Non-homologous end joining (NHEJ) is a pathway that repairs double-strand breaks in DNA. It is called "non-homologous" because the break ends are directly ligated without the need for a homologous template, in contrast to homology directed repair (HDR), which requires a homologous sequence to guide repair. NHEJ is active in both non-dividing and proliferating cells, while HDR is not readily accessible in non-dividing cells. The term "non-homologous end joining" was coined in 1996 by Moore and Haber.

NHEJ is typically guided by short homologous DNA sequences called microhomologies. These microhomologies are often present in single-stranded overhangs on the ends of double-strand breaks. When the overhangs are perfectly compatible, NHEJ usually repairs the break accurately. Imprecise repair leading to loss of nucleotides can also occur, but is much more common when the overhangs are not compatible. Inappropriate NHEJ can lead to translocations and telomere fusion, hallmarks of tumor cells.

NHEJ implementations are understood to have been existent throughout nearly all biological systems and it is the predominant double-strand break repair pathway in mammalian cells. In budding yeast (Saccharomyces cerevisiae), however, homologous recombination dominates when the organism is grown under common laboratory conditions.

When the NHEJ pathway is inactivated, double-strand breaks can be repaired by a more error-prone pathway called microhomology-mediated end joining (MMEJ). In this pathway, end resection reveals short microhomologies on either side of the break, which are then aligned to guide repair. This contrasts with classical NHEJ, which typically uses microhomologies already exposed in single-stranded overhangs on the DSB ends. Repair by MMEJ therefore leads to deletion of the DNA sequence between the microhomologies.

Many species of bacteria, including Escherichia coli, lack an end joining pathway and thus rely completely on homologous recombination to repair double-strand breaks. NHEJ proteins have been identified in a number of bacteria, including Bacillus subtilis, Mycobacterium tuberculosis, and Mycobacterium smegmatis. Bacteria utilize a remarkably compact version of NHEJ in which all of the required activities are contained in only two proteins: a Ku homodimer and the multifunctional ligase/polymerase/nuclease LigD. In mycobacteria, NHEJ is much more error prone than in yeast, with bases often added to and deleted from the ends of double-strand breaks during repair. Many of the bacteria that possess NHEJ proteins spend a significant portion of their life cycle in a stationary haploid phase, in which a template for recombination is not available. NHEJ may have evolved to help these organisms survive DSBs induced during desiccation. It preferentially uses rNTPs (RNA nucleotides), possibly advantageous in dormant cells.

The archaeal NHEJ system in Methanocella paludicola have a homodimeric Ku, but the three functions of LigD are broken up into three single-domain proteins sharing an operon. All three genes retain substantial homology with their LigD counterparts and the polymerase retains the preference for rNTP. NHEJ has been lost and acquired multiple times in bacteria and archaea, with a significant amount of horizontal gene transfer shuffling the system around taxa.

Corndog and Omega, two related mycobacteriophages of Mycobacterium smegmatis, also encode Ku homologs and exploit the NHEJ pathway to recircularize their genomes during infection. Unlike homologous recombination, which has been studied extensively in bacteria, NHEJ was originally discovered in eukaryotes and was only identified in prokaryotes in the past decade.

In contrast to bacteria, NHEJ in eukaryotes utilizes a number of proteins, which participate in the following steps: