Trinucleotide repeats can form secondary structures, whose inappropriate repair or replication can lead to repeat expansions. There are multiple loci within the human genome where expansion of trinucleotide repeats leads to disease. Although it is known that expanded repeats accumulate double-strand breaks (DSBs), it is not known which DSB repair pathways act on such lesions and whether inaccurate DSB repair pathways contribute to repeat expansions. Using Saccharomyces cerevisiae, we found that CAG/CTG tracts of 70 or 155 repeats exhibited significantly elevated levels of breakage and expansions in strains lacking MRE11, implicating the Mre11/Rad50/Xrs2 complex in repairing lesions at structure-forming repeats. About two-thirds of the expansions that occurred in the absence of MRE11 were dependent on RAD52, implicating aberrant homologous recombination as a mechanism for generating expansions. Expansions were also elevated in a sae2 deletion background and these were not dependent on RAD52, supporting an additional role for Mre11 in facilitating Sae2-dependent hairpin processing at the repeat. Mre11 nuclease activity and Tel1-dependent checkpoint functions were largely dispensable for repeat maintenance. In addition, we found that intact homologous recombination and nonhomologous end-joining pathways of DSB repair are needed to prevent repeat fragility and that both pathways also protect against repeat instability. We conclude that failure of principal DSB repair pathways to repair breaks that occur within the repeats can result in the accumulation of atypical intermediates, whose aberrant resolution will then lead to CAG expansions, contractions, and repeat-mediated chromosomal fragility.
CAG/CTG trinucleotide repeat (TNR) expansions form the basis of 14 inherited genetic disorders including Huntington's disease (HD), myotonic dystrophy type 1 (DM1), and several types of spinocerebellar ataxias (Pearson et al. 2005; Orr and Zoghbi 2007). Normal individuals carry an allele length of 6–35 repeats, which are stable. When the repeats expand beyond a threshold length of 36–40 repeats, they are unstable and become prone to subsequent expansions, and sizes of up to 135 and ∼2000 CAG/CTG repeats have been documented in individuals with HD and DM1 disorders, respectively (Pearson et al. 2005). Increases in repeat length during intergenerational transmission causes increased disease severity or earlier age of onset in the offspring. In addition, expansions during the lifespan of an individual may hasten disease progression. Thus, it is of interest to elucidate the underlying mechanisms that cause CAG/CTG repeat expansions to better understand the phenomena of disease inheritance, onset, and progression. In addition, because TNRs can inhibit DNA replication and repair, they serve as a good model to study DNA repair pathways that operate at structure-forming sequences.
Due to the intrinsic nature of single-stranded CAG and CTG repeat sequences to spontaneously form stable DNA secondary structures (McMurray 1999), they can interfere with DNA replication (Lenzmeier and Freudenreich 2003), and studies in model organisms have implicated DNA replication in CAG/CTG repeat expansions and contractions (Lenzmeier and Freudenreich 2003; Pearson et al. 2005; Mirkin 2007). When trinucleotide repeats are replicated such that the CTG strand is on the lagging strand template, deletions are common, suggesting that hairpins form on template DNA that is transiently single-stranded during passage of the replication fork (Freudenreich et al. 1997). CTG hairpins that form on the 5′ flap of an Okazaki fragment during replication or during gap repair are poor substrates for the FEN1 5′ flap-cleaving endonuclease, which can lead to ligation of the unprocessed flaps and repeat expansion (Spiro et al. 1999; Henricksen et al. 2000; Liu et al. 2004; Refsland and Livingston 2005). Slippage at the 3′ end of a strand during DNA replication or postreplication repair is another mechanism that can promote repeat expansions (Daee et al. 2007). Expanded CAG/CTG repeats can also act as replication fork pause sites or sites of fork reversal (Samadashwily et al. 1997; Pelletier et al. 2003; Fouche et al. 2006; Kerrest et al. 2009), and expansions are hypothesized to occur during Rad51-dependent fork restart processes or during recombination events that occur subsequent to fork breakdown (Mirkin 2007; Kerrest et al. 2009). In human HD patients, repeat expansions have been observed in replicating, premeiotic male germline cells (Yoon et al. 2003). Therefore, cell proliferation and by extension DNA replication is linked to repeat instability in humans.
Trinucleotide repeat expansions are also observed in nonproliferating cell types such as neural and muscle cells, implicating DNA repair processes unlinked to chromosomal replication in generating repeat length changes. CAG/CTG repeat expansions can occur during inaccurate repair of oxidative base damage in somatic cells of the mouse by a mechanism dependent on OGG1-mediated base excision repair and gap filling (Kovtun et al. 2007). In human cell extracts, nick repair is less efficient and more error prone if a CTG or CAG hairpin is nearby (Panigrahi et al. 2005). Ironically, proteins involved in repair may facilitate expansions by inappropriately binding to CAG or CTG hairpin intermediates. For example, the Msh2/Msh3 complex can bind CAG hairpins (Owen et al. 2005), and the presence of Msh2 is required for most expansion events in transgenic mouse models of CAG/CTG instability (Kovtun and McMurray 2008). Repeat instability is also dramatically enhanced by transcription across the locus and is modulated by both transcription-coupled and global-genomic nucleotide excision repair (NER) pathways in Escherichia coli, Drosophila, and human cells (Parniewski et al. 1999; Lin et al. 2006; Jung and Bonini 2007; Lin and Wilson 2007). Thus, inappropriate single-strand break (SSB) repair is another mechanism that generates expansions.
An unligated nick or a collapsed fork can result in the production of a double-stranded break (DSB). Indeed, expanded CAG/CTG repeat sequences act as fragile sites on a yeast chromosome (Freudenreich et al. 1998; Jankowski et al. 2000; Callahan et al. 2003). Since it is known that CAG/CTG repeats are sites of chromosomal breakage, it is reasonable to expect that DSB repair could play an important role in preserving repeat integrity. Despite this expectation, a clear role for DSB repair in preventing instability of trinucleotide repeat sequences has not been established. In addition, it is not known which DSB repair pathways are most important for repairing the breaks that occur naturally within TNRs, which may contain structures such as hairpins at the broken ends. Therefore, we set out to test the role of DSB repair in a system where both repeat instability and repeat fragility could be measured and in which DSB repair was not required for recovery of broken chromosomes. Our goals were to determine which DSB repair pathways are important in repairing the lesions that occur at an expanded trinucleotide repeat and to determine whether inaccurate DSB repair plays a role in the generation of CAG/CTG repeat expansions or contractions.
In this study, we have investigated the role of DSB repair pathways in preventing repeat instability and fragility at an expanded CAG-70 or CAG-155 repeat, originally cloned from a myotonic dystrophy patient and located on a yeast artificial chromosome (YAC) in S. cerevisiae. We report that homologous recombination (HR) and nonhomologous end-joining (NHEJ) repair pathways as well as the Mre11/Rad50/Xrs2 (MRX) complex cooperate to repair breaks that occur at these long repeats. In addition, we found that MRE11-deficient strains accumulate large expansions that mimic the size range seen in the triplet repeat diseases, and which are caused by a combination of aberrant HR and inefficient SAE2-dependent processing. We conclude that DSBs must occur frequently within long repeat tracts and that DSB repair is a critical mechanism for healing those breaks in a manner that prevents repeat instability.
MATERIALS AND METHODS
Yeast strains and YACs:
Most haploid yeast strains used were the BY4742 (MATα, haploid) background, and single gene deletions were obtained from the Research Genetics yeast deletion set. The wild-type (WT), sae2Δ, and mre11D56N strains were in the isogenic BY4705 MATα background. The MRE11 gene was replaced with the mre11D56N allele by a two-step gene replacement procedure (Alani et al. 1987) using plasmid pSM444 (Llorente and Symington 2004): transformed yeast strains were plated onto YC −Ura plates to score for Ura+ integrants, Ura+ clones were grown in YEPD overnight at 30° and plated on FOA plates to select for URA3 pop-outs, and then Ura− FOAR colonies were patched onto YEPD +0.035% MMS plates and scored for MMS sensitivity to identify clones that carried only the nuclease-deficient allele of MRE11. The replacement of wild-type MRE11 with the mre11D56N nuclease-deficient allele was verified by sequencing. The mre11Δrad52Δ double mutant was generated by crossing the MATα BY4742 mre11Δ∷KANMX4 and the MATa BY4741 rad52Δ∷KANMX4 strains. The RAD52 gene was deleted in dnl4Δ∷KANMX4 and sae2Δ∷KANMX4 backgrounds using a targeted one-step gene disruption approach with a HIS3MX6 fragment obtained from the pFA6a plasmid to generate the double mutants (Baudin et al. 1993; Longtine et al. 1998). The yeast artificial chromosomes (YACs) used in this study were identical to those used in Callahan et al. (2003), and consist of 41 kb of λ-DNA in the central region with ∼10 kb of non-λ DNA on each end (see Figure 1A) for a total of ∼61 kb (Schulz and Zakian 1994). YACs carrying the CAG-70, or CAG-155 repeat and a control YAC without the CAG tract (CAG-0) were introduced into the wild-type and each mutant background by cytoduction (Dutcher 1981; Callahan et al. 2003). For each experiment (CAG fragility, stability, and microcolony assays), at least two independent haploid cytoductants were used. See supporting information, Table S3 for a list of strains used in this study.
CAG/CTG repeat stability and fragility assays:
Yeast strains carrying the CAG tracts on YAC-CF1 were plated for single colonies onto YC −Ura −Leu plates, grown at 30° for 3 days, and tract length was determined by colony PCR (Callahan et al. 2003). For stability assays, a portion of a parent colony containing an intact CAG-70 repeat was used to inoculate a 1-ml YC −Leu culture, grown for 6.5–7.5 doublings, plated on YC −Leu media, and allowed to form daughter colonies (∼2 days) at 30°. Approximately 50 daughter colonies were analyzed for CAG repeat length by colony PCR, and the experiment was repeated at least three times for each strain studied. Due to limitations in PCR amplification of tracts greater than ∼200 repeats, this analysis was performed on the medium length CAG-70 tract. CAG-70 tracts were amplified from a portion of individual daughter colonies using Taq polymerase with primer pairs T720 and CTGrev2 flanking the CAG repeat (P1 and P2 in Figure 1A; sequences available upon request). PCR products were separated on a 2% Metaphor gel (Cambrex Bio Science Rockland) and sized. The frequency of repeat expansions and contractions in each strain background was calculated and statistical significance determined by Fisher's exact test (See Table S2 for raw data). Repeat lengths of up to ∼190 CAG repeats (an addition of ∼120 repeats to the starting CAG-70 repeat), with an accuracy of +/−3 repeats were obtained by this method (Figure 2; Figure S1).
For fragility assays, a colony with a repeat of desired length (CAG-0, CAG-70, or CAG-155) was used to inoculate 10 individual YC −Leu cultures and grown for 6.5–7.5 doublings at 30°. Dilutions were subsequently plated on 10 individual FOA −Leu plates to select for colonies that had lost or mutated the URA3 gene and were thus 5-FOA resistant (FOAR), and on YC −Leu plates to obtain a total cell count. Mutation rate, i.e., rate of FOAR, was calculated using the method of maximum likelihood (Zheng 2002) and statistical significance determined by a pooled variance t-test using Systat software. YAC structure for a subset of FOAR colonies in each strain background was determined by Southern hybridization (data not shown).
Microcolony experiment to determine the survival frequency of strains carrying CAG/CTG repeats:
Single, unbudded, normal-sized cells carrying either the CAG-70 or CAG-155 repeat tract or a no-tract control were isolated on YC −Leu plates using a micromanipulator, and their growth into microcolonies was recorded. Cells that failed to grow into microcolonies, defined as obtaining a size of ≤0.001 mm2 area after 30 hr of growth, were designated nonsurvivors. The area measurement was performed using National Institutes of Health ImageJ software of pictures taken at 10× magnification using an Olympus microscope. Approximately 40 cells were analyzed per strain (range, 22–73). Cells that showed no evidence of growth or division were excluded from the experiment since they could have been damaged during micromanipulation. Statistical significance of survival differences between strains with and without a CAG repeat was analyzed by Fisher's exact test.
To determine the role of DNA repair in maintaining repeat stability, we monitored expansions, contractions, and chromosomal breakage events at expanded CAG/CTG repeats (abbreviated here as CAG repeats) contained on a 61-kb yeast artificial chromosome (YAC) in haploid yeast strains (Callahan et al. 2003). The YAC contained either no CAG repeat (CAG-0), a CAG-70 repeat, or a CAG-155 repeat, sizes that are within the unstable and disease-causing range for most known human loci. If lesions or DSBs within the repeats are recognized and subsequently repaired by either the HR or NHEJ machinery, deletion of genes central to these pathways may affect the fidelity of repair at the repeats. Hence, altered repeat stability (deviation from the expected expansion and contraction frequencies) would occur, which we measured by PCR using primers flanking the CAG repeat tract (Figure 1A). Additionally, some CAG-specific lesions may be inefficiently repaired in mutant backgrounds, resulting in a loss of DNA distal to the repeat tract, which contains the URA3 gene (referred to herein as fragility). A G4T4/C4A4 sequence proximal to the CAG repeat can act as an efficient seed for addition of a new telomere to rescue the broken chromosome. These events were recovered as Leu+ 5-flouroorotic acid-resistant (FOAR) colonies (Figure 1A). The rate of FOAR is not an absolute level of breakage, as it measures only those events that failed to heal by a conventional DSB repair pathway yet successfully healed by the telomere addition pathway. Since DSB repair is not required for rescue of the broken chromosome, unlike an earlier assay using CAG repeats located on yeast chromosome II (Freudenreich et al. 1998), the requirement for different repair pathways in healing repeat-mediated breaks can be assessed.
HR and NHEJ repair are required to prevent CAG/CTG repeat fragility and repeat contractions:
To determine which of the two principal pathways of DSB repair, HR or NHEJ, plays a more significant role in repairing repeat-mediated breaks, strains defective in each pathway were tested for repeat fragility and instability. The Dnl4 ligase (homologous to human Lig4) is central to the NHEJ pathway of DSB repair, which involves rejoining of DNA ends flanking a break using as little as 0–4 bp microhomology (Lees-Miller and Meek 2003). In the fragility assay, dnl4Δ strains carrying either the CAG-70 or -155 repeat-containing YAC showed rates of FOAR of 30 × 10−6 and 77 × 10−6, which represent 3.2- and 2.8-fold increases over wild-type rates, respectively (Figure 1B, Table S1). These results indicate that the Dnl4 ligase, and by extension NHEJ, plays a role in healing breaks that occur at the expanded repeats. In the PCR-based stability assay, the dnl4Δ strain showed an increase in both expansions and contractions of the CAG-70 tract with the 3.4-fold increase in contraction frequency being highly significant (Table 1). Therefore, in the absence of Dnl4, repair occurs less efficiently and by an error-prone pathway that favors contractions. The size and spectrum of repeat length changes in the dnl4Δ strain was similar to that of wild type (Figure 2B).
Homologous recombination, where the DSB is sealed using an intact homologous template, is the preferred pathway of repair during the S and G2 phases of the cell cycle. The strand invasion step of HR, which is important in the search for homology, is mediated by the Rad52, Rad51, and Rad54 subsets of proteins (West 2003; Krogh and Symington 2004). In the fragility assay, rad52Δ strains carrying the medium CAG-70 tract on the YAC showed a 2.7-fold increase in the rate of FOAR over wild type (to 25 × 10−6), and the longer CAG-155 tract showed a significant 3.3-fold increase (to 95 × 10−6; Figure 1B, Table S1). Note that the fold increase in FOAR is about the same for rad52Δ strains with and without the repeat (Figure 1B). This indicates that the Rad52 protein is not specific for repairing repeat-mediated breaks, which is not surprising since it is well established that Rad52 is a general repair factor that is important for the repair of spontaneous lesions throughout the genome. This pattern is true for most of the proteins studied herein. Nonetheless, a significant increase in fragility at the repeat compared to that observed in wild-type cells demonstrates that a Rad52-dependent process is important for repairing breaks within the repeat. The rad51Δ mutant exhibited repeat fragility rates strikingly similar to the rad52Δ strain, supporting a role for homology-driven strand invasion in repairing lesions at CAG tracts (Figure 1B). The increased repeat breakage in a rad54Δ strain further implicates the recombination process in maintaining CAG repeats (Figure 1B). Altogether, these results indicate that both HR and NHEJ pathways are involved in repairing the breaks that occur within a repetitive sequence.
To test whether the Dnl4- and Rad52-dependent pathways are operating independently to heal breaks that occur at the CAG repeat, we tested fragility in a rad52Δdnl4Δ mutant. Indeed, the rate of FOAR was additive in the double mutant compared to the dnl4Δ and rad52Δ single mutants for the CAG-70 construct (Figure 1B). Interestingly, the rad52Δdnl4Δ strain with the CAG-155 repeat was extremely slow growing and frequently accumulated a growth suppressor phenotype; thus we were unable to obtain a consistent FOAR rate. The dependence on these two proteins for normal viability in the presence of a long repeat lends additional support for the conclusion that both HR and NHEJ pathways operate independently and are critical for healing repeat-induced breaks.
To determine whether the HR pathway was important in maintaining repeat stability, we measured frequencies of expansions and contractions in the HR mutant backgrounds. The CAG-70 tract showed a consistent increase in contractions of ∼2.5-fold in rad52Δ, rad51Δ, and rad54Δ strains (Table 1). Interestingly, we also observed a significant five- to sevenfold increase in expansion frequency relative to wild type (Table 1). The spectrum of repeat additions (+10 to +80 repeats with a median of +22 repeats) and deletions (−5 repeats to a deletion of the entire repeat) in the rad52Δ mutant revealed that large-scale repeat expansions and contractions can occur when HR is compromised (Figure 2C). To test whether these events were occurring during end joining, we tested repeat stability in a rad52Δdnl4Δ double mutant. Indeed, CAG-70 repeats in the double mutant showed a reduced repeat expansion frequency compared to a rad52Δ single mutant (Table 1), although the frequency was not reduced to wild-type levels (Table S1, P = 0.2). However contractions were not DNL4 dependent and were in fact significantly increased in the double mutant compared to the rad52Δ single mutant (Table 1; Table S2, P = 0.005). Therefore, somewhat surprisingly, although breaks that are evident in the absence of Rad52 protein may be healed by end joining, this process does not create the contractions observed. In summary, the high levels of instability and fragility in HR-deficient strains suggest that lesions within the CAG repeats that require Rad52-mediated HR for repair are fairly common. These observations are consistent with the expectation that HR repair might be crucial to repairing repeat-associated DNA damage, since homology is readily available in this context. Taken together, our results indicate that both HR and NHEJ DSB repair pathways are required to accurately repair lesions at expanded CAG repeats and prevent repeat contractions, expansions, and chromosomal breakage.
Deficiency of Mre11 results in increased CAG/CTG repeat fragility and expansion-prone repair:
The MRX complex is among the first proteins to bind to a DSB, where it has multiple functions that include regulation of DNA end processing, a structural role in bridging the ends, and a role as a sensor and mediator of the cellular checkpoint response (Williams et al. 2007). Absence of the Mre11 protein had a dramatic effect on repeat fragility, as mre11Δ strains carrying a CAG-70 or CAG-155 repeat showed 10-fold and 3-fold increases in the rate of YAC breakage over the wild type, respectively (Figure 1C, Table S1). Thus, presence of the Mre11 protein is crucial for repairing breaks at an expanded repeat sequence. Unlike the trend seen in the wild type where longer repeats have a higher rate of breakage, the CAG-70 repeat exhibited a rate of FOAR comparable to the CAG-155 repeat (93 × 10−6 and 84 × 10−6, respectively), suggesting a greater requirement for Mre11 at the CAG-70 repeat length relative to the number of breaks that are presumed to occur. Control experiments indicated that despite DSBs being processed more slowly in strains lacking a functional MRX complex (Sugawara and Haber 1992), the rate of FOAR obtained for the CAG-155 YAC is likely not underestimated relative to CAG-70 (see Figure 3 and File S1). Comparison of breakage rates among mre11Δ, dnl4Δ, rad52Δ, and mre11Δrad52Δ strains for CAG-155 (Figure 1, B–D) suggests that a primary role of Mre11 at the CAG-155 repeat is to facilitate NHEJ and/or HR double-strand break repair. However at CAG-70, Mre11 may have an additional role since the rate of FOAR for the mre11Δ strain was significantly higher than for either rad52Δ or dnl4Δ single mutants.
Strikingly, CAG/CTG repeats in the mre11Δ mutant showed a bias toward expansions with a dramatic 11-fold increase in the expansion frequency (P < 0.0003), the most seen in any repair mutant tested, while the contraction frequency remained similar to wild-type levels (Table 1). Large repeat additions of up to +70 repeats (median, +30 repeats) were observed, underscoring the significance of the MRX complex in limiting the size and frequency of CAG expansions (Figure 2D). The expansion phenotypes reported for mre11Δ and rad52Δ strains are specific to expanded CAG repeats, as expansions of the endogenous (CAG)6 repeat at the KIN1 locus or nonrepeat sequences adjacent to the CAG-70 repeat showed no instability in these backgrounds (78–96 colonies tested; data not shown).
To better understand the role of Mre11 in preventing CAG expansions and fragility, we tested mutants that abolish or attenuate the various functions of the MRX complex. Mre11 has an asymmetric hairpin loop cleaving activity (Paull and Gellert 1998; Trujillo and Sung 2001) and the E. coli MRX homolog, SbcCD, has been shown to cleave CAG/CTG hairpins in vitro (Connelly et al. 1999). A point mutation in the phosphoesterase motif II of MRE11, mre11D56N, which eliminates the nuclease activity (Moreau et al. 1999), was used to determine whether the ability of Mre11 to cleave hairpins was important in either creating an initial nick or repairing breaks subsequent to their formation at CAG repeats. A modest but tract-specific 1.7-fold increase in repeat fragility at CAG-70 was observed in the mre11D56N mutant, but fragility rates at CAG-155 remained similar to wild type (Figure 1C). We conclude that the Mre11 nuclease may have a minor role in repairing lesions at CAG-70, for example by cleaving secondary structures that arise at a nick or break to facilitate ligation, but that the majority of lesions can be resolved by other means. CAG-70 repeats in the mre11D56N mutant showed expansion and contraction frequencies similar to those of wild type, indicating that the nuclease activity of the MRX complex is not required to prevent CAG repeat instability (Table 1). Since the large-scale CAG expansions seen in the mre11Δ strain were not observed in the mre11D56N mutant that retains the ability to form a functional MRX complex in vivo (Krogh et al. 2005), we conclude that the integrity of the complex rather than its nuclease activity is most critical for preventing CAG repeat instability.
In addition to its intrinsic nuclease activity, the Mre11 protein also interacts with the Sae2 endonuclease, and specifically stimulates its cleavage of the single-stranded region at the base of hairpin-capped ends (Lengsfeld et al. 2007). To test whether the function of the MRX complex at CAG repeats was to stimulate Sae2-dependent hairpin cleavage, we tested a sae2Δ strain. Fragility was increased about twofold relative to wild type, indicating that Sae2-dependent cleavage likely plays a role in processing ends so that breaks may be efficiently healed. Intriguingly, like the mre11D56N mutation, the increase was somewhat tract specific (2.1-fold for CAG-70 vs. 1.8-fold for CAG-0). Yet comparison of the rate of FOAR in sae2Δ and mre11Δ strains indicates that this role cannot be the only (or even primary) function of Mre11 at CAG repeats. Repeat expansions were also selectively increased in the sae2Δ strain (5.9-fold), although not as much as in the mre11Δ strain, and the size of the expansions were smaller (+5 to +30 repeats, median of +15) compared to those in the mre11Δ strain (Table 1, Figure 2E).
Exo1 is a 5′ to 3′ exonuclease that is involved in DSB processing during meiosis and mitosis (Llorente and Symington 2004). Because the Exo1 nuclease collaborates with Mre11 in end resection at a DSB (Mimitou and Symington 2008; Zhu et al. 2008), its role at CAG repeats was determined. CAG fragility was significantly increased in the exo1Δ mutant, and similar to the mre11Δ strain, fragility was increased to a greater extent for CAG-70 (3.5-fold), compared to CAG-155 (1.8-fold, Figure 1C, Table S1). In the stability assay, the exo1Δ mutant showed a 4.5-fold increase in the frequency of CAG expansions with no changes in contraction frequency compared to the wild type (Table 1). This selective increase in CAG expansions is also similar to what was observed in the mre11Δ strain, suggesting that these two share a related function at CAG repeats, such as processing of DSB ends (Mimitou and Symington 2008; Zhu et al. 2008). We attempted to test this hypothesis by creating an exo1Δ mre11Δ double mutant, but although the CAG-0 strain was viable, we were unable to obtain the double mutant with an expanded repeat tract, suggesting that it may be inviable. These results are all consistent with a role for the Exo1 nuclease in CAG/CTG lesion repair.
Since the MRX complex is known to mediate checkpoint signaling by forming a complex with Tel1/ATM, we tested a tel1Δ strain to determine whether a Tel1-dependent checkpoint-signaling role of MRX was important in preventing repeat instability. No significant change in repeat stability compared to wild type was observed in the tel1Δ mutant (Table 1). In addition, repeat fragility rates for the CAG-155 YAC remained similar to wild type in the tel1Δ strain, although a modest 1.7-fold increase in repeat fragility was observed for CAG-70 (Figure 1C). On the basis of the magnitude of the increase in fragility rates obtained in the tel1Δ and mre11Δ strains, it can be inferred that the primary role of Mre11 at the CAG-70 tract is largely independent of its Tel1-associated checkpoint function. However, we showed in a previous study (Lahiri et al. 2004) that repeat fragility was increased in a mec1Δsml1Δ strain, therefore indicating that one function of Mre11 at CAG breaks may be to facilitate repair by signaling to Mec1.
Repeat expansions that occur in the absence of MRE11 can be generated by both homologous recombination and defects in hairpin processing:
To determine whether the expansions that occurred in the absence of the Mre11 protein were due to aberrant HR, an mre11Δ rad52Δ double mutant was made. About two-thirds of the expansions were dependent on the presence of the Rad52 protein, as expansions decreased from 8.6% in the mre11Δ single mutant to 2.6% in the mre11Δ rad52Δ double mutant (Table 1; P = 0.036, Table S2). These results suggest that the predominant pathway involved in rescuing lesions in an mre11Δ background is Rad52-dependent HR, and that this pathway is not working with fidelity in the context of a repeat sequence. The Rad52-dependent pathway appears to account for the large expansions that occur in the absence of MRE11, since these were selectively lost in the mre11Δ rad52Δ double mutant (Figure 2). On the basis of the sae2Δ results, the remaining expansions that occur in the absence of Mre11 could be due to a defect in recruiting Sae2 to process hairpin-capped ends. If this were true, then the expansions that occur in the sae2Δ background should not be dependent on Rad52-mediated HR. Indeed this was the case, as there were 4.7% expansions in sae2Δ cells and 6% in the sae2Δ rad52Δ strain (Table 1; Table S2, P = 0.77). Although these two pathways can account for the expansions that occur in the absence of MRE11, they still do not account for the full function of Mre11 at CAG repeats, because repeat fragility is still elevated in the mre11Δ background compared to a sae2Δ rad52Δ double mutant (Figure 1). Therefore, the MRX complex is likely playing an additional structural role that prevents chromosome fragility (see discussion).
Absence of MRE11 or RAD52 increases permanent arrest of cells containing an expanded CAG repeat tract:
Our previous results suggested that CAG/CTG repeats accumulate lesions that are capable of activating the DNA damage checkpoint (Lahiri et al. 2004). Therefore we wanted to determine whether the presence of structure-forming expanded CAG repeats decreased cell survival, and also whether any of the rates of FOAR determined in the fragility assay could be an underestimate due to failure of cells with a broken YAC to grow into a colony. Single cells containing a YAC either with or without an expanded CAG repeat tract were micromanipulated onto YC −Leu media and their growth monitored for 30 hr as they divided to form small colonies (microcolonies). Because no essential genes required for cell survival are located in close proximity to the repeats on the YAC, the growth differences observed in cells with or without the CAG tract can be attributed to a repeat-mediated effect. A bimodal distribution of growth was observed in the wild-type strain, with some microcolonies arresting within the first few divisions at an area of ≤0.001 mm2 (“nonsurvivors”), and the remainder dividing normally and/or arresting at later time points to attain microcolony sizes of ≥0.002 mm2 (“survivors”; analysis of this class will be presented elsewhere). In wild-type cells carrying a CAG-70 or CAG-155 repeat tract, presence of the repeat made no difference in the number of nonsurvivors that arrested within the initial few divisions (∼9%, Figure 3). Similarly, the frequency of initial arrests in dnl4Δ strains carrying the CAG repeats was similar to both the no-tract control and the wild type (∼10–15%, Figure 3). Therefore repeat-containing cells are not being selectively lost in these backgrounds. However, 30–35% of rad52Δ cells carrying either the CAG-70 or -155 repeat had an arrest within the first few divisions that persisted even at 30 hr, a 3.3-fold higher frequency of nonsurvivors than the CAG-0 control cells (Figure 3, P = 0.03). Also, 36–38% of the CAG repeat-containing mre11Δ cells arrested within the first few divisions, compared to 15% of the no-tract control cells (Figure 3, P = 0.002). This high proportion of initial arrests indicates that a significant fraction of cells incur a lesion within the CAG repeat that requires Mre11 or Rad52 for repair. In the absence of Mre11 or Rad52, many of these cells apparently fail to recover from repeat damage-induced checkpoint arrest. These results also indicate that the rates of fragility of the repeat-containing YACs compared to the no repeat control in the rad52Δ and mre11Δ backgrounds are likely an underestimate. However since the percentage of nonsurvivors is similar for CAG-70 and CAG-155 in each case, the fragility rates of the two tract lengths can be directly compared. More importantly, these results highlight a crucial role for both the MRX complex and HR in recruiting an appropriate repair response to damage at expanded CAG repeats.
Expanded CAG repeats have been shown to be sites of chromosomal DSBs in yeast (Freudenreich et al. 1998; Jankowski et al. 2000; Callahan et al. 2003) and to expand in both humans and model systems. In addition, they can interfere with nick repair and flap ligation (Spiro et al. 1999; Henricksen et al. 2000; Panigrahi et al. 2005), leading to the potential situation where an unrepaired nick is converted to a DSB through further processing or encounter with a replication fork. In this study we show that indeed, the presence of intact DSB repair pathways are important for preventing chromosome arm loss distal to an expanded CAG repeat. Surprisingly, despite the homology available on either side of a break within a repetitive sequence, both HR and NHEJ pathways contributed about equally to repairing breaks within the CAG repeat. In addition, we identified the Mre11 protein as having a critical role in preventing both repeat fragility and expansions. Indeed, we found that when key DSB repair proteins are absent, repeat instability was uniformly increased, implicating proper DSB repair as important in maintaining the length of repetitive tracts of DNA. Our results are likely also relevant for understanding the repair pathways that operate at other complex structure-forming lesions that occur in genomes.
When increased fragility is observed in a mutant background in our assay, it can be explained in one of the two ways. First, the protein may be important for repairing breaks that normally occur within the repeat independent of that protein. In the absence of the protein, the breaks may persist and be repaired by alternative pathways with less fidelity, leading to expansions or contractions (Figure 4, B and C). If repair is inefficient or the end distal to the repeat is lost, the break will be resected and healed by telomere addition, resulting in an increase in FOAR in our assay (“CAG fragility”) (Figure 4C). Second, the absence of that protein may increase the chance that a stalled or reversed fork or a nick will be converted to a DSB (Figure 4A). In this case, absence of the protein could also lead to an increase in FOAR, and again expansions or contractions could occur during repair. We cannot formally distinguish between these two possibilities. However, monitoring both the fidelity of repair (expansions, contractions) and efficiency of repair (fragility, cell survival) and making comparisons between the different strains and between single and double mutants, allows for some conclusions to be drawn about mechanisms governing repair at CAG repeats.
HR and NHEJ repair pathways are both important for repairing breaks within repetitive DNA and protecting against repeat contractions:
Interestingly, a defect in either HR or NHEJ led to increased levels of contractions, with absence of the Dnl4 ligase resulting in a particularly strong bias toward contractions. Since contractions persisted in the dnl4Δrad52Δ double mutant, we conclude that they occur by a Rad52-independent mechanism. One candidate is microhomology-mediated end joining (MMEJ) (McVey and Lee 2008). Dnl4 provides both scaffolding and ligase functions that are critical for NHEJ repair. Zhang et al. (2007) reported that binding of the Ku heterodimer to DSB ends is less stable in the absence of Dnl4, thereby leading to enhanced end resection, which could explain the bias toward repeat contractions in the dnl4Δ strain (Figure 4B, left arrow). In support of this model, mutation of the Exo1 or Mre11 nucleases that are involved in DSB end resection did not increase repeat contractions. Repeat contractions are also observed in other situations where DSB repair is delayed or compromised, such as in HR-deficient strains (this study) and in strains with mutations in DNA damage checkpoint proteins implicated in the recruitment of DSB repair factors (Freudenreich and Lahiri 2004; Lahiri et al. 2004). If hairpins are at the very end of the break, they would inhibit ligation (Henricksen et al. 2000; Liu et al. 2004), leading to increased CAG fragility (Figure 4C) as was observed in dnl4Δ as well as rad52Δ and mre11Δ mutants. Altogether, our results are most consistent with end resection at a long-lived DSB as a primary mechanism for generating contractions. Since the size and spectrum of contractions observed in the mutants were similar to wild type, the same mechanism may occur in wild-type yeast, although at a lower frequency.
We found that rad52Δ, rad51Δ, and rad54Δ mutants all displayed elevated frequencies of fragility, showing that the Rad52-mediated HR pathway plays a central role in repairing breaks within repetitive DNA. The importance of Rad52-mediated repair at CAG repeats is further underscored by the repeat-specific growth defect observed in the rad52Δ background (Figure 3). In addition to repairing DSBs, Rad51/Rad52-mediated strand invasion is known to play an important role in fork restart (Courcelle et al. 2001; Bjergbaek et al. 2005; Liberi et al. 2005; Gangavarapu et al. 2007); therefore, this is another mechanism by which the HR pathway could prevent the formation of breaks (Figure 4A, left; Figure 4C). On the basis of previous studies of replication through CAG repeats (Pelletier et al. 2003) and the greater requirement for checkpoint sensing by the fork restart protein Mrc1 at the longer CAG-155 tract (Freudenreich and Lahiri 2004), we propose that the CAG-155 repeat interferes with replication more strongly than CAG-70, thus requiring fork restart processes more frequently. The more significant requirement for the Rad52 protein in preventing fragility at the CAG-155 tract compared to CAG-70 is consistent with a role in facilitating fork restart, and this interpretation is also supported by the additive increase in CAG-155 fragility seen in the mre11Δrad52Δ strain, indicating an Mre11-independent function for Rad52.
In contrast to this study, some previous studies in yeast and mice with CAG/CTG repeats cloned in various chromosomal locations failed to observe increased repeat instability in the absence of RAD52 pathway members (Freudenreich et al. 1998; Miret et al. 1998; Schweitzer and Livingston 1999; Savouret et al. 2003). We observed a significant increase in repeat expansions and contractions in all three HR mutants tested, and contractions (yet not expansions) were also elevated twofold at the same locus when RAD52 was deleted in another strain background (R. Anand and C. H. Freudenreich, unpublished data). Thus, the exact level of instability is variable in this mutant and depends on both location and other factors, which we do not yet understand. Also unexpectedly, some of the expansions (more than half) that we observed in the absence of RAD52 may occur via an end-joining pathway, as they were eliminated in dnl4Δrad52Δ and mre11Δrad52Δ double mutants (Table 1). This result is intriguing because it could explain the greater bias toward repeat expansions in mammalian cells compared to yeast, since unlike yeast, mammalian cells have a greater bias toward end-joining than HR. Because a break is long lived in the rad52Δ strain with extensively resected 3′ single-stranded tails (Wang and Haber 2004), one mechanism by which CAG expansions could occur in this background is misalignment of such 3′ single-stranded ends followed by gap filling with slippage, as depicted in Figure 4B. We conclude that the breaks that are present in a rad52Δ background can often be healed with fidelity, but there are some situations, for example when more extensive resection occurs, that can lead to a repair event that produces expansions or contractions.
The Mre11 protein is required to heal breaks within CAG repeats and to prevent repeat expansions that occur by both recombination and processing defects in its absence:
The absence of MRE11 had a dramatic effect on CAG/CTG stability, resulting in an 11-fold increase in repeat expansions over wild type and high levels of repeat fragility (10-fold over wild type). In addition, repeat-containing cells lacking MRE11 were more than twice as likely to fail to divide than no-repeat control cells, indicating that breaks or secondary structures capable of interfering with normal division are quite frequent at expanded CAG repeats, and that repair of these lesions by Mre11 is vital for survival of cells with expanded repeats. Previous studies have shown that the Mre11 endonuclease and its E. coli homolog SbcCD can cleave hairpin structures (Connelly et al. 1999; Trujillo and Sung 2001) and Mre11 nuclease activity is required in vivo for the processing of hairpin-capped DSBs (Lobachev et al. 2002) and to facilitate gene conversion when the template contains a CAG repeat (Richard et al. 2000). However, we observed no increase in frequency or size of expansions in the nuclease-deficient mre11D56N background, and fragility was increased only ∼1.7-fold at CAG-70 and not at all at CAG-155. We conclude that the Mre11 nuclease activity plays a relatively minor role in facilitating repair of the lesions that occur at CAG-70 and that other functions of the MRX complex are more important for preventing repeat expansions and fragility.
Two-thirds of the CAG expansion frequency, including large-scale repeat additions (up to +70 repeats, median +30) observed in the mre11Δ strain were Rad52-dependent, indicating that aberrant recombination is a mechanism that can generate large-scale expansions. The conclusion that repeat length changes can occur during HR is consistent with observations in yeast and E. coli, which showed that CAG/CTG expansions and contractions can occur during Rad52- and RecA-dependent recombination (Jakupciak and Wells 1999, 2000; Richard et al. 1999; Nag et al. 2004), with results in yeast that showed repair of an endonuclease-induced break during gene conversion using a homologous CAG/CTG repeat-containing donor template led to high levels of expansions and contractions (Richard et al. 2000, 2003), and with recent results that unrestrained recombination in an srs2Δ strain leads to repeat expansions (Kerrest et al. 2009). The MRX complex has a structure that allows it to bind and align DNA ends, and the Rad50 coiled-coil domain provides a scaffold that can stabilize DSB ends or tether sister chromatids during DSB repair or replication (Williams et al. 2007). We propose that in the presence of the MRX complex, breaks at CAG repeats will be repaired by Rad52-dependent strand invasion events such as gene conversion by synthesis-dependent strand annealing as previously proposed (Richard and Paques 2000); however in the absence of the MRX complex, loss of one of the two ends of a DSB will result in the formation of a one-ended DSB (Figure 4C). The very large increase in CAG fragility (FOAR) in the mre11Δ strain relative to the wild type suggests that one-ended DSBs occur frequently in this mutant, supporting this idea. One ended-DSBs could be rescued by Rad52-dependent strand invasion of a sister chromatid, which can occur out of register within the repeat, followed by break-induced replication (BIR) until the end of the chromosome (Krishna et al. 2007; Mirkin 2007; Aguilera and Gomez-Gonzalez 2008) (Figure 4C). In support of this model, studies have shown that the error-prone BIR mechanism contributes significantly to repair of DSBs in the mre11Δ strain (Li and Heyer 2008), and the Rad52-dependent rescue of lesions observed in mre11Δ, although expansion prone, was able to prevent a significant degree of CAG-155 fragility (this study, Figure 1D). The lesser increase of fragility in the mre11Δ rad52Δ CAG-70 strain compared to the mre11Δ single mutant could either be because of noise in the assay or a decreased efficiency of BIR at this repeat due to unprocessed hairpin-capped ends (see below). In addition, presence of the Mre11 protein is likely important for preventing collapse of forks stalled or reversed at the CAG repeat tract (Kerrest et al. 2009), as Mre11 has been shown to prevent accumulation of DSB ends during replication and promote replication fork restart in Xenopus extracts (Costanzo et al. 2001; Trenz et al. 2006), and it locates to and stabilizes replisomes stalled by hydroxyurea treatment in yeast (Tittel-Elmer et al. 2009). Thus, the large increase in fragility in the mre11Δ strain is likely a reflection of roles both in preventing fork collapse and in repairing breaks that have occurred by other means. In summary, we conclude that the MRX complex is critical for both prevention and repair of DSBs within the expanded CAG tract, likely by providing a scaffold that prevents end loss and aberrant HR repair events that lead to CAG repeat expansions.
The remaining one-third of expansions that are observed in the absence of MRE11 were due to Rad52-independent repair events. Since another function of Mre11 is to stimulate Sae2 cleavage of hairpin-capped ends, and the increased expansions observed in a sae2Δ mutant were not Rad52 dependent, some of these expansions are likely due to a defect in Sae2 activity in the absence of Mre11. If a hairpin has been formed by strand displacement, either during replication or repair, ligation of the nick or a subsequent double-strand break without hairpin removal will result in an expansion (Figure 4A, right). Consistent with this model, the expansions that occurred in the sae2Δ and mre11Δrad52Δ mutants were smaller, with a median of +15 and +22 repeats, respectively. On the basis of its known role in DSB end processing (Mimitou and Symington 2008; Zhu et al. 2008), and the similar expansion and fragility profile of the exo1Δ mutation to both mre11Δ and sae2Δ strains, the Exo1 protein likely also aids in end processing at nicks or breaks within CAG repeats to allow ligation and prevent expansions. Interestingly, the MRX complex appeared to have a role separate from its function in facilitating HR or NHEJ at CAG-70, and its nuclease function had some role (albeit minor) in preventing CAG-70 fragility that was not evident at CAG-155. On the basis of all these results, it appears that ends that require processing by a nuclease may be more common at a CAG-70 tract compared to the longer CAG-155 tract. Altogether, our observations support a model whereby the MRX complex prevents smaller-scale CAG expansions by facilitating end processing by Sae2, and larger-scale expansions by restraining aberrant recombinational repair of DSBs within the repeats.
In summary, we found that the MRX complex and the HR and NHEJ pathways cooperate to repair breaks that occur in repetitive sequences and to prevent their instability. Therefore, DSB repair pathways normally provide a protective function at repeat sequences as they do at nonrepetitive sequences. However, in the absence of any one pathway, the compensatory pathways do not act with fidelity within repetitive DNA, and repeat expansions or contractions result. In particular, we propose that CAG expansions can occur either during aberrant HR-mediated strand invasion processes, by end joining combined with gap filling and slippage or due to a defect in processing of hairpin-capped ends by the Sae2 nuclease. A primary mechanism for contractions is likely end resection at a long-lived DSB followed by annealing. Finally, efficient DSB repair is required to prevent repeat-specific cytotoxicity. The use of mitotically dividing yeast cells to study the effect of DSB repair pathways on repeat stability and cell survival can be extrapolated to the conditions occurring in human premeiotic germ cells, early embryos, or dividing somatic cells, where both replication and repair processes are active and where repeat expansions are observed. Other types of genomic instability that are frequently observed in cancers such as segmental duplications and interstitial and terminal deletions may occur by similar mechanisms.
We thank Lorraine Symington for the gift of plasmid pSM444; Ranjith Anand for sharing unpublished data on instability in rad52Δ strains; and Mitch McVey, Sergei Mirkin, and Carol Kumamoto for helpful comments. R.S. performed experiments and data analyses for all strains except tel1Δ (R.M.Z.), sae2Δ, and sae2Δ rad52Δ (L.G.).This research was supported by National Institutes of Health grant GM063066 and a Tufts University faculty research award (to C.H.F.).
Supporting information is available online at http://www.genetics.org/cgi/content/full/genetics.109.111039/DC1.
↵1 Present address: Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720.
Communicating editor: N. M. Hollingsworth
- Received October 16, 2009.
- Accepted November 4, 2009.
- Copyright © 2010 by the Genetics Society of America