SLC10A7 (Q0GE19) — curation notes

Deep research was not available in this container (Falcon returned HTTP 402, OpenAI 401,
perplexity not registered), so no -deep-research-*.md file was generated. This file holds the
literature synthesis, assembled from the cached publications in publications/ plus UniProt
(SLC10A7-uniprot.txt), QuickGO and OLS lookups.

1. What the protein is

SLC10A7 (originally C4orf13) is the seventh and most divergent member of the SLC10
("sodium/bile acid cotransporter") family. Its only published molecular characterization is
Godoy et al. 2007: a 340–343 residue protein, >85% identity between human, mouse, rat and frog,
with a membrane topology unlike the rest of the family
PMID:17628207.
Unlike every other SLC10 member it has pre-vertebrate relatives
PMID:17628207.

It is not a bile acid or steroid sulfate carrier. The only transport assays ever reported were
negative PMID:17628207,
and UniProt records this explicitly ("Does not show transport activity towards bile acids or steroid
sulfates"). As of 2026 the substrate is still unknown
PMID:30082715.

Consistent with its divergence, SLC10A7 does not associate with the family's prototype: NTCP
co-immunoprecipitates and co-localizes with SLC10A4 and SLC10A6 but
PMID:22029531.

2. Where it is — the localization conflict

Three compartments are annotated, from three papers, and they are not of equal weight:

• Plasma membrane comes only from heterologous over-expression of epitope-tagged protein.
  Godoy detected HA/FLAG-tagged SLC10A7 at the surface of transfected HEK293 cells and Xenopus
  oocytes [PMID:17628207 as quoted above]. Dubail et al. re-asserted it from c-myc-tagged protein
  transfected into HEK293F cells, with no plasma-membrane co-marker, and the staining they describe
  is punctate rather than rim-like
  PMID:30082715,
  and they cite Godoy for the claim
  PMID:30082715.
• Golgi is where the protein sits when asked about the human protein directly, and it is the
  compartment the functional data point to
  PMID:29878199,
  with loss of SLC10A7 producing a secretory-pathway defect
  PMID:29878199.
  Note that Ashikov et al. frame this as a human/yeast difference, which also tells us the
  fungal orthologs (Rch1p, CaRch1p) are the plasma-membrane ones.
• Endoplasmic reticulum / ER membrane is attributed by UniProt and GOA to PMID:17628207
  (ECO:0000269|PubMed:17628207), but that paper is abstract-only in our cache and its abstract
  mentions only the plasma membrane. The ER data must be in the full text (Godoy's figures), which
  I could not read. Per SKILL.md this is exactly the UNDECIDED case — not a removal.

Resolution adopted: Golgi is the physiological site (ACCEPT); plasma membrane is real but an
over-annotation of the endogenous protein, driven by over-expression of tagged constructs
(MARK_AS_OVER_ANNOTATED, including the IBA, whose donor set is fungal/bacterial — see §5); the two
ER rows are UNDECIDED pending Godoy's full text. GAG assembly begins at the ER exit and continues
through the Golgi PMID:30082715,
so an early-secretory-pathway pool is biologically plausible; it is simply unverifiable here.

3. Calcium

The yeast and Candida orthologs are calcium regulators, which is what prompted the human experiment
PMID:30082715.
Patient fibroblasts behave the same way:
PMID:30082715,
i.e. loss of SLC10A7 raises Ca2+ entry, so the wild-type protein is a negative regulator of
Ca2+ influx. This is the one mechanistic function that is both conserved from yeast to human and
directly measured in human cells, so GO:0006874 intracellular calcium ion homeostasis is accepted
as core. (A more specific "negative regulation of calcium ion import" term was considered but not
proposed: the measurement is a whole-cell influx readout in a patient line, and the transported
species is unknown, so the directionality is inferred from a loss-of-function phenotype only.)

4. The GAG phenotype — necessity vs participation

What was measured in Dubail et al.:
- PMID:30082715
- PMID:30082715
- Sulfation itself is normal: PMID:30082715
- And the authors' own mechanism is indirect: PMID:30082715

So: heparin was never assayed (heparin is the hypersulfated, mast-cell/serglycin GAG;
GO:0030210 is defined as formation of heparin proteoglycans specifically), and SLC10A7 catalyses
no step of chain initiation, elongation, epimerization or sulfation. It supplies a permissive
lumenal ion environment.

Comparator check (QuickGO, human+mouse):
- GO:0030210 holds 47 annotations, essentially all of them the HS/heparin biosynthetic enzymes
— EXT1, EXT2, NDST1-4, GLCE, B3GALT6, B4GALT7, CSGALNACT1 — plus SLC10A7 itself. No other
transporter is there.
- TMEM165 is the closest comparator of all: a Golgi Ca2+/Mn2+ transporter whose loss causes a
congenital disorder of glycosylation. GO gives it GO:0032472 Golgi calcium ion transport,
GO:0032468 Golgi calcium ion homeostasis, GO:0071421 manganese ion transmembrane transport —
and no glycosylation or glycan-biosynthesis process term. The agent-of-the-step convention is
being applied.
- SLC35D1 (Golgi UDP-GlcA/UDP-GalNAc transporter, Schneckenbecken dysplasia) likewise carries
only transport terms.
- SLC35B2 (Golgi PAPS transporter) is annotated GO:0050650 chondroitin sulfate proteoglycan
biosynthetic process by IMP — so the convention is not absolute, and a transporter that delivers
the actual donor substrate into the lumen can carry the biosynthetic term.

Verdict: do not REMOVE (the requirement is solid and reproduced in two species), but the term as
written is wrong on specificity. MODIFY to GO:0015012 heparan sulfate proteoglycan biosynthetic
process — the GAG that was actually measured — and treat it as a permissive/upstream contribution
rather than a core catalytic role. If a curator prefers the strict TMEM165 convention, the
alternative would be to drop the biosynthesis term entirely and keep only the calcium term; that is
noted as a question for experts rather than acted on, because SLC10A7's substrate is unknown and it
may yet turn out to be a donor transporter in the SLC35B2 mould.

5. Bone

Human: short stature, skeletal dysplasia with multiple dislocations, scoliosis, decreased bone
mineral density, amelogenesis imperfecta
PMID:29878199.
Mouse: PMID:30082715
and PMID:30082715.
Zebrafish: PMID:29878199

SLC10A7 is widely expressed (UniProt TISSUE SPECIFICITY; liver highest), and the skeletal
phenotype is the tissue in which a general cartilage/bone ECM defect becomes visible. This is the
textbook pleiotropic-developmental-consequence case, so GO:0060348 bone development is kept but
marked non-core rather than accepted as a core function.

6. Interactors

All three GO:0005515 protein binding rows are the same interactor, Q7Z388 = DPY19L4 (UniProt
INTERACTION: Q0GE19; Q7Z388: DPY19L4; NbExp=4), reported by three high-throughput AP-MS surveys:
BioPlex 2.0 PMID:28514442, BioPlex 3.0 PMID:33961781 and the RESOLUTE SLC interactome
PMID:40355756. None of the three cached full texts mentions SLC10A7 by name, and none assigns it
a molecular function. DPY19L4 is annotated by UniProt as a "Probable C-mannosyltransferase that mediates
C-mannosylation of tryptophan residues on target proteins" (multi-pass membrane protein), so the
pairing is suggestive for the glycosylation story, but an AP-MS co-precipitation supports no informative
MF. Per SKILL.md "Quality Standards", these are REMOVE as uninformative (not
MARK_AS_OVER_ANNOTATED); removal does not assert the interaction is false.

7. Open questions

• What does SLC10A7 transport? Nothing has been tested since the 2007 bile-acid/steroid-sulfate
  panel. Candidates worth testing in a Golgi-oriented assay: Ca2+, Mn2+, sulfate, PAPS, nucleotide
  sugars.
• Is the Ca2+ effect at the plasma membrane (store-operated entry, the yeast Rch1 model) or inside
  the Golgi (the TMEM165 model)? The published measurement cannot distinguish them.
• Is the PM pool real at endogenous levels, or an over-expression artefact? Needs an endogenous
  tag / validated antibody with surface biotinylation.