DNA Repair in Haploid Context

DNA repair is a well-covered topic as alteration of genetic integrity underlies many pathological conditions and important transgenerational consequences. Surprisingly, the ploidy status is rarely considered although the presence of homologous chromosomes dramatically impacts the repair capacities of cells. This is especially important for the haploid gametes as they must transfer genetic information to the offspring. An understanding of the different mechanisms monitoring genetic integrity in this context is, therefore, essential as differences in repair pathways exist that differentiate the gamete’s role in transgenerational inheritance. Hence, the oocyte must have the most reliable repair capacity while sperm, produced in large numbers and from many differentiation steps, are expected to carry de novo variations. This review describes the main DNA repair pathways with a special emphasis on ploidy. Differences between Saccharomyces cerevisiae and Schizosaccharomyces pombe are especially useful to this aim as they can maintain a diploid and haploid life cycle respectively.


Introduction
Maintenance of genome integrity is a permanent cell challenge as both intra-and extracellular conditions can lead to chemical alterations of nucleotides or their sequence. Proper repair mechanisms have evolved so as to maintain a balance between maintenance of cellular function and adaptative processes improving fitness. For obvious reasons, germline cells must be especially proficient at this task as diversity must be transmitted while maintaining the gametes' integrity through the many differentiation steps. In addition, anisogamy, whereby the two sexes produced highly different and specialized haploid gametes, enhances diversity. Indeed, while diploidy allows gene versions compensation and error-free repair, haploidy is expected to create more variation as there is no templated repair for homologous recombination. Ploidy is therefore a key element to consider for genome plasticity or stability. The following review discusses the DNA repair and genome maintenance processes linked to the ploidy state by comparing two unicellular eukaryotes models, yeasts Saccharomyces cerevisiae (S. cerevisiae) and Schizosaccharomyces pombe (S. pombe), while considering the case of the mammalian gametes.

General Consideration Regarding Ploidy States
Since the turn of the century, knowledge regarding the advantages of different ploidy states has been increasing thanks to the study of lower eukaryotes. Thus, fission and budding yeasts have proved to be valuable models because they, respectively, represent the impact of haploid and diploid genomes on evolution and genetic integrity. The ploidy choice for lifespan appears to be strongly related to the evolution and maintenance of genetic integrity [1][2][3][4]. It is well known that diploidy confers greater fidelity to replication of both genomic and mitochondrial DNA, in particular, to attenuate single-nucleotide mutations. However, large chromosomal rearrangements or deletions occur more frequently in the diploid budding yeast impacting cellular fitness. Although there is twice the amount 2 of 26 of genetic material in diploid cells, therefore statistically twice as many possible mutations [5,6], observations show that there are only 1.4 times more mutations in diploid cells than in haploid cells [7]. Such a bias is perhaps linked to the faster replication in the haploid state allowing a much shorter repair time. Errors mainly involve AT or CG transversions in late replicating genes [8]. In addition to this mutation bias, acquisition of a recessive mutation that would clearly alter the cell survival or mutation load is reduced by diploidy because two such similar recessive mutations must happen over both alleles to alter the phenotype. However, such recessive mutations will be eliminated in the population, a process coined "heterozygote advantage".
On the other hand, this would prevent diploid cells from positively adapting to a changing environment [9]. Diploid cells can nevertheless accumulate heterozygous mutations, creating a pool of sequence variants that can be useful in reaction to a changing environment, whereas haploid cells either die or survive. Larger chromosome rearrangements in the diploid cells should be explained by the homologous recombination between chromosomes, which represents the main difference from haploid cells. However, the deletion of RDH54, a protein necessary for recombination between homologous chromosomes, does not alter the mutation rate of diploid cells [7]. This observation may be reconciled when considering the fission yeast S. Pombe, which maintains a haploid life cycle. S. pombe seems to be protected against the instability of the haploid life cycle by spending most of its time in the G2 phase. Indeed, in the G2 phase, sister chromatids are present allowing error-free repair of Double Strand Break (DSB), Homologous Recombination (HR). Although the alternative to HR is the error-prone Non-Homologous End-Joining (NHEJ). S. pombe promotes HR through stalling at the G2 phase. Deletion of NHEJ repair protein (PKu70 Sp or Lig4 Sp ) displays low sensitivity to DSB-inducing genotoxins, indicating that even outside the G2 phases, S. pombe may not rely on NHEJ for damage repair. However, these observations are in contrast to the higher mutation rate observed in haploid cells compared to diploids. For example, DNA microsatellites are 100 times more stable in diploid cells [10]. Another example is the higher rate of mutation found in mitochondrial DNA of haploid cells [7].
One potential explanation is the higher number of mitochondria in diploid cells while mitochondria often mate, mixing alleles versions, rapidly reducing heteroplasmia. In addition, polyploid cells evolve faster than haploids, in relation to chromosomal rearrangement and aberrant condition of polyploid cells [11]. It, therefore, appears that the difference in the amount of DNA between the ploidy states only partially explains the difference in the mutation rate. Although haploidy allows a faster evolution, this condition is protected by some elements preventing excessive genome instability. Similarly, anisogamy in mammals results in very numerous haploids spermatozoa versus a single diploid ovum before fertilization. As stated above, genetic stability of the ovum is provided by diploidy and a complete array of repair systems, while the haploid character of sperm and the condensed chromatin may provide the proper context for DNA repair errors at the origin of the well-known male bias in the transmission of de novo mutations. In that sense, the ovum stands as the guardian of heredity while spermatozoa are providers of diversity. Nucleotide excision repair (NER) resolves various substrates resulting in structural alterations of DNA helix ( Figure 1) [12]. NER can be divided into two sub-pathways: transcription-coupled NER (TC-NER) and global-genome NER (GG-NER). Although GG-NER uses self-assessment of DNA damage, TC-NER uses hindrance of the RNA polymerase transit to detect bulky adducts ( Figure 2). This transcription-coupled mechanism ensures the elimination of mutations during transcription by the so-called transcriptional mutagenesis. Whereas TC-NER is dedicated to transcribed strands of active genes, GG-NER detects DNA helix disturbing damages in both transcribed and non-transcribed strands [13].  . Transcription Coupled NER (TC-NER) starts with RNA pol II stalling in front of a lesion. Then CSB recruits CSA and UVSSA-USP7 allowing the TFIIH docking. Global Genome NER (GG-NER) is initiated by XPC lesion recognition. Through a cascade of events, TFIIH is recruited. Both NER pathways converge when TFIIH is recruited. Thanks to its helicases, XPB/XPD, and XPA, TFIIH creates a DNA repair bubble exposing the lesion. Then ERCC1-XPF and XPG cleave DNA around the lesion. The gap is then filled by DNA polymerase. In BER, glycosylase targets the lesion and cleaves base-deoxyribose glycosidic bounds. When Glycosylase is released from DNA, APE1 can now bind to cleave the faulty deoxyribose. DNA polymerase fills the gap and displaces some of the damaged strands. The resulting 3′ tail is then removed by FEN1. In MMR the mismatch is targeted by MutS, then recruiting MutL.  ase transit to detect bulky adducts ( Figure 2). This transcription-coupled mechanism ensures the elimination of mutations during transcription by the so-called transcriptional mutagenesis. Whereas TC-NER is dedicated to transcribed strands of active genes, GG-NER detects DNA helix disturbing damages in both transcribed and non-transcribed strands [13]. TC-NER is triggered by RNA polymerase blockage. RNA polymerases are exceptionally processive, and cover between 50 up to 90% of the genome depending on the cell's transcriptional state, providing a reliable scanning process for damage. In addition to bulky adducts, pol II arrest can result from a wide variety of potential hindrances ( Figure 1) including nicks, gaps, abasic sites, collisions with the replication complex, stable DNA-RNA dimers, or non-B DNA [14]. CSB Hs (Cockayne Syndrome B protein) is the first NER factor to be recruited. When the pol II stops, the CSB Hs -pol II interaction becomes stronger and CSB Hs wraps DNA around itself hence decreasing pol II-DNA interaction [15]. Then CSB Hs recruits CSA Hs (Cockayne Syndrome A protein) [16] at the lesion site allowing the docking of UVSSA-USP7 Hs . The latter is a ubiquitin ligase complex that activates CSA Hs , in contrast to CSB Hs which is ubiquitinated and targeted for degradation [17]. Following repair, USP7 Hs removes ubiquitination of all NER components so they can be recycled. The second step in TC-NER is done by the transcription factor TFIIH Hs , which binds DNA downstream of pol II Hs and provides an "assessment" of the lesion [18].
GG-NER targets larger lesions, typically more than one nucleotide in size, as this dramatically changes the DNA helix thermodynamics [19]. XPC-Rad23b Hs (XPC Hs complex, Xeroderma Pigmentum C protein) is the first complex to bind the DNA lesion. The XPC Hs complex may detect DNA lesions as they create a single strand local transition [20]. Other than the scanning mechanism, alternative recognition mechanisms by XPC Hs do exist [21][22][23][24]. XPC Hs SUMOylation, and re-ubiquitinoylation, are then required for GG-NER progression [25,26]. Once activated, XPC Hs complex recruits TFIIH Hs . TFIIH Hs can help XPC Hs to discriminate between lesion types since XPC Hs is easily misled by a base mismatch [27].
TC-NER and GG-NER repair pathways converge once TFIIH Hs is recruited. TFIIH Hs core subunits XPB Hs and XPD Hs create a "repair bubble" of 20-30 nucleotides surrounding the lesion [28]. A second proofreading process involves XPA Hs . XPA Hs normally stimulates the helicase activity of TFIIH Hs but can decrease this activity in the presence of a lesion. XPA Hs recruits RECCI-XPF Hs [29] a nuclease that will cleave 5 of the lesion [30]. From its interaction with the TFIIH Hs complex, XPG Hs endonuclease can then cleaves 3 of the lesion [31,32] and replaces XPC Hs . XPG Hs plays a structural role in complex with TFIIH Hs but acquires endonuclease activity once ERCC1-XPF Hs is recruited by XPA Hs 5 to the lesion. ERCC1-XPF Hs is the first to nick DNA followed by XPG Hs [33]. After removal of the damaged nucleotide, DNA polymerases can fill the gap, while ligase seals the nick.

NER in Yeast
As for mammals, NER in yeasts is divided into two sub-pathways ( Figure 2). However, Sc. pombe NER is less documented as redundancy with UVDER (UV Damage Excision Repair, see below) leads to phenotypic confusion. In S. cerevisiae, TC-NER is triggered after RNA polymerase stops, activating either Rad26 Sc (CSB Hs homolog) or Rpd9 Sc , an RNA pol II Sc subunit-specific of budding yeast [34]. Rad26 Sc deletion decreases mRNA synthesis, in agreement with its transcription promoting activity over direct damage repair (8-oxoG, 3-MeA, abasic site) [35]. This may, however, lead to "transcriptional mutagenesis". If the lesion creates a major hindrance, Rad26 Sc then initiates chromatin remodelling and the recruitment of NER partners. An alternative TC-NER mechanism involves RNA pol II subunit Rpb9 Sc , which can also recruit the NER machinery. In Sc. pombe, Rhp26 Sp (CSB Hs homolog) deletion displays an important increase in UV sensitivity [36], demonstrating the prominent role of the TC-NER pathway in UV resistance. In S. cerevisiae, GG-NER starts, with the Rad23-Rad33 Sc (Rad23b Hs homologs) complex that can transit over DNA until a UV lesion is encountered. Rad4 Sc (XPC Hs homolog) is then recruited to initiate NER [37,38]. While TFIIH Sc is already positioned at the lesion in TC-NER, a control step is needed in GG-NER. This control step uses RPA Sc (Replication protein A), which binds intact DNA, and Rad14 Sc (XPA Hs homolog) which binds damaged DNA [39]. Without such a control step, NER will abort before the final dual incision [40]. GG-NER in Sc. pombe uses two XPC Hs homologs: Rhp41 Sp and Rhp42 Sp [41,42]. Whereas Rhp41 Sp has a major role in Cyclo-Pyrimidine Dimers (CPDs) removal in both transcribed and non-transcribed genes, Rhp42 Sp is only active over non-transcribed genes [41][42][43]. Following damage detection by either GG-NER or TC-NER, Rad2 Sc /Rad13 Sp (XPG Hs homologs) then cleaves 3 of the lesion, whereas Rad1 Sc -Rad10 Sc /Rhp16 Sp -Swi10 Sp (ERCC1-XPF Hs homologs) cleave in 5 , respectively in S. cerevisiae and in Sc. pombe [44][45][46].

General Mechanism
Base excision repair (BER) is the DNA damage repair pathway in charge of resolving chromatin lesions not associated with structural alterations of the DNA helix [47]. BER mainly recognizes oxidized base (8-oxoG), deaminated base (uracil), and alkylated base ( Figure 1). BER also takes charge of the abasic sites. Apuric or apyrimidic sites (AP) are indeed the most frequent lesions encountered in a stationary state [48,49]. BER pathway uses glycosylases which target abnormal bases and cleave β glycosidic bonds between riboses and bases. Glycosylases are generally evolutionary conserved, from bacteria to higher eukaryotes [50]. In mitochondria, BER also exists and uses different isoforms, resulting in the most robust DNA repair in mitochondria [51,52].
Glycosylases are generally small proteins harbouring a unique catalytic domain with high substrate specificity. Although most of these enzymes have unique glycosylase activity (monofunctional), some also have β lyase, or β/∆ lyase activities able to cleave in 3 or 5 P, or both (bi-trifunctional) [51,52]. Glycosylases first bind the minor groove, inducing a kink in DNA, then flip the faulty base out of the helix and separate the base from the ribose [53]. Irrespective of their host organism, glycosylases are fast to catalyse the glycosidic bond cleavage, but then disengage from DNA with much slower kinetics [54,55]. This dual kinetic mode may therefore hide AP sites, which can lead to apoptosis when reaching a given threshold. If the glycosylase is monofunctional, AP site-specific endonuclease will assist in cleaving 3 of the deoxyribose, and DNA pol will remove the 5 deoxyribose phosphate (dRP) during DNA neosynthesis [51,56,57]. On the other hand, if the glycosylase also possesses a lyase activity, this will result in DNA nicking 5 to the damaged deoxyribose. The 3 dRP is then processed by AP endonuclease, leaving a functional 3 OH, the initiation site for DNA neosynthesis, and a 5 -diphosphate (5 -PP) thereby facilitating DNA pol binding ( Figure 2) [58]. BER comprises two sub pathways: short and long patch repair. Either short or long patch repair pathways are used by the cell depending on the lesion size or the glycosylase involved. One single nucleotide is excised in short patch repair whereas long patch BER involves two or more nucleotides [59]. The main BER pathway is usually the short patch as it does not require the replication machinery typical of long patch repair. The short patch process uses glycosylase, the AP endonuclease, DNA polymerase, and DNA ligase. Long patch repair is mainly used during replication and uses a glycosylase and the AP endonuclease, but usually relies on DNA polymerase, PCNA (Proliferating Cell Nuclear Antigen), flap endonuclease from the replication machinery, as well as DNA ligase [51]. In a long patch repair pathway, the basic catalytic pocket of DNA polymerase can contain some ssDNA while the repair synthesis bypasses the lesion site creating a 5 overhang tail with the old DNA. The Flap endonuclease (mammalian FEN1 Hs , S. cerevisiae Rad27 Sc , Sc. pombe Rad2 Sp ), which possesses a 5 flap endonuclease activity and 5 to 3 exonuclease activity, will then remove this 5 tail [60].

BER in Yeast
Six DNA glycosylases have been characterized in S. cerevisiae: monofunctionals, Ung1 Sc and Mag1 Sc , Mag2 Sc ; and bifunctionals, Ngt1 Sc , Ngt2 Sc , and Ogg1 Sc . In Sc. Pombe, monofunctionals, Ung1 Sp , Thp1 Sp , Mag1 Sp , Mag2 Sp , and Myh1 Sp , and the bifunctional, Nth1 Sp have been characterized (Table 1). Ung1 Sc , Ung1 Sp , and Thp1 Sp are Uracil DNA Glycosylases (UDG) recognising uracil resulting from cytosine deamination [61]. Thp1 Sp also recognizes T:G or U:G mismatches, but is also known to process 5-fluorouracil, 3,N4ethanocytosine and 5-hydroxyuracil [62]. In addition, Thp1 Sp recognizes deaminated purines (xanthine and oxanine from guanine, hypoxanthine from adenine) [63]. S. cerevisiae's Mag1 Sc and Mag2 Sc are expressed after exposure to alkylating agents [64]. Mag1 Sc is also able to cleave normal bases, usually guanine [65]. In Sc. Pombe, Mag1 Sp and Mag2 Sp share 44,8% identity. Although Mag1 Sp initiates the alkylation repair, Mag2 Sp may also plays a role in the process [66]. Unlike S. cerevisiae Mag1 Sc , deletion of Mag1 Sp only slightly increases alkylating agents' sensitivity [67]. However, Rad13 Sp , Rad16 Sp or Rhp51 Sp deletion mutants are hypersensitive to alkylation, suggesting an important contribution of NER or HR to this type of damage in Sc. pombe [67,68]. Although, mutants of both Sc. pombe BER endonucleases, Apn2 Sp or Nth1 Sp , are hypersensitive to alkylation, confirming BER in alkylation repair [69][70][71]. Surprisingly, Mag1 Sp deletion leads to a 3-fold increase in intrachromosomic recombinations [68], suggesting that HR is somehow involved in alkylation repair in Sc. pombe. In S. cerevisiae Ntg1 Sc is a bifunctional glycosylase located in the mitochondrion whereas Ntg2 Sc is nucleolar. However, only Ntg1 Sc is upregulated during oxidative stress [72]. These enzymes can process a wide variety of the oxidized pyrimidine and purine lesions [73]. However, Ntg1 Sc is the sole enzyme able to resolve 8-oxoG whereas Ogg1 Sc processes only Fapy-G and 7,8 dihydro-8-oxoG [74]. Surprisingly, an Ogg1 Sc homolog is not found in Sc. Pombe [75]. Instead, Nth1 Sp is in charge of resolving 8-oxoG in a mismatch situation. Nth1 Sp seems to process a wide variety of substrates including 8-hydroxycytosine, thimidineglycol, and 8-hydroxyuracil [76]. Due to its bifunctional activity, Nth1 Sp acts upstream of the Apn2 Sp AP endonuclease, as it leaves a 3 protected end, which is removed by the phosphoesterase activity of Apn2 Sp [69]. In fission yeast, Nth1 Sp also resolves AP sites resulting from the activity of both Mag1 Sp and Mag2 Sp , therefore playing a central role in BER [71]. However, in contrast to mammals NTHL1 Hs , or S. cerevisiae Ntg1 Sc , Nth1 Sp is seemingly absent from mitochondrion. Finally, Myh1 Sp , is a glycosylase specific to Sc. pombe and in charge of mismatches found at adenines (A:G, A:8-oxoG, 2-aminopurine:G, 2-aminopurine:A) [77]. Myh1 Sp may also process G:8-oxoG, preventing C:G to G:C transversion [78]. A prominent role for Myh1 Sp is binding to PCNA Sp and the PCNA-like heterotrimer Rad9 Sp /Rad10 Sp /Hus1 Sp [79,80], which are involved in DNA lesion checkpoint. After base eviction, the AP site is processed by the two AP endonucleases found in yeasts. Apn1 Sc and Apn2 Sc are the two AP endonucleases found in S. cerevisiae although Apn1 Sc is responsible for >95% of the total in vivo activity [81]. Apn1 Sc cleaves 5 to the lesion and possess a 3 phospho-esterase activity able to resolve 3 dRP, 3 phospho-glycolate (3 PGA) or 3 P 3 dirty ends. Both S. cerevisiae AP endonucleases are 3 tyrosyl DNA phospho-esterase with a weak 3 exonuclease activity so they can resolve Topo1 Sc -DNA adducts [82]. In Sc. pombe, Apn2 Sp is the major AP endonuclease. Unlike Apn2 Sp , Apn1 Sp does not need specific subdomains to process PCNA Sp or Top3 Sp catalytic hindrance, Apn1 Sp may represent a proper backup to Apn2 Sp at 3 blocked end, but to a limited extent as Apn1 Sp is present in both nucleus and cytoplasm, unlike Apn2 Sp [83]. Uve1 Sp , an UVDER enzyme, may also be involved in the processing of AP sites [84], especially as it is present in both the nucleus and mitochondrion [85].

General Mechanism
DNA polymerases can induce from 10 −5 to~10 −4 mismatch per base given their own exonucleolytic proof-reading activity. The mismatch repair (MMR) system increases repair fidelity by about~100-1000 times [86]. The MMR pathway is in charge of mismatches but also resolves intra-strand loops. The MMR pathway was first described in Escherichia coli (E. coli) which represented a key model to decipher the sequential events of MMR: MutS first detects the mismatch, MutL confirms the helix deformation induced by the mismatch and MutH endonuclease cleaves the faulty base. DNA resection then occurs, followed by DNA polymerase gap filing and nick sealing by DNA ligase (Figure 2). While eukaryotic MutS and MutL exist as heterodimers [87,88], the MMR-specific MutH endonuclease is seemingly absent in eukaryotes and some bacteria. Lesion recognition by MMR is achieved by two heterodimeric homologs: MutSα and MutSβ. Whereas MutSα is in charge of small mismatches and helix extrusions of about 1-4 nt in size, MutSβ exclusively recognizes larger extrusions of 2-10 nt [89]. Loss of MutSβ can easily be compensated for by MutSα. Lesion recognition by MutSα is indeed prominent in the MMR pathway [90]. Once the MutS complex becomes bound to DNA, it recruits MutLα or MutLβ. During recombination, MMR proteins act to repair mismatch in DNA heteroduplexes, remove 3 non-complementary tails produced during resection, and prevent recombination between divergent sequences. MutLγ and MutLβ have been shown to play a role in meiotic HR [91]. MutLγ is mainly used during meiotic HR, where MutSγ activates it. In the mismatch repair, the endonucleolytic cleavage is done over the weakest strand, admitting the one containing a mismatch, in both 3 and 5 flanking the lesion. The mechanism directing strand selection is still unknown but may involve PCNA, which is still bound to the neosynthetised strand or may result from hypomethylation of the new strand [92]. Although a eukaryotic MutH has yet to be identified, the human PMS2 Hs , part of MutLα Hs , have been shown to possess a latent endonuclease activity. Removing its endonuclease motif abolishes the mitotic function of MutLα Hs [93]. After incision, a fragment of DNA is excised by Exo1 Hs . Resection by Exo1 Hs is greatly enhanced by association with MutSα Hs and ATP [94], as it removes only a few nucleotides when acting independently. After resection, DNA polymerase fills in the gap while DNA ligase ligates the 3 end to complete the process. An alternative pathway to MMR involved DNA polymerase, whereby the faulty strand is displaced during DNA synthesis [95].

MMR in Yeast
In S. cerevisiae, while MutSα Sc targets mismatches, MutSβ Sc is mainly in charge of long insertion-deletion loops (IDLs), although redundancy exists between both as MutSβ Sc may substitute for defective MutSα Sc but to a limited extent [96]. MutSα Sc initiates repair of all mismatches except C:C [97]. IDLs are repaired more efficiently than mismatches as MutS functional redundancy facilitates the repair of small IDLs [98,99]. Sc. pombe possesses four MutS homologs (Msh1 Sp , Msh2 Sp , Swi4 Sp , Msh6 Sp ) [100]. MutSα Sp (Msh2 Sp -Msh6 Sp ) is sufficient to support Sc. pombe MMR, regardless of the lesion type. However, Swi4 Sp , likely acting in complex with Msh2 Sp , is involved in mating-type switching and some recombination processes, as demonstrated for MutSβ Sc in S. cerevisiae [101,102]. Nevertheless, although Sc. pombe MutSβ Sp is solely implicated in recombination, the MMR function, typical of the S. cerevisiae MutSβ Sc , has apparently been lost. The MutSγ Sc complex in S. cerevisiae is only expressed during meiosis and is linked to recombination [103,104]. In yeast, MutLγ Sc is the main Holliday Junctions resolvase [105], symmetrically nicking Holiday Junctions. MutL homologs in budding yeast includes MutLα Sc (Mlh1 Sc -Pms1 Sc ), MutLβ Sc (Mlh1 Sc -Mlh2 Sc ) and MutLγ Sc (Mlh1 Sc -Mlh3 Sc ) [106]. While MutLα Sc interacts with MutSα Sc and MutSβ Sc to process most of MMR, MutLγ Sc only interacts with MutSγ Sc to resolve meiotic crossing-over. Sc. pombe contains only one MutL homolog (Mlh1 Sp -Pms1 Sp ) [100]. Thus, MutSα Sp and MutLα Sp should be sufficient for a functional MMR machinery in fission yeast, as it seems to be the case in Drosophila and Caenorhabditis. In S. cerevisiae, Exo1 Sc mutants were found to be slightly mutagenic, owing to its "error-free damage bypass" function, as described by Tan et al. [107]. Considering the weak impact of Exo1 Sc deletion, one may surmise that other nucleases may be involved such as Rad27 Sc or from the proofreading activity of DNA polymerases ∆ or ε Sc [107,108]. In contrast, Exo1 Sp deletion in Sc. pombe generates a strong mutator phenotype, probably as a result of its role in GT repeat maintenance in an MMR-independent pathway [109].
A short patch MMR pathway is described in fission yeast and involves the NER machinery [110]. Indeed, Rhp14 Sp , Swi10 Sp , Rad13 Sp , Rad15 Sp , and Rad16 Sp are important NER effectors components in this short patch repair pathway [111][112][113]. The NER machinery may therefore play a broader role and supports MMR pathways in specific cases. The first clear evidence is the NER requirement for C:C mismatches, which cannot be repaired by MMR [111,113]. Hence, in the absence of MMR, C:C mismatches are quickly repaired, so NER seemingly processes such damage with high efficiency. Interestingly, the NER Rhp14 Sp binds to various mismatches with similar affinity to C:C mismatches [112]. Defective NER has also been shown to induce a slight increase in reversion of GT repeats, suggesting a role for NER in Sc. pombe microsatellites maintenance [112]. Short patch MMR exists in S. cerevisiae but is seemingly independent of NER [114]. In S. cerevisiae, NER mutants show an increase in reversion rate but are mainly associated with flanking nucleotides substitutions, suggesting a more complex mechanism, likely originating from a bypass by the error-prone DNA pol ζ [115]. In vitro evidence indicated that NER may also play a role in mismatch repair in other species including drosophila [116,117], E. coli, and humans [118].

DSB Regulation in Mammals
In humans, about 1 double-strand break (DSB) arises in 10 8 base pairs, so around 10-30 DSBs are generated per cell, per cell cycle, resulting from endogenous or exogenous factors [119,120]. The two main DSBs repair pathways are Homologous Recombination (HR) and Non-Homologous End-Joining (NHEJ). These pathways are quite contrasted in terms of quality and timing of repair during the cell cycle. HR relies on sister chromatids to repair DNA, providing an almost error-free mechanism. HR is therefore dependent on the cell cycle, being most active during S-G2 phases. In contrast, NHEJ is an error-prone system, inducing sequence variations, and is mainly active during G1-G2 phases, although its components are found during the whole cell cycle [121,122]. Whereas HR needs hours to complete, NHEJ can repair DSBs within half an hour given the high processivity of its enzymes [123]. Upon inactivation of NHEJ (mutation of Ku, Lig4, or XRCC4 proteins), HR becomes the prominent DSBs repair pathway [124]. If sequence homologies however exist at DSBs ends, other repair pathways may be used such as Single Strand Annealing (SSA), Micro-homology End-Joining (MMEJ), and an Alternative NHEJ (Alt-NHEJ) pathway. SSA requires at least a repetition of two nucleotides to be present at both ends of the DSB. However, the two-nucleotides homology can also be used by MMEJ or alt-NHEJ. MMEJ is frequently used even though HR proteins are present or in cases of defective HR [125]. A choice must be made between NHEJ, HR, SSA, MMEJ, and alt-NEHJ ( Figure 2). Initially, the DSB may undergo a direct repair by NHEJ, or resection may occur. If resection is chosen, then a competition takes place between HR, SSA, MMEJ, and alt-NHEJ depending on resection size [126][127][128]. Resection is inhibited either by the NHEJ Ku Hs binding at DNA ends or by both the p53 binding protein 1 (53BP1 Hs ) or the Replication timing regulatory Factor 1 (RIF1 Hs ), which are known to be resection antagonists [129]. Many such down regulators of resection lead to the predominance of NHEJ. If a resection still occurs, a competition between other repair systems begins. The shift between HR and SSA, MMEJ, and Alt-NHEJ is dependent on the cell cycle, and particularly the cyclin-dependent kinases (CDK Hs ). Indeed, during S or G2 phases, CDK Hs can phosphorylate CtIP Hs (C-terminalbinding protein-interacting-protein) thereby facilitating its interaction with BRCA1 Hs to promote HR [130]. In the case of resection, it is done before the S phase and DSB repair will be achieved by SSA or MMEJ [131]. Although most of the MMEJ promoting factors also promote SSA repair [132], SSA will be preferred for smaller resection sizes. Alternative NHEJ also needs a micro-homology of 2-20 nt [133].

DSB Regulation in Yeast
Yeast regulation of DSBs repair is relatively similar to that of mammals. However, in this case, the presence of sister chromatid is not the main factor in choosing between HR and NHEJ. Indeed, HR is regulated through the cell cycle by post-translational modifications, degradation, and re-localization of its effectors. In S. cerevisiae, some HR genes are even regulated by the presence of DNA damage itself [134,135]. In contrast, YKu70 Sc is mainly transcribed in early G1 and almost undetectable in the S phase. In post-replicative cells, HR proteins are downregulated by Clb-CDK-inhibitors Sc or Sic1 Sc overexpression [136]. A few minutes after DSB formation, YKu Sc heterodimers and MRX complexes become recruited independently [137]. MRX complexes are essential to maintain both DNA ends physically closed and protected [138], whereas YKu Sc inhibits DNA end resection [139]. As in mammals, HR-NHEJ balance is determined by resection initiation, with YKu Sc competing with MRX complex and Sae2 Sc endonuclease (CtIP Hs homolog) [140]. Moreover, MRX is ambivalent regarding YKu Sc . It seems to stabilise YKu Sc on DSBs, but also destabilising it [141]. In addition, YKu Sc dissociation from unrepaired DSBs is dependent on the formation of the MRX complex, suggesting that the latter can complete the NHEJ repair process [137]. However, YKu Sc is able to prevent extensive resections in the absence of either the MRX complex or Sae2 Sc , while the MRX complex is dispensable to initiate HR in the absence of YKu Sc since Exo1 Sc will carry out resection [139]. In the particular case of the fission yeast, despite being a haploid organism, it spends most of its time in the G2 phase, where sister chromatids are present allowing preferential selection of HR. Thus, in contrast to mammals, HR is the major DSBs repair pathway in yeast. Unlike NHEJ mutants (PKu70 Sp or lig4 Sp mutants), HR mutants were shown to be much more sensitive to genotoxins [142]. However, in G1 synchronised cells, NHEJ is 7 to 10 times more important than in asynchronous cells, and PKu70 Sp mutants become far more sensitive to ionising radiations [143]. This demonstrates the primary role of NHEJ in G1 phasis when sister chromatids are absent although HR remains the predominant DSBs repair pathway in yeast. However, loss of PKu70 Sp increases chromosomal rearrangements from DSBs without changing the gene conversion ratio, suggesting competition between HR and NHEJ despite the lack of PKu70 Sp [144]. In contrast to S. cerevisiae, MRX Sp mutation (deletion of Rad32 Sp or Rad50 Sp ) does not alter NHEJ activity [145]. Moreover, Rad32-Rad50 Sp double mutants are unable to proceed beyond the S-phase DNA damage checkpoint [146] showing a clear prevalence of HR. In S. cerevisiae, some proteins become expressed in order to promote HR. The rad55-Rad57 Sc complex is also found to promote HR in S. cerevisiae [147]. In S. pombe, two Rad52 Sc homologs are known: Rad22 Sp and Rti1 Sp . Rad22 Sp is required for mating-type switching and DNA repair [148] and shares high sequence homology with Rad52 Hs [143]. By itself, Rad22 Sp is able to bind DSB. Rti1 Sp mutant increases sensitivity to ionising radiations as much as in Rti1/Rad22 Sp [149,150] double mutants. Thus, although NHEJ, SSA, MMEJ, and Alt-NHEJ do exist in yeast, HR remains the major repair mechanism regardless of the ploidy state. Details of DSB repair processes are provided in the following sections.

General Mechanism
The first step in HR is the formation of a single strand 3 tail by resection from the original DSB (Figure 3). This initial step is mediated by the Mre11 Hs -Rad50 Hs -NBS1 Hs (MRN Hs ) complex and CtIP Hs in humans. The 3 tail resulting from large 5 ->3 DNA resection is performed by Exo1 Hs [151]. The main characteristic of HR is the single strand invasion over the sister chromatid. Rad51 Hs is the RecA bacterial orthologue in eukaryotes with a DNA-dependent ATPase activity able to form a nucleoprotein-DNA filament [152]. HR proteins form supersized structures containing hundreds to thousands of proteins. Rad51 Hs may bind B-DNA formed by trinucleotide repeats in a single strand 3 tail. These triplets could then form Watson-Crick interactions with complementary dsDNA present in sister chromatids. After ATP hydrolysis, Rad51 Hs releases a stabilized triplex DNA. By competing with Rad51 Hs , RPA Hs (Replication Protein A) binds to single-stranded DNA with high affinity providing protection against nucleases [153]. Strand invasion can lead to crossover resolution and or to non-crossover resolution (patch) sequence exchange at Holiday junctions [154]. tion is performed by Exo1 Hs [151]. The main characteristic of HR is the single strand invasion over the sister chromatid. Rad51 Hs is the RecA bacterial orthologue in eukaryotes with a DNA-dependent ATPase activity able to form a nucleoprotein-DNA filament [152]. HR proteins form supersized structures containing hundreds to thousands of proteins. Rad51 Hs may bind B-DNA formed by trinucleotide repeats in a single strand 3′ tail. These triplets could then form Watson-Crick interactions with complementary dsDNA present in sister chromatids. After ATP hydrolysis, Rad51 Hs releases a stabilized triplex DNA. By competing with Rad51 Hs , RPA Hs (Replication Protein A) binds to single-stranded DNA with high affinity providing protection against nucleases [153]. Strand invasion can lead to crossover resolution and or to non-crossover resolution (patch) sequence exchange at Holiday junctions [154]. Artemis endonuclease is then recruited in order to clean dirty DNA ends, allowing the direct ligation by Ligase IV. Resection is carried out by exonuclease I (EXO1) but initiated by the MRN complex, assisted by CtIP. Single strand DNA is then protected by RPA. In HR, Rad52 will displace some RPA to allow the formation of Rad51 filament. Rad51 filament will find a homologous sequence and start strand invasion. DNA polymerase then fills gaps by using homologous chromatid as a template. Resolution of Holliday Junction leads to Chromosomal Exchange (CO) or Non-Chromosomal Exchange (NCO). SSA also uses Rad52 and Rad51, but the filaments target homologies surrounding the DSB. When hybridization occurs, unpaired 5′ overhangs are removed by ERCC1/XPF. Finally, polymerase fills the gap. Alt-NHEJ simply uses thermodynamics for annealing close to the DSB. Then DNA polymerase θ fills the gap.

HR in Yeast
As in mammals, HR is initiated by resection, which is a two steps process. Initially, a 50-200 nt resection is performed by the MRX Sc (Mre11 Sc /Rad50 Sc /Xrs1 Sc ) complex and Sae2 Sc in S. cerevisiae, and MRN Sp (Mre11/Rad50/Nbs1 Sp ) complex and Ctp1 Sp in Sc. pombe, which are respectively MRN Hs and CtIP Hs homologs [155,156]. Then, resection extends to kilobases through the action of Exo1 Sc or Sgs1 Sc -Top3 Sc -Rmi1 Sc (STR Sc ) complex with Dna2 Sc endonuclease in S. cerevisiae [157]. Exo1 Sp is likely the HR exonuclease in Sc. pombe, but Exo1 Sp deletion does not increase sensitivity to ionizing radiation, suggesting functional Artemis endonuclease is then recruited in order to clean dirty DNA ends, allowing the direct ligation by Ligase IV. Resection is carried out by exonuclease I (EXO1) but initiated by the MRN complex, assisted by CtIP. Single strand DNA is then protected by RPA. In HR, Rad52 will displace some RPA to allow the formation of Rad51 filament. Rad51 filament will find a homologous sequence and start strand invasion. DNA polymerase then fills gaps by using homologous chromatid as a template. Resolution of Holliday Junction leads to Chromosomal Exchange (CO) or Non-Chromosomal Exchange (NCO). SSA also uses Rad52 and Rad51, but the filaments target homologies surrounding the DSB. When hybridization occurs, unpaired 5 overhangs are removed by ERCC1/XPF. Finally, polymerase fills the gap. Alt-NHEJ simply uses thermodynamics for annealing close to the DSB. Then DNA polymerase θ fills the gap.

HR in Yeast
As in mammals, HR is initiated by resection, which is a two steps process. Initially, a 50-200 nt resection is performed by the MRX Sc (Mre11 Sc /Rad50 Sc /Xrs1 Sc ) complex and Sae2 Sc in S. cerevisiae, and MRN Sp (Mre11/Rad50/Nbs1 Sp ) complex and Ctp1 Sp in Sc. pombe, which are respectively MRN Hs and CtIP Hs homologs [155,156]. Then, resection extends to kilobases through the action of Exo1 Sc or Sgs1 Sc -Top3 Sc -Rmi1 Sc (STR Sc ) complex with Dna2 Sc endonuclease in S. cerevisiae [157]. Exo1 Sp is likely the HR exonuclease in Sc. pombe, but Exo1 Sp deletion does not increase sensitivity to ionizing radiation, suggesting functional redundancy. Exo1 Sp -Rad50 Sp double mutant displays enhanced sensitivity than individual mutant, showing distinct activities [156]. After resection, single-strand tails are coated by RPA homologs (RPA Sc in S. cerevisiae and Rad11 Sp in Sc. pombe) in order to protect them and prevent untimely Rad51 Hs homologs (Rad51 Sc in S. cerevisiae and Rhp51 Sp in Sc. pombe) binding [158]. In Sc. pombe, deletion of the Rad11 Sp gene is lethal, and the temperature-sensitive mutant (Rad11-404) shows high sensitivity to genotoxins [159]. In S. cerevisiae, Rad52 Sc displaces some RPA Sc molecules to allow Rad51 Sc binding to ssDNA and to initiate the nucleoprotein filament formation [160]. Rad52 Sc can be assisted by Rad51 Sc paralogues, including Rad55 Sc -Rad57 Sc , which has been shown to stabilize Rad51 Sc filaments [161]. The ssDNA invasion by Rhp51 Sp nucleofilaments, in Sc. pombe, results from a competition between Rhp51 Sp and Rad11 Sp for ssDNA binding. Deletion of the Rhp51 Sp gene leads to much higher sensitivity to ionizing radiations [162]. Rph51 Sp binds to ssDNA to form the nucleoprotein filaments in an ATP-dependent manner [163]. As in mammals, the Rad51 Sc /Rhp51 Sp filament promotes strand invasion to find a DNA duplex with sufficient homology to form a synapse. Although very complex, the process can take from 20 to 60 min to complete [164,165]. In Sc. pombe, HR is involved in processing stalled replication forks [166]. Not surprisingly, mutations of HR proteins lead to alteration of DNA replication. Moreover, Segurado et al. [167] have shown that DNA replication is associated with physical interactions and annealing between sister chromatids. Such structures are not found in Rad22 Sp (Rad52 Hs homolog), or Rhp51 Sp mutants, clearly linking HR to replication. Finally, Rhp51 Sp and Rad22 Sp foci induced by fork stalling, result in chromosomal rearrangements from an increase in recombination activity [168].
3.6. Non-Homologous End-Joining 3.6.1. General Mechanism In mammals, Non-Homologous End-Joining (NHEJ) is a DSBs repair pathway that allows direct ligation of DNA ends in an error-prone manner (Figure 3). NHEJ proceeds in three tightly regulated steps. First Ku70/80 Hs heterodimers bind to free DNA ends and act as protein hubs [169]. DNA-Pkc Hs (DNA dependant protein kinase catalytic subunit) is recruited given its high affinity for the Ku-DNA complex [170]. DNA-Pkc Hs undergo an auto-phosphorylation that recruits and activates other repair partners [171]. The second step consists of DNA end processing by the Artemis Hs endonuclease [172]. A third step involves XRCC4 Hs and XLF Hs (XRCC4-Like factor or Cernunnos Hs ), which interact together to form a small filament [173]. This allows interaction between the two DNA ends and the action of DNA ligase IV Hs . These canonical NHEJ proteins are sufficient to repair some DSBs, but a number of DSBs require more proteins including nucleases, DNA polymerases, and some functional homologs of XRCC4 Hs and XLF Hs [174].

General Mechanism
Single Stranded Annealing (SSA) arises when a DSB is formed between two homologous regions. Once the repair is completed, the sequence between repeats is lost. Repeats must be separated by at least 400 bp and up to 15 kbp [183]. The first step in SSA depends on CtIP Hs and Mre11 Hs resection (Figure 3). This resection is around 100 nt in length and could be targeted by MMEJ [184]. Once generated, a 3 single-strand tail is covered by RPA Hs and Rad51 Hs [185]. RPA Hs is progressively displaced to allow Rad51 Hs filaments formation. The scanning for strand homology is provided by the activity of Rad52 Hs . After annealing, the two 3 tails emerging from the hybridized region are targeted by ERCC1 Hs -XPF Hs for cleavage with the help of Rad52 Hs [186]. Gaps between newly hybridized and the original dsDNA are then filled by a combination of DNA polymerase and ligase [133].

SSA in Yeast
S. cerevisiae and Sc. pombe SSA involves a similar sequence of events: DNA resection, annealing, 3 overhang tail removal by Rad1 Sc -Rad10 Sc /Rad16 Sp -Swi10 Sp flap endonucle-ases ( Figure 3). Sc. pombe SSA is functionally linked to Rad16 Sp and Swi10 Sp [187], as well as Rad22 Sp but not to Rhp51 Sp , suggesting a specific SSA pathway in this model. However, in Sc. pombe, mutation of Rad16 Sp -Swi10 Sp results in a very small increase in sensitivity to ionizing radiations [188,189], suggesting that SSA is not a major repair pathway in this experimental model. UV-induced damages can also be resolved by an alternative pathway discovered in yeast Sc. pombe [190]. Since NER genes deletion, mutants of either Rad16-Swi10 Sp , Rad13 Sp , Rad15 Sp , or Rad16 Sp show unaltered sensitivity to UV, an alternative system has been proposed and termed "UV Damage Excision Repair" (UVDER) that relies on the UV Damage Endonuclease, UVDE. Uvde Sp enzyme can perform 5 incision just next to CPD, 6-4 photoproducts, or to Dewar isomers [191]. In contrast to photolyases or glycosylases, Uvde Sp recognises both CPD and 6-4 photoproducts and may process other substrates, as it is compatible with a nuclease activity reported over apurinic/apyrimidinic sites. However, the deletion of UVDE shows a weak increase in sensitivity to chemical genotoxins [192,193]. In Sc. pombe, the Rad2 Sp protein, a homolog of mammalian FEN1 Hs or S. cerevisiae Rad27 Sc , could be involved in a related NER alternative pathway. Indeed, Rad2 Sp mutants are far more sensitive to UV than Uvde Sp mutants whereas Uvde-Rad2 Sp double mutants are less sensitive than single Rad2 Sp mutants [194]. This is compatible with a mechanism whereby Rad2 Sp could help to process Uvde Sp nicked DNA, and where the lack of Rad2 Sp leads to a potentially lethal accumulation of Uvde Sp -lesions [195]. Whatever the activation mechanism, UVDER-induced lesions can be repaired by HR through the activity of both Rhp51 Sp and Rad18 Sp [196]. Moreover, it has been shown that Exo1 Sp endonuclease may also process Uvde Sp nicked intermediates, as Exo1 Sp and Exo1-Uvde Sp mutants have similar phenotypes [197].

Photolyases
In contrast to mammals and fission yeast, budding yeast possesses two unique DNA repair enzymes: Phr1 Sc and Mgt1 Sc . Phr1 Sc is a DNA photolyase, or a so-called CPD-specific photolyase [198,199]. It is produced in response to UV-C, or to exposure to alkylating agents [200]. Phr1 Sc is a monomeric protein with two prosthetic residues and uses near-UV photon (300-500 nm) to split off the cyclobutane ring of CPD, repairing DNA [201]. The second unique enzyme, Mgt1 Sc , is a methyltransferase, targeting O6-MeG and O4-MeT [202]. O6-MeG can mismatch with thymine whereas O4-MeT mismatches with a guanidine, leading to G:C->A:T and to A:T->G:C transversions. In a suicide reaction, Mgt1 Sc transfers the faulty methyl to its cysteine, leading to its inactivation [202,203]. Despite its repair abilities, Mgt1 Sc has a higher affinity for O6-MeG than O4-MeT, resulting in a bias in repair kinetics.

Alternative DNA Repair Approaches in Yeast
Despite their evolutionary distance, NER repair systems are relatively similar between the two yeast models. However, in Sc. pombe, the presence of UVDER conceals certain phenotypes of NER proteins mutants, suggesting a strong functional redundancy. However, this is nuanced in the presence of UV, because of S. cerevisiae Phr1 Sc photolyase which processes NER substrates and somewhat hides NER deficient phenotypes. The TC-NER sub-pathway is also favoured in S. cerevisiae by the Rpb9 Sc subunit of RNA polymerase, allowing rapid identification of RNA pol II arrests leading to redundancy in TC-NER initiation. On the other hand, Sc. pombe has two homologs of XPC Sp (Rhp41 Sp and Rhp42 Sp ). Although their targets are slightly different, this redundancy seems to indicate the greater importance of GG-NER.
The case of the BER system is more difficult to compare. Although the general process is similar, specificity in glycosylases creates subtle differences in processes. Thus, even if Sc. pombe has two UDG, against one in S. cerevisiae, it does not have Ogg1 Sc . Oddly enough, Ogg1 Hs is in charge of the most common damage, 8-oxoG, in mammals, but this does not appear to be the case in budding yeast. This useful function is performed by Ntg1 Sc /Ntg2 Sc and by Nth1 Sp , respectively in S. cerevisiae and Sc. pombe. The redundant Ntg Sc homologs in budding yeast are required because of their different subcellular localization and not to improve repair fidelity. Bifunctional glycosylases seem to be more efficient for cells as it reduces the number of steps in the BER process. Thus, S. cerevisiae has 3 bifunctional enzymes, compared to only one in fission yeast. However, Nth1 Sp , the bifunctional enzyme in Sc. pombe, is a key component of BER, apparently acting upstream of AP endonucleases. Regarding alkylation, although Sc. pombe has two glycosylases dedicated to this task, Mag1 Sc from S. cerevisiae seems much more efficient. Perhaps this is due to the prominent role of NER and HR in the repair of alkylation in fission yeast. However, in addition to having a more efficient enzyme, S. cerevisiae has a photolyase, Mgt1 Sc , specific to this damage, and provides a more efficient BER activity in this model. Sc. pombe has one more enzyme, Myh1 Sp , which is in charge of resolving mismatches in front of an adenine. This may be related to a simpler MMR mechanism in this model. Few differences seem to emerge regarding AP endonucleases, although abasic sites may require processing from activation of Uve1 Sp , in the Sc. pombe UVDER system to complete BER more efficiently.
As outlined previously, the MMR system in fission yeast seems much simpler considering that MutSβ Sp has lost its function in MMR and that MutLα Sp is the only homolog of MutL. This observation was also made in other species, such as Drosophila and Caenorhabditis, suggesting a lesser role for MMR in maintaining genomic integrity. On the other hand, Sc. pombe can rely on alternative pathways, for example using NER machinery. However, this is compensated for by the existence of short patches repair in S. cerevisiae, although it does not involve the machinery of NER.
The two yeasts respond to UV damage in different ways. Response to UV in S. cerevisiae is elegant because it uses the energy of the toxic UV, to repair the damage with Phr1 Sc photolyase. This however is too specific to CPDs, as UV can induce other damage such as 6-4 photoproducts and Dewar isomers. Sc. pombe adapted by using the UVDER system which can recognize all these different substrates. UVDER is so efficient that it compensates for a defective NER and can support BER in the elimination of abasic sites.
The HR system is predominant in both yeasts and appears to work in similar ways. However, the STR Sc complex confers a greater resection activity to S. cerevisiae. The filament also appears to be better stabilized by Rad51 Sc homologs. In Sc. pombe, stalls of the replication fork are repaired by HR, so it ensures proper replication in this model. The NHEJ systems are seemingly different between these two models. In S. cerevisiae, the MRX Sc complex seems very important to NHEJ. Since MRX Sc is known to direct resection, its role in NHEJ is rather unexpected. In S. cerevisiae, MRX Sc is used mainly to process dirty DNA ends, as Artemis Hs does in mammals. This suggests that resection is important in S. cerevisiae for repairing DSBs regardless of the repair mechanism. Recently, MRN Sp has been shown to influence the NHEJ system in Sc. pombe, albeit with limited action, as shown by mutants of Rad50 Sp , which do not change NHEJ repair capacity. It is likely that the lesser dependence on MRN Sp in Sc. pombe protects against extensive resection which, in a haploid cell, could be dangerous.
SSA systems of the two yeast models appear to also be different but have been studied to a much lesser extent. In fact, in Sc. pombe SSA was shown not to rely on Rhp51 Sp , the core protein of HR. However, since SSA does not depend on the presence of sister chromatids, its regulation is somehow opposite to the activities of HR proteins. An HRindependent pathway, like SSA, should therefore be more valuable in the haploid fission yeast. Unfortunately, as the activation of SSA is related to the presence of MRN Sp , but this protein is mostly expressed in the presence of sister chromatid, the moment where SSA competes with HR, SSA is relegated to a secondary system in fission yeast.

Gametes
Most vertebrates are anisogametic whereby gametes from the opposite sex are specialized and differ in morphology and number. They are therefore expected to respond differently to DNA damage inducers. This is especially the case for mammalian gametes where spermatozoa are ex-soma cells produced in large numbers whereas the oocyte is a resident cell. One additional major difference is the much greater level of DNA compaction provided by protamines in mature sperm following the eviction of most histones whereas oocytes maintain histone-based chromatin.
In addition to the greater number of cell divisions required to produce mature sperm, the peculiar steps of the sperm differentiation program may be more vulnerable to both exogenous and endogenous DNA damage inducers. For instance, endogenous DNA breaks may arise from the change in chromatin structure in post-meiotic spermatids coincident with a decreased DNA repair activity of these cells, whereas oocytes possess a far greater DNA repair capacity and harbour much fewer DNA breaks.

Spermatozoa
Male gametogenesis is a complex differentiation program producing mature spermatozoa from spermatogonia. During meiosis, HR takes place during the late phase of prophase I. At the beginning of prophase I, programmed DSBs allow parental alleles to recombine [204]. After meiosis, spermiogenesis undergoes a Golgi phase whereby spermatozoa head and axoneme are formed. Then chromatin is remodelled with Transition Proteins (TP1 and TP2) followed by protamination where histones are replaced by protamines (PRM1 and PRM2) to achieve the highest level of DNA compaction known to the eukaryotic world.
In spermatozoa, DNA damages, strand breakage, or base alteration may result from (1) Faulty compaction/protamination of DNA [205]; (2) Abortive apoptosis, or "anastasis", in late spermiogenesis [206]; (3) Oxidative stress [207]; (4) Persistence of enzymatic DNA breaks induce during chromatin remodelling. The contribution of each to sperm DNA damages is still unknown but it becomes clear that an elevated number of DNA breaks is associated with male infertility [208] or with impaired development of the early embryo. Chromatin remodelling in spermatids leads to the replacement of 90-95% of histones by protamines [209]. During the eviction of histones, it is assumed that free DNA supercoils are formed that can be eliminated by the action of topoisomerases [210]. Potential hindrance of topoisomerases catalytic cycle in this context may lead to SSBs in the case of topoisomerase I, or DSBs in the case of a topoisomerase II activity. A possibility also exists that mechanical breaks resulting from the major chromatin remodelling can also be created. Protamination itself can induce damage [211,212] as the PRM1/PRM2 ratio can impact proper DNA packaging [209,211,213]. We, however, generated evidence that DSBs were indeed generated during this process in the whole cell population and harbour a 3 OH so is consistent with endonuclease digestion. In this haploid context, only direct end ligation repair mechanisms such as NHEJ are expected [214]. Although apoptosis was shown to maintain the germ cell vs Sertoli cell ratio [215,216], canonical apoptosis may not be responsible for the transient DNA breaks shown in spermatids as these are observed in 100% of spermatids. This could represent an "apoptosis-like" reversible mechanism that is reminiscent of "anastasis", or the recovery from apoptosis described recently [217]. Various degrees of recovery, or repairs, would then modulate the persistence of DNA breaks in the mature sperm. At other spermatogenesis steps, DSBs could be linked to the altered balance of anti and pro-apoptotic factors through different steps of spermatogenesis [218] and to the compartmentalisation of the mitochondria which prevents nucleus-organelles exchanges. Reactive oxygen species (ROS) are a common source of DNA damage. ROS create abasic sites, base modifications, inter-strand crosslinking, and both single and double-strand breaks. However, ROS are essential for some sperm functions including capacitation or acrosome formation and are produced in the sperm mitochondria [219,220]. Sperm and white blood cells are the main sources of ROS in semen. Sperm cells are very sensitive to ROS because of their high content in unsaturated fatty acids and their weak DNA repair activity [207,219]. Antioxidants appear to balance a high level of ROS [221], for this purpose, seminal fluid possesses catalases, superoxide dismutases, and glutathione peroxidases and is rich in vitamins C, E, A, lactoferrin, and Q10. Ultimately, prostasomes can decrease the release of ROS from leukocytes [222]. The relevance of ROS in male infertility is emphasized as 20 to 88% of sub-fertile men show an elevated level of ROS [223].
So far, evidence of NER, BER, SSA, MMR, HR, and NHEJ DNA repair processes has been reported during spermatogenesis.
Important variations in NER activity were reported throughout spermatogenesis. For instance, spermatogonia were shown to be more sensitive to UV than meiotic or post-meiotic cells [224]. 6-4 photoproducts repair activity is also reduced significantly in post-meiotic round spermatids and during aging [224]. This may be linked to a decrease in NER protein expression [225]. Various levels of GG-or TC-NER also exist as 16.8% of CPD are removed from transcribed Scp1 gene within 16 h in rat spermatocytes [226], while 50% of CPD are removed from Dhfr and Dazl transcribed genes in spermatogonia [224]. However, spermatogonia can repair both active and inactive genes on both strands, in contrast to meiotic cells where transcribed genes are repaired. TC-NER is lower in round spermatids than in spermatogonia. These differences may reflect the greater mitotic activity of spermatogonia, and the fact that they are lying outside of the blood-testis barrier and so are potentially more exposed to genotoxins. Chromatin compaction during spermiogenesis may also adversely impact NER [227] as chromatin access is known to hinder NER. NER is therefore present at all stages of the male germ cells differentiation albeit with various efficiency.
OGG1 Hs glycosylase is the only BER factor found in spermatozoa and recognizes 8-oxoG [219]. However, as none of the AP endonucleases exist in spermatozoa, AP sites will likely be processed during the first zygotic mitosis [219].
HR is known for its primary role during meiosis but is also important for DSBs and inter-strand lesions repair in spermatogonia. Outside of meiosis, HR acts in the S and G2 phases of the cell cycle, when sister chromatids are present. HR is obviously absent in haploid spermatids but resumes in the diploid zygote.
In male germ cells, some canonical components of NHEJ are missing resulting in an alternative NHEJ pathway [228]. For instance, Ku70 Hs and 53BP1 Hs are not expressed therefore preventing the DNA-PKc Hs pathway. PARP-1 Hs and XRCC1 Hs are however expressed in elongating spermatids supporting alt-NHEJ pathway as described by Ahmed et al. [228]. PARP-1 Hs may bind to DSB as a catalytic homodimer bringing together both DNA ends. Then, PARP-1 Hs may recruit the XRCC1 Hsc /DNA ligase III Hs complex at the DSB site to complete the end-joining.

Oocyte
Since a limited number of oocytes can be retrieved, the DNA damage response in this gamete is studied to a lesser extent. DNA damages arise mainly during the active life of oocytes in prenatal stages, before prophase I arrest, and during recovery from pre-ovulatory meiosis [229].
From the primary oocyte stage to meiosis II and ovulation, oocytes can actively repair DNA [230]. During oogenesis, repair genes are overexpressed, and their mRNA are accumulated to be used in the zygote [231]. NER, BER, MMR, HR or NHEJ repair pathways have been described in oocytes from humans to mice [231][232][233][234]. Proper DNA repair machinery must operate as oocytes undergo chromatin remodelling during fertilisation and early steps of a zygote. Oocytes are diploid cells, that possess homologous chromosomes until meiosis II that arises after fertilisation. Thus, the oocyte is virtually never haploid. The oocyte becomes transiently haploid between meiosis II end and the fusion of parental genomes (syngamy). Thus, following meiosis II, only NHEJ should operate as shown in mice [235], although variations exist between species as rodent oocytes have better DNA repair abilities than primate ones [232]. The repair capacity of the oocyte and mRNA levels for repair proteins also decreases with female age [236]. After fertilisation, syngamy is a critical step since (1) Chromatids from each parent exist in different cellular compartments; (2) Chromatin undergoes major remodelling associated with demethylation; (3) Male derived chromatids are repaired by the oocyte DNA repair machinery [237].
As for spermatogenesis, ROS are produced during oogenesis. ROS production is necessary for folliculogenesis, oocyte maturation, ovulation, and luteal function [238]. Maturation and ovulation are linked to inflammatory processes, characterized by high ROS production [239]. Unlike sperm, oocytes are more resistant to ROS due to an environment rich in antioxidants in the follicular fluid [240]. However, ROS imbalance in the female tract decreases fertility and oocyte quality [241]. In addition to its impact on genetic integrity, oxidative stress is known to alter microtubule function and spindle morphology leading to aneuploidy [242].
In contrast to spermatozoa, OGG1 Hs glycosylase is only weakly expressed in oocytes, as 8-oxoG damages are not present in spermatozoa after fertilization [243]. In many aspects, the early zygote may still be considered as an oocyte since only mRNAs for the oocyte are used up until the 4-cell stage [244]. This "zygotic oocyte" is in charge of repairing DNA damages in spermatozoa but may be overloaded if DNA damages occur in more than 8% of the nucleotide genome amount [245,246].
Both the number and the nature of DNA breaks are of concern for fertility. Whereas SSBs are easily repaired as in somatic cells, DSBs can overload the DNA repair capacity of the oocyte. After fertilisation, the persistence of DSBs in spermatozoa will delay DNA replication or can even be lethal for the zygote [247]. Finally, DSBs are mainly repaired by NHEJ during zygote [214], as for somatic cells.

Conclusions
Ploidy is a key element to consider in the regulation of DNA repair systems. In view of the aforementioned details about the two yeast models, it seems that species can however adapt their uses of DNA repair systems to overcome the weaknesses of either haploidy or diploidy. Thus, many redundancies exist within and between DNA repair systems that are linked together into a complex and integrated overarching regulatory mechanism that ensures the maintenance of genomic integrity regardless of the ploidy state. This may of course represent a functional bias for a single-cell organism as they are solely responsible to generate genetically intact progeny. In fact, major differences in repair efficiency between diploid and haploid cells are mainly observed among mammalian metazoan gametes. Hence, unlike oocytes, mature sperm clearly lacks proper DNA repair activity. However, the many replication/differentiation steps, change in DNA topology and highly condensed and specialized chromatin structure observed during spermatogenesis offers an appropriate context to generate adaptative variations.