C1GALT1C1 (Cosmc) — Golgi membrane (GO:0000139) function-assignment review

Gene: C1GALT1C1 (Cosmc), Homo sapiens, UniProt Q96EU7
Seed hypothesis: C1GALT1C1 has Golgi membrane (GO:0000139)
Evidence type / reference under review: TAS, Reactome:R-HSA-1964505 (co-ref PMID:37216524)
Focus: function-assignment — does the gene product directly have this CC term?

Executive Judgment

Verdict: Refuted / over-annotated (for the direct-localization claim).

C1GALT1C1 encodes Cosmc, the C1GALT1-specific molecular chaperone. Multiple
independent primary studies establish that Cosmc is an endoplasmic-reticulum
(ER)-resident single-pass type II membrane protein, and one study maps a
specific ER-retention determinant to its transmembrane domain. It is the
client enzyme it chaperones — T-synthase / C1GALT1 — that resides in the
Golgi, not Cosmc itself.

The GO:0000139 (Golgi membrane) annotation is TAS from Reactome:R-HSA-1964505,
a pathway model of Golgi-lumenal O-glycosylation. It is best explained as a
compartment carry-over: Reactome assigns the Golgi compartment to the
glycosylation reaction and its participants, and Cosmc — a regulator/client-folding
factor mentioned in that pathway context — inherited the Golgi-membrane location.
UniProt's own subcellular-location statement is deliberately conservative ("Membrane;
Single-pass type II membrane protein") and does not assert Golgi.

Most important caveat: No paper directly demonstrates Golgi residence of Cosmc,
so the annotation is not experimentally grounded; conversely, direct ER localization
is well supported. The safest curation action is to replace GO:0000139 with the
ER membrane term (GO:0005789). A minor caveat is that UniProt itself has not (yet)
attached an ER-membrane GO CC term to Q96EU7, so a curator should add it from the
primary literature.

Evidence Matrix

GO Decision Table (leads — require curator verification)

(Also saved as go_decision_table.csv.)

GO Curation Implications

• Term class: Cellular Component.
• Lead (requires curator verification): Do not retain GO:0000139 (Golgi
  membrane) as a direct localization for Cosmc. Replace with
  GO:0005789 (endoplasmic reticulum membrane) — consistent with type II
  single-pass topology and the mapped TMD ER-retention motif.
• Rationale: The Golgi term is TAS carried from a Reactome pathway whose
  compartment reflects the glycosylation reaction (Golgi lumen) and the client
  enzyme C1GALT1/T-synthase, not Cosmc's own residence. Multiple primary
  papers place Cosmc in the ER.
• If retention is preferred by policy (because Reactome TAS terms are not
  auto-removed): at minimum add GO:0005789 with experimental support
  (e.g., IDA from PMID:21262965 / PMID:18695044) and flag GO:0000139 as
  non-core / pathway-context. GO:0070062 (extracellular exosome, HDA) is likely
  a proteomics secondary observation and non-core.
• Avoid defaulting to "protein binding"; the informative, better-supported term
  is ER membrane (CC) plus the established MF/BP of chaperone activity for
  T-synthase folding (e.g., GO:0051082 unfolded protein binding / GO:0006486
  context via its client).

Mechanistic Scope

• Direct molecular function: Cosmc is an ER-lumenal-facing type II membrane
  chaperone that binds nascent/partly denatured T-synthase (C1GALT1) and
  promotes its folding, preventing aggregation and ubiquitin-mediated
  degradation (PMID:18695044, PMID:19923218).
• Direct location: ER membrane, retained via its TMD (PMID:21262965).
• Downstream / pathway consequence (NOT Cosmc's own location): Properly
  folded T-synthase traffics to and functions in the Golgi, synthesizing the
  core 1 O-glycan (T-antigen). Golgi is where the product enzyme acts — this is
  the effect that Reactome's Golgi compartment captures, and the likely origin of
  the Golgi mis-assignment to Cosmc.
• Disease manifestation (downstream): Loss/mutation of Cosmc → Tn syndrome /
  congenital disorder of glycosylation (PMID:37216524) — a phenotype of impaired
  chaperone function, not evidence of Golgi residence.

Conflicts and Alternatives

• Compartment carry-over (most likely): Reactome models the O-glycosylation
  reaction in the Golgi; Cosmc inherits "Golgi membrane" by pathway association.
• Client/chaperone conflation: T-synthase (Golgi) vs Cosmc (ER) are routinely
  discussed together; the Golgi location of the client can be mis-attributed to
  the chaperone.
• Exosome annotation (GO:0070062): high-throughput proteomics; not a
  functional localization and does not support Golgi.
• Paralog note: A paralog C1GALT1C1L (Q6BAM2) exists but is a distinct gene;
  the seed accession Q96EU7 is canonical Cosmc, so the Golgi term is not a paralog
  mixup — it is a same-gene pathway-compartment carry-over.
• Reactome confirmation (this run): R-HSA-1964505 is explicitly the C1GALT1
  Golgi galactosyltransferase reaction (compartment: Golgi membrane/lumen),
  confirming the Golgi CC term reflects the client enzyme's reaction, not Cosmc.
• No isoform/organism conflict found: ER localization is reported for human
  Cosmc; the mechanism (TMD ER-retention) is intrinsic to the protein sequence.
• No primary paper places Cosmc in the Golgi — the competing hypothesis (Golgi
  residence) lacks direct experimental support.

Knowledge Gaps

1. Absence of an ER-membrane GO CC term in UniProt for Q96EU7. Checked UniProt
   JSON: only Golgi (TAS) and exosome (HDA) CC terms are present. Matters because a
   curator replacing the Golgi term needs a positive ER term with proper evidence
   codes. Resolution: attach GO:0005789 with IDA citing PMID:21262965 / PMID:18695044.
2. Whether any Cosmc pool cycles to the ERGIC/cis-Golgi. No evidence found for
   this; the TMD retention data argue against significant Golgi steady-state pools.
   Resolution: colocalization microscopy (Cosmc vs ER markers/calnexin vs Golgi
   markers/GM130) — already effectively addressed by prior ER-retention work.
3. Policy question: whether GO/Reactome TAS compartment terms should be
   auto-propagated to regulatory participants. Matters broadly for chaperones of
   Golgi enzymes. Resolution: curator/consortium policy decision.

Discriminating Tests

• Confocal colocalization / IF: endogenous or tagged Cosmc vs ER (calnexin,
  PDI) and Golgi (GM130, TGN46) markers — expect ER, not Golgi (confirmatory).
• Glycosidase sensitivity: Endo H sensitivity of Cosmc N-glycans (if any) →
  ER-type high-mannose vs Golgi-processed complex glycans distinguishes compartment.
• TMD swap experiments (already reported, PMID:21262965): replacing Cosmc TMD
  with a non-ER TMD relocalizes it — direct causal test of ER retention.
• Proximity labeling (BioID/APEX) from an ER-anchored vs Golgi-anchored bait
  to define Cosmc's steady-state neighborhood.

Curation Leads (require curator verification)

• Action change: Replace GO:0000139 (Golgi membrane, TAS:Reactome) →
  GO:0005789 (endoplasmic reticulum membrane) as the direct CC; or, if
  retaining, downgrade GO:0000139 to non-core/pathway-context and add GO:0005789.
• Candidate references + exact snippets to verify:
• PMID:18695044 — "Cosmc is an endoplasmic reticulum (ER)-localized adenosine
  triphosphate binding chaperone that binds directly to human T-synthase";
  "Formation of active T-synthase, which resides in the Golgi apparatus,
  requires a unique molecular chaperone, Cosmc".
• PMID:21262965 — "the 18 amino acid transmembrane domain (TMD) of Cosmc is
  essential for ER localization and confers ER retention to select chimeras".
• PMID:19923218 — "an endoplasmic reticulum-localized molecular chaperone termed
  Cosmc".
• Candidate replacement/added GO terms: CC GO:0005789 (ER membrane);
  supporting MF GO:0051082 (unfolded protein binding); BP GO:0006457 (protein
  folding) / regulation of O-glycan biosynthesis via client.
• Suggested curator questions: (1) Is Reactome TAS compartment appropriate as
  a direct CC for a regulator whose client lives in that compartment? (2) Should
  GO:0070062 (exosome, HDA) be treated as non-core?
• Suggested experiments: ER/Golgi colocalization; Endo H assay; TMD-swap
  relocalization (confirmatory of published result).

Provenance

• UniProt Q96EU7 programmatic fetch (execute_code, requests) returned:
  SL comment "Membrane; Single-pass type II membrane protein" (evidence
  ECO:0000305, i.e., curator inference); CC keyword only "Membrane" (no ER, no
  Golgi); function "Probable chaperone"; topology cyto 1–6 / TM 7–26 / lumenal
  27–318; GO CC = GO:0000139 (TAS:Reactome), GO:0070062 (HDA).
• Human Protein Atlas API (search_download) for C1GALT1C1 returned subcellular
  main location "Vesicles" (antibody-based IF) — not Golgi.
• QuickGO/EBI endpoint was unreachable during this run (connection timeout); the
  full GO annotation set could not be independently enumerated beyond UniProt's.
• Reactome ContentService fetch of R-HSA-1964505 (this run) returned displayName
  "C1GALT1 transfers Galactose to the Tn antigen forming Core 1 glycoproteins (T
  antigens)", schemaClass "Reaction", compartment ['Golgi membrane','Golgi lumen'].
• GO decision table saved as go_decision_table.csv.
• PubMed retrievals: PMID:18695044, 21262965, 19923218, 37216524.

Independent-resource cross-check (Iteration 2): neither UniProt (agnostic
"Membrane", inferred) nor HPA (vesicular) supports Golgi residence; only Reactome
TAS does. This reinforces the over-annotation conclusion.

All localization conclusions are drawn from primary literature and the UniProt
record; the Golgi-inheritance interpretation is inference from the Reactome
compartment model and is labeled as such.

Artifacts

• OpenScientist final report
• OpenScientist final report
• OpenScientist go decision table