TLR3 (human, O15455) curation notes

Reviewer: annotation-reviewer (INNATE_IMMUNITY batch 1, Toll/TLR axis).
Deep research: TLR3-deep-research-falcon.md arrived after the review was drafted
(the earlier run had failed); see the cross-check section at the end.
The review itself rests on the cached publications in publications/, the
UniProt record, and OLS/QuickGO term checks. No file named
-deep-research-<provider>.md was written by hand.

Core biology

TLR3 is the endosomal dsRNA-sensing Toll-like receptor. Architecture: N-terminal
LRR ectodomain (horseshoe), single TM helix (705-725), cytoplasmic TIR domain
(754-897) [UniProt O15455 features].

• Direct dsRNA binding to the ectodomain, saturable/specific/reversible, affinity
  rising with acidic pH and ligand length; 40-50 bp dsRNA is the smallest that
  forms a stable complex and activates the receptor
  PMID:18172197.
• Ligand-induced dimerisation is the minimal signalling unit
  PMID:18172197.
• Binding site mapped to H539/N541 on the glycan-free lateral surface
  PMID:16720699.
• Specificity: dsRNA yes; ssRNA and dsDNA no
  PMID:12054664.

Trafficking / location

• Resting TLR3 in ER; moves to dsRNA-containing endosomes on stimulation
  PMID:16858407.
• ER -> Golgi -> endosome, chaperoned by UNC93B1; cathepsin B/H cleavage (aa 252-346)
  yields the signalling-competent, predominant endosomal form
  [PMID:22611194 "newly synthesized endogenous TLR3 is transported through the ER and Golgi apparatus to endosomes, where it is rapidly cleaved"; "TLR3 proteolytic processing is essential for its function"].
• Ligand interaction requires acidic pH
  PMID:16144834.
• Surface expression is cell-type dependent: endosomal only in myeloid DCs, but both
  surface and endosomal in fibroblasts, macrophages, epithelial cells
  PMID:21266579.
  This is why I split the plasma-membrane rows: located_in (TAS) kept as non-core,
  but is_active_in (IBA/IDA/IEA) marked over-annotated because signalling needs
  endosomal acidification.
• Termination: c-Src activates ZNRF1, which K63-ubiquitinates TLR3 at K813 for
  lysosomal degradation PMID:37158982.

Signalling

• MyD88-independent; uses TRIF/TICAM1 via the TIR domain
  [PMID:12471095 "Dominant-negative TRIF inhibited TLR3-dependent activation of both the NF-kappaB-dependent and IFN-beta promoters."; "TRIF associated with TLR3 and IFN regulatory factor 3."].
• Outputs: type I/III IFN (IRF3), canonical NF-kappaB, MAPK/JNK2, and (when
  cIAPs/caspase-8 blocked) RIPK3 necroptosis PMID:21737330.
• Human fibroblasts express only TLR3 and respond to poly(I:C) with IFN-beta,
  IRAK4-independently PMID:16286015.

Disease / physiology

• Inherited TLR3 deficiency -> herpes simplex encephalitis; TLR3 largely redundant
  for most microbes, vital for anti-HSV-1 immunity in CNS
  PMID:17872438.
• Also sensor of endogenous dsRNA in sterile skin-wound inflammation via JNK2
  PMID:27830702.

Project-question findings

• PRR status (Q from TLR_FAMILY.md): TLR3 is a bona fide PRR - it binds the
  microbial product (dsRNA) directly, unlike Drosophila Toll (binds Spaetzle
  cytokine). GO:0038187 accepted. It is on the GO:0002224 branch (GO:0034138 TLR3
  pathway), not GO:0008063 (fly Toll pathway).
• No TLR-specific MF term exists; GO:0038187 PRR activity + GO:0003725 dsRNA
  binding + GO:0004888 transmembrane signalling receptor activity are the right set.
  No new MF term proposed.

Curation decisions of note

• Many generic GO:0005515 protein binding IPI rows (PIK3R1, TICAM1, WDFY1, UNC93B1):
  REMOVE as uninformative (per project policy); interactions not disputed. The TIR->TRIF
  interaction is captured by GO:0004888.
• GO:0042802 identical protein binding (homodimer): MODIFY -> GO:0042803 protein
  homodimerization activity, since ligand-driven homodimerisation is functional. The two
  isolated-TMD self-association papers (PMID:23155421, PMID:25217833) kept as non-core
  (in vitro peptide behaviour, not full-length receptor).
• GO:0007250 activation of NIK activity (NAS): MODIFY -> GO:0043123 canonical NF-kappaB;
  TLR3 does not activate the non-canonical NIK pathway.
• GO:0006972 hyperosmotic response (NAS, PMID:12054664): REMOVE - cited abstract has
  nothing on osmotic stress; likely an error.
• GO:0042742 defense response to bacterium (TAS, PMID:10426995): REMOVE - cited paper
  is about TLR2/lipoproteins; TLR3 redundant for antibacterial defence.
• GO:0071260 cellular response to mechanical stimulus (IEP, PMID:19593445): UNDECIDED -
  the cited full text is about BAD/prostate cancer and does not mention TLR3 at all.
  Flagged in references (relevance NONE) and as a suggested question. Not removed per the
  "do not overrule experimental annotation from incomplete evidence" rule (though IEP is
  weak; a curator should recheck the identifier).
• GO:0032729 positive regulation of type II IFN production (IDA, PMID:16286015):
  MARK_AS_OVER_ANNOTATED - IFN-gamma appears only in the paper's intro citing IRAK4
  literature; the paper's poly(I:C) experiments measure type I/III IFN. IFN-gamma is an
  indirect, cell-extrinsic effect.
• Transcription-regulation IEA rows (GO:0045944): MARK_AS_OVER_ANNOTATED - TLR3 acts far
  upstream of transcription factors.
• GO:0045766 positive regulation of angiogenesis (IEA): UNDECIDED - could not trace the
  mouse source or find supporting cached evidence.

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

Compared the completed review against TLR3-deep-research-falcon.md.
No annotation action, term, description or core function was changed; two
suggested_questions were added for claims the report makes that could not be verified
against a primary paper.

Agreement:
- Endosomal/lysosomal dsRNA sensing with binding favoured at acidic pH (<=6.5), UNC93B1
chaperoning from the ER, and cell-type-dependent surface expression on fibroblasts and
epithelial cells. This is exactly the basis for splitting the plasma-membrane rows into
located_in (non-core) and is_active_in (over-annotated).
- Structure-based, sequence-independent recognition of the ribose-phosphate backbone by the
23-LRR ectodomain with sites at both ends of the horseshoe; ~40-50 bp minimum for a
stable dimer. Matches the accepted dsRNA-binding rows and the H539/N541 mapping.
- TRIF/TICAM1 recruitment to the TIR domain with TBK1-IKKepsilon-IRF3 for type I interferon
and TRAF6/RIPK1-TAK1-IKK for NF-kappaB and AP-1; RIPK3/MLKL necroptosis when caspase-8 is
blocked. Matches the accepted interferon-beta and NF-kappaB rows and the non-core
necroptotic row.
- Endogenous/damage-associated RNA as a ligand, consistent with keeping
inflammatory response to wounding as non-core rather than core.

Additions not acted on (the report cites only secondary reviews; no primary paper was
available to quote, so nothing was asserted):
- Lateral multimerisation of TLR3 dimers along long dsRNA (reported 2023 cryo-EM), with

=90 bp needed for robust signalling. Recorded as a suggested question; the review still
follows PMID:18172197 in treating one dimer as the minimal signalling unit.
- A reported TRIF-dependent MyD88 contribution to TLR3-driven NF-kappaB in mouse
macrophages (2025). Recorded as a suggested question; the MyD88-independent model
(PMID:12471095, PMID:16286015) is retained, and the description was not altered.
- TRIM56 as a scaffolding positive regulator of TRIF, and ligand-induced Tyr759/Tyr858
phosphorylation. Both concern partners or modifications of TLR3, not new TLR3 functions.
- Disease associations (autoimmunity, allergy, calcific aortic valve disease) and
poly(I:C)/poly-ICLC therapeutics. Indirect and pleiotropic; no NEW process terms.

Conflicts: the report describes TLR3 surface expression on fibroblasts and epithelial cells
as functional for detecting extracellular dsRNA, which sits against this review marking the
is_active_in plasma membrane rows as over-annotated. The report supports this only from
secondary reviews, so the decision stands and the existing suggested question on
cell-surface TLR3 signalling covers it.