ACAP2 (centaurin-beta-2) is a multidomain GTPase-activating protein that terminates ARF6 signalling in the cell periphery. Its ARF-GAP domain, with a C4-type zinc finger and the catalytic arginine at position 442, accelerates GTP hydrolysis on ADP-ribosylation factor 6 in preference to ARF1 and ARF5, and that catalysis is gated by membrane lipid - it requires phosphatidylinositol 4,5-bisphosphate together with phosphatidic acid, so ACAP2 is only active where those lipids are. A separate module at the other end of the protein, the third of three ankyrin repeats plus a following extended helix, binds the GTP-loaded form of the small GTPase RAB35 - and only RAB35, none of the other roughly sixty mammalian Rabs. ACAP2 is not a GAP for RAB35. The two modules make ACAP2 a coincidence device - RAB35 recruits it onto a membrane, the local phosphoinositide licenses its catalysis, and ARF6 is switched off at that spot. Because ARF6 drives cortical actin remodelling and the fast recycling route back to the plasma membrane, the consequence of ACAP2 arriving is that ARF6-dependent membrane protrusion and ruffling are damped. At steady state most of the protein is soluble, with pools on the tubular ARF6 endosome and on plasma-membrane protrusive buds and ruffles; it does not affect the Golgi. The same RAB35-ACAP2-ARF6 relay has been captured in several settings - nerve-growth-factor-induced neurite outgrowth, formation of the actin-dependent phagocytic cup during FcgammaR-mediated phagocytosis, and oligodendrocyte differentiation and myelination, where ACAP2 acts as a brake. A separate line of work identifies ACAP2 as the human counterpart of Caenorhabditis elegans CNT-1 and reports a pro-apoptotic activity and direct phosphoinositide binding, although ACAP2 itself is not cleaved by caspases. Two close paralogs, ACAP1 and ACAP3, share the same domain architecture and ARF6 preference; ACAP2 and ACAP3 are recovered together in proteome-scale interaction datasets.
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0005886 plasma membrane | IBA GO_REF:0000033 | ACCEPT | Summary: Phylogenetic inference from PANTHER node PTN001142372, and the WITH/FROM field contains UniProtKB:Q15057 - ACAP2 itself. A self-referential IBA is not circular reasoning; it records a PAN-GO curator inspecting the human protein and judging that the plasma membrane is where its activity happens. That judgement is right and is independently supported here - the crystallography paper's cell biology puts ACAP2 at the membrane with RAB35, UniProt carries Cell membrane with an ECO:0000269 tag, and the founding paper saw it at plasma-membrane protrusive buds. The is_active_in qualifier is the correct one; this is a site of action, not merely a compartment the protein can be found in. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q15057 · ACAP2 (human, self-reference) SUPPORTS TRANSFER Self-referential IBA - a PAN-GO curator judging this location core for the human protein. PANTHER:PTN001142372 · pan-eukaryotic centaurin/ArfGAP ancestral node SUPPORTS TRANSFER Node is very deep (see the GO:0030036 row), but peripheral-membrane action is common to the clade and the human protein has its own direct evidence. WB:WBGene00000565 · C. elegans cnt-1 (UniProtKB:Q9XXH8) SUPPORTS TRANSFER Holds its own experimental evidence for the donated term - QuickGO returns IDA x3 and EXP x2 for GO:0005886 on Q9XXH8, so this is not a family-level inference being recycled. PomBase:SPBC17G9.08c · S. pombe csx2 (UniProtKB:Q9UUE2) SUPPORTS TRANSFER Carries GO:0005886 IDA in its own right. AGI_LocusCode:AT5G13300 · Arabidopsis AGD3/VAN3/SFC ArfGAP (UniProtKB:Q5W7F2) SUPPORTS TRANSFER Carries GO:0005886 IDA in its own right. AGI_LocusCode:AT5G61980 · Arabidopsis AGD1 ArfGAP (UniProtKB:Q9FIT8) SUPPORTS TRANSFER Carries GO:0005886 IDA in its own right. dictyBase:DDB_G0279649 · Dictyostelium Q54WI0 (BAR-PH-ArfGAP-ANK) SUPPORTS TRANSFER Verified genuine ACAP-like architecture, PTHR23180:SF160, and carries GO:0005886 IDA. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC -!- SUBCELLULAR LOCATION: Cell membrane {ECO:0000269|PubMed:30905672}. PMID:11062263 In addition, ACAP1 and ACAP2 were recruited to peripheral, tubular membranes, where activation of Arf6 occurs to allow membrane recycling back to the plasma membrane. |
| GO:0005096 GTPase activator activity | IBA GO_REF:0000033 | ACCEPT | Summary: Correct, and this is the core molecular function. Again self-referential (UniProtKB:Q15057 appears in WITH/FROM), so a PAN-GO curator has affirmed it for the human protein, and it is redundant with the direct biochemistry in PMID:11062263. The term is retained unchanged rather than made more specific for an ontology reason, not a biological one: GO:0008060 "ARF GTPase activator activity" no longer exists as a distinct class. It is one of nine substrate-specific GAP terms (with GO:0005097 Rab, GO:0005098 Ran, GO:0005099 Ras, GO:0005100 Rho, GO:0005101, GO:0017123, GO:0030675, GO:0046582) merged into GO:0005096, which now has no substrate-specific children at all. So GO:0005096 is already the most precise term available, and ACAP2's genuine ARF6-over-ARF1/ARF5 preference has to be recorded as a has_input annotation extension or in a GO-CAM instead - which is what this review does, on the IDA row for the same term. Note also that MGI:MGI:2153589 in the WITH/FROM resolves via the Alliance of Genome Resources to mouse Acap3, a paralog rather than the mouse Acap2 ortholog (which is Q6ZQK5) - harmless for a family-level IBA of a function all ACAPs share, but worth recording. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q15057 · ACAP2 (human, self-reference) SUPPORTS TRANSFER Self-referential IBA marking the function core; independently supported by IDA. MGI:MGI:2153589 · mouse Acap3 SUPPORTS TRANSFER Resolves via the Alliance of Genome Resources to the paralog Acap3, not to mouse Acap2 (Q6ZQK5). It is nonetheless a sound donor rather than a weak link - mouse Acap3 (UniProtKB:Q6NXL5) carries GO:0005096 by IMP in its own right - so the point is only that a paralog, not the ortholog, is doing the work in this support set. SGD:S000002932 · S. cerevisiae AGE1 (UniProtKB:Q04412) SUPPORTS TRANSFER Carries GO:0005096 IDA in its own right. AGI_LocusCode:AT5G13300 · Arabidopsis AGD3/VAN3/SFC ArfGAP (UniProtKB:Q5W7F2) SUPPORTS TRANSFER Carries GO:0005096 IDA in its own right. dictyBase:DDB_G0279649 · Dictyostelium Q54WI0 (BAR-PH-ArfGAP-ANK) SUPPORTS TRANSFER Carries GO:0005096 IDA in its own right. Supporting Evidence: PMID:11062263 In vitro, ACAP1 and ACAP2 preferred Arf6 as a substrate, rather than Arf1 and Arf5, more so than did ASAP1. file:human/ACAP2/ACAP2-uniprot.txt CC -!- FUNCTION: GTPase-activating protein (GAP) for ADP ribosylation factor 6 file:human/ACAP2/ACAP2-bioinformatics/RESULTS.md | `MGI:MGI:2153589` | Q6NXL5 Q6NXL5_MOUSE (Acap3) | Mus musculus | 833 | yes | GO:0005096 (IBA) | Knowledge gap: GO cannot express which GTPase a GAP acts on. ACAP2's defining specificity - ARF6 in preference to ARF1 and ARF5, and explicitly not RAB35 - is invisible in the annotation. ONTOLOGYCURATION |
| GO:0010008 endosome membrane | IBA GO_REF:0000033 | ACCEPT | Summary: Correct and core. Unlike the other three IBA rows this one has an ortholog-grade support set - RGD:1562939 resolves to rat Acap2 (Q5FVC7), the true ortholog, and WB:WBGene00000565 to C. elegans cnt-1 - so the inference does not depend on the deep node alone. It matches the human data directly, since ACAP2 sits on the tubular ARF6 endosome, which is where ARF6 is activated for recycling, and UniProt records Endosome membrane as a peripheral-membrane association. is_active_in is appropriate because this is the membrane on which ACAP2 inactivates ARF6. Propagation Review Root cause: NO FAILURE CORE Sources checked: RGD:1562939 · rat Acap2 (UniProtKB:Q5FVC7) SUPPORTS TRANSFER True ortholog, not a paralog, and it carries GO:0010008 IDA in its own right. WB:WBGene00000565 · C. elegans cnt-1 (UniProtKB:Q9XXH8) SUPPORTS TRANSFER Carries GO:0010008 IDA x2 in its own right. PANTHER:PTN001142372 · pan-eukaryotic centaurin/ArfGAP ancestral node SUPPORTS TRANSFER Supporting Evidence: PMID:11062263 In untreated cells, although ACAP1 and ACAP2 were primarily cytosolic, they could be observed on tubular portions of the Arf6 endosomal compartment file:human/ACAP2/ACAP2-uniprot.txt CC Endosome membrane {ECO:0000250}; Peripheral membrane protein |
| GO:0030036 actin cytoskeleton organization | IBA GO_REF:0000033 | MODIFY | Summary: The conclusion that ACAP2 has something to do with actin is right, but both the term and its evidence chain are wrong in ways worth separating. On the term. ACAP2 is not part of the actin machinery. It is a GAP that switches ARF6 off, and ARF6 is what remodels cortical actin; the readouts are suppression of PDGF-induced dorsal ruffles and of ARF6-dependent protrusions, both abolished when GAP activity is removed. That is regulation of actin cytoskeleton organization (GO:0032956), one level up in role, not the process itself. On the evidence chain. This IBA's four supporting entities are PANTHER:PTN001142372 plus two Dictyostelium proteins and Drosophila blow. The two Dictyostelium entries check out - Q54WI0 (1333 aa) and Q551Q8 (843 aa) both carry Arf-GAP domains in genuine ACAP-like architectures, and Q54WI0 holds its own experimental actin annotations, so the term itself is properly supported. blow (blown fuse) does not check out. All four of its UniProt entries have exactly one annotated domain, a PH domain - A1Z714 204-307, P91678 204-307, E1JGZ4 197-300, Q8MSU1 92-195 - with no BAR domain, no ArfGAP catalytic domain and no ankyrin repeats on any isoform, and none of the four carries a PANTHER cross-reference in UniProt. PANTHER nonetheless assigns blow to PTHR23180 subfamily SF399, so its experimental actin annotation - which comes from myoblast fusion - is propagated into ACAP2's support set. Whatever blow does to actin, it cannot be doing it as an ARF GAP. The subfamily assignment is the sharper form of the problem. SF399 is not a domain-less outlier group that happens to contain blow; the cached PTHR23180 member table shows its other reviewed member is C. elegans cnt-1 (Q9XXH8, 826 aa, "Arf-GAP with ANK repeat and PH domain-containing protein"), a bona fide ACAP-family protein that this same review treats as ortholog-grade support on other rows. So a PH-domain-only protein has been placed inside a real ACAP subfamily, and the subfamily name PANTHER gives it - "BLOWN FUSE, ISOFORM A-RELATED" - takes its label from the member that lacks the catalytic domain rather than from the one that has it. PTN001142372 is also much deeper than the ACAP1/2/3 node. Querying QuickGO for annotations carrying that node returns 1000 annotations of the same four terms, on Brassica, Cucumis, Medicago, Theobroma, Zostera, Mimulus, Manihot, Sorghum and Helianthus genes, on the charophyte alga Klebsormidium, on yeast AGE1, and on Xenopus acap2.L - i.e. it predates the plant/fungal/animal split. GAP activity survives that depth; a specific actin process does not. Every identifier and domain assignment above was resolved programmatically; see ACAP2-bioinformatics/RESULTS.md and the script that generates it. MODIFY rather than REMOVE because the human protein has its own evidence for the regulatory role, so the right fix is the term and not deletion. Reason: ACAP2 acts upstream of actin remodelling as an ARF6 GAP rather than participating in actin cytoskeleton organization itself, so the regulation term is the accurate role. The IBA's own support set is additionally weak - a pan-eukaryotic node plus Drosophila blow, a PH-domain-only protein with no ArfGAP domain whose actin annotation comes from myoblast fusion. Propagation Review Root cause: TERM SCOPING PROBLEM Failure modes: ROLE CONFLATION SOURCE EVIDENCE WEAK Sources checked: FB:FBgn0004133 · Drosophila blow (blown fuse) SOURCE WEAK OR INFERRED PH domain only on all four isoforms (204-307, 204-307, 197-300, 92-195); no BAR, no Arf-GAP, no ANK, and no PANTHER xref in UniProt. PANTHER places it in PTHR23180:SF399 anyway, alongside the genuine ACAP protein cnt-1. Its own GO:0030036 annotation is experimental (IMP on A1Z714 and E1JGZ4), so the objection is not evidence quality - it is that a protein with no ArfGAP domain cannot be doing to actin what ACAP2 does, so its myoblast-fusion phenotype is not transferable support for ACAP2. dictyBase:DDB_G0279649 · Dictyostelium Q54WI0 SUPPORTS TRANSFER Verified BAR-PH-ArfGAP-ANK architecture, and it carries GO:0030036 by both IMP and IGI in its own right - this is the source that actually justifies the term. dictyBase:DDB_G0276395 · Dictyostelium Q551Q8 SUPPORTS TRANSFER Verified PH-ArfGAP architecture, PTHR23180:SF414, and carries GO:0030036 by IGI. PANTHER:PTN001142372 · pan-eukaryotic centaurin/ArfGAP ancestral node SUPPORTS SOURCE BUT NOT TARGET Node spans plants, fungi and animals (same four terms appear on Brassica, Medicago, Zostera, yeast AGE1 and Xenopus acap2.L), too deep to carry a specific actin process term. Proposed replacements: regulation of actin cytoskeleton organization Supporting Evidence: PMID:11062263 Overexpression of either ACAP inhibited the formation of PDGF-induced dorsal ruffles PMID:11062263 The effects of ACAP1 and ACAP2 were dependent on GAP activity. file:human/ACAP2/ACAP2-bioinformatics/RESULTS.md - `A1Z714` A1Z714_DROME, 644 aa, domains: PH 204-307, PANTHER xrefs in UniProt: NONE file:human/ACAP2/ACAP2-bioinformatics/RESULTS.md - PANTHER assignment: PTHR23180 / PTHR23180:SF399 file:human/ACAP2/ACAP2-bioinformatics/RESULTS.md - `Q9XXH8` cnt-1 (Caenorhabditis elegans), 826 aa - Arf-GAP with ANK repeat and PH domain-containing protein cnt-1 file:human/ACAP2/ACAP2-bioinformatics/RESULTS.md - QuickGO reports 1000 annotations citing this node. |
| GO:0005096 GTPase activator activity | IEA GO_REF:0000002 | ACCEPT | Summary: InterPro2GO mapping from IPR001164 (ArfGAP domain) and IPR045258 (ACAP1/2/3-like). Both signatures are appropriate and well scoped, and the domain they detect is real and intact in ACAP2 - residues 399-520, with the C4-type zinc finger at 414-437 and the catalytic arginine at 442 whose mutation abolishes activity. This is one of the cases where a domain-derived IEA and the characterised activity coincide, because the ArfGAP domain is the module that performs the function rather than merely accompanying it. Redundant with the IDA but not wrong. Propagation Review Root cause: NO FAILURE CORE Sources checked: InterPro:IPR001164 · ArfGAP domain SUPPORTS TRANSFER Present at 399-520 with an intact catalytic Arg442; R442Q abolishes GAP activity. InterPro:IPR045258 · ACAP1/2/3-like SUPPORTS TRANSFER Subfamily-scoped signature, so no risk of leaking across the wider ArfGAP superfamily. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt FT DOMAIN 399..520 file:human/ACAP2/ACAP2-uniprot.txt FT /note="R->Q: Loss of GAP activity." |
| GO:0005737 cytoplasm | IEA GO_REF:0000002 | KEEP AS NON CORE | Summary: The conclusion happens to be true - ACAP2 is largely soluble at steady state, becoming membrane-associated when recruited - but the inference behind it is poor. The mapping comes from IPR004148, the BAR domain, and a BAR domain is a membrane-curvature-binding module; deducing cytoplasmic residence from it says nothing a curator would want. Kept because it is not incorrect, but non-core, because the informative locations for this protein are the plasma membrane, the ruffle and the endosome membrane, where the GAP actually works. Reason: Correct but uninformative, and derived from a signature (BAR) whose implication is membrane binding rather than cytoplasmic residence. The specific membrane locations already carry the curatorially useful information. Supporting Evidence: PMID:11062263 In untreated cells, although ACAP1 and ACAP2 were primarily cytosolic, they could be observed on tubular portions of the Arf6 endosomal compartment |
| GO:0005886 plasma membrane | IEA GO_REF:0000044 | ACCEPT | Summary: Mapping from the UniProt subcellular-location vocabulary term SL-0039. The underlying UniProt statement is experimentally supported (ECO:0000269|PubMed:30905672), so this electronic row inherits real evidence rather than a guess, and it agrees with the IDA row from the same paper and with the peripheral plasma-membrane pools seen in the founding paper. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC -!- SUBCELLULAR LOCATION: Cell membrane {ECO:0000269|PubMed:30905672}. |
| GO:0010008 endosome membrane | IEA GO_REF:0000044 | ACCEPT | Summary: Mapping from UniProt SL-0100. Weaker in provenance than the plasma-membrane row because UniProt tags the endosome-membrane statement ECO:0000250 (by similarity) rather than ECO:0000269, but it is corroborated in human cells - ACAP2 is on tubular portions of the ARF6 endosome - and by the ISS transfer from rat Acap2. Correct and core. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC Endosome membrane {ECO:0000250}; Peripheral membrane protein PMID:11062263 In untreated cells, although ACAP1 and ACAP2 were primarily cytosolic, they could be observed on tubular portions of the Arf6 endosomal compartment |
| GO:0005515 protein binding | IPI PMID:33961781 Dual proteome-scale networks reveal cell-specific remodeling... | MARK AS OVER ANNOTATED | Summary: The partner is UniProtKB:Q96P50, ACAP3 - a close paralog of ACAP2 with the same BAR-PH-ArfGAP-ANK architecture and the same ARF6 preference. This is not a paralog transfer error; it is a measured physical interaction, recovered here by BioPlex 3.0 affinity-purification mass spectrometry, and independently by OpenCell (the PMID:35271311 row below). UniProt promotes it to an INTERACTION line with NbExp=4, so the pair is very likely real. It is nevertheless recorded under a term that conveys no function. Hetero-oligomerisation through the shared N-terminal BAR domain would be the obvious mechanism, and would matter - it could let two ARF6 GAPs with different recruitment logic act as one unit - but nothing about the interface, the stoichiometry or any functional consequence has been tested, and AP-MS cannot license a dimerisation molecular function. Marked over-annotated rather than modified for that reason; proposing GO:0046982 here would be inventing a mechanism the data do not contain. Reason: Bare protein binding from a proteome-scale AP-MS interactome. The ACAP2-ACAP3 pair is probably real (two orthogonal screens, UniProt NbExp=4) but functionally untested, and the term as recorded contributes nothing to understanding either protein. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC Q15057; Q96P50: ACAP3; NbExp=4; IntAct=EBI-4401196, EBI-715726; PMID:33961781 Through affinity-purification mass spectrometry, we have created two proteome-scale, cell-line-specific interaction networks. |
| GO:0005515 protein binding | IPI PMID:35271311 OpenCell: Endogenous tagging for the cartography of human ce... | MARK AS OVER ANNOTATED | Summary: Same interaction, same partner ACAP3 (UniProtKB:Q96P50), independently detected by OpenCell endogenous tagging. Methodologically orthogonal to BioPlex, since it measures protein at near-native expression rather than after overexpression, so it genuinely strengthens the case that the pair exists - the assessment above is about informativeness, not about reality. Still an uninformative term with no functional consequence attached. OpenCell also pulled CAPZB with ACAP2, which is suggestive given the actin connection but is a single-shot screen hit and is not annotated. Reason: Second high-throughput recovery of the same ACAP2-ACAP3 interaction under the same uninformative term. Replication supports the interaction but does not make protein binding a useful annotation. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC Q15057; Q96P50: ACAP3; NbExp=4; IntAct=EBI-4401196, EBI-715726; PMID:35271311 Genome-edited cells enable to examine protein function at near-native expression levels |
| GO:0031267 small GTPase binding | IEA GO_REF:0000107 | ACCEPT | Summary: Ensembl Compara projection from UniProtKB:Q6ZQK5, which resolves to mouse Acap2 - the true one-to-one ortholog, not ACAP1 or ACAP3. So this is a clean ortholog transfer, and it is in any case superfluous, because the human protein has its own direct evidence - the Rab35/ACAP2 crystal structure (PDB 6IF3) and the mutagenesis at Asp-721, Arg-727, Met-731 and Asp-756 that weakens or abolishes RAB35 binding. Retained unchanged because GO:0031267 is already the most specific term available. GO:0017137 "Rab GTPase binding" is now a secondary id of GO:0031267, and GO:0031267 has no children at all, so the fact that ACAP2 recognises RAB35 exclusively among roughly sixty mammalian Rabs cannot be stated as a term and needs a has_input extension or a GO-CAM. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q6ZQK5 · mouse Acap2 SUPPORTS TRANSFER Verified as the mouse ortholog (ACAP2_MOUSE), not ACAP1 or ACAP3, and it carries GO:0031267 by IPI in its own right, so the projection has real evidence behind it. The human protein also has direct structural evidence, so the transfer is redundant rather than load-bearing. The accompanying ensembl:ENSMUSP00000154983 token resolves to the same entry. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC -!- SUBUNIT: Interacts (via KANK domains) with RAB35 (GTP-bound form); the PMID:25694427 Centaurin-β2 is the only Rab35-binding protein reported thus far that exclusively recognizes Rab35 and does not recognize any of the other 59 Rabs identified in mammals |
| GO:0005515 protein binding | IPI PMID:30905672 Rab35/ACAP2 and Rab35/RUSC2 Complex Structures Reveal Molecu... | MODIFY | Summary: This one should not stay as protein binding. The partner is UniProtKB:Q15286, RAB35, and the interaction is about as well characterised as a protein-protein interaction gets - a high-resolution crystal structure of the complex (PDB 6IF3, 1.50 A, ACAP2 residues 601-770), RAB35 engaging the terminal ankyrin repeat and a following extended helix, four point mutants that graded the binding, and exclusive specificity for RAB35 among the mammalian Rab family. Recording that as "protein binding" discards all of it. Two replacements are proposed rather than one because they capture different, both-verifiable facts. GO:0031267 records the identity of the partner as a small GTPase. GO:0030742 records the nucleotide dependence - ACAP2 binds the GTP-loaded form, which is what makes it an effector rather than a generic interactor, and, unusually, it binds without acting on it, since ACAP2 has no GAP activity towards RAB35. Reason: A structurally resolved, nucleotide-state-dependent effector interaction with a named small GTPase is being recorded under the least informative term in the ontology. GO:0031267 names the partner class and GO:0030742 captures the GTP dependence that defines the effector relationship. Proposed replacements: small GTPase binding GTP-dependent protein binding Supporting Evidence: PMID:30905672 In the Rab35/ACAP2 complex structure, Rab35 binds to the terminal ankyrin repeat and a C-terminal extended file:human/ACAP2/ACAP2-uniprot.txt CC -!- DOMAIN: The ANK domains are required for interaction with RAB35. file:human/ACAP2/ACAP2-uniprot.txt CC (ARF6). Doesn't show GAP activity for RAB35 (PubMed:30905672). |
| GO:0005886 plasma membrane | IDA PMID:30905672 Rab35/ACAP2 and Rab35/RUSC2 Complex Structures Reveal Molecu... | ACCEPT | Summary: Direct imaging evidence, and the basis for UniProt's ECO:0000269 Cell membrane statement. The observation is colocalisation with RAB35 at membrane protrusions of HEK293T cells, which is transfected rather than endogenous protein, so the strength is imaging of an overexpressed construct - but that caveat applies to essentially all of the ACAP2 localisation literature, and the endogenous OpenCell dataset independently places the protein in this part of the cell. It is consistent with the mechanism, in which RAB35 recruits ACAP2 to the membrane. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC -!- SUBCELLULAR LOCATION: Cell membrane {ECO:0000269|PubMed:30905672}. file:human/ACAP2/ACAP2-uniprot.txt CC interaction is direct and probably recruits ACAP2 to membranes |
| GO:0001726 ruffle | IDA PMID:11062263 ACAPs are arf6 GTPase-activating proteins that function in t... | ACCEPT | Summary: Kept, and it is the most informative location term ACAP2 carries. The founding paper documented recruitment to PDGF-induced actin-rich dorsal ruffles in fibroblasts, and it did so for the GAP-dead mutant as well as wild type - which is the useful control, because it separates targeting from catalysis. ACAP2 gets to the ruffle whether or not it can inactivate ARF6 there. The images are of overexpressed protein in NIH 3T3 (mouse) cells, so this is an ACAP-family rather than strictly human-endogenous observation, but the ruffle is exactly where ARF6-dependent actin remodelling happens and it fits the rest of the biology. Supporting Evidence: PMID:11062263 The mutant ACAPs, like mutant ASAP1, were efficiently recruited to the actin-rich ruffles |
| GO:0005096 GTPase activator activity | IDA PMID:11062263 ACAPs are arf6 GTPase-activating proteins that function in t... | ACCEPT | Summary: The primary experimental anchor for the whole review, and the best-supported annotation ACAP2 has. Recombinant ACAP2 accelerates GTP hydrolysis on ARF6 in vitro with measured preference over ARF1 and ARF5; the activity is abolished by R442Q, the conserved arginine of the ArfGAP domain; and it is lipid-gated, needing PIP2 together with phosphatidic acid, with half-maximal activity at 10-20 microM PIP2 in 380 microM PA. IDA is the right code here, because this is a direct in vitro assay on purified protein, unlike the cellular experiments in the same paper, which are overexpression-based. One curation observation worth recording, since it bears on whether these IDAs are over-called. PMID:11062263 assayed ACAP1 and ACAP2 side by side in the same figures, and by the paper's own account ACAP1 was the more potent of the two in the protrusion assay. Yet QuickGO returns three annotations from this reference for ACAP2 (GO:0005096, GO:0030029, GO:0001726, all IDA) and zero for ACAP1 (UniProtKB:Q15027). The asymmetry is a curation artifact rather than a biological difference; ACAP2's IDAs are not over-called, they are simply the half of a symmetric experiment that got recorded. Left as GO:0005096 only because no more specific term exists; GO:0008060 "ARF GTPase activator activity" has been merged into it. The measured substrate preference, which is the single most informative biochemical fact about this protein, therefore has to be captured outside the term - which is what the RO:0002233 has_input extension on this row does, mirrored by core_functions[0].substrates. Without that, a downstream consumer of this file would see only "GTPase activator activity" and lose the one fact the assay was designed to establish. Supporting Evidence: PMID:11062263 In vitro, ACAP1 and ACAP2 preferred Arf6 as a substrate, rather than Arf1 and Arf5, more so than did ASAP1. PMID:11062263 [R448Q]ACAP1 and [R442Q]ACAP2 had no detectable activity using Arf6 as a substrate PMID:11062263 ACAP1 and ACAP2 showed similar phospholipid dependencies when Arf6 was used as a substrate. |
| GO:0030029 actin filament-based process | IDA PMID:11062263 ACAPs are arf6 GTPase-activating proteins that function in t... | MODIFY | Summary: The observation is sound but the term is a grandparent-level placeholder. What was actually shown is that ACAP2 suppresses two specific actin structures - PDGF-induced dorsal ruffles in fibroblasts, and ARF6/AlF4-induced actin-rich protrusions in human HeLa cells, where ACAP2 halved the fraction of protruding cells - and that both effects need GAP activity. That is a regulatory relationship, so GO:0032956 regulation of actin cytoskeleton organization states it properly, and the qualifier should become involved_in, since ACAP2 directly performs the regulation rather than merely acting somewhere upstream of it. One evidence-code point. These are overexpression and point-mutant experiments, so IMP would be the correct code for the process rather than IDA. IDA is right for the in vitro GAP assay in the same paper, but not for this cellular phenotype. Flagged rather than actioned, since the action vocabulary here covers terms and not codes. Reason: GO:0030029 is too general and, more importantly, has the wrong role - ACAP2 regulates actin organization from upstream as an ARF6 GAP rather than participating in an actin filament-based process. GO:0032956 with involved_in states the demonstrated relationship. The supporting experiments are overexpression/point-mutant based, so IMP rather than IDA would fit the process. Proposed replacements: regulation of actin cytoskeleton organization Supporting Evidence: PMID:11062263 Overexpression of either ACAP inhibited the formation of PDGF-induced dorsal ruffles PMID:11062263 ACAP2 reduced the number of cells generating protrusions by 50%. PMID:11062263 The effects of ACAP1 and ACAP2 were dependent on GAP activity. |
| GO:0032456 endocytic recycling | ISS GO_REF:0000024 | ACCEPT | Summary: Curator-judged transfer from UniProtKB:Q5FVC7, which resolves to rat Acap2 - the true ortholog, so this is a legitimate ortholog transfer and not a paralog leak. It is also supported in human cells independently of the transfer, since ACAP2 is recruited to the peripheral tubular membranes on which ARF6 is activated to drive recycling back to the plasma membrane, and ARF6 is the GTPase whose recycling function ACAP2 terminates. RAB35, which recruits ACAP2, is itself a rate-limiting regulator of the fast recycling pathway. Core. Why this is kept as participation while the actin rows are moved to a regulation term, since the two look superficially alike. The difference is how many steps separate ACAP2 from the process. For actin, ACAP2 never touches the cytoskeleton; ARF6-GTP recruits downstream effectors which remodel actin, so ACAP2 is at least two steps removed and the observable is a modulation of rate and extent - the signature of a regulator. For recycling, the ARF6 GTPase cycle is not upstream machinery, it is part of the trafficking step itself: nucleotide hydrolysis has to occur on the recycling tubule for the round to complete, and ACAP2 is the enzyme that performs it there. A GEF or GAP that turns over the GTPase of a trafficking step is conventionally annotated as involved_in that step, in the same way a Rab GAP is treated as part of its transport process rather than as its regulator. The direction of effect also differs - ACAP1, the paralog, is required for transferrin-receptor recycling, so a signed regulation term would be hard to justify here in a way it is not for the actin phenotype, where the observed effect is unambiguously suppression. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q5FVC7 · rat Acap2 SUPPORTS TRANSFER Verified as ACAP2_RAT, the true ortholog. Its UniProt FUNCTION block is identical to the human one (ARF6 GAP, no RAB35 GAP activity), and it carries GO:0032456 by IMP in its own right, so this is a transfer of experimental evidence rather than of another inference. Supporting Evidence: PMID:11062263 In addition, ACAP1 and ACAP2 were recruited to peripheral, tubular membranes, where activation of Arf6 occurs to allow membrane recycling back to the plasma membrane. |
| GO:0016020 membrane | HDA PMID:19946888 Defining the membrane proteome of NK cells. | MARK AS OVER ANNOTATED | Summary: Detection in a bulk membrane-fraction proteome of the YTS natural-killer-like line. Two problems compound. First, the term. GO:0016020 is one step below the cellular-component root and adds nothing over the plasma-membrane and endosome-membrane annotations ACAP2 already carries. Second, the dataset. The authors explicitly stripped peripheral membrane proteins before analysis, and ACAP2 is a peripheral membrane protein by UniProt's own description, so its recovery is more likely carry-over than evidence of tight membrane residence - consistent with their own estimate that only about 40% of the 1843 identifications were plausible membrane proteins. Reason: Root-adjacent CC term from a bulk membrane-fraction proteome whose authors report that only ~40% of identifications were plausible membrane proteins and whose protocol was designed to remove peripheral membrane proteins - the class ACAP2 belongs to. Fully subsumed by the specific membrane annotations. Supporting Evidence: PMID:19946888 Isolated membranes were treated with reagents that have been reported to remove peripheral membrane proteins. PMID:19946888 approximately 40% of the identified proteins were predicted as plausible membrane proteins file:human/ACAP2/ACAP2-uniprot.txt CC Endosome membrane {ECO:0000250}; Peripheral membrane protein |
| GO:0010008 endosome membrane | ISS GO_REF:0000024 | ACCEPT | Summary: Ortholog transfer from rat Acap2 (UniProtKB:Q5FVC7), verified as the true ortholog. Redundant with the IBA and SubCell rows for the same term but independently sound, and corroborated in human cells by the ARF6 tubular-endosome localisation. Core location. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q5FVC7 · rat Acap2 SUPPORTS TRANSFER True ortholog; carries GO:0010008 IDA in its own right. Supporting Evidence: file:human/ACAP2/ACAP2-uniprot.txt CC Endosome membrane {ECO:0000250}; Peripheral membrane protein |
| GO:1990090 cellular response to nerve growth factor stimulus | ISS GO_REF:0000024 | KEEP AS NON CORE | Summary: Transferred from rat Acap2 (UniProtKB:Q5FVC7), and the rat evidence is solid and species-appropriate. The RAB35-ACAP2-ARF6 relay was worked out in NGF-stimulated rat PC12 cells, where RAB35 accumulates on ARF6-positive endosomes after NGF and recruits ACAP2 there, and where binding-deficient mutants of either partner fail to support neurite outgrowth. So the transfer is a legitimate ortholog transfer of a real finding. Kept as non-core rather than accepted because it describes one deployment of ACAP2 rather than what ACAP2 is. NGF responsiveness is a property of PC12 cells and of TrkA-expressing neurons; the same ACAP2 module is used in FcgammaR phagocytosis in macrophages and as a brake on oligodendrocyte differentiation, neither of which involves NGF. The invariant is RAB35-dependent recruitment and ARF6 inactivation; the NGF context is downstream of which cell type the protein happens to be in. Reason: A real, correctly transferred finding, but cell-type-specific context rather than a core property of the protein - ACAP2 performs the same RAB35-to-ARF6 relay in settings that have nothing to do with NGF. Propagation Review Root cause: NO FAILURE NON CORE Failure modes: CONTEXT OR TISSUE MISMATCH Sources checked: UniProtKB:Q5FVC7 · rat Acap2 SUPPORTS TRANSFER True ortholog, and the rat annotation is GO:1990090 IDA, resting on rat PC12 experiments, so the donor evidence is both experimental and species-appropriate. The context, not the transfer, is what makes this non-core. Supporting Evidence: PMID:22344257 We found that Rab35 accumulates at Arf6-positive endosomes in response to nerve growth factor (NGF) stimulation and that centaurin-β2 is recruited to the same compartment in a Rab35-dependent manner. |
| GO:0032013 negative regulation of ARF protein signal transduction | IMP PMID:11062263 ACAPs are arf6 GTPase-activating proteins that function in t... | NEW | Summary: Proposed new annotation, and the main gap this review closes. GOA gives ACAP2 four biological processes - actin cytoskeleton organization (IBA), actin filament-based process (IDA), endocytic recycling (ISS) and cellular response to NGF (ISS) - and not one of them says that ACAP2 switches ARF6 off, which is the entirety of its characterised biology. The molecular function is annotated; the process the function serves is not. GO:0032013 exists, is not obsolete, and is exactly the right term. Support in human cells is direct. In HeLa cells, co-expression of ACAP2 with ARF6 suppressed the ARF6-dependent protrusions that AlF4 induces, halving the fraction of protruding cells, and the suppression required GAP activity. Three later studies describe the same relationship in words, calling ACAP2 an ARF6 GTPase-activating protein recruited by GTP-RAB35, and the GTPase-activating protein that switches off Arf6 activity. IMP rather than IDA, deliberately. The cellular evidence is overexpression plus a GAP-dead point mutant, not a direct assay at native levels. The in vitro GAP assay in the same paper is the IDA, and it supports the molecular function row, not this process row. Reason: ACAP2's only characterised job is to terminate ARF6 signalling, and GOA records no biological process to that effect. GO:0032013 is a live, non-obsolete term that states it, and it is the process under which the existing GO:0005096 molecular function should sit. Supporting Evidence: PMID:11062263 we used HeLa cells to test for an effect of the ACAPs on Arf6 function in vivo PMID:11062263 Coexpression of Arf6 with either ACAP inhibited the formation of protrusions in response to AlF4 treatment PMID:11062263 The effects of ACAP1 and ACAP2 were dependent on GAP activity. PMID:24600047 We show that Rab35 and its effector, ACAP2, a GTPase-activating protein that switches off Arf6 activity, negatively regulate oligodendrocyte morphological differentiation. |
| GO:0035091 phosphatidylinositol binding | IDA PMID:25853217 Human ACAP2 is a homolog of C. elegans CNT-1 that promotes a... | NEW | Summary: Proposed new annotation. ACAP2 has no lipid-binding annotation at all, which is a conspicuous absence for a protein whose catalytic activity is lipid-gated and which carries a PH domain at 266-361. Two independent strands support it. First the enzymology - ACAP2's GAP activity is not merely faster in the presence of lipid but requires it, with half-maximal activity at 10-20 microM PIP2 in 380 microM phosphatidic acid, which UniProt records as an ACTIVITY REGULATION statement with an ECO:0000269 tag. Second a direct binding measurement - PMID:25853217 reports that ACAP2 has an identical phosphoinositide-binding pattern to that of tCNT-1, the caspase-cleaved form of its C. elegans counterpart. The generic parent term is used on purpose. That cache entry is abstract-only, so the individual phosphoinositide species and the assay format cannot be read off it, and GO:0035091 is defined to cover phosphatidylinositol and its phosphorylated derivatives, which is exactly the resolution the evidence supports. A child such as GO:0005546 (PI(4,5)P2 binding) would be a guess, because the PIP2 dependence in PMID:11062263 is a property of the enzymatic activity, and attributing it to direct PH-domain binding would be reading a mechanism into a domain name. Independent, weaker corroboration comes from a PI(3)P affinity proteomics screen that recovered ACAP2, though that screen pulled 681 proteins and is not on its own a basis for annotation. To be precise about what carries the annotation and what only motivates it: the IDA rests solely on PMID:25853217, the one source that reports a binding measurement. The lipid dependence of catalysis, from PMID:11062263 and the UniProt ACTIVITY REGULATION block, establishes that ACAP2 works only in the presence of PIP2 and phosphatidic acid, which is why the absence of any lipid-binding annotation is conspicuous - but a lipid that gates an enzyme need not be bound by that enzyme, so those sources are context and are deliberately not listed as support below. Reason: A protein whose GAP activity is strictly lipid-dependent and which has a direct phosphoinositide-binding measurement in the literature carries no lipid-binding annotation. The parent term is chosen because the specific lipid species cannot be established from the available evidence. The annotation rests on PMID:25853217 alone; the activity-regulation evidence is motivation, not support. Supporting Evidence: PMID:25853217 We report here that ACAP2, a homolog of C. elegans CNT-1, has a pro-apoptotic function and an identical phosphoinositide-binding pattern to that of tCNT-1, despite not being an apparent target of caspase cleavage. |
Loading supporting content…
Download this section (compressed HTML)Q: GO merged the nine substrate-specific GTPase activator terms (including GO:0008060 ARF GTPase activator activity) into GO:0005096, and merged GO:0017137 Rab GTPase binding into GO:0031267, leaving neither term with substrate-specific children. For a protein like ACAP2, whose entire identity is that of an ARF6 GAP which is not a RAB35 GAP but binds RAB35 exclusively, is the intended replacement a has_input annotation extension, a GO-CAM, or both - and should GOA rows carrying GO:0005096 without such an extension be treated as incomplete?
Q: Should ACAP2 carry an annotation to FcgammaR-mediated phagocytosis or phagocytic cup formation? PMID:22045739 shows GTP-RAB35-dependent recruitment of ACAP2 to the phagocytic cup, but the abstract states only that the work is in macrophages and the cached record is abstract-only, so the species could not be established and no transfer was proposed here. A curator with the full text could settle whether this is a human annotation, a mouse annotation transferable by ISS, or neither.
Q: Is the pro-apoptotic activity of ACAP2 reproducible, and is it separable from its ARF6-GAP function? PMID:25853217 reports that ACAP2 knockdown blocks 5-fluorouracil-induced apoptosis in human cancer cells and that ACAP2 has the same phosphoinositide-binding pattern as tCNT-1, yet also notes that ACAP2 is not a caspase substrate - so the mechanism established for the worm protein cannot be the mechanism here. Nothing in the annotation record reflects this activity.
Q: What is the functional meaning of the ACAP2-ACAP3 interaction? It is recovered by two orthogonal proteome-scale datasets and UniProt records NbExp=4, and both proteins are ARF6 GAPs with N-terminal BAR domains, but no interface, stoichiometry or functional consequence has been reported. Do ACAP paralogs hetero-oligomerise through their BAR domains, and does that pool their recruitment determinants?
Q: PANTHER assigns Drosophila blow (blown fuse) to PTHR23180 (CENTAURIN/ARF) as subfamily SF399 even though the protein has only a PH domain - no BAR, no ArfGAP catalytic domain, no ankyrin repeats - and its experimental actin annotation comes from myoblast fusion. Should a family member lacking the family's catalytic domain contribute to IBA support sets for process terms of the catalytic members?
Experiment: Express and purify the isolated ACAP2 PH domain (residues 266-361) and the full-length protein, and measure binding to a defined phosphoinositide panel by liposome flotation or surface plasmon resonance rather than by protein-lipid overlay, which is prone to false positives. In parallel, mutate the canonical basic residues of the PH-domain binding pocket and ask whether the PIP2 dependence of GAP activity on ARF6 is lost while catalysis on a soluble substrate is retained. This would convert the current inference, that a PH domain plus lipid-dependent activity implies PH-mediated lipid binding, into a measurement, and would let the proposed GO:0035091 annotation be replaced with a specific child term.
Hypothesis: ACAP2's PH domain is the phosphoinositide sensor that gates its ARF6-GAP activity, and the lipid it reads is PI(4,5)P2.
Type: in vitro lipid binding and enzymology
Experiment: Knock out or degron-tag ACAP2 in a human cell line and measure the ARF6-GTP fraction by an effector pull-down or a GTP-ARF6 conformational biosensor, alongside quantification of membrane ruffling and protrusion. Rescue with wild-type ACAP2 and with the R442Q GAP-dead mutant. The existing cellular evidence for the process is entirely gain-of-function (overexpression plus a point mutant), which is why the proposed GO:0032013 annotation is IMP from an overexpression experiment; a loss-of-function result would upgrade it and would also settle whether the paralogs ACAP1 and ACAP3 buffer the loss.
Hypothesis: Endogenous ACAP2 restrains ARF6 activity at the plasma membrane, so that loss of ACAP2 increases the ARF6-GTP pool and enhances ARF6-dependent protrusion.
Type: loss-of-function cell biology
Experiment: Co-express tagged ACAP2 and ACAP3 BAR domains (ACAP2 residues 1-226) and test for heterodimer formation by size-exclusion chromatography with multi-angle light scattering, then validate in cells with a split-luciferase or proximity-ligation assay on endogenous protein. Ask whether a BAR-interface mutant loses the interaction without losing ARF6-GAP activity, and whether ACAP3 co-recruits with ACAP2 to GTP-RAB35 sites. A positive result would replace the two uninformative protein-binding rows with a defensible dimerisation function.
Hypothesis: ACAP2 and ACAP3 form a heterodimer through their N-terminal BAR domains.
Type: biophysical reconstitution and cell-based interaction assay
Experiment: Compare recruitment of wild-type ACAP2 with the RAB35-binding-deficient mutant (Asp-721/Arg-727/Met-731/Asp-756 substitutions) and with a PH-domain lipid-binding mutant, in cells where PI(4,5)P2 can be acutely depleted with a recruitable phosphatase. If either lesion alone abolishes membrane targeting, the two inputs are ANDed; if recruitment survives loss of one, RAB35 binding and lipid binding are redundant recruitment routes. This distinguishes the coincidence-detector model asserted in the core-function description from the simpler linear RAB35-recruitment model.
Hypothesis: RAB35 binding and phosphoinositide binding are jointly required for ACAP2 recruitment, making ACAP2 a coincidence detector rather than a simple RAB35 effector.
Type: live-cell recruitment with acute lipid depletion
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)