ACAP3 (ArfGAP with coiled-coil, ANK repeat and PH domains 3; also centaurin-beta-5) is a multidomain GTPase-activating protein for the small GTPase Arf6. Its 834-residue chain carries the canonical ACAP module set: an N-terminal BAR domain, a pleckstrin-homology domain, an Arf-GAP domain built around a C4-type zinc finger, and three C-terminal ankyrin repeats. It works as a peripheral membrane protein at the cell surface and on endosomal membranes, accelerating GTP hydrolysis on Arf6 and so returning it to the GDP-bound state. In neurons this activity is required for neurite outgrowth and for the migration of newly born neurons through the developing cerebral cortex; both requirements depend on catalysis, and rescue with nucleotide-locked Arf6 shows that what the neuron needs is continuous cycling of Arf6 between its active and inactive states rather than simple suppression of Arf6, so ACAP3 is best understood as one arm of a GEF/GAP cycle rather than as an Arf6 switch-off device. In epithelial cells the same activity is directed at receptor traffic: ACAP3 restrains epidermal growth factor receptor recycling and accelerates lysosomal EGFR degradation, damping EGFR signalling and proliferation. Expression is enhanced in brain. The gene is transcriptionally silenced by Myc-driven DNA hypermethylation in lung adenocarcinoma and by HDAC2 in papillary thyroid carcinoma, and in both settings reduced ACAP3 accompanies greater tumour-cell growth, migration and invasion. ACAP3 physically associates with its paralogue ACAP2, consistent with the dimeric character of the BAR-PH module shared across the ACAP family.
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0005886 plasma membrane | IBA GO_REF:0000033 | ACCEPT | Summary: Phylogenetic propagation from the deep centaurin/ArfGAP node PANTHER:PTN001142372. All seven WITH/FROM entries were resolved: Arabidopsis AGD3 (Q5W7F2) and AGD1 (Q9FIT8), fission yeast cnt5 (Q9UUE2), human ACAP2 (Q15057), C. elegans cnt-1 (Q9XXH8) and the Dictyostelium ACAP Q54WI0. This is the best-supported of the four superfamily-level IBAs, because peripheral-membrane localisation is independently demonstrated by IDA in four separate lineages: cnt5 (GO:0005886 IDA, PMID:19076239), human ACAP2 (GO:0005886 IDA, PMID:30905672), Dictyostelium Q54WI0 (GO:0005886 IDA, PMID:23264736) and worm cnt-1 (GO:0009898 cytoplasmic side of plasma membrane IDA, PMID:25383666). Every one of these donors carries the full ACAP module set, so the transfer is architecturally sound. It is also the right compartment for ACAP3 in particular. Arf6, its substrate, operates at the cell periphery, and the mouse orthologue is localised by IDA to the growth cone (GO:0030426 IDA, PMID:27330119), a plasma-membrane-rich structure. is_active_in is the appropriate qualifier for a peripheral membrane GAP. Reason: Correct compartment, architecturally sound propagation, and independently supported by IDA in four lineages. This is where the Arf6 GAP activity is exercised. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q15057 · human ACAP2 (paralogue) SUPPORTS TRANSFER A paralogue rather than an orthologue, but with the identical BAR-PH-ArfGAP-ANK architecture and a plasma membrane IDA (PMID:30905672); the compartment claim is not paralogue-specific. PomBase:SPBC17G9.08c · S. pombe cnt5 SUPPORTS TRANSFER Plasma membrane IDA, PMID:19076239. Lacks ANK repeats but retains ArfGAP-BAR-PH. dictyBase:DDB_G0279649 · D. discoideum ACAP-A (Q54WI0) SUPPORTS TRANSFER Plasma membrane IDA, PMID:23264736 - the paper that characterises it as "Dictyostelium ACAP-A is an ArfGAP involved in cytokinesis, cell migration and actin cytoskeleton dynamics". Full ACAP architecture. WB:WBGene00000565 · C. elegans cnt-1 SUPPORTS TRANSFER GO:0005886 by EXP and GO:0009898 cytoplasmic side of plasma membrane by IDA, both PMID:25383666; full ACAP architecture. AGI_LocusCode:AT5G13300 · A. thaliana AGD3/VAN3 (Q5W7F2) SUPPORTS TRANSFER GO:0005886 IDA (PMID:21118984) and GO:0005096 IDA (PMID:15743878) - a plant ArfGAP with experimental support for both the activity and the compartment. Full ACAP architecture. AGI_LocusCode:AT5G61980 · A. thaliana AGD1 (Q9FIT8) SUPPORTS TRANSFER GO:0005886 IDA (PMID:28646092), which is what it is donating here, and the paper is specifically a plasma-membrane-recruitment analysis. Worth noting it carries no experimental GO:0005096 of its own - UniProt calls it a "Probable GTPase-activating protein (By similarity)" - so it is a sound donor for the compartment but would be a weak one for the activity. PANTHER:PTN001142372 · deep centaurin/ArfGAP node SUPPORTS TRANSFER Superfamily node spanning plants, fungi, amoebae and animals. Adequate for a compartment this generic and this widely demonstrated. Supporting Evidence: PMID:11062263 ACAP1 and ACAP2 were recruited to peripheral, tubular membranes, where activation of Arf6 occurs to allow membrane recycling back to the plasma membrane. |
| GO:0001764 neuron migration | IBA GO_REF:0000033 | ACCEPT | Summary: This is the strongest annotation on the gene, and the WITH/FROM shows exactly why. Unlike the four superfamily-level IBAs, this one comes from PANTHER:PTN002754173, the ACAP3 subfamily node, and its only gene donor is MGI:MGI:2153589 - mouse Acap3, the 1:1 orthologue ("Orthologous to human ACAP3", Alliance automated description). The mouse gene carries GO:0001764 by IMP from PMID:28919417, where in utero knockdown abrogated cortical neuronal migration and was rescued by wild-type but not GAP-inactive ACAP3. So this is a clean orthologue transfer of a catalysis-attributed, in vivo phenotype - not a paralogue transfer. The context also fits the human gene, which is brain-enhanced in HPA. Reason: Orthologue-restricted propagation of an in vivo, rescue-validated, GAP-dependent phenotype. Core biological process. Propagation Review Root cause: NO FAILURE CORE Sources checked: MGI:MGI:2153589 · mouse Acap3 (1:1 orthologue) SUPPORTS TRANSFER GO:0001764 IMP from PMID:28919417. In utero knockdown plus wild-type and GAP-dead rescue; the true orthologue, not a paralogue. PANTHER:PTN002754173 · ACAP3 subfamily node SUPPORTS TRANSFER Narrow, ACAP3-restricted node. Its WITH/FROM never contains anything but mouse Acap3, which is what makes this transfer safe. Supporting Evidence: PMID:28919417 Knockdown of ACAP3 in the developing cortical neurons of mice in utero significantly abrogated neuronal migration in the cortical layer, which was restored by ectopic expression of wild type of ACAP3, but not by its GAP-inactive mutant. file:human/ACAP3/ACAP3-uniprot.txt DR HPA; ENSG00000131584; Tissue enhanced (brain). |
| GO:0005096 GTPase activator activity | IBA GO_REF:0000033 | ACCEPT | Summary: The core molecular function, and correct. All five gene donors in the WITH/FROM (the sixth entry is the PANTHER node itself) resolve to genuine ArfGAPs: Arabidopsis AGD3, mouse Acap3 (the 1:1 orthologue), yeast AGE1 (Q04412, an ArfGAP of the ASAP rather than ACAP type), human ACAP2 (GO:0005096 IDA, PMID:11062263) and Dictyostelium Q54WI0 (GO:0005096 IDA, PMID:20062541). ACAP3 itself has an Arf-GAP domain at 403-525 with a C4-type zinc finger at 418-441. Direct orthologue evidence exists and is stronger than this IBA: mouse Acap3 carries GO:0005096 by IMP from PMID:27330119, which showed Arf6-specific GAP activity and a rise in GTP-bound Arf6 on knockdown. GOA has not transferred that, so the human gene still rests on phylogenetic inference for its central activity - an ISS or IBD from PMID:27330119 would be a straightforward upgrade. One thing this term cannot express is the specificity, which is the most interesting fact about the activity. GO:0008060 "ARF GTPase activator activity" has been MERGED into GO:0005096 - QuickGO lists it among the secondaryIds of GO:0005096, along with every other substrate-specific GAP term (GO:0005097 Rab, GO:0005098 Ras, GO:0005100 Rho, GO:0005099, GO:0005101, GO:0017123, GO:0030675, GO:0046582). GO:0005096 is now the terminal GAP term, so Arf6 specificity has to live in a has_input extension (ARF6, UniProtKB:P62330) or be lost. No MODIFY is possible: there is no more specific term to move to. Reason: Correct core molecular function, sound propagation from true ArfGAPs including the 1:1 mouse orthologue. Retained as-is because GO has merged away every substrate-specific GAP term, leaving no more precise option. Propagation Review Root cause: NO FAILURE CORE Sources checked: MGI:MGI:2153589 · mouse Acap3 (1:1 orthologue) SUPPORTS TRANSFER GO:0005096 IMP from PMID:27330119, which established the activity as specific to Arf6. This is the donor that makes the transfer safe. UniProtKB:Q15057 · human ACAP2 (paralogue) SUPPORTS TRANSFER GO:0005096 IDA, PMID:11062263; same architecture, same Arf6 preference. SGD:S000002932 · S. cerevisiae AGE1 SUPPORTS TRANSFER A real ArfGAP but not an ACAP - it has the ArfGAP domain without BAR, PH or ANK modules, and Alliance places it with ASAP1/ASAP2. Harmless for a term this generic; it would not support transfer of any BAR/PH-dependent function. dictyBase:DDB_G0279649 · D. discoideum ACAP-A (Q54WI0) SUPPORTS TRANSFER GO:0005096 IDA, PMID:20062541; full ACAP architecture. AGI_LocusCode:AT5G13300 · A. thaliana AGD3/VAN3 (Q5W7F2) SUPPORTS TRANSFER GO:0005096 IDA, PMID:15743878. Plant ArfGAP with a directly demonstrated activity, which is what makes the deep node's GAP call ancestral rather than assumed. PANTHER:PTN001142372 · deep centaurin/ArfGAP node SUPPORTS TRANSFER For this term the depth of the node is a strength rather than a weakness: four independent lineages carry the activity by IDA, so ArfGAP catalysis is securely ancestral. Supporting Evidence: PMID:27330119 Ectopically expressed ACAP3 in HEK (human embryonic kidney)-293T cells showed the GAP activity specific to Arf6. PMID:27330119 the level of GTP-bound Arf6 was significantly increased by knockdown of ACAP3 in hippocampal neurons file:human/ACAP3/ACAP3-uniprot.txt GTPase-activating protein for the ADP ribosylation factor file:human/ACAP3/ACAP3-uniprot.txt DR InterPro; IPR001164; ArfGAP_dom. |
| GO:0010008 endosome membrane | IBA GO_REF:0000033 | ACCEPT | Summary: This is the annotation where WITH/FROM resolution changes the reading, even though it does not change the outcome. The donors are PANTHER:PTN001142372 (the deep superfamily node), RGD:1562939 and WB:WBGene00000565. RGD:1562939 is rat Acap2 - a PARALOGUE, the orthologue of human ACAP2, not of ACAP3 - carrying GO:0010008 by IDA from PMID:23572513 in an NGF/endocytic-recycling context, and it is the only mammalian donor. WB:WBGene00000565 is C. elegans cnt-1, the worm's single centaurin-beta, with GO:0031901 early endosome membrane IDA from PMID:22869721; being single-copy it cannot discriminate ACAP1/2/3 sub-functions. The diagnostic detail is what is absent: mouse Acap3, the 1:1 orthologue, is NOT in this WITH/FROM, although it IS in the WITH/FROM of both neuronal terms. PAINT therefore had no orthologue evidence for the endosomal compartment and fell back to the superfamily node; the orthologue's own experimentally determined location is the growth cone (GO:0030426 IDA, PMID:27330119). So when this annotation was made, the mammalian case for it rested entirely on a paralogue. Independent human evidence has since caught up. ACAP3 impairs EGFR recycling and accelerates lysosome-mediated EGFR degradation in a GAP-activity-dependent manner (PMID:41520057) - endosomal sorting biology, on the human protein. The compartment is therefore right, and it is now justified for the human gene by human data rather than by the paralogue transfer that originally produced it. Accepted, with the propagation caveat recorded rather than buried: no study has yet imaged ACAP3 itself on an endosome, so this remains an activity-site inference. Reason: The compartment is correct and is where the newly described human function is exercised, so it is retained as a core site of action. The route by which it arrived is worth flagging: its only mammalian experimental donor is the paralogue ACAP2 and the 1:1 orthologue is absent from the WITH/FROM, so the annotation was right for the wrong reason until PMID:41520057 supplied human evidence. Propagation Review Root cause: NO FAILURE CORE Failure modes: WRONG ORTHOLOG OR PARALOG Sources checked: RGD:1562939 · rat Acap2 (paralogue) SUPPORTS SOURCE BUT NOT TARGET GO:0010008 IDA (PMID:23572513) is solid for Acap2, the family member specialised for endocytic recycling (GO:0032456 IMP in the same paper). It is the orthologue of human ACAP2, not of ACAP3, and it is the only mammalian donor here. WB:WBGene00000565 · C. elegans cnt-1 SUPPORTS TRANSFER GO:0031901 early endosome membrane IDA (PMID:22869721) in the worm's single centaurin-beta, so it cannot discriminate ACAP1/2/3 sub-functions. PANTHER:PTN001142372 · deep centaurin/ArfGAP node SOURCE WEAK OR INFERRED The superfamily node was used because the ACAP3-restricted node had nothing to offer for this compartment, which is precisely why the resulting statement is family-level rather than ACAP3-level. MGI:MGI:2153589 · mouse Acap3 (1:1 orthologue) SOURCE STALE OR MISSING Absent from this WITH/FROM although present for both neuronal terms. Its own recorded location is the growth cone (GO:0030426 IDA, PMID:27330119). Supporting Evidence: PMID:41520057 Mechanically, ACAP3 inhibits epidermal growth factor receptor (EGFR) signalling via impairing EGFR recycling and accelerating lysosome-mediated EGFR degradation in a GTPase-activating protein (GAP) activity-dependent manner. |
| GO:0010975 regulation of neuron projection development | IBA GO_REF:0000033 | MODIFY | Summary: Like the neuron migration annotation, this comes from the narrow ACAP3 subfamily node PANTHER:PTN002754173 with mouse Acap3 (MGI:MGI:2153589) as its only gene donor, so the transfer itself is an orthologue transfer and is sound. The mouse gene carries GO:0010975 by IMP from PMID:27330119. The term is nonetheless less informative than the evidence supports. PMID:27330119 determined the direction: knockdown abrogated neurite outgrowth, wild-type ACAP3 rescued it and a GAP-dead mutant did not, and the authors state the conclusion as positive regulation. GO:0010976 "positive regulation of neuron projection development" is available and is what the experiment shows. Modify to the directional child. A caveat worth recording rather than hiding: the direction should not be read as "ACAP3 keeps Arf6 off". Outgrowth was rescued by a fast-cycling Arf6 mutant but not by GTP-locked or GDP-locked Arf6, so the requirement is for Arf6 cycling. ACAP3 is one arm of a GEF/GAP cycle, and the positive regulation of outgrowth is a property of the cycle, not of Arf6 inactivation. Reason: The parent term understates a directionally determined result. PMID:27330119 establishes positive regulation by knockdown plus wild-type and GAP-dead rescue, so GO:0010976 is the correct term. Note on where the fix has to land: an IBA term is constrained to whatever the ancestral node carries, which is constrained in turn by the descendant experimental annotations - here mouse Acap3's GO:0010975 IMP. PAINT cannot refine this row on its own. The recommendation is therefore addressed to MGI first (refine the mouse annotation to GO:0010976 from the same paper), after which the human IBA follows automatically. It is recorded as MODIFY rather than ACCEPT because the term genuinely should change, not because PAINT can change it unilaterally. Propagation Review Root cause: TERM SCOPING PROBLEM Failure modes: GRANULARITY MISMATCH Sources checked: MGI:MGI:2153589 · mouse Acap3 (1:1 orthologue) SUPPORTS TRANSFER GO:0010975 IMP from PMID:27330119. The source annotation is also at the non-directional parent, so the granularity loss originates upstream rather than in the propagation step. PANTHER:PTN002754173 · ACAP3 subfamily node SUPPORTS TRANSFER Orthologue-restricted node; the transfer is safe, only the term is coarse. Proposed replacements: positive regulation of neuron projection development Supporting Evidence: PMID:27330119 In primary cultured mouse hippocampal neurons, knockdown of ACAP3 abrogated neurite outgrowth, which was rescued by ectopically expressed wild-type ACAP3, but not by its GAP activity-deficient mutant. PMID:27330119 These results demonstrate that ACAP3 positively regulates neurite outgrowth through its GAP activity specific to Arf6. PMID:27330119 Thus cycling between active and inactive forms of Arf6, which is precisely regulated by ACAP3 in concert with a guanine-nucleotide-exchange factor(s), seems to be required for neurite outgrowth of hippocampal neurons. |
| GO:0030036 actin cytoskeleton organization | IBA GO_REF:0000033 | KEEP AS NON CORE | Summary: Propagated from the deep node PANTHER:PTN001142372 with three gene donors, and one of them should not be there. FB:FBgn0004133 is Drosophila blow (blown fuse), and its entire InterPro content is IPR011993 plus IPR001849 - a PH domain and nothing else. It has no ACAP1/2/3-like signature (IPR045258), no ArfGAP domain (IPR001164), no BAR domain and no ankyrin repeats, so it cannot be an Arf GAP at all. PTHR23180 (CENTAURIN/ARF) has clustered it in on the PH domain alone. FlyBase describes blow as a cytoplasmic myoblast-fusion protein acting through WASp-Vrp1 complex stability, and its GO:0030036 is an IMP from that myoblast-fusion work - a mechanism with no ArfGAP component. Its contribution to this annotation should be discounted. The reciprocal cost of the same mis-clustering is visible on blow itself, which now carries GO:0005096 GTPase activator activity by IBA despite having no GAP domain; that is the clearer error of the two and is worth reporting to PAINT. The term survives regardless, because the other two donors are legitimate. The Dictyostelium ACAPs Q54WI0 and Q551Q8 both carry the full BAR-PH-ArfGAP-ANK architecture, and Q54WI0 has GO:0030036 by IMP (PMID:23264736) and IGI (PMID:20062541) plus GO:0031941 filamentous actin IDA and GO:0051489 regulation of filopodium assembly IMP. Arf6, ACAP3's substrate, is itself a cortical actin regulator, and the mouse orthologue's phenotypes - neurite outgrowth and the morphological change of migrating cortical neurons - are actin-driven. Kept as non-core: for ACAP3 actin reorganisation is a downstream consequence of Arf6 cycling rather than a distinct job, and the term is generic. Reason: Correct in substance but generic, and one of its three gene donors (Drosophila blow) is a PH-domain-only protein that is not an ArfGAP and should not be in this family. The Dictyostelium ACAP donors carry it legitimately, and Arf6's role in cortical actin makes the term defensible - but as a downstream consequence, not a core function. Propagation Review Root cause: NO FAILURE NON CORE Failure modes: WRONG ORTHOLOG OR PARALOG GRANULARITY MISMATCH Sources checked: FB:FBgn0004133 · D. melanogaster blow (blown fuse) SUPPORTS SOURCE BUT NOT TARGET InterPro content is IPR011993 + IPR001849 only - a PH domain, no ACAP1/2/3-like signature, no ArfGAP domain, no BAR domain, no ANK repeats. Not an ArfGAP; clustered into PTHR23180 on the PH domain alone. Its GO:0030036 IMP comes from myoblast fusion via WASp-Vrp1 stability. dictyBase:DDB_G0279649 · D. discoideum ACAP-A (Q54WI0) SUPPORTS TRANSFER Full ACAP architecture; GO:0030036 IMP (PMID:23264736) and IGI (PMID:20062541), plus filamentous actin IDA and regulation of filopodium assembly IMP. The IMP paper is titled for exactly this: ACAP-A is "an ArfGAP involved in cytokinesis, cell migration and actin cytoskeleton dynamics". dictyBase:DDB_G0276395 · D. discoideum ACAP-B (Q551Q8) SUPPORTS TRANSFER Full ACAP architecture; GO:0030036 IGI (PMID:20062541). PANTHER:PTN001142372 · deep centaurin/ArfGAP node SOURCE WEAK OR INFERRED The node has absorbed a PH-domain-only protein, so annotations drawn from it need per-donor checking rather than blanket trust. Supporting Evidence: PMID:11062263 The GTP-binding protein ADP-ribosylation factor 6 (Arf6) regulates endosomal membrane trafficking and the actin cytoskeleton in the cell periphery. PMID:28919417 Furthermore, morphological changes of neurons during migration in the cortical layer were impeded in ACAP3-knocked-down cortical neurons. |
| GO:0005096 GTPase activator activity | IEA GO_REF:0000002 | ACCEPT | Summary: InterPro2GO mapping from IPR001164 (Arf GTPase activating protein) and IPR045258 (ACAP1/2/3-like). Both signatures are correct for ACAP3: the FT table places an Arf-GAP domain at 403-525 with a C4-type zinc finger at 418-441, and the InterPro cross-reference list contains both accessions. IPR001164's only InterPro2GO mapping is GO:0005096, so the electronic route is exactly what it should be. Redundant with the IBA row, but independently correct and, unlike the IBA, based on the target's own sequence. Reason: Domain-based inference that matches the protein's actual annotated domain and is the canonical InterPro2GO mapping for that domain. Supporting Evidence: file:human/ACAP3/ACAP3-uniprot.txt DR InterPro; IPR045258; ACAP1/2/3-like. file:human/ACAP3/ACAP3-uniprot.txt DR PANTHER; PTHR23180:SF407; ARF-GAP WITH COILED-COIL, ANK REPEAT AND PH DOMAIN-CONTAINING PROTEIN 3; 1. |
| GO:0005737 cytoplasm | IEA GO_REF:0000002 | MARK AS OVER ANNOTATED | Summary: InterPro2GO mapping from a single signature, IPR004148 - the BAR domain. The inference does not follow from the premise. A BAR domain is a membrane-binding and membrane-curvature module: it dimerises into a banana-shaped scaffold that binds and bends bilayers, and in ACAP1, the paralogue where this module has been solved, the BAR-PH tandem assembles into a lattice on tubulated membrane. What a BAR domain predicts is peripheral membrane association, not residence in the cytoplasm. Mapping it to GO:0005737 discards the informative half of the signal and substitutes a near-root compartment. The statement is not false - ACAP3 has no transmembrane segment or signal peptide, so it is a soluble/peripheral protein and formally cytoplasmic - but it is uninformative. It is also partly redundant, though only partly, and the distinction is worth being exact about: GO:0005737 is an ancestor of GO:0010008 endosome membrane, so the endosomal annotation already entails it, but it is NOT an ancestor of GO:0005886 plasma membrane (whose path runs through cell periphery), so cytoplasm is not simply subsumed by the record as a whole. Marked as over-annotated rather than removed, because the compartment itself is not wrong. Reason: A BAR domain implies peripheral membrane binding and curvature generation, not cytoplasmic residence; the mapping loses the informative content and yields a near-root term already covered by the two specific compartment annotations. Not removed because the assertion is technically true. Propagation Review Root cause: TERM SCOPING PROBLEM Failure modes: GRANULARITY MISMATCH ROLE CONFLATION Sources checked: InterPro:IPR004148 · BAR domain SUPPORTS SOURCE BUT NOT TARGET A real and correctly assigned ACAP3 feature (IPR042695 ACAP3_BAR, CDD cd07637 BAR_ACAP3), but its biological implication is membrane binding and curvature generation. GO:0005737 is not what it predicts. Supporting Evidence: PMID:25284369 The BAR (Bin-Amphiphysin-Rvs) domain undergoes dimerization to produce a curved protein structure, which superimposes onto membrane through electrostatic interactions to sense and impart membrane curvature. file:human/ACAP3/ACAP3-uniprot.txt DR InterPro; IPR042695; ACAP3_BAR. |
| GO:0005515 protein binding | IPI PMID:33961781 Dual proteome-scale networks reveal cell-specific remodeling... | KEEP AS NON CORE | Summary: The WITH/FROM is UniProtKB:Q15057, which is human ACAP2 - ACAP3's own paralogue. The source is BioPlex 3.0, a proteome-scale AP-MS interactome. Judged on its own a single screen hit would be weak, but three things make this partner credible. First it replicates in an orthogonal method: the second GO:0005515 row on this gene is the same pair from OpenCell endogenous split-tag IP-MS (PMID:35271311). Second it is reciprocal - ACAP2's own GOA record carries GO:0005515 with UniProtKB:Q96P50 from the same two papers. Third UniProt curates it as ACAP3's only interaction, with four supporting experiments. There is also a structural reason to expect it. The ACAP N-terminal BAR-PH tandem is a dimerisation module: ACAP1 dimerises in solution and its BAR-PH unit assembles into a membrane-deforming lattice (PMID:31291238, PMID:25284369). ACAP2 and ACAP3 both carry that module, so a BAR-mediated ACAP2-ACAP3 heterodimer is the parsimonious interpretation. GO:0046982 protein heterodimerization activity was considered as a replacement and rejected. Its definition is "Binding to a nonidentical protein to form a heterodimer", and neither AP-MS nor IP-MS measures stoichiometry: both establish co-complex membership, which is compatible with a direct heterodimer, with co-residence in a larger assembly, and with two BAR proteins sharing a membrane surface. The structural precedent is also weaker than it first looks - what PMID:31291238 shows is ACAP1 dimerising with ITSELF, a homotypic interaction. Extending that to a heterotypic ACAP2-ACAP3 dimer is a hypothesis, and co-purification of BAR-domain paralogues is a known way to produce this signal without a direct dimer. The heterodimer belongs in suggested_experiments, where it is stated as a testable hypothesis, not in a replacement term that instructs a curator to assert stoichiometry. Retaining GO:0005515 is less lossy here than it would be for a partnerless binding row, because the partner identity travels in the WITH/FROM field: this annotation is not "binds something" but "binds UniProtKB:Q15057", which is the informative part and is already recorded. Kept as non-core because a physical association with no measured stoichiometry and no known functional consequence is a lead about ACAP3, not one of its roles. Neither cached full text names ACAP3 in its body - the pair lives in the supplementary interactome tables - so the claim is anchored on the UniProt INTERACTION line and the IntAct/GOA records rather than on a quotable sentence. Reason: The partner is real and worth keeping: replicated across two orthogonal proteome-scale methods, reciprocal in ACAP2's own GOA record, and curated by UniProt with NbExp=4, with the partner identity carried in WITH/FROM. But no more specific GO term is honestly available - GO:0046982 would assert a dimeric stoichiometry that neither method measures, and the only solved ACAP dimer is homotypic ACAP1. Keep the annotation, record the partner and the replication argument here, and leave the heterodimer as a hypothesis in suggested_experiments. Supporting Evidence: file:human/ACAP3/ACAP3-uniprot.txt Q96P50; Q15057: ACAP2; NbExp=4 PMID:31291238 Simulation studies then revealed how ACAP1, which dimerizes into a symmetrical structure in solution, is recruited asymmetrically to the membrane through dynamic behavior. |
| GO:0005515 protein binding | IPI PMID:35271311 OpenCell: Endogenous tagging for the cartography of human ce... | KEEP AS NON CORE | Summary: Same partner, same conclusion, different method - and this is the row that makes the pair believable. OpenCell tags proteins at their endogenous loci with a split fluorescent protein and pulls them down at native expression levels, so it is not subject to the overexpression artefacts that afflict transfected AP-MS such as BioPlex. Recovering ACAP2 with ACAP3 by both routes is independent replication across orthogonal methods, not one screen counted twice. Treated exactly as the BioPlex row: kept at GO:0005515 rather than replaced with GO:0046982, because endogenous split-tag IP-MS is a better method for establishing that the two proteins are in one complex but is no better than AP-MS at establishing that the complex is a dimer. It is also not a spatial assay, so it supports no localisation claim either. Reason: Independent, endogenous-level replication of the ACAP2 interaction, which is what makes the partner credible rather than an overexpression artefact. Retained at GO:0005515 for the same reason as the other row: replication establishes that the interaction is real, not that it is a dimer, and GO offers no honest intermediate between bare binding and an explicit dimerization term. Supporting Evidence: file:human/ACAP3/ACAP3-uniprot.txt Q96P50; Q15057: ACAP2; NbExp=4 |
| GO:0030426 growth cone | ISS GO_REF:0000024 | NEW | Summary: Proposed new annotation, and a straightforward discrepancy to close. The mouse orthologue Acap3 (UniProtKB:Q6NXL5, MGI:MGI:2153589) carries GO:0030426 growth cone by IDA from PMID:27330119 - the same study that established its Arf6-specific GAP activity and its requirement for neurite outgrowth. The location is therefore not an isolated observation but the site at which the characterised function is exercised. The cached record for PMID:27330119 is abstract-only, so the growth-cone imaging is in full text the curator read and we cannot quote; the annotation is cited on the mouse record and is deferred to. The human UniProt entry already carries it as an Ensembl orthology projection ("DR GO; GO:0030426; C:growth cone; IEA:Ensembl.") but the GOA record does not, so this is a location present in UniProt and missing from GOA. Given that human ACAP3 is brain-enhanced and that the neuronal terms it does carry are transferred from the same mouse orthologue, the compartment should be transferred too. Reason: A specific, functionally apposite compartment supported by an IDA on the 1:1 mouse orthologue and already present on the human UniProt entry, but absent from GOA. It is more informative than either of the compartment terms currently annotated. Supporting Evidence: file:human/ACAP3/ACAP3-uniprot.txt DR GO; GO:0030426; C:growth cone; IEA:Ensembl. PMID:27330119 In primary cultured mouse hippocampal neurons, knockdown of ACAP3 abrogated neurite outgrowth, which was rescued by ectopically expressed wild-type ACAP3, but not by its GAP activity-deficient mutant. |
| GO:0032012 regulation of ARF protein signal transduction | ISS GO_REF:0000024 | NEW | Summary: Proposed new annotation, and the only route by which GO can currently record the single most specific fact about ACAP3's activity. Its GAP activity is specific to Arf6, but GO:0008060 "ARF GTPase activator activity" has been merged into the generic GO:0005096 (it appears among the secondaryIds of GO:0005096 in QuickGO), so the molecular function aspect cannot express substrate identity at all. The biological process aspect still can: GO:0032012 regulation of ARF protein signal transduction is current and is precisely what an Arf GAP does. The evidence is the mouse orthologue's IMP for GO:0005096 from PMID:27330119, where ACAP3 showed GAP activity specific to Arf6 in a cellular assay and GTP-bound Arf6 rose on knockdown. On the evidence code: the mouse orthologue does not itself carry GO:0032012, so this is not a like-for-like ISS of an existing term. IC from the human GO:0005096 annotation was considered and rejected, because GO:0005096 no longer names a substrate - it cannot support an inference about ARF specifically, which is the whole point here. What is being transferred by orthology is the mouse experimental finding that the GAP activity acts on Arf6, expressed with the only current term that can name the substrate family; ISS with the orthologue as supporting entity is the honest coding for that. The cleaner curation route, if MGI will take it, is two-step: assert GO:0032012 on the mouse record by IMP from PMID:27330119, then ISS it across, which fixes the gap at its source instead of only at the human end. The direction is deliberately left at the non-directional parent: rescue succeeded with a fast-cycling Arf6 mutant but failed with GTP-locked and GDP-locked mutants, so ACAP3 sustains Arf6 cycling rather than simply reducing Arf6 signalling, and a negative-regulation child would misdescribe it. Reason: Records the Arf specificity of the GAP activity, which the molecular function ontology can no longer express since GO:0008060 was merged into GO:0005096. Kept at the non-directional parent because the requirement is for Arf6 cycling, not suppression. Supporting Evidence: PMID:27330119 Ectopically expressed ACAP3 in HEK (human embryonic kidney)-293T cells showed the GAP activity specific to Arf6. PMID:27330119 Thus cycling between active and inactive forms of Arf6, which is precisely regulated by ACAP3 in concert with a guanine-nucleotide-exchange factor(s), seems to be required for neurite outgrowth of hippocampal neurons. |
| GO:0001920 negative regulation of receptor recycling | IMP PMID:41520057 Myc-mediated epigenetic silencing of ACAP3 promotes lung ade... | NEW | Summary: Proposed new annotation, from the only mechanistic study performed on the human protein. In lung adenocarcinoma cells ACAP3 impairs EGFR recycling and accelerates lysosome-mediated EGFR degradation, and the effect requires GAP activity. This places the human protein in the endosomal receptor-sorting arm of Arf6 biology, alongside its paralogues ACAP1 and ACAP2, and it is the first human-cell function ACAP3 has. IMP rather than IDA: the experiments are gain- and loss-of-function manipulations of ACAP3 with EGFR trafficking as the readout, not a direct assay of ACAP3 at native levels. The cached record is abstract-only, so the quoted mechanism sentence is from the abstract. Reason: GAP-activity-dependent restraint of receptor recycling, demonstrated on the human protein and entirely absent from GOA. Currently ACAP3 has no annotation of any kind derived from a human experiment. Supporting Evidence: PMID:41520057 Mechanically, ACAP3 inhibits epidermal growth factor receptor (EGFR) signalling via impairing EGFR recycling and accelerating lysosome-mediated EGFR degradation in a GTPase-activating protein (GAP) activity-dependent manner. |
| GO:0042059 negative regulation of epidermal growth factor receptor signaling pathway | IMP PMID:41520057 Myc-mediated epigenetic silencing of ACAP3 promotes lung ade... | NEW | Summary: Proposed new annotation, the signalling-level consequence of the trafficking effect above. ACAP3 inhibits EGFR signalling, and the mechanism given is the trafficking one: reduced recycling plus accelerated lysosomal degradation lowers the surface receptor pool. The pathway-level term is worth annotating separately from the trafficking term because the two are separable claims - a receptor-trafficking change need not damp signalling - and both are supported here. Downstream proliferation and tumour-suppression phenotypes are reported in the same study (and in papillary thyroid carcinoma, PMID:39098591) but are too far downstream of a GAP activity to annotate; the AKT and p53 read-outs in the thyroid study especially so. Reason: Directly demonstrated on the human protein and mechanistically explained. Separable from, and complementary to, the receptor-recycling annotation. Supporting Evidence: PMID:41520057 Mechanically, ACAP3 inhibits epidermal growth factor receptor (EGFR) signalling via impairing EGFR recycling and accelerating lysosome-mediated EGFR degradation in a GTPase-activating protein (GAP) activity-dependent manner. PMID:41520057 ACAP3 significantly suppresses the proliferation of LUAD cells in vitro and in vivo. |
| GO:0008270 zinc ion binding | ISM file:human/ACAP3/ACAP3-uniprot.txt | NEW | Summary: Proposed new annotation, and unlike the other three proposals here it rests on ACAP3's own sequence rather than on orthology. The FT table places a C4-type zinc finger at 418-441, inside the Arf-GAP domain (403-525), called from the PROSITE ARFGAP profile (PS50115 / ProRule PRU00288). The UniProt entry still carries the resulting GO statement - "DR GO; GO:0008270; F:zinc ion binding; IEA:UniProtKB-KW." - but GOA does not, because the keyword-derived route (GO_REF:0000043) was withdrawn for cellular organisms, and InterPro2GO does not fill the gap: IPR001164's only mapping is GO:0005096. So no automatic route now supplies it, which is why it is absent rather than because anyone judged it wrong. Two limits stated plainly. First, the basis is a sequence-model call, not a metal-binding measurement on ACAP3 - hence ISM, with the UniProt record as the reference, rather than an experimental code. Second, at the family level the motif is known to be functionally required rather than merely present: the founding ArfGAP paper reported that the ARF1 GAP "predicts a protein with a zinc finger motif near the amino terminus" and that "The GAP function required an intact zinc finger", which is mutational evidence that the motif matters, though still not a demonstration that zinc is bound. Not a core function. Zinc coordination here is a structural requirement of the ArfGAP fold, not a distinct biological activity of ACAP3, so it is recorded as a molecular feature and deliberately left out of core_functions. Reason: Better founded than either of the orthology-based proposals, because it rests on a feature of ACAP3's own sequence, and it is missing from GOA for a purely mechanical reason - the withdrawal of the keyword route with no InterPro2GO replacement - rather than on any judgement about the protein. Supporting Evidence: file:human/ACAP3/ACAP3-uniprot.txt DR GO; GO:0008270; F:zinc ion binding; IEA:UniProtKB-KW. file:human/ACAP3/ACAP3-uniprot.txt DR PROSITE; PS50115; ARFGAP; 1. PMID:8533093 The GAP function required an intact zinc finger and additional amino-terminal residues. |
Loading supporting content…
Download this section (compressed HTML)Q: Should PANTHER family PTHR23180 (CENTAURIN/ARF) continue to contain Drosophila blow (FBgn0004133)? Its InterPro content is a PH domain and nothing else - no ACAP1/2/3-like signature, no ArfGAP domain, no BAR domain, no ankyrin repeats - so it cannot be an Arf GAP, yet it is a WITH/FROM donor for ACAP3's actin annotation and has itself acquired GO:0005096 GTPase activator activity by IBA. Both directions of that transfer look unsafe.
Q: Now that GO:0008060 has been merged into GO:0005096, how should the Arf6 specificity of ACAP1, ACAP2 and ACAP3 be recorded? A has_input extension naming ARF6 is available but is not used on any of the three, so the ontology currently cannot distinguish an Arf6-preferring GAP from an Arf1-preferring one anywhere in this family.
Q: Should the mouse orthologue's curated annotations be transferred to human ACAP3 by ISS? Mouse Acap3 has GO:0005096 IMP, GO:0001764 IMP, GO:0010975 IMP and GO:0030426 IDA from PMID:27330119 and PMID:28919417, while the human gene carries only IBA/IEA/IPI. The IBAs are already drawn from that same mouse gene, so the evidence exists but arrives in its weakest available form.
Q: Will MGI refine two mouse Acap3 annotations at source? Both fixes this review proposes for the human gene are blocked upstream: GO:0010975 should be GO:0010976 (PMID:27330119 determines the direction), and GO:0032012 regulation of ARF protein signal transduction should be asserted by IMP from the same paper. An IBA cannot be refined past its ancestral node, so until the mouse annotations change, the human MODIFY and the human ISS are both requests PAINT alone cannot honour.
Q: Was the withdrawal of GO_REF:0000043 meant to drop zinc ion binding for the whole ArfGAP family? The C4-type zinc finger is a defining feature of the ArfGAP domain and the UniProt entries still carry the keyword-derived GO statement, but InterPro2GO maps IPR001164 only to GO:0005096, so nothing replaced the withdrawn route. If the loss was incidental rather than intended, an InterPro2GO mapping from IPR001164 to GO:0008270 would restore it across the family in one step instead of gene by gene.
Q: What determines the division of labour between the three human ACAPs? All three are Arf6-preferring GAPs with identical domain architecture; ACAP1 and ACAP2 are broadly expressed while ACAP3 is brain-enhanced, and ACAP2 and ACAP3 physically associate. Is the specialisation set by expression, by localisation, or by which partners each one dimerises with?
Q: Is ACAP3's tumour-suppressive behaviour entirely explained by EGFR downregulation? Two independent tumour types report growth suppression on ACAP3 restoration with different proximal read-outs - EGFR trafficking in lung adenocarcinoma, AKT and p53 in papillary thyroid carcinoma - and it is unclear whether these are one mechanism seen twice or two.
Experiment: Knock in a fluorescent tag at the endogenous human ACAP3 locus in iPSC-derived neurons and in a lung adenocarcinoma line, and image against Arf6, EEA1, RAB11, LAMP1 and surface EGFR. Compare the two cell types directly: if the compartment differs between them, the current single set of compartment annotations is wrong for at least one context.
Hypothesis: ACAP3 acts on Arf6 at the growth cone plasma membrane in human neurons, not at endosomes, and the endosomal annotation is a transfer from the paralogue ACAP2.
Type: live-cell imaging with endogenous tagging
Experiment: In ACAP3-depleted hippocampal neurons, screen the known Arf6 GEFs (the cytohesin/ARNO family, EFA6 family, BRAG family) by co-depletion and by rescue with the fast-cycling Arf6 mutant, scoring neurite length and GTP-Arf6. A GEF whose loss phenocopies ACAP3 loss and whose double depletion is not additive is the cycle partner.
Hypothesis: ACAP3's requirement in neurite outgrowth reflects a GEF/GAP cycle, and the partner GEF can be identified by epistasis.
Type: genetic epistasis in primary neurons
Experiment: Purify the ACAP2 and ACAP3 BAR-PH modules separately and together, measure stoichiometry by SEC-MALS and mass photometry, map the interface by crosslinking mass spectrometry, then compare liposome tubulation and Arf6 GAP kinetics of the heterodimer against both homodimers. Test interface mutants in cells for effects on ACAP3 localisation and on the EGFR phenotype.
Hypothesis: ACAP2 and ACAP3 form a BAR-mediated heterodimer whose properties differ from either homodimer.
Type: structural biochemistry and reconstitution
Experiment: Assay recombinant full-length ACAP3 and its BAR-PH module for binding to liposomes across the phosphoinositide series, and measure Arf6 GAP kinetics with and without each lipid. Score tubulation by negative-stain EM in the same experiment, which would supply the first BAR-protein annotation to GO:0180020 membrane bending activity. Then test whether PH-domain mutants that lose the preferred lipid also lose growth-cone localisation and the neurite outgrowth rescue.
Hypothesis: ACAP3's GAP activity is stimulated by a specific phosphoinositide bound through its BAR-PH module, as it is in the paralogues.
Type: in vitro lipid binding and enzymology
Experiment: Rebuild the PTHR23180 alignment with and without blow and score whether blow places inside the ACAP clade on the ArfGAP and BAR regions or only on the PH domain, then check how many other PTHR23180 members lack IPR001164. Report the result, and blow's own GO:0005096 IBA, to the PANTHER and PAINT curators.
Hypothesis: Drosophila blow is mis-assigned to PTHR23180 and its inclusion is injecting an ArfGAP-independent actin mechanism into the family node.
Type: comparative sequence and family analysis
What is not known — curated, literature-grounded statements of the open unknowns (the inverse of core functions).
Gap: GO can no longer express which GTPase a GAP acts on, so the defining fact about ACAP3's molecular function - that its GAP activity is specific to Arf6 - cannot be recorded as a molecular function.
OPEN ONTOLOGY MF_DARK
What is known: The specificity itself is not in doubt. ACAP3 showed GAP activity specific to Arf6 in a cellular assay, GTP-bound Arf6 rises on ACAP3 knockdown, and every ACAP3 phenotype is abolished by a GAP-dead mutant. The obstacle is ontological: GO:0008060 "ARF GTPase activator activity" has been merged into GO:0005096 "GTPase activator activity" - it appears among the secondaryIds of GO:0005096 in QuickGO, together with GO:0005097 (Rab), GO:0005098 (Ras), GO:0005100 (Rho), GO:0005099, GO:0005101, GO:0017123, GO:0030675 and GO:0046582. GO:0005096 is now the terminal GAP term for every substrate.
Significance: ACAP3, ACAP1 and ACAP2 are Arf6-preferring GAPs, whereas ASAP1 in the same ArfGAP superfamily is measurably less Arf6-selective in the same side-by-side assay. After the merge all of them carry the identical molecular function term, so the ontology cannot distinguish paralogues that differ in substrate preference. Substrate identity is the main axis of variation in this family, and it is exactly what is now invisible.
What would resolve it: Annotate GO:0005096 with a has_input extension naming ARF6 (UniProtKB:P62330), which is the mechanism GO retained; alternatively GO could reinstate a substrate-partitioned grouping under GO:0005096. The GO:0032012 regulation of ARF protein signal transduction biological-process annotation proposed here is a partial workaround, since it at least records the Arf family.
Provenance (the field's own admissions):
Gap: Not one of ACAP3's ten GOA annotations derives from an experiment; every annotation is either a phylogenetic or an electronic inference, although four functional papers exist.
OPEN CURATION MF_DARK
What is known: The protein is not undetected - the UniProt entry is "PE 1: Evidence at protein level" and ACAP3 has been identified by mass spectrometry. And the functional literature is not empty: PMID:27330119 and PMID:28919417 characterise the mouse orthologue, PMID:41520057 and PMID:39098591 the human protein in tumour cells. The mouse orthologue has been curated from these papers (GO:0005096 IMP, GO:0001764 IMP, GO:0010975 IMP, GO:0030426 IDA). The human gene has not: its record consists of six IBAs, two InterPro IEAs and two high-throughput IPIs, and the UniProt entry cites none of the four functional papers.
Significance: ACAP3 looks moderately well annotated - it has a molecular function, two compartments and three biological processes - but the whole record is inference. A curation pass over the four papers, plus ISS transfer from the curated mouse orthologue, would convert the gene from inferred to evidenced without any new experiment.
What would resolve it: Curate PMID:41520057 and PMID:39098591 for the human gene, and transfer the mouse orthologue's IMP/IDA annotations by ISS. The four NEW annotations proposed in this review are that curation pass.
Provenance (the field's own admissions):
Gap: ACAP3's only known physical partner is its own paralogue ACAP2, and neither the architecture of that association nor its purpose is known.
OPEN BIOLOGY RESIDUAL_SUBGAP
What is known: The interaction itself is well supported: two orthogonal proteome-scale methods (BioPlex AP-MS, PMID:33961781; OpenCell endogenous split-tag IP-MS, PMID:35271311), reciprocal annotation on both genes, UniProt NbExp=4. Dimerisation of the ACAP BAR-PH module is established for ACAP1 by structural and simulation work, but homotypically - ACAP1 with ACAP1. What is untested for ACAP2-ACAP3 is the stoichiometry, whether the BAR domains mediate it at all, where in the cell it happens, and whether it changes anything about Arf6 specificity, membrane curvature generation or localisation. Because none of that is settled, the interaction is recorded here as a gap rather than promoted to a core function or to a dimerization term.
Significance: If ACAPs form obligate dimers and can mix, then ACAP2 and ACAP3 phenotypes are not independent, and the division of labour between the three human ACAPs may be set by which partners are co-expressed rather than by intrinsic differences between them. The question is also what stands between the current uninformative GO:0005515 annotation and a term that says something: no GO term can currently express "physically associates with a specific paralogue" without also asserting a stoichiometry nobody has measured.
What would resolve it: Measure the stoichiometry of the purified complex (SEC-MALS, mass photometry), map the interface by crosslinking mass spectrometry, and test whether BAR-domain interface mutants that abolish heterodimerisation change Arf6 GAP activity or ACAP3 localisation. A positive dimer result would justify GO:0046982; a larger-assembly result would point to GO:0032403 protein-containing complex binding instead.
Provenance (the field's own admissions):
Gap: ACAP3 has no lipid-binding annotation and no lipid-binding data, despite carrying both a PH domain and a BAR domain, and despite the family's GAP activity being phosphoinositide-dependent.
OPEN BIOLOGYCURATION MF_DARK
What is known: The modules are unambiguous: FT DOMAIN 268-363 PH, plus a BAR domain recognised as IPR042695 ACAP3_BAR and CDD cd07637 BAR_ACAP3. In the paralogues, phosphoinositide dependence is demonstrated - all ACAPs have phosphoinositide-dependent GAP activity (PMID:11062263) and the ACAP1 BAR-PH tandem binds and tubulates PI(4,5)P2-containing liposomes. Worm cnt-1 has direct binding data for PI(4,5)P2, PI(3,4,5)P3 and PI4P. For ACAP3 itself there is no binding assay of any kind, so which phosphoinositide recruits it, and whether its GAP activity is lipid-stimulated, are unknown.
Significance: Phosphoinositide dependence is how ACAP-family GAPs are targeted to the right membrane at the right time, so it is a load-bearing part of the mechanism rather than an accessory property. Its absence also means the BAR and PH domains - two of ACAP3's four modules - contribute nothing to the annotation record, and the one place a BAR domain does appear (the IPR004148 to GO:0005737 mapping) uses it to infer the wrong thing.
What would resolve it: Liposome binding and tubulation assays with recombinant ACAP3 BAR-PH across the phosphoinositide series, plus a GAP assay on Arf6 with and without PI(4,5)P2, would support a specific phosphoinositide-binding annotation (GO:0005546 or a sibling) and test lipid stimulation. A tubulation result would also be annotatable: GO:0180020 membrane bending activity is current and is defined for exactly this, yet it has only four human annotations (CHMP2A, CHMP3, OPA1) and not one BAR-domain protein among them - even ACAP1, for which liposome tubulation by the BAR-PH module has been shown directly (PMID:25284369). The term is waiting for this whole protein family.
Provenance (the field's own admissions):
Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)