AIGR Deep-Research Report — F6LAX4 (WHEAT, LOC123103357)

Focus: computational_prediction · slug prediction-animal-specific-terms Primary term in focus: GO:0043005 neuron projection (CC) Question: Are the ProtNLM2 animal-specific predictions for wheat F6LAX4 appropriate, or are they cross-kingdom / taxon-inappropriate misassignments? Independently determine whether F6LAX4 is the PP2A scaffold (A/PR65) subunit.

Executive Judgment

Verdict: REFUTED (over-annotated) for the animal-context terms; the underlying protein identity in the seed hypothesis is SUPPORTED.

F6LAX4 is, with high confidence, the wheat orthologue of the protein phosphatase 2A (PP2A) 65 kDa scaffold / structural "A" subunit (PR65 / PPP2R1). This is established independently by (i) UniProt naming it "Protein phosphatase 2A structural subunit", (ii) its ARM/HEAT-repeat α-α solenoid domain architecture (InterPro IPR051023 PP2A_Regulatory_Subunit_A, PANTHER PTHR10648:SF4, Pfam PF02985/PF22646), and (iii) computed orthology — 85–90 % global identity to the three Arabidopsis PP2A A subunits (RCN1/A1, A2, A3) and 56–59 % to human PR65 α/β, the exact conservation signature of this deeply conserved scaffold.

Given this identity, the ProtNLM2 predictions divide into two groups:

• Taxon-inappropriate (REFUTE): neuron projection (GO:0043005), neuronal cell body (GO:0043025) — plants have no neurons; protein antigen binding (GO:1990405) — plants have no adaptive immune system / antibodies. These are text-transfer artifacts from the extensive mammalian PP2A/PR65 neuronal literature. They must not be asserted on a wheat protein.
• Conserved but indirect / imprecise (QUALIFY, do not assert on this subunit): chromosome centromeric region (GO:0000775) and chromosome segregation (GO:0007059) describe a genuine, eukaryote-conserved Shugoshin–PP2A cohesion-protection function, but that role is mediated by the B56 regulatory subunit binding Shugoshin — not by the A scaffold intrinsically — so for the A subunit these are holoenzyme-context, non-core terms with no direct wheat evidence. Protein heterodimerization activity (GO:0046982) mis-describes the scaffold, which nucleates an A–B–C heterotrimer; the informative MF is protein-phosphatase regulator/PP2A-complex scaffolding, already captured by the existing IBA annotations.

Most important caveat: F6LAX4 is a TrEMBL (unreviewed) entry with no wheat-specific experimental characterization; the identity call rests on strong sequence/domain orthology (computational + database), and the "conserved-but-indirect" mitotic terms cannot be positively excluded as PP2A-complex functions — they are simply not supportable for the scaffold subunit on current evidence.

Evidence Matrix

Structural Confirmation (added Iteration 2)

AlphaFold DB models were retrieved for F6LAX4 (global pLDDT 90.4, an elongated, confidently modeled HEAT-repeat solenoid), Arabidopsis RCN1 (Q38845, pLDDT 93.0) and human PR65α (P30153, pLDDT 94.9). A Kabsch superposition of Cα atoms over identical-residue anchors gives an elevated global RMSD (F6LAX4↔RCN1 2.82 Å; F6LAX4↔human PR65 4.46 Å), but a sliding 60-residue local window RMSD is near-zero (median 0.29 Å vs RCN1, 0.92 Å vs human PR65). This pattern — individually superimposable repeats but hinge-bent ends — is the structural signature of the elastic PR65 α-α solenoid (20133745) and independently confirms the scaffold identity. Provenance: /tmp/aigr/AF-*.pdb, superposition code in the executed analysis log.

Family Disambiguation & Annotation-Status Check (added Iteration 3)

Two curator-relevant confirmations:

1. The ProtNLM2 terms are NOT curated GO annotations. UniProt's curated GO set for F6LAX4 is exactly five IBA/GO_Central terms — cytoplasm (GO:0005737), cytosol (GO:0005829), nucleus (GO:0005634), PP2A complex (GO:0000159), protein phosphatase regulator activity (GO:0019888). The intersection with the six ProtNLM2 predictions is empty. The correct curation action is therefore "reject / do not import the prediction," not "remove an existing annotation."
2. F6LAX4 is the non-catalytic A scaffold, not the catalytic C subunit. A motif scan finds none of the PPP-phosphatase catalytic metal-binding signatures (GDxHG, GDxVDRG, GNHE, RGNHE) in F6LAX4 (587 aa), whereas the genuine PP2A catalytic subunit human PPP2CA (P67775, 309 aa) contains all four. This excludes the main competing family assignment (catalytic subunit) and is consistent with the scaffold role.

GO Curation Implications (leads — require curator verification)

Net: the existing IBA annotations already capture the correct core function; the ProtNLM2 additions add nothing valid and three are actively wrong.

Mechanistic Scope

Immediate molecular function of F6LAX4: it is the structural (A) scaffold subunit of the PP2A holoenzyme — a rigid-yet-elastic α-α solenoid of ~15 HEAT repeats that simultaneously binds the catalytic (C) subunit and a variable regulatory (B/B56/B''/B''') subunit, thereby determining substrate specificity of a Ser/Thr phosphatase. It has no catalytic activity itself and no intrinsic DNA-, antigen-, or neuron-related activity. Downstream/holoenzyme-context phenomena (auxin-transport polarity, ethylene signaling, ER-stress/UPR in plants; Tau dephosphorylation, cohesion protection in animals) are properties of specific PP2A holoenzymes, not of the isolated scaffold, and must be separated from the scaffold's direct GO annotation.

Conflicts and Alternatives

• Paralog/ortholog confusion (the actual error source): ProtNLM2 transferred descriptors from mammalian PR65/PP2A entries whose partner holoenzymes act in neurons and immune cells. The wheat sequence is a bona fide ortholog (56–59 % to human), so a naive language model reuses mammalian tissue/organelle vocabulary.
• Subunit confusion: the centromere/segregation terms belong to PP2A-B56 + Shugoshin, i.e., the regulatory subunit, not the A scaffold. Assigning them to the A subunit conflates holoenzyme function with scaffold function.
• No competing identity hypothesis survives: 85–90 % identity to three independent Arabidopsis A subunits rules out any non-PP2A-A assignment. The alternative that this is a generic ARM/HEAT protein (e.g., importin, karyopherin) is excluded by the PP2A-A-specific InterPro/PANTHER family and by orthology.

Knowledge Gaps

1. No wheat experimental data. Checked: UniProt (TrEMBL, PE=3 "inferred from homology"), literature. Matters because all functional claims are homology-based. Resolved by: wheat PP2A pulldown/complex-MS, or a wheat rcn1-like mutant.
2. Which wheat PP2A A homoeolog/paralog is this? Hexaploid wheat has A/B/D homoeologs and multiple A-subunit genes; LOC123103357 is one copy. Checked: single UniProt entry only. Matters for gene-level annotation propagation. Resolved by: wheat genome synteny / homoeolog mapping.
3. Does wheat PP2A participate in Sgo-mediated cohesion (GO:0000775/0007059)? Checked: no wheat-specific paper found; conserved in animals/fungi and reported in plants via SGO1. Matters only if a curator considers those terms. Resolved by: wheat/cereal meiosis cohesion studies (SGO1–PP2A-B56).

Discriminating Tests

• Orthology/phylogeny (done, decisive): global identity to Arabidopsis A subunits (85–90 %) vs catalytic subunits (would be <25 %) confirms scaffold, not catalytic, identity.
• AlphaFold geometry: predict/inspect the AlphaFold model of F6LAX4 for the characteristic hooked HEAT-repeat solenoid and superpose on human PR65 (PDB 1B3U) — expected low RMSD over the solenoid.
• Holoenzyme co-IP MS in wheat: recovery of PP2A C and B/B56 subunits (not neuronal or immune proteins) would positively confirm scaffold function and simultaneously refute the animal-context terms.
• Motif check for B56/Sgo pocket: the Sgo-binding pocket is on B56, absent from the A subunit — confirming the centromere terms do not map to F6LAX4.

Curation Leads (require curator verification)

• Reject GO:0043005, GO:0043025, GO:1990405 on F6LAX4 as ProtNLM2 cross-kingdom errors (organism = plant; no neurons, no adaptive immunity).
• Do not propagate GO:0000775 / GO:0007059 to the A scaffold subunit; if any PP2A-complex mitotic annotation is desired, it belongs to the B56 subunit with a wheat/plant reference.
• Replace GO:0046982 with an informative scaffold MF (protein phosphatase 2A binding / structural constituent of PP2A holoenzyme) rather than generic heterodimerization.
• Retain existing IBA terms GO:0000159, GO:0019888, GO:0005737/0005829/0005634.
• Candidate references & exact snippets to verify:
• 20133745 — "PR65 is the two-layered (alpha-alpha solenoid) HEAT-repeat … scaffold of protein phosphatase PP2A."
• 38820156 — "PR65 is the HEAT repeat scaffold subunit of the heterotrimeric protein phosphatase 2A (PP2A)."
• 37290287 — RCN1 "a regulatory A1 subunit isoform of Arabidopsis PP2A" (plant ortholog exists; function class = growth/development/signaling).
• 33973335 — "a highly conserved pocket on the B56 regulatory subunit is required for hSgo1 binding and cohesion protection" (centromere terms map to B56, not A).
• Suggested question for curators: should AIGR add a taxon-constraint QC rule flagging animal-only CC/MF terms (neuron*, antibody/antigen) predicted on Viridiplantae proteins?
• Suggested experiment: wheat PP2A affinity-MS to positively identify holoenzyme partners and close the "no wheat data" gap.

Provenance / Artifacts

• Orthology identity matrix (global Needleman–Wunsch, computed this run): /tmp/aigr/orthology_identity.csv
• Per-term verdict table: /tmp/aigr/protnlm_term_verdicts.csv
• Sequences retrieved from UniProt REST: F6LAX4, P30153, P30154, Q38845, Q38950, Q38951.

Computational note: identities were computed with a simple match/mismatch (+1/−1, gap −2) global alignment rather than a BLOSUM-scored aligner; the plant-vs-plant (85–90 %) versus cross-kingdom (56–59 %) separation is large and robust to scoring scheme, so the ortholog assignment is not sensitive to alignment parameters.