Collaborations between chromatin and nuclear architecture to optimize DNA repair fidelity

Abstract

Homologous recombination (HR), considered the highest fidelity DNA double-strand break (DSB) repair pathway that a cell possesses, is capable of repairing multiple DSBs without altering genetic information. However, in “last resort” scenarios, HR can be directed to low fidelity subpathways which often use non-allelic donor templates. Such repair mechanisms are often highly mutagenic and can also yield chromosomal rearrangements and/or deletions. While the choice between HR and its less precise counterpart, non-homologous end joining (NHEJ), has received much attention, less is known about how cells manage and prioritize HR subpathways. In this review, we describe work focused on how chromatin and nuclear architecture orchestrate subpathway choice and repair template usage to maintain genome integrity without sacrificing cell survival. Understanding the relationships between nuclear architecture and recombination mechanics will be critical to understand these cellular repair decisions.

1. Introduction to homologous recombination (HR): beneficial and deleterious outcomes

By querying the genome for optimal donor sequences for conservative repair, HR can accurately repair large numbers of DSBs within a single cell (Fig. 1). This process is arguably the highest fidelity mechanism by which DSBs are repaired. HR can also be attempted in situations where an ideal repair template is not available, resulting in repair outcomes that are often deleterious. Such HR outcomes can yield gross chromosomal rearrangements where large portions of a chromosome are translocated or lost. While these repair events may promote survival in the instance of single cell organisms, such genome instabilities in higher eukaryotes are often associated with diseases such as cancer. In this respect, cell death due to failure to repair a DSB may be provide lower risk to multicellular organisms, avoiding illegitimate repair by the HR machinery [1,2]. This review provides an overview of the changes that occur in chromatin at the molecular and nuclear architectural levels during HR, and how regulation of these processes helps to guard the genome from the deleterious consequences of low fidelity HR pathways.

Fig. 1. DNA repair pathways involving HR.

During HR, DSBs (A) are resected to form 3’ single-strand tails which (B) invade at a donor site to form a D-loop. In the most common mitotic HR pathway, synthesis-dependent strand annealing (SDSA), this D-loop structure is extended by DNA polymerase (G), then unwound to reanneal to the 3’ overhang adjacent to the original DSB site. Repeats exposed by 3’ end resection can anneal by a Rad51 independent manner through single strand annealing (SSA) to form a deletion product (H). In the less common canonical double-strand break repair (DSBR) pathway, the second 3’ end of the DSB is annealed to the displaced strand of the D-loop to form a double Holliday junction (C). This double Holliday junction can be cleaved to form crossover (D) or noncrossover products (E) or undergo dissolution by the Sgs1-Topo3-Rmi1 complex to form a noncrossover (F). The rarely occurring break-induced replication (BIR) pathway involves extensive D-loop extension by DNA Polymerase δ, followed by lagging strand synthesis which is templated off the newly extended leading strand (I). The pathways are organized left to right, from high to low fidelity.

1.1. First challenge to maintaining optimal HR: subpathway choice.

DSBs can occur in a cell as a result of exogenous or endogenous agents during any stage of the cell cycle. For example, a nick in DNA encountered during replication can cause a DSB as well as replication fork collapse [3,4]. Eukaryotes use two major DSB repair pathways: HR, which generally utilizes a donor template to direct repair of a DSB, and non-homologous end joining (NHEJ), which rapidly ligates two DNA ends together. Occurring primarily in the G1 phase of the cell cycle in baker’s yeast and in all phases of the cell cycle in mammalian cells, NHEJ is considered a low fidelity pathway. During NHEJ, rapid ligation of multiple DSB ends can yield gross chromosomal rearrangements and deletions [5,6]. In contrast, in the S and G2 phases of the cell cycle, the higher fidelity process of HR utilizes a donor template to repair DSBs [5,6]. Much work has been done to understand how the HR-NHEJ pathway decision is made in mitotic cells [6,7]. Briefly, this decision involves DNA damage response factors and is dependent on the cell cycle stage as well as the abundance and activity of factors that either promote the direct joining of broken DNA or initiate the 5’ to 3’ resection of DSBs to commit to HR. In addition, different NHEJ/HR outcomes can occur depending on the organism/cell-type studied and chromatin context of the DSB [4,6–8]. In contrast, the choices that a cell makes once committed to HR are not as well characterized and are the focus of this review. The factors described in this review refer primarily to baker’s yeast proteins, with mammalian nomenclature for homologs and orthologs provided when relevant.

In initial stages of HR, DSB ends are bound by the MRX complex (Mre11-Rad50-Xrs2 in baker’s yeast, MRE11-RAD50-NBS1 in mammals), which promotes, through actions of the Mre11 exonuclease and the associated factor Sae2 (CtIP in mammals), limited resection of the DSB to yield short 3’ single-stranded tails [4,9,10]. Long range resection occurs through the coordinated actions of 5’−3’ exonucleases and helicases (Exo1, Dna2, Sgs1 [9–11]). The resulting 3’ tails are covered by the single-strand binding protein RPA, which is ultimately displaced to form a Rad51 nucleofilament that can invade a duplex homologous donor locus to form a displacement loop (D-loop) (Fig. 1B,[10]). This D-loop structure contains heteroduplex DNA comprised of the 3’ single strand end of the processed DSB and the complementary strand from the donor template. The 3’ end of the processed DSB, now part of the heteroduplex, is then extended by DNA polymerase δ.

At this stage, several sub-pathways diverge to repair the DSB by different means (Fig. 1). Most prevalent is Synthesis Dependent Strand Annealing (SDSA), a mechanism involving displacement of the invading strand, which can the reanneal back to the broken chromosome. Subsequent DNA synthesis and ligation steps complete the repair process [12–17]. SDSA is considered the lowest risk HR subpathway; while SDSA repair using a homolog or divergent template can result in gene conversion and/or loss of heterozygosity, it rarely yields chromosomal rearrangements [12–17]. In mitotic cells, a less utilized canonical double-strand break repair (DSBR) pathway can occur where the second 3’ ssDNA is captured (termed second-end capture) by annealing to the displaced ssDNA of the D-loop to the form a double Holliday structure [15,18,19]. In this mechanism, the 3’ ends of the invading strands are extended by DNA synthesis and branch migration of the Holliday junction may also occur. The double Holliday junction is resolved by structure selective nucleases to form either crossover or non-crossover products [19], or dissolved through branch migration and decatenation activities catalyzed by the Sgs1-Top3-Rmi1 helicase/topoisomerase complex [20,21]. While the canonical DSBR pathway facilitates high fidelity repair if an allelic template is used, it can cause gross chromosomal rearrangements if the donor sequence invades in a non-allelic genomic position.

In some instances, the cell may utilize HR pathways that are inherently error prone such as Break Induced Replication (BIR), or are non-conservative such as Single Strand Annealing (SSA; Fig. 1). As indicated above, resection of DSBs to form 3’ DNA overhangs represents a commitment step in HR [22]. In the rare instance that a homologous template cannot be found for both ends of a DSB, cells utilize lower fidelity HR pathways. One such pathway is BIR, where a single 3’ end extending from a DSB can initiate conservative DNA synthesis using a homologous template. BIR may extend to the end of a chromosome, but also involves frequent template switching [23]. BIR is a low fidelity, mutagenic process that has been associated with genome instability and disease. The high level of mutation associated with BIR is thought to be the result of frequent template switching and the presence of a migrating bubble structure which may also expose large stretches of single-stranded DNA which is susceptible to damage [23–26]. At best, BIR causes mutational burden, and at worst, gross chromosomal rearrangements, as well as amplification and deletion of genetic information (copy number alterations) [23–26]. Alternatively, a non-conservative HR pathway, single-strand annealing, can occur which does not require the formation of a D-loop. Single-strand annealing acts to repair DSBs through a deletion mechanism involving direct repeat sequences; complementary repeat sequences are exposed by DSB end resection and then anneal [3,4]. Non-homologous 3’ tails that result from sequences present between the repeats are then clipped off, with the final product containing a deletion of sequences between the repeats [3,4].

1.2. Second challenge to maintaining optimal HR: template choice.

Unfortunately, the very mechanism which allows HR to find an optimal donor template, the homology search process, can lead to the generation of inappropriate repair products. During the process of HR, the sister chromatid is the ideal template and should be available within the context of the replication fork (or in late S/G2); however, if the sister chromatid is damaged or has been replicated imperfectly, the homolog or other non-allelic sequences may be used as templates [19]. Mechanistically, template choice in vegetative cells is facilitated by sister chromatid cohesion (see below) and is regulated by the ability of the Rad51 nucleofilament to form a D-loop. Potential donor sequences of less than ~10% divergence can be invaded by the Rad51 nucleoprotein filament to form a D-loop which can then be unwound pre- or post- D-loop extension by means of various helicases, in a process termed heteroduplex rejection [27–36]. Recombination intermediates that involve non-allelic loci often contain mismatches or insertion/deletion loops that are present when heteroduplex DNA is formed [27–36]. Such small non-homologies are recognized by the heterodimeric mismatch recognition complexes Msh2-Msh6 or Msh2-Msh3 [27–29,37–40], which then can recruit factors such as the Sgs1-Top3-Rmi1 (STR) complex [31,33,34,41] or potentially the helicase Mph1 [42] to dismantle the D-loop (Fig 2; [43]).

Fig. 2. A model for interactions between the chromatin and heteroduplex machineries during DNA repair.

Following D loop formation (A), histone chaperones CAF-1 (red circles) and Rtt106 (purple rectangle) deposit core histones at or near heteroduplex DNA, most likely through interactions with PCNA (green ring), to stabilize the D-loop structure (B). This stabilization inhibits the actions of the mismatch recognition complexes (purple ring), which recognize non-homologies between donor and recipient strands of the heteroduplex. Successful stabilization results in a repair outcome (C), while successful recruitment of the rejection machinery such as the STR complex (blue, cyan, and green circles) leads to dismantling of the D-loop (D).

The HR and heteroduplex rejection machineries can sensitively detect non-homologies in heteroduplex DNA during repair but are unaware of the larger context or relative genomic positions of the donor and recipient loci. This is a particular concern in higher eukaryotes which contain large numbers of gene duplications and repetitive elements scattered throughout their genomes. Using similar, but not allelic loci as donor templates can result in gene conversion events and loss of heterozygosity, as well as deleterious gross chromosomal rearrangements, all of which have been linked to disease states [44]. Paradoxically, this problem cannot be solved by “tuning” the HR/heteroduplex rejection machinery to require perfect homology between a broken DNA sequence because of naturally occurring non-homologies. These non-homologies might arise between sister chromatids due to replication defects, and between chromosome homologs due to allelic differences. Failure to use these templates would result in an unrepaired DSB and cell death or chromosomal rearrangements. Thus, mechanisms that take into account overall nuclear architecture must be employed to help guide the HR machinery in donor locus selection. Recent studies summarized below provide evidence that chromatin factors act in such a manner.

Recent studies have shown that HR subpathways as well as donor template choice play important roles in the fidelity of DNA repair. As outlined in this review, individual subpathways and donor templates appear to be in competition, and hierarchies for these pathways have been identified [4,15,18,45–47]. However, the mechanisms that facilitate these choices have yet to be fully understood. Answering these questions requires an understanding of the interplay between chromatin, HR, and heteroduplex rejection machineries [25,26,40,27–34]. We discuss how chromatin, both in the local environment and at a nuclear architectural level, interacts with the repair machinery to increase the likelihood of repair events occurring through high fidelity HR.

2. Chromatin guides HR subpathway choice

2.1. Chromatin facilitates rapid DSB repair through first choice HR pathways and is accompanied by a global increase in chromatin mobility.

Cells orchestrate major changes in nuclear architecture to facilitate the rapid repair of a DSB. As first demonstrated in baker’s yeast, in the absence of DSBs, chromosomal loci within S-phase nuclei display small radii of confinement (~3–10% of the nuclear volume) (Fig. 3; [48–57] Upon induction of a DSB at a single locus, the affected locus and other undamaged genomic loci become more diffuse, sampling roughly 10% and up to 50% of the nuclear volume for the affected and undamaged loci, respectively [48–57]. This increase in mobility (Fig. 3) is dependent on the DNA damage checkpoint [48,51,52,55–57], components of the strand exchange machinery (Rad51 and the strand exchange stimulatory factor Rad52) [48,50,54,55], and the chromatin remodeler INO80 [49,51,52]. Interestingly, the magnitude of these increases in mobility scale with the number of DSBs induced [50,55]. In baker’s yeast, two potential mechanisms that may not be exclusive have been proposed to explain these behaviors. In the first, the DNA damage checkpoint kinases Mec1 and Rad53 direct un-tethering of yeast centromeres from the spindle pole body by phosphorylating the kinetochore protein Cep3 [52]. However, the importance of this phosphorylation event is somewhat unclear, as other labs have not observed either the release of yeast centromeres by DNA damage induction, or the dependence of DSB mobility upon Cep3 phosphorylation [56,58]. In the second model, the DNA damage response (DDR) proteins and the chromatin modifier INO80 initiate a pathway in which histones are degraded, resulting in global chromatin decompaction [57]. While the conservation of these mechanisms across eukaryotes remains somewhat unclear (reviewed in detail in [59]), it appears that at least in certain mammalian cell types, DNA damage response dependent mobility after DSB induction is observed [60]. This increase in global and DSB-specific mobility presumably allows for a more extensive homology search [50,58].

Fig. 3. Increase in loci mobility after DSB induction.

Upon DSB induction, an HR dependent increase in mobility is observed in the nucleus both at the site of the DSB and globally. This process is dependent upon Rad51 (cyan circles), Rad52 (not shown), and the DNA damage response (DDR) kinases Mec1 (pink hexagons), and Tel1 (orange ellipses). Global increases in mobility have been hypothesized to occur through multiple processes, including activation of the chromatin remodeler INO80 (red square) by Mec1, leading to histone eviction and subsequent proteosomal degradation (illustrated as a red cylinder). Global increases in mobility allow heterochromatic regions to expand in volume and become more diffuse (illustrated by expanding purple regions) [121].

Interestingly, the efficiency of a homology search is affected by nuclear architecture; templates located in closer proximity within the nucleus are used preferentially compared to those located further away [61,62]. Thus, an increase in mobility of both the DSB locus and the genome as a whole facilitates a more unbiased homology search during DNA repair. Along these lines, Mine-Hattab and Rothstein [50] found that the overall movement of a particular locus, as determined by measuring the mean-square displacement of foci, correlated with homolog pairing following the induction of a DSB. In support of this idea, spontaneous Rad52 foci indicative of S-phase DSBs associated with a replication fork display reduced movement [63]. Presumably, since these spontaneous DSBs have a sister chromatid to use as a template, homolog pairing, and thus an increase in mobility associated with finding the homolog, is unnecessary [63]. Another interesting observation is that diploid yeast in G1, where the homolog, but not the sister chromatid is available as a template, display DSB dependent increases in mobility, concomitant with Rad51 foci formation. Similar to the previous studies mentioned above, this observation was dependent on Rad51 and the DNA damage checkpoint [64]. In contrast, it was independent of the strand exchange stimulatory factor Rad52 that also acts in SSA [64]. More research in this area is certainly needed to discover how the cell cycle and ploidy affect this process.

2.2. Significance of repair centers?

DSBs repaired by HR are localized to specific centers in the nucleus where efficient HR processes such as SDSA and canonical DSBR are performed (Fig. 4). These repair centers were first observed in baker’s yeast, and different groups noted that the number of foci formed by HR proteins, especially Rad52, was fewer than the DSBs induced [63,65]. This indicates that multiple DSBs are brought together to form a single focus. Repair foci have been well characterized in baker’s yeast, and contain HR proteins such as Rad51, Rad52, Rad59, Rad55, Rad54, Rdh54, and Rfa1, the large subunit of the RPA complex, and components of the DNA damage response machinery that include Mre11, and Tel1, the yeast homolog of the mammalian ATM DNA damage response kinase [66]. In mammals, these repair centers also include and are usually characterized by their 53BP1 (DNA damage-dependent checkpoint protein, ortholog of the baker’s yeast Rad9 protein) foci [67–69], which have been hypothesized to link DNA damage repair to cell fate decisions [70]. The purpose of these repair centers appears to be the efficient repair of DSBs through the increased concentration of repair machinery [71], facilitating enhanced DSB end resection and recruitment of downstream factors necessary for HR. Recent research in both yeast and mammalian systems indicate that repair centers display characteristics of phase-phase separated compartments [70,72–75]. In yeast, this behavior, and the formation of repair foci, appears to be partially dependent on a long disordered region of the repair protein Rad52 [74,75]; disruption of this region, or treatment with the drug 1,6 hexanediol reduces the formation of damaged induced repair foci [74]. Similarly, the damage induced 53BP1 foci observed in higher eukaryotes are also thought to be phase-separated compartments [70,72,73].

Fig 4. Repair center formation during DSB repair.

Upon DSB induction, multiple DSBs colocalize to form repair centers in the nucleus. These repair centers are characterized by an increase in the concentration of repair factors, such as the Rad52 epistasis group proteins, as well as DNA damage signaling kinases. Repair centers, thought to be liquid-liquid phase separated compartments, facilitate efficient repair by HR.

Several hypotheses have been proposed by which these liquid droplets might form upon DNA damage. In yeast, the disordered region of Rad52 is necessary [74], but not sufficient [75] for this process, and a tentative hypothesis has been proposed that damaged induced microtubules induce a nuclear motion which lends itself toward the coalescence of multiple Rad52 foci into a single focus [74]. Hypotheses proposed for the formation of phase-phase separated 53BP1 compartments in higher eukaryotes include liquid de-mixing seeded by poly(ADP Ribose) [72] and the transcription of damage induced long non-coding RNAs [73]. In mammalian cells, localization of the DSB to such repair centers have been shown to be dependent on the actin nucleator Arp2/3 and actin filaments [76].

2.3. Nuclear localization regulates the use of lower fidelity repair pathways at the nuclear periphery.

Recent studies have shown that slow to repair or unrepairable DSBs are first targeted to the nuclear periphery (Fig. 5A). In baker’s yeast, this mechanism occurs through a tethering mechanism involving the nuclear membrane protein Mps3, a member of the SUN family of proteins that is required in baker’s yeast for spindle pole body duplication, and telomere movements in meiotic prophase [77–80]. Intriguingly, this decision to target DSBs to the nuclear envelope appears to be temporal, rather than based on the presence or absence of a donor repair template. To test this idea, Oza et al. [77] utilized single-strand annealing reporters in baker’s yeast and showed that a slow to repair construct containing 30 kb of non-homology between repeats was targeted to the nuclear envelope, whereas a rapidly repairing construct containing only 5 kb of non-homology was not. How are these slow to repair or unrepairable DSBs targeted to the nuclear membrane? In agreement with the idea that failed HR attempts localize DSBs to the inner nuclear membrane, studies showed that the targeting of unrepairable/slow to repair DSBs requires many HR proteins as well as the DNA damage response checkpoint kinases, and only occurs in the S or G2 phases of the cell cycle [77–79]. Chromatin factors were also required for this process and include the chromatin remodeling complexes INO80 and SWR1 [51,77–79]. An interesting observation which appears unique to unrepairable DSBs destined for the nuclear membrane was the presence of extensive coating of Rad51 across the damaged chromosome [78]. Given the fact that resection of ectopic or unrepairable DSBs appears to continue indefinitely until a repair outcome is reached [81], it seems possible that continued resection and Rad51 loading acts somehow through the DNA damage response to “switch” targeting of DSBs from their current position to the nuclear membrane. While the mechanism of this “switch” is unknown, it involves the mono-SUMOylating SUMO E3 ligase Mms21, a subunit of the Smc5/6 complex [80].

Fig 5. Peripheral localization of inefficiently repair DNA.

(A) DSB ends which cannot be repaired because they lack a homologous donor or are slow to repair due to the sole availability of an ectopic donor sequence, are shuttled to the nuclear periphery through two distinct mechanisms. In S and G2, targeting to the inner nuclear membrane occurs by a mechanism which requires DNA damage response kinases Mec1 and Tel1, and the chromatin remodelers SWR1 and INO80. Critical to this localization is mono-SUMOylation Mms21. At any stage of the cell cycle, unrepairable DSBs and stalled replication forks can be shuttled to the nuclear pore. This mechanism requires both mono-SUMOylation by Mms21 and poly-SUMOylation by Siz2, and the SUMO dependent ubiquitin ligase protein Slx5/Slx8. This process is associated with repair by error prone mechanisms such as BIR. (B) Heterochromatic DSBs are excluded from heterochromatin for HR. DSBs which occur in heterochromatin are resected to form 3’ overhangs (RPA in red ovals), and then excluded from these areas prior to Rad51 (cyan circles) loading. These DSBs are then targeted to the nuclear pore. This mechanism involves the formation of nuclear actin filaments (blue chevrons) by Arp2/3 and SUMOylation by Mms21 in Drosophila. In yeast, DSBs are excluded from the nucleolus in a manner requiring the SUMOylation of Rad52.

What happens to DSBs when they reach the nuclear membrane? While the answer is not yet known, studies have suggested that HR is suppressed [79,82] and repair occurs through the recruitment of the telomerase machinery to promote telomere addition [77] to thus heal the DSB. This model is not outlandish, given the association of the telomerase machinery with Mps3 [83]. Taken together, these data suggest a model where the absence of a repair template leads to extensive resection, and continued failure of HR prompts the DNA damage response factors Mec1/Tel1 to actively target unrepairable/slow to repair DSBs to the nuclear membrane where they associate with the membrane protein Mps3. Given that this process seems to only involve slow to repair DSBs, this type of localization is most likely a relatively rare event. Especially in diploid organisms, a donor should theoretically always be available for repair; in fact in baker’s yeast, recent evidence suggests that HR may proceed using the homolog as a template in interphase [64].

2.4. Peripheral localization of DSBs associated with the nuclear pore.

A second and distinct pathway by which DSBs localize to the nuclear periphery in G1 and early S phase is through association with the nuclear pore complex (Fig. 5A; [79,80,84,85]). This pathway was first characterized by Nagai et al. [85], who discovered a link between the DSB repair machinery, the SUMO-dependent ubiquitin ligase Slx5/Slx8, and the nuclear pore complex Nup84. They developed a model where stalled replication forks become SUMOylated in a Mec1/Tel1 dependent manner, leading to their localization to the nuclear pore. This in turn could lead to Slx5/Slx8 dependent ubiquitination which would promote degradation of histone proteins by the proteasome and subsequent replication fork repair and restart [85]. Along these lines, expanded (CAG) triplet repeat tracts, which presumably lead to stalled replication forks, also localize to the nuclear pore in an Slx5/Slx8 dependent manner, and failure to do so leads to greater breakage at these repeats [86]. Recent work by the Freudenreich lab indicates colocalization of Rad51 with stalled replication forks only after they reach the NPC, indicating repair of stalled forks by HR is facilitated in some way by this relocalization [87]. This model is further supported by work in systems tracking the location of induced, unrepairable DSBs tagged by LacO binding arrays. Specifically, Horigome et al. [80] showed that the association of persistent DSBs with the nuclear pore requires both the mono-SUMOylation Mms21, followed by polySUMOylation by the SUMO E3 ligase Siz2. In the case of stalled replication forks, this association seems to require only SUMOylation by Mms21 [87]. Recently, Rad52, RPA, and Rad59 have been shown to be targets of such SUMOylation, and to be key to the re-localization of stalled replication forks to the NPC [87]. This in turn promotes the association of the Slx5/Slx8 SUMO dependent ubiquitin ligase with the DSB and subsequent nuclear pore targeting [80]. Recent research in this area also implicates damage induced microtubules and their associated kinesins with nuclear pore targeting [84,88].

The mechanism(s) by which DSBs are sent either to the either the nuclear pore complex or the nuclear membrane, and the decision between these two pathways, continues to be investigated. In brief, this decision seems to be primarily governed by the cell cycle and mono/polySUMOylation status of various repair factors, as outlined by Horigome et al. [80]. Specifically, monoSUMOylation by Mms21 occurring in S/G2 targets persistent DSBs to the membrane [80]. Subsequent polySUMOylation, (requiring initial monoSUMOylation by Mms21) presumably of the same targets in G1 or S phase by the SUMO ligase Siz2 targets DSBs to the nuclear pore.

In some contexts, DSB localization to the NPC appears to promote genome stability, preventing unequal sister chromatid exchange [80,85,89]. However, localization with the nuclear pore is also associated with repair via lower fidelity mechanisms. For example, DSBs that occur in subtelomeric regions require tethering to the nuclear pore for efficient DNA repair [90,91]. Such studies showed that DSBs that occur in both telomeric and sub-telomeric regions localize to the nuclear pore, and repair of these DSBs is viewed as an error prone HR pathway [88,90–92]. Repair of these DSBs does not occur via canonical DSBR or SDSA method, but rather by BIR [88,92].

To explore the relationship between BIR, the nuclear pore complex, and subtelomeric DSBs, Chung et al. [88] created a substrate that can be placed at a subtelomeric or internal locus where two induced DSBs excise the URA3 gene. Loss of URA3, which can be selected for, indicates efficient DSB induction. Repair of the DSB by NHEJ or BIR could be distinguished. Demonstrating the relationship between nuclear pore targeting and repair by BIR, Chung et al. [88] showed that DSBs which occur in subtelomeric regions are repaired more efficiently by BIR than their internal counterparts, and that subtelomeric repair is dependent on tethering of the DSB to the nuclear pore. Similar to studies that examined stalled DNA replication forks, the subtelomeric repair examined by Chung et al. [88] was dependent on the SUMO-dependent ubiquitin ligase Slx5-Slx8 complex. Once tethered to the pore, these subtelomeric DSBs were repaired by BIR, as repair was prevented by the absence of components of the BIR machinery [88]. For the internal DSB, they demonstrated artificial targeting to the nuclear pore by use of what they term a “zip code” sequence, increased repair efficiency by BIR, while mutations in the nuclear pore significantly reduced repair [88].

2.5. A hypothesis: BIR is restricted to the nuclear pore.

Taken together, these data indicate a potential mechanism to explain how DSB localization impacts HR subpathway choice. In this model, DSBs are first sent to a repair center [63,65,66] in which rapid repair is attempted through high fidelity canonical DSBR or SDSA pathways. Unsuccessful repair of the DSB would result in ultimate targeting to the nuclear pore, or potentially the nuclear envelope, where BIR or spontaneous telomere addition would occur, respectively [77–80,84]. An important part of this model is that BIR is only permitted at the nuclear pore and is suppressed in the interior of the nucleus; such a mechanism would ensure that higher fidelity HR subpathways would first be attempted.

3. Chromatin affects choice of possible repair templates

3.1. Pre-invasion: Sister chromatid cohesion makes the sister chromatid the most likely target.

Equally important as the choice of specific HR repair pathway is the choice of donor sequence used as a template for repair. Current data indicate that this step is rate limiting for HR [61,62,92]. During this step, overall nuclear architecture and chromatin accessibility can have a significant impact on use of various potential donor templates [61,62,92]. As mentioned previously, the sister chromatid provides an ideal donor template; the corresponding locus on the undamaged sister chromatid should be identical, barring potential replication errors. Mechanistically, this is accomplished by maintaining a physical connection between sister chromatids via sister chromatid cohesion, which is established more strongly upon DSB induction [93]. Exactly how this occurs is not known; however, recent data studying the loading of cohesin as a DSB response implicate the histone deacetylases Hda1 and Rpd3 in establishing sister chromatid cohesion at a DSB occurring in the context of the replication fork [94]. While this particular study found this phenomenon to be replication fork specific, it is worth noting that other research groups have found Hda1 and Rpd3 to localize to an induced DSB at MAT during mating type switching [95]. Additionally, it is worth noting that Hda1, along with Sir2, facilitate sister chromatid cohesion at silent chromatin in S. cerevisiae [96]. The involvement of histone deacetylases in regulating such architecture is unsurprising; deacetylation of histone lysine residues “erases” acetylation marks which in turn affect the way “reader” proteins containing acetyl-lysine recognizing bromodomains interact with nucleosomes. This deacetylation thus affects a wide variety of processes in the nucleus. In addition to deacetylating histone lysines, histone deacetylases are capable of deacetylating non-histone protein targets in the nucleus, further increasing their regulatory control. While beyond the scope of this review, various histone variants and post-translational modifications certainly play a role in efficient DSB repair and signaling; for more complete reviews, see Hauer and Gasser [97], and Clouaire and Legube [98].

3.2. General increases in chromatin mobility enable homolog use as donor templates.

While most DSBs repaired by HR utilize the allelic locus on the sister chromatid as a template, allelic loci on the homologous chromosome can also be utilized as a repair template [19]. In mitotic cells, homologs are unpaired and not in physical proximity to one another [50], and thus the donor and recipient sequences must be brought into proximity to each other through the homology search. The global increase in chromatin mobility and accessibility outlined in the previous section is thought to facilitate this process; homolog pairing can be observed immediately following this damage dependent increase in mobility [50].

3.3. “Tricky” heterochromatic DSBs are sequestered away from other repeat sequences.

Global increases in chromatin mobility and opening of chromatin structure carry with them the potential discovery and use of a non-allelic donor templates for repair, thus increasing the risk of non-allelic homologous recombination and its associated genome instability. As the repetitive element content of a genome increases, so does the risk of such recombination (Fig. 5B). In higher eukaryotes, the risks associated with such recombination may outweigh the benefits of cell survival in such cases. Accordingly, differences between yeast and mammalian cell systems have been noted with regard to the nuclear volume a DSB can probe [99,100]. These differences seem to be to some extent cell type specific in higher eukaryotes.

While the global changes in nuclear architecture may be cell type specific, it seems that one common way to reduce the risk associated with a genome wide homology search process is to sequester heterochromatic DSBs away from heterochromatic areas. This observation has been studied most extensively in Drosophila, where heterochromatic DSBs have been observed to rapidly relocate from their typical DAPI-dense HPA1 domains, into nearby euchromatic regions prior to Rad51 loading [101,102], where they eventually localize to the nuclear pore [103]. Once relocated outside their native heterochromatic domains, they then load Rad51 and presumably resume repair away from heterochromatic repeat sequences which might provide tempting ectopic donor sequences. This mechanism of relocation resembles that required for persistent DSBs to localize to the pore; the SUMO ligase Mms21, a subunit of the Smc5/6 complex, is required for this localization [103]. The mechanism by which this occurs in Drosophila is the topic of much recent research. In short, the process involves recruitment of Arp2/3, an actin assembly factor, to sites of DNA damage, which results in the assembly of nuclear F-actin filaments [76,104,105]. Nuclear myosins are then involved in relocating the DSB out of its heterochromatic domain to be repaired, either at a repair center, or to the periphery [76,104,105]. This phenomenon has been observed across several species in higher eukaryotes, and evidently mirrors the behavior of budding yeast nucleolar DSBs, which also relocate out of the nucleolus to complete recombinational repair away from potentially dangerous rDNA repeats [106]. Concurrent with the mechanism of relocation in Drosophila, relocation out of the yeast nucleolus requires SUMOylation of the repair protein Rad52 [106]. How conserved this mechanism is, and whether this system relates to the relocation of other DSBs, such as unrepairable DSBs or those associated with stalled replication forks to the nuclear pore remains unclear.

3.4. Post-invasion: Deletion of chromatin factors appears to affect the repair-rejection decision.

In addition to guiding DSBs before strand invasion to those donors most likely to be allelic, chromatin factors also affect the probability of use of a potential donor template post strand invasion (Fig. 2). Recent research indicates the fluidity of the steps and subpathways of HR; the helicases Sgs1, Srs2, and Mph1 have all been shown to unwind different joint molecules, allowing reversibility of steps during HR [15,31,42,107,108]. These unwinding events can occur pre- or post- 3’ end extension to create cycles of invasion, potential extension, and unwinding [4,18,107]. If the heteroduplex between the donor and invading single strands contains mismatches, these D-loops are actively targeted for disassembly. This occurs by use of the mismatch repair system, which recognizes such non-homologies and is thought to recruit the STR complex to unwind the D-loop structure in a process called heteroduplex rejection [27–30,32–34,36]. This process is affected by various chromatin factors including factors involved in replication-coupled nucleosome deposition, and several histone deacetylases [36]. Specifically, the histone chaperone CAF-1 and Rtt106 suppress heteroduplex rejection [36]. To explain this observation, histone deposition by CAF-1 and Rtt106 in or around the heteroduplex may act to stabilize the structure and prevent attempts at unwinding by helicases such as Sgs1 [36,109]. This D-loop stabilization by these histone chaperones would thus antagonize the efforts of the mismatch machinery to reject heteroduplexes containing non-homologies. In support of this, an antagonistic relationship between CAF-1 and related histone chaperone Asf1 with the mismatch repair machinery has been previously reported in vitro [110], and the replication coupled nucleosome assembly factors CAF1 and Asf1 have been implicated in multiple stages during recombination [111,112].

Histone deacetylases are another class of chromatin modifier that affect heteroduplex rejection. Specifically, the Sir2/3/4 deacetylase complex was shown to suppress rejection, while the Rpd3 complex promoted it [36]. Interestingly, these two deacetylase complexes have been previously shown to localize to DSBs [95,113,114]. Whether these factors act to promote or suppress rejection through direct deacetylation of nearby histones, repair proteins, or by another mechanism entirely remains to be discovered. One insight which may help us to understand this relationship is to develop a timeline for heteroduplex rejection. When a heteroduplex is formed, non-homologies may exist at the 3’ end of the invading strand. These non-homologous tails have been shown to be critical for rejection in single-strand annealing and break-induced replication models, and are clipped off during repair by Rad1-Rad10 [34,115–117]. This clipping activity is thought to reflect a commitment step for DNA repair, after which mismatch repair mediated heteroduplex rejection cannot act [34,36,115]. If this is the case, it may be that histone deposition by factors such as the replication-coupled nucleosome deposition pathway stabilize repair intermediates for efficient repair, after the opportunity for rejection has ended (Fig. 2; [36]).

4. Conclusions and implications for our understanding recombinational repair decisions

This review is focused on understanding how cells choose specific HR subpathways. To answer this, one must understand how the recombination machinery and accompanying changes in nuclear architecture are coordinated. Work presented in this review indicates that the high-fidelity repair of a DSB by HR is an orchestrated set of events that involve changes in chromatin and nuclear architecture. By shuttling DSBs to different subnuclear compartments, the cell can prioritize fast and efficient repair. We are attracted to the idea that cells suppress error prone repair processes such as BIR by restricting them to specific areas of the nucleus where they can act as last resort mechanisms. By shuttling only those DSBs which are unrepairable by other means to the nuclear periphery where BIR can occur, the cell can prioritize the choice of repair pathways. Continued research investigating the mechanisms by which DSBs localize to different parts of the nucleus, and how this correlates with repair outcomes will shed light on how these processes are coordinated. The mis-regulation of high and low fidelity HR mechanisms is thought to be associated with the formation of complex genome rearrangements seen in cancers. Thus, by continuing to analyze HR pathway choice through analyses that combine mechanistic and chromosome architecture approaches, researchers can obtain a better understanding of events that underlie cancer biogenesis.

4.1. Common factors and structures could be regulated by nuclear location.

Sub-nuclear compartmentalization allows for a change in the local concentration of repair factors across different compartments. This allows for a significant fine-tuning of how and when recombination occurs in the cell. For example, the exclusion of Rad51 from heterochromatin has been proposed as a mechanism (Fig. 5B) to prevent aberrant recombination between repeat elements [102]. This concept applied to the mechanics of recombination could possibly explain the discrepancies seen between different repair systems. For example, cells seem to display various levels of stringency with regard to heteroduplex rejection of non-homologous donor sequences between different repair systems. In an inverted repeat based recombination assay thought to be repaired by high-fidelity HR pathways in the context of a replication fork during S phase, even very small non-homologies dramatically reduce repair efficiency [30,31]. In contrast, systems which repair via BIR are far more tolerant of nonhomologous heteroduplexes [28–34,36,38,115]. If different repair systems localize to separate sub-nuclear domains, a model to explain these differences across systems may lie in variation in concentration or activity of rejection machinery across different compartments.

4.2. Fluidity within a pathway may be heavily influenced by chromatin environment.

D-loop structures that form by strand invasion are dynamic, often dissolving before or even after 3’ end extension. Subsequent invasion and synthesis steps often lead to a pattern of short conversion tracts from different donor sequences [118]. The information presented in this review supports the hypothesis that chromatin environment affects these dynamics; current work indicates a potential antagonistic relationship between factors acting to deposit histones and the mismatch repair machinery and helicases which act to unwind D-loop intermediates. It will be important to investigate this relationship further to fully understand the repair/rejection decision. At present, a study that investigated the relationship between the mismatch repair machinery involved in heteroduplex rejection and histone chaperones suggests that histone chaperones such as CAF-1, Asf1, and Rtt106 act to promote the stability of repair intermediates as well as rapid and stable repair. Acting in an antagonistic manner, the MMR machinery promotes rejection by the recruitment of helicases to unwind recombination intermediates in the presence of mismatches.

There is much still to learn about how chromatin, nuclear architecture, the DNA damage response, and the HR machinery act in a concerted effort to choose both the repair template and HR subpathway. Work in the coming years will be critical for us to understand how these processes act alone and in concert. The dependency of some of these architectural changes upon checkpoint kinases Mec1 and Tel1, as well as these kinases signaling roles in the phosphorylation of H2A (H2A.X in mammals), implies such crosstalk exists. Further evidence for such a relationship is evidenced by the recent discovery of a phospho-peptide recognizing H-BRCT domain in Sir4, which is thought to regulate varying processes in which the Sir2/3/4 complex is involved [119,120]. Hopefully, further work in this area will help map these relationships to really understand the regulation of DSB movement, subpathway choice, and template usage during HR.

Highlights.

Homologous recombination is considered the highest fidelity DNA double-strand break repair pathway that a cell possesses, but careful regulation of this process is critical to maintain genome integrity.

In “last resort” scenarios, homologous recombination can be directed to low fidelity subpathways, most likely in distinct sub-nuclear compartments.

Chromatin and nuclear architecture help to direct donor template usage during homologous recombination.

We describe work focused on how chromatin and nuclear architecture orchestrate subpathway choice and repair template usage to maintain genome integrity.