Innate Immune System Pathways Across Animals

IN_PROGRESS BIOLOGY_DOMAINFLAGSHIP

Species: human, mouse, CHICK, DANRE, XENTR, DROME, ANOGA, worm, NEMVE, TACTR

Innate Immune System Pathways Across Animals

Bottom line: this project reviews the GO annotation of the innate immune
signaling machinery of animals — Toll and Toll-like receptors, their TIR-domain
adaptors and kinase relays, NLRs and inflammasomes, cytosolic nucleic-acid
sensors (RIG-I-like receptors, cGAS-like receptors and STING), C-type lectin and
complement-type recognition, and the Rel/NF-kappaB and IRF transcriptional
outputs — organised by gene family and compared across ten species from
sea anemone to human. Batch 1, the Toll/TLR axis, is reviewed: 37 gene
products in six species (all ten human TLRs with their co-receptors and
adaptors, mouse, chicken and zebrafish lineage-specific TLRs, the Drosophila
Toll pathway from Spätzle to Dorsal, C. elegans TOL-1 and Nematostella
MyD88), covering 2,734 existing annotations (plus 14 proposed new ones). Together with 28 genes reviewed
earlier by other projects, 65 of the 202 candidates
(candidates-resolved.tsv) are done.
The main results: the fly Toll-pathway term has leaked onto human IRAK4 and
MYD88 by phylogenetic inference; LPS and lipopeptide receptor terms were
spread across TLR partners that do not bind those ligands; and one mis-cited
paper (PMID:19593445) supports a "response to mechanical stimulus" row on six
TLR-axis genes. Details are in Batch 1 results and on the
TLR family sub-page.

We scoped this because innate immunity is where GO annotation is most exposed
to cross-species over-propagation. The same families (TLRs, NLRs, TIR proteins,
lectins) are massively and independently expanded in some lineages and lost in
others, and the canonical names hide real functional divergence — the
Drosophila Toll receptor binds an endogenous cytokine, not a microbial ligand;
chicken lacks RIG-I; fish TLR22 has no GO pathway term at all. A
family-by-lineage view is the right unit for checking whether IBA, ISO and
InterPro2GO annotations follow the biology.

Scope

In scope: germline-encoded recognition, signaling and transcriptional output of
the innate immune response in Metazoa (GO:0045087 innate immune response and
its signaling-pathway descendants), plus a small set of effectors used as
end-points.

Out of scope, and handled elsewhere:

Topic Where
cGAS-STING in human detail CGAS_STING_PATHWAY
NLRP3 inflammasome in human detail NLRP3_INFLAMMASOME
C. elegans surveillance immunity and the IPR CAEEL_SURVEILLANCE_IMMUNITY
Prokaryotic immunity (CBASS, Thoeris, CRISPR) PROKARYOTIC_IMMUNITY_TERM_PREDICTION
Autoimmune disease genes AUTOIMMUNE
Adaptive immunity (TCR, BCR, MHC, VLRs) not covered
Plant NLRs and pattern-triggered immunity not covered; a natural follow-on

This project overlaps the two human-pathway projects deliberately: here those
genes appear as one member of a family traced across species, not as a pathway
in its own right.

Species

Chosen to span the main animal lineages that have well-studied innate immune
systems and a usable proteome in UniProt. Species codes follow repository
convention.

Code Species Lineage Why included
human Homo sapiens Mammalia Reference system; most experimental annotation
mouse Mus musculus Mammalia Tlr11, Tlr12, Tlr13, which have no counterpart among human TLR1–10; NAIP/NLRC4 and NLRP1B genetics
CHICK Gallus gallus Aves Bird-specific TLR15 and TLR21
DANRE Danio rerio Teleostei Fish-specific TLRs (TLR5b, TLR18–22); inflammasome GO-CAM exists
XENTR Xenopus tropicalis Amphibia Amphibian comparator (thin coverage)
DROME Drosophila melanogaster Insecta Toll and Imd pathways; the origin of the Toll paradigm
ANOGA Anopheles gambiae Insecta Complement-like TEP1 system against malaria parasites
worm Caenorhabditis elegans Nematoda TLR-independent p38/PMK-1 immunity
NEMVE Nematostella vectensis Cnidaria Early-branching animal with MyD88, NF-kappaB, cGAS and STING
TACTR Tachypleus tridentatus Chelicerata LPS- and beta-glucan-triggered protease cascade (factor C, factor G)

Lineages with very large expansions of innate receptor families — sea urchin
(Strongylocentrotus purpuratus), amphioxus (Branchiostoma floridae) and
oyster (Crassostrea gigas) — are not in the candidate list, because almost all
their entries are unreviewed TrEMBL records without gene names. They are
tracked at the family level instead. A count of UniProt entries carrying the
TIR domain (InterPro IPR000157) or the NACHT domain (IPR007111), taken
2026-09-30, shows the scale:

Taxon TIR-domain entries NACHT-domain entries
sea urchin (7668) 271 264
oyster (29159) 234 13
amphioxus (7739) 138 103
lamprey (7757) 40 94
Nematostella (45351) 17 15

These are entry counts, not gene counts: redundant and fragmentary TrEMBL
entries inflate them. They are shown only to mark where IBA and InterPro2GO
propagation will act on hundreds of paralogs at once.

Gene families

The organising unit. Each family is listed with the lineage-specific points
that matter for curation; per-gene accessions are in the checklist below.

Toll and Toll-like receptors (TLR)

Leucine-rich-repeat ectodomain plus a cytoplasmic TIR domain. In vertebrates,
each TLR recognises a class of microbial molecule (lipopeptide, dsRNA, LPS,
flagellin, ssRNA, CpG DNA) and signals through MYD88 and/or TICAM1 (TRIF). In
Drosophila, Toll (Tl) is activated by the cleaved cytokine Spätzle, produced by
a protease cascade downstream of the real pattern recognition proteins (PGRPs,
GNBPs). GO carries a numbered signaling-pathway term for most vertebrate TLRs,
including the lineage-specific TLR11, 12, 13, 15 and 21, but none for fish TLR22.
Detail: TLR family sub-page.

TIR-domain adaptors

MYD88, TIRAP (MAL), TICAM1 (TRIF), TICAM2 (TRAM) and SARM1 in human; Myd88 in
Drosophila; TIR-1 (the SARM1 orthologue) in C. elegans, which signals to the
p38 MAPK cascade rather than NF-kappaB. SARM1 has NADase activity, so its GO
annotation should reflect an enzyme, not only an adaptor. MYD88 is present in
Nematostella.

IRAK and Pelle kinases

IRAK1, IRAK2, IRAK3 (IRAK-M, a pseudokinase) and IRAK4 in human; Tube
(adaptor, death domain only) and Pelle (kinase) in Drosophila. IRAK3 is already
reviewed and is a test case for PSEUDOENZYMES: kinase
activity annotations propagated from active paralogs.

TRAF ubiquitin ligases and kinase relays

TRAF3 and TRAF6 (both reviewed), MAP3K7 (TAK1) with TAB1/TAB2, the IKK complex
(CHUK, IKBKB, IKBKG) and the IKK-related kinases (TBK1 reviewed, IKBKE). The
Drosophila Imd branch uses Tak1, Tab2 and an IKK formed by IKKbeta (ird5) and
Kenny (key); Drosophila IKKepsilon already has a folder from another project.

Rel/NF-kappaB and IkappaB

RELA and NFKB1 with the inhibitor NFKBIA in human; Dorsal (dl) and Dif with
the inhibitor Cactus (cact) downstream of Toll, and Relish (Rel) downstream of
Imd, in Drosophila; REL2 in Anopheles; a single NF-kappaB in Nematostella.
C. elegans has no NF-kappaB; its immune transcriptional output runs through
ATF-7, ZIP-2 and others.

Interferon regulatory factors (IRF)

IRF3, IRF5 and IRF7 as the TLR/RLR/STING outputs in vertebrates. IRFs are a
vertebrate-centred family and serve here as the check that type I interferon
terms are not propagated to invertebrates.

NOD-like receptors (NLR) and inflammasomes

NOD1, NOD2, NLRP1, NLRP3 (reviewed), NLRC4, NAIP, NLRP6 and CARD8 in human;
mouse Naip5, Nlrp1b and Nlrc4 for the genetics that defined NAIP/NLRC4 and NLRP1
ligand sensing; fish nod1/nod2 and the fish inflammasome caspases caspa and
caspb. Outputs: PYCARD (ASC), CASP1, CASP4 (reviewed), gasdermin D (reviewed),
IL1B and IL18. NACHT-domain expansions in sea urchin, amphioxus and lamprey are
tracked at family level only.

RIG-I-like receptors (RLR)

RIGI (formerly DDX58), IFIH1 (MDA5, reviewed) and DHX58 (LGP2) signal through
MAVS, with TRIM25 as the E3 ligase activator. Chicken has IFIH1 and MAVS but no
RIGI entry, consistent with the reported loss of RIG-I in chicken. The C.
elegans RIG-I homologue DRH-1 acts in antiviral RNAi, not via a MAVS pathway,
and is the test case for whether RLR signaling terms are propagated to it.

cGAS-like receptors (cGLR) and STING

Human CGAS and STING1 (reviewed), with fish, chicken, fly (cGlr1, cGlr2, Sting)
and Nematostella orthologues. Nematostella cGAS and STING are Swiss-Prot entries
(A7SFB5, A7SLZ2). Human biology is covered by
CGAS_STING_PATHWAY; this project asks whether
annotations of the invertebrate members reflect their own biochemistry, since
invertebrate STING proteins are ancestral and signal differently.

C-type lectin receptors (CLR)

CLEC7A (Dectin-1), CLEC4E (Mincle) and CD209 (DC-SIGN) with the SYK–CARD9
relay (CARD9 also connects to BCL10–MALT1). Lectin domains are among the most
widely propagated domains in GO; C. elegans CLEC-60 (reviewed) is a
non-receptor example.

Complement, thioester proteins and soluble recognition

MBL2 (reviewed), FCN2 and MASP2 (lectin pathway) and C3 in human; the
thioester-containing proteins TEP1 in Anopheles (reviewed, with its partners
LRIM1 and APL1C) and Tep1/Tep2 in Drosophila; human PGLYRP1 and PGLYRP2 as the
vertebrate peptidoglycan recognition proteins, compared with the Drosophila
PGRPs.

Peptidoglycan recognition proteins and GNBPs (insect)

Drosophila PGRP-SA, PGRP-SD, GNBP1 and GNBP3 feed the Toll pathway through the
proteases ModSP, Grass and SPE; Persephone (psh) senses microbial proteases and
Necrotic (nec) inhibits it. PGRP-LC and PGRP-LE are the Imd pathway receptors.
These are the true pattern recognition receptors of the fly, which is why
pattern recognition receptor activity on Toll itself would be an error.

Horseshoe crab clotting cascade

Factor C (LPS-activated serine protease), factor B, proclotting enzyme and
coagulogen form the LPS-triggered cascade; factor G (alpha and beta subunits)
is the beta-glucan branch. All six entries are Swiss-Prot. Factor C is the
case of a pattern recognition receptor that is itself a protease — a check on
whether GO has an MF for recognition-coupled protease activation.

Effectors

A small set of end-point genes: antimicrobial peptides (human DEFB4A, CAMP,
LYZ; Drosophila Drosomycin, Diptericin A, Metchnikowin, Attacin A, Cecropin A1,
Defensin; C. elegans LYS-7, SPP-1, NLP-29), interferon effectors (OAS1,
RNASEL, EIF2AK2, MX1) and Drosophila JAK-STAT (hop, Stat92E, upd3, TotA). They
are included to check the rule that an effector is not a participant in the
signaling pathway that induces it (see Do not add what curators deliberately
declined to add in CLAUDE.md).

Existing modules and GO-CAMs

Draft pathway modules already exist and should be grounded against these
reviews:

Two of those GO-CAMs, 5fb9cc0600000727 (TLR1-TLR2) and 5fce9b7300000030
(TLR2-TLR6), have GoCamReview files recording why their receptor activity is
wrong; the two TLR4 models (5f46c3b700001031, 6413ac9800000654) type TLR4 only
with generic signaling receptor activity and omit MD-2.

Production GO-CAMs in gocams/index.tsv relevant to the first batch
(human unless stated):

Model Title
5f46c3b700001031 MYD88-dependent TLR4 signaling pathway leading to NF-kappa-B activation
6413ac9800000654 MYD88-independent TLR4 signaling pathway leading to interferon production
5fb9cc0600000727 Triacyl lipopetide activation of TLR1-TLR2 complex
5fce9b7300000030 Diacyl lipopeptide and bacteriocin activation of TLR6_TLR2 receptor
663d668500002770 Bacterial flagellin recognition by TLR5
6918f23700000328 TLR9 signaling pathway leading to IRF5 activation
5f46c3b700002102 SPModule-TIRAP-MYDDOSOME
641ce4dc00000050 NOD2 signaling in response to muramyl dipeptide (MDP) leading to NF-kappa-B and MAPK kinase activation
5fadbcf000001101 SPModule IFIH1-MAVS
60418ffa00001019 NLRP3 inflammasome in pyroptosis via nlrp3, caspa, caspb etc. (zebrafish)
568b0f9600000284 Antibacterial innate immune response in the intestine via MAPK cascade (C. elegans)

No production GO-CAM in the index models the Drosophila Toll or Imd pathway.

GO terms in play

All ids checked against OLS on 2026-09-30.

Id Label Note
GO:0045087 innate immune response Root process for the project
GO:0002221 pattern recognition receptor signaling pathway Parent of the receptor-class pathways
GO:0002224 toll-like receptor signaling pathway Vertebrate TLRs
GO:0008063 Toll signaling pathway Drosophila Toll–Spätzle pathway
GO:0061057 peptidoglycan recognition protein signaling pathway Drosophila Imd pathway
GO:0035872 nucleotide-binding domain, leucine rich repeat containing receptor signaling pathway NLRs
GO:0002753 cytoplasmic pattern recognition receptor signaling pathway Cytosolic sensors
GO:0039529 RIG-I signaling pathway RLRs
GO:0140896 cGAS/STING signaling pathway cGLR/STING
GO:0038187 pattern recognition receptor activity MF for PRRs; no TLR-specific MF exists
GO:0001875 lipopolysaccharide immune receptor activity TLR4 complex
GO:0035591 signaling adaptor activity TIR adaptors
GO:0061702 canonical inflammasome complex Inflammasomes
GO:0140367 antibacterial innate immune response Used in the C. elegans GO-CAM

GO:0039528 (cytoplasmic pattern recognition receptor signaling pathway in
response to virus) is obsolete; any existing annotation to it should be read
against the current replacement.

Curation questions to test

  1. Is Drosophila Toll a pattern recognition receptor? Toll binds cleaved
    Spätzle, not a microbial molecule. Check whether GO:0038187 or
    GO:0002224-branch terms reach Tl, 18w or Toll-7 by IBA or InterPro2GO, and
    whether GO:0008063 is kept distinct from GO:0002224 in practice.
  2. Numbered TLR pathway terms across species. GO has terms for TLR1–13, 15
    and 21, but not TLR14, TLR16–20 or TLR22. Check how fish TLR22 and the
    TLR18–20 genes are annotated and whether a new term, or a general term, is
    right.
  3. No TLR-specific molecular function. TLRs are annotated to generic PRR
    activity (GO:0038187) or, for TLR4, LPS immune receptor activity
    (GO:0001875). Decide whether ligand-class-specific MFs are needed or whether
    the generic term plus has_input in GO-CAM is the intended pattern (compare
    the inflammasome sensor activity proposal in the NLRP3 review).
  4. TIR-domain propagation. TIR domains occur in adaptors, receptors and the
    NADase SARM1/TIR-1. Check whether InterPro2GO or IBA gives adaptor or
    receptor terms to TIR proteins that are enzymes, and vice versa.
  5. Interferon terms outside vertebrates. IRFs and type I interferons are
    vertebrate-specific; check for type I interferon terms on invertebrate STING,
    cGLR or NF-kappaB orthologues.
  6. Expanded families. With hundreds of TLR- and NACHT-domain entries in sea
    urchin and amphioxus, check what ligand-specific terms (e.g. TLR4 signaling)
    are propagated to lineage-specific paralogs that cannot share the ligand.
  7. Effectors versus participants. Antimicrobial peptides and interferon
    effectors are outputs of these pathways; apply the substrate/output test in
    CLAUDE.md before accepting a signaling-pathway term on them.
  8. Recognition by a protease. Horseshoe crab factor C recognises LPS and is
    itself the first protease of the cascade; record how GO represents this.

Deliverables

  1. Gene reviews for the candidate list, TLR axis first.
  2. A per-family summary of what the reviews changed, by lineage.
  3. Grounding of the five draft modules against the reviewed genes.
  4. Term requests arising from questions 2, 3 and 8, if the reviews support
    them.

STATUS

2026-09-30 — batch 1 (Toll/TLR axis) reviewed: 37 genes, all status: COMPLETE, all passing just validate, each with a history record. 65 of 202
candidates now have reviews. Next: batch 2 (see below).

Accessions come from scripts/resolve_candidates.py, which queries UniProt by
primary gene name per taxon (Swiss-Prot preferred, then the reference-proteome
entry) or verifies an accession pinned in candidates.tsv; rerun it after
editing candidates.tsv. "TrEMBL" marks unreviewed entries.

Batch 1 results

Actions per gene. "Rows" counts every annotation entry in the review: the
seeded GOA rows (exact duplicates in the GOA file are merged) plus the NEW
proposals added by the reviewer.

Species Gene Accession Rows Accept Keep non-core Modify Remove Over-annot. Undecided New
human TLR1 Q15399 81 58 11 4 4 3 0 1
human TLR2 O60603 146 86 21 8 20 9 1 1
human TLR3 O15455 126 76 32 3 6 6 3 0
human TLR5 O60602 33 28 2 2 0 0 1 0
human TLR6 Q9Y2C9 120 70 32 8 6 4 0 0
human TLR7 Q9NYK1 108 84 15 3 2 3 1 0
human TLR8 Q9NR97 54 32 5 8 4 4 1 0
human TLR9 Q9NR96 136 74 53 1 2 5 1 0
human TLR10 Q9BXR5 41 21 12 2 4 0 0 2
human CD14 P08571 151 93 47 6 4 1 0 0
human LY96 Q9Y6Y9 104 59 32 10 3 0 0 0
human UNC93B1 Q9H1C4 65 25 11 2 27 0 0 0
human MYD88 Q99836 229 112 36 41 35 4 1 0
human TIRAP P58753 109 41 29 19 15 4 1 0
human TICAM1 Q8IUC6 120 91 13 8 7 1 0 0
human IRAK4 Q9NWZ3 126 51 40 11 21 2 0 1
mouse Tlr11 Q6R5P0 9 1 3 2 0 0 2 1
mouse Tlr12 Q6QNU9 8 2 3 2 0 0 0 1
mouse Tlr13 Q6R5N8 21 10 4 6 0 0 0 1
DROME spz P48607 89 75 11 1 0 1 1 0
DROME Tl P08953 104 73 21 4 3 1 2 0
DROME Toll-7 Q7KIN0 24 18 4 0 0 0 2 0
DROME tub P22812 55 38 1 10 4 0 1 1
DROME pll Q05652 76 56 4 9 3 2 1 1
DROME cact Q03017 47 33 5 4 3 0 2 0
DROME dl P15330 114 79 20 6 7 0 0 2
DROME Dif P98149 81 63 11 4 3 0 0 0
CHICK TLR15 A0A8V0Z0H8 7 2 3 1 1 0 0 0
CHICK TLR21 A0A8V0ZKW5 6 0 2 3 0 0 0 1
DANRE tlr5b A0ACM8R384 5 1 2 2 0 0 0 0
DANRE tlr21 F1QMN8 10 1 4 3 1 1 0 0
DANRE tlr22 A0A2R8RTN4 4 1 1 1 0 0 1 0
DROME 18w A1ZBR2 20 14 4 0 0 2 0 0
DROME Myd88 A1Z7T8 32 25 2 1 3 1 0 0
worm tol-1 Q9N5Z3 7 4 1 1 0 0 0 1
human TLR4 O00206 275 145 72 5 30 22 1 0
NEMVE MyD88 A7RHZ4 5 5 0 0 0 0 0 0
all 37 genes 2748 1647 569 201 218 76 23 14

Findings

Deep-research cross-check

Most reviews were written before their Falcon deep-research report arrived, so
each was later compared against its report. A report claim could change a
review only after it was verified in a primary paper and quoted from it. The
reports largely agreed with the reviews. Across 37 genes the pass changed no
existing action and added two NEW annotations, both verified and
comparator-checked: positive regulation of interferon-beta production on
chicken TLR21 (PMID:37951324; human and mouse TLR9 carry the term) and defense
response to Gram-positive bacterium on mouse Tlr13 (PMID:32209688,
PMID:38358825; mouse Tlr2 carries it). Other changes were to descriptions and
supporting references, for example correcting the relationship between worm
TOL-1 and IKB-1 (IKB-1 opposes TOL-1; PMID:26279230). Several report claims
were rejected because cached primary papers contradict them (e.g. that
Spätzle binds Toll with low affinity, that TLR5 and TLR6 signal only through
MyD88, and that TLR4 binds amyloid-beta directly). Reviews whose report adds
nothing citable keep a validator warning that the deep-research file is not
cited; the Tlr11 and Tlr12 reports use the swapped literature names, so they
are deliberately not cited.

Batch 2 (proposed)

The Drosophila recognition and Imd modules (PGRP-SA, PGRP-SD, GNBP1, GNBP3,
modSP, Grass, SPE, psh, nec; PGRP-LC, PGRP-LE, imd, Fadd, Dredd, Diap2, Tak1,
Tab2, IKKbeta, key, Rel), the remaining human TLR-axis genes (LBP, TICAM2, IRAK1, IRAK2,
MAP3K7, TAB1, TAB2; TRAF3, TRAF6 and TBK1 are already reviewed) and the zebrafish/chicken TLRs not
yet covered (tlr3, tlr4ba, tlr5a, tlr18, tlr19.1, tlr20.2; chicken TLR3, TLR4,
TLR7).

Full checklist by module

Toll-like receptor signaling (29/51 reviewed)

NF-kappaB / IRF output (3/17 reviewed)

Drosophila Toll pathway (10/19 reviewed)

Drosophila Imd pathway (0/11 reviewed)

C. elegans p38 PMK-1 pathway (9/10 reviewed)

C. elegans epidermal/TGF-beta immunity (3/4 reviewed)

NLR signaling and inflammasomes (3/24 reviewed)

Cytosolic nucleic-acid sensing (4/23 reviewed)

C-type lectin receptors (0/5 reviewed)

Complement and soluble PRRs (2/12 reviewed)

Horseshoe crab coagulation cascade (0/6 reviewed)

Drosophila JAK-STAT (0/4 reviewed)

Interferon effectors (0/4 reviewed)

Antimicrobial effectors (2/12 reviewed)

NOTES

2026-09-30

2026-09-30 — batch 1