GPC3 (donkey, Equus asinus; UniProt A0A9L0J2B1) - review notes

Sources used

Donkey UniProt record (unreviewed, Ensembl ENSEASP00005043430.1), 565 aa: GPC3-uniprot.txt.
Donkey GOA: 8 IEA rows (5 TreeGrafter GO_REF:0000118 from PANTHER:PTN001327382; 2 InterPro2GO GO_REF:0000002 from IPR001863; plasma membrane via GO_REF:0000120).
Human GPC3 (P51654) UniProt record fetched 2026-10-14: GPC3-bioinformatics/human-GPC3-P51654-uniprot.txt.
Alignment: GPC3-bioinformatics/RESULTS.md.
Cached publications (already present in publications/): PMID:24496449 (GPC3 binds Frizzled; abstract only), PMID:14610063 (furin processing of GPC3; full text), PMID:17549790 (GPC3 inhibits CD26/DPP4; abstract only).
Human QuickGO (2026-10-14) comparator: IDA positive regulation of canonical Wnt signaling (PMID:24496449); IBA extracellular region, plasma membrane, cell surface, cell migration, regulation of protein localization to membrane; ISS negative regulation of smoothened signaling pathway; the extracellular matrix and regulation of signal transduction rows on human GPC3 are InterPro2GO only.

[RESULTS.md "The donkey GPC3 protein matches human GPC3 at 92.61% identity over 555 paired residues"]
[RESULTS.md "The furin-like cleavage site (human R358/S359), the two heparan sulfate attachment serines (human S495, S509) and the GPI-anchor asparagine (human N554) are all conserved in the donkey sequence."]

GPI-anchored heparan sulfate proteoglycan [human UniProt "Cell membrane {ECO:0000269|PubMed:14610063}; Lipid-anchor, GPI-anchor"].
Wnt: PMID:24496449; [human UniProt "Positively regulates the canonical Wnt signaling pathway by binding to the Wnt receptor Frizzled"].
Hedgehog: [human UniProt "Negatively regulates the hedgehog signaling pathway when attached via the GPI-anchor to the cell surface by competing with the hedgehog receptor PTC1 for binding to hedgehog proteins"] (by similarity to mouse).
Processing: PMID:14610063; cell movements PMID:14610063.
Shedding: [human UniProt "leading to its shedding from the cell surface"].

plasma membrane, cell surface, positive regulation of canonical Wnt signaling: ACCEPT.
extracellular region (shed form), cell migration, regulation of protein localization to membrane: KEEP_AS_NON_CORE.
regulation of signal transduction: TOO_GENERAL (entailed by positive regulation of canonical Wnt signaling, and by the NEW Hh term).
extracellular matrix: MARK_AS_OVER_ANNOTATED (GPC3 is GPI-anchored to the cell surface, not an ECM constituent; InterPro2GO family mapping).
NEW: negative regulation of smoothened signaling pathway (GO:0045879).

ProtNLM2 output was not opened or used. Everything below comes from the UniProt/GOA records, PAINT, cached literature and new sequence checks.

just deep-research-falcon EQUAS GPC3 --fallback perplexity-lite failed: falcon returned "Failed to authenticate" and perplexity-lite returned HTTP 401 (insufficient quota). No deep-research file was produced. The review relies on the cached literature listed below.

A0A9L0J2B1 (565 aa) is an alternative-first-exon model of donkey GPC3 (Ensembl transcript ENSEAST00005057340, gene ENSEASG00005018489). Its residues 1-43 (MSFQFAPGGCYGHARVFDCQMVMMMEARMFIFWPRFWKGTPNK) are not homologous to human P51654 1-58, and the human signal peptide (P51654 SIGNAL 1..24) is absent. From donkey 44 / human 59 (GSDLQVCLPK) onward the two are 502/522 identical (96.17%) [file:EQUAS/GPC3/GPC3-bioinformatics/RESULTS-rereview.md "Human signal peptide present verbatim in donkey: False"].
The 92.61% figure in the earlier RESULTS.md includes the non-homologous N-terminus forced into register. The site-conservation conclusions there still hold.
The canonical donkey GPC3 product is A0A9L0JGR2 (580 aa, ENSEAST00005069329). It has a human-like signal peptide and is the GPC3 entry in the UniProt one-protein-per-gene reference proteome (UP000694387). The Ensembl canonical transcript ENSEAST00005080949 (A0A9L0J383, 611 aa) has the same signal peptide plus a 31-residue net insertion. Horse A0A3Q2GY79 has the same divergent N-terminus as A0A9L0J2B1.
Consequence: the secretory-pathway annotations (plasma membrane, cell surface, GPI-anchored HSPG, Frizzled binding via HS chains) describe the canonical product of the gene. They are not established for this transcript model as such. I kept the gene-level actions and recorded the caveat in the review rows and in a suggested question.

Donkey, human (P51654) and mouse (Q8CFZ4) GPC3 are all in PTHR10822:SF4 (GLYPICAN-3).
PAINT IBDs in interpro/panther/PTHR10822/PTHR10822-paint.tsv that cover the GPC3 lineage: positive regulation of canonical Wnt (PTN000850517, Bilateria; seeds dally + human GPC3); cell surface, extracellular region, cell migration, regulation of protein localization to membrane (PTN002247069, Eumetazoa); plasma membrane (PTN002585559, Euteleostomi; seeds mouse Gpc3, rat Gpc3, human GPC3). These match the donkey TreeGrafter rows exactly.
Not inherited by the GPC3 lineage: extracellular matrix and synapse (PTN000088172), FGF binding / negative regulation of FGFR signaling (PTN002585522, seed mouse Gpc1), and the synaptic terms at PTN000088293. Human GPC3 has no IBA for these. This supports MARK_AS_OVER_ANNOTATED for the InterPro2GO extracellular matrix row.
No PAINT node carries any hedgehog term. Mouse Gpc3's hedgehog annotations are IMP/IGI acts_upstream_of_or_within [PMID:18477453; PMID:19590577] plus an IPI with Shh. I read the omission as a deliberate decision not to make this a family-wide (all-glypican) function. Following that decision, the hedgehog NEW terms are transferred only within the GPC3 subfamily (ISS from mouse, as the human GPC3 ISS rows already do), not from glypicans in general.
No PANTHER family review (PTHR10822-review.yaml) exists, so there is no subfamily-placement disagreement to report.

PMID:18477453 (abstract only): PMID:18477453; PMID:18477453. Newly cached and cited.
PMID:19590577 (abstract): Gpc3-null overgrowth is partly Ihh-dependent. Newly cached and cited.
PMID:24496449 (abstract): PMID:24496449.
PMID:14610063 (full text): convertase processing; zebrafish morphant gastrulation defects; and PMID:14610063. The sign of the canonical Wnt effect is therefore context-dependent.

Description: rewritten to state the gene-model identity (alternative first exon, no signal peptide, canonical = A0A9L0JGR2).
Existing rows: actions unchanged (plasma membrane, cell surface and positive regulation of canonical Wnt ACCEPT; extracellular region, cell migration and regulation of protein localization to membrane KEEP_AS_NON_CORE; regulation of signal transduction TOO_GENERAL; extracellular matrix MARK_AS_OVER_ANNOTATED). Reasons now cite the PAINT node placements. The cell-migration row explicitly does not use the zebrafish morphant phenotype as proof of participation.
NEW rows: frizzled binding changed from IDA to ISS (the experiment is on human GPC3). The negative regulation of smoothened signaling and hedgehog family protein binding rows are re-sourced to PMID:18477453 (mouse, same subfamily) as ISS.
Not added: peptidyl-dipeptidase inhibitor activity (human IDA, PMID:17549790, DPP4/CD26). It is a single study in a cancer-cell context, not part of the core signaling roles.
New analysis files: GPC3-bioinformatics/compare_nterm_rereview.py, nterm_rereview_output.txt, RESULTS-rereview.md. Existing bioinformatics files are untouched.

Same-gene canonical product: donkey GPC3 A0A9L0JGR2 (folder would be genes/EQUAS/GPC3__A0A9L0JGR2). It has no review yet.