AIGR Deep Research Report — ire-1 (Q09499, C. elegans)

Hypothesis: hydrolase activity, acting on ester bonds (GO:0016788)

Focus type: computational_prediction (GO-GPT / BioReason-Pro)
Term under test: GO:0016788 — hydrolase activity, acting on ester bonds
Reference context: doi:10.64898/2026.03.19.712954

Executive Judgment

Verdict: PARTIALLY SUPPORTED — but over-general and redundant; do NOT add GO:0016788 as an annotation.

ire-1 unambiguously carries a catalytic activity on nucleic-acid ester (phosphodiester)
bonds: its C-terminal KEN/RNase domain is a site-specific endoribonuclease that cleaves
xbp-1 mRNA to drive non-conventional UPR splicing. In that loose biochemical sense the
prediction "acts on ester bonds" is not wrong. However, the prediction is the wrong level
of specificity for annotation, for three concrete reasons:

1. A precise, experimentally-supported term already exists. UniProt Q09499 already carries
   GO:0004521 (RNA endonuclease activity, IMP:WormBase) and GO:0016787 (hydrolase, IEA-KW).
   GO:0016788 adds no information beyond these.
2. In the current GO ontology, GO:0016788 is NOT an ancestor of GO:0004521. QuickGO shows the
   RNA-endonuclease branch descends through catalytic activity, acting on a nucleic acid
   (GO:0140640) → nuclease (GO:0004518) → endonuclease (GO:0004519) → RNA endonuclease (GO:0004521),
   not through hydrolase, acting on ester bonds. So the seed's premise that GO:0016788 is a
   "correct high-level parent of GO:0004521" does not hold in the current ontology.
3. Mechanistic mismatch — the closest analog is a LYASE, not a hydrolase. IRE1-family RNases use a
   metal-independent transesterification (general acid/base His–Tyr pair) yielding a
   2′,3′-cyclic phosphate + 5′-OH, not a hydrolysis to a phosphomonoester. GO's own treatment of the
   enzyme class sharing this exact chemistry — tRNA-intron lyase activity (GO:0000213) — places it
   under lyase / phosphorus-oxygen lyase (GO:0016829 / GO:0016849) and under RNA endonuclease
   (GO:0004521), but explicitly NOT under GO:0016788. So GO:0016788 is not just off-branch; it is the
   mechanistically wrong high-level MF for a cyclic-phosphate-producing transesterase. (It also fails
   GO:0016788's own hydrolytic nuclease children GO:0016891/2, which require phosphomonoester products.)

Bottom line for the curator: the true, well-supported molecular function is the specific
IRE1-type endoribonuclease activity (GO:0004521 / more specifically an IRE1/tRNA-ligase-type
transesterifying endoribonuclease). GO:0016788 is a correct-but-imprecise generalization that is
subsumed and redundant and should not be added.

Evidence Matrix

GO Curation Implications

Molecular Function (MF). The relevant, supported MF term is GO:0004521 (RNA endonuclease
activity) — already annotated to Q09499 with IMP:WormBase evidence, plus the kinase term
GO:0004674. A still-more-precise child (e.g. an IRE1/tRNA-ligase-type transesterifying
endoribonuclease term) would be ideal if available.

Recommended action for the predicted term GO:0016788 (lead, requires curator verification):
- Do not add GO:0016788. It is redundant with the already-present GO:0016787 (hydrolase) and,
more importantly, is superseded by the specific GO:0004521 that captures the actual activity.
- Treat the GO-GPT prediction as "correct in spirit, too general, and structurally off-branch."
It is not a false positive at the chemistry level, but it fails the annotate-to-most-specific-term
principle and does not sit on the lineage of the experimentally supported term.
- If anything, the existing IEA GO:0016787 keyword annotation could itself be viewed as low-value
given GO:0004521 is present, but that is a separate housekeeping matter.

This is not a "protein binding" fallback; a specific, informative MF (endoribonuclease) is
well supported.

Mechanistic Scope

• Direct gene-product activity (what the term should describe): two catalytic activities on the
  cytoplasmic face — (i) a Ser/Thr protein kinase (autophosphorylation; a phosphotransferase,
  not a hydrolase) and (ii) a site-specific endoribonuclease (KEN domain) that cleaves two
  stem-loops in xbp-1 mRNA, excising an intron. The endoribonuclease is the only activity relevant
  to GO:0016788.
• Chemistry: cleavage of an RNA phosphodiester bond (an ester bond) via general acid/base
  catalysis, producing a 2′,3′-cyclic phosphate and 5′-OH (transesterification), later re-ligated by
  RtcB-type ligase to yield spliced xbp-1.
• Downstream (NOT the molecular function): production of active XBP-1 transcription factor;
  transcriptional UPR; ER homeostasis; larval development; longevity in insulin/IGF-1 mutants;
  ER-stress apoptosis. These are BP/phenotype consequences, not the MF being tested.

Conflicts and Alternatives

• Ontology-structure conflict (primary): The seed frames GO:0016788 as a valid parent of
  GO:0004521. Current GO places nuclease/endonuclease activity under GO:0140640 (catalytic activity,
  acting on a nucleic acid), not under GO:0016788. Thus GO:0016788 and GO:0004521 are in
  parallel branches, not parent–child. This weakens the "correct-but-less-precise parent" framing.
• Mechanism conflict (hydrolase vs lyase): GO:0016788 still owns hydrolytic nuclease children
  (GO:0016891/2) that specify phosphomonoester products; IRE1 produces cyclic-phosphate ends
  (transesterase/lyase-like, cf. EC reclassification of cyclizing RNases to EC 4.6.1). Decisively,
  GO classifies IRE1's mechanistic twin — tRNA-intron lyase (GO:0000213) — under lyase
  (GO:0016829)/phosphorus-oxygen lyase (GO:0016849) and RNA endonuclease (GO:0004521), not under
  GO:0016788. So IRE1 fits neither GO:0016891/2 nor the strict "hydrolase" reading; if a high-level
  chemistry parent beyond GO:0004521 were ever wanted, the lyase branch would be the better-motivated
  choice — further undercutting the GO:0016788 prediction.
• Orthology/paralog caveat: Direct biochemical cleavage assays (Calfon; Korennykh) were done on
  mouse/yeast IRE1; worm ire-1 is supported genetically (Shen) and by conserved domain architecture.
  No paralog-overannotation risk (IRE1 is the single ER UPR IRE-branch sensor); low frequency-bias risk.
• Not an in-vitro-only artifact: function is corroborated in vivo in C. elegans.

Knowledge Gaps

1. Worm-specific direct cleavage assay. Checked: worm evidence is genetic (IMP) + orthology;
   direct in-vitro cleavage was shown for mouse IRE1. Matters because MF assignment ideally rests on
   direct activity. Resolve with a purified C. elegans IRE-1 RNase cleavage assay on xbp-1 stem-loops.
2. Exact GO term granularity. Checked: a QuickGO search found no dedicated "IRE1 endoribonuclease"
   MF term; GO:0004521 (RNA endonuclease activity) is the most specific applicable term and is already
   annotated (IMP:WormBase). A mechanism-precise child (analogous to the tRNA-intron lyase term
   GO:0000213) does not exist for IRE1. Matters only if maximal precision is desired; could be a new-term
   request but is not required for correct curation.
3. Whether curators intend GO:0016788 as a mechanism-agnostic descriptor. If a curation policy
   treats phosphodiester cleavage as "acting on ester bonds," the term becomes defensible-but-still-
   redundant. Resolve via GO editorial guidance.

Discriminating Tests

• Ontology audit: confirm in the review's GO release whether GO:0016788 is an ancestor of
  GO:0004521 (it is not in current GO) — decisive for the "correct parent" claim.
• Point-mutant RNase-dead worm (KEN-domain catalytic His/Tyr equivalent) that abolishes xbp-1
  splicing while retaining kinase activity — isolates the endoribonuclease MF in vivo.
• In-vitro cleavage of a C. elegans xbp-1 stem-loop by recombinant IRE-1 cytoplasmic module,
  with 2′,3′-cyclic-phosphate product characterization — confirms mechanism class.

Curation Leads (require curator verification)

• Lead 1 — Reject/de-prioritize GO:0016788 as an added annotation. Rationale: redundant with
  existing GO:0016787 + GO:0004521; not on the lineage of the specific term in current GO;
  mechanism mismatch with its hydrolytic nuclease children.
• Lead 2 — Retain/strengthen GO:0004521 (RNA endonuclease activity) as the core MF, upgrading
  evidence with orthologous direct-assay references (PMID 11780124) if the review model permits
  ISS-with-mouse support alongside worm IMP.
• Candidate references to verify:
• PMID 11780124 — snippet: "Purified mouse IRE1 accurately cleaved XBP-1 mRNA in vitro… a direct target of IRE1 endonucleolytic activity."
• PMID 11779465 — snippet: "C. elegans requires ire-1-mediated splicing of xbp-1 mRNA for UPR gene transcription and survival upon ER stress."
• PMID 21729333 — snippet: Ire1 RNase uses His1061/Tyr1043 general acid/base and coordinates the scissile phosphate.
• PMID 36389122 — snippet: "IRE-1 endoribonuclease activity declines early in" (worm).
• Suggested curator question: "Does our GO release place GO:0016788 as an ancestor of GO:0004521?
  If not, GO:0016788 should not be represented as the specific term's parent."
• Suggested experiment: RNase-dead separation-of-function ire-1 allele + in-vitro xbp-1
  stem-loop cleavage to nail the worm-specific direct MF and mechanism class.

Provenance

Computed checks run (executed code + outputs retained in the job log):
UniProt Q09499 REST fetch (domains + 16 GO annotations); QuickGO ancestry queries for GO:0004521,
GO:0004518, GO:0004540, and GO:0000213; children of GO:0016788; QuickGO term search for
IRE1/endoribonuclease/tRNA-splicing terms (no dedicated IRE1 MF term exists). Literature via PubMed
(PMIDs above).
Where worm-specific direct biochemistry was unavailable, it is stated as such and not fabricated.

Artifact files (computed provenance):
- go_decision_table.csv — per-term GO curation decision table (predicted vs annotated terms, recommended actions, rationale).
- ontology_relationship_tests.csv — computed QuickGO ancestry tests (GO:0016788 is not an ancestor of GO:0004521; the mechanistic analog GO:0000213 is a lyase, not under GO:0016788).

Artifacts

• OpenScientist final report
• OpenScientist final report
• OpenScientist go decision table
• OpenScientist ontology relationship tests