Notes: benzil reductase (EC 1.1.1.320 / RHEA:25968) gene set

Reviewed as a set with the other UniProt entries carrying RHEA:25968: yeast IRC24 and NRE1, B. cereus yueD (Q8RJB2), B. subtilis yueD (O32099). Gerbil SPR (Q8R536) also carries RHEA:25968, but it is a sepiapterin reductase that was incidentally tested on benzil, so it is out of scope.

Why this set was reviewed

GO:0102306 "benzil reductase [(S)-benzoin-forming] activity" was obsoleted
(geneontology/go-ontology#30498, "not a physiologically relevant substrate"), and
RHEA:25968 was then attached as a skos:narrowMatch to the EC 1.1.1.- grouping
class GO:0016616. Consequences visible in GOA:

• The experimental benzil annotations (SGD IDA PMID:37602278; UniProt EXP
  PMID:11796169, PMID:11745140) now sit on GO:0016616, which only says "CH-OH
  donor, NAD(P) acceptor".
• New IEA rows (GO_REF:0000120, 2026-07-27, with/from RHEA:25968) are generated
  to the same grouping class.

Biology

• B. cereus YueD: found by expression screen; NADPH-dependent stereospecific benzil ->
  (S)-benzoin; broad aromatic carbonyl specificity (diketones, 1,4-naphthoquinone,
  p-nitrobenzaldehyde); GFP shows cytoplasm PMID:11796169.
• Yeast Irc24 (YIR036C) and its tandem paralog Nre1 (YIR035C, 52% identical) both reduce benzil and
  prefer NADPH 2:1 over NADH PMID:37602278.
• All substrates are xenobiotic; no physiological substrate is known for any member.
• Sister reaction RHEA:31891 (1-phenyl-1,2-propanedione, also EC 1.1.1.320) is
  already narrowMatch to GO:0004090 carbonyl reductase (NADPH) activity, which yields the
  GO_REF:0000116 IEA to GO:0004090 on IRC24 and on B. cereus YueD.

Decisions

• GO:0016616 rows -> MODIFY to GO:0004090. (S)-benzoin is a secondary alcohol, and
  benzil -> benzoin is ketone -> secondary alcohol with NADPH: exactly the
  GO:0004090/EC 1.1.1.184 reaction. EC 1.1.1.184's synonym list includes "xenobiotic ketone reductase".
• GO:0050664 (NAD(P)H, O2 acceptor) on IRC24 (SGD IDA 2013) and its IBA onto NRE1
  -> MODIFY to GO:0004090; the acceptor is benzil, not O2.
• Bacillus sepiapterin reductase / BH4 biosynthesis (PTHR44085 IBA/IEA): BH4
  process REMOVE (pathway absent in Bacillus); SPR activity over-annotated or
  undecided (never tested).
• No new GO term is needed: a benzil-specific term would reverse #30498, and
  GO:0004090 already covers the reaction class.

Proposed GO-side change

Move xref: RHEA:25968 {source="skos:narrowMatch"} from GO:0016616 to GO:0004090.
More generally, flag any RHEA xref on an EC-incomplete (x.x.x.-) grouping term:
currently 9 reactions on GO:0016616, GO:0008241 and GO:0016712.

Structural + kinetic analysis (OpenScientist, AlphaFold)

The one run of the five that delivered the measurement I asked for, and it
qualified my own quinone speculation.

What it measured. AlphaFold model AF-Q8RJB2-F1 v6 (the v4 URL in my
seed context 404s), global pLDDT 96.8. SDR catalytic tetrad
Asn86/Ser140/Tyr154/Lys158 at pLDDT 98.0-98.9, Rossmann motif TGTSQGLG at
residues 7-14. The substrate pocket is 33 residues, mean pLDDT 97.1, and
enclosed (only 7% of 400 probe rays escape). Candidate sterics from PubChem
3D: benzil is 10.0 A across and twisted (plane-RMSD 0.39), 1,4-naphthoquinone
is 5.3 A and perfectly flat (0.00), methylglyoxal 3.7 A. So the pocket does
discriminate sterically, and in the direction the reported affinities predict:
bulky twisted benzil binds worst.

Where it corrected me. I had read the low naphthoquinone Km as hinting at
quinone detoxification. Four arguments against making that the function:

1. Affinity is not efficiency. By kcat/Km the best substrate is
   1-phenyl-1,2-propanedione (3.93 min-1 uM-1), about 15x better than
   1,4-naphthoquinone (0.268) and 47x better than benzil (0.084). The best
   substrate is another synthetic diketone.
2. Family context points elsewhere. Foldseek's nearest named neighbours are
   B. subtilis YueD, yeast Irc24, then sepiapterin reductases and SDR
   ketoreductases. No quinone reductase appears at all.
3. The real bacterial quinone does not fit. Menaquinone is a large
   lipophilic isoprenoid; this pocket is small and soluble.
4. The mechanism may be the opposite of detoxification. In human SPR,
   quinone handling is NADPH-dependent redox cycling at the cofactor site,
   distinct from sepiapterin reduction and separable from it by the D257H
   substrate-site mutation
   PMID:23640889.
   That generates ROS rather than clearing an electrophile.

Both points are now knowledge_gaps on the core function rather than claims.

Quality signal. The run self-corrected mid-analysis: iteration 2 misread
UniProt's "in decreasing order" phrasing as a preference ranking; iteration 3
fetched the actual Km/kcat table and reversed its own judgment, flagging the
change explicitly. Its honest limitations are also stated: the model is apo, so
pocket metrics are geometric rather than energetic, and no docking or MD was run.

Corroborations. It independently endorses GO:0004090 as the best-supported
general core MF, and independently flags GO:0004757 (sepiapterin reductase) and
GO:0006729 (BH4 biosynthesis) as TreeGrafter-only inferences with no BH4 pathway
known in B. cereus - the same calls this review makes.