Bioinformatics analysis: A0A3B6GK97 (wheat patatin/PNPLA protein)

Question. (1) Which plant patatin-related phospholipase A (pPLA) subfamily does
this protein belong to (to refine the functional inference beyond the domain-level
"lipid metabolic process"), and (2) is its catalytic machinery intact (i.e., is it a
genuine acyl hydrolase, a pseudoenzyme, or a truncated model)?

TL;DR.
- By sequence identity and tree placement over the region it retains, A0A3B6GK97 belongs
to the pPLAII subfamily (42–50% identity to pPLAII / rice pPLAs; ~23% to pPLAIII;
~16% to pPLAI). It is not a pPLAIII (galactolipase/growth) or pPLAI (iPLA2-like) protein.
- However, the modeled 302-aa sequence lacks the entire N-terminal half of the patatin
catalytic domain — both the oxyanion glycine-rich block (DGGG) and the catalytic
serine nucleophile elbow (G-T-S-T-G) are absent. It retains the C-terminal portion
including the catalytic Asp. As deposited, the protein is therefore predicted to be
catalytically inactive (no nucleophilic serine).
- Most parsimonious explanation: an incomplete/incorrect gene model (TraesCS3D02G033600
is ~100–130 aa shorter than full-length orthologs, missing exactly the N-terminal catalytic
exon region). A genuine degenerate pseudo-enzyme cannot be excluded from sequence alone.

This is a caveat to the lipase interpretation: the GO_Central IBA annotations
(glycerophospholipase GO:0004620, monoacylglycerol lipase GO:0047372) are phylogenetic
propagations that assume an intact active site and do not verify it; the modeled
sequence cannot support those activities as-is.

Methods

All sequences fetched live from the UniProt REST API; no sequences or results are
hardcoded. Reproduce with just all (recipes: fetch, analyze, test-control).

• Query: A0A3B6GK97 (wheat, TraesCS3D02G033600), 302 aa.
• Reference set (13): characterized plant pPLAs spanning all three subfamilies plus a
  structural anchor — Arabidopsis pPLAI (F4HX15), pPLAIIα–ε (O48723, O23181, O23179, O23180,
  Q9FIY1), pPLAIIIα–δ (O80959, Q9SV43, Q8H133, Q93ZQ3 — the last annotated "probable
  inactive", used as a pseudoenzyme control), rice PLP1/PLP2 (Q84QY3, Q6ZJD3), and potato
  patatin (P15478; the Ser-Asp dyad family of PMID:12779324). Provenance and the UniProt
  queries used are in data/reference_accessions.tsv.
• Tools: Python 3.11; pyfamsa 0.7.0 (FAMSA MSA, UPGMA guide tree); BioPython 1.87
  (identity distance matrix, neighbour-joining tree). Catalytic motifs detected two ways:
  (a) MSA-independent regex scan of raw sequences for the nucleophile G-x-S-x-G and
  oxyanion DGGG; (b) MSA column conservation among the core active single-domain enzymes
  (pPLAI and the inactive PLP9 excluded from the catalytic-consensus set only; they are still
  in the placement analysis).
• Outputs: results/alignment.fasta, pairwise_identity_to_query.tsv, nj_tree.newick,
  motif_scan.tsv, catalytic_site.tsv, catalytic_columns_all_seqs.tsv, summary.json.

Results

1. Subfamily placement → pPLAII

results/pairwise_identity_to_query.tsv (%identity to query, over ~98% query coverage):

The whole pPLAII subfamily (+ rice pPLAs) ranks at 42–50%, with a clear gap to pPLAIII
(~23%) and pPLAI (16%). The NJ tree (nj_tree.newick) places the query within the
pPLAII/rice/patatin clade, separate from pPLAIII and the long-branch pPLAI. Nearest
neighbours by patristic distance: rice PLP2, At pPLAIIα, At pPLAIIγ, At pPLAIIδ.
Conclusion: pPLAII-type (the "classic" lipid acyl hydrolase clade — defense/wounding/
stress-associated in Arabidopsis), over the portion of the domain the model retains.

2. Catalytic machinery is missing from the modeled sequence

MSA-independent motif scan (results/motif_scan.tsv):

Every active reference carries the canonical G-T-S-T-G nucleophile elbow; the query has
no G-x-S-x-G anywhere in 302 aa and no DGGG oxyanion block. The MSA
(results/catalytic_columns_all_seqs.tsv, and the alignment around cols 540–600) shows the
query fully gapped through the N-terminal catalytic core — its modeled N-terminus begins
downstream of where the catalytic Ser sits. The query does retain the catalytic-Asp
region (Asp121, in NLIDSG), matching the active references.

Interpretation: the model has the C-terminal ~⅔ of the patatin domain (incl. catalytic Asp)
but is missing the N-terminal ~⅓ bearing the oxyanion and the catalytic serine. Without the
nucleophilic Ser, acyl-ester hydrolysis is mechanistically impossible — predicted inactive
as modeled.

3. Control (pipeline is input-driven, not hardcoded)

Re-running with potato patatin relabelled as the query (just test-control,
results_control/) correctly recovers intact motifs (GTSTG pos 77, DGGG pos 35) and a
pPLAII/patatin placement — confirming the query's "motifs absent" result is a property of the
A0A3B6GK97 sequence, not an artifact.

Interpretation & caveats

• Subfamily = pPLAII is well supported (large identity gap to other subfamilies; consistent
  tree placement). pPLAII proteins are soluble lipid acyl hydrolases (sn-1/sn-2 acyl ester
  hydrolysis of glycerolipids), associated in Arabidopsis with defense, wounding, phosphate
  starvation and abiotic stress — not membrane trafficking. This (plus the single-domain
  architecture) makes a trafficking role very unlikely for this protein.
• The deposited protein lacks the catalytic serine and oxyanion block → predicted
  catalytically inactive as modeled.
• Two non-exclusive explanations, which sequence alone cannot distinguish:
• Incomplete/incorrect gene model (most parsimonious): 302 aa vs ~400–432 aa for
  orthologs; the missing segment is a clean N-terminal block consistent with a missing
  5′ exon — a common error in automated hexaploid-wheat annotation. Resolving this
  requires checking the genomic locus, the 3B/3A/3D homoeologs, and alternative gene
  models / RNA-seq — not done here.
• A genuine degenerate pseudo-patatin that has lost the nucleophile.
• Consequence for GO: the IBA lipase MF annotations were propagated phylogenetically and
  assume an intact active site; they are not verifiable on the modeled sequence. The MF
  inference (lipase / carboxylic ester hydrolase) should carry this caveat, and the gene model
  itself is a candidate for revisiting.

Reproducibility checklist

• [x] Scripts use no hardcoded inputs or outputs — all sequences fetched live from UniProt
  by accession; all numbers computed from the data at runtime. Conclusions are only in this
  RESULTS.md, not in code.
• [x] Pipeline tested on another input — (a) it runs over 13 references simultaneously and
  correctly recovers GTSTG/DGGG in all active enzymes while flagging absence in the
  annotated-inactive PLP9; (b) an explicit control (test-control) with active patatin as
  the query recovers intact motifs + correct placement.
• [x] Analyses completed as expected; full pipeline re-runs cleanly from scratch
  (just all).
• [x] Direct script outputs are in results/ and results_control/.
• [x] Summary includes detailed provenance and justification, and is explicit about
  uncertainty (truncated model vs true pseudoenzyme is not resolved).
• [?] Not resolved: whether the missing N-terminus is an annotation artifact or biological
  — would need genomic/homoeolog/transcript evidence.

Provenance

• Reference accessions + UniProt query rationale: data/reference_accessions.tsv.
• pPLA subfamily framework: Scherer et al. 2010 (PMID:20961799); review PMID:39645102.
• Patatin Ser-Asp catalytic dyad / GTSTG nucleophile: PMID:12779324.
• Tools: UniProt REST (rest.uniprot.org), pyfamsa 0.7.0, BioPython 1.87.