TLR15 (chicken, UniProt A0A8V0Z0H8) — 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)

A0A8V0Z0H8_CHICK is an unreviewed Ensembl-derived entry, 924 aa, named
"Toll-like receptor 15" with gene name TLR15, cross-referenced to
Ensembl ENSGALG00010016958, RefSeq NP_001385167 and GeneID 421219. Ensembl gives
that gene as "TLR15 toll like receptor 15 [Source:NCBI gene; Acc:421219]" on
chromosome 3, so this entry is the right gene. This matters because the project's
candidate table listed a different accession, Q2XQ10, whose submitter name is
"Toll-like receptor 2"; the entry fetched here carries the matching name and the
NCBI/Ensembl cross-references.

Architecture is canonical for a TLR: leucine-rich repeats (Pfam LRR_1, LRR_8), two
predicted transmembrane segments (59..80, 709..732) and a TIR domain (761..902),
PANTHER family PTHR24365 (subfamily SF17, named after TLR2 — consistent with the
family placement below rather than with orthology).

Established biology

• TLR15 is restricted to birds and reptiles and is not the orthologue of any
  mammalian TLR: PMID:23066147 and
  reptilian genomes, which places its origin after the bird/reptile–mammal split.
  Phylogenetically it falls near, but not inside, the TLR1/2/6/10 group:
  PMID:16495540.
• It is a cell-surface receptor:
  PMID:21383168.
• Its activation mechanism is unique among TLRs: it is not activated by binding a
  microbial molecular pattern but by proteolytic cleavage of its own ectodomain by
  secreted microbial proteases —
  PMID:21383168
  and stimulation of NF-kappa-B-dependent transcription; removing the whole
  ectodomain makes the receptor constitutively active. The paper's own controls
  exclude the classical ligands: a panel including
  PMID:21383168
  with and without candidate chicken heterodimerisation partners, and
  PMID:21383168.
• An independent study found activation by yeast lysates:
  PMID:23066147
  PMID:23066147
  agonists; the activity was abolished by heat or by the serine-protease inhibitor
  PMSF, which is consistent with the protease-activation mechanism rather than with
  recognition of a yeast cell-wall carbohydrate (zymosan preparations did not
  activate).
• Transcript induction after Salmonella infection was the finding that first
  described the receptor (PMID:16495540); that is expression data, not evidence of
  ligand recognition.

Annotation decisions (summary)

The interesting rows are the two IBA annotations transferred from human TLR2
(via PANTHER:PTN002808136).

• GO:0042497 triacyl lipopeptide binding (IBA from UniProtKB:O60603/TLR2) —
  REMOVE. This is exactly the ligand-specificity transfer the project set out to
  check. TLR15 is not a TLR2 orthologue but a lineage-specific receptor, and the
  cleanest experimental test available reports that di- and triacylated lipopeptides
  do not activate it, with or without candidate chicken co-receptors.
• GO:0043235 signaling receptor complex (IBA, part_of, same donor) — kept as
  non-core. The donor basis is the TLR2/TLR1 heterodimer, which TLR15 does not form
  as far as anyone has shown, but TLRs signal as dimers and the truncated receptor's
  autoactivation was interpreted as self-dimerisation, so complex membership is
  defensible at this generic level while the identity of the complex is unknown.
• GO:0002224 (IBA) modified to GO:0035681 toll-like receptor 15 signaling pathway.
  The receptor-specific term exists, it names this receptor, and two independent
  studies show NF-kappa-B activation initiated at this receptor.
• GO:0038023 signaling receptor activity (IBA) — accepted deliberately, and not
  upgraded to GO:0038187 pattern recognition receptor activity. GO:0038187 requires
  combining with a pathogen-associated molecular pattern; TLR15 is activated by
  cleavage of its ectodomain, and the proteases that cleave it are enzymes rather
  than patterns bound by the receptor. The generic receptor term is the accurate one
  here, which is worth recording explicitly because every other reviewed TLR in this
  batch takes the GO:0038187 upgrade.
• GO:0005886 plasma membrane (IBA) accepted: unusually for an IBA location call in
  this family, it is confirmed experimentally for this receptor by surface
  biotinylation and confocal microscopy.
• GO:0007165 and GO:0016020 kept as non-core.

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

Compared TLR15-deep-research-falcon.md against the finished review. The report is about
the right gene and is unusually useful here, because it surfaces four papers later than the
two the review rested on. Each was fetched and read before use.

Agreement: cell-surface localisation; activation by proteolytic cleavage of the ectodomain
by secreted microbial proteases as the best-supported mechanism; MyD88-dependent signalling
to NF-kappa-B; no mammalian orthologue; explicit warning that the Salmonella, Clostridium
and viral associations are expression-level and should not be treated as direct ligand
recognition. The report's own summary line, that
[file:CHICK/TLR15/TLR15-deep-research-falcon.md "pathogen-associated protease sensing remains its best-supported molecular function"],
coincides with the review's deliberate refusal to upgrade GO:0038023 to GO:0038187.

Additions taken up, each verified in a primary paper:

• The receptor is not confined to birds and reptiles (publications/PMID_36605191.md).
  PMID:36605191;
  it is an ancient gnathostome lineage lost or pseudogenised in many groups, including mammals.
  description corrected. The same paper strengthens the REMOVE of GO:0042497 rather than
  weakening it:
  PMID:36605191
  and PMID:36605191
  — the lipopeptide-binding site of the TLR1/TLR2 pair is built in the three-domain
  architecture this receptor does not have. Both quotes added to that row's supported_by.
• The signalling complex is better constrained (publications/PMID_37079400.md). The
  cytoplasmic TIR domain was crystallised as a dimer with a TLR1-subfamily-unlike interface,
  PMID:37079400,
  and interface mutation abolished the dimer in solution. The GO:0043235 row stays
  KEEP_AS_NON_CORE, but its reason no longer says the composition is unknown: a homodimer is
  now the likely composition, while the isolated domain remains mostly monomeric and no
  full-length dimer has been shown. The same paper argues structurally for the adaptor,
  PMID:37079400,
  added to the GO:0002224 row.
• Independent receptor-dependence in a chicken cell background
  (publications/PMID_33514434.md, abstract-only). Overexpression in DF-1 fibroblasts
  augmented and PMID:33514434 on
  NF-kappa-B and NLRP3/IL-1-beta activation by Eimeria tenella sporozoites. Added to the
  GO:0002224 row.
• The one candidate protein ligand is itself a protease
  (publications/PMID_40639002.md): an E. tenella aspartyl protease
  PMID:40639002 co-immunoprecipitation. Cited in
  the GO:0038023 reason, because it reinforces protease activation rather than pattern binding.

Considered and not taken up:

• No NEW defence-response or inflammatory-response process term. The Eimeria evidence is
  strong enough to be worth recording but not to annotate: the causal readouts are signalling
  intermediates and cytokine transcripts in an overexpressing fibroblast line, the candidate
  ligand paper calls the receptor assignment "potential" and explicitly leaves other pattern
  recognition receptors open, and the cecal expression correlation with oocyst output is
  correlative. Raised as a suggested question instead.
• The Mycoplasma synoviae diacylated-lipopeptide reports do not reopen GO:0042497. They
  are induction and siRNA studies, the term removed is a binding term for triacyl
  lipopeptides, and the one direct test of acylated lipopeptides on this receptor was negative.
  Noted here as the main caveat a curator might raise against the REMOVE.
• Viral associations: the report itself says the receptor should not be annotated as a viral
  sensor, which matches the review.

Other change: the references list contained each of its three PMIDs twice; the duplicates
were removed.

Actions: unchanged for every existing annotation. Changes were confined to description, four
reason fields, six new supported_by quotes, five new references (four PMIDs plus the Falcon
file) and one new suggested question.