)",
    "MLL-EEN": "0.00214 (27|27)",
    "MLL-ENL": "0.000618 (21|24)",
    "MLL-GAS7": "0.0157 (20|18)"
  },
  "in_paper_functional_result": "Named in the Results as an interactor with a KNOWN ESSENTIAL FUNCTION whose knockdown inhibited growth of both MLL-fusion (MOLM-13) and MLL-wild-type (K562, HL-60) leukaemia cells - the paper's own non-selectivity control, contrasted with MEN1.",
  "source": "analyses/skucha-mll-fusion-interactome/results.json, re-derived from the cached MOESM4 xlsx on 2026-09-21; the functional sentence read from the PMC5959866 full text on 2026-09-21"
}
cell types and states
[
  "Jurkat (human T-ALL) - AP-MS host, nuclear lysate",
  "MOLM-13 (MLL-AF9) - shRNA knockdown, growth inhibition (NOT selective)",
  "K562 and HL-60 (MLL-wild-type) - shRNA knockdown, growth inhibition (the non-selectivity result)",
  "cultured human cells - SF3a subunit RNAi, splicing block, speckle reorganisation",
  "no KMT2A::AFF1 cell of any kind"
]
bounded literature
{
  "query": "SF3A3[tiab] AND (KMT2A[tiab] OR MLL[tiab] OR \"MLL-AF4\"[tiab] OR \"MLL-AF9\"[tiab])",
  "querytranslation": "\"SF3A3\"[Title/Abstract] AND (\"KMT2A\"[Title/Abstract] OR \"MLL\"[Title/Abstract] OR \"MLL-AF4\"[Title/Abstract] OR \"MLL-AF9\"[Title/Abstract])",
  "count": 0,
  "date": "2026-09-21",
  "tool": "NCBI esearch db=pubmed",
  "names_tried": [
    "SF3A3"
  ],
  "positive_control": "the identical query shape with MEN1 returns 85",
  "note": "Zero in titles and abstracts even though the Skucha paper itself names SF3A3 in its Results - a clean illustration of the standing rule that a PubMed zero is bounded by abstracts, not by papers.",
  "bound": "Titles and abstracts only; supporting figures and body text not searched. SF3a60 and SAP61 were not run as separate strings."
}
functionally screened in the source paper
{
  "_added": "king, cycle 48, 2026-09-21, after Q86 flagged that cycle-47 entries understate the record. Verified by me directly from the supplementary file, not from the paper's prose.",
  "screened": true,
  "library_symbol": "SF3A3",
  "shrnas_for_this_gene": 6,
  "design": "Supplementary Table 6 / MOESM5 is 768 data rows, 128 unique gene symbols, EXACTLY 6 shRNAs each -- counted with openpyxl on 2026-09-21. Negative-selection RNAi screen in MLL-AF9-expressing MOLM-13.",
  "what_this_DOES_establish": "Every one of the 128 conserved MLL-fusion interactors, this protein included, was functionally tested against an MLL fusion. An entry saying 'nothing in the paper validates this prey' is true of BINDING and understates the record.",
  "what_this_does_NOT_establish": "A per-gene result. Scores are presented as heatmaps (Fig. 2c/2d) and no per-gene depletion value is tabulated, so for any individual gene the outcome is measured-not-reported. It is also MLL-AF9 MOLM-13, a myeloid line, not the index fusion and not B-lineage.",
  "absence_class_for_the_per_gene_score": "measured-not-reported"
}
How this could bite us back
  • SF3A3 is constitutively essential for splicing in every human cell tested; targeting it has no therapeutic window in an infant and the source paper says so.
  • Splicing modulators already in the clinic act on SF3B1, a different subcomplex; borrowing that chemistry for an SF3a-based idea would be targeting a different particle with the wrong tool.
  • Because SF3A3 is a plausible co-purification contaminant AND a genuine essential gene, it is exactly the kind of entry that inflates a target list without adding a target. It is recorded here so that it cannot.
Still open
  1. Are other U2 snRNP or SF3a/SF3b subunits in the Skucha 960-protein network? If the whole particle is there, the row is a spliceosome co-purification and should be read as such rather than as a protein-protein contact.
  2. Does KMT2A::AFF1 occupy nuclear speckles, and if so is the SF3A3 co-purification simply compartment co-residence? This graph already has a speckle hypothesis and this is a concrete way to test it.
  3. The paper's non-selectivity result is against an MLL-AF9 index in myeloid cells. Does it hold in a B-lineage t(4;11) comparison, where pan-essentiality may still leave a selectivity window?
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    SF3A3 co-purified with all seven Strep-HA-tagged MLL fusions in isogenic Jurkat cells, with the MLL-AF4 (= KMT2A::AFF1) call at p = 5.95e-4 and 15|12 spectral counts, against a negative control of 0|3.

    MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity. · Skucha A, Ebner J, Schmoellerl J, Roth M, Eder T, Cesar-Razquin A, Stukalov A, Vittori S, Muhar M, Lu B, Aichinger M, Jude J, Mueller AC, Gyorffy B, Vakoc CR, Valent P, Bennett KL, Zuber J, Superti-Furga G, Grebien F · Nature Communications 2018
    PMID 29777171DOI
    What this evidence cannot show

    SYSTEM: the AP-MS host is JURKAT, a T-ALL line, whereas KMT2A::AFF1 disease is B-lineage infant leukaemia, so a partner that needs a B-cell or infant proteome cannot appear here and a T-lineage partner can appear spuriously. The bait is a Strep-HA-tagged, SINGLE-COPY, doxycycline-INDUCED transgene, not the endogenous fusion. Prey was recovered from bulk nuclear lysate, so co-purification is not evidence of a direct contact. INCLUSION RULE: the paper kept the top 300 interactions per bait by combined p-value rather than applying a p threshold (Methods, verbatim: "The 300 most significant interactions per bait were retained"). Measured by this beat by reading the cached Supplementary Table 5 xlsx directly on 2026-09-21: 545 of the 2,049 per-bait p-values in the table (26.6%) exceed 0.05, to a maximum of 0.175. This graph's own king analysis ana-the-index-fusion-interactome-is-not-distinctive-and-two-earlier-answers-were-artifacts found the MLL-AF4 column reaches deepest into the weak tail of the seven, so an MLL-AF4 p-value is systematically weaker than the same protein's p-value under another bait and should not be read as a weaker interaction. The negative control is not clean (0|3), and spliceosomal proteins are among the commonest contaminants of nuclear-lysate affinity purifications. Values re-read from the cached xlsx on 2026-09-21.

    Evidence that argues against this (1)
    • IntAct has never curated this dataset (pubid:29777171 returns 0 rows against a working positive control), so no independent curator has re-examined these calls, and the raw data at ProteomeXchange PXD009338 have not been reanalysed here. Nothing in the paper validates this particular prey by an orthogonal assay: of the twelve proteins in this beat, only SETD2 is mentioned anywhere in the paper's body text (counted by regex over the PMC full text, 2026-09-21; SF3A3 appears once, as a named non-selective control). ProteomeXchange/PRIDE 2018
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the same study, SF3A3 knockdown inhibited the growth of both MLL-fusion-expressing MOLM-13 cells and the MLL-wild-type leukaemia lines K562 and HL-60 - the authors' own named example of an interactor with a known essential function and no fusion selectivity, in explicit contrast to MEN1, which spared the MLL-wild-type lines.

    MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity. · Skucha A, Ebner J, Schmoellerl J, Roth M, Eder T, Cesar-Razquin A, Stukalov A, Vittori S, Muhar M, Lu B, Aichinger M, Jude J, Mueller AC, Gyorffy B, Vakoc CR, Valent P, Bennett KL, Zuber J, Superti-Furga G, Grebien F · Nature Communications 2018
    PMID 29777171DOI
    What this evidence cannot show

    shRNA in human MOLM-13 (MLL-AF9), K562 and HL-60 - NOT in a KMT2A::AFF1 line, so the non-selectivity is demonstrated against an MLL-AF9 index rather than the index fusion of this project. Full text read from PMC5959866 on 2026-09-21; the sentence is quoted in the section above rather than paraphrased. A pan-essential gene will kill every line, so this result bounds selectivity and says nothing about whether the physical interaction is real.

    Evidence that argues against this (1)
    • Non-selectivity in a myeloid MLL-AF9 comparison does not formally exclude a KMT2A::AFF1-selective effect in B-lineage cells, and this graph's own DepMap work records that floor-compressed pan-essential genes have almost no dynamic range in which selectivity could appear. Tanackovic G et al. 2005
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Each of the three human SF3a subunits is individually essential for cell viability and its knockdown produces a general block in splicing; in SF3a-depleted cells the 17S U2 snRNP is undetectable and U2 snRNA and U2-B'' are significantly reduced in nuclear speckles and the nucleoplasm while remaining present in Cajal bodies.

    PMID 15647371DOI
    What this evidence cannot show

    RNAi in cultured human cells, not a haematopoietic or fusion system. Read from the abstract. The essentiality is the reason SF3A3 cannot be a selective target and is recorded here as such.

    Evidence that argues against this (1)
    • Constitutive-splicing essentiality has not stopped the field from drugging SF3B1 with splicing modulators in myeloid disease; 'essential' is an argument about therapeutic index, not about undruggability. Lin PC et al. 2012
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The SF3a complex is an evolutionarily conserved heterotrimer whose core assembles as a bifurcated three-subunit structure, with the SF3A3 orthologue harbouring the major binding site for U2 snRNA stem-loop IIa.

    PMID 22314233DOI
    What this evidence cannot show

    The crystal structure is of the YEAST core complex (Prp9, Prp11, Prp21); SF3A3 is the human orthologue of Prp9 and the residue-level detail does not transfer. Read from the abstract. Cited for architecture only.

    Evidence that argues against this (1)
    • A yeast core structure cannot establish the human subunit boundaries, and this graph's standing rule on residue numbering warns against reading spans off a deposit in another organism. Tanackovic G et al. 2005
Connected to