PDZD7 (human, Q9H5P4) — curation notes
2026-09-04 — annotation review of all GOA rows
Protein architecture
- 1033 aa, three PDZ domains (86–168, 210–293, 862–934), a harmonin-N-like (HN-like) domain
(InterPro IPR042786), a central proline-rich/disordered region, and extensive disorder in the
C-terminal third (UniProt feature table, PDZD7-uniprot.txt). No catalytic domain.
- Paralog of the two other Usher PDZ scaffolds: PMID:25406310; the 2009 discovery paper
makes the same point for harmonin/whirlin PMID:19028668.
- Three annotated isoforms (Q9H5P4-3 displayed, -1, -2); the deep-research report notes shorter
transcripts terminating before PDZ3 dominate adult retina while the developing cochlea expresses
N-terminal and rarer full-length forms. Isoform biology is unresolved and I made no
isoform-specific annotation calls.
Core function: ankle-link / USH2 complex scaffold
- Site of action: cytoplasmic face of the stereociliary membrane at the ankle region, i.e. the
tapered base of developing hair-cell stereocilia — explicitly not the tip/tip-link apparatus
[file:human/PDZD7/PDZD7-deep-research-falcon.md "Immunofluorescence and tagged-protein studies placed
PDZD7 immediately above the tapered stereociliary base, peripheral to the actin core. It overlaps
USH2A, ADGRV1 and WHRN but is distinct from the upper tip-link insertion site where MYO7A clusters."]
UniProt says the same [file:human/PDZD7/PDZD7-uniprot.txt "Note=Localizes at the ankle region of the
stereocilia."].
- Complex composition and colocalization PMID:25406310.
- Assembly logic (the key mechanistic result): PMID:25406310 and PMID:25406310. The complex is non-obligate — PMID:25406310 — and partner preference differs PMID:25406310.
- Homodimerization: PMID:25406310, matching UniProt [file:human/PDZD7/PDZD7-uniprot.txt "Homodimerizes (via PDZ2 domain).
Component of USH2 complex,"]. This multivalency is what the 2023 phase-separation work (Wang et al.,
Nat Commun, not in the publication cache) builds on.
- Direct partner mapping from the 2010 paper: PMID:20440071 and PMID:20440071.
- Requirement for partner localization: PMID:25406310; [file:human/PDZD7/PDZD7-deep-research-falcon.md "PDZD7 loss
disrupts or redistributes USH2A, ADGRV1 and WHRN in developing cochlear hair cells."]
- Loss of function: [file:human/PDZD7/PDZD7-deep-research-falcon.md "-null mice show congenital profound
deafness by auditory brainstem response"]; PMID:25406310. Human: DFNB57 PMID:19028668, with expression confirmed in the target tissue
PMID:19028668.
Retina / cilium — real but secondary
Nucleus — over-annotated
- UniProt lists Nucleus from PMID:20440071, but that paper qualifies its own nuclear signal:
PMID:20440071 followed by PMID:20440071. Only the perinuclear observation survives
PMID:20440071. No nuclear partner, NLS or function
has been described. Both nucleus rows (IEA and IDA) → MARK_AS_OVER_ANNOTATED (not REMOVE: a curator
read the full text and cultured-cell signal is reported).
Judgement calls worth recording
- GO:0032426 stereocilium tip (IBA) → REMOVE. The WITH/FROM of this row is mouse whirlin
(MGI:2682003) + rat whirlin (RGD:631330) only. Contrast the ankle-link-complex row on the same node
PTN000563297, which is seeded by mouse Pdzd7 (MGI:3608325). Whirlin is the canonical row-1
tip-elongation scaffold; PDZD7 has never been reported at tips. Classified as PROPAGATION_BAD /
WRONG_ORTHOLOG_OR_PARALOG + COMPARTMENT_OR_COMPLEX_MISMATCH — the sources are right for whirlin, the
transfer is what fails. Donor identities resolved through the GO API
(api.geneontology.org/api/bioentity/<id>): MGI:1919338 = harmonin, MGI:2682003 = whirlin,
MGI:3608325 = Pdzd7, RGD:1303329 = harmonin, RGD:631330 = whirlin.
- GO:0005576 extracellular region (HDA, PMID:22664934) → REMOVE. Pooled tear-fluid MALDI-TOF-TOF de
novo proteomics from a breast-cancer biomarker study; PDZD7 is never mentioned in the text. No signal
peptide, no TM segment; contradicted by every targeted localization study.
- GO:0005515 protein binding. Three rows. The PMID:20440071 (USH2A, ADGRV1) and PMID:19028668
(USH1G/SANS) rows → MODIFY to GO:0030674 protein-macromolecule adaptor activity, which is what the
assays actually demonstrate. The PMID:36115835 row → MARK_AS_OVER_ANNOTATED: >30 partners imported
from one holdup PDZ-PBM affinity matrix, including HPV E6 / HTLV1 Tax and non-co-expressed human
proteins, with the authors themselves noting PMID:36115835 and PMID:36115835.
- GO:0050910 detection of mechanical stimulus … sound (ISS) → MARK_AS_OVER_ANNOTATED. Reduced MET
currents in Pdzd7 mutants are secondary to bundle disorganization; PDZD7 has no described association
with the MET channel complex (TMC1/2, TMIE, LHFPL5, PCDH15) and is absent from tip-link insertion
sites. acts_upstream_of_or_within keeps it defensible, so not REMOVE.
- GO:0045184 establishment of protein localization (ISS) → MODIFY to GO:1990778 protein localization
to cell periphery. GO has no "protein localization to stereocilium/ankle link" term (checked
QuickGO: only stereocilium GO:0032420, stereocilium base GO:0120044, ankle link GO:0002141 exist on
the CC side). GO:1990778 is the closest informative existing term; a stereocilium-specific term would
be the right long-term fix.
- GO:0060113 inner ear receptor cell differentiation (NAS, ComplexPortal) → MODIFY to GO:0060088.
Hair cells differentiate; what fails in mutants is bundle organization.
Open questions
- Do human PDZD7 isoforms differ in ankle-link vs periciliary function? Mouse data suggest
tissue-specific isoform usage but human isoform-level data are absent.
- Is the "blocks inhibition of adenylate cyclase activity mediated by ADGRV1" statement in the UniProt
FUNCTION line (ECO:0000250 from mouse E9Q9W7) supported well enough to warrant a signalling-related GO
annotation? Not annotated in GOA; left alone here.
- Is there any evidence for the ankle-link condensate (LLPS) in vivo at physiological concentrations?
The 2023 Nat Commun work is in vitro/ex vivo only; no GO annotation was proposed on that basis.