AIGR Gene Hypothesis Deep Research — Final Report

Target: S. pombe gaa1 (UniProt Q9US48) · GO:0004197 (cysteine-type endopeptidase activity)

Focus type: computational_prediction · Hypothesis slug: prediction-cysteine-endopeptidase Prediction source: GO-GPT (via BioReason-Pro) · Reference context: doi:10.64898/2026.03.19.712954

Summary

The GO-GPT/BioReason-Pro pipeline predicts cysteine-type endopeptidase activity (GO:0004197) for Schizosaccharomyces pombe gaa1 (Q9US48), a subunit of the GPI transamidase (GPI-T) complex that attaches glycosylphosphatidylinositol (GPI) anchors to the C-termini of substrate proteins. This report evaluates, independently, whether gaa1 is itself the catalytic cysteine protease of the complex, or whether the catalytic transamidation/cysteine-protease chemistry resides in a different subunit — Gpi8 (mammalian PIG-K) — with gaa1 acting as a non-catalytic recognition/scaffolding component.

The prediction is REFUTED. Every tier of evidence — classical yeast genetics and biochemistry, active-site mutagenesis, human and fungal cryo-EM structures, and direct protein-domain/annotation queries — places the catalytic cysteine-protease/transamidase activity in Gpi8/PIG-K, a caspase/legumain-related Peptidase C13 enzyme with a defined Cys–His(–Asn) catalytic triad. gaa1 carries no peptidase domain, no annotated catalytic residues, and no EC number; its role is to recognize/bind the GPI anchor and stabilize the pentameric complex. The prediction is therefore a textbook subunit-level misassignment: it attributes the catalytic chemistry of one subunit (Gpi8) to a different, non-catalytic subunit (gaa1).

Two practical points sharpen the curation guidance. First, gaa1 does not currently carry GO:0004197, so the prediction would be a new, incorrect molecular-function addition rather than a correction of an existing annotation — the correct action is simply to reject it. Second, even the most generous alternative reading (that Gaa1/GPAA1 assists the amide-bond-forming second half-reaction) does not rescue the prediction: the only peptidase-like fold anywhere in the Gaa1 family is a Zn/M28 metallopeptidase superfamily hit (seen in human GPAA1, absent even from S. pombe gaa1), which is mechanistically incompatible with cysteine-type catalysis.

Key Findings

Finding 1 — gaa1 is a non-catalytic GPI-T subunit; the catalytic cysteine protease is Gpi8/PIG-K

The GPI transamidase reaction proceeds through a covalent thioester (carbonyl) intermediate: an active-site cysteine attacks the carbonyl of the substrate's ω-site residue after cleavage of the C-terminal GPI-attachment signal peptide, then the GPI anchor's ethanolamine amino group displaces the enzyme to form the amide-linked anchor. The seed hypothesis asks which subunit supplies that catalytic cysteine.

Classical genetics/biochemistry answers this directly. Ohishi et al. (2000) showed that GAA1-knockout cells fail to accumulate the carbonyl intermediate, but the essential cysteine and histidine residues that generate the intermediate belong to Gpi8p, not Gaa1p: "cysteine and histidine residues of Gpi8p, which are conserved in members of a cysteine protease family, are essential for generation of a carbonyl intermediate... Gpi8p is a catalytic component that cleaves the GPI attachment signal peptide" (PMID:10793132). The GAA1 requirement is therefore indirect — it reflects the need for an intact, functional complex, not a catalytic activity intrinsic to Gaa1.

Meyer et al. (2000) independently mapped Gpi8p's active site and identified it as a caspase-related cysteine protease whose catalytic Cys199 and His157 (S. cerevisiae numbering) are essential: "gpi8 alleles mutated at Cys199 or His157 are nonfunctional" (PMID:10727241). This directly localizes the catalytic cysteine/histidine to Gpi8p.

Structural biology confirms this across kingdoms. The human GPI-T cryo-EM structure places catalysis in PIGK (the Gpi8 ortholog): "The PIGK subunit functions as the catalytic component, in which we identified a C206-H164-N58 triad that is critical for the transamination reaction" (PMID:35165458); GPAA1 — the direct ortholog of S. pombe gaa1 — is non-catalytic. The fungal GPI-T cryo-EM structure reinforces this, describing "the dynamic accommodation of catalytic subunit Gpi8" and casting Gaa1/Gab1 as GPI-anchor-binding/scaffold components (PMID:41085069).

Finally, direct UniProt provenance shows the domain asymmetry cleanly. S. pombe gaa1 (Q9US48) carries only Pfam PF04114 "Gaa1" (IPR007246) — no Peptidase_C13 domain, no annotated active site, no EC number, no cysteine-endopeptidase MF term — whereas S. pombe gpi8 (Q9USP5) carries Pfam PF01650 "Peptidase_C13" (IPR001096), annotated ACT_SITE His145/Cys187, EC 3.-.-.-, and GO:0003923. Genetics, mutagenesis, structure, and annotation all converge: the cysteine-protease catalysis is Gpi8's, not gaa1's.

Finding 2 — Domain analysis: gaa1 has zero cysteine-protease signatures; the only peptidase-like fold in the Gaa1 family is a Zn/M28 metallopeptidase, not cysteine-type

To exclude a "hidden" or divergent protease activity in gaa1, direct InterPro API queries were run against the target and its orthologs. S. pombe gaa1 (Q9US48) returns only Gaa1-component family signatures across all member databases — InterPro IPR007246, Pfam PF04114, PIRSF036762, PANTHER PTHR13304 — and no peptidase/protease domain and no catalytic-triad signature of any kind. By contrast, S. pombe gpi8 (Q9USP5) returns the full catalytic signature set: InterPro IPR001096 "Peptidase C13, legumain," Pfam PF01650 "Peptidase C13 family," PIRSF019663 "Transamidase/peptidase, legumain type," PIRSF500138 "catalytic/GPI8 component," and CATH G3DSA:3.40.50.1460 (the caspase/legumain-like fold).

The human ortholog GPAA1 (O43292) carries the same Gaa1-family entries plus a CATH-Gene3D structural superfamily hit G3DSA:3.40.630.10, "Zn peptidases" — an M28 metallopeptidase-like fold. This matters for two reasons. First, a metallo/Zn-dependent peptidase fold is mechanistically incompatible with a cysteine-type endopeptidase (GO:0004197): one uses a metal ion and an activated-water nucleophile, the other uses a catalytic cysteine thiolate. Second, S. pombe gaa1 did not even return this metallopeptidase CATH hit, so there is no protease-like structural signal in the target at all. The domain analysis thereby closes the loophole: if any enzymatic role for Gaa1 exists (some models propose Gaa1/GPAA1 contributes to the amide-bond-forming second half-reaction), it would be metallo/synthetase in character, not cysteine-type. Either way, GO:0004197 is contradicted at the domain level.

Finding 3 — Current annotation status: gaa1 has no GO:0004197; its sole MF term is GO:0003923 (TAS) — the prediction would be a novel, incorrect addition

A QuickGO annotation query for Q9US48 confirms that GO:0004197 is not present in gaa1's current record. Its only molecular-function annotation is GO:0003923 (GPI-anchor transamidase activity), evidence code TAS, reference 26563290. Its other annotations are biological-process terms (GO:0016255 attachment of GPI anchor to protein, IBA/NAS; GO:0006506 GPI anchor biosynthetic process, IEA; GO:0031505 fungal-type cell wall organization, NAS) and cellular-component terms (GO:0042765 GPI-anchor transamidase complex; GO:0005789 ER membrane; GO:0005783 ER; GO:0098553 lumenal side of ER membrane, IC). No cysteine-protease/endopeptidase term exists anywhere in the record.

The curation implication is concrete: the prediction is not a duplicate of an existing annotation — it is a proposal to add a new molecular-function term. Because that term is refuted by all available evidence, the correct action is to reject the addition. Separately (and not as justification for the prediction), the existing GO:0003923 TAS annotation on gaa1 attributes whole-complex catalytic activity to a non-catalytic subunit; by the same subunit-assignment logic it is a candidate for curator review, though annotating whole-complex activity to each subunit is an accepted GO convention. The bottom line for this hypothesis is unchanged: cysteine-type endopeptidase activity is not supported for gaa1.

Mechanistic Model / Interpretation

The GPI transamidase (GPI-T) is a pentameric, ER-membrane–embedded enzyme that catalyzes the second major post-translational step in GPI anchoring: it cleaves the C-terminal GPI-attachment signal peptide of a nascent substrate at the ω-site and, in the same catalytic cycle, replaces it with a preassembled GPI glycolipid via an amide (transamidation) bond. The five subunits and their roles are summarized below.

        GPI Transamidase (GPI-T) complex — subunit roles
  ------------------------------------------------------------------
  Subunit (S. pombe / human)   Role
  ------------------------------------------------------------------
  Gpi8  / PIG-K                CATALYTIC cysteine protease (Peptidase
                               C13 / legumain-caspase-like). Active-site
                               Cys-His(-Asn) triad forms the thioester
                               carbonyl intermediate.  <-- GO:0004197 HERE
  Gaa1  / GPAA1  (TARGET)      Non-catalytic. Binds/recognizes the GPI
                               anchor; stabilizes the complex; proposed
                               role in the amide-bond (second) half-
                               reaction. NO cysteine-protease domain.
  Gpi16 / PIG-T                Scaffold; disulfide-links to Gpi8.
  Gpi17 / PIG-S                Scaffold / substrate handling.
  Cdc91 / PIG-U                Accessory.
  ------------------------------------------------------------------

  Reaction (catalysis at Gpi8/PIG-K, NOT Gaa1):

   Substrate-omega-C(=O)-signal-peptide
              |  Gpi8 active-site Cys thiolate attacks
              v
   Substrate-omega-C(=O)-S-Gpi8   (thioester "carbonyl intermediate")
              |  GPI-ethanolamine-NH2 nucleophile (Gaa1 assists binding)
              v
   Substrate-omega-C(=O)-NH-GPI   +   Gpi8-SH (regenerated)

The prediction under review takes a property that is real for the complex and for one specific subunit (Gpi8/PIG-K) and mis-attributes it to a different subunit (Gaa1). All four evidence tiers place the catalytic cysteine in Gpi8: (1) GAA1-null cells lose intermediate formation only because the complex is non-functional, while the actual catalytic Cys/His are Gpi8's (10793132); (2) direct mutagenesis pinpoints Gpi8's Cys199/His157 (10727241); (3) human and fungal cryo-EM place the Cys–His–Asn triad in PIGK/Gpi8 and label Gaa1/GPAA1 non-catalytic (35165458, 41085069); and (4) domain databases show gaa1 has no protease signature while gpi8 carries the full Peptidase C13/legumain catalytic signature.

A subtle but useful refinement: even the possibility that Gaa1 contributes catalytically to transamidation is constrained to be non-cysteine in nature. The only peptidase-like structural fold detected anywhere in the Gaa1 family is a Zn/M28 metallopeptidase superfamily hit (in human GPAA1, and absent even from S. pombe gaa1). A metallopeptidase-like fold cannot support GO:0004197, which is defined by cysteine-thiolate catalysis. So both the "Gaa1 is purely structural/recognition" model and the "Gaa1 assists the amide-bond second half-reaction" model are consistent with refuting cysteine-type endopeptidase activity.

Evidence Base / Evidence Matrix

GO Curation Implications

Current live annotation set for gaa1 (Q9US48): the only MF term is GO:0003923 (GPI-anchor transamidase activity), evidence TAS, ref 26563290. GO:0004197 is NOT currently annotated. BP: GO:0016255, GO:0006506, GO:0031505. CC: GO:0042765, GO:0005789, GO:0005783, GO:0098553.

Leads (curator verification required):

• Reject the addition of GO:0004197 to gaa1 (Q9US48). This would be a new, incorrect MF term — a subunit-level misassignment; the catalytic cysteine-protease/transamidase activity belongs to Gpi8/PIG-K.
• Review (lower priority) the existing GO:0003923 (TAS) on gaa1, which attributes whole-complex catalytic activity to a non-catalytic subunit. This is a judgment call (whole-complex activity is sometimes annotated to each subunit by convention), not a clear error.
• The catalytic MF term (GO:0003923; mechanistically GO:0004197) correctly belongs to gpi8 (Q9USP5), where GO:0003923 is already annotated.
• MF for gaa1: No catalytic MF is supported by direct evidence. If a molecular-function statement is desired, the defensible level is complex-membership/recognition, not enzymatic activity. Avoid "protein binding" as the terminal recommendation; the informative characterization is non-catalytic structural/recognition subunit of the GPI-T complex.

Mechanistic Scope

The immediate molecular function under test is cysteine-type endopeptidase (peptide-bond hydrolysis via a catalytic cysteine). The relevant chemistry — nucleophilic attack by an active-site Cys thiolate on the substrate ω-site carbonyl to form a covalent thioester intermediate, followed by aminolysis by the GPI ethanolamine to form the amide-linked anchor — is a direct gene-product activity of Gpi8/PIG-K. For gaa1, the tested activity is absent: gaa1's direct contributions are (i) recognition/binding of the GPI-attachment signal and/or the GPI glycolipid, and (ii) structural stabilization of the pentameric complex. The observed loss of intermediate formation in GAA1-null cells (10793132) is a downstream, loss-of-function consequence of complex dysfunction — not evidence of a catalytic activity intrinsic to gaa1. Distinguishing these is the crux: mutant phenotypes of a non-catalytic subunit can mimic loss of catalysis without the subunit being catalytic.

Conflicts and Alternatives

• Apparent conflict (resolved): GAA1 knockout abolishes carbonyl-intermediate formation (10793132), which could naively suggest Gaa1 is catalytic. The same paper resolves this: the essential catalytic Cys/His are in Gpi8p; Gaa1 is a required partner, not the enzyme.
• Paralog/subunit confusion (likely source of the prediction): gaa1 is a subunit of a cysteine-protease-containing complex; association/frequency bias can propagate the complex's catalytic term onto any member. This is precisely the misassignment the seed hypothesis anticipates.
• Alternative activity for GAA1 (M28-like/second half-reaction): Some literature proposes GPAA1 assists amide-bond formation and resembles M28 metallopeptidases; this is corroborated here at the fold level (human GPAA1 carries a CATH "Zn peptidases"/M28-like hit). Even if catalytically relevant, this is a metallo/amide-synthetase character, mechanistically incompatible with GO:0004197. S. pombe gaa1 returns no protease fold signature at all, so the prediction is unsupported even under the most generous reading.
• Organism-specific differences: None that rescue the hypothesis. The Cys-protease identity of GPI8/PIG-K is conserved from yeast to human to plant (10727241 35165458 27208238), and the distinctness of GAA1 vs GPI8 is conserved even in trypanosomes (12958211).
• Database carry-over: No existing GO:0004197 annotation on gaa1 to carry over; the current record's MF is GO:0003923 only.

Limitations and Knowledge Gaps

1. Does S. pombe gaa1 have any weak intrinsic proteolytic/amide-forming activity? Checked: no domain, active site, or assay supports this. Matters because a positive result could justify a different (metallo/synthetase) MF, but never cysteine-endopeptidase. Resolution: reconstituted GPI-T activity assays with catalytically dead gpi8 vs gaa1 point mutants.
2. Structural triad absence for S. pombe gaa1. Orthologs (human PIGK, yeast) were assessed structurally; pombe gaa1 has not been individually solved. Resolution: map Q9US48 onto the human/yeast GPI-T cryo-EM structures (or AlphaFold model) to confirm no Cys–His–Asn triad in the Gaa1 fold.
3. Provenance of the GO-GPT prediction. The reference DOI (2026.03.19.712954) was not programmatically accessible in this run; its internal reasoning could not be inspected. This does not change the verdict.
4. Positive MF for gaa1 remains under-specified. The evidence robustly refutes GO:0004197 and supports a non-catalytic recognition/stabilization role, but a precise, primary-evidence-backed "GPI-attachment-signal/GPI-anchor binding" MF term was not pinned to a specific GO ID here.

Discriminating Tests

1. Active-site swap/mutagenesis: Reconstitute GPI-T with catalytically dead gpi8 (Cys→Ala/Ser) vs wild-type gaa1, and vice versa; only gpi8 active-site mutation should abolish transamidation — already demonstrated in orthologs (10727241).
2. Thioester-intermediate trapping / activity-based probes: Cysteine-protease-directed covalent probes on isolated GPI-T, with subunit identification by MS. Prediction: label maps to Gpi8/PIG-K, not gaa1.
3. Structural triad mapping: Superpose an AlphaFold model of Q9US48/gaa1 onto the human PIGK catalytic pocket; confirm the absence of a nucleophilic cysteine positioned for catalysis.
4. Domain/HMM scan (done at annotation level): Confirm gaa1 lacks Peptidase_C13 (PF01650) and any MEROPS clan CD cysteine-protease signature — only PF04114 is present.

Proposed Follow-up Actions / Curation Leads (require curator verification)

• Action: Reject the GO:0004197 prediction for gaa1 (Q9US48). Do not add cysteine-type endopeptidase activity.
• Candidate references to attach: 10793132, 10727241, 35165458, 41085069.
• Verify snippet (10793132): "cysteine and histidine residues of Gpi8p, which are conserved in members of a cysteine protease family, are essential for generation of a carbonyl intermediate ... Gpi8p is a catalytic component that cleaves the GPI attachment signal peptide."
• Verify snippet (10727241): "gpi8 alleles mutated at Cys199 or His157 are nonfunctional."
• Verify snippet (35165458): "The PIGK subunit functions as the catalytic component, in which we identified a C206-H164-N58 triad that is critical for the transamination reaction."
• Verify snippet (41085069): "the dynamic accommodation of catalytic subunit Gpi8."
• Retain existing gaa1 annotations: GO:0042765 (CC), GO:0016255 (BP), GO:0006506 (BP), GO:0005789 (CC).
• Route the catalytic term correctly: the cysteine-protease/transamidase MF belongs to gpi8 (Q9USP5), where GO:0003923 is already annotated.
• Suggested curator question: Should gaa1 carry any positive MF term beyond complex membership? Current evidence supports only a non-catalytic recognition/scaffolding role.

Provenance / Artifacts

• Direct UniProt fetches (executed): Q9US48 (gaa1) and Q9USP5 (gpi8) domain/active-site/GO lines.
• InterPro API queries (executed): gaa1 (Q9US48), gpi8 (Q9USP5), human GPAA1 (O43292).
• QuickGO annotation snapshot (executed): Q9US48 MF/BP/CC terms.
• Literature: PMIDs 10793132, 10727241, 35165458, 41085069, 27208238, 12958211.

Bottom line: The cysteine-type endopeptidase (GO:0004197) prediction for gaa1 is REFUTED — a subunit-level misassignment of Gpi8/PIG-K's catalytic activity. gaa1 is a non-catalytic GPI-transamidase subunit that binds the GPI anchor and stabilizes the complex; it should not receive GO:0004197.