Is Non-Homologous End-Joining Really An Inherently Error-Prone ...

The end-joining accuracy of directly ligatable DNA ends.

We will first discuss NHEJ in two biological models, Paramecium and mammalian cells, in which this pathway is of particular importance. Paramecium provides a physiological example of the efficient contribution of C-NHEJ to the precise repair of thousands of developmentally programmed DSBs [45]. In mammalian cells, C-NHEJ is a prominent DSB repair mechanism, and it is essential in fundamental processes establishing the immune repertoire. The accuracy of NHEJ will then be addressed in yeast, bacteria, and plants, and during cut-and-paste transposition.

Similar to other ciliates, Paramecium harbors two different nuclei in its cytoplasm. During vegetative growth, the diploid micronucleus (MIC) divides through mitosis but remains transcriptionally silent, whereas the highly polyploid macronucleus (MAC) ensures gene expression. During sexual processes, the MAC is fragmented and eventually lost. Subsequent divisions of the zygotic nucleus produce the new MICs and MACs of the next sexual generation (Figure 2A). During the MAC development, the germline genome is amplified to a final ploidy of ∼800 n. Concomitantly, massive genome rearrangements occur [46]: i) repeated sequences, including transposons or minisatellites, are eliminated in a heterogeneous manner and ii) at least 45,000 short, non-coding intervening sequences, the IESs (Internal Eliminated Sequences), are excised [47] (Figure 2B). IESs excision generates one chromosomal DSB every 1–2 kb within a defined time window [48]. Thus, because endoduplication occurs during rearrangements, an estimated 106 DSBs must be repaired in each developing MAC [49]. Despite this huge number, DSB repair preserves the linear organization of the MAC chromosomes. The highly precise repair of the IES excision sites occurs through the C-NHEJ pathway, as evidenced by the absolute requirement for ligase IV and Xrcc4 [50], but requires limited processing of DSBs (Figure 2C). Because 47% of the genes are interrupted by at least one IES in the MIC [47], the precision of end-joining is essential for the recovery of functional genes in the new MAC and, therefore, for cell survival.

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Figure 2. End-joining accuracy of ligation-compatible ends.

A) The Paramecium sexual cycle. Two types of sexual processes are induced through starvation in Paramecium: autogamy, a self-fertilization process (shown in the figure), and conjugation between compatible mating types (not shown). During autogamy, the two germline diploid MICs (red) undergo meiosis to generate eight haploid nuclei (pink), and a single nucleus migrates to a specialized cell compartment, dividing once to produce two identical gametic nuclei. The remaining seven meiotic products are degraded, and the old MAC (black) becomes fragmented. During karyogamy, the two gametic nuclei fuse to form a diploid zygotic nucleus. The zygotic nucleus subsequently undergoes two successive mitotic divisions; after the second division, the two nuclei become the new MICs of the sexual progeny (red), whereas the other two differentiate into new developing MACs (red and gray) and undergo programmed genomic rearrangements. At the first cell division, the new MICs divide mitotically, and each of the two developing new MACs segregates into a daughter cell where it continues to amplify the rearranged genome to a final ploidy of ∼800 n. During conjugation, MIC meiosis is triggered through the mating of two compatible sexual partners, which undergo a reciprocal exchange of their haploid gametic nuclei. Consequently, the zygotic nucleus in each partner is formed through the fusion of a resident and a migratory haploid nucleus. Exconjugants separate between the first and second divisions of the zygotic nucleus, and MAC development occurs as described for autogamous cells. B) General structure of MIC and MAC chromosomes. On the MIC chromosomes, genes (black boxes) and non-coding regions (thin lines) are interrupted by short internal eliminated sequences (IESs in red). Repeated germline sequences (e.g., transposons and minisatellites) are indicated with a yellow double-headed arrow. During MAC development, each MIC chromosome is amplified ∼400-fold to generate a population of heterogeneous MAC chromosomes. The imprecise elimination of repeated DNA is associated with the following alternative rearrangements: i) chromosome fragmentation and telomere addition to new MAC chromosome ends (gray squares) and ii) imprecise joining of the two chromosome arms that flank the eliminated germline region. C) Mechanism of IES excision. The successive DNA intermediates formed during IES excision are displayed, with IESs shown in red and flanking MAC-destined DNA shown in black. The first step of the reaction is the introduction of 4-base staggered double-strand breaks at each IES end, depending upon the PiggyMac domesticated transposase. The molecular steps that lead to the repair of the chromosomal junction are shown on the left, which might occur within a paired-end intermediate through the annealing of the central TAs within each 5′ overhang. The removal of the 5′-terminal nucleotide was demonstrated in vivo (dotted arrow), but the nuclease(s) involved has not been identified. For the 3′-processing step, ligase IV (Lig4) recruits or activates a gap-filling DNA polymerase, which adds one nucleotide to the recessive 3′-end prior to the final ligation. A similar mechanism has been proposed for the circularization of excised linear IES molecules (right), provided that these molecules are sufficiently long. IES circles do not replicate and are actively degraded. D) End-joining of fully versus non–fully complementary ends. 1) I-SceI sites in direct orientation (arrows). The cleavage generates 3′-overhangs (red nt), which are fully complementary. C-NHEJ promotes accurate ligation (left panel), and A-EJ deletes the four protruding nucleotides, leading to the deletion of at least 4 bp at the resealed junction (right panel) [1], [2], [11]. 2) I-SceI sites in an inverted orientation (arrows). The cleavage generates 3′overhangs (red nt), which are not fully complementary. Similarly, A-EJ deletes the 3′-protruding nt, resulting in the deletion of at least 4 bp at the resealed junction (left panel). C-NHEJ anneals two of the four protruding nt (red nt), according to three classes of events (right panel). This imperfect annealing generates gaps (in blue in class I), mismatches (in blue in classes I and II), or 3′-single-stranded tails (in blue in class III) [1], [2], [11].

https://doi.org/10.1371/journal.pgen.1004086.g002

In mammalian cells, different studies analyzing the end-joining of plasmids that are cleaved by restriction endonucleases in either acellular extracts or in transfected cells have all concluded that NHEJ is accurate [6], [51]–[53]. Defects in any C-NHEJ component resulted in error-prone end-joining [6], [52], [53], corroborating that mutagenic end-joining results, at least in part, from A-EJ. However, in these studies, the DSB repair was not monitored in a chromosomal context. Therefore, different systems, based on the use of intrachromosomal substrates containing cleavage sites for the meganuclease I-SceI, have been studied. Notably, these experiments facilitated the characterization of A-EJ at a precise molecular level in the context of chromosomes in living cells [1], [2], [11], [21], [23], [54], [55]. The conclusions drawn from these studies (see below) were confirmed in vivo in mice in the context of physiological processes, such as class switch or V(D)J recombination [5], [10].

Because the I-SceI cleavage site is not palindromic, the use of two cis sites (Figure 2D.1) has been informative [2]; with the sites in direct orientation, the I-SceI–mediated cleavage generates two fully complementary ends that can be readily ligated, whereas in the inverted orientation, only partially complementary ends are generated (Figure 2B). Note that in these cases, the DNA ends are not chemically modified and, thus, are competent for the ligation machinery; the differences arise from the annealing of the four protruding nucleotides, which are fully complementary or not. In the latter case, end-joining cannot restore the initial sequence, generating an apparently mutagenic event. Notably, the efficiency of the joining of imperfectly complementary ends is similar to that of fully complementary ends, underlying the adaptation capabilities of NHEJ [2].

With fully complementary ends (I-SceI sites in direct orientation), an error-free event restores one I-SceI cleavage site. Thus, with these substrates, the frequency of error-free end-joining is underestimated because residual I-SceI protein can re-cleave the repaired junction, increasing the possibility for error-prone repair and introducing a bias in favor of inaccurate repair. The frequency of error-free events in wild-type cells consistently varies from 35% to 75%, according to the level of I-SceI expression and the half-life of the expressed I-SceI protein [1], [2], [11]. Nevertheless, the high frequency of error-free events (up to 75%, which is likely underestimated) shows that C-NHEJ should not be primarily error-prone in mammalian cells. Deficiencies in Ku80 or Xrcc4 abolish error-free events, showing that accurate end-joining events result from C-NHEJ. The remaining end-joining events (i.e., A-EJ) correspond to deletions at the junctions, with the frequent use of microhomologies distant from the DSB site [1], [2], [11]. Mutagenic events exhibit a similar signature in wild-type cells, suggesting that they result from A-EJ and, thus, that C-NHEJ is not responsible for error-prone DSB repair.

With non–fully complementary ends, end-joining cannot restore a cleavable I-SceI site; therefore, this substrate monitors a single cleavage/joining event. Interestingly, 90–95% of the end-joining events involve 3′-protruding nucleotides that are generated by I-SceI cleavage (Figure 2D.2) [2]. Strikingly, although the four 3′-protruding nucleotides at each end are not complementary, the annealing of two out of the four 3′-protruding nucleotides is observed, corresponding to the maximum possible complementarities. Thus, C-NHEJ adapts to imperfectly complementary ends with minimal genetic modifications. A systematic in vitro analysis of most of the DNA end possibilities in human cell extracts consistently yielded similar results [6]. Importantly, in Ku80- or Xrcc4-deficient mammalian cells, the repair events involve none of the 3′-protruding nucleotides, all the resulting products exhibiting deletions at the repair junctions with the frequent use of microhomologies that are distant from the DSB.

These data can be summarized as follows: regardless of the structure of the DNA ends (fully complementary or not), A-EJ removes at least all of the 3′-protruding nucleotides (and generally more), whereas C-NHEJ retains at least one of the 3′-protruding nucleotides, therefore accounting for 90–95% of the events using the 3′-protruding ends.

Importantly, these analyses have revealed that there are two different types of microhomologies (MHs): 1) MHs at the DSB itself that guide the annealing process of imperfectly complementary ends; end-joining is then processed by C-NHEJ in a conservative manner; and 2) MHs distant from the break that are involved in A-EJ, generating extended deletions at the repair junctions (Figure 1B).

These combined data show that C-NHEJ is not error-prone per se but is rather versatile and capable of adapting to non–fully complementary ends, maximizing the annealing process of potentially complementary nucleotides, which, in turn, limits genetic alterations. Thus, at the repair junction, C-NHEJ is conservative and the precision of end-joining is dictated by the structure of the DNA ends.

In Saccharomyces cerevisiae, sequence analysis of the end-joining events on transfected linearized plasmids revealed that NHEJ is very accurate. In contrast, extended deletions are recovered in yku70 mutant strains [19], [56], [57]. An alternative end-joining pathway (MMEJ), which increases upon Ku loss, has also been described in a chromosomal context [7]. In addition, NHEJ can generate reciprocal translocations, but in the absence of yKu80, the breakpoint junctions are associated with deletions [58]. An alternative end-joining pathway has also been identified in fission yeast [59].

The continuous expression of endonucleases, such as HO or I-SceI, consistently leads to multiple cycles of cleavage/repair in an essential chromosome, resulting in only 0.1% survival. This result suggests that NHEJ is at least 99.9% error-free because it restores a re-cleavable site [60]–[62]. NHEJ is also adaptable in S. cerevisiae. Indeed, the large majority of ends generated by HO are repaired by events involving the four 3′-protruding nucleotides, and the ligation of imperfect overhangs acts in a Ku-dependent manner [62], [63]. Finally, Tdp1, a yeast DNA 3′-phosphatase, has been proposed to increase the accuracy of the NHEJ machinery by preventing the modification of DNA ends [64].

C-NHEJ and A-EJ have also been described in bacteria [65]. In Mycobacterium smegmatis, the vast majority of Ku-independent junctions harbor microhomology-mediated deletions, indicating that A-EJ substituted for C-NHEJ during DSB repair [66]. Ku and ligase D are absent in the classical bacterial model Escherichia coli, and A-EJ is the most active end-joining pathway in this species [67]. Finally, evidence for conservative C-NHEJ and mutagenic A-EJ pathways has also been presented in plants [12], [68]–[70].

A large number of class II transposons transpose through a cut-and-paste mechanism in which the transposon is excised from its donor site and integrates into another locus where a target site duplication (TSD) is generated on both sides of the newly integrated element. Transposon excision leaves a DSB at the donor site with one copy of the initial TSD at each broken end; DSB repair through end-joining generally yields a characteristic footprint in which the two TSDs are separated by a few bp from the transposon (reviewed in [71]). The excision of cut-and-paste transposons, such as Sleeping Beauty [72], [73], Mos1 [74], or the P element [75], has been used in different hosts to induce DSBs at defined genomic loci. These studies have revealed that, in the absence of Ku, large deletions of the flanking sequences are recovered at transposon excision sites. This result confirms that Ku-dependent C-NHEJ is a conservative but versatile repair pathway in mammals, C. elegans, and, to a certain extent, Drosophila. In the latter, however, a chromosomal assay indicated that the most active end-joining pathway is independent of ligase IV [35].