CAO-1 substrate specificity: structural analysis of the co-crystal H-bond network

Objective

Test whether the CAO-1 (Q7S860) co-crystal structures explain the empirical substrate
specificity reported by Díaz-Sánchez et al. 2013 (PMID:23893079): CAO-1 cleaves resveratrol and
piceatannol but not trans-stilbene, 4-monohydroxystilbene, pinosylvin (3,5-diOH), trismethoxy-
resveratrol, or a 4′-methoxy-stilbene glucoside — the authors concluding a requirement for "a minimal
number of unmodified hydroxyl groups."

Method

Analysis of existing co-crystal structures (not de-novo docking):
- 5U90 — Co-CAO1 · resveratrol (PDB ligand STL)
- 5U97 — Co-CAO1 · piceatannol (PDB ligand PIT)

For each ligand oxygen we enumerate protein polar atoms (N/O), water, and metal within
H-bonding distance (≤3.5 Å), and measure the scissile interphenyl alkene relative to the active-site
metal. Reproducible: uv run python analyze_specificity.py → hbond_contacts.tsv.

Note: these are cobalt-substituted structures (catalytically inert Co(II) replaces the native
Fe(II) so a Michaelis-like substrate complex can be trapped for crystallography); the substrate
poses are taken as representative of the productive binding mode.

Result: a two-ring-anchor recognition model

Ligand O···O distances confirm the ring assignment (O2–O3 = 4.8 Å, meta on the resorcinol ring;
O1 is ~11 Å away on the opposite ring; O1–OAD = 2.8 Å, adjacent on that ring):

The shared 3,5,4′ hydroxyls contact identical residues in both structures; piceatannol's extra 3′-OH
adds a Thr151 H-bond (independently matching the Sui et al. 2017 text). The scissile alkene (C7=C8)
sits ~4.6 Å from the metal — reproducing the "substrates bind ~4.7 Å from the iron" statement in
PMID:28493664. The recovered binding residues (positions 133, 164, 383) match the UniProt
BINDING annotations exactly, validating the pipeline.

Interpretation: productive binding uses two anchors on opposite rings — a 4′-OH → Tyr133/Lys164
(ring A) and a 3/5-OH → Glu383 (ring B) — which together clamp the substrate with its interphenyl
double bond positioned over the metal/O2 site.

This explains the empirical SAR

The two-anchor model accounts for all seven panel compounds. Crucially it resolves the puzzle the
gene-review text flagged: a free 4′-OH is necessary but not sufficient — 4-hydroxystilbene has the
4′-OH (ring A anchor) yet is not cleaved because it lacks a ring-B (Glu383) anchor. So the requirement
is not "a 4′-OH" per se but a hydroxyl on each ring engaging the two anchors, which is why ≥2 well-
placed free hydroxyls are needed and fully-blocked stilbenes fail.

Status of the claim (important)

This is a retrospective structural rationalization, not a validated predictor:
- Only the two substrates were co-crystallized; the five non-substrates were not — their loss
of anchors is inferred from which hydroxyls are absent/blocked, not observed.
- No docking or binding-energy calculation was performed; this is H-bond geometry only.
- Co(II)-substituted structures; single active-site copy (chain A) analyzed.

So the honest status advances from "empirical + structure-consistent" to "structure-explained" —
the co-crystals contain a coherent two-anchor mechanism that reproduces the entire SAR — but a
predictive test would require docking/assaying the non-substrates (and diagnostic new probes below).

Falsifiable predictions

• 4,4′-dihydroxystilbene and 3,4′-dihydroxystilbene (one OH per ring, both anchors satisfiable)
  should be cleaved, unlike the single-ring pinosylvin/4-hydroxystilbene.
• Tyr133Phe / Lys164 / Glu383 mutants should selectively abolish cleavage of substrates relying on
  the corresponding anchor.

Provenance

• Structures: RCSB 5U90, 5U97 (Sui et al. 2017, PMID:28493664).
• Empirical SAR: Díaz-Sánchez et al. 2013 (PMID:23893079); non-substrate list per UniProt Q7S860.
• Script + raw contacts: analyze_specificity.py, hbond_contacts.tsv (this folder).