Oncofusion

start / graph / gene-rpa3

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q192-the-rpa-heterotrimer

RPA3 (RPA14) - the 121-residue subunit of replication protein A, which has no annotated domain, no HPA localisation call and no connection to this disease at all

also called RPA14 · REPA3 · Replication protein A 14 kDa subunit

RPA3 is the smallest of the three proteins in the clamp that coats exposed single-stranded DNA - so small that databases record almost nothing about it beyond its sequence. It exists to complete the three-part core; nobody has ever connected it to leukaemia fusions.

The same thing, for a scientist

RPA3 (UniProt **P35244**, RFA3_HUMAN, **121 aa**, HGNC:10291) is the smallest RPA subunit and contributes DBD-E to the trimerisation core, where its OB fold packs against RPA1's DBD-C and RPA2's DBD-D and trimerisation is mediated by three parallel C-terminal helices [E1, E3]. **UniProt's feature table annotates no domain and no DNA-binding region for it at all** - the only feature is N-acetylvaline at position 2 - so this entry deliberately records no domain coordinates [E1]. Localisation: **Nucleus** on two ECO:0000269 attributions (PubMed 17596542, 9430682); RPA3 interacts with BRIP1/FANCJ *via the RPA1 subunit* and the two colocalise in nuclear foci after DNA damage [E1]. **The Human Protein Atlas has no subcellular location for RPA3 at all** - reliability 'Approved', location field null - which is recorded here as a measured absence of a localisation call rather than as a nuclear assumption [E2]. `RPA3[tiab] AND MLL[tiab]` and `RPA3[tiab] AND KMT2A[tiab]` both return 0, and RPA3 is absent from all seven baits of the only MLL-fusion AP-MS network [E4]. **This entry exists for completeness of the complex and to make the subunit-coherence argument in the Skucha adjudication checkable; it makes no disease claim.**

Go deeper
The honest content of this entry is how little there is

RPA3 is the clearest case in this beat of an entry whose value is in bounding the ignorance rather than in reporting a finding.

  • No annotated domain. UniProt P35244's feature table contains exactly one feature, an N-terminal acetylation 1. The DBD-E designation comes from the structural literature, where the entire RPA14 subunit was crystallised as part of the 2.8 Angstrom human trimerisation core 3 - but no residue range is quoted here, because the paper's numbering was not cross-checked against a fetched coordinate file.
  • No HPA localisation. The HPA record exists (ENSG00000106399, reliability 'Approved') and its subcellular-location field is null 2. That is different from 'HPA says nucleus'. A nuclear call for RPA3 rests on UniProt's two coded attributions and on the complex, not on imaging of this subunit.
  • No disease connection of any kind. Zero literature records against MLL or KMT2A, absent from the AP-MS network, and it appears in only 5 node files in this corpus, all of them either the Q180 adjudication or a Fanconi-module analysis 4.
  • One thing it is good for. RPA3 is the second half of the subunit-coherence test used to decline the Skucha RPA2 row. Having its accession and length written down means the test can be re-scored. At 121 aa, RPA3 offers roughly 0.45x RPA2's tryptic peptide space, so RPA3's absence from a shallow AP-MS list is far weaker evidence than RPA1's - a distinction the decline does not draw.
The alternative complex, which is a fact about composition worth having on file

UniProt records that RPA1 and RPA3 are also components of aRPA, an alternative trimeric complex in which RPA1 and RPA3 associate with RPA4 instead of RPA2 (PubMed 19116208, 7760808) 1.

That matters for exactly one reason and it is not the one a reader might reach for. It does not support a free-RPA2 pool - the substitutable subunit in aRPA is RPA2, so aRPA is an RPA1/RPA3 species lacking RPA2. What it does establish is that 'obligate heterotrimer' is a simplification: the complex has at least two compositions, and any argument of the form 'subunit X must co-purify with subunit Y' inherits that imprecision.

RPA4 has no entry in this corpus and was not investigated by this beat. It is recorded here as a named, un-followed thread rather than left invisible.

Facts and figures
uniprotP35244
length AA121
sequence md5B538C1B4F1CF8092ABAB50AE595D5861
gene symbolRPA3
hgnc IDHGNC:10291
protein nameReplication protein A 14 kDa subunit
uniprot resolutionFetched by accession 2026-09-24; primary gene name read back as RPA3, uniProtkbId RFA3_HUMAN. UniProt synonyms REPA3, RPA14. HGNC alias symbol REPA3 only; no previous symbol.
symbol collision checkNo row in data/symbol-collisions.json (220 rows, 2026-09-24). RPA3 is NOT an HGNC alias of POLR1C, unlike RPA1/POLR1A and RPA2/POLR1B - verified at rest.genenames.org the same day. The nearest hazard is RPA3-AS1 / RPA3OS, which are aliases of the separate gene UMAD1 (HPA search on 'RPA3' returns it), so an antisense-transcript name can return a protein-coding gene that is not RPA3.
domains noteDELIBERATELY EMPTY. UniProt P35244's feature table annotates no domain, no DNA-binding region and no zinc finger; the only feature is N-acetylvaline at position 2. RPA3 is structurally DBD-E and the whole subunit was crystallised in the human trimerisation core (PMID 11927569), but no residue range was confirmed against a fetched coordinate file, and this project's rule is that a domain with an unverified or null boundary is invisible to every figure and should be left out rather than guessed.
epigenetic readoutNONE, class never-measured. No histone mark is written, read or erased by RPA3, and no chromatin-mark readout has ever been reported for it in any cell in this disease.
shown for fusionNO - nobody has looked in any fusion, in any lineage. RPA3 is absent from all seven baits of the only MLL-fusion AP-MS dataset (PXD009338 / PMID 29777171), which is the strongest form this negative can take here, and it is bounded by that deposit's published top-300-per-bait cut rather than by its raw data.
fetched2026-09-24
measurements
count0.000count: 0count0.000count: 0length121.0length: 121
domains
[]
localization
{
  "subcellular": "Nucleus, on two ECO:0000269 attributions (UniProt P35244, PubMed 17596542 and 9430682). NO HPA subcellular-location call exists for RPA3 - the record is present with reliability 'Approved' and a null location field.",
  "subnuclear": "NOT MEASURED for RPA3 itself. Damage-induced nuclear foci are recorded on RPA3's UniProt record but explicitly as colocalisation with BRIP1/FANCJ mediated VIA RPA1, so they are a complex-level observation. No PML-body call, unlike RPA1 and RPA2.",
  "measured_in_a_blood_cell": false,
  "measured_in_a_kmt2a_rearranged_cell": false
}
cell type
{
  "answer": "NOT LINEAGE-SPECIFIC and nothing beyond that was found. No haematopoietic measurement, no marker association, no lineage requirement.",
  "haematopoietic_measurement": "none found"
}
cell state
{
  "state": "S-PHASE AND DAMAGE, inherited from the complex. No state-specific measurement of RPA3 alone was found, and that is recorded as an absence rather than filled in from RPA2's behaviour.",
  "kmt2a_relevance": "none documented"
}
essentiality
{
  "depmap_common_essential": null,
  "chronos_mean": null,
  "note": "NOT FETCHED. No number recorded rather than an approximate one."
}
measurements — the numbers behind the chart
[
  {
    "quantity": "length",
    "subject": "RPA3|P35244",
    "value": 121,
    "units": "aa",
    "n": 1,
    "universe": "UniProtKB canonical, fetched 2026-09-24, md5 B538C1B4F1CF8092ABAB50AE595D5861",
    "null": null,
    "note": "0.45x RPA2's 270 aa and 0.20x RPA1's 616 aa. This is why RPA3's absence from a shallow spectral-count list is much weaker evidence than RPA1's, a distinction the Q180 subunit-coherence argument does not draw."
  },
  {
    "quantity": "count",
    "subject": "annotated domains in the UniProt feature table for RPA3",
    "value": 0,
    "units": "features of type Domain/DNA binding/Zinc finger",
    "n": 1,
    "universe": "UniProtKB P35244 complete feature table, fetched 2026-09-24",
    "null": null,
    "note": "for comparison, the same fetch returns 1 DNA-binding OB fold plus 1 zinc finger for RPA1 and 1 OB fold plus 2 regions for RPA2"
  },
  {
    "quantity": "count",
    "subject": "papers naming RPA3 with MLL or KMT2A in title or abstract",
    "value": 0,
    "units": "records",
    "n": 2,
    "universe": "PubMed, RPA3[tiab] AND MLL[tiab] and RPA3[tiab] AND KMT2A[tiab], both querytranslations read back, 2026-09-24",
    "null": 197,
    "note": "null is RPA3[tiab] alone (197 records), the denominator that makes the zero interpretable"
  }
]
How this could bite us back
  • RPA3 is a core subunit of a complex required in every dividing cell, so it inherits the whole therapeutic-window problem of RPA with none of the compensating evidence: there is no disease-specific rationale for it whatsoever.
  • Because RPA3 is small and has no annotated domain, it is the subunit most likely to be silently dropped from a proteomic or structural analysis, and its absence from such an analysis should never be treated as biological information without a detectability control.
Still open
  1. Was RPA3 detected in PXD009338 below the published rank cut? At 121 aa it is the subunit least likely to be seen in a shallow list, so its absence carries much less information than RPA1's, and the raw deposit can settle both at once.
  2. RPA4 replaces RPA2 in the alternative aRPA complex with RPA1 and RPA3 (UniProt P35244, PubMed 19116208, 7760808) and has NO entry in this corpus. Is RPA4 expressed in haematopoietic cells, and does an aRPA-type complex exist in blast cells? This is a named, un-followed thread.
  3. Is there any human structure - not the Ustilago maydis full-length trimer - from which DBD-E boundaries for P35244 could be recorded with a PDB id? This entry leaves the domain array empty rather than guess.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    RPA3 is UniProt P35244 (RFA3_HUMAN), primary gene symbol RPA3, 121 aa, 'Replication protein A 14 kDa subunit', with no annotated domain, DNA-binding region or zinc finger in its feature table (the sole feature is N-acetylvaline at position 2); it is annotated to Nucleus on two ECO:0000269 attributions (PubMed 17596542, 9430682), is a component of the canonical RPA heterotrimer and also of the alternative aRPA complex in which RPA1 and RPA3 associate with RPA4 instead of RPA2, and interacts with BRIP1/FANCJ via the RPA1 subunit with which it colocalises in nuclear foci following DNA damage.

    What this evidence cannot show

    Resolved by accession with the primary gene name read back; sequence md5 B538C1B4F1CF8092ABAB50AE595D5861. The 'no annotated domain' statement is a statement about this UniProt release's feature table and not about the protein - RPA3 demonstrably has an OB fold, which is why the structural literature calls it DBD-E; the absence is in the annotation, and it is recorded because it is the reason this entry carries no coordinates. The BRIP1/FANCJ interaction is curated on RPA3's record but is explicitly stated to occur VIA RPA1, so it is not an RPA3 contact. Rule 91: the two coded localisation citations were not opened.

    Evidence that argues against this (1)
    • An empty feature table is the sort of absence that invites a later agent to 'fix' it by importing boundaries from a structure paper. Do not: the most-cited full-length RPA-ssDNA structure is a USTILAGO MAYDIS heterotrimer (PMID 23070815, MeSH 'Fungal Proteins', 'Ustilago') and its numbering does not transfer to the 121-residue human protein. If DBD-E boundaries are ever needed, they must come from a human structure with the PDB id stated. Fan J et al. 2012
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The Human Protein Atlas has a record for RPA3 (ENSG00000106399) with reliability 'Approved' and NO subcellular location value at all, in contrast to RPA1 (Nucleoplasm, Supported) and RPA2 (Nucleoplasm and Nuclear bodies, Supported) returned by the same API on the same day.

    What this evidence cannot show

    A null field in an API response is not the same as 'HPA looked and saw nothing'; it can mean the immunofluorescence experiment was never run, or was run and not released, or that the column was not populated for this record. This item asserts only what the response contains. The expected_absent_terms field records that the ABSENCE of a location string is the claim, so an auditor must not flag the missing term. Three queries on one day from one API; not re-fetched.

    Evidence that argues against this (1)
    • The natural inference - that RPA3's localisation is simply unmeasured - is not safe either. RPA3 is 121 aa and 14 kDa, which makes raising a specific antibody and calling a compartment genuinely hard, so a missing HPA call may reflect an assay limit rather than an absence of interest. Either way a reader should not upgrade 'nuclear' for RPA3 from the complex to the subunit without saying so, which is what this item exists to prevent. UniProt Consortium 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The 2.8 Angstrom structure of the human RPA trimerisation core comprises RPA70's DBD-C, RPA32's DBD-D and the entire RPA14 subunit, each built around a central OB fold flanked by a C-terminal helix, with trimerisation mediated by three parallel C-terminal helices; the core does not contribute to binding 10-nucleotide substrates but interacts with 24-nucleotide oligonucleotides, which is how RPA switches from its 8-10 nt to its ~30 nt binding mode.

    Structure of the RPA trimerization core and its role in the multistep DNA-binding mechanism of RPA. · Bochkareva E, Korolev S, Lees-Miller SP, Bochkarev A · EMBO Journal 21(7):1855-63 2002
    PMID 11927569DOI
    What this evidence cannot show

    Abstract level; HUMAN protein, which is why it is the reference used. 'The entire RPA14 subunit' is the paper's own description and is the basis for saying RPA3 is structurally accounted for in full, but no residue range and no PDB id are quoted here because neither was fetched. The 10-nt and 24-nt figures come from time-resolved trypsin proteolysis on purified protein and naked oligonucleotides.

    Evidence that argues against this (1)
    • This structure is the trimerisation CORE, not the whole complex, so it cannot speak to what the flexible RPA1 N-terminus or the RPA2 phospho-tail do - and those are the two parts that matter for protein recruitment, including the menin contact mapped to RPA2 aa 43-171 and the ATRIP contact on RPA1's N-terminus. A reader should not take a core structure as a structure of RPA. Searched for a human full-length RPA-ssDNA structure and did not find one cited in this beat's sources; the full-length structure that exists is fungal. Zou L et al. 2003
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    RPA3 has no literature and no dataset connection to KMT2A: RPA3[tiab] AND MLL[tiab] returns 0 and RPA3[tiab] AND KMT2A[tiab] returns 0, both genuine zeros with non-empty querytranslations, and RPA3 is absent from all seven MLL-fusion baits of the 960-protein Skucha 2018 network; the symbol is word-boundary matched in only 5 of 3,086 node files in this corpus, none of which is about RPA3 as a subject.

    What this evidence cannot show

    Bounded by titles and abstracts (rule 80) and by a short name list: RPA3 and RPA14 were not both run against the fusion terms - only RPA3 was, plus the complex-level forms 'replication protein A' and 'RPA complex' which are recorded in the beat's absence entry. NOT tried: REPA3. RPA3[tiab] AND "RNA polymerase I"[tiab] returns 0, so unlike RPA1 and RPA2 this symbol appears NOT to be contaminated by the RNA-polymerase-I nomenclature in the literature, and HGNC confirms RPA3 is not an alias of POLR1C (whose aliases are RPA40, RPA39, RPA5, AC40, RPAC1, RPC40).

    Evidence that argues against this (1)
    • The complex-level query is the one that finds RPA at this locus, and it did: RPA occupancy at the MLL 11q23 breakpoint cluster region after etoposide is published (PMID 21048951) and is invisible to every subunit-symbol query including this one, because the paper writes 'Replication protein A (RPA)'. So this zero bounds RPA3 as a NAMED SUBJECT and must not be quoted as 'RPA is absent from the KMT2A literature'. Sun J et al. 2010
Connected to