Is Non-Homologous End-Joining Really an Inherently Error-Prone Process?

For decades, every molecular biology textbook taught that your cells have two ways to fix a broken chromosome — one careful and one sloppy. Homologous recombination was the careful one, using an intact template to restore the sequence faithfully. Non-homologous end-joining was the sloppy one, the repair of last resort, stitching broken ends back together and leaving a mess. That was the consensus. Then the data started saying something different. Bétermier, Bertrand, and López ask the question in the title of their paper: is non-homologous end-joining really inherently error-prone? The answer they build toward is no, not exactly. The reason why turns out to hinge on a conflation that ran through decades of experimental work. Start with the problem itself. A DNA double-strand break, which is often abbreviated as DSB, is one of the most dangerous things that can happen inside a cell. Both strands of the double helix are severed, leaving you with two loose ends that have no template, no tether, and no obvious way back to each other. The authors describe DSBs as "highly toxic lesions" that, if mishandled, produce genomic instability. But here's the complication: not all DSBs are accidents. During normal physiology, cells deliberately create double-strand breaks during immune system development and certain forms of chromosome compaction, and they repair them using the same machinery that fixes stress-induced damage. So DSB repair sits, as Bétermier and colleagues put it, "at the crossroads between genetic variability and instability." The cell has to be both faithful and flexible. That tension is what makes the repair system so interesting. The two canonical repair strategies divide the labor differently. Homologous recombination, or HR, uses a sister chromatid as a template and reconstructs the broken sequence accurately. It's restricted to the S and G2 phases of the cell cycle, when that sister chromatid is actually present. Non-homologous end-joining, abbreviated as NHEJ, works without a template; it joins ends directly and operates throughout the cell cycle. The textbook verdict is that HR is accurate, while NHEJ is mutagenic. Bétermier and colleagues argue that this verdict is too simplistic. The reason becomes clear once you realize that "NHEJ" has been used as a label for two distinct processes that have radically different fidelity profiles. This is the conceptual hinge the whole paper turns on. Canonical non-homologous end-joining, where the "C" stands for classical, is driven by the Ku70 and Ku80 protein heterodimer, which clamps directly onto broken DNA ends. Ku recruits DNA-PKcs and end-processing factors, and the final ligation is carried out by the Xrcc4 and ligase IV complex. That's one system. Alternative end-joining, abbreviated as A-EJ, also called microhomology-mediated end-joining or B-NHEJ, is a different beast entirely. It operates independently of Ku and Xrcc4. Instead of protecting ends, A-EJ initiates resection: the ends are chewed back, exposing short stretches of matching sequences called microhomologies, typically two to four nucleotides long, which align the ends before ligation. Parp1 initiates the process, Mre11 and CtIP carry out the resection, and Xrcc1 and ligase III seal the result. A-EJ is genuinely error-prone. It always leaves deletions at repair junctions, often uses microhomologies far from the original break, and actively promotes chromosome translocations. CtIP and ligase III have both been shown to drive translocations via A-EJ, and loss of Ku — the protein that normally protects ends — increases deletion sizes, end mobility, and rearrangement risk. The key insight from Bétermier and colleagues is this: much of the mutagenic end-joining historically blamed on "NHEJ" was actually A-EJ, unmasked in experiments where C-NHEJ components were absent or overwhelmed. Now look at what the evidence shows when C-NHEJ is actually allowed to operate. The paper draws on data from multiple organisms to make one central claim: the accuracy of repair is dictated by the structure of the DNA ends, not by the C-NHEJ machinery itself. The most striking example comes from Paramecium. During development of its macronucleus, the Paramecium genome is amplified to an extraordinary 800-fold ploidy, and at least 45,000 short internal sequences are excised, producing an estimated one million double-strand breaks per developing macronucleus. Every one of those breaks requires repair. The repair, which depends on ligase IV and Xrcc4, proceeds with very limited end processing. The junctions are restored with high precision across tens of thousands of events. When ends are competent for ligation, C-NHEJ is massively conservative. Mammalian cell experiments using the meganuclease I-SceI allow researchers to set the terms precisely. When I-SceI cuts to produce fully complementary three-prime overhangs, the fraction of error-free C-NHEJ events in wild-type mammalian cells runs from thirty-five to seventy-five percent, and the authors note that even these figures are probably underestimates because a perfectly repaired site can be re-cleaved and re-repaired repeatedly before it gets counted. When the overhangs are not fully complementary, C-NHEJ adapts. Ninety to ninety-five percent of end-joining events retain at least one protruding nucleotide, minimizing sequence loss. A-EJ, by contrast, removes all protruding nucleotides and typically more. Remove Ku80 or Xrcc4 from these cells and the error-free events disappear entirely, replaced by A-EJ signature deletions built on distant microhomologies. In yeast, Saccharomyces cerevisiae data suggest that when the ends are clean, NHEJ restores a re-cleavable site at least ninety-nine point nine percent of the time. So where do the mutations come from? Two sources. First, when DNA ends are chemically damaged, as they are after ionizing radiation, they cannot be ligated directly. The ends have to be cleaned up first, and that upstream processing step is where nucleotides are lost. The mutations belong to the cleaning, not to the ligation machinery. Second, in cells or contexts where C-NHEJ is compromised, A-EJ takes over and produces the deletion-riddled junctions that gave NHEJ its bad reputation. This reframing has immediate implications for biology we care about. The immune system is the best example. V(D)J recombination — the process that generates antibody and T-cell receptor diversity — works by deliberately creating double-strand breaks and then rejoining them imprecisely enough to produce a vast range of sequences. The RAG1 and RAG2 enzymes cut DNA and release coding ends as hairpins. Hairpin resolution produces variable end structures, and then terminal deoxynucleotidyl transferase, or TdT, adds non-templated N-nucleotides at coding junctions — random bases inserted to maximize sequence diversity. That's where the diversity comes from: hairpin resolution plus TdT activity, both of which are accessory mechanisms layered on top of C-NHEJ. The signal joints — the other product of the same reaction — are largely error-free because the signal ends are blunt and directly ligatable. C-NHEJ handles them cleanly. The same machinery generates diversity in one context and precision in another because the ends it receives are different. Ionizing radiation tells a complementary story. Radiation-induced breaks carry chemically altered ends — damaged bases and strand breaks with missing phosphate groups — that must be processed before ligation. The mutagenesis at resealed radiation-induced junctions reflects that processing, not an inaccurate ligation machine. Crucially, Ku itself carries a specialized enzymatic activity: a five-prime dRP and AP lyase that targets damaged ends at DSBs and restricts nucleotide loss. Without that activity, organisms would be far more sensitive to radiation even at low doses. C-NHEJ's versatility is what enables resistance to ionizing radiation, including the environmental and medical exposures most of us encounter. There are also direct practical consequences for genome engineering. Strategies that use site-specific nucleases, like CRISPR-style approaches, rely on the cell's repair machinery to process the resulting cut. In wild-type cells, C-NHEJ dominates. A-EJ is less efficient and produces larger, less predictable deletions through uncontrolled resection. The paper notes that ectopic expression of TdT can introduce nucleotides at I-SceI generated ends in a Ku and ligase IV dependent manner even in non-lymphoid cells, suggesting C-NHEJ can be deliberately tuned to introduce diversity at defined sites. Strategies that steer repair through C-NHEJ rather than A-EJ should minimize the risk of collateral genomic instability. The conclusion Bétermier, Bertrand, and López reach is clean: C-NHEJ is conservative but adaptable. At the chromosome level, it limits break mobility and large rearrangements. At the nucleotide level, when ends are directly ligatable, it is largely error-free. When ends require processing, that processing introduces mutations, but that's an upstream constraint imposed by DNA chemistry, not a deficiency of the C-NHEJ machinery. The error-prone reputation belongs to A-EJ and to the conditions that force end processing. The broader point is this: how you label a process shapes how you study it. If NHEJ was never truly error-prone, if the errors came from a distinct parallel pathway and from the chemistry of the breaks, not from the joining machinery, then decades of experiments interpreted through the wrong label need reinterpretation. For cancer biology, radiation therapy, and genome editing, the distinction between C-NHEJ and A-EJ is not a technical footnote; it is the difference between a pathway that protects the genome and one that destabilizes it. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.