Pcbd2 (mouse, Q9CZL5) — curation notes

Why this review exists

Triggered by geneontology/go-annotation#6591, which asks for the molecular
function term on MGI:1919812 (Pcbd2) and MGI:94873 (Pcbd1) to be replaced with
"4-alpha-hydroxytetrahydrobiopterin dehydratase activity", noting that in
PMID:15182178 both enzymes complement P. aeruginosa PhhB, a
pterin-4-alpha-carbinolamine dehydratase. The term currently annotated is
GO:0004505 phenylalanine 4-monooxygenase activity (IDA); the requested
replacement is GO:0008124. Companion review: genes/mouse/Pcbd1/, whose notes
set out the term confusion and its origin in full.

No deep-research file: just deep-research mouse Pcbd2 fails here because no
provider API key is configured. Evidence base is the cached publications, the
UniProt snapshot, the PANTHER PAINT table, and a bioinformatics analysis written
for this review (Pcbd2-bioinformatics/).

The requested fix, and why Pcbd2's case is stronger than Pcbd1's

Same error, same paper, but for Pcbd2 UniProt has recorded the enzymology as
experimental on this very protein:

DE            EC=4.2.1.96 {ECO:0000269|PubMed:15182178};
CC   -!- CATALYTIC ACTIVITY:
CC       Reaction=(4aS,6R)-4a-hydroxy-L-erythro-5,6,7,8-tetrahydrobiopterin = ...
CC         Evidence={ECO:0000269|PubMed:15182178};

ECO:0000269 is experimental evidence from the cited publication — i.e. UniProt
read PMID:15182178 and concluded it demonstrates EC 4.2.1.96 for mouse Pcbd2.
(For Pcbd1 the same EC is ECO:0000250, by similarity.) So UniProt and the GOA
row disagree outright on the same reference: one says hydro-lyase from this
paper, the other says monooxygenase from this paper. The issue's request resolves
the contradiction in UniProt's favour, and the abstract agrees:

Like DCoH, DCoH2 forms a tetramer, displays pterin-4alpha-carbinolamine
dehydratase activity, and binds HNF1alpha in vivo and in vitro.
PMID:15182178

"pterin-4alpha-carbinolamine dehydratase activity" is an exact synonym of
GO:0008124 (OLS-verified). So MODIFY GO:0004505 → GO:0008124, keeping IDA and
the reference.

As on Pcbd1, three annotations are collateral:

GO:0006571 L-tyrosine biosynthetic process (IEA, GO_REF:0000108) — WITH field
is literally GO:0004505. Inferred from the wrong MF; REMOVE.
GO:0004505 (IEA, ARBA:ARBA00029278) — a separate ARBA route to the same
wrong term, so correcting the IDA will not clear it; REMOVE and flag the
rule.
Note the asymmetry in the GOC inference: Pcbd1's GO:0006571 row is
ECO:0000366 while Pcbd2's is ECO:0000364. Both trace to GO:0004505.

An unexploited paper: PMID:24389307

UniProt's FUNCTION and SUBCELLULAR LOCATION for Pcbd2 both rest on
PMID:24389307 (Kazgan et al., Gastroenterology 2014), yet GOA has no
annotation from it at all. It is mouse-specific, full text is cached, and it
contains exactly the data GO wants:

Coactivation, assayed on the mouse protein in a mouse cell line:
PMID:24389307
and
PMID:24389307
Mechanism: PMID:24389307 and
PMID:24389307

Two consequences for this review:

The direct nucleus and cytoplasm observations are better evidence than the
SubCell-keyword IEAs currently carrying those terms. Recorded in the reasons
via additional_reference_ids so a curator can upgrade the evidence codes.
GO:0003713 transcription coactivator activity is proposed as NEW. See
the comparator check below, because this one needed it.

Comparator check on the proposed NEW GO:0003713

Project rules say not to add what curators deliberately declined to add, so I ran
the check before proposing. QuickGO for GO:0003713 over the PCD family returns:

Gene	Species	Evidence	Reference
PCBD1	human (P61457)	TAS	PMID:1763325
Pcbd1	mouse (P61458)	ISO	PMID:1763325
Pcbd1	rat (P61459)	IDA	PMID:1763325
PCBD2	any species	absent	—

So GO:0003713 is on PCBD1 in three species and on PCBD2 in none. That pattern is
exactly the shape that usually means "the term does not mean what you think" —
but here the alternative explanation is visible and better supported: human
PCBD2 carries involved_in GO:0045893 with IDA from PMID:11980910, whose
abstract states

DCoH stabilizes HNF1alpha as a dimer and enhances its transcriptional activity
on the beta-fibrinogen promoter reporter, and DCoHm had similar activity.
PMID:11980910

(DCoHm is PCBD2.) Curators therefore did not decline the function — they recorded
the process for PCBD2 and both the process and the molecular function for
PCBD1. The absence is a missing MF row, not a judgment that PCBD2 is not a
coactivator.

Applying the participation test: the entity doing the work is Pcbd2 itself. It
binds Hnf1a, increases Hnf1a dimerization, and a point-mutant that mimics
constitutive acetylation loses reporter activation — so it is a component of the
activating complex, not merely upstream of or required for transcription.
GO:0003713 is the term for that — its definition is "A transcription
coregulator activity that activates or increases the transcription of specific
gene sets via binding to a DNA-binding transcription factor at a specific genomic
locus, either on its own or as part of a complex", which is DCoH2 on Hnf1a
exactly. It is neither an ancestor nor a descendant of the GO:0045893 the gene
already has (one is MF, the other BP). The four coactivator mechanisms the
definition goes on to list are introduced as examples, so dimer stabilisation not
being among them is not an exclusion.

Mitochondrion: the interesting annotation

GO:0005739 mitochondrion HDA from PMID:18614015 (MitoCarta) is the one row that
does not fit the characterised biology — the dehydratase substrate is made by
cytosolic hydroxylases and the coactivator partner is nuclear — and MitoCarta's
own paper concedes MS-based inventories carry "up to 41% false positive rates".
UniProt has not adopted it: SUBCELLULAR LOCATION lists only Cytoplasm and
Nucleus.

It would be easy to call this a contaminant. I wrote
Pcbd2-bioinformatics/nterm_mts_analysis.py to test the mechanistic
prerequisite instead, because Pcbd2 has something Pcbd1 does not: ~33 extra
N-terminal residues (136 aa vs 104 aa). Scored against 259 reviewed mouse
proteins with annotated mitochondrial transit peptides and 300 reviewed mouse
cytosolic proteins:

mouse Pcbd2 residues 1–32: net charge +4 (presequence median +4), 4 Arg
(median 4), 0 acidic (median 0), max ⟨µH⟩ 0.423 (median 0.416).
mouse Pcbd1 residues 1–32, as internal control: net charge −1 (0.2nd
percentile of presequences), 6 acidic (99.8th percentile). Opposite case.
Confounder, stated honestly: the extension is 40.6% Ala, above both
reference distributions, so the zero acidic count is partly low-complexity bias
rather than selection. What survives is the positive charge and amphipathicity.

Consistent with cleavage, the two Pcbd2 crystal structures (1RU0, 4WIL) cover
33/34–136 — Rose et al. assayed the protein without the extension — and
UniProt flags both source mRNAs Erroneous initiation because they began at the
internal Met34.

And the organellar evidence is thicker than one HDA row:

Human PCBD2 is in MitoCoP: QuickGO has Q9H0N5 located_in GO:0005739 with
HTP from PMID:34800366, a study built to exclude contaminants
PMID:34800366
Mouse Pcbd2's K120/K124/K131 acetylation comes from PMID:23576753, a SIRT3
study of the liver mitochondrial acetylome (ECO:0007744 on the UniProt
MOD_RES lines), with succinylation at the same lysines from PMID:23806337
(SIRT5). Both sirtuins are mitochondrial.
The family has a genuine mitochondrial member elsewhere: PANTHER's PAINT table
for PTHR12599 has a GO:0005739 IBD at PTN000972177 (Embryophyta), seeded by
AGI_LocusCode:AT1G29810 = Q6QJ72, "Pterin-4-alpha-carbinolamine dehydratase 2,
mitochondrial".

All of that is the same class of evidence (detection in organellar preparations),
so none of it substitutes for imaging or fractionation. But four preparations
across two species, one explicitly contaminant-filtered, plus a presequence-like
extension, is not the profile of a single artefact. Action: KEEP_AS_NON_CORE —
retain, do not treat as a site of either characterised function, and see
RESULTS.md for the experiments that would settle it.

PANTHER note

Pcbd2 is in PTHR12599 subfamily SF15 ("PTERIN-4-ALPHA-CARBINOLAMINE
DEHYDRATASE 2"). It does not receive the IBAs that Pcbd1 gets from
PTN002650414, even though that node is placed at taxon:117571 (Euteleostomi) and
both paralogs descend from the vertebrate ancestor. The node therefore appears to
sit below the Pcbd1/Pcbd2 duplication. Since GO:0008124 is experimentally
supported for Pcbd2 itself (UniProt ECO:0000269|PubMed:15182178), this looks
like a case where the dehydratase IBD could be asserted deeper. Raised as a
question for PAINT curators, not actioned.

Summary of actions

Term	Evidence	Action
GO:0004505 phenylalanine 4-monooxygenase activity	IDA PMID:15182178	MODIFY → GO:0008124 (the issue's ask)
GO:0004505	IEA ARBA:ARBA00029278	REMOVE (independent ARBA fix needed)
GO:0006571 L-tyrosine biosynthetic process	IEA GO_REF:0000108	REMOVE (inferred from the wrong MF)
GO:0005515 protein binding (w/ Hnf1a)	IPI PMID:15182178	MODIFY → GO:0140297
GO:0005515 protein binding (w/ Pcbd1)	IPI PMID:15182178	UNDECIDED (paralog hetero-oligomer; as on Pcbd1)
GO:0042802 identical protein binding (w/ self)	IPI PMID:15182178	KEEP_AS_NON_CORE
GO:0005634 nucleus ×2, GO:0005737 cytoplasm	IEA / IPI	ACCEPT (direct mouse evidence available to upgrade)
GO:0005739 mitochondrion	HDA PMID:18614015	KEEP_AS_NON_CORE
GO:0006729 tetrahydrobiopterin biosynthetic process	IEA InterPro	ACCEPT (regeneration, not de novo)
GO:0008124	IEA GO_REF:0000120	ACCEPT (core)
GO:0045893 positive regulation of DNA-templated transcription	ISO GO_REF:0000119	ACCEPT (core)
GO:0003713 transcription coactivator activity	—	NEW from PMID:24389307

2026-10-12 Ortholog/family audit

GOA refetched (no new rows); review already COMPLETE and unchanged. PANTHER PTHR12599 confirmed from UniProt (Pcbd1 SF13, Pcbd2 SF15). Family treatment is consistent between the two mouse paralogs: both MODIFY the GO:0004505 phenylalanine-4-monooxygenase IDA to GO:0008124 (dehydratase, per go-annotation#6591), REMOVE the ARBA-sourced GO:0004505 IEA and the GO:0006571 tyrosine-biosynthesis inference, ACCEPT GO:0008124 and GO:0006729, MODIFY the Hnf1a protein-binding IPI to GO:0140297, mark the paralog self-binding IPI UNDECIDED, and KEEP_AS_NON_CORE identical-protein-binding. Core functions in both: GO:0008124 dehydratase + GO:0003713 transcription coactivator (DCoH/DCoH2).

Human PCBD1 (status INITIALIZED, do-not-complete) agrees on GO:0004505->GO:0008124 MODIFY, GO:0004505 IEA REMOVE, GO:0006571 REMOVE, GO:0008124 ACCEPT, and the GO:0003713 coactivator core. Noted mismatch (not fixed, human is a stub): human PCBD1 marks its Hnf1a protein-binding rows KEEP_AS_NON_CORE whereas mouse Pcbd1 MODIFYs the Hnf1a IPI to GO:0140297 DNA-binding transcription factor binding. No human PCBD2 review exists.