Non-Homologous End Joining

Non-homologous end joining (NHEJ, specifically canonical NHEJ or c-NHEJ) is the predominant pathway for repairing DNA double-strand breaks (DSBs) in multicellular eukaryotes, including human cells. Unlike homologous recombination (HR), which requires an intact sister chromatid as an extensive sequence template and is therefore restricted to the S and G₂ phases of the cell cycle, NHEJ directly synapses, processes, and ligates two broken DNA ends back together without requiring sequence homology. Consequently, NHEJ operates throughout the cell cycle—serving as the primary rapid DSB DNA repair mechanism during G₀ and G₁, when the majority of differentiated somatic cells reside—and repairs approximately 75–85% of ionizing-radiation-induced DSBs in human cells. Because incompatible or chemically damaged DNA termini must often be trimmed by nucleases or filled in by polymerases prior to ligation, NHEJ frequently introduces small insertions or deletions (indels of 1–10 base pairs) at the repair junction, a property exploited both physiologically by the adaptive immune system during V(D)J recombination and technologically in CRISPR gene editing.

A DNA double-strand break—in which both complementary phosphodiester backbones of the duplex are severed within a few base pairs of each other—is a cytotoxic lesion. An unrepaired DSB can trigger permanent G₁/G₂ arrest or apoptosis, whereas misjoining two breaks on different chromosomes generates dicentric chromosomes, acentric fragments, or oncogenic chromosomal translocations that drive cancer. Endogenous and exogenous DSBs arise from:

• Ionizing radiation and radiomimetic drugs (X-rays, γ-rays, bleomycin), which shatter deoxyribose rings via hydroxyl free radicals, leaving unligatable ends bearing 3′-phosphoglycolate (3′-PG), 3′-phosphate, or 5′-hydroxyl (5′-OH) groups instead of standard 3′-OH and 5′-phosphate termini.
• Reactive oxygen species (ROS) generated during aerobic cellular respiration and topoisomerase II poisons (etoposide, doxorubicin).
• Programmed physiological breaks introduced by the RAG1/RAG2 recombinase during lymphocyte V(D)J recombination and by activation-induced cytidine deaminase (AID) during immunoglobulin class-switch recombination (CSR).
• Targeted genome engineering nucleases, including Cas9, Cas12a, zinc-finger nucleases (ZFNs), and TALENs used in genetic engineering.

Whether a cell repairs a DSB via canonical NHEJ or homologous recombination is governed at the earliest step by a molecular competition over 5′→3′ DNA end resection:

• In G₀ and G₁ phase: Cyclin-dependent kinase (CDK) activity is low, and the chromatin reader 53BP1 (recruited to H4K20me2 and RNF168-ubiquitinated H2AK15ub surrounding the break) recruits RIF1 and the Shieldin complex (SHLD1/2/3–REV7). This 53BP1–RIF1–Shieldin axis blocks nucleases from resecting the 5′ DNA strand, preserving blunt or short-overhang ends for rapid capture by the Ku70/Ku80 heterodimer and channeling repair into c-NHEJ.
• In S and G₂ phase: High CDK1/2 activity phosphorylates CtIP, enabling BRCA1–BARD1 to antagonize 53BP1 and activate the MRN complex (MRE11–RAD50–NBS1) and long-range nucleases (EXO1 and DNA2–BLM). Extensive 5′→3′ resection generates long (>100–1,000 nt) 3′ single-stranded DNA tails coated by RPA and RAD51, which cannot be bound by Ku70/Ku80 and are committed to homologous recombination.

Canonical NHEJ proceeds through four sequential biochemical stages coordinated by a dynamic multiprotein synapse:

Stage 1: Rapid End Binding by the Ku70/Ku80 Heterodimer

Within 1–2 seconds of DSB formation, the nuclear Ku heterodimer (~400,000 molecules per human cell), composed of Ku70 (70 kDa, XRCC6) and Ku80 (86 kDa, XRCC5), binds the exposed DNA terminus in a sequence-independent manner with picomolar affinity (K_d ≈ 10⁻¹⁰ M). As revealed by John Walker and Stephen Jackson's crystal structure (2001), Ku70 and Ku80 form an asymmetrical toroidal (ring-shaped) basket whose positively charged central channel encircles two turns (14–20 bp) of B-DNA, protecting the break from exonucleases and serving as the docking platform for downstream NHEJ enzymes.

Stage 2: DNA-PKcs Recruitment and Long-Range to Short-Range Synapsis

Once threaded onto the DNA end, Ku translocates inward by ~10–14 bp and recruits the DNA-dependent protein kinase catalytic subunit (DNA-PKcs, 469 kDa, encoded by PRKDC) via the conserved C-terminal peptide of Ku80 to assemble the active DNA-PK holoenzyme (a PIKK-family serine/threonine protein kinase alongside ATM and ATR). Two DNA-PK holoenzymes initially bridge opposing DSB ends in a protective long-range synaptic complex (~115 Å apart) and undergo DNA-stimulated autophosphorylation at the ABCDE (Thr2609–Thr2647) and PQR (Ser2056) serine/threonine clusters, releasing DNA-PKcs from the break apex and transitioning the assembly into an aligned short-range synaptic complex stabilized by XRCC4, XLF, and PAXX.

Stage 3: Chemical Cleansing, Nuclease Trimming, and Polymerase Gap Filling

While clean restriction breaks can be ligated directly without mutation, biological DSBs bearing damaged sugars or incompatible overhangs are processed into ligatable 5′-phosphate and 3′-OH termini by three enzyme groups: Polynucleotide kinase 3′-phosphatase (PNKP) (phosphorylating 5′-OH and dephosphorylating 3′-P ends), TDP1/2, and Aprataxin (APTX); the DNA-PKcs-activated endonuclease Artemis (DCLRE1C), which trims 5′/3′ overhangs and opens covalent V(D)J coding hairpins; and the X-family DNA polymerases Pol λ (template-dependent gap filling), Pol μ (template-independent synthesis across discontinuous ends), and lymphoid-specific TdT (adding non-templated N-nucleotides).

Stage 4: Ligation by the XRCC4–DNA Ligase IV–XLF Complex

The final step is catalyzed by ATP-dependent DNA ligase IV (LIG4) bound constitutively in a tight complex with a homodimer of the scaffolding protein XRCC4, which stabilizes LIG4 against proteasomal degradation and stimulates adenylation of the catalytic Lys273 residue. XLF (Cernunnos) and PAXX form alternating helical filaments with XRCC4 that bridge and align the two DNA ends in the short-range synapse, allowing XRCC4–Ligase IV to seal both phosphodiester nicks across short gaps or mismatches before the ubiquitin-selective p97 (VCP) segregase extracts ubiquitylated Ku80 from the restored DNA duplex.

When c-NHEJ or HR is compromised, cells utilize two resection-dependent error-prone backup pathways: Alternative End Joining (alt-EJ / MMEJ / Theta-Mediated End Joining [TMEJ]) and Single-Strand Annealing (SSA).

Note that Pol θ (POLQ)-driven TMEJ is synthetically lethal with BRCA1/2 deficiency. Physiologically, c-NHEJ is required to resolve RAG1/RAG2 hairpin coding ends during lymphocyte V(D)J recombination (where biallelic mutations in DCLRE1C, PRKDC, LIG4, or NHEJ1 cause radiosensitive RS-SCID), and in CRISPR gene editing, c-NHEJ-generated frameshift indels at Cas9 cuts produce targeted gene knockouts.

Is canonical non-homologous end joining inherently error-prone?

A widespread textbook misconception states that c-NHEJ is intrinsically inaccurate at every break. Biochemical and sequencing studies demonstrate that when two DNA ends are chemically clean and compatible (such as blunt ends or complementary 5′/3′ restriction overhangs bearing undamaged 5′-phosphate and 3′-OH groups), c-NHEJ ligates them accurately without losing or adding a single nucleotide. Indel mutations arise primarily when the breaking agent (such as ionizing radiation or persistent Cas9 recutting) damages the terminal deoxyribose sugars or bases, forcing processing enzymes (Artemis, PNKP, Pol μ/λ) to trim or fill in a few nucleotides before XRCC4–DNA Ligase IV can seal the phosphodiester backbone.

Why do cells use NHEJ more often than homologous recombination if homologous recombination is error-free?

Two biological constraints favor c-NHEJ over homologous recombination in mammalian cells. First, homologous recombination requires an identical sister chromatid nearby to serve as a repair template, which exists only after DNA replication in late S and G₂ phases; post-mitotic neurons, cardiomyocytes, quiescent stem cells, and G₁ proliferating cells lack a sister chromatid, and using the homologous chromosome inherited from the other parent would risk loss of heterozygosity (LOH). Second, the human genome consists of >50% repetitive DNA (such as Alu and LINE-1 elements); initiating homologous recombination during G₁ across non-allelic repeats would cause deleterious chromosomal rearrangements.

Why do inherited mutations in Artemis (DCLRE1C) or DNA-PKcs (PRKDC) cause combined immunodeficiency (SCID)?

During B- and T-lymphocyte development in the bone marrow and thymus, the RAG1/RAG2 recombinase cuts V, D, and J gene segments and leaves the two coding DNA ends sealed as covalent DNA hairpins. Opening these hairpins requires DNA-PKcs to bind Ku70/Ku80 at the hairpin apex, autophosphorylate, and activate the Artemis endonuclease. Without functional DNA-PKcs or Artemis, developing lymphocytes cannot open the coding hairpins to assemble immunoglobulin or T-cell receptor genes, arresting lymphocyte differentiation and producing T⁻B⁻NK⁺ combined immunodeficiency alongside cellular sensitivity to ionizing radiation.

1. Lieber, M. R. (2010). The mechanism of double-strand DNA break repair by the nonhomologous DNA end-joining pathway. Annual Review of Biochemistry, 79, 181–211. https://doi.org/10.1146/annurev.biochem.052308.093131
2. Walker, J. R., Corpina, R. A., & Goldberg, J. (2001). Structure of the Ku heterodimer bound to DNA and its implications for double-strand break repair. Nature, 412(6847), 607–614. https://doi.org/10.1038/35088000
3. Zhao, B., Rothenberg, E., Ramsden, D. A., & Lieber, M. R. (2020). The molecular basis and disease relevance of non-homologous DNA end joining. Nature Reviews Molecular Cell Biology, 21(12), 765–781. https://doi.org/10.1038/s41580-020-00297-8
4. Chang, H. H. Y., Pannunzio, N. R., Adachi, N., & Lieber, M. R. (2017). Non-homologous DNA end joining and alternative pathways to double-strand break repair. Nature Reviews Molecular Cell Biology, 18(8), 495–506. https://doi.org/10.1038/nrm.2017.48