.
F U L L T E X T S O U R C E : Cell
Highlights
Both the age of the father and the age of the mother are positively correlated with the number of de novo mutations (DNMs) in offspring, with the effect size of paternal age being larger.
DNMs associated with paternal and maternal aging each have unique mutational signatures based on their nucleotide substitution spectrum and genomic locations.
DNM clusters have characteristics distinct from nonclustered DNMs, suggesting different underlying mutational causes.
Post-zygotic mutations arising during early embryonic development are a frequent phenomenon and differ from nonmosaic germline DNMs in their mutational spectrum.
A subset of presumed DNMs can be traced back as low-level mosaic mutations in the somatic tissue of a parent. These mutations have a distinct mutation spectrum and can recur in future offspring.
Human germline de novo mutations (DNMs) are both a driver of evolution and an important cause of genetic diseases. In the past few years, whole-genome sequencing (WGS) of parent–offspring trios has facilitated the large-scale detection and study of human DNMs, which has led to exciting discoveries. The overarching theme of all of these studies is that the DNMs of an individual are a complex mixture of mutations that arise through different biological processes acting at different times during human development and life.
De novo mutations
Human de novo mutations (DNMs, see Glossary) are germline mutations that newly occurred within one generation. While the vast majority of the genome has been inherited from earlier generations, DNMs provide new genetic variation. The consequences of the new genetic mutation can vary widely. While neutral or advantageous mutations might become established in the genome of our species and thereby contribute to human evolution, changes to crucial genetic sequences can also result in misfunctioning of biological systems, resulting in severe disease. One of the earliest known examples of this was Down syndrome, which is caused by a de novo trisomy of chromosome 21 [1, 2, 3]. In recent years DNMs have been found to be a prominent cause of neurodevelopmental diseases, including intellectual disability, autism, and schizophrenia [4, 5]. The unbiased study of de novo point mutations in humans was for many years hampered by the lack of techniques to scan the entire genome in a cost-effective way. The introduction of next-generation sequencing (NGS) technologies has spurred investigations of DNMs in humans [6]. DNMs can refer to a variety of different mutation types, such as single-nucleotide substitutions, insertions, deletions, and copy-number variants (CNVs). In this review we focus on single-nucleotide mutations and review the progress made in this field since the introduction of WGS, exploring their biology and possible underlying mechanisms (Figure 1, Key Figure), but not the potential pathological consequences.
Figure 1Key Figure. Overview of Distinct Classes of Germline De Novo Mutations (DNMs)
(A) Schematic depiction of the biological niches where DNMs arise. The earliest DNMs of an individual arise during the embryonic development of both parents, starting with the specification of primordial germline cells during gastrulation (top schemes; green bolts). During aging of the father, DNMs arise in spermatogonial cells, which replicate continuously to produce sperm (mid-left scheme; blue bolts). Aging of the oocytes in the mother is not accompanied by genome replications, but mutations still arise (mid-right; red bolt). Finally, post-zygotic mutations during early embryonic development of the offspring can affect a large fraction of the offspring’s cells (bottom scheme; orange bolt). (B) While aging-associated mutations are more predominant in offspring from parents of advanced age, the number of embryonic DNMs is independent of parental age. Parent–offspring trio sequencing experiments do not allow the analysis of mutations arising during childhood; therefore, uncertainty remains about DNM accrual before the fertile period. © Each class of DNM has a distinct mutational spectrum. All numeric values are based on observations by Sasani et al. [7].
The Paternal Age Effect
One of the most prominent findings concerning germline DNMs is the fact that their number increases steadily with the age of the father at conception. The concept of mutations being associated with age was first described by the human geneticist Wilhelm Weinberg in 1912, who observed that sporadic achondroplasia is more common in last-born children than in first-born children [8]. For several sporadic genetic diseases, the association with paternal age is stronger than the association with maternal age [9]. It has been hypothesized that the underlying cause of these associations could be unprecise genome copying during the large number of cell divisions required for continuous sperm cell production in males [8, 9, 10]. The first direct observation of DNM rates being higher for offspring of fathers of advanced age was made by WGS of 78 Icelandic parent–offspring trios [11]. This study described that the number of DNMs in the offspring is dependent on the age of the father, with a slope of two additional DNMs per year of life of the father at the time of conception. This association of increased paternal age with increased numbers of DNMs in offspring is generally called the paternal age effect.
The main hypothesis of the underlying cause for the paternal age effect is that mutations arise from incidental copying errors during genome replications [8, 9, 10, 11]. Genome replications are frequent in the male germline to ensure continuous generation of sperm cells. The stem cells for sperm production are spermatogonia, located in the seminiferous epithelium. Spermatogonia divide to renew themselves and give rise to spermatocytes, which will eventually differentiate into sperm cells [12]. During aging, the mean number of divisions per spermatogonium continues to increase, with each replication potentially introducing new mutations due to copying errors. Due to this, spermatogonia are believed to accumulate mutations during aging, thereby explaining the paternal age effect.
Next to this genome-wide paternal age effect, an additional paternal age effect has been observed for highly specific mutations. Due to particular activating mutations in specific genes, affected spermatogonia lineages can generate more sperm cells than other lineages. As a consequence, the effective germline mutation rate in these genes is up to four orders of magnitude higher than the genome-wide rate. This effect has been termed selfish spermatogonial selection (Box 1). The association of these mutations with paternal age is stronger than the linear genome-wide paternal age effect [13].
Box 1.
Selfish Spermatogonial Selection
The mutations that accumulate in spermatogonia during paternal aging can also have an effect on the fitness of the spermatogonia themselves. While mutations in essential genes might lead to the eradication of the spermatogonial lineage, other mutations can result in selective advantages and lead to clonal expansion; that is, the outgrowth of the specific cell lineage, a process often linked to precancerous lesions. Clonal expansions have been observed in human tissues like bone marrow [14], skin [15], and esophageal epithelium [16], as well as many others [17]. In the human testis, clonal expansion has been observed to be driven by oncogenic mutations of growth factor receptors and components of the RAS–MAPK pathway [13, 18]. As a consequence of the expansion of mutated spermatogonia that occurs in the testes of healthy men during aging [19, 20, 21], the number of sperm carrying these mutations grows continuously at rates much higher than the general, genome-wide paternal age effect. If these mutations are passed on to offspring, they will cause developmental disorders. One prominent example is achondroplasia, a condition characterized by short stature that is caused by mutation of the fibroblast growth factor receptor 3 (FGFR3) gene. Achondroplasia occurs more frequently than expected based on average mutation rates and its occurrence is strongly associated with paternal age [10, 22]. It was shown that these pathogenic mutations are often present as high-level mosaicisms in the testes of the fathers of achondroplasia patients [20, 22].
Recent studies have confirmed the observation of the general paternal age effect [23, 24, 25, 26, 27, 28]. The generation of large-scale datasets allowed deeper characterization of the DNMs, including the analysis of mutational spectra and mutational signatures (Box 2). The mutational spectrum of human DNMs has been found to mainly comprise two mutational signatures that are known from somatic mutations in cancer [24, 29]. These two signatures occur in a wide range of somatic tissues and accumulate mutations in a clock-like manner (i.e., at a constant rate). Therefore, in somatic tissues these signatures are likely to represent the impact of aging on DNA. In line with the replication-error hypothesis, one of these signatures is associated with genome replication [29]. Surprisingly, however, the mutational spectrum of DNMs is not constant and changes with the age of the father at conception. With advancing paternal age, more T>G substitutions and fewer non-CpG C>T substitutions are found in the offspring [27, 28]. The existence of such differences suggests that, besides the paternal aging-associated mutations, there is another type of mutation that is independent of paternal age. With increasing age of the father, the balance of these types of mutation shifts and the mutational spectrum becomes biased towards more paternal aging-associated mutations.
Box 2.
Mutational Signatures
From the analysis of cancer genomes it is known that exposure to mutagens or failure of DNA repair pathways can result in mutations with specific patterns. An example is the light-associated skin cancers, in which C>T and CC>TT nucleotide substitutions are very common [30]. The advent of NGS technologies allowed the systematic study of mutational patterns in large collections of well-annotated cancer genomes [31, 32, 33]. In these analyses, the mutational signatures were analyzed by differentiating the six nucleotide substitution types and stratifying by the nucleotides surrounding the affected base pair, resulting in spectra of 96 mutation types. By mathematically deconvoluting archetypical patterns in the spectra of cancer mutations, these efforts identified 49 single-nucleotide-substitution mutational signatures. These signatures are well annotated and a subset could be linked to underlying mutational mechanisms. The mutational signatures are available online (https://cancer.sange...smic/signatures). Analysis of a mutational spectrum and comparison with the known signatures can hint at the underlying mutational mechanisms.
The Paternal Age Effect and Estimations of the Number of Germline Genome Replications
According to the genome-replication hypothesis for the paternal age effect, the number of replications should be directly proportional to the number of DNMs in offspring. The established model for estimating the number of genome replications that occurred during development from zygote to sperm cell includes the following factors [10, 34, 35]. First, during very early embryogenesis, about ten cell divisions occur until germline cells are specified and the germline is established. Second, during sex organ development, the seminiferous ducts become colonized with germ cells, which requires about 24 germline cell divisions after germ cell specification. During the further fetal period and childhood, the seminiferous epithelium rests and no cell divisions occur. Starting in puberty, spermatogonia will divide continuously to generate sperm cell precursors. It has been estimated that the average spermatogonium divides 23 times per year during this period [36]. Finally, an age-independent number of four genome replications is needed to proceed from spermatogonium to sperm cell. To give an example, the number of genome replications of an average sperm cell of a 30-year-old male can be estimated as ten very early embryonic divisions, plus 24 early embryonic divisions, plus 17 years of sperm production each with 23 divisions (assuming a sperm production onset at the age of 13 years [37]), and finally four replications for spermatogenesis. This is in total 429 replications. Fathers at the age of 20 and 60 years would generate sperm after 199 and 1119 replications, respectively.
However, these estimates for genome replications are not proportional with the observations on DNM accrual (Figure 2). While the number of replications increases by a factor of more than five between the parental ages of 20 and 60 years, the number of DNMs increases by less than a factor of three, as the estimated numbers would be 40 and 91, respectively [28]. Supporting the observations on DNMs, the epidemiological risk of giving birth to a child affected by a developmental disorder (of which a large fraction is caused by DNMs [38]) increases with paternal age only by a factor of less than two (from 0.24% for fathers and mothers both below 22 years up to 0.47% for fathers and mothers both above 42 years) [39]. This mismatch of the observations with the genome replication data has been reported several times and remains unresolved [40, 41, 42, 43, 44].
Figure 2Estimated Genome Replications in the Male Germline versus Number of De Novo Mutations (DNMs) in Offspring.
During the reproductive period, the number of genome replications is estimated to grow by a higher factor than the number of observed DNMs (see text). Black dots indicate observations of DNM numbers in parent–offspring trios; the black line indicates a linear fit to the data [28]. Green line and green axis indicate the estimated number of genome replications in the male germline. Dark-gray area, prenatal period; light-gray area, childhood.
A crucial factor for the genome-replication model is the estimate of the annual post-pubertal spermatogonium division rate. This estimate of 23 divisions per year is based on observations in the 1960s by Heller and Clermont [36], who radiolabeled the seminiferous epithelium of the testicular ducts and showed that one division cycle lasts 16 days, which would extrapolate to 23 divisions per year. This extrapolation implicitly assumes that there are no interruptions or resting phases between the divisions. However, more recent studies have observed the nature of spermatogonial self-renewal to not be a unidirectional process but rather be driven by stochastic cell transitions, including cell stages without replicative activity [12, 45, 46]. Such a nonhierarchical model of spermatogonial self-renewal with stochastic, continuous oscillation between cell stages is supported by recent single-cell studies in two ways. First, analyses of single spermatogonial transcriptomes have shown that there is a continuous distribution of cells that do not express replication markers on the one hand and cells that are expressing markers of both replication and sperm differentiation on the other hand [47, 48, 49, 50, 51]. Second, spermatogonia have been found to be able to adapt their transcriptomes into both directions, either towards the replication-inactive state or into the amplification and differentiation state [47]. A spermatogonium division model that incorporates such oscillation with dormant cell stages would imply a slower rate of spermatogonium divisions and would therefore provide a better fit with the observations on DNM accrual. Such a model has been proposed and its parameters have been estimated by a fit to DNM accumulation data [43]. A mean spermatogonium division time parameter of 300 days provided the best concordance with the observations. According to this model, the yearly increase of spermatogonium divisions would be not 23 divisions per year, but 0.8. Subsequently, the number of germline divisions for sperm of 20-year-olds and 60-year-olds would be 44 and 76, respectively. This constitutes an increase by a factor of 1.7, which would be similar to the increase in DNM numbers. In addition, a lower rate of spermatogonium divisions would also imply that the human spermatogenesis per-replication mutation rate would be more similar to the same rate in the mouse as well as to the per-replication mutation rates for other phases of human germline biology [52, 53]. Taken together, recent observations of DNM accumulation and single-cell transcriptome analyses of the spermatogenic epithelium suggest that estimated spermatogonium division rates may have been based on wrong assumptions.
.../...
.
