The two pathways for DSB repair differ in their mechanisms and enzymatic requirements. The repair of DSBs via NHEJ is characterized by binding of the Ku70/80 heterodimer to the broken ends followed by recruitment of the Mre11-Rad50-Xrs2 (MRX) complex (Figure 1). The main function of the Ku70/80 complex is to protect DNA ends against nucleolytic degradation and to recruit additional NHEJ proteins. Meanwhile, MRX complex bridges the ends and prevents their separation. In some cases, the ends require removal of damaged nucleotides to allow conjugation by NHEJ-specific DNA ligase IV and its associated factor, Lif1 (reviewed in [8]; Figure 1).

In the second DSB repair pathway, homologous recombination (Figure 1), the broken ends need to be nucleolytically processed to produce 3’ single-stranded DNA (ssDNA) overhangs. During HR, MRX complex binding to the broken ends catalyzes removal of short oligonucleotides from the 5’ ends in collaboration with Sae2 nuclease. Two alternative pathways then extensively process the short 3’ overhangs. One is characterized by an action of Exo1 (5’–3’ exonuclease), while the other is dependent on the activities of Dna2 endonuclease and the Sgs1-Top3-Rmi1 complex [9]. During resection, the resulting ssDNA strand is rapidly bound by replication protein A (RPA), which not only promotes end resection but also prevents formation of secondary structures. In addition, RPA creates a barrier for the binding of Rad51 recombinase, thus prohibiting the formation of Rad51 presynaptic filament. To overcome this inhibitory effect, the action of recombination mediators, in particular Rad52 and the Rad55–57 complex, is required [10]. These assemble Rad51 on the RPA-coated ssDNA and promote the formation of Rad51 filament. Upon Rad51 filament formation, HR continues by searching for a homologous sequence, followed by DNA-strand invasion, and results in a D-loop formation. These reactions are catalyzed by Rad54 protein [11].

The invading strand of the D-loop is at this point extended by a polymerase, which is followed by resolution in one of the three possible sub-pathways. The first pathway, designated double-strand break repair (DSBR, [12]), proceeds by capturing the second DSB end to the extended D-loop and formation of double Holliday junctions (dHJ) (Figure 1B). dHJs are then resolved into either crossover or non-crossover products characteristic of meiotic recombination (recombination in meiosis will be discussed later). The second pathway, synthesis-dependent strand annealing (SDSA, [13]), is characterized by displacement of the extended strand from the D-loop and its consequent annealing with the complementary strand of the other resected DSB end (Figure 1A). This pathway eliminates formation of crossovers and is therefore typical for the repair of DSBs during mitosis. In the third alternative pathway, the D-loop assembles into a full-fledged replication fork in a process called break-induced replication (BIR) (Figure 1C).This mechanism can lead to a loss of heterozygosity and is also often used to repair broken or shortened telomeres [14]. For further details about HR, see additional review articles [15,16,17,18].

DSBs can alternatively be repaired by a mechanism known as single-strand annealing (SSA). SSA is used for DSB repair if directly repeated DNA sequences are present (Figure 1D). After the resection of DSBs, the generated single-stranded DNA overhangs can anneal to the complementary DNA strand with the help of Rad52 and Rad59 proteins. The 3’ non-homologous tails, produced during annealing as an intermediate, are removed by the Rad1-Rad10 endonuclease followed by gap filling and ligation (reviewed in [17]).

Sumoylation is a post-translational modification characterized by an attachment of SUMO (small ubiquitin-like modifier) peptide to target proteins. SUMO, known as Smt3 in S. cerevisiae, is an 11 kDa protein that modifies many proteins participating in diverse cellular processes, including DNA repair, replication, basic metabolism, gene transcription, ion and protein transport, and others [19,20,21,22] The conjugation of SUMO to a target protein involves a 3-step mechanism analogous to ubiquitylation (reviewed in [19,20,21,22]). It is initiated by an ATP-dependent activation of the SUMO protein by SUMO-activating enzyme (E1), a heterodimer consisting of Aos1 and Uba2. SUMO is then transferred to SUMO-conjugating enzyme (E2), Ubc9, which catalyzes the conjugation of SUMO to a target protein. To accomplish efficient and specific sumoylation of the substrate, however, the presence of SUMO ligases (E3) is usually necessary [20,23]. To date, four SUMO E3 ligases have been identified in budding yeast: Siz1, Siz2, Mms21, and meiosis specific Zip3 [20,23,24,25,26]. Attachment of SUMO can block interactions occurring at or near the attachment site, or, more often, it provides a binding surface for new protein interactions or stimulates the existing ones. In the latter case, the binding protein contains a SUMO-specific binding site, known as SIM (SUMO-interacting motif). The SIM has been shown to play essential role in multi-step enzymatic processes, and affect the assembly and disassembly of dimeric and multimeric protein complexes (reviewed in [27]). The crucial component of SIM is a hydrophobic amino acid core, which provides an interface for non-covalent interaction with SUMO. However, also position of acidic residues juxtaposed to SIM can further stimulate SIM-SUMO interaction [27,28,29,30,31]. Alternatively, negative charge in the SIM can be introduced by phosphorylation of serine or threonine residues and can additionally lead to increased selectivity of the protein interaction [30,32,33]. At functional level, SUMO has been shown to change protein interaction, localization, stability or activity [19,20,21]. Importantly, sumoylation is a reversible process and SUMO can be rapidly deconjugated from the target protein by the action of SUMO-specific proteases (Ulp1 and Ulp2 in yeast), which makes sumoylation ideal for regulatory purposes [19].

As mentioned above, SUMO modification has been implicated as a possible key player in the regulation of DSB repair. Mutations or deletions of the components of SUMO machinery lead to severe defects, including recombination abnormalities, thus implicating sumoylation as a potential regulator of recombinational repair [25,34,35,36,37]. Ubc9 mutant cells as well as cells carrying a SUMO ligase-deficient allele of MMS21 exhibit increased sensitivity to DNA-damaging agents [25,35]. Moreover, during replication of damaged template both mutants accumulate cruciform structures in a Rad51-dependent manner [34]. In addition, a strong hyper-recombination phenotype has been observed in a mutant of SUMO protease Ulp1. This mutant has also been shown to be synthetically lethal with mutations in genes involved in HR (such as srs2, rad51, rad52, rad54, rad55, rad50, and mre11) [37].

Recently, a breakthrough study from the Zhao laboratory has revealed a comprehensive role of sumoylation in maintaining genome stability [38]. Using a biochemical screen in yeast, they identified a large group of proteins participating in DNA repair and undergoing sumoylation, mainly in response to DNA damage. This revelation greatly broadens the potential roles of sumoylation in genome maintenance. Additional discussion on this work will be described in another review article by Zhao et al. in the next special issue focusing on “DNA damage response”.

In HR, the spectrum of SUMO-modified proteins includes all steps indicating SUMO’s substantial role in HR regulation (see Table 1). However, the exact role of sumoylation in HR regulation is often elusive, due to the problems in identification of modified sites and their possible redundancy. Identification of the downstream SUMO-interacting partners, analysis of sumo-deficient alleles as well as permanent sumo-fusion of target proteins that can only partially mimic the effect of sumoylation further hinder the task.

Nevertheless, the available data of selected examples described below show the diversity of the effects of SUMO modification, including its ability to regulate the intracellular localization, stability, and conformation of target protein as well as their interactions or biochemical activities. Future studies will be needed to uncover the molecular mechanism and biological function of sumoylation in HR.

The role of SUMO in DSB repair is not restricted to HR, as proteins involved in NHEJ undergo sumoylation as well. Interestingly, a SIM-containing peptide has been found to inhibit NHEJ in humans cell lines, though, the underlying mechanism remains unknown [63]. Ku70 protein, which forms a heterodimer with Ku80, has been shown to be sumoylated in both yeast and humans and can also interact with the SIM-containing peptide after radiation [25,63]. Since the main role of Ku70 is to recognize and protect DSBs as well as load other NHEJ factors, it is possible that SUMO-SIM mediated interaction can either affect the dynamics of the recruitment of additional NHEJ proteins or regulate the removal of the Ku70/80 heterodimer from DNA ends. Also sumoylation of Lif1 and MRX complex was recently observed, but the biological function is not known [38]. Interestingly, the sumoylation of NHEJ factors (Ku70, Ku80 and Lif1) was not influenced by deletion of Mre11 in contrast to proteins involved in recombinational repair, suggesting that resection contributes to the HR proteins sumoylation induction [38]. Therefore it might be intriguing to speculate if sumoylation might play important role in pathway choice between NHEJ–HR, or consequent amplification of the decision signal.

The rDNA is localized in the nucleolus and in S. cerevisiae is formed by 150 tandem repeats encoding the 35S and 5S ribosomal RNAs. Sumoylation plays an essential role in regulation of rDNA recombination as the loss of sumoylation severely impairs rDNA stability [71,72]. The important role of SUMO is further suggested by the striking localization of sumoylated proteins in the nucleolus, when SUMO deconjugation is blocked [73]. One main target of SUMO regulation of HR in nucleolus is the recombination mediator Rad52. In wild type cells, sumoylated Rad52 is excluded from the nucleolus and the DSB is repaired in the nucleoplasm. In contrast, SUMO-deficient mutant of Rad52 forms foci within the nucleolus resulting in hyperrecombination at the rDNA locus and rDNA marker loss [72]. Rad52 sumoylation therefore seems to specifically inhibit formation of Rad52 foci in the nucleolus in order to preserve rDNA integrity. Similar phenotype have been also observed in the Smc5–Smc6 mutant cells [72]. Smc5-Smc6 (structural maintenance of chromosomes) heterodimer forms a core of a multimeric complex containing also six non-Smc elements (Nse1–Nse6) [74]. The complex regulates sister chromatid cohesion, HR, chromatin structure and its dynamics, however, many aspects of its function still remain unclear [74,75]. Smc5–Smc6 is highly enriched at the rDNA and other repetitive sequences and is required for their proper segregation. The complex associates with E3 SUMO ligase Mms21 (Nse2), which activity is required for the integrity of repetitive sequences [25,76]. However, Mms21 in the nucleolus probably targets other factors than Rad52, as Rad52 sumoylation is independent of the Smc5–Smc6 complex and they rather act synergistically in limiting recombination at rDNA [72]. Another protein complex that is enriched in the nucleolus and is involved in the rDNA recombination regulation is the Slx5–Slx8 SUMO-targeted ubiquitin ligase. The slx8Δ cells exhibit increased rDNA recombination and nucleolar Rad52 foci formation [67]. The slx8Δ cells also accumulate Smt3 foci in the nucleolus, suggesting the role of Slx5–Slx8 in proteasomal degradation of sumoylated nucleolar proteins [71]. Altogether these findings demonstrate SUMO’s central role in controlling HR at the rDNA locus. However, further studies are required to determine specificity of the DNA-damage response in nucleolus, other nucleolus-specific SUMO targets, dynamics and molecular mechanism of protein re-localization.

Telomeres-the structures located at the ends of chromosomes, are vital for the stability and complete replication of the genome. In S. cerevisiae they consist of 300 base pairs of telomeric repeats terminated by short 3’ single-stranded overhangs. Their resemblance to DSBs necessitates specific set of proteins that protects telomeres from recognition as DSBs and against exonucleolytic cleavage [77].

SUMO is an important regulator of the telomeres and many telomeric proteins undergo sumoylation [78]. Sumoylation limits telomere length both in budding and fission yeast [25,78,79,80,81]. A main target of sumoylation is the Cdc13 protein, an important telomerase regulator. Siz-dependent sumoylation of Cdc13 strengthens its interaction with the telomerase inhibitor Stn1 and thus suppresses telomerase function [78]. Siz2 was also shown to sumoylate Ku70/80 and Sir4 proteins and thus stimulate anchoring of telomeres to the nuclear envelope [82]. The observations that Siz2Δ mutant displays telomerase–dependent telomere extension and elongating telomeres shift away from the nuclear envelope, led to the hypothesis that sumoylation represses telomerase by tethering telomeres to the nuclear periphery, whereas their release is connected to telomere elongation [82]. The importance of sumoylation in directing telomeres to nuclear envelope is also supported by impaired telomere clustering in sumoylation deficient Mms21 cells [25]. Though the mechanism of SUMO-dependent telomere anchoring to the nuclear envelope remains elusive, it possibly involves multiple SUMO–SIM interactions, as both structures are associated with profound sumoylation. The Smc5–Smc6 complex is enriched at the telomeres in the budding and fission yeast [83,84,85]. The whole complex as well as Mms21 activity are particularly important for telomere maintenance in telomerase deficient cells, as their absence cause accumulation of HR intermediates at telomeres, aberrant recombination between sister telomeres and growth termination [86,87].

Contradictory to SUMO’s role in telomere length restriction, SUMO was also found to play an important role in telomere length increase in the telomerase-deficient cells. Though in most cells lacking telomerase the telomere length gradually decreases and finally leads to cell cycle arrest [88], some cells are able to maintain the telomere length by employing recombination-mediated pathways, which are in humans referred to as alternative lengthening of telomeres (ALT) [89]. Sumoylation of Sgs1 was shown to specifically promote telomeric recombination in telomerase-deficient cells [60]. Similarly, sumoylation of Rqh1, the RecQ homologue in S. pombe, stimulates ALT-like recombination events in the taz1Δ and taz1Δ trt1Δ cells [61]. Though the mechanism by which sumoylation of Sgs1 and Rqh1 increases the telomeric recombination is unclear, it was suggested that sumoylation mediates localization of the RecQ helicases to telomeres where they facilitate restart of collapsed replication forks that subsequently lead to telomere-telomere recombination [60,61]. Whether sumoylation of human RecQ homologues has a similar importance in the telomeric recombination remains to be determined. However, conserved sumoylation and the role in telomere maintenance among RecQ helicases indicate their roles may be maintained [68,69,90].

The situation observed in the telomerase-deficient yeasts resembles the one occurring in human ALT cancer cell, and importantly in both sumoylation plays a central role in the promoting of telomeric recombination. Similarly to the absence of telomerase in the tlc1Δ yeasts, telomerase is downregulated in most human cells. However, the cancer cells are able to elongate their telomeres and achieve unlimited replicative potential either by upregulating telomerase transcription (85% of cancers), or by the ALT mechanism using recombination between telomeres (15% of cancers) [89,91,92]. The telomeres of the ALT cells usually associate with PML (promyelocytic leukemia) nuclear bodies, which are in this context referred to as ALT-associated PML bodies (APBs) [93]. SUMO plays a central role in the PML bodies as sumoylation and subsequent noncovalent SUMO–SIM interaction between PML subunits is absolutely necessary for PML body formation [94,95]. The association of telomeres and APBs is thought to facilitate recombination between telomeric repetitive sequences, as APBs are known to contain various HR proteins and artificially created APBs cause telomere elongation by a DNA repair mechanism [96,97]. The Smc5–Smc6 complex is also localized in APBs and is required for the telomere–PML colocalization [98]. The necessity of Mms21-dependent sumoylation of telomere-binding proteins for APB formation [98] suggests that the telomere–PML interaction may be stimulated by multiple noncovalent SUMO–SIM interactions, as is the case other PML-interacting partners [68,99,100]. The observation that Smc5–Smc6 depletion inhibits HR at telomeres and their elongation further supports its major role in ALT cells [98]. Moreover, the Smc5–Smc6 complex and Mms21 activity is also important for de-novo formation of PML bodies on telomeric DNA [96,101].

The above-mentioned roles of sumoylation in the telomere length restriction and elongation nicely illustrate how sumoylation of the same substrates occurring in different cellular conditions can lead to completely different outcomes.