SIR3 curation notes

2026-09-02 Update

Audited existing_annotations for oversights.

• GO:0006303 (double-strand break repair via nonhomologous end joining), IMP,
  PMID:9501103: was marked REMOVE ("SIR3 itself is not a NHEJ component or
  catalyst"). This is contradicted by the cited paper's own abstract, which
  reports a direct in vivo functional assay: "using an in vivo plasmid rejoining
  assay, we demonstrate that SIR2, SIR3 and SIR4, three genes shown previously to
  function in TPE, are essential for Ku-dependent DSB repair" PMID:9501103. SIR3 is not core catalytic NHEJ machinery, but the
  IMP evidence directly demonstrates a genuine secondary requirement. Changed
  REMOVE to KEEP_AS_NON_CORE.
• GO:0070481 (nuclear-transcribed mRNA catabolic process, non-stop decay), IMP,
  PMID:17660569: was marked REMOVE with a specific mechanistic claim
  ("indirect through secondary effects on gene expression or cell stress
  responses") that isn't supported by anything in the cached source. The abstract
  is not gene-specific; I force-refetched full text (ai-gene-review fetch-pmid 17660569 --force, now cached with full_text_available: true), but the body/
  discussion text never names SIR3, and the results table (Table 2, listing all
  15 genes with per-gene phenotype values) is not recoverable from the cached
  page — table rows were not captured by the HTML extraction. Since the specific
  supporting detail for SIR3 cannot be verified, and per project guidance an
  experimental annotation should not be REMOVEd on an unverified assumption,
  changed action to UNDECIDED.
• All other annotations reviewed; no other genuine, evidence-backed oversights
  found.

2026-09-05 Update (review follow-up)

Follow-up to PR review feedback on the GO:0006303 entry: the previous reason
attributed the NHEJ requirement to SIR3's "telomere/chromatin maintenance role",
which was an unevidenced mechanism swapped in for another. Looked up the actual
literature (PMIDs resolved via ai-gene-review fetch-pmid, not guessed) and the
established mechanism is silencing-dependent control of NEJ1 expression:

• Deleting SIR genes derepresses the silent mating-type cassettes, and it is
  that derepression — not a Sir role at the break — that accounts for the NHEJ
  defect PMID:10421582. In haploids that retain their
  mating type, sir deletions left plasmid end-joining unaffected and reduced
  chromosomal NHEJ only two- to threefold.
• The downstream target is NEJ1, a haploid-specific NHEJ factor whose promoter
  carries an a1/α2 repressor site: "transcription of NEJ1 was completely
  repressed in a/alpha diploid and sir haploid strains. The NEJ1 promoter
  contained a consensus binding site for the a1/alpha2 repressor" PMID:11676923.
• Decisive test: restoring Nej1p bypasses the Sir requirement entirely
  PMID:11676923.

Given the rescue result, SIR3 is not a participant in end-joining; it acts
upstream, silencing HML/HMR so that NEJ1 stays expressed. Changed the action
from KEEP_AS_NON_CORE to MODIFY, proposing GO:2001032 (regulation of
double-strand break repair via nonhomologous end joining; id verified via OLS,
already used elsewhere in this repo). Chose the general regulation term over the
directional GO:2001034 because SIR3's contribution is permissive — maintaining
NEJ1 expression — rather than active modulation of the repair reaction itself.

Also addressed two smaller review points:

• The UNDECIDED GO:0070481 entry's supported_by quoted only the paper title,
  which reads as support for a claim that PR showed cannot be supported. Swapped
  in the recoverable methods sentence PMID:17660569 and added a breadcrumb naming
  Table 2 of that paper (or the SGD annotation detail) as the concrete unblocker.
• Flagged, without editing out of scope: the identical GO:0006303 / IMP /
  PMID:9501103 annotation — from one experiment assaying SIR2, SIR3 and SIR4
  together — is currently adjudicated three different ways across the repo
  (SIR2 REMOVE, SIR3 this entry, SIR4 MARK_AS_OVER_ANNOTATED). Those sibling
  entries should be reconciled with the NEJ1 rationale above in a separate pass.

2026-09-29 Update

Aligned SIR3 with the IBA propagation review:

• The GO:0006270 transfer from the ORC1/CDC6 PAINT family still carries a
  PROPAGATION_BAD/WRONG_ORTHOLOG_OR_PARALOG propagation review because Sir3
  is an ORC1-derived silent-chromatin scaffold that suppresses, rather than
  initiates, MCM loading at euchromatic origins. The GO:0003688 transfer is now
  TERM_SCOPING_PROBLEM: SIR3 itself is in that seed list via SGD's
  over-scoped origin-binding IDA from PMID:29795547, whose data support Sir3
  binding to origin-adjacent nucleosomes rather than origin DNA itself.
• The GO:0033314 checkpoint-signaling IBA points to the same PTN in the cached
  GOA row but is no longer present in the local 2026 PTHR10763 PAINT export, so
  it was marked SOURCE_STALE_OR_MISSING.
• The five remaining GO:0005515 protein binding rows from RAP1/SIR4
  interaction and high-throughput complex papers are now REMOVE: the physical
  interactions are not disputed, but no more-specific RAP1- or Sir4-binding MF
  term exists and generic protein binding should not be retained.

The newer full-text SIR3 literature does not reopen the IBA calls. Brothers and
Rine 2022 refine Sir3 recruitment and spread by Sir3-M.EcoGII/Nanopore mapping
but do not support origin-DNA binding or initiation. Bordelet et al. 2022 do
require a correction to the September NHEJ rationale: the original NEJ1
rescue literature still shows an indirect SIR contribution through HML/HMR
silencing, but SIR3 also has a direct Sae2-binding role that limits
MRX-dependent resection and promotes NHEJ PMID:34817085. The existing
GO:0006303 action therefore remains MODIFY, but the replacement is now the
directional GO:2001034 positive regulation of double-strand break repair via nonhomologous end joining. Because the same paper explicitly describes Sir3 as
"a direct negative regulator of Sae2", the proposed Sae2-directed molecular
function is the direction-preserving GO:0140678 molecular function inhibitor activity.

2026-10-16: GO:0070481 nonstop decay row resolved (issue #4027)

The September note left GO:0070481 UNDECIDED pending Table 2 of
PMID:17660569 or the underlying SGD detail. The table is still not in the
cached text (the PMC copy is not open access and the publisher page is
paywalled), but SGD's phenotype curation for this reference
(https://www.yeastgenome.org/backend/reference/17660569/phenotype_details,
retrieved 2026-10-16) lists sir3-null in the systematic mutation set as
"RNA accumulation: increased", together with SKI3, SKI7, SKI8, SSN2, SSN3,
MED1, NUP2, HTZ1, IPK1 and YGR122W; PRE9, UMP1 and RKR1 are scored instead for
increased protein accumulation. So SIR3 was one of the eleven hits that raised
the Protein A-nonstop reporter mRNA PMID:17660569. The paper gives no mechanism for SIR3, and notes that
most new hits did not behave like general exosome factors PMID:17660569.

Decision: MARK_AS_OVER_ANNOTATED. The phenotype is verified, so REMOVE is
not justified, but a deletion-screen hit shows necessity rather than
participation, and Sir3 is a nuclear heterochromatin protein with no known
activity in cytoplasmic mRNA surveillance. An indirect route through loss of
silencing is plausible (compare the indirect NEJ1/HML-HMR contribution to NHEJ
discussed above) but untested.

Follow-up after PR #5072 review (2026-10-16): the reason now quotes and answers
the authors' own reading of the eleven RNA-increased strains PMID:17660569. sir3-null passed the heterologous
Protein A reporter control, so the HIS3-promoter caveat (which the paper raises
for SSN2/SSN3/MED1) was dropped as support. The reviewer suggested sir3-null
must be one of the "eight other mutants" with a longer Protein A-nonstop
half-life in Figure 6. The count does not close: SGD's eight non-SKI,
non-proteasome RNA-increased strains (SIR3, YGR122W, HTZ1, SSN2, SSN3, NUP2,
MED1, IPK1) plus ylr021w, which the text puts in Figure 6 but SGD does not list
in that phenotype set, give nine candidates for eight slots. Either way the
action stands, because a half-life increase is also a necessity readout.