CA8 (P35219) — review journal

The contested annotation, and the thing that makes this case unusual

GOA carries GO:0004089 carbonate dehydratase activity twice, and I checked the
evidence codes myself in CA8-goa.tsv:

There is no IDA, IMP or EXP row for either term anywhere in the file. So the catalytic
assignment rests on (a) a fold/domain match and (b) a single traceable author statement.

The TAS contradicts its own source

PMID:8977131 is Sjöblom et al., FEBS Lett 1996;398(2-3):322-5,
"Two point mutations convert a catalytically inactive carbonic anhydrase-related protein
(CARP) to an active enzyme", DOI 10.1016/s0014-5793(96)01263-x — verified against PubMed,
abstract-only in cache (full_text_available: false).

The title alone says it, and the abstract is unambiguous:
PMID:8977131

The whole point of the paper is that you have to engineer two substitutions
(Arg117→His, Glu115→Gln) to get activity:
PMID:8977131

And the same sentence that establishes zinc binding establishes it for the mutant, by
contrast with the wild-type:
PMID:8977131

So the PINC TAS for GO:0004089 is a misannotation of the very paper it cites: the source
reports the absence of the activity in the wild-type protein. That also disposes of the
GO:0008270 zinc-ion-binding IEA — unmodified CARP was not isolated as a zinc complex, and
the fold-based signature that produced the IEA is exactly what the experiment refutes.

(The paper works on murine CARP, the CA8 ortholog. That is not a problem for the
argument: the human protein has the identical active-site substitution, UniProt flags it,
and PINC used this paper for the human entry in the first place.)

Independent confirmation that CA8 is acatalytic

• UniProt P35219 FUNCTION: Does not have a carbonic anhydrase catalytic activity. and
  CAUTION: Although it belongs to the alpha-carbonic anhydrase family, Arg-116 is present instead of the conserved His which is a zinc-binding residue. It is therefore expected that this protein lacks carbonic anhydrase activity.
• HGNC name: "carbonic anhydrase 8 (inactive)".
• PMID:32316137 and PMID:32316137
• PMID:42268453

Decision: GO:0004089 (both the IEA and the TAS rows — same term must carry the same
action) and GO:0008270 → REMOVE. This is not second-guessing an experimental
annotation: there is no experimental annotation to second-guess, and the one cited paper
says the opposite of what the TAS asserts.

What CA8 actually does

The fold is retained and repurposed as a protein-interaction surface. The primary finding
is Hirota et al., Biochem J 2003 (PMID:12611586, verified; abstract-only in cache):

• PMID:12611586
• PMID:12611586
• PMID:12611586

Mechanism, restated by the structural work:
PMID:32316137

The interaction is reciprocally validated from the receptor side by Ando et al., PNAS
2018 (PMID:30429331):
PMID:30429331 and
PMID:30429331
This is the strongest part of the case: mutations on either side of the interface
converge on the same functional readout.

Disease: recessive CA8 mutations cause cerebellar ataxia (SCAR34 / CAMRQ3).
PMID:19461874

Directionality: CA8 is a negative regulator —
PMID:42268453, and it acts
on the receptor, not on the ligand:
PMID:42268453

Choosing the replacement molecular function

IP3R1 (ITPR1) is a ligand-gated intracellular calcium-release channel. CA8 binds it and
reduces its activity. I looked up candidates rather than guessing:

There is no GO term for "IP3 receptor binding" specifically; GO:0019855 is the most
informative MF available and it is in the molecular_function branch (checked via QuickGO),
so it passes the strict branch validation on core_functions.

Other GOA rows

• GO:0005515 protein binding × 5 IPI rows (PMIDs 25416956, 25910212, 27107012, 31515488,
  32296183) — all systematic binary-interactome/Y2H screens. Partners are CRX, GGA2,
  HSD17B14, INTS7, KLHL8, LMNB2, LNX1, MAGED1, RAB34, SPDL1, TBX3. ITPR1 — the one
  partner that carries the biology — does not appear in GOA at all. Marked
  MARK_AS_OVER_ANNOTATED per project guidance (same action on every row of the same
  term).
• GO:0005737 cytoplasm (IEA, GO_REF:0000107, Ensembl Compara from mouse P28651) —
  ACCEPT. CA8 is a soluble cytosolic protein engaging the cytoplasmic modulatory domain of
  an ER-membrane channel; nothing more specific is supported.

Honest uncertainties

• Is CA8 a complete pseudoenzyme, or does it retain a vestigial/alternative activity?
  The literature is uniform that it cannot hydrate CO2, but "no measurable CA activity" is
  not the same as "no catalytic activity of any kind", and no one has screened for one.
• Quantitative stoichiometry and affinity of the CA8–IP3R1 interaction are not
  established in a purified reconstituted system, so the "inhibitor" call rests on
  IP3-binding assays and cell-based calcium readouts rather than on single-channel
  recordings.
• Non-neural roles. The 2026 review notes reported roles in tumour progression and
  metabolic regulation but calls them less well characterised; I did not annotate them.
• PMID:26399641 reports CA8 overexpression stunting Purkinje dendritic growth without
  detectable direct binding to IP3R1 in that system — a dissenting data point on the
  universality of the interaction. Noted; it does not overturn the two reciprocal
  mutational studies, but it belongs in the record.

Validation notes

• references: titles filled with fill_refs.py.
• All supporting_text verified as normalised verbatim substrings of the cached
  publications/PMID_*.md files.
• GO ids for core_functions (GO:0019855, GO:0051280, GO:0005737) and the CL id
  CL:0000121 Purkinje cell were each looked up (QuickGO / OLS), not written from memory.