AIGR Deep Research — S. pombe alo1 (Q9HDX8): L-gulonolactone oxidase prediction

Hypothesis slug: prediction-gulonolactone-oxidase
Focus type: computational_prediction
Term under test: GO:0050105 L-gulonolactone oxidase activity (+ GO:0019853 L-ascorbic acid biosynthetic process)
Model source of prediction: BioReason-Pro SFT (ref doi:10.64898/2026.03.19.712954)

Executive Judgment

Verdict: REFUTED (over-annotated — paralog/substrate misassignment).

The BioReason-Pro predictions of L-gulonolactone oxidase activity (GO:0050105) and
L-ascorbic acid biosynthetic process (GO:0019853) for S. pombe alo1 (Q9HDX8) are
not supported and should not enter the review as-is. Multiple independent lines of
evidence identify alo1 as D-arabinono-1,4-lactone oxidase (EC 1.1.3.37), the terminal
enzyme of the fungal D-erythroascorbate pathway — a distinct-substrate, distinct-product
paralog within the same ALO/GULO/GLDH aldonolactone-oxidoreductase (FAD, vanillyl-alcohol-oxidase)
superfamily:

• UniProtKB Q9HDX8 curated recommended name = D-arabinono-1,4-lactone oxidase, EC 1.1.3.37;
  curated reaction produces dehydro-D-arabinono-1,4-lactone + H₂O₂, and the assigned pathway
  is D-erythroascorbate biosynthesis (step 2/2) — not L-ascorbate.
• PomBase/GO_Central assigns MF GO:0003885 (D-arabinono-1,4-lactone oxidase activity, IBA);
  GO:0050105 is absent.
• The mislabel is explainable: the enzyme sits in PANTHER family PTHR43762, whose family-level
  name is "L-gulonolactone oxidase," while alo1 maps to subfamily SF1 = D-arabinono-1,4-lactone oxidase.
  A model keying on the family name (frequency/name bias) will over-call GULO.
• Biochemistry of orthologs confirms the fungal substrate/product: S. cerevisiae ALO1 was purified
  as D-arabinono-1,4-lactone oxidase making D-erythroascorbate and shares only 32 % identity with rat GULO
  (PMID:10094636); C. albicans ALO1 null mutants lose the activity and D-erythroascorbate (PMID:11349062, 18282465).

Most important caveat: Members of this family have measurable in vitro promiscuity toward related
aldonolactones (the Q9HDX8 alt-name "L-galactono-γ-lactone oxidase" reflects this), so a weak in vitro
turnover of L-gulono-1,4-lactone cannot be excluded. That would still not justify a GULO/L-ascorbate
annotation, because the physiological substrate and product are D-arabinono-lactone → D-erythroascorbate.

Evidence Matrix

(Provenance: provenance/evidence_matrix.csv, provenance/go_decision_table.csv, provenance/sequence_identity_matrix.csv; UniProt fetch + alignment executed via MCP execute_code.)

Computed identity note (this run): A Gotoh global alignment (BLOSUM62) shows alo1 is essentially
equidistant from the fungal ALO (S. cerevisiae ALO1, 30.5 %) and the mammalian GULO (rat, 31.0 %) —
both in the ~30 % alignment "twilight zone". Global % identity therefore does not discriminate the two
functions and marginally favors GULO, which is exactly why a homology/embedding-based model can
over-predict L-gulonolactone oxidase. Discrimination instead rests on the PANTHER subfamily assignment
(PTHR43762:SF1 = D-arabinono-lactone oxidase), the curated EC 1.1.3.37 / D-erythroascorbate pathway, and
fungal-clade biology — all of which point to ALO, not GULO. This nuance strengthens, rather than weakens,
the refuted verdict.

GO Curation Implications (leads — require curator verification)

Verified against the live QuickGO annotation set for Q9HDX8 (13 annotations). Both model-predicted
terms are absent from the curated set, and the curated terms are already correct and more specific:

Core function is fully captured by the curated MF (GO:0003885) plus the specific product BP (GO:0070485);
"protein binding" is neither needed nor recommended. The model prediction is strictly less accurate than
existing curation — a textbook paralog/substrate over-annotation.

Mechanistic Scope

Direct molecular function tested: a mitochondrial-membrane, covalent-FAD flavoprotein oxidase that
oxidizes an aldono-1,4-lactone using O₂, releasing H₂O₂. The specific, physiological reaction for
alo1 is D-arabinono-1,4-lactone → dehydro-D-arabinono-1,4-lactone, the last step yielding
D-erythroascorbate (a 5-carbon L-ascorbate analog). Downstream/pleiotropic effects (oxidative-stress
resistance, virulence, alternative-oxidase induction seen in Candida) are pathway consequences of the
D-erythroascorbate product, not the enzyme's direct activity, and must not be conflated with the MF term.
The seed's GULO/L-ascorbate claim describes a different substrate (L-gulono-1,4-lactone) and different product
(L-ascorbate) belonging to the mammalian pathway.

Conflicts and Alternatives

• Paralog confusion (primary conflict): GULO (mammals), GLDH/L-galactono-lactone dehydrogenase (plants),
  and ALO (fungi) are homologous but substrate-distinct. ~32 % identity between fungal ALO1 and rat GULO
  (PMID:10094636) is exactly the zone where automated methods over-transfer the better-known family label.
• Family-name bias: PANTHER PTHR43762 carries the name "L-gulonolactone oxidase" at family level; the
  correct assignment is subfamily SF1 (D-arabinono-lactone oxidase).
• In-vitro promiscuity: The UniProt alt-name "L-galactono-γ-lactone oxidase" and known broad aldonolactone
  activity mean a low-level in vitro L-gulonolactone turnover is possible; this is an artifact-level caveat,
  not physiological evidence for L-ascorbate biosynthesis in S. pombe.

Knowledge Gaps

1. No S. pombe-specific enzymatic assay was located. Checked PubMed (no S. pombe ALO papers returned);
   the S. pombe assignment rests on curated by-similarity + IBA. Matters because direct S. pombe kinetics
   would definitively fix substrate preference. Resolve with: in-vitro substrate-panel kinetics on recombinant Q9HDX8.
2. Exact GO BP ID for D-erythroascorbate biosynthesis not confirmed here — curator should map the UniProt
   pathway string to the current ontology term.
3. Whether S. pombe actually accumulates D-erythroascorbate (vs. only enzyme presence) — resolve by
   metabolite LC-MS of wild-type vs. alo1Δ.

Discriminating Tests

• Substrate-panel oxidase assay on recombinant Q9HDX8: compare V/Km for D-arabinono-1,4-lactone,
  L-gulono-1,4-lactone, and L-galactono-1,4-lactone (O₂ consumption / H₂O₂ release). Physiological substrate
  should dominate.
• Metabolomics of alo1Δ S. pombe: loss of D-erythroascorbate (not L-ascorbate) confirms pathway identity.
• Phylogenetic placement / reciprocal-best-hit of Q9HDX8 vs. GULO (rat), GLDH (plant), ALO1 (Sc/Ca):
  expect clustering with fungal ALO subfamily, away from GULO.

Curation Leads (require curator verification)

• Action change: reject BioReason-Pro GO:0050105 and GO:0019853 for alo1; keep GO:0003885; consider adding
  the D-erythroascorbate biosynthetic-process BP term.
• Candidate references / snippets to verify:
• PMID:11349062 — "ALO1, the gene encoding D-arabinono-1,4-lactone oxidase, which catalyzes the final step of D-erythroascorbic acid biosynthesis."
• PMID:10094636 — "shared 32% … identity with … L-gulono-1,4-lactone oxidase from rat," establishing paralogy, not identity.
• UniProtKB Q9HDX8 — EC 1.1.3.37; pathway "D-erythroascorbate biosynthesis; step 2/2."
• Suggested question for curator: Is a fungal-ascorbate-analog BP term already applied? If not, add and
  remove the L-ascorbate BP.
• Suggested experiment: recombinant-enzyme substrate panel + alo1Δ metabolomics (above).

Limitations

Evidence is predominantly curated-database and ortholog-biochemistry; no S. pombe-specific assay was found in
the literature search. Network access limited the run to UniProt REST + PubMed; a direct sequence/phylogeny
computation was not executed. The conclusion (paralog misassignment) is nonetheless robust because it is
concordant across UniProt curation, GO_Central phylogenetics, PANTHER subfamily mapping, and ortholog
biochemistry.

Artifacts

• OpenScientist final report
• OpenScientist final report
• OpenScientist evidence matrix
• OpenScientist go decision table
• OpenScientist sequence identity matrix