Classical non-homologous end joining of DNA double-strand breaks — Ku70/Ku80 end binding, DNA-PKcs recruitment and autophosphorylation, Artemis/PNKP/APTX/APLF end processing, pol mu/pol lambda/TdT gap filling, and XRCC4-LIG4 ligation with XLF and PAXX; XRCC6/XRCC5/PRKDC/DCLRE1C/PNKP/APTX/APLF/POLM/POLL/DNTT/XRCC4/LIG4/NHEJ1/PAXX

Classical non-homologous end joining (c-NHEJ) rejoins the two ends of a DNA double-strand break (DSB) directly, without a homologous template. It operates throughout the cell cycle, is the predominant DSB repair pathway in G1 and in non-dividing cells, and is the joining step of V(D)J recombination and class switch recombination in lymphocytes. The Ku heterodimer (Ku70/XRCC6 and Ku80/XRCC5) threads onto each broken end through a preformed ring and recruits the DNA-dependent protein kinase catalytic subunit (DNA-PKcs, PRKDC) to form the DNA-PK holoenzyme, which tethers the two ends in a synaptic complex. DNA-PKcs is activated by the DNA end, autophosphorylates (which remodels the synapse and releases the ends for processing) and phosphorylates other NHEJ factors, including the nuclease Artemis (DCLRE1C). Incompatible or damaged ends are processed by enzymes chosen by end chemistry: the Artemis-DNA-PKcs complex opens hairpins and trims 5' and 3' overhangs; the polynucleotide kinase-phosphatase PNKP restores 5'-phosphate and 3'-hydroxyl termini; aprataxin (APTX) removes 5'-adenylate left by abortive ligation; and APLF, recruited by poly(ADP-ribose) and by Ku, acts as a scaffold that retains the ligase complex and has reported nuclease activity. Gaps are filled by the family X polymerases pol mu (POLM) and pol lambda (POLL), and in developing lymphocytes the template-independent terminal deoxynucleotidyl transferase (TdT, DNTT) adds non-templated nucleotides to V(D)J coding ends. The ends are finally sealed by DNA ligase IV (LIG4) in a constitutive complex with its scaffold XRCC4; the XRCC4 paralogs XLF (NHEJ1) and PAXX bind XRCC4 and Ku respectively, bridge and align the ends, and stimulate ligation, with XLF and PAXX partly redundant. Upstream, the 53BP1-RIF1 axis of ATM signalling protects DSB ends from resection in G1 and so favours c-NHEJ over homologous recombination.

Derived from the RGD Pathway Ontology diagram PW:0000203 through the RGD_PATHWAY_ONTOLOGY project's diagram->module route (seed in projects/RGD_PATHWAY_ONTOLOGY/data/module_seeds.yaml, tier related_module_only; the only related module found by the seed was bacterial_nonhomologous_end_joining). RGD curates 16 human genes to PW:0000203: APLF APTX DCLRE1C DNTT LIG4 NHEJ1 PNKP POLL POLM PRKDC RBBP8 RIF1 TP53BP1 XRCC4 XRCC5 XRCC6. Thirteen are modelled as annotons here; TP53BP1 and RIF1 are included by module_ref (part 1), and RBBP8 is excluded (decision 3). One gene product not on the RGD list is added: PAXX, the XRCC4/XLF paralog identified in 2015, after the RGD diagram was drawn. Molecular functions were taken from each protein's experimental GO annotations in QuickGO (November 2026) and, where a gene review exists (DCLRE1C, PRKDC), checked against its core functions. Boundary decisions. (1) Upstream signalling is not re-modelled. Part 1 is a module_ref to atm_dsb_signaling#tp53bp1_rif1_arm, the 53BP1-RIF1 arm of that module's pathway-choice node, which already carries TP53BP1 and RIF1 with their activities and GO-CAM links (gomodel:617b481400000460, the seed's best-overlap GO-CAM for PW:0000203, whose only overlap with this pathway is TP53BP1; it is not linked again here). PW:0001361, which RGD links from the NHEJ diagram, is mentioned here but not listed as an evidence source. The part is optional because Ku-dependent joining does not itself require 53BP1. Shieldin and CST, the downstream effectors of RIF1, are not on the RGD list and are not modelled. (2) Alternative end joining is excluded and recorded as context only. RGD's gene list decides scope, and none of the alt-EJ machinery (POLQ, PARP1, LIG3 with XRCC1, LIG1) is curated to PW:0000203; the RGD description and diagram describe only the Ku-DNA-PK-XRCC4-LIG4 pathway. Alt-EJ is mechanistically distinct (it needs resection and microhomology, is Ku-independent and is suppressed by Ku), has its own GO term (GO:0097681), and would be a separate module. For the same reason the module concept is GO:0097680 (classical NHEJ), the child of the mapped term GO:0006303. (3) RBBP8 (CtIP) is excluded although RGD curates it to this pathway: its activity is the initiation of end resection, which commits a break away from c-NHEJ toward HR or alt-EJ; it is a participant of modules/homologous_recombination_dsb_repair.yaml. Its presence on the RGD list probably reflects its role in pathway choice. (4) End processing is a ROUTE variant set of distinct activities for distinct end chemistries (ONE_OR_MORE) in an optional part. PNKP has two annotons, one per catalytic activity. APLF is given DNA endonuclease activity, its experimental IDA row (PMID:17396150) and the activity RGD states, but its best-documented NHEJ role is as a poly(ADP-ribose)- and Ku-binding scaffold that retains XRCC4-LIG4 (PMID:21211721), and no cited study shows that the nuclease is needed for joining in cells; this is recorded as a knowledge gap rather than resolved. APTX activity was demonstrated on single-strand nicks and gaps; its DSB role rests on its XRCC4 interaction. (5) The Pol X polymerases are a REALIZER variant set in an optional part. DNTT (TdT) is on the RGD list and is included as a lymphocyte-restricted variant because its N-addition is part of NHEJ-mediated V(D)J coding-joint formation; V(D)J recombination itself (RAG1/RAG2) is not modelled. (6) Ligation is one part with the XRCC4-LIG4 complex and the XRCC4 paralogs XLF and PAXX as separate annotons. XLF carries DNA end binding (IDA) and PAXX the generic molecular adaptor activity (IDA), the only non-binding experimental MF rows each has; XRCC4 carries protein-macromolecule adaptor activity (IMP). (7) DNA-PKcs is modelled as a single annoton with DNA-dependent protein kinase activity and its two functionally established targets (itself and Artemis); its many other substrates, its H2AX kinase role (backup to ATM, see atm_dsb_signaling) and its non-repair roles are outside the module. (8) Ku is modelled with DNA end binding; Ku70's 5'-dRP/AP lyase activity is noted only in a role. (9) No substrate is modelled as a participant; every annoton is an enzyme or a DNA-binding scaffold. (10) modules/bacterial_nonhomologous_end_joining.yaml is the bacterial analogue (Ku with the multifunctional LigD) and is not referenced; it shares the Ku end-binding step but not the eukaryotic DNA-PK, XRCC4-LIG4 or Pol X machinery. GO-CAM and GOA observations. No human GO-CAM in gocams/index.tsv models the core c-NHEJ machinery (Ku, DNA-PKcs, Artemis, Pol X, XRCC4-LIG4, XLF, PAXX) in GO:0006303 or GO:0097680; the activities that are in that process in the index are 53BP1, RNF168, RAP80 and TOPBP1, which belong to the ATM module. The cached models that contain members do so for other processes (XRCC4 in gomodel:63c0ac2b00002148, DSB attachment to the nuclear envelope; APLF in gomodel:627d95ee00002067, PARP-dependent recruitment; PRKDC in cGAS and TONSL-MMS22L models), so no gocam_associations are asserted. POLM has no experimental biological-process row in human GOA, and DCLRE1C has no experimental process row for NHEJ or V(D)J recombination although both are well established. PANTHER family ids and PTNs are deliberately not asserted.