AGFG2 (O95081) — ArfGAP Catalysis vs. Trafficking Cofactor

Focus type: function_assignment
Hypothesis slug: arfgap-catalysis-versus-trafficking-cofactor
Question: Does human AGFG2/HRBL directly possess GTPase-activator (ArfGAP) activity, warranting a molecular-function annotation, or is its established role a non-catalytic vesicle-trafficking cofactor?

Executive Judgment

Verdict: UNRESOLVED (leaning "unsupported as an experimental MF; not refuted at the sequence level").

There is no direct biochemical evidence that AGFG2 accelerates GTP hydrolysis on any ARF (or other) GTPase. The molecular-function claim of catalytic ArfGAP activity therefore cannot be asserted as experimentally supported. At the same time, it cannot be refuted on sequence grounds: AGFG2 retains an intact Arf-GAP zinc-finger domain and, critically, the catalytic arginine finger (Arg75) that the transition-state structure of ASAP3–Arf6 identifies as the key catalytic residue (PMID20510928). The one divergence highlighted by the seed — Thr89 where ASAP3 has Asp484 — is not, by itself, diagnostic of inactivity, because the GIT-family ArfGAPs lack that same aspartate-homologous site (PMID23433073) yet display measured ARF GAP activity (PMID10788515).

The only functional data on AGFG2 (PMID34369554) is a knockdown secretion phenotype (von Willebrand factor exocytosis from Weibel–Palade bodies), which is not a GTPase-hydrolysis assay. The "GAP-dead" R75Q construct in that paper still promoted secretion, but those data are marked "not shown." Thus the evidence-supported role of AGFG2 is a trafficking cofactor, while catalytic ArfGAP activity remains an untested computational inference. Absent assays are not equivalent to functional refutation.

Most important caveat: the central factual question — has anyone ever purified AGFG2 (or its paralog AGFG1) and measured GAP activity? — appears to be answered "no" in the retrievable literature. The MF annotation therefore rests on domain homology (IEA/ISS), not on experiment.

Evidence Matrix

GO Curation Implications (leads — require curator verification)

Curation baseline (verified from UniProt O95081, iteration 2): The existing MF term GO:0005096 "GTPase activator activity" is annotated IEA:InterPro (from ArfGAP domain IPR001164 / PF01412) — i.e., pure domain-based inference, no experimental support. AGFG2's curated BP terms are all IBA:GO_Central (acrosome assembly, intermediate filament organization, spermatid nucleus differentiation); the experimentally demonstrated regulated-exocytosis role (PMID34369554, vWF/Weibel-Palade body) is NOT yet in GO. The paralog AGFG1 (P52594) also carries GO:0005096 only as IEA:InterPro, while its experimental annotations are nucleus / cytoplasmic vesicle / RNA binding / mRNA export — so neither AGFG paralog has an experimentally supported GAP annotation.

Molecular Function — "ARF GTPase activator activity" (GO:0008060) / "GTPase activator activity" (GO:0005096):
- Recommended action: retain only as computational/IEA (or ISS) — do NOT upgrade to an experimental (EXP/IDA) MF annotation. No direct assay supports catalysis for AGFG2; the current term is already IEA:InterPro, which is appropriately conservative.
- Do not remove the term solely on the Thr89 (Asp→Thr) substitution: the GIT precedent (PMID10788515 + PMID23433073) shows loss of that aspartate is not diagnostic of inactivity, and the catalytic arginine finger (Arg75) is retained.
- If an experimental MF annotation currently exists based on PMID34369554, it is too strong — that paper provides a cellular phenotype, not GAP activity; consider re-scoping to a BP annotation.

Biological Process (evidence-supported, core):
- "regulated exocytosis" / "von Willebrand factor secretion" / "Weibel-Palade body" (e.g., GO:0060627 regulation of vesicle-mediated transport; GO:0002576-adjacent secretion terms; GO:0033093 Weibel-Palade body) — supported by IMP/IGI from PMID34369554 (siRNA knockdown). This is the best-supported functional annotation.

Cellular Component: AGFG family carries FG-repeat / nucleoporin-related regions; localization annotations should follow direct evidence, not be inferred from the MF claim.

GO Decision Table (leads — require curator verification)

Structural provenance of the catalytic scaffold (iteration 3)

Direct inspection of UniProt sequences shows the AGFG2 ArfGAP zinc finger (res 47–70) has four cysteines at 47, 50, 67, 70 with inter-cysteine gaps [3, 17, 3] — identical spacing to canonical active ARFGAP1 (Cys 22, 25, 42, 45; gaps [3, 17, 3]; the C‑x2‑C‑x16‑C‑x2‑C ArfGAP zinc knuckle). The catalytic arginine finger sits immediately C‑terminal to the last cysteine in a conserved motif — AGFG2 SGLLRGL (R75), ARFGAP1 SGRHRGL (R50), ASAP3 SGVHREL (R469). The zinc-coordinating scaffold and arginine finger of AGFG2 are therefore structurally canonical and intact, so the MF claim cannot be refuted on structural grounds; divergence is confined to accessory positions (Thr89↔Asp484 and a non-conserved tryptophan).

Bottom line for the curator: the ArfGAP MF term is best treated as non-core / homology-only, and the gene's evidence-backed core is a trafficking cofactor in ARF-regulated regulated secretion. Avoid defaulting to "protein binding."

Mechanistic Scope

• Immediate molecular activity tested: acceleration of GTP hydrolysis on an ARF substrate by the AGFG2 Arf-GAP zinc-finger domain (an MF, GO:0008060).
• What is actually demonstrated: a downstream cellular phenotype — loss of AGFG2 reduces stimulated vWF exocytosis (PMID34369554). This is separated from direct catalysis by at least two inferential steps (knockdown → pathway effect; not enzyme kinetics).
• Structural inference: catalysis in this family is driven by a GAP-supplied arginine finger acting on an Arf-supplied glutamine (PMID20510928); AGFG2 retains the arginine finger, so the catalytic machinery is not obviously broken, but this is inference, not measurement.

Conflicts and Alternatives

1. Non-catalytic cofactor model (favored by available data). The R75Q "GAP-dead" construct reportedly still promoting secretion (PMID34369554, data not shown) suggests the trafficking role may not require catalysis. Catalysis and cofactor roles are not mutually exclusive, but only the cofactor role has functional support.
2. Paralog / family carry-over. ArfGAP MF annotations propagate by InterPro domain (Arf-GAP, PF01412) across the whole family; AGFG2's term may be database carry-over rather than gene-specific evidence.
3. Evolutionary divergence argument (PMID23433073). Predicts AGFG catalytic weakening (Asp + Trp divergence). This is a prediction, self-limited by the authors' own GIT counterexample.
4. Substrate ambiguity. Even if AGFG2 is an active GAP, the target Arf isoform is unknown (GIT proteins prefer ARF6 over ARF1; AGFG2 substrate untested).
5. AGFG1 context. AGFG1/HRB was historically characterized as an HIV-Rev cofactor / FG-repeat nucleoporin-like protein, not as an assayed GAP — reinforcing that the family's "ArfGAP" label is largely domain-based.

Knowledge Gaps

Discriminating Tests

1. Single-turnover GAP assay with recombinant AGFG2 Arf-GAP domain (res ~27–153) against ARF1/ARF5/ARF6 loaded with GTP; include ASAP3 (positive) and a catalytic-Cys-mutant AGFG2 (negative). Direct readout of the MF claim.
2. Structure-guided mutagenesis: AGFG2 R75Q (arg-finger dead) and T89D (restore Asp) tested in the GAP assay to determine whether Thr89 limits activity.
3. Rescue specificity in cells: re-express WT vs R75Q vs zinc-finger-dead AGFG2 in AGFG2-depleted endothelial cells and quantify vWF exocytosis — resolves catalytic vs scaffold requirement.
4. Comparative alignment/AlphaFold3 modeling of AGFG2–ARF6 to check whether the catalytic geometry (arg finger, Ca²⁺/Asp pocket) is preserved.

Curation Leads (require curator verification)

• Candidate references to attach:
• PMID34369554 — AGFG2 in regulated vWF exocytosis (BP/IMP lead; note "GAP-dead" data not shown).
• PMID20510928 — defines ArfGAP catalytic residues (arg finger; Asp not essential) — supports NOT refuting on Thr89.
• PMID10788515 — GIT ArfGAP activity without the Asp-homologous site — counters over-reliance on the Asp substitution.
• PMID23433073 — evolutionary divergence of AGFG catalytic residues — context for weakening (but self-caveated).
• Candidate GO actions:
• MF GO:0008060 / GO:0005096 → keep as IEA/ISS only; flag as unverified; do not assign experimental evidence codes.
• BP → add/retain regulated exocytosis / vWF secretion / Weibel-Palade body with IMP from PMID34369554.
• Suggested curator questions:
• Is any existing experimental MF ArfGAP annotation on AGFG2 traceable to a real assay, or is it phenotype-based (PMID34369554)? If the latter, re-scope to BP.
• Should the review add a "no direct catalytic assay available" note to prevent paralog-driven over-annotation?
• Suggested experiments: the single-turnover GAP assay + R75Q/T89D mutants above.

Limitations

• Analysis is literature- and sequence-based; no wet-lab data were generated.
• The pivotal R75Q rescue in PMID34369554 is "data not shown," so its interpretation is provisional.
• The pairwise NW alignment reproduces the seed's MAFFT coordinates but is a two-sequence global alignment, not a full family MSA; conclusions about the arginine finger and Asp positions are robust because they are anchored on the ASAP3 crystal structure numbering.

Artifacts

• OpenScientist final report
• OpenScientist final report