AGGF1

Existing Annotations Review

GO Term	Evidence	Action	Reason
GO:0001525 angiogenesis	IBA GO_REF:0000033	ACCEPT	Summary: Sound propagation. The one protein donor resolves to the zebrafish orthologue aggf1, which carries its own IMP for this exact term, and the human protein has independent IDA evidence for the same biology. Reason: Resolving the WITH/FROM column settles this row rather than leaving it as a family-level guess. ZFIN:ZDB-GENE-060929-836 is zebrafish aggf1 (Q08CK9, 774 aa, TrEMBL) - a true orthologue, not a paralogue - and QuickGO shows it holds GO:0001525 itself by IMP from PMID:23197652, where morpholino knockdown produced defective vasculogenesis and angiogenesis. The propagation therefore lands at the same term the donor holds, not above it, and no heterogeneous clade is forcing a broad call. Human AGGF1 independently carries GO:0045766 by IDA from the CAM assay in PMID:14961121, so the target has its own evidence for the same process. Core: whatever else AGGF1 does in the nucleus, promotion of new vessel growth is a real, independently replicated activity of the secreted protein. Propagation Review Root cause: NO FAILURE CORE Sources checked: ZFIN:ZDB-GENE-060929-836 · Danio rerio aggf1 (Q08CK9, TrEMBL, 774 aa) - orthologue SUPPORTS TRANSFER Holds GO:0001525 by IMP from PMID:23197652 in its own right. Unreviewed (TrEMBL) entry, so its protein NAME is an automatic label, but its GO evidence is a real curated IMP - those are two different things. PANTHER:PTN002749826 · PANTHER tree node, not a protein NOT RELEVANT The inference structure rather than an evidence source. Supporting Evidence: PMID:23197652 Overexpression of AGGF1 led to increased angiogenesis and increased lumen diameter of veins, whereas knockdown of AGGF1 expression resulted in defective vasculogenesis and angiogenesis.
GO:0001570 vasculogenesis	TAS PMID:15905966 Biomedicine and diseases: the Klippel-Trenaunay syndrome, va...	KEEP AS NON CORE	Summary: The term is defensible, but on evidence this annotation does not cite. Its actual source is a review written by the authors of the discovery paper, restating that paper - which measured angiogenesis, not vasculogenesis. Reason: PMID:15905966 is a survey of vascular-anomaly genes, not a primary study, and what it says about AGGF1 is a restatement of PMID:14961121: that the gene encodes a potent angiogenic factor and that KTS-associated mutations enhance that activity. Vasculogenesis - de novo vessel formation from angioblasts - is a distinct process from angiogenesis and was not assayed in either paper. The term nevertheless survives, because zebrafish aggf1 knockdown was later shown to cause defective vasculogenesis (PMID:23197652) and to act upstream of scl, fli1 and etsrp in hemangioblast specification (PMID:24277077). So this row is right for reasons published seven years after it was made, and it rests on organismal knockdown in fish rather than on anything measured on the human protein. Kept, but not core - and it needs re-sourcing rather than simply keeping: the term should be retained while the reference and code are changed to PMID:23197652 with ISO (or an IBA from the zebrafish orthologue), because TAS to a review that discusses angiogenesis cannot support a vasculogenesis annotation on the human protein. That is why PMID:15905966 is marked MISCITED in the reference list: the term survives, this particular (term, evidence, reference) triple does not. Supporting Evidence: PMID:15905966 AGGF1 encodes a potent angiogenic factor, and KTS-associated mutations enhance angiogenic activity of AGGF1, defining 'increased angiogenesis' as one molecular mechanism for the pathogenesis of KTS. PMID:23197652 knockdown of AGGF1 expression resulted in defective vasculogenesis and angiogenesis
GO:0001938 positive regulation of endothelial cell proliferation	IDA PMID:14961121 Identification of an angiogenic factor that when mutated cau...	ACCEPT	Summary: Direct assay of purified AGGF1 on human endothelial cells, extended and given a mechanism by later work from the same group. One of the best-supported statements about this protein. Reason: Purified recombinant AGGF1 applied to HUVECs increased thymidine uptake from 524 to 680 counts (P = 0.007, n = 8) - a direct assay of the protein's activity on the annotated organism's cells, which is exactly what IDA asserts. The result has since been extended and given a mechanism: AGGF1 treatment of HUVECs activates FAK, Src and AKT through integrin alpha5beta1 (PMID:34551592), and blocking autophagy abolishes AGGF1-driven HUVEC proliferation (PMID:27513923). Both of those papers share this one's senior author, so they are extensions rather than independent replications - the activity is well characterised but has been characterised almost entirely by one group. It is nonetheless a direct assay of purified protein on human cells, which is what the code asserts. Core. Supporting Evidence: PMID:14961121 VG5Q can bind to endothelial cells and promote cell proliferation, suggesting that it may act in an autocrine fashion.
GO:0003676 nucleic acid binding	IEA GO_REF:0000002	MODIFY	Summary: A domain-name mapping, not a measurement - no reviewed human G-patch protein holds this term on anything but IEA. The conclusion nonetheless survives on AGGF1's own RNA-immunoprecipitation data, at the more informative child term. Reason: Two separate questions, and they come out differently. First, what is the InterPro2GO inference worth? Querying the whole reference class shows it is worth nothing on its own: of the eleven reviewed human G-patch proteins the QuickGO query returns in full - NKRF, GPKOW, RBM17, SUGP1, CMTR1, GPATCH1, GPATCH4, RBM5, PINX1, ZGPAT and AGGF1 - every single one carries GO:0003676 by IEA and by nothing else, and not one carries it with any experimental code. The mapping is a uniform output of IPR000467, not anybody's finding. That is also what the structural literature now says about the domain: the DHX15 crystal structure shows the G-patch acting as an extended tether that clamps the helicase's C-terminal domains to its catalytic core, a helicase-activation peptide rather than an autonomous nucleic-acid-binding module. Second, is the conclusion true of AGGF1 anyway? Yes, and on better evidence: RNA immunoprecipitation shows AGGF1 binding the lncRNA NEAT1 (PMID:35608889), and six of the ten other G-patch proteins in the panel carry GO:0003723 RNA binding with real experimental codes while none of them has experimental support for the generic parent. So the right action is to move the row to GO:0003723 RNA binding, which is both more informative and the term the class's own measurements land on - not to remove it. The domain itself is intact and canonical: AGGF1 matches 8 of 8 columns conserved at 80 percent or better across 34 annotated G-patch domains in 33 reviewed human proteins, and all 8 are identical to the corresponding residues of NKRF, whose G-patch is the one resolved in complex with DHX15 - including the invariant glycines AGGF1 G631 (NKRF G563) and AGGF1 G639 (NKRF G570), each conserved in 100 percent of the panel. Propagation Review Root cause: TERM SCOPING PROBLEM Failure modes: GRANULARITY MISMATCH Sources checked: InterPro:IPR000467 · G-patch domain signature SUPPORTS SOURCE BUT NOT TARGET The signature is genuinely present and intact in AGGF1 - this is not a false domain call. What does not follow is the activity: the mapping IPR000467 -> GO:0003676 is applied uniformly across the family and is never corroborated experimentally in any member, while the structural work on the domain describes a helicase-activation tether rather than a nucleic-acid-binding module. Proposed replacements: RNA binding Supporting Evidence: PMID:35608889 AGGF1 interacts with NEAT1, which may be another possible mechanism underlying PMID:32179686 conformation across the helicase surface. It tethers the catalytic core to the PMID:32179686 ATPase activity of DHX15 are increased when G-patch is bound. file:human/AGGF1/AGGF1-bioinformatics/RESULTS.md 11 carry GO:0003676 nucleic acid binding;
GO:0005515 protein binding	IPI PMID:14961121 Identification of an angiogenic factor that when mutated cau...	MODIFY	Summary: The best-supported interaction AGGF1 has - three orthogonal assays in the original paper and an independent co-IP in human endothelial cells 21 years later - and the least informative term available for it. Reason: TNFSF12/TWEAK was pulled out of a yeast two-hybrid screen with AGGF1 as bait, then confirmed by GST pull-down with recombinant proteins and by co-immunoprecipitation from cells, and the two proteins co-localise in HUVECs. It is the only AGGF1 partner in GOA with orthogonal in vitro and in vivo support, and notably it is absent from IntAct entirely - all 742 IntAct records for Q8N302 come from high-throughput screens, none of which found it. It was independently re-confirmed by co-immunoprecipitation in human retinal microvascular endothelial cells in PMID:39905000. Bare protein binding states none of this. TWEAK is a secreted cytokine, and AGGF1 is a secreted protein that acts on endothelium in what the discovery paper proposed to be an autocrine fashion, so GO:0019955 cytokine binding captures the relationship and is definitionally correct. GO:0043120 tumor necrosis factor binding was considered and rejected on its definition, which names TNF specifically as a proinflammatory cytokine produced by monocytes and macrophages - that does not cover TNFSF12 - even though GO itself places GO:0038057 TNFSF11 binding beneath it. That inconsistency is raised separately in suggested_questions, together with a proposal for the missing TNFSF12 binding term. GO:0019955 is also defensible by usage and not only by its definition: asking each of the 92 human holders whether its own record carries GO:0038023 signaling receptor activity gave 83 receptors and 9 non-receptors, and among the non-receptors is BGN (biglycan) - a secreted extracellular proteoglycan that binds secreted cytokines, which is the same situation as AGGF1 and TNFSF12. Proposed replacements: cytokine binding Supporting Evidence: PMID:14961121 We also demonstrate a direct interaction of VG5Q with another secreted angiogenic factor, TWEAK (also known as TNFSF12). PMID:39905000 AGGF1 (V5GQ) is known to promote angiogenesis by interacting with TNFSF12 (TWEAK)
GO:0005515 protein binding	IPI PMID:16189514 Towards a proteome-scale map of the human protein-protein in...	KEEP AS NON CORE	Summary: A real two-method interaction with BLOC1S6, but with no functional follow-up in twenty years and no shared biology to point at. Kept as uninformative rather than refuted. Reason: Q9UL45 resolves to reviewed canonical BLOC1S6 (pallidin, 172 aa), a BLOC-1 subunit involved in biogenesis of lysosome-related organelles - no published connection to AGGF1's vascular or nuclear biology. The evidence is better than a single screen: one Rual CCSB-HI1 yeast two-hybrid hit here plus an anti-tag co-immunoprecipitation in human HEK293T cells from BioPlex (PMID:33961781, the second GOA row for this partner), i.e. two independent methods. That is enough not to call it over-annotation, and not enough to say anything more specific than protein binding. Both proteins carry coiled-coil regions - AGGF1 at residues 18-88 - which is the commonest route to a reproducible but physiologically inert two-hybrid pair, so the interaction is recorded without being promoted.
GO:0005515 protein binding	IPI PMID:31515488 Extensive disruption of protein interactions by genetic vari...	KEEP AS NON CORE	Summary: MAB21L3 from a CCSB-pipeline two-hybrid assay. Real and reproduced in a second method, but the partner is itself uncharacterised, so nothing more specific can be said. Reason: Q8N8X9 is reviewed canonical MAB21L3 (362 aa), a mab-21-like protein with no FUNCTION comment in UniProt at all - an uncharacterised member of a nucleotidyltransferase-fold family. AGGF1-MAB21L3 appears in five IntAct records across three publications, but those collapse to two independent experiments: the CCSB two-hybrid pipeline (this row and PMID:32296183, logged as array plus prey-pooling plus validated two hybrid, which is one screen counted three ways) and an anti-tag co-immunoprecipitation in human U2OS cells from PMID:40205054. Two methods make it a genuine interaction; an uncharacterised partner makes it uninformative. No more specific term is available, and inventing a functional interpretation for a partner nobody has characterised would be the error this review is trying to avoid elsewhere.
GO:0005515 protein binding	IPI PMID:32296183 A reference map of the human binary protein interactome.	MODIFY	Summary: The single most mechanistically predicted partner a G-patch protein can have, sitting under the least informative term in the ontology. The evidence is one two-hybrid screen, so the term is refined without asserting any activity. Reason: O43143 is reviewed canonical DHX15 (795 aa), the DEAH-box RNA helicase that is the archetypal client of G-patch cofactors: the crystal structure of DHX15 bound to the NKRF G-patch shows the motif tethering the helicase's C-terminal domains to its catalytic core and increasing its RNA affinity, helicase and ATPase activity (PMID:32179686). AGGF1 has an intact G-patch - 8 of 8 panel-conserved columns retained, every one identical to the corresponding NKRF residue - so of all the high-throughput partners in this gene's record, this is the one with a prior. The evidence, however, is a single HuRI yeast two-hybrid experiment: all three IntAct records for this pair come from PMID:32296183 and are the sub-methods two hybrid array, two hybrid prey pooling approach and validated two hybrid of one screen. MODIFY to GO:0017151 DEAD/H-box RNA helicase binding therefore restates the measured fact more informatively and adds nothing: it says AGGF1 binds a DEAH-box helicase, which is what was observed, and specifically does not say that AGGF1 activates it. A peer-reviewed 2026 survey of the family still lists AGGF1's interacting helicase as unknown (PMID:42052570). Whether the G-patch functions as a DHX15 activator is filed as a suggested experiment, not annotated. On the term itself: DHX15 is a DEAH-box rather than a DEAD-box helicase, but GO has no DEAH-specific binding term at all - an ontology search for DEAH returns zero terms, and GO:0017151 has no comment, no synonyms and no children - so it is the only term available for this superfamily. Its ten human annotations name four partners: DDX5, and via POT1 the RecQ DNA helicases BLM and WRN, so GO already applies it well outside a strict DEAD-box reading. Proposed replacements: DEAD/H-box RNA helicase binding Supporting Evidence: PMID:32179686 ATPase activity of DHX15 are increased when G-patch is bound. file:human/AGGF1/AGGF1-bioinformatics/RESULTS.md \| O43143 DHX15 \| 3 \| 1 — HuRI Y2H (PMID:32296183) as array + prey-pooling + "validated two hybrid" \|
GO:0005515 protein binding	IPI PMID:32296183 A reference map of the human binary protein interactome.	KEEP AS NON CORE	Summary: The same MAB21L3 interaction as the PMID:31515488 row, from the HuRI screen. Same verdict. Reason: This is the HuRI instance of the AGGF1-MAB21L3 pair; PMID:31515488 is the variant-interactome instance and PMID:40205054 the orthogonal co-IP. Kept for the same reason as that row: two independent methods support the interaction, but MAB21L3 has no characterised function, so protein binding is the most that can be said. Recorded per-row rather than merged because GOA records them separately and each carries its own provenance.
GO:0005515 protein binding	IPI PMID:32296183 A reference map of the human binary protein interactome.	KEEP AS NON CORE	Summary: One two-hybrid screen, but a nuclear partner that coheres with the nuclear biology the rest of this gene's GO record omits. Kept, not promoted. Reason: Q96EZ8 is reviewed canonical MCRS1 (462 aa), a nucleoplasmic and nucleolar protein and a component of the NSL and INO80 complexes - and, like AGGF1, an FHA-domain protein (its FHA sits at 363-419). The evidence is thin: three IntAct records, all sub-methods of the single HuRI screen, no orthogonal assay, no functional follow-up. It is kept rather than marked over-annotated for a reason that matters beyond this row. Three of the four partners HuRI found for AGGF1 - DHX15, MCRS1 and FBXO28 - carry GO:0005634 nucleus with experimental evidence (EXP or IDA), on a gene whose entire GOA localisation record is cytoplasm plus extracellular region. The fourth, MAB21L3, has no cellular-component annotation in GOA at all and no SUBCELLULAR LOCATION comment in UniProt, so it neither supports nor weakens the point. The interaction evidence in AGGF1's own GO record already points at a compartment its localisation rows do not mention, which is corroborating context rather than noise.
GO:0005515 protein binding	IPI PMID:32296183 A reference map of the human binary protein interactome.	KEEP AS NON CORE	Summary: One two-hybrid screen with a nuclear SCF substrate receptor. Weakest of the HuRI set, kept on the same nuclear-coherence grounds. Reason: Q9NVF7 is reviewed canonical FBXO28 (368 aa), the substrate-recognition subunit of an SCF E3 ubiquitin ligase that localises to the nucleus and kinetochore. Three IntAct records, all sub-methods of the one HuRI screen; no orthogonal assay and no follow-up anywhere. On its own this is the profile of an unreplicated screen hit. It is kept rather than called over-annotation because it is topologically coherent - a nuclear protein binding a protein that is demonstrably nuclear (PMID:33069768, PMID:35608889) - and because an E3 substrate receptor is a plausible route to the proteasomal control of AGGF1 levels that nobody has looked for. Nothing more specific is asserted.
GO:0005515 protein binding	IPI PMID:33961781 Dual proteome-scale networks reveal cell-specific remodeling...	KEEP AS NON CORE	Summary: The BioPlex co-immunoprecipitation instance of the BLOC1S6 interaction. It is what makes the PMID:16189514 two-hybrid row credible. Reason: Anti-tag co-immunoprecipitation in human HEK293T cells, an orthogonal method to the yeast two-hybrid that produced the other BLOC1S6 row. Two methods from two laboratories is the reason neither row is marked over-annotated. Still uninformative: no shared pathway, no functional experiment, and coiled-coil-mediated association remains the likeliest explanation.
GO:0005515 protein binding	IPI PMID:40205054 Multimodal cell maps as a foundation for structural and func...	KEEP AS NON CORE	Summary: The orthogonal co-immunoprecipitation instance of the MAB21L3 interaction, in human U2OS cells. Carries an unflagged Publisher Correction that does not affect it. Reason: Anti-tag co-immunoprecipitation in human U2OS cells, which is what lifts AGGF1-MAB21L3 above the two-hybrid-only partners. The partner is still uncharacterised, so protein binding stands. Checking this reference for post-publication notices - which a publication-type search does not find, because a Publisher Correction has to be read from CommentsCorrections on the cited article's own record - turned up PMID:41039152. Reading it shows the correction is a subscript typo in a Methods loss function, Ty/Sy to Tx/Sx, with no effect on the interaction data. Recorded because the check was run, not because it changes anything.
GO:0005576 extracellular region	IBA GO_REF:0000033	ACCEPT	Summary: A self-referential IBA - the only protein in the WITH/FROM column is AGGF1 itself - recording a PAINT curator's judgement that extracellular activity is core for this node. Valid, and correct. Reason: UniProtKB:Q8N302 in the WITH/FROM field is this gene. That is not circular evidence; it is how PAINT records a curator deciding that a function seen in one member is the node's core function. The underlying evidence is AGGF1's own IDA from PMID:14961121, and it holds up: competitive ELISA on media from untransfected HUVEC matrigel cultures detects secreted AGGF1, metabolic labelling recovers it from the medium, purified extracellular AGGF1 engages a cell-surface receptor (PMID:34551592), and AGGF1 protein is measurable in human vitreous fluid (PMID:39905000). The is_active_in qualifier is the right one - this is where the protein does its receptor-ligand work, not merely where it can be found. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q8N302 · AGGF1 itself (self-referential IBA) SUPPORTS TRANSFER Self-reference, which for an IBA is a curator judgement rather than a defect. Its own evidence is the GO:0005576 IDA from PMID:14961121. PANTHER:PTN008590572 · PANTHER tree node, not a protein NOT RELEVANT The inference structure rather than an evidence source.
GO:0005576 extracellular region	IDA PMID:14961121 Identification of an angiogenic factor that when mutated cau...	ACCEPT	Summary: Secretion measured three ways on endothelial cells, and confirmed by an independent group's detection of AGGF1 in human vitreous fluid. Unusual in that the protein has no signal peptide. Reason: The evidence is a competitive ELISA on conditioned medium from untransfected HUVECs plated on matrigel - endogenous protein, not an overexpression artefact - supported by 35S metabolic labelling of transfected HUVECs and by immunostaining showing redistribution to the cell periphery and outside the cell as tube formation begins. AGGF1 has no FT SIGNAL feature in UniProt, and the authors say so, proposing a non-classical route like FGF-2's. That is a mechanistic gap, not a reason to doubt the localisation. Three later studies act on extracellular AGGF1, two of them applying purified protein to cells - but PMID:34551592 and PMID:27513923 share this paper's senior author, so they are continuations rather than replications. The independent evidence is PMID:39905000, from a different group, which measures AGGF1 protein in the vitreous of patients with proliferative diabetic retinopathy. That one result is what makes the extracellular pool safe to call core. Supporting Evidence: PMID:14961121 VG5Q shows strong expression in blood vessels and is secreted as vessel formation is initiated.
GO:0005576 extracellular region	IEA GO_REF:0000044	ACCEPT	Summary: A faithful transfer of UniProt's curated Secreted subcellular location, which itself rests on the PMID:14961121 IDA reviewed above. Reason: GO_REF:0000044 maps UniProt's SUBCELLULAR LOCATION vocabulary onto GO, and SL-0243 Secreted is asserted in AGGF1's entry with ECO:0000269\|PubMed:14961121 - the same experimental basis as the IDA row. The mapping is correct and the source is sound, so this is redundant rather than wrong. Worth noting what UniProt's own Note adds and the GO term cannot express: the protein is cytoplasmic in microvascular endothelial cells and is secreted only when endothelial tube formation is initiated, so the extracellular pool is conditional on an angiogenic stimulus. Supporting Evidence: file:human/AGGF1/AGGF1-uniprot.txt CC -!- SUBCELLULAR LOCATION: Cytoplasm {ECO:0000269\|PubMed:14961121}. Secreted
GO:0005737 cytoplasm	IDA PMID:14961121 Identification of an angiogenic factor that when mutated cau...	ACCEPT	Summary: Immunostaining of endogenous AGGF1 in human microvascular endothelial cells. Correct, and now known to be only half the picture. Reason: Endogenous AGGF1 was detected in the cytoplasm of HMVECs by immunostaining with a peptide-raised antibody, and the cytoplasmic pool is where the protein sits before an angiogenic stimulus triggers its export. The row is right. What it is not is complete: AGGF1 also occupies the nucleus, where it coats paraspeckles and regulates splicing, and it shuttles between the two compartments under the control of its FHA domain and an internal nuclear localisation region (PMID:33471274). That is proposed as a NEW GO:0005634 row rather than as a correction to this one. Supporting Evidence: PMID:14961121 VG5Q protein was detected mostly in the cytoplasm and around the nuclei of HMVECs
GO:0005737 cytoplasm	IEA GO_REF:0000044	ACCEPT	Summary: Faithful transfer of UniProt's curated Cytoplasm location, resting on the same IDA. Redundant but correct. Reason: SL-0086 Cytoplasm is asserted in AGGF1's UniProt entry with ECO:0000269\|PubMed:14961121, so this electronic row and the IDA row above have one experimental source between them. No objection; it simply adds no information beyond the IDA. Supporting Evidence: file:human/AGGF1/AGGF1-uniprot.txt CC -!- SUBCELLULAR LOCATION: Cytoplasm {ECO:0000269\|PubMed:14961121}. Secreted
GO:0007155 cell adhesion	IDA PMID:14961121 Identification of an angiogenic factor that when mutated cau...	ACCEPT	Summary: Measured directly - endothelial cells adhere to AGGF1-coated wells - and since given a receptor. Kept as-is, with the mechanism proposed additively. Reason: In calcein-AM adhesion assays HUVECs bound significantly better to wells coated with purified AGGF1 than to BSA, and trypsinised cells lost that binding, which led the authors to predict a proteinaceous receptor. The prediction was borne out: integrin alpha5beta1 is that receptor, neutralising antibodies against alpha5, beta1 and alpha5beta1 block HUVEC adhesion to AGGF1, and GST-AGGF1 precipitates integrin alpha5 from HUVEC extracts (PMID:34551592). The BP row is therefore correct and is left alone; what is missing is the molecular function underneath it, proposed here as a NEW GO:0005178 integrin binding row rather than as a replacement. Additive, not corrective. Supporting Evidence: PMID:14961121 VG5Q can bind to endothelial cells and promote cell proliferation, suggesting that it may act in an autocrine fashion.
GO:0042802 identical protein binding	IPI PMID:22365833 Dynamic protein-protein interaction wiring of the human spli...	KEEP AS NON CORE	Summary: A yeast two-hybrid self-hit, and the screen it came from is itself informative: AGGF1 was included because it copurifies with spliceosomes. Reason: Hegele and colleagues screened an all-versus-all two-hybrid matrix of the proteins that copurify with human spliceosomes, and AGGF1's presence in that bait set is independent corroboration - from a laboratory with no stake in AGGF1's vascular biology - that the protein is spliceosome-associated. The self-interaction itself is credible because a two-hybrid self-hit requires the DNA-binding and activation-domain fusions genuinely to associate, and AGGF1 carries a 70-residue coiled coil at 18-88 that would do it. It is not core: no structure, no stoichiometry, no function assigned to the dimer, and the more interesting claim from a related line of work - that AGGF1 drives liquid-liquid phase separation into nuclear condensates (PMID:35608889) - is a different and unannotated phenomenon. Supporting Evidence: PMID:22365833 More than 200 proteins copurify with spliceosomes, the compositionally dynamic RNPs catalyzing pre-mRNA splicing.
GO:0042802 identical protein binding	IPI PMID:25416956 A proteome-scale map of the human interactome network.	KEEP AS NON CORE	Summary: A second, independent two-hybrid self-hit from the HI-II-14 interactome. Together with the PMID:22365833 row this is real. Reason: Two-hybrid array self-interaction from a completely separate screen and laboratory to PMID:22365833. Two independent two-hybrid self-hits, plus a 70-residue coiled coil, is reasonable support that AGGF1 self-associates. Non-core because nothing has been built on it - no interface, no stoichiometry, no measured consequence.
GO:0042802 identical protein binding	IPI PMID:39251607 Systematic identification of post-transcriptional regulatory...	MARK AS OVER ANNOTATED	Summary: An artefact of the assay. This row comes from a proximity-labelling screen in which AGGF1 is the bait, and a biotin-ligase-fusion bait labels itself whether or not it dimerises. Reason: Querying IntAct for the underlying record settles it: the AGGF1-AGGF1 entry from PMID:39251607 has detectionMethod proximity-dependent biotin identification and experimentalRoleA bait with experimentalRoleB prey, and AGGF1 is the bait in all 635 proximity records in the dataset. A BirA fusion biotinylates its own polypeptide constitutively, so recovering the bait in its own streptavidin pull-down is an expected property of the method and carries no information about self-association. This is a different situation from the two two-hybrid self-hits above, which are kept: there, the DNA-binding and activation-domain fusions have to associate for a signal to appear. The same screen is also what makes AGGF1 look promiscuous - 634 of the 742 IntAct records for this protein, and most of its 695 distinct partners, come from this one experiment. Marked over-annotated rather than removed because the term itself is supported by the other two rows; it is this row's evidence that is vacuous. Supporting Evidence: file:human/AGGF1/AGGF1-bioinformatics/RESULTS.md IntAct gives AGGF1 `experimentalRoleA=bait` in all 635 proximity records
GO:0045766 positive regulation of angiogenesis	IDA PMID:14961121 Identification of an angiogenic factor that when mutated cau...	ACCEPT	Summary: Purified AGGF1 induces new vessels in the chick CAM assay as potently as VEGF, and knockdown blocks endothelial tube formation. The gene's best-replicated biological activity. Reason: Two directions in the same paper: purified recombinant protein applied to the chorioallantoic membrane produced new radiating vessels comparable to VEGF, and two independent siRNAs plus an antisense oligonucleotide inhibited tube formation by HMVECs and HUVECs. Everything since has reinforced it - the receptor and signalling pathway are mapped (PMID:34551592), autophagy is required downstream (PMID:27513923), and AGGF1 is used therapeutically to drive angiogenesis in ischaemia models. Core. One caveat belongs on the record but does not touch this row: the disease claim in the same paper, that the E133K variant is a gain-of-function KTS mutation, has since been refuted (PMID:16443853), and the variant is classified Benign/Likely benign in ClinVar. The angiogenic activity of the wild-type protein was measured separately and stands. Supporting Evidence: PMID:14961121 VG5Q protein acts as a potent angiogenic factor in promoting angiogenesis, and suppression of VG5Q expression inhibits vessel formation.
GO:0048471 perinuclear region of cytoplasm	IDA PMID:14961121 Identification of an angiogenic factor that when mutated cau...	ACCEPT	Summary: Faithful to what was seen in 2004 - endogenous AGGF1 around the nuclei of human microvascular endothelial cells - and probably the earliest sighting of the nuclear pool this gene's GO record still lacks. Reason: The observation is exactly as described: AGGF1 was detected mostly in the cytoplasm and around the nuclei of HMVECs, and TWEAK co-localised with it around the nuclei of HUVECs. The row is correct for what the images showed. It is worth recording that later work resolved the same region differently - AGGF1 is nuclear in HCT116 cells where it co-localises with p53 (PMID:33069768), coats paraspeckles inside the nucleus (PMID:35608889), and moves from nucleus to cytoplasm in endothelial cells as angiogenesis begins (PMID:33471274). Kept as an accurate record of a perinuclear pool, with the nuclear pool proposed additively rather than as a correction. Supporting Evidence: PMID:14961121 VG5Q protein was detected mostly in the cytoplasm and around the nuclei of HMVECs
GO:0005634 nucleus	IDA PMID:33069768 Angiogenic factor AGGF1 acts as a tumor suppressor by modula...	NEW	Summary: Proposed. AGGF1 has no nucleus annotation anywhere in GOA, against three separate lines of evidence - and against three of the four partners its HuRI two-hybrid screen returned, which carry GO:0005634 experimentally themselves. Reason: This is the largest single gap in AGGF1's GO record. The whole localisation set is cytoplasm, extracellular region and perinuclear region of cytoplasm, all traceable to one 2004 immunostaining experiment. Since then: immunostaining in human HCT116 cells places AGGF1 in the nucleus co-localised with p53 (PMID:33069768, which also notes that the nuclear pool had been reported by the authors' earlier work and confirmed by another group); AGGF1 forms an outer rim around paraspeckles, which are by definition subnuclear (PMID:35608889); and a separate laboratory - headed by the first author of the original discovery paper - maps a nuclear localisation region to residues 260-288 and shows AGGF1 moving from nucleus to cytoplasm in endothelial cells during angiogenesis (PMID:33471274). Independently, three of the four partners the HuRI two-hybrid screen found for AGGF1 - DHX15, MCRS1 and FBXO28 - carry GO:0005634 nucleus experimentally, while the fourth, MAB21L3, has no cellular-component annotation anywhere; and AGGF1 was included in a spliceosome-copurifying bait set (PMID:22365833). Nothing in GOA reflects any of this. One of those lines is independent of the rest, which matters because the others are not: PMID:33168501 (senior author Liao S, no authorship on any paper from the dominant group) shows AGGF1 co-localising with gammaH2AX by immunofluorescence in human HCT116 cells - the same cell line as PMID:33069768, from a different group. It is cited here for localisation only; this review makes no DNA-repair claim. IDA is the right code: the HCT116 imaging is a direct assay in human cells, though it uses tagged constructs, and a curator should note that the cached records for PMID:35608889 and PMID:33471274 are abstract-only and do not name their cell systems. Supporting Evidence: PMID:33069768 HA-tagged p53 expression plasmids, and immunostaining showed that AGGF co-localized with p53 in the nucleus PMID:33471274 Angiogenic factor with G-patch and FHA domains 1 (AGGF1) exhibits a dynamic PMID:35608889 of paraspeckles that induces paraspeckle formation, forms an outside rim of PMID:33168501 Immunofluorescence assay showed the co-localization of AGGF1 and γH2AX.
GO:0042382 paraspeckles	IDA PMID:35608889 Identification and characterization of a special type of sub...	NEW	Summary: Proposed. AGGF1 is reported as a structural component of paraspeckles, forming an outer rim around the NONO/PSF/PSPC1/NEAT1 core, and GO records none of it. Reason: RNA-FISH combined with immunostaining places AGGF1, NONO and NEAT1_2 together in 20.58 percent of NEAT1_2-positive paraspeckles, with AGGF1 wrapped around the core rather than inside it, and knockdown of NEAT1_2 reduces the number and size of AGGF1-positive paraspeckles. The term is exactly what GO:0042382 describes - discrete subnuclear bodies in the interchromatin nucleoplasmic space. This is the specific form of the nucleus proposal above, and it matters for curation because paraspeckle membership is a functional statement about AGGF1's nuclear role rather than a bare compartment call. Two caveats a curator should weigh. The cached record is abstract-only and does not name the cell system, so the human-cell assumption rests on the study targeting the human protein and the 23 kb human NEAT1_2 isoform. And the finding comes from the same laboratory that discovered AGGF1, so it would benefit from independent replication - which is why it is proposed alongside, not instead of, the better-corroborated nucleus row. Supporting Evidence: PMID:35608889 of paraspeckles that induces paraspeckle formation, forms an outside rim of PMID:35608889 AGGF1 interacts with NEAT1, which may be another possible mechanism underlying
GO:0000381 regulation of alternative mRNA splicing, via spliceosome	IMP PMID:40035560 Angiogenic factor AGGF1 is a general splicing factor regulat...	NEW	Summary: Proposed. AGGF1 silencing in human endothelial cells changes alternative splicing of 436 genes, and the gene has no splicing annotation of any kind. Reason: Full-length transcriptome sequencing of human pulmonary artery endothelial cells with AGGF1 silenced identified AGGF1 as a general splicing factor regulating alternative splicing of 436 genes, with the SRSF6 exon-3 skipping event characterised in detail and shown to be the functionally relevant one: full-length SRSF6 rescues the proliferation, migration and tube-formation defects of AGGF1-silenced cells. IMP is the correct code - the perturbation is siRNA knockdown, and the cells are human. A second study reports that AGGF1 decreases the exon skipping/inclusion ratio in a CD44 minigene (PMID:35608889), giving a second, independent assay of the same activity. This is also the process the protein's domain architecture predicts: OCRE plus FHA plus G-patch is the architecture of RBM5/RBM10-class splicing regulators, and AGGF1 appears in a spliceosome-copurifying bait set (PMID:22365833). Neither paper produced a GO annotation on any gene product. Supporting Evidence: PMID:40035560 genes, and showed that AGGF1 is a general splicing factor regulating AS of 436 PMID:40035560 skipping of exon 3 that produces the full-length SRSF6 protein, an PMID:35608889 splicing as it decreases the exon skipping/inclusion ratio in a CD44 model.
GO:0005178 integrin binding	IPI PMID:34551592 Receptor and Molecular Mechanism of AGGF1 Signaling in Endot...	NEW	Summary: Proposed. The receptor for secreted AGGF1 was identified in 2021 and GO records nothing from that paper - not on AGGF1, not on ITGA5, not on anything. Reason: This is the molecular function underneath AGGF1's existing cell-adhesion and endothelial-proliferation BP rows, and it has been directly measured in human cells. GST-AGGF1, but not GST alone, precipitates integrin alpha5 from HUVEC extracts in a dose-dependent manner, and of ten integrin-subunit neutralising antibodies tested only anti-alpha5, two anti-beta1 antibodies and an anti-alpha5beta1 antibody blocked HUVEC adhesion to AGGF1. Serial deletion mapping localises the activity to residues 604-613 (FQRDDAPAS) - a motif that lies outside both named domains of the protein, which is worth noting given the gene's name. IPI with UniProtKB:P08648 (human ITGA5) is the right form: the pull-down detected alpha5 specifically, while the beta1 requirement rests on siRNA and antibody blocking rather than on direct binding. The integrin alpha7 interaction reported in smooth muscle (PMID:35202649, PMID:37081014) is deliberately not used here - those experiments were done in mouse MOVAS-1 cells and in mice, so they would support ISO, not IPI, in human. Supporting Evidence: PMID:34551592 GST pull-down assays showed that GST-AGGF1, but not control GST, precipitated integrin-α5 from HUVEC protein extracts PMID:34551592 Of the many integrin antibodies tested (α1, α2, α3, α4, α5, α6, α5β1, αV, β1, and β3), only one anti-α5 antibody, two different anti-β1 antibodies (β1a and β1b) and one anti-integrin α5β1 antibody disrupted the binding of HUVECs to AGGF1
GO:0048018 receptor ligand activity	IDA PMID:34551592 Receptor and Molecular Mechanism of AGGF1 Signaling in Endot...	NEW	Summary: Proposed. AGGF1's entire molecular-function record is one InterPro IEA and a set of bare binding rows, yet the defining activity of the secreted protein - engaging a receptor and changing its signalling output - has been measured. Reason: GO:0048018 is defined as the activity of a gene product that interacts with a receptor to effect a change in the activity of the receptor, and that is precisely what was shown. Purified AGGF1 applied to HUVECs increases phosphorylation of FAK Tyr397, Src Tyr416 and AKT Ser473 in a time-dependent manner; two independent siRNAs against ITGA5 and two against ITGB1 abolish the AKT response; and the effect requires the 604-613 motif, since a 15-mer peptide containing it blocks AGGF1 function while the same peptide with RDD changed to RGD does not. The specificity control is the strongest part: cells with low integrin alpha5/beta1 do not respond, and HEK293T, 3T3-L1 and HeLa cells adhere to AGGF1 without activating AKT, so adhesion and signalling are separable. IDA rather than IMP: the primary assay is application of purified protein to human cells, a direct assay of the protein's own activity, with the siRNA experiments serving as controls on the receptor rather than on AGGF1. Supporting Evidence: PMID:34551592 Western blot analysis showed that AGGF1 treatment increased the phosphorylation levels of FAK (Tyr397), Src (Tyr416), and AKT (Ser473), the downstream signaling events for integrins, in HUVECs in a time-dependent manner PMID:34551592 A 40-mer or 15-mer peptide containing the angiogenic domain blocks AGGF1 function, however, a 15-mer peptide containing a single amino acid mutation from -RDD- to -RGD- (a classical RGD integrin-binding motif) failed to block AGGF1 function.
GO:0051897 positive regulation of phosphatidylinositol 3-kinase/protein kinase B signal transduction	IDA PMID:34551592 Receptor and Molecular Mechanism of AGGF1 Signaling in Endot...	NEW	Summary: Proposed. AKT activation is the most consistently reported downstream consequence of AGGF1 across cell types, species and disease models - though almost all of those reports come from one group - and GO has no record of it. Reason: In HUVECs, purified AGGF1 raises phospho-AKT Ser473, and the response is lost when ITGA5, ITGB1, PTK2/FAK or SRC is knocked down or Src is inhibited, placing AGGF1 upstream of an integrin-FAK-Src-AKT cascade. The same dependency was reproduced in human brain microvascular endothelial cells in the same study. Independent support: AGGF1 activates the p110alpha catalytic and p85alpha regulatory subunits of PI3K leading to AKT, GSK3beta and p70S6K activation (PMID:27522498), and constitutively active AKT rescues the loss of venous identity caused by aggf1 knockdown in zebrafish (PMID:23197652). Independent corroboration exists and is worth naming, because it is what makes the summary's "almost all of those reports come from one group" a checkable statement rather than a hedge: PMID:29885663, from a group with no authorship overlap in the senior position with the dominant one (senior author Zhang JH), shows that recombinant AGGF1 raises PI3K and phospho-Akt and that the effect is abolished by the PI3K inhibitor LY294002. That work is in Sprague-Dawley rats, so it corroborates the pathway without bearing on the evidence code here. IDA is appropriate for the human evidence because the assay is purified protein applied to human cells; the rat, zebrafish and mouse results are cited as corroboration, not as the evidence for this row. Supporting Evidence: PMID:34551592 Two siRNAs for ITGA5 (encoding integrin α5) significantly reduced the increased phosphorylation level of AKT (Ser473) after AGGF1 stimulation compared with control scramble siRNA in HUVECs PMID:34551592 Altogether, these data suggest that AGGF1 activates AKT signaling via integrin α5β1, FAK and Src.
GO:0010508 positive regulation of autophagy	IDA PMID:27513923 Angiogenic Factor AGGF1 Activates Autophagy with an Essentia...	NEW	Summary: Proposed. Purified AGGF1 induces autophagy in human endothelial cells dose-dependently, and autophagy is required for every angiogenic readout AGGF1 drives. The paper produced no GO annotation on any gene. Reason: HUVECs treated with purified AGGF1 show concentration-dependent, saturable accumulation of LC3-II and loss of p62; GFP-LC3 puncta rise from 10 percent of GFP-positive cells to 31 percent; and electron microscopy shows autophagosomes directly. The causal direction is established both ways - pharmacological block (bafilomycin A1, chloroquine) and ATG5 knockdown each abolish AGGF1-induced proliferation, migration, tube formation and aortic-ring sprouting - so this is not an incidental correlate but a required step in the protein's angiogenic activity. IDA because the assay is purified protein applied to human cells at defined doses; the Aggf1 heterozygous mouse heart data in the same paper would be ISO and are not what this row rests on. Supporting Evidence: PMID:27513923 (A) Western blot analysis showing that AGGF1 at different concentrations induces autophagy (increased LC3-II/β-actin and LC3-II/I; decreased p62 expression) in HUVECs under a normoxic condition for 4 h (n = 3/group). PMID:27513923 These data suggest that autophagy is required for AGGF1-mediated angiogenesis.
GO:0010595 positive regulation of endothelial cell migration	IDA PMID:27513923 Angiogenic Factor AGGF1 Activates Autophagy with an Essentia...	NEW	Summary: Proposed. Migration is one of the three endothelial behaviours AGGF1 drives, it is measured by two assays in human cells, and only the proliferation component of that trio reached GO. Reason: GOA records AGGF1 promoting endothelial proliferation (GO:0001938) and angiogenesis (GO:0045766) but not migration, although the migration data are as direct as either. In HUVECs, AGGF1 treatment significantly induced migration in wound-healing scratch assays and drew 1.64-fold more cells through Transwell filters than IgG control, and both effects were abolished by autophagy inhibition or ATG5 knockdown. The same activity is independently reported for the 604-613 motif in PMID:34551592, where deleting the angiogenic domain eliminates AGGF1-driven endothelial migration. IDA: purified protein applied to human cells. Supporting Evidence: PMID:27513923 Wound-healing scratch assays for migration showed that AGGF1 treatment significantly induced HUVEC migration, but this effect was blocked by Baf or CQ (Fig 2B). PMID:34551592 The angiogenic domain is required for EC adhesion and migration, capillary tube formation, and AKT activation.
GO:0003723 RNA binding	IDA PMID:35608889 Identification and characterization of a special type of sub...	NEW	Summary: Proposed alongside the GO:0003676 MODIFY, so that AGGF1's only direct measurement of RNA binding is recorded with its own experimental code rather than inherited by an electronic row. Reason: RNA immunoprecipitation recovers the lncRNA NEAT1 with AGGF1. That is the only experiment anyone has done on this protein that measures nucleic-acid binding at all, and as the record stands it produces no annotation. The GO:0003676 row is separately modified to GO:0003723 because the generic term is uninformative, but that MODIFY alone would leave the experimental result behind an IEA code whose stated provenance - InterPro2GO from IPR000467 - does not map to GO:0003723 and could not have produced it. This row fixes that: same term, correct code, correct reference. A curator who prefers to leave the electronic row at its generic parent can do so and still keep the specific term, because this row stands on its own evidence. Two caveats a curator should carry forward. RNA immunoprecipitation cannot distinguish direct contact from recruitment through the paraspeckle core proteins AGGF1 also binds (NONO, PSF, HNRNPK), and AGGF1 has no recognisable RRM or KH domain - so the corresponding knowledge gap is recorded against the core function rather than being silently resolved by this row. And the cached record is abstract-only and does not name the cell system, so the human-cell assumption rests on the study targeting the human protein and the 23 kb human NEAT1_2 isoform. Supporting Evidence: PMID:35608889 AGGF1 interacts with NEAT1, which may be another possible mechanism underlying

Gene Description

Proposed New Ontology Terms

TNFSF12 binding

Existing Annotations Review

Core Functions

References

Suggested Questions for Experts

Suggested Experiments

Deep Research

Affinage

📚 Additional Documentation

Notes

Bioinformatics Results

📄 View Raw YAML