Reviewer journal. 188 existing annotations over 56 distinct GO terms; every decision was taken
once per term and applied to all instances of that term regardless of evidence code.
Two 2026 papers give partly conflicting accounts of where and in what oligomeric state NLRP3 acts.
Mateo-Tórtola et al., Nat Commun 2026 (PMID:42215451) - two parallel pathways:
PMID:42215451 and
PMID:42215451.
Wu & Wu, PNAS 2026 (PMID:42378282) - everything converges on the centrosome:
PMID:42378282,
PMID:42378282, and the bypass result
PMID:42378282.
The curation discipline that matters here: almost all of this is about localisation and
assembly state. It is CC and BP evidence, not MF. Neither paper measures a new molecular
activity of NLRP3, and neither was used to change a molecular-function action. What they did
change is the reason text on the MTOC (GO:0005815), trans-Golgi network membrane
(GO:0032588) and condensate-scaffold (GO:0140693) annotations, which now say explicitly
that the MTOC/dTGN route is a major but non-obligatory pathway.
The NEK7 disagreement is recorded in suggested_questions rather than adjudicated. The two
results may be compatible - both involve NLRP3 abundance thresholds - but nothing in either
abstract settles it, and PMC access to the PNAS paper was blocked, so only its abstract could
be read (recorded as full_text_unavailable: true).
GO:0140299 is defined as "Binding to a molecule and eliciting a change in the protein's
activity in response to the intracellular level of that molecule" (QuickGO). That definition
requires the sensed species to be bound. NLRP3 has no established direct agonist ligand at
all - it is thought to read a common cellular perturbation.
Worse, the papers cited as the IDA support argue against direct binding themselves:
PMID:18604214 and
PMID:18604214.
The convergence model is stated most cleanly by Chen & Chen:
PMID:30487600.
Call: MARK_AS_OVER_ANNOTATED, not REMOVE (these are experimental annotations recording
real activation events) and not ACCEPT (the term asserts a ligand that does not exist). The one
NLRP3 binding event that does satisfy the definition - PtdIns4P on the dispersed TGN - is
already separately annotated as GO:0070273, and that is where the ACCEPT went:
PMID:30487600.
Because this leaves NLRP3 with no ligand-independent sensor MF, I proposed a new term,
inflammasome sensor activity, under GO:0038187 pattern recognition receptor activity. The
gap is general (NLRP1, NLRP6, NLRC4, pyrin).
GO:0042834 peptidoglycan binding, TAS from PMID:15967716, a 2005 NLR review. The abstract
contains no peptidoglycan claim, and no primary study ever demonstrated NLRP3-peptidoglycan
binding. This descends from the mid-2000s proposal that NALP3 senses bacterial peptidoglycan and
MDP; the dedicated peptidoglycan sensors turned out to be NOD1/NOD2. REMOVE is appropriate here
and only here: it is a curator assertion from a review, not an experimental call, so no
curator's reading of a primary dataset is being second-guessed. Leaving a ligand-binding MF
asserted for a protein with no known ligand is the kind of machine-readable falsehood downstream
tooling believes.
GO:0043565 (IBA + ISS), GO:0140297 DNA-binding transcription factor binding (IBA + ISS),
GO:0005634 nucleus, GO:0045944, and the four Th2 terms (GO:0002830, GO:0032753,
GO:0045630, GO:2000553) - nine terms, one source. Traced through the GOA WITH/FROM
columns: the IBAs come from node PANTHER:PTN000648032 with donor MGI:2653833 (mouse Nlrp3);
every ISS has donor UniProtKB:Q8R4B8 (mouse Nlrp3). Both trails end at Bruchard et al. 2015:
PMID:26098997 and
PMID:26098997.
Against it: single laboratory; mouse only; a published corrigendum (PMID:26580508, verified in
PubMed as Nat Immunol 2015;16(12):1292); no human experimental evidence; and no NLRP3 DNA-binding
domain or NLRP3-DNA structure in the decade since, despite very heavy structural work on this
protein. In its favour: the NACHT-LRR architecture is shared with CIITA, a genuine
transcriptional activator, so it is not biologically absurd.
Call: KEEP_AS_NON_CORE across all nine, with the provenance spelled out. Not REMOVE -
surprise is not evidence. Not challenged on propagation grounds either: the IBD node placement
follows correctly from the mouse experimental annotation, so no propagation_review was
fabricated; the issue is the strength of the source, not the phylogeny. The question of whether
these should be retired is put to experts in suggested_questions.
Four annotations invert the direction of the biology:
- GO:0032691 negative regulation of IL-1beta production, IMP from PMID:12483741 (NOMID)
- GO:0050728 negative regulation of inflammatory response, IMP from PMID:12483741,
PMID:16531551, PMID:17178985
- GO:0002674 negative regulation of acute inflammatory response, IMP from PMID:11687797
All four are CAPS genotype-phenotype studies. CAPS mutations are dominant gain-of-function
alleles causing constitutive inflammasome activity and excess IL-1beta - which is precisely why
two of these papers are reports of clinical improvement on IL-1 receptor antagonism. GOA already
carries the positive counterparts with 13 (GO:0032731) and 9 (GO:0050729) annotations.
Call: MODIFY with the positive term as replacement, not REMOVE. The patient data are
sound; only the direction of the inference is wrong, so a sign correction is the right
instrument. One alternative reading is recorded in the reason for completeness: early work did
report that wild-type cryopyrin suppresses NF-kappaB (PMID:14662828), which could motivate a
negative annotation - but that is captured separately by GO:1901223, and these references are
not NF-kappaB assays.
GO:1901223 negative regulation of non-canonical NF-kappaB (IBA + IDA from PMID:14662828:
"Transfection of full-length CIAS1 or either of two shorter, naturally occurring isoforms
dramatically inhibited TNF-alpha-induced activation of NF-kappaB reporter activity") and
GO:1901224 positive regulation of the same pathway (IPI from PMID:15817483, WITH/FROM
UniProtKB:Q9ULZ3 = ASC), consistent with PMID:11786556.
Both are overexpression reporter assays from the early 2000s; neither has been tested at
endogenous levels. Both KEEP_AS_NON_CORE, deliberately, so the contradiction stays visible
in the record rather than being silently resolved by a reviewer who has no basis to pick a side.
GO:0060090 molecular adaptor activity and GO:0030674 protein-macromolecule adaptor activity
are both IDA from PMID:1189953. Fetched and checked: that identifier resolves to
"[Profanities and the profane person].", Alva Quiñones J, Acta Psiquiatr Psicol Am Lat 1975 - a
Spanish-language psychiatry abstract. It is a digit-dropped PMID:31189953 (Sharif et al.,
Nature 2019, NEK7-licensed NLRP3 activation), which supports the neighbouring annotations made on
the same dates. Recorded as correctness: WRONG_IDENTIFIER / relevance: NONE in
reference_review; the two annotations are MODIFYed to GO:0035591 signaling adaptor activity
on their merits (both are under-specific parents).
GOA carries mitochondrion (IEA + EXP PMID:21124315 + IDA PMID:23582325) and ER (IEA + ISS)
alongside dTGN and MTOC. The mitochondrial model is real experimental work -
PMID:23582325 and
PMID:21124315 - but
it has been largely displaced by the dTGN/MTOC account, which explains activation by
structurally unrelated stimuli, and neither 2026 paper invokes mitochondria.
KEEP_AS_NON_CORE for both, with the displacement stated.
Worth flagging because it feeds the 2026 dispute. PMID:35114687 versus PMID:35254907 - the latter paper describes a dodecamer. This does not change any action
(GO:0042802 and GO:0051260 are ACCEPTed either way) but it is recorded in those reason
fields and raised in suggested_questions.
Five, each a distinct mechanistic step rather than five names for one thing:
1. GO:0035591 signaling adaptor activity - PYD filament nucleates ASC polymerisation.
2. GO:0140608 cysteine-type endopeptidase activator activity - caspase-1 activation, the output.
3. GO:0070273 phosphatidylinositol-4-phosphate binding - dTGN recruitment; the only
well-established direct binding event.
4. GO:0016887 ATP hydrolysis activity - the NACHT nucleotide switch between the ADP-bound cage
and the ATP-bound active disc; also the MCC950 site.
5. GO:0140693 molecular condensate scaffold activity - centrosomal condensation, with the
MTOC-distal caveat.
Deliberately not core: molecular sensor activity (see above), bare protein binding (24 IPI
annotations, MARK_AS_OVER_ANNOTATED), apoptotic process (NAS, a relic of the "Apaf1-like" naming
era - NLRP3's death modality is pyroptosis), and the entire Th2/transcription cluster.