C18orf32

Summary: Interaction with GLP1R (P43220) from a membrane yeast two-hybrid (MYTH) screen using unliganded GLP-1 receptor as bait; bare protein binding carries no functional information.

Reason: Generic protein binding is uninformative and no more specific molecular function can be derived. C18orf32 is not among the interactors the study followed up functionally, and it is not named in the cached main text, so the hit is a screen-level observation. A hydrophobic membrane-anchored microprotein is a typical MYTH prey. There is no other evidence linking C18orf32 to GLP-1 receptor signalling. Removal does not imply that the interaction is false.

Summary: Co-purification of S-tagged C18orf32 with derlin-1 (DERL1, Q9BUN8) in the context of its own ER-associated degradation.

Reason: The interaction is real but reflects C18orf32 being an ERAD client that derlin-1 handles; it does not describe an activity that C18orf32 performs. Bare protein binding is uninformative, and no more specific MF (for example an adaptor activity) is supported, so the row is removed rather than rewritten.

Summary: Co-purification of S-tagged C18orf32 with the ERAD E3 ubiquitin ligase gp78 (AMFR, Q9UKV5), which is required for its degradation.

Reason: C18orf32 is a substrate of gp78; binding to the ligase that degrades it is a property of the ligase's substrate recognition, not a molecular function of C18orf32. Bare protein binding is uninformative and no more specific MF is supported.

Summary: Immunofluorescence localisation of C18orf32 to the ER in the study describing the GPIBD25 frameshift variant.

Reason: The ER localisation agrees with two independent studies (PMID:29275994, PMID:29255114) and with the ER-anchoring N-terminal hydrophobic region. The cached record is abstract-only; it reports the mutant protein in the ER, and the curator, who read the full text, presumably also saw the wild-type protein there. The ER is where GPI-inositol deacylation takes place, so this is the functionally relevant location. ORGANELLES check (2026-11-05, ER batch 2): ER residence is constitutive, not biosynthetic transit; the protein has no signal peptide, is held by its own N-terminal hydrophobic region, and is degraded in the ER by gp78/derlin-1 ERAD. Root row kept; the membrane sub-compartment is added as a separate NEW GO:0005789 row.

Summary: GFP-tagged and endogenous C18orf32 are in the ER under basal conditions; the N-terminal hydrophobic domain is necessary and sufficient for ER targeting.

Reason: Direct imaging plus fractionation and a domain-deletion analysis, in a full-text paper; consistent with constitutive ERAD of the protein. ORGANELLES check (2026-11-05, ER batch 2): the assay (GFP-fusion imaging, endogenous staining, calnexin co-staining, pellet versus buoyant fractions) shows the organelle; the membrane sub-compartment is recorded in a separate NEW GO:0005789 row rather than on this row.

Summary: UniProt subcellular-location mapping to ER, derived from the experimental localisations in PMID:29275994 and PMID:35107634.

Reason: Redundant with the experimental rows but correct. ORGANELLES check (2026-11-05, ER batch 2): the SL-0095 keyword rests on curated UniProt text tagged ECO:0000269 to PMID:29275994 and PMID:35107634, both read here, so the organelle claim holds. UniProt gives no ER sub-location, which is why the mapping stops at the root; the membrane is added as a NEW GO:0005789 row.

Summary: Proposed sub-compartment. C18orf32 is held in the ER by its N-terminal hydrophobic region, a membrane anchor on the cytosolic face; the existing annotations stop at the organelle.

Reason: ORGANELLES rule 4 (2026-11-05, ER batch 2). Lines that agree on the membrane rather than the lumen. (1) Sequence and targeting - the 76-residue chain has no signal peptide, so it cannot enter the lumen, and its N-terminal hydrophobic region (1-37) is necessary and sufficient for ER and lipid-droplet targeting; without it the protein is diffuse (PMID:29275994). The same region moves the protein onto the lipid-droplet monolayer, which only a protein embedded in the cytosolic leaflet of the ER bilayer can do, and PMID:41351175 models it as a helix whose residues 7-15 insert into the hydrophobic phase. (2) Fractionation - the endogenous, knockout-validated band is only in the pellet (ER membrane) and buoyant fractions, not the cytosol (PMID:29275994). (3) Turnover, supportive - degradation needs the membrane ERAD machinery gp78 and derlin-1, which dislocate membrane-embedded substrates. This is a constitutive membrane-anchored pool, not a cytosolic protein recruited to the ER surface, so the root-only rule for recruited pools does not apply. IC because UniProt annotates no TM segment and no carbonate extraction or protease-protection assay was done; the anchor may be monotopic (hairpin or amphipathic) rather than spanning; PMID:41351175 describes it as an "N-terminal amphiphilic helix", which favours a monotopic, membrane-embedded anchor. The membrane row still holds for a monotopic anchor, but not a bilayer-spanning topology. core_functions are unchanged.

Summary: Identified as a high-confidence lipid droplet protein by APEX2 proximity labelling; endogenous protein visualised on LDs with a knockout-validated antibody, found in buoyant fractions, and recruited to nascent LDs.

Reason: Strong, multi-method evidence with controls for overexpression and for ER-wrapping artefacts. Independently supported by PMID:41351175.

Summary: UniProt subcellular-location mapping to lipid droplet, derived from PMID:29275994.

Reason: Redundant with the IDA row but correct.

Summary: C18orf32 was one of ~300 cDNAs that activated an NF-kappaB luciferase reporter when overexpressed in HEK293 cells in a genome-scale screen of 150,000 cDNAs.

Reason: The only evidence is a hit in an overexpression reporter screen, which shows that excess protein can activate a reporter, not that the endogenous protein acts in NF-kappaB signalling. The cached record is abstract-only, so gene-level data could not be inspected, and the result is not contradicted. However, no study in the 23 years since has linked C18orf32 to NF-kappaB signalling. Every later functional study places it in the ER and on lipid droplets, in GPI-anchor processing, ERAD and lipid handling. Overexpression of an ER membrane protein that is normally degraded by ERAD could plausibly activate NF-kappaB indirectly through ER stress. The term is kept but flagged as not representing the gene's function.