Research notes: A0A3B6GK97 (WHEAT) — PNPLA/patatin domain-containing protein

Deep-research providers (falcon, perplexity) were not available in this environment
(no API keys; the falcon attempt failed with a template/credentials error). Per project
guidance, the manual research below is recorded in this notes file rather than a
-deep-research-{provider}.md file. No PMIDs are cited in the GOA (annotations are all
homology-based IEA), so fetch-gene-pmids returned nothing to cache.

Identity

Domain / family

Functional inference (family-level, no direct evidence for this protein)

GOA annotations to review (from QuickGO; both IEA)

  1. GO:0006629 lipid metabolic process — IEA, GO_REF:0000002 (InterPro2GO),
    WITH InterPro:IPR002641 (PNPLA domain). Defensible: the PNPLA domain is a lipid hydrolase
    module; lipid metabolic process is the correct (if general) BP. → ACCEPT (non-core / keep;
    this is the most specific safe BP given only a domain match).
  2. GO:0016787 hydrolase activity — IEA, GO_REF:0000117 (ARBA), WITH ARBA:ARBA00026276.
    Correct but very general MF. The patatin domain specifically supports
    acylglycerol/lipid acyl-hydrolase / phospholipase activity, so a more informative MF
    (e.g. GO:0052689 carboxylic ester hydrolase activity, or GO:0004620 phospholipase activity)
    would be preferable. However it is not wrong. → MODIFY toward a more specific hydrolase
    term, or KEEP_AS_NON_CORE if remaining conservative for a homology-only entry.

IBA lipase annotations missing from QuickGO GOA pull (confirmed in AmiGO) — IMPORTANT

The UniProt flat file lists two IBA molecular-function terms that are absent from the
QuickGO -goa.tsv pull used to seed the review (which returned only the 2 IEAs above).
I verified these directly against AmiGO/GOlr (via the GO MCP search_annotations on
UniProtKB:A0A3B6GK97; web record:
http://amigo.geneontology.org/amigo/gene_product/UniProtKB:A0A3B6GK97):

GO term Label Evidence Reference Assigned by Date
GO:0004620 glycerophospholipase activity IBA (ECO:0000318) GO_REF:0000033 GO_Central 2017-02-28
GO:0047372 monoacylglycerol lipase activity IBA (ECO:0000318) GO_REF:0000033 GO_Central 2017-02-28

QuickGO and AmiGO are complementary here: QuickGO returned only the 2 IEAs
(InterPro2GO + ARBA); AmiGO/GOlr returned only the 2 GO_Central IBAs. These IBA
annotations are manually-reviewed phylogenetic (PAINT) assertions that this protein is a
lipase (glycerophospholipase + monoacylglycerol lipase), which is stronger and more
specific than the generic IEA hydrolase term.

Handling decision (per maintainer): do NOT add these IBAs to existing_annotations —
the CI consistency check requires existing_annotations to mirror the seeded GOA tsv, so
adding rows it doesn't contain would fail CI. Instead they are recorded as out-of-band
knowledge: cited via GO_REF:0000033 (with the AmiGO URL) in the references list and
woven into the GO:0016787 MODIFY reasoning and core_functions. They reinforce (but do
not change) the MODIFY of the generic hydrolase IEA toward carboxylic ester hydrolase
activity, the immediate informative parent of both IBA lipase terms.

The UniProt flat file additionally lists GO:0006952 defense response (IEA:UniProtKB-KW),
also not in the GOA pull; defense response for this specific locus is unsupported beyond
keyword propagation and is not asserted here.

Bioinformatics analysis (subfamily + catalytic-site integrity) — IMPORTANT

Reproducible analysis in A0A3B6GK97-bioinformatics/ (see RESULTS.md). MSA of the query
against 13 characterized plant pPLAs (Arabidopsis pPLAI/II/III, rice PLP1/2, potato patatin)
with FAMSA + BioPython; input-driven control included.

Two findings:
1. Subfamily = pPLAII. The query is 42–50% identical to the pPLAII subfamily and rice
pPLAs, vs only ~23% to pPLAIII and ~16% to pPLAI; NJ tree places it in the pPLAII/rice/
patatin clade. So it is the soluble acyl-hydrolase (defense/wounding/stress) clade, NOT
the pPLAIII galactolipase/growth clade or the large iPLA2-like pPLAI. Single-domain
architecture + this placement make a membrane-trafficking role very unlikely (answers
the trafficking question raised in review).
2. The deposited 302-aa model lacks the catalytic serine. It is fully gapped through the
N-terminal patatin catalytic core: no oxyanion DGGG block and no catalytic-Ser
G-T-S-T-G nucleophile elbow (zero G-x-S-x-G motifs in 302 aa); it retains only the
C-terminal portion incl. the catalytic Asp (D121). Every active reference has GTSTG+DGGG;
the annotated-inactive PLP9 control also lacks GTSTG. Predicted catalytically inactive
as modeled. Most parsimonious = incomplete/incorrect gene model (~100–130 aa shorter
than orthologs, missing a clean N-terminal block), though a true degenerate pseudo-enzyme
cannot be excluded from sequence alone (would need genomic/homoeolog/RNA-seq checks).

Consequence: the GO_Central IBA lipase calls (GO:0004620, GO:0047372) were propagated
phylogenetically and do NOT verify active-site integrity, so they are not supported by the
current sequence. The review keeps the MODIFY → carboxylic ester hydrolase activity as the
family-level term but adds this explicit caveat, and flags the gene model for curation.

Bottom line

A plausibly-real but entirely uncharacterized wheat patatin/PNPLA-family lipid acyl
hydrolase. All evidence is homology/profile-based (IEA). No experimental data, no
gene-specific publications. Reviews should stay conservative: the two IEA annotations are
biologically reasonable; the MF (hydrolase activity) is under-specific and is the main
candidate for refinement.

2026-10-20 — Full re-review, independent of the ProtNLM2 predictions

ProtNLM2 output was not opened or used. New analysis: A0A3B6GK97-bioinformatics/RESULTS-rereview.md
(scripts gene_model_rereview.py, sf103_motif_scan_rereview.py; outputs in results_rereview/).
Earlier files in the bioinformatics folder were left unchanged.

Identity / gene model

Phylogeny (PANTHER 19 tree API, PTHR32176)

Annotation decisions (changed from previous review)

Family review / modules

Next relative