Contested function

Proteins that keep an enzyme's fold but not its activity

AI Gene Review · projects/CONTESTED_FUNCTION · 2026

Bottom line

  • Pseudo-enzymes keep a family's domain but have lost catalysis, so IBA, IEA and sometimes experimental rows give them an activity biochemistry contradicts.
  • We reviewed four cases in depth: Epe1 (fission yeast) and three biosynthetic-cluster proteins, eryCII, pqsB, actI-ORF2.
  • Epe1's 4 electronic catalytic rows are REMOVE (none has a replacement term; binding functions are separate NEW rows); the 2 IDA/EXP rows and metal ion binding are UNDECIDED. Finding more candidates has not started.

Why these cases

  • They need several evidence types at once: domain homology, activity assays, catalytic-site mutants, and the alternative mechanism.
  • A wrong catalytic term on a pseudo-enzyme is re-propagated by IBA and domain IEA until something records the loss.
  • The Epe1 case was presented at the GO Consortium meeting, October 2025.

Epe1: eraser or recruiter?

Evidence against demethylase activity

  1. The JmjC iron triad is H297-E299-Y370: position 370 is a Tyr where canonical JmjC demethylases have the third iron-ligating His. Yet Y370A loses function (Raiymbek 2020): the Tyr is required, though required is not catalytic.
  2. Mass spectrometry shows no demethylation of H3K9me2/me3 peptides.
  3. H297A keeps prevention of de novo H3K9me (nearly fully), so part of the anti-silencing activity does not need H297 (Sorida 2019). One experiment read two ways: its other arm, lost removal of established H3K9me, is on the next slide.
  4. The C-terminus alone, without JmjC, partly disrupts heterochromatin (a hypomorph).

Sources named on the project page: Ayoub 2003, Trewick 2007, Wang 2013 (Bdf2, PMID:24013502), Audergon 2015, Bao 2019, Sorida 2019, Raiymbek 2020.

Evidence for a catalytic contribution (not excluded)

  • The JmjC domain is essential for Epe1 activity (Ayoub 2003) and for its effect on Pol II accessibility (Zofall & Grewal 2006, who note the mechanism may differ from demethylases).
  • Y307A is recorded as loss of function, but from the same Ayoub paper, possibly the same experiment. Y307 is a 2-oxoglutarate-site residue, not an Fe ligand (Raiymbek; Sorida): cofactor pocket implicated, catalysis not isolated.
  • Sorida 2019, independent of Ayoub: under H297A, removal of established H3K9me is lost. (Retained ectopic H3K9me is shown for H297A, Y307A and Y370A by Raiymbek too.) One experiment read two ways: its other arm, retained prevention, is on the previous slide.
  • Wang 2015 (Mst2/Epe1): epe1-H374A and epe1-Y307A read as enzymatically dead, redundant with Mst2 (residue 374 of O94603 is Thr, so the His cannot be identified).
  • Against: no study has measured demethylation directly, and the same mutations weaken Swi6 binding and localization.

Sources: Ayoub 2003 (PMID:12773576), Zofall & Grewal 2006 (PMID:16762840), Sorida 2019 (PMID:31206516), Wang 2015 (Mst2/Epe1, PMID:25774602), Raiymbek 2020 (PMID:32195666).

Four pseudo-enzymes

Contested claims inside a sound paper

  • Some disputes are about one finding, not the whole reference.
  • The schema's finding_review records these: finding_status: DISPUTED + superseded_by.
  • Example, eryCIII: a 2004 claim that EryCIII is highly active on its own (PMID:15303858) is DISPUTED, superseded by the 2012 structure study (PMID:22056329) showing it is inactive without EryCII.

Status and next steps

  • ✅ Epe1, eryCII, pqsB, actI-ORF2 reviewed.
  • ⬜ Identify suspected pseudo-enzymes (Priority 2): the Top-Nots pseudo-enzyme pattern already names PLD5, AKTIP, AIP, CG6051, CPT1C, Pld4.
  • ⬜ Screen reviewed genes for over-annotated domain-based IBA/IEA rows (Priority 3).

Read more: projects/CONTESTED_FUNCTION.md · genes/SCHPO/Epe1/ · projects/TOP_NOTS.md