Stilbene Cleavage Oxygenases (SCO / LSD family)

Bottom line: stilbene cleavage oxygenases are non-heme iron enzymes that
split resveratrol-type stilbenes into two aromatic aldehydes; they share a fold
and a PANTHER family (PTHR10543) with the carotenoid cleavage oxygenases. We
reviewed seven members: six stilbene cleavers from fungi (Neurospora cao-1,
Ustilago Rco1) and sphingomonad bacteria (NOV1, NOV2, LSD-I, LSD-III), plus the
Neurospora carotenoid cleaver cao-2 as a control. We did this because
family-level propagation assigns carotenoid terms to the whole clade. In all
six stilbene cleavers, GO:0010436 carotenoid dioxygenase activity (IBA or
TreeGrafter IEA) was marked REMOVE, and GO:0016121 carotene catabolic process
was either modified to GO:0046272 stilbene catabolic process (four genes) or
removed (the two LSD isozymes); the same terms were accepted on cao-2. The
project also tracks GO's July 2026 revision of this area (new GO:7770086
resveratrol dioxygenase activity) and proposes a hydroxystilbene α,β-dioxygenase
activity grouping term. All seven reviews are still status IN_PROGRESS, and
the grouping term exists so far only as a proposed new term in the cao-1 review.

Overview

Stilbene cleavage oxygenases (SCOs), also called lignostilbene α,β-dioxygenases (LSDs,
EC 1.13.11.43), are non-heme iron enzymes that oxidatively cleave the interphenyl Cα–Cβ double
bond of stilbenes, converting one stilbene into two aromatic aldehydes. They are one branch of the
larger carotenoid cleavage oxygenase (CCO) family: they share the seven-bladed β-propeller fold
and the four-His mononuclear Fe(II) center of carotenoid cleavers, but have a substrate-binding cleft
adapted to the smaller, hydroxylated stilbene scaffold instead of a long polyene.

This project scopes the SCO/LSD class as a whole — fungal and bacterial members, the chemistry and
substrate recognition, and the Gene Ontology representation of the activity — building on the
Neurospora cao-1 review that opened the thread.

The reaction and the family

• Chemistry: a stilbene + O₂ → two phenolic/aromatic aldehydes (dioxygenase; both oxygen atoms
  incorporated). E.g. resveratrol → 3,5-dihydroxybenzaldehyde + 4-hydroxybenzaldehyde;
  lignostilbene → 2 vanillin.
• Founding member: lignostilbene α,β-dioxygenase (LSD) of Sphingomonas paucimobilis, the first
  enzyme shown to cleave the central stilbene double bond, acting in lignin-derived stilbene
  catabolism.
• CCO-family context: the carotenoid-cleaving sister enzymes (e.g. Neurospora CAO-2, a
  torulene dioxygenase in the neurosporaxanthin pathway) share the fold but act on carotenes — a
  substrate-class split that is the source of a recurring IBA over-annotation (see below).

Members reviewed / in scope

Substrate recognition: a two-ring-anchor model

A structure-based analysis of the CAO-1 co-crystals (5U90 resveratrol, 5U97 piceatannol) —
cao-1 bioinformatics RESULTS — recovers a
two-anchor recognition model: a 4′-hydroxyl → Tyr133/Lys164 anchor on one ring and a
3/5-hydroxyl → Glu383 anchor on the other, clamping the substrate with its scissile alkene ~4.6 Å
over the metal. This retrospectively explains CAO-1's entire empirical substrate panel (PMID:23893079),
including why a free 4′-OH is necessary but not sufficient (4-hydroxystilbene has it but lacks the
second anchor). Per-member specificity varies across the family: CAO-1 requires several free hydroxyls,
whereas bacterial LsdA reportedly cleaves even 4-hydroxystilbene — a difference a follow-up
bacterial/fungal anchor-residue comparison should examine structurally.

Gene Ontology representation (a live issue)

The GO representation of this activity was actively revised in July 2026, overlapping exactly with
this work:

• GO:1905594 "resveratrol binding" is being obsoleted (go-ontology #32321/#32333) as ill-defined —
  its only two experimental annotations were CAO-1 (Q7S860, where resveratrol is a substrate) and
  NQO2 (P16083, where resveratrol is an inhibitor), which the single term conflated.
• GO:7770086 "resveratrol dioxygenase activity" was added (go-ontology
  #32332, merged 2026-07-17; RHEA:73735;
  parent GO:0016702), explicitly avoiding the carotenoid-dioxygenase branch. The term is live and
  resolvable —
  QuickGO returns it as
  non-obsolete with definition "trans-resveratrol + O2 = 3,5-dihydroxybenzaldehyde +
  4-hydroxybenzaldehyde" and xrefs RHEA:73735 / PMID:28493664. It is not yet in the oaklib
  ontology snapshot this repo validates against, so structured molecular_function slots in the
  gene reviews still carry the validated parent GO:0016702 and name GO:7770086 in prose; they
  should be switched once the snapshot catches up.
• This project proposes a grouping term — "hydroxystilbene α,β-dioxygenase activity", with
  "stilbene α,β-dioxygenase" as the literature-conventional synonym — as the parent of
  GO:7770086 and GO:0050054 (lignostilbene α,β-dioxygenase), to organize the currently-flat set of
  stilbene-cleavage leaf terms and to serve family-level (IBA) annotation. The hydroxystilbene scope is
  deliberate: these enzymes require free phenolic hydroxyl groups, so a term over all stilbenoids would
  overstate the demonstrated family specificity.

The IBA lesson

CAO-1 is a flagship example in the IBA Annotation Quality project (Pattern 11:
substrate over-propagation from a multi-specificity family). Because SCOs and carotenoid cleavers share
the PANTHER PTHR10543 family, carotenoid-dioxygenase terms over-propagate to the stilbenoid-cleaving
members. The N. crassa paralog pair is the decisive positive control: the same family IBA terms
are wrong for cao-1 (a stilbene cleaver) and right for cao-2 (a genuine torulene/carotenoid
cleaver). Getting the family node right — a stilbene-cleavage subfamily annotated with the grouping
term above — would fix the whole clade at once.

The same error recurs across the whole clade and through multiple propagation routes. Every SCO
reviewed here carries the carotenoid-dioxygenase over-annotation:

• manual IBA (GO_REF:0000033): cao-1, Rco1 — both experimentally refuted stilbene cleavers;
• automated TreeGrafter IEA (GO_REF:0000118): NOV1, NOV2, LSD-I, LSD-III.

For the two LSD isozymes the contradiction is starkest — the automated carotenoid terms sit alongside
the genes' own experimental (IDA) lignostilbene-dioxygenase and lignin-catabolism annotations. So
phylogenetic inference and tree-grafting make the identical substrate-class mistake on this family,
because it mixes carotenoid and stilbene cleavers. (See also the TreeGrafter Inference Evaluation and
IBA_REVIEW projects.)

Paralog positive controls in two organisms. The carotenoid/stilbene split appears as a paralog pair
in both reviewed fungi: Neurospora CAO-2 (carotenoid/torulene) vs cao-1 (stilbene), and
Ustilago maydis Cco1 (β-carotene) vs Rco1 (stilbene). In each, the same family annotation is
correct for the carotenoid paralog and wrong for the stilbene paralog — only target-specific
experimental evidence separates them.

Open questions

• What is CAO-1's natural physiological stilbenoid substrate (plant-host vs microbial-competitor
  origin)? (See the cao-1 knowledge gaps.)
• Do the bacterial (NOV1/NOV2/LsdA) and fungal (CAO-1/Rco1) members differ structurally in the
  two-anchor architecture in a way that explains their differing hydroxylation requirements?
• Should GO adopt the proposed hydroxystilbene α,β-dioxygenase activity grouping as the parent of the
  reaction-specific leaves?

Source: ai4curation/ai-gene-review