TLR21 (chicken, UniProt A0A8V0ZKW5) — curation notes

These notes rest on the UniProt record, the GOA rows and the cached publications;
the Falcon deep-research report arrived after the review was written and is
cross-checked at the end of this file.

Entry identity (TrEMBL, unreviewed)

A0A8V0ZKW5_CHICK is an unreviewed, Ensembl-derived entry of 1083 aa, submitter
name "Toll like receptor 21", gene name TLR21, whose only cross-reference is
Ensembl ENSGALG00010023066 / ENSGALP00010034014. Ensembl gives that gene as
"TLR21 toll like receptor 21 [Source:NCBI gene; Acc:415623]" on chromosome 11, so
the entry is the right gene even though it carries no RefSeq or GeneID link of its
own. (The project's candidate table listed another accession, A0A8V0ZYL3; both are
Ensembl models of chicken TLR21.) The entry has a TIR domain (915..1055), five
LRR_8 repeats and one predicted transmembrane segment; its PANTHER subfamily is
PTHR24365:SF545, the same subfamily as zebrafish tlr21, which is consistent with
the TLR21/TLR22/TLR13 branch rather than with any mammalian TLR9 orthology.

Caveat worth recording: this Ensembl model is longer than the cDNAs used in the
functional literature, and the functional work was done on cloned chicken TLR21
rather than on this accession; the review therefore treats the published work as
evidence about the chicken TLR21 gene, which is what GO annotations are made to.

Established biology

Annotation decisions (summary)

GOA has only five electronic rows for this entry; none is experimental, and none
mentions the CpG-DNA function.

Ligand-specificity question for the project: the transfer that would have been wrong
here — a TLR9 pathway or ligand term — is absent from GOA, and the correct
receptor-specific term (GO:0035682) exists and is now proposed. The general pattern
across the batch is that the numbered pathway terms are safe to use where the
receptor identity is unambiguous, whereas donor-specific ligand terms
(triacyl lipopeptide binding on chicken TLR15) are not.

Deep-research cross-check (2026-09-30)

Compared TLR21-deep-research-falcon.md against the finished review. The report is about the
right gene and the right accession, and it agrees with the review on everything the review
asserted: unmethylated CpG DNA as the ligand class, endosomal site of action, functional rather
than orthologous relationship to mammalian TLR9, pattern recognition receptor rather than enzyme,
and the reservation that MyD88 dependence for this receptor is inferred rather than directly
demonstrated.

One substantive addition, verified and taken up as the review's only new annotation:

The two tests in CLAUDE.md were run before proposing it. Participation: the receptor performs
the sensing step that switches the output on, rather than being a substrate of it. Comparator:
the mammalian receptors in the same role already carry this exact term from equivalent
experiments — human TLR9 and mouse Tlr9 hold GO:0032728 by IDA (PMID:16286015 and
PMID:15356140 respectively, checked in QuickGO), and human TLR3 holds it by IDA and IMP — so
the absence of an interferon term on the avian sensor is a gap in this species' annotation
rather than a convention. GO:0032728 is neither an ancestor nor a descendant of the proposed
GO:0035682, so it is not redundant with it.

Not taken up: the CpG-adjuvant and in-ovo vaccination work, the intravaginal CpG study and the
CpG-K3 whole-animal study are all applications of the ligand rather than evidence about the
gene product's own activity, and the behavioural and metabolic readouts are far downstream. The
report also repeats the (correct) point that no new primary localisation study exists, so the
endosome call still rests on the 2009-2010 papers.

Actions: every existing annotation unchanged; one NEW row added. Other changes are to
description, core_functions.directly_involved_in, two new references (PMID:37951324 and the
Falcon file, the latter cited only for the endosome consensus) and one new supported_by quote.