**8. The role of Spo11 in promoting accurate DNA repair can also facilitate proper chromosome segregation**

In the budding yeast *S. cerevisiae,* synapsis (pairing of homologous chromosomes) and synaptonemal complex formation depend on Spo11, a nuclease related to type II topoismerases. Spo11 induces DSBs leading to HRR events of the CO type that form the physical association between homologs (chiasmata) needed for synaptonemal complex formation and proper disjunction of non-sister homologs at the first meiotic division. On the basis of these properties of Spo11, it is sometimes assumed that the primary function of meiotic recombination is to promote synapsis. However, as reviewed by Barzel & Kupiec (2008), this theme cannot be generalized, as synapsis occurs independently of Spo11 induced recombination in the nematode worm *C. elegans* and the fruitfly *D. melanogaster*. In *C. elegans,* synapsis between homologs occurs normally in a *spo-11* mutant (Dernburg et al., 1998). The *D. melanogaster* gene *mei-W68* encodes a *spo11* homolog (McKim & Hayashi-Hagihara, 1998). In *D. melanogaster* females, meiotic chromosome synapsis occurs in the absence of *mei-W68* mediated CO recombination (McKim et al., 1998). Electron microscopy of oocytes from females homozygous for *mei-W68* mutations that eliminated meiotic recombination revealed normal synaptonemal complex formation. In *D. melanogaster* females, meiotic recombination does not appear to be necessary for synapsis. Since the role of Spo11 is of substantial interest in current discussions of the adaptive significance of meiotic recombination, we offer a speculation on its possible role consistent with the DNA repair hypothesis. As shown in Figure 1, both the DHJ and SDSA models for HRR start with a DSB. During meiosis in *S. cerevisiae*, DSBs are formed by a process that usually depends on Spo11. In *S. pombe*, Spo11 homolog Rec12 generates meiotic recombinants and meiosis specific DSBs. In *C. elegans*, a Spo11 homolog seems to have a similar role. We propose that DNA damages of various types are converted to DSBs, a "common currency," in order to initiate their recombinational repair (see also H. Bernstein et al., 1988). Spo11 appears to be employed in this process. Our reasoning is based on the precedents of the well-established pathways of nucleotide excision repair and base excision repair. In nucleotide excision repair, the initial steps of the pathway involve recognition of a wide variety of bulky damages followed by their removal to generate a single-strand gap, the "common currency" which is then repaired by a gap filling process. In base excision repair, a variety of altered bases are recognized by a corresponding variety of DNA glycosylases that generate an intermediate apurinic/apyrimidinic site, the "common currency" for further repair. On this reasoning, formation of DSBs by a Spo11-dependent process is part of an overall DNA repair sequence. In those species where the resolution of meiotic HRR by CO recombination is beneficial in promoting proper chromosome segregation at the first meiotic division, we think this benefit arose secondarily to the primary benefit of accurate DNA repair.

Meiosis as an Evolutionary Adaptation for DNA Repair 367

**10. DNA repair likely provides the strong short-term advantage that maintains** 

Evolutionary explanations for sex have often assumed that the adaptive advantage of meiosis arises from the genetic variation produced. A variety of models and reviews have been presented in this active area of research (e.g. Barton & Charlesworth, 1998; Otto & Gerstein, 2006; Agrawal, 2006). However, Otto & Gerstein (2006) have also pointed out that in a fairly stable environment, individuals surviving to reproductive age have genomes that function well in their current environment. They raise the question of why such individuals should risk shuffling their genes with those of another individual, as happens during meiotic recombination. This consideration, and others, have led many investigators to question whether production of genetic diversity is the principal adaptive advantage of sex. Heng (2007) and Gorelick & Heng (2010) reviewed evidence that sex actually decreases most genetic variation. Their view is that sex acts like a coarse filter, weeding out major changes, such as chromosomal rearrangements, but allowing minor variation, such as changes at the nucleotide or gene level (that are often neutral), to flow through the sexual sieve. Thus, they consider that sex acts as a constraint on genomic variation, thereby limiting

We consider that the major adaptive advantage of meiosis is enhanced recombinational repair. In contrast to the variation hypothesis, DNA repair provides an appropriate explanation for the adaptive advantage of sex (and meiosis) in the short-term, since its benefits are large enough (removal of DNA damages that would be deleterious/lethal to gametes or progeny) to plausibly balance the large costs of sex. The large costs of sex include the "cost of males" (Maynard Smith, 1978; Williams, 1975), "recombinational load" that arises from the randomization of genetic information during sex and loss of coadapted gene complexes (Shields, 1982), the cost of mating (Bernstein et al., 1985b), and cost of sexually

The hypothesis that meiosis is an adaptation for DNA repair can be consistently applied to all organisms that have sex, including the facultative sexual organisms discussed above, as well as species that undergo meiosis but experience little or no outcrossing, as described below. If, in the long-term, the genetic variation produced by sex increases the rate of adaptation, as proposed by a number of authors (Goddard et al., 2005; Colegrave et al., 2002; Kaltz & Bell, 2002; Cooper et al., 2005; de Visser & Elena, 2007; Peters & Otto, 2003), this would be an added benefit. However, in the short-term, we consider it unlikely that the

In nature, many organisms that undergo meiosis outcross only rarely or not at all. In these cases, meiosis generates little or no genetic variation. In the budding yeast *S. cerevisiae,* outcrossing sex, in contrast to inbreeding sex, appears to be very infrequent in nature. Ruderfer et al. (2006) estimated that the ancestors of three *S. cerevisiae* strains outcrossed in nature only about once every 50,000 generations. On the other hand, mating between closely related yeast cells is likely to have been much more common in nature. Mating can occur when haploid cells of opposite mating types, MATa and MATα, come into contact. As pointed out by Zeyl & Otto (2007), mating between closely related cells is common for two reasons; (1) the close physical proximity of cells of opposite mating type from the same ascus (the sac that contains the products from a single meiosis), and (2) homothallism, the ability of haploid cells of one mating type to produce daughter cells of the opposite mating type. Thus, in nature, the meiotic events that produce little or no recombinational variation

**meiosis, while genetic variation may provide a long-term advantage** 

adaptive evolution.

transmitted disease (Michod et al., 2008).

benefit of variation is large enough to maintain sex.

The function of recombination as a repair process may have arisen very early in the evolution of life [perhaps in the RNA world (H. Bernstein et al., 1984)], and the function of promoting synapsis during meiosis probably arose later in evolution in some eukaryotic lineages. If, in mammals, a major function of meiotic CO recombination, as distinct from NCO recombination, is to promote synapsis and proper chromosome segregation, then one might expect CO events to be localized to specific hot-spot sequences. Hot-spot determinants may also include specific proteins that bind to hot-spot sequences and facilitate CO recombination such as Prdm9 (Hochwagen and Marals, 2010). It is estimated that, in humans, the average number of endogenous DNA DSBs per somatic cell occurring at each cell generation is about 50 (Vilenchik & Knudson, 2003). This rate of DSB formation likely reflects unprogrammed damages, such as may be caused by ROS, and can be taken as an indication of the level of unprogrammed DSBs present in cells undergoing meiosis as well. In the human genome 25,000 hotspots for meiotic recombination have been identified (Myers et al., 2006). The average number of CO recombination events per hotspot is one CO event per 1,300 meioses. The large number of recombination hotspots is consistent with a wide distribution of sites vulnerable to unprogrammed DNA damage as well as specific sites where recombination would need to be induced to promote synapsis. A challenge for future research is the identification of the types of natural damages and programmed damages, and their frequencies, that are removed by CO recombinational repair during meiosis.

#### **9. During meiosis, CO recombination can repair DNA damages independently of Spo11**

In a *spo11* mutant of *S. cerevisiae*, the meiotic defects in recombination and synapsis are alleviated by X-irradiation, indicating that X-ray induced DNA damages can initiate CO recombination leading to synapsis independently of Spo11 (Thorne & Byers, 1993). Also, in *C. elegans*, Spo11 is required for meiotic recombination, but radiation induced-breaks alleviate this dependence (Dernberg et al., 1998). These findings indicate that unprogrammed DNA damages induced by X-rays can be repaired by HRR during meiosis independently of Spo11. In both *S. pombe* and *C. elegans*, mutants deficient for Spo11 undergo meiotic CO recombination when single base lesions of the type dU:dG are produced in their DNA (Pauklin et al., 2009). This recombination does not involve production of large numbers of DSBs, but does require uracil DNA-glycolylase, an enzyme that removes uracil from the DNA backbone and initiates base excision repair. These authors proposed that base excision repair of a uracil base, an abasic site, or a single-strand nick are sufficient to initiate meiotic CO recombination in *S pombe* and *C. elegans*.

In a Rec12 (Spo11 homolog) mutant strain of *S. pombe*, meiotic recombination can be restored to near normal levels by a deletion in *rad2* that encodes an endonuclease involved in Okazaki fragment processing (Farah et al., 2005). Both CO and NCO recombination were increased, but DSBs were undetectable. On the basis of the biochemical properties of Rad 2, these authors proposed that meiotic recombination can be initiated by non-DSB lesions, such as nicks and gaps, which accumulate during premeiotic DNA replication when Okasaki fragment processing is deficient.

In general the findings reviewed in this section indicate that DNA damages arising from a variety of sources can be repaired by meiotic HRR of the CO type, and that this repair may occur independently of Spo11.

The function of recombination as a repair process may have arisen very early in the evolution of life [perhaps in the RNA world (H. Bernstein et al., 1984)], and the function of promoting synapsis during meiosis probably arose later in evolution in some eukaryotic lineages. If, in mammals, a major function of meiotic CO recombination, as distinct from NCO recombination, is to promote synapsis and proper chromosome segregation, then one might expect CO events to be localized to specific hot-spot sequences. Hot-spot determinants may also include specific proteins that bind to hot-spot sequences and facilitate CO recombination such as Prdm9 (Hochwagen and Marals, 2010). It is estimated that, in humans, the average number of endogenous DNA DSBs per somatic cell occurring at each cell generation is about 50 (Vilenchik & Knudson, 2003). This rate of DSB formation likely reflects unprogrammed damages, such as may be caused by ROS, and can be taken as an indication of the level of unprogrammed DSBs present in cells undergoing meiosis as well. In the human genome 25,000 hotspots for meiotic recombination have been identified (Myers et al., 2006). The average number of CO recombination events per hotspot is one CO event per 1,300 meioses. The large number of recombination hotspots is consistent with a wide distribution of sites vulnerable to unprogrammed DNA damage as well as specific sites where recombination would need to be induced to promote synapsis. A challenge for future research is the identification of the types of natural damages and programmed damages, and their frequencies, that are removed by CO recombinational

**9. During meiosis, CO recombination can repair DNA damages independently** 

In a *spo11* mutant of *S. cerevisiae*, the meiotic defects in recombination and synapsis are alleviated by X-irradiation, indicating that X-ray induced DNA damages can initiate CO recombination leading to synapsis independently of Spo11 (Thorne & Byers, 1993). Also, in *C. elegans*, Spo11 is required for meiotic recombination, but radiation induced-breaks alleviate this dependence (Dernberg et al., 1998). These findings indicate that unprogrammed DNA damages induced by X-rays can be repaired by HRR during meiosis independently of Spo11. In both *S. pombe* and *C. elegans*, mutants deficient for Spo11 undergo meiotic CO recombination when single base lesions of the type dU:dG are produced in their DNA (Pauklin et al., 2009). This recombination does not involve production of large numbers of DSBs, but does require uracil DNA-glycolylase, an enzyme that removes uracil from the DNA backbone and initiates base excision repair. These authors proposed that base excision repair of a uracil base, an abasic site, or a single-strand

nick are sufficient to initiate meiotic CO recombination in *S pombe* and *C. elegans*.

In a Rec12 (Spo11 homolog) mutant strain of *S. pombe*, meiotic recombination can be restored to near normal levels by a deletion in *rad2* that encodes an endonuclease involved in Okazaki fragment processing (Farah et al., 2005). Both CO and NCO recombination were increased, but DSBs were undetectable. On the basis of the biochemical properties of Rad 2, these authors proposed that meiotic recombination can be initiated by non-DSB lesions, such as nicks and gaps, which accumulate during premeiotic DNA replication when Okasaki

In general the findings reviewed in this section indicate that DNA damages arising from a variety of sources can be repaired by meiotic HRR of the CO type, and that this repair may

repair during meiosis.

fragment processing is deficient.

occur independently of Spo11.

**of Spo11** 

#### **10. DNA repair likely provides the strong short-term advantage that maintains meiosis, while genetic variation may provide a long-term advantage**

Evolutionary explanations for sex have often assumed that the adaptive advantage of meiosis arises from the genetic variation produced. A variety of models and reviews have been presented in this active area of research (e.g. Barton & Charlesworth, 1998; Otto & Gerstein, 2006; Agrawal, 2006). However, Otto & Gerstein (2006) have also pointed out that in a fairly stable environment, individuals surviving to reproductive age have genomes that function well in their current environment. They raise the question of why such individuals should risk shuffling their genes with those of another individual, as happens during meiotic recombination. This consideration, and others, have led many investigators to question whether production of genetic diversity is the principal adaptive advantage of sex.

Heng (2007) and Gorelick & Heng (2010) reviewed evidence that sex actually decreases most genetic variation. Their view is that sex acts like a coarse filter, weeding out major changes, such as chromosomal rearrangements, but allowing minor variation, such as changes at the nucleotide or gene level (that are often neutral), to flow through the sexual sieve. Thus, they consider that sex acts as a constraint on genomic variation, thereby limiting adaptive evolution.

We consider that the major adaptive advantage of meiosis is enhanced recombinational repair. In contrast to the variation hypothesis, DNA repair provides an appropriate explanation for the adaptive advantage of sex (and meiosis) in the short-term, since its benefits are large enough (removal of DNA damages that would be deleterious/lethal to gametes or progeny) to plausibly balance the large costs of sex. The large costs of sex include the "cost of males" (Maynard Smith, 1978; Williams, 1975), "recombinational load" that arises from the randomization of genetic information during sex and loss of coadapted gene complexes (Shields, 1982), the cost of mating (Bernstein et al., 1985b), and cost of sexually transmitted disease (Michod et al., 2008).

The hypothesis that meiosis is an adaptation for DNA repair can be consistently applied to all organisms that have sex, including the facultative sexual organisms discussed above, as well as species that undergo meiosis but experience little or no outcrossing, as described below. If, in the long-term, the genetic variation produced by sex increases the rate of adaptation, as proposed by a number of authors (Goddard et al., 2005; Colegrave et al., 2002; Kaltz & Bell, 2002; Cooper et al., 2005; de Visser & Elena, 2007; Peters & Otto, 2003), this would be an added benefit. However, in the short-term, we consider it unlikely that the benefit of variation is large enough to maintain sex.

In nature, many organisms that undergo meiosis outcross only rarely or not at all. In these cases, meiosis generates little or no genetic variation. In the budding yeast *S. cerevisiae,* outcrossing sex, in contrast to inbreeding sex, appears to be very infrequent in nature. Ruderfer et al. (2006) estimated that the ancestors of three *S. cerevisiae* strains outcrossed in nature only about once every 50,000 generations. On the other hand, mating between closely related yeast cells is likely to have been much more common in nature. Mating can occur when haploid cells of opposite mating types, MATa and MATα, come into contact. As pointed out by Zeyl & Otto (2007), mating between closely related cells is common for two reasons; (1) the close physical proximity of cells of opposite mating type from the same ascus (the sac that contains the products from a single meiosis), and (2) homothallism, the ability of haploid cells of one mating type to produce daughter cells of the opposite mating type. Thus, in nature, the meiotic events that produce little or no recombinational variation

cell cycle.

replication in mitotic cells.

due to apoptosis.

phase), NHEJ is elevated and HRR is in decline.

unrepairable DSBs which arise before DNA replication.

Meiosis as an Evolutionary Adaptation for DNA Repair 369

repaired by an inaccurate process, non-homologous end-joining (NHEJ), which generates mutation. Double strand damages arising after DNA replication, may be repaired during mitosis by HRR between sisters (Tichy et al., 2010). However, meiotic recombination can cope in a non-mutagenic way with double strand damages which arise at any point in the

Meiotic G1 phase cells appear to be more resistant to the lethal effects of X-irradiation than mitotic G1 phase cells (Kelly et al., 1983). This finding suggests that repair of DSBs is more efficient during meiotic than mitotic G1 phase, as DSBs are a common consequence of Xirradiation. We speculate that during meiosis, in contrast to mitosis, double-strand damages occurring prior to DNA replication may be accurately repaired by HRR because pairing occurs between non-sister chromosomes. If this is so, meiotic cells have the advantage, compared to mitotic cells, of being able to accurately and efficiently repair double-strand damages that occur both before and after replication. As a result, germ cells would tend to be protected against the mutagenic effect of inaccurate NHEJ that typically occurs prior to

Mao et al. (2008) presented evidence that one type of somatic cell, human fibroblasts, utilizes error-prone NHEJ as the major DSB repair pathway at all cell cycle stages. In these cells, HRR is nearly absent prior to replication (G1 phase) and is used, when it occurs, primarily in the S phase. Even after the S phase when two sister-chromosomes are present (the G2/M

The situation is somewhat different in mammalian embryonic stem (ES) cells compared to differentiated somatic cells (Tichy et al., 2010). ES cells give rise to all of the cell types of an organism. Because mutations at this early embryonic stage are passed on to all clonal descendents, they can be seriously detrimental to the organism as a whole. Therefore robust mechanisms are needed in ES cells for reducing DNA damages (or eliminating damaged cells) in order to reduce mutations. Mouse ES cells were found to predominantly use high fidelity HRR to repair DSBs, compared to somatic cells that predominantly used NHEJ (Tichy et al., 2010). Furthermore mouse ES cells lack a G1 checkpoint and do not undergo cell-cycle arrest upon receiving DNA damage prior to DNA replication. Rather, they undergo p53-independent apoptosis in response to DNA damage (Aladjem et al., 1998). Consistent with these findings, mouse ES stem cells have a mutation frequency about 100 fold lower than that of isogenic mouse somatic cells (Cervantes et al., 2002), but, as discussed next, at a likely cost resulting from somatic selection against cells with

These results imply that a low mutation rate is achievable in mitotic cells by using apoptosis to remove cells with DNA damages that are present prior to replication, and using HRR, rather than NHEJ, to remove double-strand damages present subsequent to DNA replication. The non-sister chromosomes present in every diploid somatic cell during mitosis, in principal, might pair and undergo accurate HRR (as in meiosis), but this does not ordinarily occur, presumably because, in somatic cells, the benefit is outweighed by costs [e.g. loss of heterozygosity and expression of deleterious recessive alleles including those leading to cancer]. Meiosis is therefore unique, in that DNA damages occurring both prior to and after DNA replication can be subject to high fidelity HRR between non-sister homologs. This would avoid the high costs of both deleterious mutation and loss of potential gametes

In humans at each cell division, 30,000-50,000 DNA replication origins are activated (Mechali et al., 2010). Thus the chromosome is ordinarily replicated in segments. We

are much more frequent than meiotic events that do produce recombinational variation. This disparity is consistent with the idea that the primary adaptive function of meiosis in *S. cerevisiae* is HRR of DNA damages, since this benefit is realized in meiosis resulting from either inbreeding or outcrossing. If the primary adaptive function of meiosis were to generate genetic variation, it is difficult to understand how the complex process of meiosis could be selectively maintained in *S. cerevisiae* during the many generations in which there is no outcrossing.

Various levels of inbreeding due to consanguineous mating are known in many species. One extreme, but well studied, example among vertebrate species is the Mangrove Killifish, *Kryptolebias marmoratus*, which inhabits brackish water mangrove habitats from Brazil to Florida. These fish produce sperm and eggs by meiosis and reproduce routinely by selffertilization. Each hermaphroditic individual normally fertilizes itself when a sperm and egg that it has produced by an internal organ unite inside the fish's body (Sakakura et al., 2006; for review see Avise, 2008). In this highly inbred hermaphroditic species meiotic recombination does not produce significant allelic variation, suggesting that meiosis is retained for some other adaptive benefit.

In higher plants, outcrossing sexual reproduction is the most common mode of reproduction, but about 15% of plants undergo meiosis and are principally self-fertilizing (C. Bernstein & H. Bernstein, 1991). We infer from these examples that the generation of genetic variation is not likely to be the adaptive benefit maintaining meiosis in these organisms. However, meiosis may be maintained by the adaptive benefit of HRR of DNA damage, since this benefit does not depend on outcrossing, nor that the participating chromosomes carry different alleles.

The meiotic function of repairing DNA damages primarily acts to preserve the existing genome. The generation of new genomic variants, a consequence of recombinational repair processes, appears to be a secondary effect that may provide a benefit in the long-term.

As discussed above, most HRR events during meiosis are of the NCO type, which generate minimal genetic variation compared to the CO type. This is consistent with the DNA repair hypothesis, since both the CO and NCO types of recombination can repair DNA. On the assumption that the generation of variation is the primary benefit of meiosis, the majority of HRR events, those of the NCO type, provide no significant benefit and hence are wasteful.

Even though, during meiosis, the frequency of CO recombination is ordinarily substantially less than the frequency of NCO recombination, during mitosis the frequency of CO compared to NCO recombination is even lower (e.g. Virgin et al., 2001; Prado et al., 2003). The higher frequency of CO recombinants during meiosis compared to mitosis may reflect the role of CO recombinants in promoting synapsis during meiosis (see section 8, above), a process distinct to meiosis.

#### **11. During meiosis, HRR may remove a class of damages that cannot be accurately repaired during mitosis**

HRR during meiosis offers unique advantages compared to HRR during mitosis, based on the opportunity for non-sister homologs to pair and recombine during meiosis, which does not happen during mitosis. In mitosis, HRR involves interaction between the sisterchromosomes formed upon DNA replication. Thus, in mitosis, HRR is limited to the phases of the cell cycle during DNA replication (S phase) and after DNA replication (G2/M). Prior to DNA replication (G1 phase) in mitosis, double-strand DNA damages, such as DSBs, are

are much more frequent than meiotic events that do produce recombinational variation. This disparity is consistent with the idea that the primary adaptive function of meiosis in *S. cerevisiae* is HRR of DNA damages, since this benefit is realized in meiosis resulting from either inbreeding or outcrossing. If the primary adaptive function of meiosis were to generate genetic variation, it is difficult to understand how the complex process of meiosis could be selectively maintained in *S. cerevisiae* during the many generations in which there is

Various levels of inbreeding due to consanguineous mating are known in many species. One extreme, but well studied, example among vertebrate species is the Mangrove Killifish, *Kryptolebias marmoratus*, which inhabits brackish water mangrove habitats from Brazil to Florida. These fish produce sperm and eggs by meiosis and reproduce routinely by selffertilization. Each hermaphroditic individual normally fertilizes itself when a sperm and egg that it has produced by an internal organ unite inside the fish's body (Sakakura et al., 2006; for review see Avise, 2008). In this highly inbred hermaphroditic species meiotic recombination does not produce significant allelic variation, suggesting that meiosis is

In higher plants, outcrossing sexual reproduction is the most common mode of reproduction, but about 15% of plants undergo meiosis and are principally self-fertilizing (C. Bernstein & H. Bernstein, 1991). We infer from these examples that the generation of genetic variation is not likely to be the adaptive benefit maintaining meiosis in these organisms. However, meiosis may be maintained by the adaptive benefit of HRR of DNA damage, since this benefit does not depend on outcrossing, nor that the participating

The meiotic function of repairing DNA damages primarily acts to preserve the existing genome. The generation of new genomic variants, a consequence of recombinational repair processes, appears to be a secondary effect that may provide a benefit in the long-term. As discussed above, most HRR events during meiosis are of the NCO type, which generate minimal genetic variation compared to the CO type. This is consistent with the DNA repair hypothesis, since both the CO and NCO types of recombination can repair DNA. On the assumption that the generation of variation is the primary benefit of meiosis, the majority of HRR events, those of the NCO type, provide no significant benefit and hence are wasteful. Even though, during meiosis, the frequency of CO recombination is ordinarily substantially less than the frequency of NCO recombination, during mitosis the frequency of CO compared to NCO recombination is even lower (e.g. Virgin et al., 2001; Prado et al., 2003). The higher frequency of CO recombinants during meiosis compared to mitosis may reflect the role of CO recombinants in promoting synapsis during meiosis (see section 8, above), a

**11. During meiosis, HRR may remove a class of damages that cannot be** 

HRR during meiosis offers unique advantages compared to HRR during mitosis, based on the opportunity for non-sister homologs to pair and recombine during meiosis, which does not happen during mitosis. In mitosis, HRR involves interaction between the sisterchromosomes formed upon DNA replication. Thus, in mitosis, HRR is limited to the phases of the cell cycle during DNA replication (S phase) and after DNA replication (G2/M). Prior to DNA replication (G1 phase) in mitosis, double-strand DNA damages, such as DSBs, are

no outcrossing.

retained for some other adaptive benefit.

chromosomes carry different alleles.

process distinct to meiosis.

**accurately repaired during mitosis** 

repaired by an inaccurate process, non-homologous end-joining (NHEJ), which generates mutation. Double strand damages arising after DNA replication, may be repaired during mitosis by HRR between sisters (Tichy et al., 2010). However, meiotic recombination can cope in a non-mutagenic way with double strand damages which arise at any point in the cell cycle.

Meiotic G1 phase cells appear to be more resistant to the lethal effects of X-irradiation than mitotic G1 phase cells (Kelly et al., 1983). This finding suggests that repair of DSBs is more efficient during meiotic than mitotic G1 phase, as DSBs are a common consequence of Xirradiation. We speculate that during meiosis, in contrast to mitosis, double-strand damages occurring prior to DNA replication may be accurately repaired by HRR because pairing occurs between non-sister chromosomes. If this is so, meiotic cells have the advantage, compared to mitotic cells, of being able to accurately and efficiently repair double-strand damages that occur both before and after replication. As a result, germ cells would tend to be protected against the mutagenic effect of inaccurate NHEJ that typically occurs prior to replication in mitotic cells.

Mao et al. (2008) presented evidence that one type of somatic cell, human fibroblasts, utilizes error-prone NHEJ as the major DSB repair pathway at all cell cycle stages. In these cells, HRR is nearly absent prior to replication (G1 phase) and is used, when it occurs, primarily in the S phase. Even after the S phase when two sister-chromosomes are present (the G2/M phase), NHEJ is elevated and HRR is in decline.

The situation is somewhat different in mammalian embryonic stem (ES) cells compared to differentiated somatic cells (Tichy et al., 2010). ES cells give rise to all of the cell types of an organism. Because mutations at this early embryonic stage are passed on to all clonal descendents, they can be seriously detrimental to the organism as a whole. Therefore robust mechanisms are needed in ES cells for reducing DNA damages (or eliminating damaged cells) in order to reduce mutations. Mouse ES cells were found to predominantly use high fidelity HRR to repair DSBs, compared to somatic cells that predominantly used NHEJ (Tichy et al., 2010). Furthermore mouse ES cells lack a G1 checkpoint and do not undergo cell-cycle arrest upon receiving DNA damage prior to DNA replication. Rather, they undergo p53-independent apoptosis in response to DNA damage (Aladjem et al., 1998). Consistent with these findings, mouse ES stem cells have a mutation frequency about 100 fold lower than that of isogenic mouse somatic cells (Cervantes et al., 2002), but, as discussed next, at a likely cost resulting from somatic selection against cells with unrepairable DSBs which arise before DNA replication.

These results imply that a low mutation rate is achievable in mitotic cells by using apoptosis to remove cells with DNA damages that are present prior to replication, and using HRR, rather than NHEJ, to remove double-strand damages present subsequent to DNA replication. The non-sister chromosomes present in every diploid somatic cell during mitosis, in principal, might pair and undergo accurate HRR (as in meiosis), but this does not ordinarily occur, presumably because, in somatic cells, the benefit is outweighed by costs [e.g. loss of heterozygosity and expression of deleterious recessive alleles including those leading to cancer]. Meiosis is therefore unique, in that DNA damages occurring both prior to and after DNA replication can be subject to high fidelity HRR between non-sister homologs. This would avoid the high costs of both deleterious mutation and loss of potential gametes due to apoptosis.

In humans at each cell division, 30,000-50,000 DNA replication origins are activated (Mechali et al., 2010). Thus the chromosome is ordinarily replicated in segments. We

Meiosis as an Evolutionary Adaptation for DNA Repair 371

outcrossing (H. Bernstein et al., 1985a, 1987; Michod, 1995). However, more explicit population genetic models have raised some issues that are in need of further clarification. In population genetics terms, the basic effect of outcrossing is to bring populations to Hardy-Weinberg (HW) equilibrium. Thus, outcrossing can be beneficial if there is another force that pushes the population away from HW equilibrium (generating either an excess or a deficit of heterozygotes) and if it's advantageous to go closer to HW equilibrium. One possible force that generates departure from HW equilibrium is dominance: for example if deleterious alleles tend to be recessive, after selection there will be an excess of heterozygotes (and a deficit of homozygotes). However in this case outcrossing is costly in the short term (because it tends to expose deleterious alleles), but beneficial in the long term (because purging them becomes more efficient). Otto (2003) showed that under this scenario high rates of outcrossing are favored only if deleterious alleles are weakly recessive (dominance close to 0.5). Another potential force pushing away from HW equilibrium considered by Roze and Michod (2010) is gene conversion which creates homozygosity. Gene conversion could result from mitotic HRR between sister chromosomes as discussed above. In this case (and if deleterious alleles tend to be partially recessive) outcrossing is beneficial in the short term (because it masks deleterious alleles) but disadvantageous in the long term (because purging is less efficient). The magnitude of this force may be estimated from rates of loss of heterozygosity during development [discussed in Roze and Michod (2010)]. The few estimates which exist indicate that the loss of heterozygosity is low, and thus this selective force for outcrossing may be weak. Clearly, we need more estimates of this critical

Another consequence of outcrossing is the generation of new genetic variants which may

Bdelloid rotifers are common invertebrate animals. They are apparently obligate asexuals that reproduce by parthenogenesis. These organisms are extraordinarily resistant to ionizing radiation (Gladyshev and Meselson, 2008). This resistance appears to be a consequence of an evolutionary adaptation to survive desiccation in ephemerally aquatic habitats. Such desiccation causes extensive DNA breakage, which they are able to repair. Bdelloid primary oocytes are in the G1 phase of the cell cycle and thus lack sister chromatids. Welch et al. (2008) proposed a mechanism of repair involving interaction of non-sister co-linear chromosome pairs, which are maintained as templates for repair of DNA DSBs caused by the frequent desiccation and rehydration. Thus although these organisms apparently lack sex and meiosis, an essential feature of meiosis, HRR between non-sister homologs appears

Sex appears to be universally based on RecA-like proteins. RecA-like proteins play a key role in HRR, and the HRR machinery and its mechanism of action appear to be highly conserved among eukaryotes. The r*ad51* and d*mc1* genes in the eukaryotic yeasts *S. cerevisiae* and *S. pombe* are orthologs of the bacterial *recA* gene. The *dmc1* gene is found in

**15. Conservation among eukaryotes of RecA-like proteins as key** 

**components of the HRR machinery acting during meiosis** 

parameter to know how large this force for outcrossing may be.

**14. The special case of asexual bdelloid rotifers** 

provide an additional long-term advantage.

to be retained.

postulate that any segment containing a DSB will fail to complete its replication until the DSB is repaired. This limited and temporary blockage of replication may result directly from the break itself, or occur as a response to regulatory events set off by proteins that specifically bind to the broken ends. In any case, HRR can be carried out during the subsequent prophase I stage of meiosis, when the segment containing a DSB pairs with a non-sister homologue. This repair would then allow chromosome replication to be completed.
