U2AF1 - the 3' splice-site recognition factor, a speckle protein whose BioID neighbourhood contains LEDGF
also called U2 small nuclear RNA auxiliary factor 1 · U2AF35 · U2AFBP · RNU2AF1 · Splicing factor U2AF 35 kDa subunit
U2AF1 is one of the two proteins that finds the end of an intron so the splicing machine knows where to cut. In a 2024 map of RNA-handling proteins in leukaemia cells, LEDGF - the chromatin tether this project cares about - turned up in U2AF1's neighbourhood, which is one row and no more than that.
The same thing, for a scientist
U2AF1 (HGNC:12453, ENSG00000160201, UniProt Q01081, 240 aa, 21q22.3) is the small subunit of the U2 auxiliary factor, required for 3' splice-site recognition; it mediates the protein-protein and protein-RNA contacts that recruit U2 snRNP to the branch point and is localised experimentally to the **nucleus** and **nuclear speckle** [E3]. It enters this project's core-complex interactome on one curated row: IntAct IM-30059, **MI:1314 BioID2**, MI:2364 proximity, MI:1060 spoke expansion, **U2AF1 bait / PSIP1 prey**, MI-score 0.27, host K562 [E1][E2]. **This is a weak row and this entry says so plainly.** U2AF1 earns an entry not because the row is strong but because of what it is a member of: PSIP1 is recovered as prey by a large number of RBP baits in this deposition, and the splicing-factor class is the biggest single block among the unwritten partners of this project's baits - U2AF1, U2AF2, SRSF4, SRSF5, SRSF6, SF3B4 and EFTUD2 all sit in the same uncovered list. **Writing one of them as a real entry, honestly, is more useful than a summary entry covering all seven, which the standing granularity rule forbids.**
What the interaction record actually is
One IntAct row, IM-30059: MI:1314 BioID2, MI:2364 proximity, MI:1060 spoke expansion, U2AF1 bait (MI:0496) / PSIP1 prey (MI:0498), MI-score 0.27, host taxid:9606(Homo sapiens K562 myeloid leukemia cell), curator annotation Data file S6 2.
One row, one paper, one method. No co-IP, no reconstitution, no orthogonal validation. K562 carries no KMT2A rearrangement. Anyone citing this as a LEDGF-splicing link is over-reading it.
Why the class is interesting even though the row is not
LEDGF/PSIP1 reads H3K36me2/3 over the bodies of transcribed genes, and H3K36 methylation is one of the best-established chromatin inputs to co-transcriptional splicing - the mark is deposited behind elongating polymerase and is read by splicing-associated PWWP and chromodomain proteins. A LEDGF that is near U2AF1 in a nucleus is therefore exactly what the co-transcriptional splicing literature would predict, which cuts both ways: it makes the row plausible and it makes it uninformative, because any abundant gene-body chromatin protein would be near the splicing machinery.
The useful version of this question is not 'does LEDGF touch U2AF1'. It is whether KMT2A::AFF1 target genes show altered splicing, and whether that is a consequence of the fusion's elongation defect. This graph has an elongation beat and a set of entries on the fusion's effect on polymerase; splicing is the readout nobody has joined to them.
Subnuclear address - the one axis where this row is informative
U2AF1's UniProt record carries nucleus speckle (ECO:0000269) 3. Nuclear speckles are the storage and assembly compartment for splicing factors, and this project has a subnuclear beat and existing entries that turn on speckle residence. A speckle-resident bait recovering PSIP1 is a weak but real datum on whether LEDGF has a speckle pool.
Caution, recorded in this graph's own registry: speckle* is a contaminated search token - it matches speckle-tracking echocardiography and inflates any bounded query built on it. This entry makes no bounded absence claim about speckles and the token was not used as a search term.
Disease context, stated carefully
U2AF1 carries recurrent hotspot mutations in myelodysplastic syndrome and secondary AML, and splicing-factor mutations are an established class of myeloid driver. This entry does not cite a primary source for that and therefore does not rest on it, and it is in any case a statement about a different disease: paediatric KMT2A-rearranged ALL is not a splicing-factor-mutant disease. No U2AF1 mutation has been reported in KMT2A::AFF1 infant ALL as far as this beat checked, and this beat did not run a bounded query on that question, so the absence is undocumented rather than established.
| hgnc ID | HGNC:12453 |
|---|---|
| symbol | U2AF1 |
| ensembl gene ID | ENSG00000160201 |
| uniprot | Q01081 |
| length AA | 240 |
| chromosomal location | 21q22.3 |
| identifier provenance | HGNC REST fetch/symbol/U2AF1 and UniProtKB Q01081 retrieved live 2026-09-23. The alias U2AF35 was checked against the corpus and matched only gene-zranb2, which is not a U2AF1 entry - no duplicate exists. |
| subnuclear location | nuclear speckle - an experimentally coded call, and the reason this row is worth anything |
| cell type of the observation | K562 (BCR::ABL1 CML) - not KMT2A-rearranged, not B-lineage, not infant |
| epigenetic role | None directly. The link runs the other way: H3K36me2/3, the mark PSIP1 reads, is an established chromatin input to co-transcriptional splicing. |
| shown for fusion | NO - nobody has looked in any fusion. The row is wild-type PSIP1 in K562. |
| evidence class | single publication, single method, spoke-expanded BioID prey row, no orthogonal validation |
prev and alias symbols
[ "U2AFBP", "U2AF35", "RNU2AF1", "RN" ]
interaction with project baits
[
{
"bait": "PSIP1",
"bait_uniprot": "O75475",
"partner_experimental_role": "bait",
"scaffold_experimental_role": "prey",
"method": "MI:1314 proximity-dependent biotin identification (BioID2)",
"interaction_type": "MI:2364 proximity",
"expansion": "MI:1060 spoke",
"mi_score": 0.27,
"imex": "IM-30059",
"pmid": "39251607",
"host_cell": "K562 (BCR::ABL1 CML, no KMT2A rearrangement)",
"n_rows": 1,
"n_publications": 1
}
]subcellular location uniprot
[ "Nucleus (ECO:0000269)", "Nucleus speckle (ECO:0000269)" ]
sibling splicing partners still unwritten
[ "U2AF2", "SRSF4", "SRSF5", "SRSF6", "SF3B4", "EFTUD2" ]
- Splicing-factor inhibitors exist and are toxic; a rationale built on a single proximity row would not survive the first toxicity review, and this entry should not be cited as one.
- If chromatin-coupled splicing is genuinely downstream of the fusion, then correcting the fusion corrects the splicing, and the splicing machinery is not the target - the fusion is.
- Do KMT2A::AFF1 target genes show altered splicing relative to their expression change, and is that a consequence of the fusion's elongation phenotype? This is the version of the splicing question worth this project's time, and it needs RNA-seq that already exists rather than a new pull-down.
- Does LEDGF have a nuclear-speckle pool in a haematopoietic cell, by imaging rather than by proximity labelling?
- Are the seven unwritten splicing-factor partners of these baits one observation - an abundant nuclear class recovered by every proximity experiment - or seven? A single length- and abundance-matched analysis over the IM-30059 deposition would answer it for all seven at once and would justify writing or declining them as a class.
- Has any U2AF1 or other splicing-factor mutation ever been reported in KMT2A-rearranged infant ALL? This beat did not run the bounded query and the absence is undocumented.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Khoroshkin et al. generated in vivo proximity-dependent biotinylation datasets for 50 human RNA-binding proteins in K562 cells and deposited the curated bona fide pair list to IMEx as IM-30059; U2AF1 is one of the 50 baits and PSIP1 appears in its prey list.
Systematic identification of post-transcriptional regulatory modules. · Khoroshkin M, Buyan A, Dodel M, et al. · Nat Commun 2024What this evidence cannot show
Abstract read here. This graph's ana-rbp-bioid2-proximity-map-k562 records that the full text (PMC11385195) was read in an earlier cycle and that the elongation factors appear nowhere in the body - these pairs are reachable only through the supplementary data and the IMEx deposition. K562 is a BCR::ABL1 chronic myeloid leukaemia line with NO KMT2A rearrangement. The authors note that RBP over-expression, BioID2 fusion and the use of a single isoform per protein may bias the data.
Evidence that argues against this (1)
- No SEC component, no KMT2A-complex component and no fusion protein is a BAIT in this deposition; the scaffold proteins appear only as prey among 2,682 preys, and the median prey is recovered by five of the fifty baits. A single prey row here is weak by the deposition's own internal scale. source 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
IntAct serves one curated human U2AF1-PSIP1 row: MI:1314 proximity-dependent biotin identification, MI:2364 proximity, MI:1060 spoke expansion, U2AF1 experimental role MI:0496 bait and PSIP1 MI:0498 prey, MI-score 0.27, host taxid:9606 Homo sapiens K562 myeloid leukemia cell, IMEx IM-30059, publication PMID 39251607, curator annotation 'Data file S6'.
IntAct molecular interaction database (EMBL-EBI), PSICQUIC tab27 record for id:Q01081 AND id:O75475 2026What this evidence cannot show
Retrieved live 2026-09-23. Rule 93: PSICQUIC is a floor relative to IntAct REST and cannot reach processed-chain identifiers. MI-score 0.27 is the deposition-wide value for this screen and carries no pair-specific information.
Evidence that argues against this (1)
- A single spoke-expanded row in a 19,403-row, 2,682-prey deposition is at or below the noise floor of that dataset by its own internal scale - the median prey there is recovered by five of the fifty baits, and PSIP1's recovery by one bait places it in the least-supported tier. source 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProt Q01081 is 240 aa and records U2AF1 as playing a critical role in constitutive and enhancer-dependent splicing by mediating the protein-protein and protein-RNA interactions required for accurate 3' splice-site selection and recruiting U2 snRNP to the branch point, with experimental (ECO:0000269) subcellular locations nucleus and nucleus speckle.
What this evidence cannot show
Fetched live 2026-09-23. Rule 91 applies to the ECO codes: the curator read an experiment in a cited paper; that paper was not opened and may not be the founding observation for the speckle call.
Evidence that argues against this (1)
- Nuclear speckles are where splicing factors are STORED, not only where they act, so speckle residence of the bait does not establish that a prey recovered near it is engaged in splicing. The same caution applies in reverse to reading a LEDGF speckle pool out of this row. source 2026
- co-occurs with →PSIP1 / LEDGF p75 — the tether with the best therapeutic index in the whole interactome
U2AF1 bait / PSIP1 prey, one spoke-expanded BioID2 row, IntAct IM-30059, K562 - proximity only
one of the 50 RBP baits in this deposition
H3K36me2/3, the mark PSIP1 reads, is an established chromatin input to co-transcriptional splicing - the mechanistic reason this row is plausible and also the reason it is uninformative