Repair of DNA breaks by break-induced replication (BIR)

Abstract

Double-strand DNA breaks (DSB) are the most lethal type of DNA damage, making DSB repair critical for cell survival. However, some DSB repair pathways are mutagenic and promote genome rearrangements, leading to genome destabilization. One such pathway is break-induced replication, which repairs primarily one-ended DSBs, similar to those formed by collapsed replication forks or by telomere erosion. BIR is initiated by invasion of a broken DNA end into a homologous template, synthesizes new DNA within the context of a migrating bubble, and is associated with conservative inheritance of new genetic material. This mode of synthesis is responsible for a high level of genetic instability associated with BIR. Eukaryotic BIR was initially investigated in yeast, but now it is also actively studied in mammalian systems by employing reporters, and by investigating alternative lengthening of telomeres. Additionally, a significant breakthrough has been made in our understanding of the role of microhomology-mediated BIR in the formation of complex genomic rearrangements that underly various human pathologies, including neurological diseases and cancer.

1.0. Introduction.

Over the course of everyday life, human cells are confronted with instances of DNA damage, potentially on the order of 10⁵ instances per cell per day (1). DNA double strand breaks (DSBs) represent the most dangerous type of DNA damage where both strands of a DNA molecule are broken at the same place. Repair of DSBs is essential for genome stability and cell viability and can occur by several mechanisms, which are divided in two types: non-homologous end joining (NHEJ) and homologous recombination (HR) (reviewed in (2, 3)). Usually, HR completes error-free repair of the lesion; however, sometimes HR can result in mutations and chromosomal rearrangements. Some HR pathways are especially error-prone and frequently lead to genome destabilization that can promote human diseases, including cancer. One such mechanism is break-induced replication (BIR), which repairs so-called “one-ended” DSBs, and is associated with a high frequency of mutations and chromosomal rearrangements.

1.1. DNA damage leading to DSBs.

DSBs can be introduced into DNA by exposure to various endogenous and exogenous DNA damaging agents. As examples, DSBs can be formed directly (Figure 1a) via transposon excision, ionizing radiation, site-specific endonucleases, or as a result of topoisomerase failures (see (4) for more details). Alternatively, multiple events during DNA replication can result in the formation of a DSB, including collapse of the replication fork upon encountering lesions (Figure 1b_i) (e.g., inter- or intra-strand DNA crosslinks, protein-DNA crosslinks, bulky adducts, etc. (see (4) for more details)) or copying through a ssDNA nick (Figure 1b_ii) resulting from excision of a DNA lesion during mismatch repair (MMR) (5), base excision repair (BER) (6), or nucleotide excision repair (NER) (7). Problems that arise within replication often produce one-ended DSBs, where only one side of the break is available as a substrate for HR (8, 9), which can become two-ended upon the convergence with a replication fork from the opposite side. Telomere shortening due to the end-replication problem is another important source of one-ended DSBs (Figure 1b_iii), ultimately leading cells to interpret the unprotected chromosome end as a DSB (reviewed in (10)).

Figure 1. DSB repair formation and repair pathways.

(a) Formation of a two-ended DSB. The lightning bolt symbolizes the effect of site-specific endonucleases, gamma-rays, or other type of damage producing DSB. (b) Formation of one-ended DSB via one of three processes: (i) replication fork stalling followed by resolution; (ii) replication through a ssDNA nick; (iii) telomere erosion leading to formation of critically short telomere. (c) Pathways of two-ended DSB repair. (i) NHEJ leading to ligation of two broken ends; (ii) MMEJ initiated by end resection followed by annealing at microhomology, removal of flaps, filling of the gaps and re-ligation. (iii) SSA, initiated by 5’-to-3’ resection of broken ends followed by annealing between homologous regions, removal of 3’-flaps, filling of gaps and re-ligation. (iv) and GC, which can result either from SDSA (left) or dHJ (right). During SDSA one end of the DSB invades a homologous template and initiates copying from the donor, followed by unwinding of the newly synthesized strand from its template, and annealing to the opposite broken end, where it serves as a template for the second strand. dHJ is initiated similarly to SDSA, but D-loop is “captured” by annealing to second broken DNA, leading to formation of a Holiday Junction, which is resolved to produce crossover or non-crossover outcome. (d) One- ended breaks are repaired by BIR proceeding by strand invasion into homologous donor followed initiation of bubble-migration DNA synthesis that can proceed until the end of the chromosome.

1.2. Mechanisms of DSB repair.

The choice of DSB repair pathway is influenced by various factors, including the cell and organism type, the cell cycle stage, as well as the structure and location of the DSB (2, 3). During classical NHEJ (Figure 1c_i), the two broken ends are brought together and re-ligated (2, 11, 12). When one or more of the proteins in these complexes cannot participate, an alternative end-joining repair mechanism, microhomology-mediated end joining (MMEJ) (Figure 1c_ii), can be utilized (13). MMEJ is initiated by end resection that exposes microhomologous sequences as short as 2 base pairs (bps) or up to 20 bps, on both sides of the DSB, which anneal with each other followed by ligation. The remaining single-strand DNA gaps are then filled in by DNA synthesis to complete repair of the DSB. If larger regions of homology are exposed following resection of the DSB, repair can proceed via single-strand annealing (SSA) (Figure 1c_iii), in which Rad52 anneals the complementary homologous regions, followed by gap DNA synthesis and ligation. Both types of NHEJ (classical and MMEJ) as well as SSA (which is often classified as homologous recombination (14)) lead to the loss or alteration of genetic information, which can be deleterious to the cell.

HR mechanisms employ homologous DNA sequences located either on a sister chromatid or homologous chromosome as a template (2, 3, 15). HR can repair both two-ended DSBs, as well as one-ended DSBs, with the two different types of lesions requiring different repair pathways. HR mechanisms for the repair of two-ended DSBs include synthesis-dependent strand annealing (SDSA) (Figure 1c_iv) and double Holliday junction (dHJ) (Figure 1c_iv) pathways, which both lead to gene conversion (GC) (Figure 1c_iv) (3, 14). SDSA is initiated by resection of DSB ends and subsequent Rad51-mediated strand invasion that then proceeds with copying of a donor (Figure 1c_iv), dissociation of the newly synthesized DNA strand from its donor template and annealing to the opposite end of the DSB. The dHJ pathway proceeds similarly but, following Rad51-mediated DNA strand invasion producing D-loop and some synthesis, the opposite end of the break is captured by the D-loop (Figure 1c_iv), which leads to the formation of a double Holliday junction that is subsequently resolved, and depending on the orientation of resolution, can lead to crossover or non-crossover outcomes. SDSA and dHJ repair outcomes are structurally similar and hard to differentiate, and they are collectively referred to as GC. Because both GC pathways prevent the loss of genetic material, they are often termed “error-free” DSB repair.

BIR repairs DSBs that contain only one broken end capable of finding homology in the genome for strand invasion (Figure 1d). During BIR, the single invading end is used as a primer for templated repair synthesis that may continue through the telomere to effectively reconstitute the lost portion of the chromosome, reviewed in (16–19). BIR repair is essential to deal with the frequently occurring one-ended lesions but use of this pathway comes at a cost, as BIR is associated with a high frequency of mutations and gross chromosomal rearrangements (GCRs). In this review, we discuss the mechanism of BIR and genomic instabilities associated with this repair pathway.

1.3. Experimental systems employed to study BIR

BIR was first described in bacteriophage T4, where it was called recombination-dependent replication (RDR) (20–23). Following infection, early replication of the T4 bacteriophage genome is initiated, but is inactivated shortly after it begins. Late replication proceeds via RDR, where a single end formed by an interrupted early replication invades another phage genome to initiate synthesis and replicate the genome.

An RDR mechanism also plays an important role in the restart of collapsed replication forks and DSB repair in E. coli (24–28). Characterization of RDR in bacteriophage T4 and E. coli, which has been pivotal to our understanding of BIR, has been summarized in several previous reviews (23, 28–31) and will be not described here. This review will primarily focus on BIR in eukaryotes, with an emphasis on recent progress that has been made in the analysis of BIR in yeast and mammals.

The first investigations of eukaryotic BIR were conducted in yeast, Saccharomyces cerevisiae, where several important yeast experimental systems have contributed to our understanding of BIR. BIR was observed following transformation of yeast cells with linear DNA fragments bearing a telomere-seeding sequence on one end and sequence homology to a genome location on the other end (32). The latter sequence initiates BIR by strand invasion into the homologous sequence in the yeast genome to prime synthesis that can proceed for hundreds of kilobases through the opposite telomere to stabilize the fragment, thereby converting it to a miniature chromosome. A second approach to study BIR in yeast utilizes the HO endonuclease recognition site (33–36), normally used by yeast cells to initiate mating-type switching (reviewed in (37)), to induce a site specific DSB. Here, HO endonuclease, under the control of a galactose-inducible promoter, promotes a DSB such that only one broken end can find homology in the genome to initiate HR, resulting in highly efficient BIR repair for these lesions. These systems have enabled both genetic and physical monitoring of BIR, which led to the discovery of Rad51-dependent (36, 38) and Rad51-independent BIR (34). Additionally, these approaches have allowed studies of DNA synthesis during BIR (35, 39, 40), as well as molecular characterization of various genome-destabilizing events associated with BIR (41–46). More recent BIR assays have employed a site-specific ssDNA nick that is converted into a one-ended DSB after the passage of a replication fork (47–50). These systems closely mimic the type of damage inflicted at a stalled, or collapsed, replication fork, thus yielding critical insights into the repair of these common genomic lesions. In addition, it has been demonstrated that replication stalling and/or collapse at positions of unusual DNA structures, including R-loops (three-stranded hybrid DNA-RNA structures), cruciforms, hairpins or foldback structures formed by inverted DNA repeats, hairpins formed by (CAG)_n trinucleotide repeats, and triplex structures formed by (GAA)_n trigger initiation of BIR (51–55). Studies of the repair of eroded telomeres through ALT in yeast have also contributed to our understanding of BIR (56–59). The telomeres erode following inactivation of telomerase until they reach a minimum telomere length that initiates BIR using another longer telomere as a template. In this context, BIR supports maintenance of telomere length and allows cells to continue to divide, even without telomerase.

Studies in yeast have served as the basis of our understanding of the mechanism of BIR, and our learnings in these models have been directly applied to help the entire field study one-ended DSB repair and telomere maintenance in “higher” eukaryotic organisms. One such study employed egg extracts from frog, Xenopus laevis, containing all cellular components to support DNA repair and replication, and protein dependencies can be established by removing individual proteins using antibodies. Studies in X. laevis egg extracts have been used to study replication restart after a one-ended DSB formation during replication (60). Specifically, DNA replication was challenged using aphidicolin to form ssDNA that was subsequently cut using S1 or mung bean ssDNA-specific nucleases to form a one-ended break. It was determined that repair of the one-ended break involved reestablishment of the replication fork via a mechanism that resembled yeast Rad51-dependent BIR in its genetic requirements.

In human and other mammalian cells, BIR has been studied in reporter-based systems containing various split-gene cassettes (e.g., split EGFP cassette). These systems induce a DSB in the DNA using a site-specific endonuclease (usually I-SceI (61, 62) or CRISPR/Cas9 (62)) or a similar tool, which introduce a site-specific nick that is converted to a one-ended DSB by replication (62). Also, binding of the bacterial replication termination protein Tus/Ter at a recognition site introduced into the mammalian genome can initiate replication fork stalling and collapse (63, 64) that will initiate BIR-like synthesis. Another system that has yielded important knowledge about BIR in mammals investigates ALT by inducing an ALT-like process at telomeres using Fok1 nuclease (65, 66) fused to telomere-bound protein TRF1 to cut within telomere sequences and induce break repair. Finally, investigation of replication restart at common fragile sites where replication is frequently interrupted led to the discovery of MIDAS (67, 68), a BIR-like repair DNA synthesis that takes place during mitosis. Together, all of these systems have informed the mechanism of BIR in many different organisms, which we will describe in detail in this review.

2.0. The mechanism of BIR in yeast.

2.1. The mechanism of BIR progression.

The mechanism of BIR has been studied in yeast S. cerevisiae using experimental systems similar to other yeast HR systems that rely on time and site-specific DSB induction using HO endonuclease (33–36, 40). To favor repair using BIR, the HO cut sequence is placed such that only one broken DSB end can find homology for strand invasion, which could occur either on the same chromosome as the DSB or elsewhere within the genome. Similar to other HR pathways, BIR begins with 5’-to-3’ resection of the broken DNA end to form 3’single-stranded DNA (ssDNA) (41) (Figure 2a) that is then bound by replication protein A (RPA) (69), which is critical for the protection of persistent ssDNA regions that often become very long. Next, RPA is displaced by Rad52 and forms a Rad51 nucleoprotein filament that carries out the homology search and initiates strand invasion (36, 38, 70). This invaded filament serves as the substrate to prime DNA synthesis that can continue for hundreds of kilobases through the end of the chromosome. DNA synthesis during BIR is mechanistically distinct from S-phase synthesis and is carried out by a migrating bubble (39, 40), in which the newly synthesized leading strand is constantly dissociated from its template and displaced behind the migrating bubble (39). To complete BIR, lagging-strand synthesis occurs asynchronously (i.e., delayed relative to synthesis of the leading strand) and uses the newly synthesized leading strand as its template. Due to this unique mechanism, BIR results in conservative inheritance of the newly synthesized DNA(35, 39) (Figure 2a).

Figure 2. Progression of BIR and its associated instabilities.

(a) The progression of BIR. BIR begins with 5’ to 3’ resection of the one-ended DSB exposing a 3’ ssDNA end that invades at sequence homology to form a D-loop. BIR synthesis is primed by the invaded 3’-end and synthesis progresses via a migrating bubble through the end of the chromosome, resulting in conservative inheritance. (b) Genetic Instabilities associated with BIR. (i) frameshift mutations resulting from replication slippage inside the bubble; (ii) mutations and mutation clusters (red stars) resulting from DNA damage of ssDNA accumulated behind BIR bubble; (iii) translocations resulting from ectopic invasion; (iv) half-crossovers resulting from resolution of BIR intermediate.

While the main features of the BIR mechanism have been elucidated, our understanding of some of the details of BIR initiation and progression are still evolving. For example, a PCR assay to detect strand invasion and downstream BIR synthesis indicated that initiation of BIR synthesis lagged behind successful interaction of the nucleoprotein filament with the homology site by between 3 and 4 hours (36). Further, assessment of the completion of BIR by using contour-clamped homogeneous electric field (CHEF) gel electrophoresis suggested that, after BIR synthesis starts, it proceeds very quickly to the end of the chromosome, at a rate that approximates the rate of S-phase replication (36). This observation was confirmed by several research groups using various experimental systems (42, 58), until Donnianni and Symington (35) obtained results suggesting that the PCR method used to detect the beginning of BIR synthesis could detect newly synthesized DNA only after it becomes double-stranded, but fails to detect the nascent ssDNA formed by BIR. Later, the development of new methods by the Heyer lab enabled detection of nascent ssDNA, which demonstrated that BIR synthesis could begin as early as one hour after strand invasion (71). Also, recently the Malkova lab used droplet-digital PCR to compare the kinetics of BIR synthesis at different positions along the template, which determined that BIR synthesis proceeds at a rate of 0.5 kb/minute, 6 times slower than S-phase replication (72). Even with this new information, there have been some experimental observations regarding BIR kinetics that are still difficult to interpret. For example, it is unclear why, using the original PCR method, initiation of BIR synthesis has been observed by Haber lab much earlier in sgs1Δ mph1Δ mutant cells (73) as compared to wild type cells. This may be explained by the lack of disassembly of extended D-loops in sgs1Δ mph1Δ mutant cells allowing earlier conversion of ssDNA to newly synthesized dsDNA. This could also explain why template switching events that are frequently observed at the beginning of BIR (46, 74) are reduced in mph1Δ mutants (75), and also the shift of the competition between GC and BIR in the direction of BIR in mph1Δ mutants (76). Another important insight into the priming of BIR synthesis came from the study performed by Haber lab, where BIR was initiated by strand invasion within 108bp of homology and the donor carried different densities of single-base-pair mismatches. It was observed that even at high density of mismatched bases BIR was still efficient and that nearly all mismatch correction depended on proofreading activity of polymerase δ (70).

The identity of the polymerases that drive BIR has been of continued interest in the field. It has been proposed that all three replicative polymerases, δ, α, and ε, are involved in BIR synthesis (58), but their exact roles remain incompletely understood. The involvement of polymerase δ (Polδ) is supported by findings of defective BIR in several POL3 mutants (42, 58, 77), an effect that is especially pronounced in the absence of Pol32 (42, 58, 77), a Polδ subunit that mediates displacement DNA synthesis and is dispensable for S-phase DNA synthesis (78). Because displacement synthesis during BIR would occur within the migrating D-loop of the leading strand, it was proposed that Polδ may drive leading-strand synthesis. Recently, experiments performed by the Symington and Kunkel labs (79) employed mutant forms of replicative polymerases with increased propensity for incorporating ribonucleotides into newly synthesized DNA, which led to the finding that both leading and lagging strands are synthesized by Polδ during BIR. In addition, they did not find any evidence of participation by polymerase ε (Polε) during BIR synthesis, which contrasts with an earlier model proposed by Lydeard et al (58). Their data suggested that Polε was dispensable for BIR initiation but is required for copying the template further downstream of the break. It is possible that these conflicting conclusions are influenced by differences in the strain backgrounds or by the different lengths of synthesis completed by BIR in the two systems. The role of the polymerase α (Polα)-primase complex in BIR synthesis is also not clear. While there is wide agreement that this complex is required to synthesize the lagging strand (58, 79), Lydeard et al. (58) proposed that initiation of leading-strand synthesis may also requires Polα-primase, even though it would be expected that the 3’-ssDNA invaded strand could serve as a primer. Recent experiments performed by the Malkova lab (72) suggests that leading-strand synthesis can successfully initiate in the absence of primase; however, it can proceed for no more than approximately 20–25 kb without accompanying lagging strand synthesis. This suggests that the requirement for primase in leading strand synthesis may not be to initiate or extend the leading strand directly, but its role in lagging-strand synthesis is required to stabilize the ssDNA produced during leading strand synthesis when it proceeds for longer distances.

Finally, some questions remain regarding the identity of helicases driving BIR. It is established that Pif1 helicase, which is not essential for S-phase replication, is required for BIR (39, 40). Pif1 has been proposed to play two different roles: unwinding the DNA duplex to allow the progression of BIR and removing the newly synthesized DNA from the D-loop. Both of these functions are required for forward migration of the D-loop, but the actual function of Pif1 in BIR remains poorly understood. Additionally, whether MCM2–7, the main S-phase replicative helicase, has any role in BIR remains unclear. Lydeard et al. (80) observed that MCM2–7 is required for BIR, but Wilson et al. 2013 (40) did not observe any defect when MCM2–7 activity was eliminated. Finally, the helicase Srs2 is required to prevent accumulation of Rad51-mediated toxic joint molecules during BIR (81) by preventing the formation of long Rad51 filaments that subsequently re-invade behind the D-loop, as well as by removing joint molecules after they form.

2.2. Instabilities associated with BIR.

Repair by HR has been regarded as an error-free mechanism, but BIR is inherently error prone. BIR has been associated with a high mutation rate, gross chromosomal rearrangements (GCRs), and loss of heterozygosity, all of which can lead to deleterious effects (43, 45, 46, 74, 77, 82). The high level of genetic instabilities associated with BIR was first reported by the Symington lab (74), where the authors observed a very high level of template switching associated with ~ 10kb at the beginning of BIR synthesis. A similar phenomenon was also observed by (46) in a different experimental system.

The unusual mechanism of BIR synthesis underlies its increased frequency of mutations. The frequent dissociation and reconnection of the nascent DNA with the template during BIR leads to mistakes in pairing that are then translated into frameshift mutations (82) (Figure 2b_i). BIR-associated mutagenesis is further elevated by inefficient mismatch repair (MMR) (82), as mismatches in this context are short lived due to the conservative mode of BIR synthesis. Additionally, the asynchronous synthesis of leading and lagging strands during BIR results in large tracts of ssDNA that can accumulate both endogenous (e.g., reactive oxygen species) and exogenous (e.g., alkylating agents) DNA damage (Figure 2b_ii), as these lesions cannot be easily repaired in the absence of the complementary strand and, ultimately, led to base substitutions (39, 83). Also, it has been demonstrated that the presence of the alkylating agent, methyl methanesulfonate (MMS), induces mutation clusters (Figure 2b_ii) specifically in DNA synthesized by BIR (43), similar to those formed in cancer cells where persistent ssDNA is exposed to APOBEC enzymes (cytidine deaminases) (reviewed in (84). Importantly, when APOBEC3A was expressed in cells undergoing BIR repair, it led to the formation of mutation clusters, the length and number of mutations of which was significantly increased in an ung1Δ background in which excision of uridines resulting from APOBEC-induced deamination was lacking (83). Thus, it was proposed that ssDNA created by BIR serves as an ideal substrate for APOBEC to create mutation clusters and that APOBEC-induced mutation clusters observed in cancer cells might also result from BIR. Indeed, recent studies by the Gordenin lab (85) have demonstrated that a significant fraction of mutation clusters in humans likely results from BIR. Finally, BIR proved to be also highly mutagenic when instead of a site-specific DSB it was induced in yeast by the DNA nick generated in regions of the genome devoid of replication origins and within a MUS81-null background (49).

BIR has also been linked with GCRs. When BIR repair uses homology in a sister chromatid or a homologous chromosome, completed repair does not create GCRs (Figure 2a). However, if the homology utilized is within an ectopic location (Figure 2b_iii), BIR will then copy a sequence that is dissimilar from the genetic material present before the break (Figure 2b_iii), leading to the formation of a GCRs (33, 42–46, 77). GCRs can be compounded when BIR synthesis is disrupted, leading to the resolution of the D-loop forming what is called a half crossover (42, 77) (Figure 2b_iv), in which the recipient chromosome is joined to the donor chromosome at the point of resolution. This leaves a newly formed one-ended DSB on the donor chromosome (Figure 2b_iv), which can subsequently invade at sequence homology to initiate a new BIR event. Depending on the factors that resulted in the first half crossover, the cell may be prone to the formation of additional half crossovers until the end is stabilized, altogether a highly genome-destabilizing process referred to as a half crossover cascade (43, 45). Even when a homologous chromosome is used as a template, heterozygosity between the recipient and donor chromosomes will be eliminated as BIR duplicates the template (Figure 2b_v). The consequences of this are most severe when a heterozygous donor template contains a deleterious mutation that was compensated by a partially or completely dominant wild type allele in the lost recipient molecule. In sum, although BIR is an important pathway to stabilize broken chromosomes, the repaired chromosomes contain a high frequency of mutations, and the repair process itself can trigger larger-scale genomic destabilization, making BIR a highly error-prone pathway.

2.3. Rad51-independent break-induced replication

The discussion above is specific to the mechanism of canonical Rad51-dependent BIR. However, a less efficient, but well documented, BIR mechanism does not require Rad51(34). Like canonical BIR, Rad51-independent BIR also requires the function of Rad52 (34, 86, 87), making it an HR mechanism, but differences in their genetic requirements have been identified (34, 88). For example, Rad51-independent BIR requires Rad59, Rdh54, and the MRX complex (Mre11, Rad50, and Xrs2) (88). As Rad59 aides Rad52 in the annealing of ssDNA (89, 90), it was proposed that Rad51-independent BIR proceeds by an annealing mechanism after extensive resection. Interestingly, Rad51-independent BIR promotes interaction, and potential annealing, with highly repetitive regions within Ty or delta elements (86, 87, 91), but the details of these interactions (e.g., the amount of homology or microhomology used) are unknown. What is known is that Rad51-independent BIR, in a majority of cases, leads to genomic instability, mainly due to the formation of GCRs (34, 86, 87). Studies of Rad51-independent BIR have been limited by the lack of convenient experimental systems. However, interest in understanding the mechanism of Rad51-independent BIR in yeast is growing, likely due to the recent discovery of Rad51-independent BIR-like events in mammals (65, 92).

3.0. The mechanism of BIR in mammals.

The first observation of BIR in mammalian cells resulted from a screen aimed to identify genes, and define pathways, for replication fork repair (61), by overexpressing Cyclin E to induce replication stress to promote segmental duplications. Previously, in yeast, it was demonstrated that Pol32 was required for the formation of segmental duplications that are attributed to BIR (93). The screen performed in mammalian cells by Halazonetis lab (61) uncovered that deletion of either of two orthologs of yeast Pol32, POLD3 or POLD4, reduced the formation of segmental duplications, which suggested that BIR may occur during repair of replication forks in mammals. Next, the authors developed a split GFP reporter system, where only one of the two broken ends created by a site-specific I-SceI endonuclease cut could invade another, intact copy of GFP to initiate BIR-like repair. Consistent with their earlier findings, depletion of POLD3 decreased the frequency of repair in this system (61). Importantly, while they observed BIR synthesis that proceeded for distances longer than those synthesized during GC, BIR synthesis was typically interrupted after synthesizing just several kilobases of DNA before the newly synthesized strand was rejoined to the opposite end of the break, likely through MMEJ. This is an important contrast to the extended length of DNA – up to hundreds of kilobases – that can be copied during BIR in yeast, where it can continue all the way through telomeres to stabilize broken chromosomes (35, 36, 38, 58).

The Wu lab investigated BIR involving DNA repeats using a system in which I-SceI or CRISPR-Cas9 was used to create a break between the repeat sequences (62), prompting the broken end to invade the other repeat located in the same chromosome. Upon successful repair, the region between the repeats was deleted (62). Prior to these experiments, SSA was believed to be the repair mechanism employed when a DSB occurs between repetitive DNA sequences (14). However, here the authors observed that, even when the repeats are separated by a relatively short distance (just ~2 kb), BIR can compete with SSA (62). In particular, they noted that when the break occurred at the approximate center between the repeats, SSA was favored. Conversely, when the DSB occurred asymmetrically, repair tended to proceed through a POLD3-dependent BIR mechanism (62). This finding suggests that, in the time required for resection to reach the opposite repeat (~2 kb of resection), BIR can efficiently proceed through invasion of the resected end into the repeat sequence, and required the function of POLD3, RAD51, RAD52, and the ATM-mediated DNA damage response, mirroring the genetic requirements of BIR in yeast. When the distance between interacting repeats was increased to 400 kb, repair only proceeded through BIR (62). Additionally, BIR was successfully induced during S-phase when a replication fork encountered a Cas9-generated nick (ssDNA break), which was transformed into a one-ended DSB upon replication. Lastly, overexpression of Cyclin E, to induce replication stress, further increased the efficiency of BIR in the same system (where BIR involved two repeats separated by ~ 2kb and was initiated by I-SceI-mediated site-specific DSB) (62), which suggested that oncogenic stress can “prime” cells to use BIR, which in turn can promote genomic destabilization fueling tumorigenesis.

The Scully lab studied, in depth, the competition between short-tract gene conversion (STGC) and long-tract gene conversions (LTGC) ((63, 94), also reviewed in (12)) to repair a break introduced into a GFP/RFP reporter. The breakage in this reporter was induced either by I-SceI or by a site specific collapse of a replication fork at the place where bacterial Tus replication termination protein bound a Ter binding site. Both types of breakage promoted STGC and LTGC outcomes, and the ratios between the two were affected by inactivation of various DNA repair genes, suggesting distinct genetic requirements for the two repair types. For example, inactivation of BRCA1, BRCA2, CTIP, BLM, and FANCM, or paralogues of RAD51 ((63, 64, 94–96), and reviewed in (12)) lead to an increase in the frequency of LTGC. Based on their observations, the authors speculated that LTGCs represent longer repair synthesis similar to BIR. Another BIR-like process that has been recently discovered in mammalian cells is mitotic DNA synthesis (MIDAS) (67) that is initiated at common fragile sites (genome regions where DNA frequently remains un-replicated after S phase). MIDAS is RAD51-independent and is mediated by RAD52 and POLD3, and also requires the structure-specific nuclease MUS81-EME1, which is likely involved in the processing of stalled replication forks ((67, 68), and also reviewed in (92)).

In sum, a diverse body of evidence supports that BIR occurs in human and other mammalian cells. Mammalian BIR requires POLD3 and generates tracts of DNA synthesis that are shorter than those described for BIR in yeast. To more fully understand the BIR mechanism, it will be important for future studies to determine whether mammalian orthologs of other known yeast BIR proteins, such as Pif1, are important for BIR in mammals.

4.0. Microhomology mediated break-induced replication (MMBIR)

One mechanism related to BIR that is thought to mediate complex genomic rearrangements (CGR) in several disease contexts is microhomology-mediated break-induced replication (MMBIR) (97–99). CGRs include rearrangements, such as translocations that often contain several break-points, and are frequently associated with copy-number variations (CNVs). In contrast to canonical BIR, where substantial homology is required for strand invasion, MMBIR requires only microhomology (as short as 1 to 3 bp), or regions of limited homology (up to approximately 20 bp), to initiate and proceed with repair synthesis. MMBIR is also associated with frequent template switching events, where each event uses different sites of microhomology, or limited homology, to initiate synthesis. This can result in the repair site being highly variegated, containing adjacent pieces of genetic information from multiple regions of the genome that are frequently amplified.

MMBIR was first proposed by Hastings et al. (97) to explain CGRs observed by the Lupski research group in patients with Pelizaeus-Merzbacher disease (PMD) (100), and also associated with MECP2 duplication syndrome (101). Both of these human diseases are characterized by GCRs with highly complex breakpoints, microhomology at breakpoint junctions, and CNVs on the order of duplications to quadruplications (100–104). An important mechanistic insight drawn from sequencing of causal CGRs from patients with PMD or MECP2 duplication syndrome was that MMBIR may be induced by BIR (101–103). In this scenario, it was postulated that breakage of a replication fork at a chromosome fragile site would lead to BIR producing a copy number gain and subsequent MMBIR events could result in further template switching at microhomologies to produce even more complex amplifications of varying sizes. Support for this idea of BIR transitioning to MMBIR came from studies in yeast in which Anand et al. observed template switching that required less homology than the initial priming for BIR synthesis (105). A study by Sakofsky et al. also reported a switch from BIR to MMBIR in pif1Δ cells, which are deficient in long-tract BIR synthesis and prone to replication stalling (44). This latter study resulted in a more detailed model of MMBIR in which BIR stalling induced by pif1Δ mutation lead to template switching at microhomologies, followed by synthesis and additional template switching mediated by microhomology (Figure 3e).

Figure 3. A comprehensive model of MMBIR.

Molecular mechanism of MMBIR. (a) Initiation of MMBIR by fork stalling (left) or by DNA breakage leading to a DSB (right). (b) Replication fork stalling promotes template switching guided by microhomologies, which are used to prime synthesis by translesion polymerases, or other replicative polymerases. (c) Additional template switching either restarts fork progression or continues to an ectopic location to produce rearrangements. (d) DSB initiates BIR. (e) BIR switches to MMBIR by promoting template switching at microhomologies and engaging in synthesis driven by translesion or other replicative polymerases. (f) Additional template switching may restart BIR synthesis or continue to an ectopic location to produce rearrangements.

MMBIR as the mechanism responsible for CGRs observed in congenital diseases (106) offered an alternative explanation for instances of chromothripsis that had been observed in cancer (107–110). Chromothripsis, a mutational process characterized by a massive number of chromosomal rearrangements typically isolated to one or just a few chromosomes, was previously understood to be the result of re-stitching of broken chromosomal pieces by NHEJ (108, 111). However, while this mechanism could produce CGRs, it did not explain copy number increases that were sometimes associated with chromothripsis. Alternatively, the MMBIR model of chromothripsis, which proposes that DNA synthesis proceeds through many rounds of template switching, could yield the copy-number gains observed in many of the CGRs identified in cancer genomes (see for example in (112)). While early sequencing of chromothriptic cancer genomes did not identify templated insertions at breakpoints that are characteristic of MMBIR (107, 108), a recent study of chromothriptic cancer genomes selected from the Pan-Cancer Analysis of Whole Genomes (PCAWG) consortium revealed some breakpoint junctions that were associated with copy number increases and contained templated insertions, which were proposed to be mediated by MMBIR (112).

Though these and other studies have identified putative MMBIR events by characterizing the sequence of large rearrangements and CNVs, the direct identification of all MMBIR events within a genome based on their unique molecular signature - irrespective of whether the event led to a CGR - remains a challenge. Recently, Osia et al. translated the DNA signature associated with MMBIR in yeast (44) to interrogate human cancer and non-cancer genomes (113). This analysis determined that, while MMBIR events appeared only in the germline and did not mediate CGRs in normal cells, some of the MMBIR events in cancer genomes mediated CGRs. Furthermore, the pattern of some MMBIR events in cancer cells indicated that MMBIR had initiated cascades of instability, resulting in a chromothripsis-like pattern of CGRs. These conclusions are consistent with studies reported by Zhang et al. where chromothripsis was induced via formation of micronuclei (109). Analysis of the resultant chromothriptic genomes identified templated insertions consistent with MMBIR. In sum, these studies support that MMBIR is a likely mechanism by which chromothripsis can proceed in cancer, and it may be the initiator of some instances of chromothripsis.

Based on the characterization of MMBIR outcomes in a number of disease contexts (97, 98, 101–103, 109, 112, 113), a basic model for MMBIR can be proposed. The current literature supports that MMBIR is initiated by either DNA breakage, or replication stress, (Figure 3a) that leads to a free ssDNA end formed by resection of the DSB or dissociation during ongoing synthesis, respectively (Figure 3b,d,e). This free end can anneal to the exposed ssDNA, or potentially invade dsDNA, using microhomology to begin synthesis (Figure 3b,e). Additional template switching that returns to the original template, or goes elsewhere, determines whether the event resolves as a simple MMBIR insertion or creates a more complex multi-template MMBIR event (Figure 3c,f). This model acknowledges that MMBIR may initiate from an S-phase replication fork directly or from BIR synthesis initiated by a collapsed replication fork or DSB, which converges with the model proposed based on analyses of patients with PMD or MECP2 duplication syndrome (101–103), where pre-established homology-driven BIR was theorized to switch to MMBIR.

The study by Sakofsky et al. made the important molecular observation that the switch from BIR to MMBIR is governed by a switch from Polδ to Polζ, a translesion polymerase that often introduces errors during synthesis (44). This insight recognizes the possibility that MMBIR, in some contexts, may be executed by polymerases other than those driving the S-phase replication fork or the migrating D-loop that synthesizes homology-driven BIR. This idea was further supported by the analysis of MMBIR in human cells by (113), who observed that some MMBIR events were extended from mismatched 3’-ends with microhomology behind the 3’-terminal nucleotide. Such imperfect microhomologies, that were also termed “microhomeology”, were also observed by (114) who reported the presence of microhomologies containing mismatches at presumed MMBIR breakpoints. In humans, one attractive candidate polymerase that could initiate synthesis from imperfect primers is polymerase θ (polθ), a polymerase known to mediate synthesis from primers annealed at microhomologies during the process known as theta-mediated end-joining (TMEJ) (115–117). Polθ was shown to work at DSBs, preferring to anneal to templates nearest to the DSB (118), and also to promote repair at collapsed replication forks (119). Polθ is also capable of strand displacement activity (120), another important feature making the synthesis performed in TMEJ similar to that of BIR. If polθ is indeed a polymerase capable of mediating MMBIR in mammals, it prompts the hypothesis that TMEJ could play a critical role in MMBIR, either initiating or resolving MMBIR events.

Future studies of MMBIR may seek to understand the spectrum of possible MMBIR-initiating events, and what factors determine whether instances of replication stress or ongoing repair (e.g., BIR or TMEJ) transition to MMBIR. More work is also needed to understand the broader consequences of MMBIR in various contexts, including the initiation of chromothripsis. One major outstanding question is what genetic backgrounds promote a high level of MMBIR events in human cells. Particularly in the context of disease, understanding whether MMBIR is a product of dysregulated DNA repair, or whether MMBIR may occur frequently in some tissues but produce genomic changes that rarely survive selective pressures under normal circumstances, may eventually lead to important clinical insights. For example, understanding whether MMBIR is generally suppressed by RAD51 in human cells, as proposed by (97), may help to contextualize MMBIR’s role in genome instability in RAD51-deficient cancers. Also, comprehensively characterizing the MMBIR signature in cancer genomes may uncover new therapeutic concepts or new diagnostic or prognostic biomarkers.

5.0. Alternative lengthening of telomeres (ALT)

Telomeres are nucleoprotein structures on the end of each chromosome that prevent the cell from interpreting chromosome ends as DSBs. However, with each cell division, telomeres shorten due to the end-replication problem, until the telomere can no longer prevent the DNA damage response, reviewed in (121–125). Thus, in human cells, telomeres are potent tumor suppressors that limit the number of divisions a cell can undergo before the onset of senescence. To overcome the onset of senescence and resume cellular divisions, tumors must maintain their telomere lengths by one of two mechanisms, reviewed in (10, 122, 123, 125). A majority of cancers (85% - 90%) reactivate telomerase, while approximately 10% to 15% promote alternative lengthening of telomeres (ALT) (reviewed in (122, 123)). The mechanism of ALT is closely related to BIR as ALT exploits the one-ended break formed at shortened deprotected telomeres to invade a sufficiently long telomere, which is used as a template to extend the telomere through a BIR-like process. S. cerevisiae has served as an important model to study ALT for decades (59, 121, 124), where experimental systems have been designed to dissect the entire process, from telomere erosion to establishment of ALT through one-ended telomere invasion all the way to formation of ALT survivors. The discovery of ALT in mammals (126–128) has prompted investigation of ALT in mammalian systems, which has revealed new insights into ALT in this context.

5.1. Alternative lengthening of telomeres (ALT) in yeast.

ALT was first discovered in telomerase-deficient yeast “survivor” cells (59) that use a BIR-like synthesis mechanism to maintain telomeres (58) and support long-term cell division. These survivor cells have been used to model ALT ever since their discovery, as they allow the study of the entire process of survivor formation, from telomere erosion through survivor cell establishment, reviewed in (121, 124). Studies in survivor cells determined that ALT proceeds through an HR mechanism. Because survivors fail to form in the absence of both RAD52(59) and POL32(58), the primary HR and BIR genes, respectively, it is accepted that ALT proceeds through BIR. Further analysis suggested the existence of two distinct ALT pathways (Figure 4a), Type I and Type II (56, 57, 59, 129), and an extensive genetic screen for ALT genes identified over 280 genes that were potentially involved in the formation of Type I or Type II ALT (130). Type I ALT results in formation of survivors that are unstable and cycle between periods of active cell division and cell cycle arrest, reviewed in (121). Structural analysis of telomeres in Type I survivors was performed by digesting genomic DNA within the Y’ sub-telomeric sequence using XhoI (56, 57, 59, 129, 131), followed by Southern analysis using a Y’-specific probe. This identified the presence of two distinct bands correlating to the sizes of the short and long Y’ sub-telomeric sequences, which indicated that Type I survivors maintain their telomeres by adding many tandem Y’ repeats (56, 57, 59, 129, 131), while the terminal telomeres on their chromosome ends remain short (Figure 4a). The short terminal telomeres in Type I survivors explained the slow rate of cell divisions, as the short telomeres cannot sustain many divisions before the onset of another cell cycle arrest. Genetic analysis of Type I ALT was performed by identifying gene deletions that inhibit formation of Type I survivors. The fact that Type II survivors could still form in these genetic contexts further supported the model of two distinct ALT pathways (Figure 4a). Deletion of RAD51, RAD54, RAD57 and RAD55, all of which are required for DNA strand invasion at sites of homology, prevented Type I survivor outcomes, leading to the proposed model (Figure 4a), reviewed in (121), that Type I ALT proceeds via Rad51-mediated BIR to form chromosome ends that contain many tandem Y’ sequences and a short telomere end.

Figure 4. Two models of ALT in yeast.

(a) Two independent pathway model. Short telomeres resulting from telomere erosion either (left) invades into other telomeres or internal telomere sequences (ITS), which leads to formation of Type I survivors, or (right) anneals to telomere ssDNA (e.g. telomere circles) followed by rolling circle replication to form Type II survivors. The proteins required for Type I or Type II formation listed on the right. (b) Unified ALT pathway model. ALT precursors are formed by Rad51-mediated invasion of eroded telomeres into other telomeres. Rad59-mediated annealing of ssDNA in precursors to (right) telomere circles or other substrates, followed by telomere extension resulting in the formation of Type II survivors. Precursors annealing to (left) ITS sequences form Type I telomeres that are spread throughout the cell by Rad51. The Type I telomeres are stabilized by (right) annealing at telomere circles or other substrates followed by extension to form a “hybrid” telomere structure.

Type II ALT survivors are believed to occur less frequently compared to Type I, and they maintain a continuous, faster rate of cell divisions ((56, 57), reviewed also in (121)). Structural characterization of Type II survivor telomeres performed by Southern analysis, similar to the approach used to study Type I outcomes, revealed multiple bands of variable lengths in a ladder-like pattern (57, 129, 131–133). This indicated that Type II outcomes stabilize their telomeres by adding long stretches of telomeric sequence (Figure 4a), in some cases upwards of 12 kb, consistent with utilization of telomere circles (134) for rolling-circle replication. Genetic analysis identified that deletion of RAD59 (129) (encoding a protein required to facilitate Rad52-mediated ssDNA annealing), SGS1 (132, 133) (an ortholog of the human BLM helicase), and RAD50 (57, 131) (a member of the MRX complex) eliminated Type II ALT survivor formation, but Type I outcomes were still observed. Significantly, RAD51 is not required for Type II outcomes (57, 129, 131), indicating that they form via Rad51-independent BIR. Further, and consistent with interpretation of the Southern analysis of Type II outcomes, the requirement for Rad59 supports that Type II ALT proceeds via ssDNA annealing at, potentially, telomere circles, that would then be followed by rolling circle replication to form long telomeres.

ALT is a stochastic event, and formation of ALT survivors is extremely rare. In fact, for decades, it was believed that the frequency of ALT could not be determined experimentally (reviewed in (121)), and our knowledge was based on genetic analyses. Specifically, the finding that rad51Δ, rad59Δ, or rad50Δ single mutants produce survivors, whereas no survivors were observed in rad51Δ rad59Δ or rad51Δ rad50Δ double mutants (57, 129, 131), leant very strong support to the model that two genetically distinct ALT pathways yield distinct survivor outcomes (Figure 4a). Recently, Kockler et al. (135) provided the first quantitative analysis of ALT, and determined the frequency of ALT (combined Type I and Type II) to be 2×10⁻⁵ per senescent tlc1Δ cell. Their quantitative approach was further applied to genetic studies, which uncovered that survivor frequency was drastically decreased in tlc1Δ rad51Δ and tlc1Δ rad59Δ mutants by 18- and 15-fold, respectively. Strikingly, the sum of ALT frequencies from the two mutant constructs was much lower than the overall frequency of ALT in tlc1Δ cells, prompting the authors to hypothesize that Rad51 and Rad59 collaborate in a single ALT pathway (Figure 4b).

Technological advances in sequencing have made possible much more detailed structural analysis of ALT survivor chromosomes. The earlier studies using Southern blot analysis were able to determine the basic composition of telomere ends, but fine detailed structures were not able to be attributed using this method. Further, detailed analysis using whole genome sequencing methods by Illumina sequencing was not fruitful as the short reads of Illumina could not be attributed to specific chromosome ends. Kockler et al. (135) employed ultra-long Oxford nanopore sequencing technology to study individual chromosome ends from both survivor types. This revealed the presence of what the authors termed “hybrid” telomere structures (Figure 4b) that contained tandem Y’ repeats (the feature of Type I survivors) adjacent to long terminal telomeres (the feature of Type II survivors). This same approach was used to study telomeres in tlc1Δ rad51Δ and tlc1Δ rad59Δ mutants, which determined that Rad51 and Rad59 participated in early and late steps, respectively, of a unified ALT pathway, which aligned with the ALT frequencies the authors had noted in these same double mutant constructs. The methodologies used for these studies are anticipated to prompt further research that will unravel the molecular mechanism of the unified ALT pathway and factors that influence ALT survivor formation and telomere structures in yeast.

5.2. Alternative lengthening of telomeres in human cells

ALT in mammalian cells was originally discovered in cancer cell lines that maintained their telomeres without telomerase (127). Shortly thereafter, ALT was identified in human cancer tumors (126). Later, it was demonstrated that approximately 10% to 15% of tumors maintain their telomeres via ALT (reviewed in (122, 123)), predominantly in tumors of neuroepithelial or mesenchymal origin (136, 137).

Beginning with the prediction that ALT proceeds through HR, the Reddel lab placed a genetic marker on a single telomere and observed that, through ALT, this marker was transferred to multiple telomeres within the cell (138). Further, it was observed that synthesis of ALT telomeres is associated with PML bodies (APBs) in G2 cells (139, 140) and, in mitosis, at APB-like foci (141). Later, it was found that the telomeres in ALT cancer cells contained very long telomeres that more closely matched the classical Type II ALT survivors formed in yeast via Rad51-independent BIR (reviewed in (122)).

To better understand the ALT mechanism in humans, the system to track DNA synthesis during ALT was developed. Resulting studies yielded numerous breakthroughs in the field. Specifically, it was determined that after a FokI induced telomere DSBs, ALT-like synthesis is initiated that requires RAD52 as well as POLD3 and POLD4 (orthologs of yeast Pol32), and results in conservative inheritance of newly synthesized DNA (65), which all represent findings suggesting that BIR underlies ALT synthesis. Interestingly though, ALT DNA synthesis in human cells does not require RAD51 (65), which led to the conclusion that human ALT proceeds via Rad51-independent BIR (which the authors termed break-induced telomere synthesis (BITs)), a process resembling the yeast Type II ALT pathway. However, the same research lab demonstrated that a RAD51 knockout was deficient in the ability to cluster telomeres prior to synthesis a required early step in the ALT pathway (66), which highlights a potential role for Rad51-dependent BIR in human ALT. This idea is further supported by recent findings that RAD51-associating protein, RAD51AP1, is required for ALT synthesis (142), while depletion of Rad54 decreases the level of ALT synthesis (143). More research is required to fully understand the roles for both Rad51-independent and -dependent pathways to execute ALT in humans. Based on our current knowledge, it cannot be determined whether ALT in humans requires multiple pathways, or whether a unified pathway similar to what was recently described for yeast may underlie all ALT outcomes. Distinguishing between these possibilities will require new approaches for quantitative and structural studies of the specific steps of ALT in mammals, from ALT establishment through chromosome stabilization and survivor formation.

Further questions regarding the mechanism of ALT remain. First, the source of the one-ended breaks at mammalian telomeres remains obscure. The most obvious source is, simply, the erosion of telomeres following many cellular divisions in the absence of telomerase. This would allow, over time, the shortening of telomeres until shelterins are no longer able to protect the end, or form a T-loop, leading to recognition of the chromosome end as a one-ended break. Recent evidence suggests an additional source of one-ended breaks: replication fork stalling resulting in DSBs in telomeres. This possibility was raised when ASF1, a histone chaperone, was depleted (144), resulting in an increase in replication stress at telomeres in addition to inducing ALT characteristics, that include APB formation, telomere recombination, long telomeres, and telomere circle formation. The idea that replication stress may activate ALT was further supported by the findings that ALT telomere clustering and synthesis is increased upon Cyclin E overexpression and G-quadruplex introduction at telomeres in a RAD52, CHK1, and MRE11 dependent manner (145). A source for such replication stress could be R-loops that could be formed by telomere repeat containing RNA (TERRA) known to be increased in ALT cells and to induce telomere instability (146). This was supported by demonstration that depleting FANCM, known to suppress R-loops, increased replication stress as well as ALT activity and synthesis (147, 148). Therefore, an increase in replication stress, potentially via TERRA-promoted R-loops, could promote ALT, which could in turn fuel tumorigenesis in pre-cancerous cells, or tumor progression in established disease.

Another outstanding question for ALT in human cells is: what do the final sequences of ALT outcomes look like? A better understanding of telomeres from ALT survivors will be critical for our ability to more definitely characterize the ALT mechanism in humans, and this will require methodologies in mammalian systems similar to what has been recently developed in yeast. Recently, the PCAWG consortium successfully sequenced ALT cancer genomes and identified that ALT cancers are associated with the formation of novel telomere repeats (149). This important finding may allow tracing of intermediate and final ALT structures, which would be integral to understanding the mechanism of ALT. Unfortunately, the short-read nature of the most common sequencing methods made the attribution of telomeres to individual chromosome ends impossible, and extra-long sequencing methods will be required to mitigate this problem in the future.

6.0. Summary and Perspectives.

BIR remains a very interesting, important, and not yet fully understood pathway of DSB repair. The research conducted in the last several years has led to many advances in our understanding of BIR, including: (i) the characterization of BIR kinetics, mode of conservative DNA synthesis, and the identification of many participating proteins for BIR in yeast; (ii) development of several reporter systems allowing the characterization of the efficiency and processivity of BIR in mammals; (iii) understanding the ALT mechanism in yeast and humans; and (iv) characterization of multiple types of DNA damage and DNA processes triggering BIR (e.g., chromosome fragile sites, R-loops, etc); (iv) progress in understanding the mechanisms of MMBIR.

An important advancement that is expected from future studies is to more clearly define how lagging-strand BIR synthesis is carried out. For example, whether, when, and where Okazaki fragments are formed and how long they persist remain unclear. It will also be important to further our understanding of the relationship between BIR and other DSB repair pathways, including GC, NHEJ, and MMEJ. Understanding how these different repair pathways compete with one another at different stages of the cell cycle, and in different organisms, will greatly advance our understanding of DNA metabolism and repair in normal and pathological contexts. We also currently do not know whether BIR can occur in meiosis. If it does, determining whether it increases meiotic mutagenesis will be a question of critical importance. Similarly, if it is determined that BIR is associated with increased mutagenesis in mammalian cells, a more comprehensive characterization of genetic and physiological contexts that induce or suppress mammalian BIR may have implications for human disease.

Rad51-independent BIR remains poorly understood, and additional research is required to more completely elucidate how this pathway is initiated, how the synthesis proceeds, and which proteins are involved. The structural characteristics of chromosomal regions that serve as donor templates for Rad51-independent BIR are also unknown at this time. Future research in this area should aim to define the homology requirements, or lack thereof, for Rad51-independent BIR, the mechanistic role of Rad52 in this process, and which factors determine whether Rad51-dependent or -independent BIR is used for repair. The urgency of understanding the detailed molecular mechanism of Rad51-independent BIR has been underscored by multiple recent reports of this pathway being active in mammalian systems (65, 68). Likewise, many important gaps in our knowledge of MMBIR persist, although compelling data regarding its relevance in human disease are rapidly emerging. It is anticipated that clarifying the mechanism of MMBIR initiation and DNA synthesis, in particular the potential use of error-prone polymerases, and its tissue specificity in mammals will bring a new understanding to how MMBIR participates in the development of congenital diseases and cancer.

We were unable to describe in this review the pathways associated with restart of collapsed replication forks at fission yeast RTS sequences (reviewed in (150)). RTS sequences initiate a BIR-like process that proceeds by semiconservative DNA synthesis without inducing a DSB intermediate. This example demonstrates that some variants of BIR-like processes do exist with different substrates and protein dependencies as compared to BIR described in S. cerevisiae during G2/M repair of site-specific DSBs. In addition, many questions remain regarding the events that trigger ALT and the mechanism(s) used to stabilize chromosomes, as well as which factors may make a cell more or less likely to use ALT versus reactivating telomerase or entering a senescent state.