Mini reviews
Spontaneous de novo germline mutations in humans and mice: rates, spectra, causes and consequences
2019 Volume 94 Issue 1 Pages 13-22
Details
2019 Volume 94 Issue 1 Pages 13-22
Germline mutations are the origin of genetic variation and are widely considered to be the driving force of genome evolution. The rates and spectra of de novo mutations (DNMs) directly affect evolutionary speed and direction and thereby establish species-specific genomic futures in the long term. This has resulted in a keen interest in understanding the origin of germline mutations in mammals. Accumulating evidence from next-generation sequencing and family-based analysis indicates that the frequency of human DNMs varies according to sex, age and local genomic context. Thus, it is likely that there are multiple causes and drivers of mutagenesis, including spontaneous DNA lesions, DNA repair status and DNA polymerase errors. In this review, recent studies of human and mouse germline DNMs are discussed, and the rates and spectra of spontaneous germline DNMs in the mouse mutator lines Pold1exo/exo and TOY-KO (Mth1−/−/Ogg1−/−/Mutyh−/− triple knockout) are summarized in the context of endogenous causes and mechanisms.
In mammals, a certain number of mutations occur spontaneously and accumulate in both somatic and germline cells throughout the lifetime of the individual. In somatic cells, the accumulated mutations increase the risk of cancer and other diseases. Crucially, only mutations that occur in germline cells can be transmitted to offspring and may spread in the population in later generations. Important questions must be addressed. Are there any differences between cell types regarding the maintenance of genome integrity? What are the causes and mechanisms of spontaneous mutation in both cell types? The answers to these questions may provide clues to further understanding genome evolution in mammals. In theory, soma and germ cells are expected to share some factors and mechanisms of DNA replication and repair because all cells in a body are derived from a single zygote. Both types of cells further undergo multiple rounds of DNA replication and mitosis during development. Consequently, environmental conditions or physical states that promote somatic mutations may also increase the incidence of germline mutations. However, even under similar mutagenic conditions, the outcomes may differ because the DNA damage response and the regulation of cell cycle checkpoints and cell death pathways may vary between cell types (Vermezovic et al., 2012; Bailly and Gartner, 2013). Although it is unclear whether germline cells possess specific systems that prevent mutations, several studies have reported that germline cells acquire mutations less frequently than somatic cells (Kohler et al., 1991; Walter et al., 1998; Milholland et al., 2017).
Eukaryotic cells possess a highly accurate DNA replication system that involves nucleotide selectivity and proofreading activity of DNA polymerases and a highly efficient mismatch repair system, such that the mutation rate is maintained at a very low level, typically ~10−9 - 10−10 mutations per nucleotide pair per replication (Drake et al., 1998; Arana and Kunkel, 2010). However, despite this highly accurate replication system, a considerable number of somatic mutations are generated because numerous DNA replications occur in order to produce a large number of cells over the lifetime of a mammal. For example, the human body consists of approximately 1014 cells and undergoes approximately 1016 cell divisions in a lifetime, resulting in over 1015 cumulative mutations per individual (Frank, 2014). The mutation burden borne by organs in a human body over its lifetime was previously summarized (Lynch, 2010a). For example, if 106 stem cells in intestinal tissue generate transient daughter cells once a week with a mutation rate of approximately 10−9 per nucleotide per cell division, the intestinal epithelium of a 60-year-old human would have accumulated more than 109 independent mutations. Thus, nearly every genomic site is likely to be mutated in at least one cell in this organ (Lynch, 2010a, 2010b).
Recent studies have revealed that even cancer-free tissues in humans accumulate multiple somatic mutations with age (Blokzijl et al., 2016). These mutations can increase the possibility of physical dysfunction and increase the risk of cancer (Weinberg, 2007; Negrini et al., 2010; Tomasetti and Vogelstein, 2015), neuronal disorders and other diseases (Poduri et al., 2013; Genovese et al., 2014; Jaiswal et al., 2014; Campbell et al., 2015). In addition, organs in the adult body that contain many dividing cells acquire mutations more frequently than non-proliferative tissues. The lifetime tumor incidence in human tissues is strongly correlated with the number of stem cell divisions; therefore, DNA replication error has been considered to be a major source of spontaneous mutations that drive cancer development (Tomasetti and Vogelstein, 2015). However, there is evidence that spontaneous mutations also occur in post-mitotic non-replicative cells as a result of DNA repair synthesis (Foster and Rosche, 1999; Kim et al., 2006; Brégeon and Doetsch, 2011), which suggests that spontaneous DNA damage is also a source of spontaneous mutations.
In the large intestine, small intestine and liver, mutations have been observed to accumulate at a steady rate of approximately 40 novel mutations per year, despite differences in proliferation rate, with the liver possessing a lower proliferation rate than intestines (Blokzijl et al., 2016). However, the mutation spectra varied among tissues. Collectively, these observations suggest that the origin of mutations and the mechanisms of mutagenesis vary according to cell state. If the cancer risk increases with the number of cell divisions and increased somatic mutation burden, large and long-lived animals such as elephants or whales should have a high cancer incidence rate. However, these species exhibit a much lower cancer incidence rate than humans (Nagy et al., 2007; Abegglen et al., 2015; Tollis et al., 2017). These observations suggest that species such as elephant and whale have developed effective mechanisms for cancer prevention and that there is not always a clear correlation between the number of cell divisions and cancer incidence. Further understanding of the basic mechanism of maintaining genome integrity in soma and germ cells may be derived from knowing the germline and somatic mutation rate in these cancer-resistant animals.
Germline de novo mutation (DNM) is defined as sequence change that arises during a generation of parental germ cells. It can be experimentally identified by next-generation sequencing (NGS), as mutations present in all the tissues of the offspring but not in the somatic tissues of parents (Roach et al., 2010; O’Roak et al., 2012). Knowing the lineage history of germ cells is key to understanding the origin of germline DNM. During a single generation of the germ cell lineage, defined as one cycle from zygote to gamete, both male and female gametes will have undergone multiple mitoses and one meiosis; the total number of mitotic cell divisions and DNA replications differs between male and female (Fig. 1). The father actually transmits more DNMs to the offspring than the mother does, as will be discussed in detail in a later section.
Human germ cell lineage. De novo germline mutations occur when female and male gametes undergo multiple mitoses in the course of a generation. A dot in the nucleus represents a de novo mutation that arose in a cell of the paternal germline (arrowhead) and was transmitted to the offspring. Such a mutation is detected in somatic cells of the offspring, but not of the father. The number of chromosome replications in female and male germ cells, at each developmental stage, is summarized in the bottom panel. The developmental stages before and after differentiation of primordial germ cells are designated pre-PGC and post-PGC, respectively. Post-puberty encompasses all life after puberty. In the male germline, spermatogonial stem cell divisions occur every 16 days post-puberty, for a total of ~23 replications per year, in addition to four replications during differentiation (Crow, 2000; Rahbari et al., 2016).
Early in development, approximately 10 cell divisions from the zygote, primordial germ cells (PGCs) appear and become committed to the germ lineage. Mutations acquired prior to this stage persist in both somatic and germ cells, a phenomenon called mosaicism, while mutations acquired by committed PGCs are restricted to germline cells, resulting in germline mosaicism (Freed et al., 2014; Campbell et al., 2015). In humans, it is difficult to determine germline mosaicism experimentally because of the ethical and technical issues of obtaining multiple tissue samples from a single donor without citing specific reasons in advance. Therefore, animal models are required for such studies. Indeed, a similar analysis of several generations of a mouse mutation accumulation line (Fig. 2) is particularly relevant to studies of the origins, molecular mechanisms, rates and spectra of DNMs.
Pedigree-based germline DNM analysis. To detect human germline DNMs, samples from the father, mother and child are often analyzed by NGS trio analysis (left, upper panel). Samples from multi-sibling families are also used to determine the origin of mutations (left, lower panel). A model mouse mutation accumulation line is described in the right panel. This mouse line is expanded by intra-generational crossing from one pair of mice. Germline DNMs occur at each generation and are transmitted and accumulated in subsequent generations. DNMs are efficiently identified by comparing the genomes of ancestor and offspring mice, enabling calculation of the average mutation rate per generation.
Recent progress in NGS, including whole-genome or exome sequencing, has enabled direct detection of DNMs by pedigree-based analysis of parent-offspring trios, with samples from other family members also used occasionally (Fig. 2). Based on such studies, the average rate of mutations that result in single-nucleotide variations (SNVs) was estimated to range from 0.96 to 1.29 × 10−8 per nucleotide per generation in the human germline; on average, 1.2 × 10−8 per nucleotide per generation (Roach et al., 2010; Conrad et al., 2011; Campbell et al., 2012; Kong et al., 2012; Michaelson et al., 2012; Ségurel et al., 2014; Besenbacher et al., 2015; Lynch, 2016; Rahbari et al., 2016; Wong et al., 2016; Jónsson et al., 2017). The rate for small insertions and deletions (indels) was estimated to be 0.53 to 1.5 × 10−9 per nucleotide per generation (Kondrashov, 2003; Lynch, 2010a; Campbell and Eichler, 2013; Ramu et al., 2013; Besenbacher et al., 2015), and 0.16 per generation for structural variants (large indels of >20 bp) (Kloosterman et al., 2015). Unfortunately, it is quite difficult to determine the mutation rates for repeat sequences such as microsatellite sequences because they vary by motif and number of repeats, and rate estimates are strongly influenced by frequent PCR errors. However, Sun et al. carefully estimated the mean DNM rate for dinucleotide repeats as 2.73 × 10−4 per locus per generation (Sun et al., 2012).
The majority of DNMs in offspring, especially SNVs, are transmitted from the paternal germ line, and the number of DNMs also significantly correlates with the father’s age (Kong et al., 2012). This is direct molecular evidence for the observed clinical link between advanced paternal age and increased risk of genetic disorders, such as achondroplasia, in the offspring (Crow, 2000). One plausible explanation for the difference in DNM content between the father and mother is that male gametes undergo DNA replications much more frequently than female gametes (Gilbert, 2000), therefore acquiring more germline mutations. Indeed, human male gametes undergo as many as 380 and 610 DNA replications by age 30 and 40, respectively (Crow, 2000), while female gametes undergo only 23 (Crow, 2000) or ~30 (Rahbari et al., 2016) regardless of age, because DNA replication does not occur postnatally (Fig. 1). Accordingly, the ratio of the frequency of male germ cell divisions to that of female germ cell divisions increases with age. In an adult male, spermatogonial stem cells divide every 16 days, for a total of 23 divisions per year and for an average of 1–2 additional DNMs per year (Campbell et al., 2012; Kong et al., 2012; Michaelson et al., 2012; Rahbari et al., 2016; Jónsson et al., 2017). On the other hand, the average number of additional DNMs per year of maternal age was reported to be about 0.25 to 0.37. Thus, the effect of maternal age on the rate of de novo SNVs in the post-replicative stage of female germ cells is small, but significant (Goldmann et al., 2016; Jónsson et al., 2017).
If a large proportion of the DNMs results from DNA replication errors, then the DNM rate per generation must be proportional to the number of cell divisions. However, some observations do not fit this assumption. The cell division frequency for paternal germ cells is over 10 times higher than that for maternal germ cells although the number of paternally derived DNMs is only several times more than that of maternally derived DNMs (Ségurel et al., 2014). These results suggest that DNMs are derived not only from replication-based processes but also as a result of spontaneous base damage and subsequent repair. In fact, age-dependent accumulation of DNMs was observed in post-replicative maternal germ cells (Blokzijl et al., 2016; Goldmann et al., 2016; Jónsson et al., 2017). However, other interpretations should be considered, such as the mutation rate per replication not remaining constant during development (Walter et al., 1998; Rahbari et al., 2016), inaccuracy in the estimated number of cell divisions (Scally and Durbin, 2012), or the presence of selection mechanisms for mutation-free germ cells.
Base substitution mutations are classified either as transitions, in which a purine base replaces another purine base or a pyrimidine base replaces another pyrimidine base, or as transversions, in which a purine base replaces a pyrimidine base or vice versa. Transitions are less likely to result in an amino acid change than transversions. Note that each of the above base-to-base substitutions may be caused by different mechanisms of mutation.
The spectra of germline base substitutions obtained in several pedigree-based NGS studies are similar (Fig. 3). Transitions were more frequently observed than transversions, accounting for more than 60% of total substitutions (Campbell et al., 2012; Kong et al., 2012; Michaelson et al., 2012; Goldmann et al., 2016; Rahbari et al., 2016; Wong et al., 2016). The most commonly observed transitions, G:C to A:T, are generally attributed to the high mutability of CpG sites as a result of spontaneous oxidative deamination of 5-methylcytosine (Coulondre et al., 1978; Cooper and Krawczak, 1989; Kong et al., 2012). Indeed, 5-methylcytosine and cytosine in mammalian DNA are well known to undergo deamination to thymine and uracil, resulting in G:T and G:U mismatches. Usually, these mismatches are efficiently repaired by base excision repair.
Spectra of germline de novo mutations. The relative ratio of germline de novo one-base substitutions in humans (upper panel) and mice (lower panel) is plotted in stacked bars, corresponding to separate studies. For human data, A) Campbell et al., 2012, B) Michaelson et al., 2012, C) Kong et al., 2012, D) Goldmann et al., 2016, E) Rahbari et al., 2016 and F) Wong et al., 2016 are cited. For mouse data, Uchimura et al., 2015 and Ohno et al., 2014 are cited.
Additional mechanisms of G:C to A:T mutations must exist, since a high incidence of this mutation is also observed at non-CpG sites. While the frequency of DNMs at CpG sites varies greatly among studies (Kong et al., 2012; Francioli et al., 2015; Rahbari et al., 2016; Wong et al., 2016), the rate of transitions at CpG sites is ~18-fold greater than at non-CpG sites, after adjusting for site variability (Nachman and Crowell, 2000; Kondrashov, 2003; Kong et al., 2012; Wong et al., 2016). G:C to A:T mutation constitutes approximately 40% of the total DNM, even though CpG sites only account for approximately 1% of the human genome. Transversions also occur at 2.5-fold higher incidence at CpG sites than at non-CpG sites, indicating that the mutability of CpG sites is not exclusively due to deamination of 5-methylcytosine. The primary cause of A:T to G:C transitions, which constitute approximately 25% of total substitutions (Fig. 3) in the human germline, remains unclear. Several have been proposed, including hypoxanthine generated by the spontaneous deamination of adenine or 2-hydroxyadenine generated by the oxidation of adenine, both of which are able to pair with cytosine as well as thymine to form hypoxanthine:C and 2-hydroxyadenine:C pairs and could thus promote A:T to G:C transition (Basu et al., 1989; Kamiya and Kasai, 1997; Masaoka et al., 2001; Valentine and Termini, 2001).
Transversions include four types of base substitutions, G:C to T:A, G:C to C:G, A:T to C:G and A:T to T:A. Each of the first two accounts for ~10% of all DNMs, and each of the latter two constitutes less than 8% (Fig. 3). In human DNMs, transversions are observed approximately 50% less frequently than transitions (Fig. 3), since purine-purine or pyrimidine-pyrimidine pairs in the DNA double helix are energetically unfavorable. One major cause of the G:C to T:A mutation is thought to be 8-oxoguanine (8-oxoG). This oxidatively modified guanine base can form a Hoogsteen pair with adenine as well as a canonical Watson-Crick pair with cytosine (Kasai and Nishimura, 1984; Grollman and Moriya, 1993). Thus, during DNA replication, incorporation of adenine opposite 8-oxoG leads to an 8-oxoG:A pair, promoting the G to T mutation, while incorporation of 8-oxoG opposite adenine leads to an A:8-oxoG pair, promoting the A to C mutation (Kasai and Nishimura, 1984; Cheng et al., 1992; Grollman and Moriya, 1993; Egashira et al., 2002; Sekiguchi and Tsuzuki, 2002; Nakabeppu et al., 2006).
The mutation spectra of paternally and maternally derived DNMs are comparable, although the former contain a marginally higher frequency of G:C to T:A and A:T to C:G and a marginally lower frequency of G:C to A:T mutations (Goldmann et al., 2016; Rahbari et al., 2016). Moreover, paternally and maternally derived DNMs tend to occur in different sequence contexts, which are now termed mutational signatures (Alexandrov et al., 2013; Goldmann et al., 2016). Interestingly, the mutational signature of DNMs in the youngest fathers is more similar to that in mothers than to that in older fathers (Goldmann et al., 2016), suggesting that different factors affect the mutation spectra and rates as male and female germ cells develop.
The local mutation rate varies across the human genome and is correlated with the state of the local chromatin. For instance, the density of human single-nucleotide polymorphisms, as well as the sequence divergence between humans and other primates, is higher in late-replicating genomic regions than in early-replicating genomic regions (Watanabe et al., 2002; Stamatoyannopoulos et al., 2009; Koren et al., 2012, 2014). Similarly, somatic mutations are more frequent in late-replicating regions in cancer genomes (Liu et al., 2013). In general, regions that replicate early in S-phase are GC-rich, gene-rich, and reside in a transcriptionally active open-chromatin state. In contrast, regions that replicate at late S-phase are AT-rich, gene-poor, and reside in a transcriptionally inactive closed-chromatin state (Rhind and Gilbert, 2013). Similarly, germline DNMs detected from analysis of human trios are not randomly distributed throughout the genome. Indeed, the local DNM rate correlates with replication timing at megabase resolution, with late-replicating regions containing more DNMs than early-replicating regions (Francioli et al., 2015). Notably, such differences were statistically significant in DNMs from younger fathers but were not significant in DNMs from older fathers, suggesting that the mechanism driving local DNM rates undergoes age-related changes (Francioli et al., 2015).
Several possible mechanisms underlying the increased mutation rate in late-replicating regions can be inferred. The fidelity of DNA polymerases and the efficiency of mismatch repair are influenced by the absolute and relative concentrations of dNTPs (Buckland et al., 2014; Watt et al., 2016). It is known that the absolute concentration of the dNTP pool decreases with progression of S-phase (Kumar et al., 2011; Stillman, 2013); also, the relative ratio of dNTPs fluctuates (Kenigsberg et al., 2016). Both base substitutions and indels are increased by an imbalanced dNTP pool (Herrick, 2011; Kumar et al., 2011; Watt et al., 2016). Spontaneous DNA lesions may also increase at late-replicating regions due to the formation of single-stranded DNA, which is more susceptible to chemical modifications such as oxidation, deamination, depurination, depyrimidination and alkylation (Stamatoyannopoulos et al., 2009). Indeed, both single-stranded and double-stranded DNA breaks are formed at stalled or blocked replication forks, which occur more frequently in late-replicating heterochromatic regions than in early-replicating euchromatic regions (Sogo et al., 2002; Mirkin and Mirkin, 2007; Zeman and Cimprich, 2014). Differences in the effectiveness of DNA repair during S-phase may also impact the local mutation rate. Nucleotide excision repair and base excision repair are both more efficient in euchromatin than in heterochromatin (Sanders et al., 2004; Amouroux et al., 2010). In addition, loss of DNA mismatch repair was found to decrease the variability of regional mutation rates within euchromatin and heterochromatin, suggesting that less efficient repair occurs in late-replicating regions (Lang and Murray, 2011; Lujan et al., 2014; Supek and Lehner, 2015).
Some DNMs also form clusters, which consist of ~2 - 3 closely spaced mutations in the same allele, named clustered mutations (Schrider et al., 2011; Campbell et al., 2012; Michaelson et al., 2012; Francioli et al., 2015; Goldmann et al., 2016). Interestingly, clustered mutations were found more frequently in DNMs from older fathers and mothers, although there was no significant difference in the frequencies and spectra of mutations derived from either parent (Goldmann et al., 2016), and C:G to G:C and C:G to A:T transversions are more frequent in clustered mutations than in sporadic mutations (Francioli et al., 2015; Goldmann et al., 2016). Both types of substitutions are generated by unrepaired 8-oxoG (Shibutani et al., 1991; Kino and Sugiyama, 2000), which is spontaneously formed in the nuclear genome by oxidative stress and accumulates with age (Nie et al., 2013). Indeed, the increase of oxidative stress with age (Kryston et al., 2011) may explain why clustered DNMs are more frequent in the offspring of older parents. It has also been pointed out that some clustered DNMs are attributed to mutagenesis mediated by APOBEC family enzymes (Francioli et al., 2015; Goldmann et al., 2016). These enzymes convert cytosine bases to uracil, are known to introduce clustered mutations within a DNA strand, and are implicated in tumorigenesis (Roberts et al., 2013).
Nevertheless, limited information is available for elucidation of the mechanisms that generate these local differences in genomic fate. This topic deserves thorough investigation, since small differences in de novo mutability may be sufficient to confer divergent evolutionary futures on these genomic regions.
In contrast to the extensive studies of human germline mutations, few NGS reports of germline DNMs in laboratory mice have been published. The average de novo base substitution rate in wild type C57BL/6J mice was reported to be 5.4 × 10−9 per nucleotide per generation (Uchimura et al., 2015), roughly half of the rate observed in humans and much lower than previous estimates of 1.1 × 10−8 (Drake et al., 1998) and 3.7 × 10−8 (Lynch, 2010a). It should be noted that the previous estimates were obtained from specific locus tests or in vivo reporter gene assays, in contrast with the new result obtained by whole-genome sequencing and family-based analysis of one female-male pair bred for more than 20 generations by repeated intra-sib mating.
The difference between humans and mice in the rate of DNMs cannot be attributed solely to the frequency of DNA replication during germ cell development. The frequency of cell division per generation is approximately five times greater in humans than in mice, while the DNM rate per generation in humans is only twice that in mice. The number of divisions that the germ cells undergo has been estimated to be 62 in male mice and 25 in female mice at age 9 months (Drost and Lee, 1995), but 380 in male humans and 30 in female humans at age 30 (Crow, 2000; Rahbari et al., 2016). Thus, the mutation rate per replication may be lower in humans than in mice (Milholland et al., 2017), or it may not be constant throughout germ cell development (Rahbari et al., 2016).
Transitions are more frequent than transversions in mice, with G:C to A:T mutations being the most common (Fig. 3). The base substitution rate at CpG sites is also 16.8 times higher than at non-CpG sites. As these observations are consistent with the characteristics of human DNMs, similar factors are probably involved in generating DNMs in human and mouse germ cells.
DNA replication errors and spontaneous DNA lesions are considered to be major causes of spontaneous DNMs in mammals, along with DNA repair activity. Hence, experimental studies on laboratory mice will help researchers to elucidate the natural causes of DNMs, the relative contribution of each pathway to genome integrity in germ cells, and the impact of DNMs on phenotype. To my knowledge, there are a limited number of reports of NGS- and family-based analysis of spontaneous germline DNMs in mice available at the present time. Here, the focus is on two studies, based on data obtained from the mutator mouse lines Pold1exo/exo (Uchimura et al., 2015) and TOY-KO (Mth1-/-/Ogg1-/-/Mutyh-/- triple knockout) (Ohno et al., 2014). Pold1exo/exo mice lack 3′-5′ exonuclease proofreading activity in DNA polymerase delta and exhibit a high rate of spontaneous mutation due to increased nucleotide misincorporation during DNA replication. A deficiency in DNA polymerase proofreading activity increases the tumor incidence in mice and humans (Goldsby et al., 2002; Albertson et al., 2009; Yoshida et al., 2011; Palles et al., 2013). To clarify the effect of this deficiency on the germ cell genome, homozygous Pold1exo/exo mice in a C57BL/6J genetic background were generated and maintained for more than 20 generations by intra-sib mating, in order to analyze DNMs (Uchimura et al., 2015). In comparison with wild type mice, the mutator mice exhibited a high frequency of abnormal phenotypes, lower breeding efficiency and elevated mortality. The average base substitution rate was, at approximately 9.4 × 10−8 per nucleotide per generation, 17 times higher than in wild type mice. All types of substitutions were more frequent in Pold1exo/exo mutator mice than in the wild type. In particular, transversions were remarkably increased, and no G:C to A:T bias was observed (Fig. 3). These results suggest that the deficiency in proofreading activity of DNA polymerase delta alters the DNM rate and spectrum, not only in somatic cells but also in germ cells, and despite fully functional DNA mismatch repair.
Oxidative DNA lesions are also considered to be a major natural cause of mutations in many organisms, especially since reactive oxygen species are generated during normal metabolism and during exposure to environmental factors such as radiation or chemicals (Bhattacharyya et al., 2014; Krumova and Cosa, 2016). Of the four bases, guanine is the most susceptible to oxidation. Its major oxidized form is 8-oxoG, known as a potent inducer of G:C to T:A transversions (Shibutani et al., 1991; Smith, 1992). Interestingly, 8-oxoG is spontaneously and constantly generated, and it accumulates unrepaired in specific genomic regions in vivo (Ohno et al., 2006). In humans and mice, it is known that the three enzymes MTH1, OGG1 and MUTYH cooperate to minimize G:C to T:A mutations induced by 8-oxoG in somatic cells, and that deficiencies in these enzymes increase somatic mutation rates and cancer susceptibility (Tsuzuki et al., 2001, 2007; Sakumi et al., 2003; Sakamoto et al., 2007). However, the impacts of 8-oxoG and its repair systems on germline mutations had not previously been elucidated. Thus, we generated Mth1-/-/Ogg1-/-/Mutyh-/- triple knockout (TOY-KO) mice, which we maintained to the 8th generation by intra-generational mating (Ohno et al., 2014). These mice exhibited a high somatic mutation rate, high tumor incidence, short life span and low reproductive rate. Several phenotypic variations, such as hydrocephalus and belly white spot, were observed among progeny and were presumed to arise from DNMs. To identify germline DNMs, mice from the 5th, 7th and 8th generations were analyzed by whole-exome sequencing, and candidate SNVs were re-sequenced in the progenitor mice. Finally, a total of 262 SNVs were confirmed, for an average base substitution rate of 20 × 10−8 per nucleotide per generation, a rate 37 times greater than in wild type mice, with a rate of 5.4 × 10−8 per nucleotide per generation (Uchimura et al., 2015). Notably, 96% of mutations were G:C to T:A transversions (Fig. 3), presumably due to spontaneously generated 8-oxoG, which accumulated and remained unrepaired. Taken together, the data suggest that DNA oxidation occurs spontaneously in germline cells in vivo, but efficient repair systems minimize and stabilize the DNM rate under normal environmental conditions. Accordingly, excessive oxidative stress or compromised DNA repair due to endogenous or exogenous factors may alter the rate or spectra of DNMs, even in wild type animals.
Recent data obtained on human DNMs suggest that the causal mechanisms of spontaneous germline mutations vary by genomic region, mutation type, sex and age of parents. For a comprehensive understanding of the causes and the mechanisms of DNM in mammals, mouse mutation accumulation lines (Fig. 2) are uniquely advantageous because of large litter size, short generation time, availability of tissue samples from all members of a family, and observable phenotypic effects. Indeed, we could determine the origin of each DNM found in the TOY-KO mouse line by tracing back the family tree from the founders. For example, in our previous study, 11 of 262 DNMs detected in TOY-KO mice were presumed to arise from germline mosaicism (mutations that occurred at the post-PGC stage) in either father or mother, since these mutations were found in tail tissues from multiple litters, but not in tail tissues from parents (Ohno et al., 2014). Another three DNMs on the X chromosome appear to have occurred in the pre-PGC stage embryos of the male offspring: these were detected as somatic mosaics in their tail tissues but were not detected in the tails of either parent (Ohno et al., 2014). The male has only one X chromosome, which always comes from the mother. Thus, if more than one allelic sequence is found in the son, it is likely to be somatic mosaicism. Also, if the mutant allele frequency is more than a quarter, the mutation is likely to have occurred at the very early embryonic stage. This is direct evidence for the occurrence of germline mutations at different developmental stages of the germ cell lineage. Note that the rate and spectrum of DNMs at each developmental stage may be affected by multiple factors (Rahbari et al., 2016), so that the mutation rate and spectrum may not be constant through germ cell development.
DNA polymerase proofreading activity and DNA repair status are the primary determinants of the rates and spectra of germline DNMs. Therefore, species-specific or strain-specific sequence polymorphisms in DNA polymerase or DNA repair genes affect DNM rates and spectra and may lead to different genomic futures in the long term. Moreover, rates for mutations beyond base substitutions, including large indels, mutations in repeat sequences, copy number change and genome rearrangements, are compelling to study as these mutations are reportedly consequences of radiation exposure, a major environmental mutagen. While these sequence variations remain challenging to detect by short-read NGS, the use of new sequencing technologies, such as long-read or single-cell sequencing, will serve to advance research in this area by enabling detailed genomic analysis and illumination of the environmental or intrinsic factors that affect molecular evolution and impact human genetic diseases.
I thank Dr. K. Sakumi, Dr. Y. Gondo, Dr. A. Uchimura, Dr. T. Ikemura and Dr. Y. Nakabeppu for collaborating on mouse germline mutation studies. I also thank Dr. T. Tsuzuki for continuous support and discussion and Dr. Y. Nakatsu, Dr. N. Takano, Ms. F. Sasaki and all members of our laboratory for their support and useful advice. This work was supported by JSPS KAKENHI Grant Numbers 22300144, 25650130, 26281022, 25241016.
