Pcbd2

Summary: This is the annotation that geneontology/go-annotation#6591 asks to correct. GO:0004505 is the activity of phenylalanine hydroxylase (Pah), one step upstream of Pcbd2 in the BH4 cycle. PMID:15182178 measured pterin-4-alpha-carbinolamine dehydratase activity, which is GO:0008124.

Reason: The two terms name consecutive steps of one cycle. GO:0004505 is "Catalysis of the reaction: L-phenylalanine + tetrahydrobiopterin + O2 = L-tyrosine + 4-alpha-hydroxytetrahydrobiopterin"; Pcbd2 instead dehydrates that 4a-hydroxy-BH4 product, which is GO:0008124, a term whose synonym list contains "pterin-4-alpha-carbinolamine dehydratase activity" verbatim - the exact phrase the abstract uses. For Pcbd2 the contradiction is sharper than for Pcbd1, because UniProt attributes the dehydratase chemistry to this same reference with experimental evidence: the entry carries EC=4.2.1.96 and a RHEA:11920 CATALYTIC ACTIVITY block, both tagged ECO:0000269|PubMed:15182178. Two databases therefore read one paper as supporting two mutually exclusive EC classes (4.2.1.- hydro-lyase versus 1.14.16.- monooxygenase), and the issue resolves it in UniProt's favour. The experiment, evidence code and reference are sound; only the term is wrong, so MODIFY keeps IDA and PMID:15182178.

Proposed replacements: 4-alpha-hydroxytetrahydrobiopterin dehydratase activity

Summary: An ARBA machine-learning rule assigning phenylalanine 4-monooxygenase activity independently of the MGI IDA. The same wrong term by a different route, so correcting the IDA will not clear it.

Reason: Pcbd2 has no monooxygenase chemistry - no iron centre, no pterin-dependent hydroxylase domain - and UniProt assigns it EC 4.2.1.96 with the Lyase keyword on the strength of a direct experiment. ARBA00029278 fires on both mouse PCD paralogs, which suggests it has learned an association from the BH4/phenylalanine-hydroxylation context surrounding this family rather than from the family's own catalytic signature. An electronic inference contradicted by the curated EC and reaction; remove, and revisit the rule's condition set independently of the MGI correction requested in geneontology/go-annotation#6591.

Summary: Interaction with PR:P22361, mouse Hnf1a. The identity of the partner is the informative part and the bare protein binding term discards it.

Reason: Per project curation guidelines GO:0005515 carries no functional information and should be replaced with a term naming what is bound. Hnf1a is a sequence-specific DNA-binding transcription factor, so GO:0140297 ("Binding to a DNA-binding transcription factor, a protein that interacts with a specific DNA sequence ... to modulate transcription") states the experiment exactly. This binding is the molecular basis of the coactivator activity proposed as NEW below and of the GO:0045893 process already annotated; the three are complementary rather than redundant, GO:0140297 recording the recognition event.

Proposed replacements: DNA-binding transcription factor binding

Summary: Interaction with PR:P61458, the paralog Pcbd1 (DCoH). The mirror image of the Pcbd1 annotation that cites PR:Q9CZL5, and it raises the same term problem.

Reason: A bare protein-binding row should be replaced by a term naming what is bound, and the candidate here is GO:0046982 protein heterodimerization activity ("Binding to a nonidentical protein to form a heterodimer"), since Pcbd1 is a paralog and GO:0042802 identical protein binding therefore cannot apply. I cannot justify that replacement from what I can see: the cached record for PMID:15182178 is abstract-only, and the abstract reports DCoH2 disproportionating and forming a 2:2 complex with HNF1 without describing a mixed DCoH/DCoH2 species. The MGI curator read the full text and recorded the interaction, so the interaction is not in doubt; what I cannot determine is whether the data show a discrete heterodimer, a mixed tetramer, or subunit exchange between separately assembled tetramers, and these map to different terms. UNDECIDED pending full text: MODIFY to GO:0046982 if a defined mixed species is shown, otherwise a complex term or GO-CAM is the right home, as raised in suggested_questions.

Summary: Nucleus from the UniProt SubCell keyword mapping (SL-0191). Correct, and now directly observable in mouse: Kazgan et al. report strong nuclear staining of DCoH2 in a mouse hepatic cell line.

Reason: The underlying UniProt statement is itself experimental (SUBCELLULAR LOCATION Nucleus, ECO:0000269|PubMed:24389307), so this keyword mapping reproduces a direct observation rather than a guess. The nucleus is where the coactivator function operates. Worth noting for curators that PMID:24389307 is not currently used in any Pcbd2 GOA row despite being UniProt's source for both FUNCTION and SUBCELLULAR LOCATION; it would support a direct evidence code here in place of the keyword mapping.

Summary: Nuclear localization inferred from the in vivo interaction with Hnf1a (PR:P22361), a nuclear transcription factor.

Reason: The abstract states the HNF1alpha interaction was shown in vivo as well as in vitro, so co-residence with a nuclear transcription factor is a reasonable basis for the compartment. Weaker than imaging, but independently confirmed by the direct immunofluorescence in PMID:24389307 and redundant with the keyword-mapped nucleus row above.

Summary: Cytoplasm from the UniProt SubCell keyword mapping (SL-0086). This is the compartment where the dehydratase half of the protein works, since the aromatic amino acid hydroxylases that generate its substrate are cytosolic.

Reason: Directly observed in mouse by Kazgan et al., who found DCoH2 in both nucleus and cytoplasm of Hepa1-6 cells; the UniProt Cytoplasm statement carries ECO:0000269|PubMed:24389307. Pcbd2 has no cytosol (GO:0005829) annotation, unlike Pcbd1 which has three - a plausible gap, since the enzymatic function requires the cytosol specifically, but I am not proposing the child term as NEW because no cited experiment resolves cytosol from other cytoplasmic compartments for this paralog.

Summary: Mitochondrion from the MitoCarta mouse mitochondrial compendium. The one compartment that does not fit either characterised function, but it is supported by more than this single row and by a plausible import mechanism, so it is retained as non-core rather than removed.

Reason: The easy call would be contaminant: MitoCarta is MS-based and its own paper concedes that published reports exhibit up to 41% false positive rates, UniProt has not adopted the compartment (SUBCELLULAR LOCATION lists only Cytoplasm and Nucleus), and neither characterised function belongs there - the dehydratase substrate is produced by cytosolic hydroxylases and Hnf1a is nuclear. Three things argue against dismissing it. First, Pcbd2 has roughly 33 N-terminal residues that Pcbd1 lacks, and scored against 259 reviewed mouse transit peptides and 300 reviewed mouse cytosolic N-termini they sit at the median of real presequences on every classical feature, while the same window of Pcbd1 sits at the opposite extreme; the honest caveat is a 40.6% alanine content that partly explains the acidic-residue depletion. Second, the extension is absent from both Pcbd2 crystal structures (1RU0, 4WIL cover 33/34-136) and UniProt flags both source mRNAs Erroneous initiation, consistent with a cleaved presequence yielding the mature species that was assayed. Third, the MitoCarta detection is not alone: QuickGO records human PCBD2 (Q9H0N5) as located_in mitochondrion with HTP evidence from PMID:34800366, the contaminant-filtered MitoCoP proteome, and mouse Pcbd2's K120/K124/K131 acetylation comes from a liver mitochondrial acetylome (PMID:23576753), with succinylation at the same lysines from PMID:23806337 and SIRT3 and SIRT5 both being mitochondrial sirtuins. This is still one class of evidence - detection in organellar preparations - and so does not substitute for imaging or fractionation. A plant-side PAINT signal adds context rather than mouse evidence: PANTHER's PTHR12599 PAINT table carries a GO:0005739 IBD at the Embryophyta node seeded by Arabidopsis Q6QJ72, named "Pterin-4-alpha-carbinolamine dehydratase 2, mitochondrial", showing that this enzyme family can be organelle-targeted; however, the plant and mammalian extensions are not homologous presequences, and the vertebrate node PTN002650414 carries only nucleoplasm, cytosol and GO:0008124 with no mitochondrial component. Three mammalian organellar preparations, one explicitly built to exclude contaminants, plus a presequence-like extension, do not fit the profile of a single co-purification artefact. Retained as non-core; experiments that would settle it are in suggested_experiments.

Summary: An inter-ontology logical inference whose WITH field names GO:0004505 as its source. A mechanical consequence of the mis-annotated molecular function, and it inherits the error.

Reason: The rule is sound - a phenylalanine 4-monooxygenase does participate in tyrosine biosynthesis - but its premise is false for this gene. Tyrosine is made by Pah; Pcbd2 acts after the hydroxylation, on the spent cofactor. Under the project rule that a gene product is involved_in a process only if it performs some of the work of that process, Pcbd2 contributes nothing to tyrosine formation: the carbon skeleton of tyrosine never passes through it. Comparator: Qdpr, which performs the other BH4 regeneration step, is likewise not annotated as a tyrosine biosynthetic enzyme. Should disappear once GO:0004505 is corrected, but needs separate removal because the ARBA-sourced GO:0004505 would otherwise keep regenerating it.

Summary: InterPro2GO annotation from the pterin dehydratase signatures. Pcbd2 regenerates BH4 from the 4a-hydroxy-BH4 carbinolamine formed during aromatic amino acid hydroxylation.

Reason: Strictly, Pcbd2 acts in BH4 regeneration rather than de novo synthesis from GTP (the Gch1/Pts/Spr route), so the term is broader than the precise role, and GO has no regeneration child to move to. Pcbd2 does perform a chemical step whose product re-enters the BH4 pool, so it is a genuine participant rather than a substrate or a mere requirement. IPR001533/IPR036428 are the correct family signatures for this protein. Supported directly by UniProt's FUNCTION statement for this entry, which cites the experimental paper.

Summary: The correct molecular function, currently carried only by an electronic annotation. This is the term geneontology/go-annotation#6591 asks MGI to use for its experimental annotation.

Reason: Four lines of support converge: the InterPro family signatures, RHEA:11920, EC 4.2.1.96, and UniProt's curated CATALYTIC ACTIVITY block for this entry - which, unusually for an IEA-supported term, is itself backed by experimental evidence on this protein (ECO:0000269|PubMed:15182178). This row is also the direct refutation of the GO:0004505 annotations on the same gene product, since the two EC classes are mutually exclusive. Once the MGI IDA is corrected the gene will carry this term with an experimental code, which is the right outcome; the IEA is correct in the meantime.

Summary: Self-interaction. Pcbd2 is a homotetramer, but one that behaves quite differently from Pcbd1's: PMID:15182178 reports that it readily disproportionates.

Reason: Factually correct and experimentally grounded on this protein. Self-association is a structural property enabling both functions rather than a function in its own right - the PCD active site is formed at a subunit interface and the DCoH-HNF1 assembly is built from dimers - so it is supportive, not core. The paralog-specific instability of this tetramer is biologically interesting (the paper proposes it places DCoH2 regulation under thermodynamic rather than kinetic control) but that is a property of the oligomer, not a distinct molecular function.

Summary: Transferred from human PCBD2 (Q9H0N5), whose source annotation is an IDA from PMID:11980910 showing that DCoHm enhances HNF1alpha transcriptional activity on a beta-fibrinogen promoter reporter. A core process for this gene.

Reason: The orthology is unambiguous and the source evidence is direct, so the transfer is sound. Mouse-specific evidence for the same conclusion now exists independently: Kazgan et al. show that an acetylation-mimicking DCoH2 mutant loses the ability to activate a mouse FXR promoter reporter in the mouse Hepa1-6 line, and that DCoH2 increases Hnf1a dimerization. Pcbd2 is a physical component of the activating complex rather than merely upstream of it, so involved_in is the right qualifier - note this is stronger than the acts_upstream_of_or_within used for the equivalent Pcbd1 annotation. The molecular function underlying this process is missing from GOA and is proposed as NEW below.

Summary: Proposed new annotation. The molecular function underlying the GO:0045893 process Pcbd2 already carries: it binds Hnf1a, increases Hnf1a dimerization, and an acetylation-mimicking mutant loses reporter activation in a mouse hepatic line. GOA currently has no annotation at all from PMID:24389307 despite that paper being UniProt's source for both FUNCTION and SUBCELLULAR LOCATION.

Reason: The term fits the definition directly: "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." Pcbd2 binds Hnf1a, a DNA-binding transcription factor, and increases transcription of its targets; the four coactivator mechanisms the definition lists afterwards are introduced as examples, so dimer stabilisation not being among them is not an exclusion. On the participation test, the entity doing the work is Pcbd2 itself - it is a component of the activating assembly, not an upstream requirement - and the KQ-mutant result is loss of coactivation by the protein, not loss of an input to someone else's activity. Comparator check, because the absence is systematic: QuickGO returns GO:0003713 for PCBD1 in human (TAS), mouse (ISO) and rat (IDA), and for PCBD2 in no species. That pattern usually means the term does not mean what one assumes, but here the better explanation is visible - human PCBD2 carries involved_in GO:0045893 with IDA from PMID:11980910, the paper that demonstrates its coactivation, so curators recorded the process for PCBD2 and both process and function for PCBD1. The function was not declined; the MF row is simply missing. GO:0003713 is neither ancestor nor descendant of GO:0045893 (one MF, one BP), so this is not redundancy. Conservatively this could be left as a question for MGI; it is proposed because the mouse-specific experimental evidence exists, is cached in full text, and is currently unused.