The role of somatic mosaicism in brain disease

Abstract

In this review we discuss the importance of genetic somatic mosaicism and its impact on brain diseases. We start from introducing the different types of somatic mutations, their frequencies and abundances across development and lifespan. We then describe how weakness in DNA repair mechanisms influence their prevalence. Finally, we address their functional consequences in the brain and review recent research showing their unsuspected importance in several neurodevelopmental, psychiatric, and neurodegenerative diseases.

1. Somatic mutations and genomic mosaicism

Population genetic heterogeneity is defined by germline and de novo variants inherited by individuals from their parents. Germline variants are present in all parental tissues while de novo variants arise in parental germ cells. Another form of genetic heterogeneity develops after the formation of the zygote from mutations that can occur in each cell as the organism develops. Such post-zygotic mutations occur from the very first cell division, throughout embryogenesis and until the end of life, making each individual a genetic mosaic. The absolute majority of post-zygotic mutations occur in the soma after formation of the germline lineage – thus are called Somatic Mutations (SM) – and are not transmitted to the offspring. The later SM occur, the smaller the fraction of cells that will carry them. In fact, some variants are shared between tissues, while others will be confined to a particular organ or region [1]. Common types of SM are single nucleotide variants (SNVs), copy number variants (CNVs), structural variants (SVs), mobile element insertions (MEIs) and chromosomal aneuploidies (Box 1).

Box1: Origin of different types of genomic variations.

SNVs (Single nucleotide variants) result from single nucleotide sequence substitutions that can occur at any location of the genome. They can result from errors in nucleotide incorporation during DNA replication, spontaneous alteration (e.g. de-amination or oxidation), or promoted by a variety of external agents including UV light, radiations, and a variety of chemicals (Table 1).

Indels, short for insertions and deletions events, can span between one and thousands of nucleotides. They occur at DNA double stranded breaks (DSBs) after non-homologous end joining (NHEJ) repair, and they are often present near breakpoints of structural variations (SVs). DSBs have been identified at recurrent locations in primary neural stem/progenitor cells around long, transcribed, and late-replicating genes. DSB could originate from replication stress, involving the collision of replication fork with transcription of long genes, as it has been recently suggested [23,24].

MEIs (Mobile Element Insertions) are caused by ancient viral genomic sequences (e.g. Alu or L1 elements) that constitute about 50% of the genome and can potentially self-transcribe and re-insert in a new genomic location. Active L1 elements are transcribed into RNA and assemble in a multi-protein complex with reverse transcriptase and endonuclease cutting activities to insert new L1 elements into the genome.

CNVs (Copy Number Variations) and SVs (structural variants) are deletion or duplications of genomic segments of different possible lengths. Kilobase-scale CNVs likely arise as a result of a variety of mechanisms, including alterations and slippage of replication fork during DNA replication. Megabase-scale CNVs often results from non-allelic homologous recombination (NAHR), especially in highly repetitive segments of DNA. Finally, very large CNVs and loss of entire chromosomes might occur as a result of alterations in the mitotic spindle.

Aneuploidy is characterized by an abnormal number of chromosomes. It is caused by segregation errors during mitosis, resulting in uneven division of chromosomes into daughter cells.

2. Methods to detect, call and quantify SM in the brain

Detecting SM in brain cells is especially challenging due to limited access of brain tissue. Moreover, neuronal cells are mostly postmitotic, which precludes amplifying their DNA by cloning, a method used for other cell types [2,3]. Therefore, direct quantification of brain SM has relied mostly on the analysis of postmortem or post-surgical brain samples by high coverage whole genome or exome sequencing of tissue in bulk (WGS and WES, respectively) [4,5], clonal expansion of progenitor cells [6], or single-cell DNA enzymatic amplification [7].

Another approach is to identify SM in accessible peripheral tissues (e.g. blood, saliva or skin biopsies). Pre-gastrulation SM results in a high variant frequency in endoderm, mesoderm and ectoderm lineages and are therefore likely to be present in the neuroectoderm lineage. However, their actual frequency in the brain remains unknown, which limits the inference of causality between detected peripheral SM and a co-occurring brain disorder. Furthermore, SM exclusive to the brain would not be detected in peripheral tissues. Empirical estimation of genetic distance between tissue sources is necessary to be able to project occurrence of detected mosaic mutations in the saliva, blood, or urine to brain regions [8], and would be of great diagnostic help for SM screening.

3. Frequency of somatic mutations types across development and lifespan

Currently, several studies estimated the rates and burden of somatic SNVs in brain cells across the human lifetime. However, such estimates have only been partially accomplished for other variant types (Figure 1).

Figure 1.

Rate of somatic mutagenesis over the life span. Summary of literature quantifying rates of somatic single nucleotide variations (SNVs), structural variation (SVs), and mobile element insertions (MEI) over different epoch of the human life span.

SNVs are the most abundant form of SM. The rate of somatic SNV formation is highest during development. Our studies in clones of fetal brain progenitors [6], coupled with other studies in single neuron nuclei [7,9], suggested that there are already about 500–1000 somatic SNVs per neuron at birth. The rate of accumulation is about 1.3 SNVs per division per cell during cell divisions of the early human embryos with a likely increase (estimates range from 1.6 to 21 SNVs per division per cell) during neurogenesis. Normally, rates of SNV accumulation decline by orders of magnitude postnatally, to 0.1 SNVs per cell per day in adult somatic tissues [10].

CNVs/SVs are considerably less common than SNVs and are often generated during cell division. Since they involve long stretches of DNA potentially containing many genes, they can have direr functional consequences. The rate of CNVs has not been formally established, but there are on average 0.2 to 3.4 CNVs per cell, affecting between 5% to 30% of the neurons [11–13]. Surprisingly, their abundance seems to decrease with age, suggesting a compensatory mechanism where cells with deleterious CNVs undergo apoptosis [11–13].

Somatic MEIs seems present in 10 to 40% of adult neurons [14,15] but their time of origin is presently unclear. The fact that they are mostly not shared among cells suggests an origin in post-mitotic cells. However, the difficulty in their detection in single cells and validation in original tissue has precluded an unbiased assessment of their overall allele frequency and cell sharing in tissues. Although L1 retrotransposition occurs in all tissues of healthy individuals, these events preferably take place in neuronal tissues [16]. Previously, the rate of L1 induced somatic insertions was estimated based on experiments using PCR amplification [17,18]. These numbers were much higher than what was demonstrated in recent single neuron WGS studies using MDA [14,15]. These data suggest the importance of validation experiments, in particular when studying those highly repetitive L1 elements.

4. DNA repair mechanisms and associated diseases

A variety of exogenous chemicals and physical agents, together with endogenous processes such as DNA replication, transcription, chromosome segregation, as well as products of cellular metabolism, like reactive oxygen and nitrogen species (ROS and NOS), can cause DNA damage. To protect the human genome, a complex network of repair processes known as DNA damage response (DDR) pathways have evolved and each is directed to a specific type of damage [19,20] (see Table 1 for a description of the major DNA repair mechanisms). The importance of DDR mechanisms is demonstrated by the growing number of human syndromes that are characterized by defects in DNA repair processes. These result mostly from single and double strand breaks (SSBs and DSBs) and nucleotide excision repair (NER) deficiency (reviewed in [21]). DNA damage can impact the nervous system at all stage of development. During neurogenesis, a major source of mutations is DNA strand breaks due to replication stress. Proliferating neural progenitors rely on both DNA DSBs repair mechanisms: homologous recombination (HR) and non-homologous end joining (NHEJ) repair. Ten to fifty DNA DSBs have been estimated to occur per day in a dividing mammalian cell [22], and often occur at recurrent sites [23,24]. In mature neurons that have exited the cell cycle, NHEJ mechanisms become crucial for DSBs repair. DSBs formation and repair have been also implicated in the gene expression and function of post-mitotic neurons [25,26]. The majority of them occurs in long, late transcribed genes involved in synaptogenesis and neuronal function [24]. In the adult brain oxidative stress is the major cause of DNA SSBs, which is repaired by NER or SSBs. Defective DNA SSB and DSB repair due to germline mutations lead to severe immunodeficiency and neurological disorders [27–34], and could potentially be compounded by the increased burden of SM [35]. The extent to which DNA repair defects lead to increase somatic mutagenesis has been largely unexplored. However, recently defects in NER in Cockayne syndrome and Xeroderma Pigmentosum with neurodegeneration and microcephaly have been linked to increased burden of somatic SNVs in cortical neurons [9].

Table 1.

Overview of the major DNA repair mechanisms and their DNA lesions target

DNA Repair mechanism		DNA damage	Genotoxic source
Direct repair		Methylated (O6 or N7) Guanine	Alkylating agents
Base excision repair (BER)		Non-bulky lesions (Oxidized/Deaminated bases)	ROS Alkylating agents X-rays
Mismatch repair (MMR)		Mismatched bases A-G, T-C mismatch Small insertion or deletion	Replication error
Nucleotide Excision Repair (NER)		Helix-distorting lesions due to bulky DNA lesions (thymidine dimers, pyrimidine dimers), DNA-protein adducts	UV light Polycyclic aromatic Hydrocarbons, ROS and chemicals
Global Genomic Repair NER	Transcription Coupled Repair NER
Single Strand Break Repair (SSBR)		Single strand DNA breaks	X-rays ROS
Double Strand Break Repair (DSBR)		Double strand DNA breaks, DNA interstrand crosslinks	Ionizing radiations, UV light, X-rays, ROS Hydroxyurea (Hu) Chemotherapy Anti-tumor agents
Homologous recombinational repair (HR)	Non-homologous end-joining (NHEJ)

5. Brain diseases where somatic mutations have been shown to play a role

a. Cortical malformations and epilepsies

Hemimegalencephaly (HME) and Focal cortical dysplasia Type II (FCD2) are two malformations of cortical development (MCD) where the involvement of mosaic post-zygotic mutations has been demonstrated. While HME presents with an overgrowth of up to an entire hemisphere, FCD2 presents more localized lesions. Both diseases are often associated with medically refractory epilepsy and require surgical resection, allowing direct histological and genomic analysis of affected neural tissues. Lesions present cortical dyslamination with dysmorphic neurons and sometimes balloon cells. Somatic mutations in genes of the mTOR (mechanistic target of rapamycin) pathway, which is involved in cellular growth, proliferation, and metabolism, have been found in multiple cohorts of MCD and epilepsy related disorders. Hyperactivation of mTOR has also been simultaneously observed in brain specimen [36–38]. Recently, targeted capture deep sequencings on panels of mTOR-related genes coupled to tissue or cell microdissection have demonstrated that the abnormal cells are indeed enriched in causal mosaic variants [39,40]. Earlier mutations during development lead to more affected cells and more severe malformations, with clear correlation between the abundance of the causal variants in the tissue, the density of abnormal cells, the size of the lesion, and the severity of the disease [37,39–41]. A similar correlation between phenotype severity and deleterious variant abundance was described in migration disorders [42]. Interestingly, MCDs can arise from convergent alterations in multiple genes linked to mTOR, and also in unrelated genes that could modulate the impact, such as a somatic double-hit in mTOR and RPS6 which affects cell proliferation [43,44].

In nonlesional focal epilepsy, the absence of a detectable lesion by brain imaging renders the identification of causal mutations and pathological mechanism more complex, as in the X-linked galactose transport gene SLC35A2 in cases of FCDI and mild MCD [45–47], and in voltage-gated sodium channels SCN1A/2A in Dravet syndrome [48]. PCDH19-related epilepsy is an interesting case where the mosaic characteristic of the mutation is pathological and not the mutation by itself. Mutation in the X-linked PCDH19 protocadherin gene triggers epilepsy in heterozygous females and mosaic males but not in hemizygous males despite the fact that the latter expresses only the causative variants [49,50]. Mosaic alterations of PCDH19 in neural progenitors – either due to somatic mutations or random X-inactivation – could cause heterogeneity in cell-cell adhesion affinities, resulting in aberrant tissue architecture and epileptogenicity [51].

b. Psychiatric diseases

Autism spectrum disorder (ASD) is a highly inheritable neurodevelopmental disorder with multiple genetic and environmental risks factors and no convergent cellular etiology identified to date. Establishing a genotype-phenotype correlation for germline variants is making progress [52]. The contribution of somatic events to disease etiology has been recently investigated for both idiopathic ASD and many neurodevelopmental diseases with comorbid ASD (e.g. TSC, NF1, Fragile X, and Rett syndrome). Targeted deep sequencing in postmortem brain of patients with ASD revealed somatic mutations in risks genes (e.g. SETD2, SCN1A) [4,53]. Recently, focused re-analysis of WES from large cohorts of ASD families from the Simons Simplex Collection (SSC) found that a significant proportion of mutations identified as de novo are actually mosaic, although the actual proportion varied between the studies (i.e. 5.4% [54], 7.5% [53] and 22% [55] due to different methods for calling, filtering and validating the variants (reviewed in [56]). Overall, somatic mutation could contribute to 3 to 5% of ASD simplex risk. A recent in vitro study of neural progenitor cells from macrocephalic ASD revealed hyperproliferation and a sensibility to replication stress leading to increased DSB, correlating it with aberrant adherent junction and migration [57]. Altogether, it will be important to establish if ASD can be in part triggered by genetic sensitivity to genomic alteration and subsequent somatic mutations in causative genes.

Like ASD, syndromic intellectual disability has been linked to somatic mosaicism, for instance in the NIPBL gene in Cornelia de Lange Syndrome [58–60]. More risks genes and causal variants of severe intellectual disability are progressively being identified [61,62]. Re-analysis focused on mosaic events revealed that a significant proportion of de novo calls (6.5% of cases) are actually post-zygotic events [63].

Data are considerably sparser regarding the burden of SVs, CNVs, and MEIs in neuropsychiatric diseases given their general lower frequency. SVs have been proposed to be more prevalent in developmental disorders [64], and low-level mosaic aneuploidy has been observed in postmortem brains of schizophrenia and intellectual disability with ASD [65]. With regards to MEIs, several studies reported potential changes in number or pattern of somatic MEIs in Rett syndrome [66], schizophrenia [67], and several other neurodevelopmental disorders [68], but these data need replication.

c. Neurodegeneration

The burden of SM variants was found to be higher in early onset neurodegenerative disorders compared to normal aging brains [5,9], which perhaps could be attributed to disruption of endogenous DNA damage repair mechanisms (Table 1). Recently, it has also been reported, that a subset of Alzheimer’s disease brains showed SM in genes related to AMPK, MAPK and PI3K-AKT pathways, which are known to be involved in hyperphosphorylation of the microtubule-associated Tau protein [5] as well as in other autosomal dominant genes implicated in the disorder [69,70]. Several studies also propose a role of somatic copy number gains in alpha-synuclein and amyloid precursor protein (APP) in the etiology of Parkinson’s [71,72] and Alzheimer’s [73] diseases, respectively, although these data need confirmation.

Other neurodegenerative disorders, including Huntington’s chorea, fragile X syndrome, amyotrophic lateral sclerosis (ALS), and frontotemporal dementia (FTD) have been associated with polymorphic tandem repeats, which are reoccurring pattern of nucleotides (single to hexanucleotides) and are also prevalent during normal development (reviewed in [74]). However, to what extent somatic tandem repeats contribute to those disorders remains unclear, as a correct interpretation of data is still challenging. Numerous studies have discussed a potential link between aneuploidy and brain disorders [75–78]. These results remain controversial due to the advancement of single cell analysis, where aneuploidy was predicted at a much lower rate (< 5%) [79,80], compared to previous estimates of 10% by using DNA probes [81,82]. Indeed, a direct link between Alzheimer’s and aneuploidy in the frontal cortex has not been yet confirmed [80].

6. Conclusion and perspectives

A likely causal relationship exists between somatic SNVs and idiopathic epilepsies, as well as between somatic mutations in DNA repair genes and diseases of premature aging. Similarly, an increased burden of somatic SNVs has been observed in both blood and brain of idiopathic ASD. While SM could play a widespread role in developmental and/or neurogenerative brain disorders, this has not yet been clearly established, partially due to their rare occurrence and difficulty in detection and mapping. Technological and informatic advances in the near future promise to shed light into the contribution of SM to disease.