LIG4 (human, P49917) review notes

Literature search (manual; no deep-research provider available)

Sources checked:
- All 38 GOA-cited PMIDs, cached with just fetch-gene-pmids human LIG4.
- PubMed E-utilities searches (2026-11-04): "Frank KM[au] late embryonic lethality DNA ligase
IV 1998" -> PMID:9823897; "Barnes DE[au] targeted disruption ligase IV 1998" ->
PMID:9889105; "Lieber MR[au] mechanism double-strand break repair nonhomologous
end-joining 2010" -> PMID:20192759. Fetched into publications/.
- Reactome cached entries for the NHEJ ligation reactions and HIV 2-LTR circle formation.
- UniProt P49917 record.
- Context: telomere end-joining pathway choice (module and GO-CAM review proposed in ai4curation/ai-gene-review#5904; not on this branch and not used as evidence): LIG4 as the ligase of the
classical-NHEJ route that fuses telomeres lacking TRF2 PMID:28216226.

Key findings

• ATP-dependent nick sealing; purified from HeLa nuclei PMID:8798671. Crystal structures of catalytic core on nicked DNA
  PMID:29980672. Mg2+ cofactor (PMID:29980672
  metal screen).
• Obligate partner XRCC4 (BRCT-linker interface, PMID:19332554), stimulates ligation
  PMID:9242410. XRCC4 imports and stabilizes LIG4
  (PMID:21982441) and LIG4 reciprocally controls XRCC4 nuclear localization (PMID:24984242).
• Ligase of classical NHEJ and V(D)J recombination PMID:9809069. Flexible
  ligase that ligates incompatible ends and across gaps (PMID:17290226). Component of NHEJ
  supercomplexes with Ku, DNA-PKcs, XRCC4, XLF (PMID:34352203, PMID:25941166).
• LIG4 is dispensable for viability in human HCT116 cells; needed for TRF2-dominant-negative
  telomere fusions PMID:23275564; most end joining in human cells is LIG4-dependent, LIG3
  only for Ku-suppressed A-NHEJ PMID:24837021.
• Mouse: Lig4 knockout is late-embryonic lethal with neuronal apoptosis and blocked
  lymphopoiesis [PMID:9823897; PMID:9889105]. These underlie the many ISS rows (development,
  proliferation, neuron apoptosis), which are downstream consequences of unrepaired DSBs.

Curation decisions of note

• GO:0003909 and GO:0016874 rows MODIFY -> GO:0003910.
• GO:0006297 "nucleotide-excision repair, DNA gap filling" rows: the cited papers
  (PMID:12517771, PMID:17290226) describe gap filling/ligation during NHEJ of DSBs, not NER.
  The IDA rows are repointed to GO:0006303; the IBA (seeded only by these LIG4 rows) is
  removed. Recorded as source defects.
• GO:0071479 IGI PMID:29463814: the paper is entirely on C. elegans (LEM-3 with lig-4
  alleles); the annotation belongs to worm lig-4, not human LIG4. REMOVE + source defect.
• Base-excision repair IDA (PMID:19589734) and single strand break repair IDA (PMID:8798671)
  are in vitro nick-ligation results; LIG4 has no in vivo role in BER/SSBR -> over-annotated.
• Mouse phenotype ISS rows: developmental ones kept as non-core; proliferation/apoptosis/
  neurogenesis/stem-cell/chromosome-organization rows marked over-annotated.
• 25 bare protein-binding rows REMOVE; partner relationships are captured by the DNA ligase
  IV complex, DNA-PK-LIG4 complex and NHEJ complex CC terms.

PR #5944 review follow-up

• Rule applied across the NBN and LIG4 reviews: a developmental or tissue-level process term whose only basis is the knockout phenotype of a housekeeping DNA repair factor (lethality, apoptosis, progenitor loss) records necessity, not participation, and is MARK_AS_OVER_ANNOTATED; it is kept as non-core only where the gene performs a step inside the process itself (as LIG4 does in V(D)J joining during lymphocyte differentiation). Applied to GO:0001701 and GO:0007417 (changed from KEEP_AS_NON_CORE); GO:0002328 pro-B cell differentiation, GO:0033077 T cell differentiation in thymus and GO:0045190 isotype switching stay non-core because LIG4 performs the joining step inside them.