Hotspots of Human Mutation.

Mutation of the human genome results in three classes of genomic variation: single nucleotide variants; short insertions or deletions; and large structural variants (SVs). Some mutations occur during normal processes, such as meiotic recombination or B cell development, and others result from DNA replication or aberrant repair of breaks in sequence-specific contexts. Regardless of mechanism, mutations are subject to selection, and some hotspots can manifest in disease. Here, we discuss genomic regions prone to mutation, mechanisms contributing to mutation susceptibility, and the processes leading to their accumulation in normal and somatic genomes. With further, more accurate human genome sequencing, additional mutation hotspots, mechanistic details of their formation, and the relevance of hotspots to evolution and disease are likely to be discovered.

Introduction

In 1983, the HTT gene was mapped to human chromosome 4 [1]. Single allele loss-of-function mutations (see Glossary) at this locus result in autosomal dominant Huntington’s disease. It was later found that a CAG tract within HTT can expand to more than 35 times the normal length, and that the expanded allele encodes a nonfunctional protein [2]. This trinucleotide expansion is seen in ~10% of Huntington’s disease cases and occurs due to slippage of the DNA polymerase during replication of the tandem repeat [3,4]. The HTT CAG tract expansion is one example of the many different mutation hotspots scattered throughout our genome. These hotspots can lead to a ~100-fold difference in germline mutation rates across human DNA, and are typically generated by distinct mutational processes [5]. Understanding these processes is not only important for basic biology and evolutionary dynamics, but can also inform preventative and therapeutic approaches to genetic disease. In this review, we discuss mutation hotspots in the human genome, with an emphasis on mechanistic inferences.

Classes and Mechanisms of Genetic Mutation and Variation

Mutational processes and consequences are commonly observed by examining variation present between two individuals in a population. Genomic variants are largely defined by their size and mechanism of formation (Figure 1). Single nucleotide variants (SNVs) are one base pair substitutions in DNA, which are further classified as transitions (i.e., purine to purine or a pyrimidine to pyrimidine change) or transversions (i.e., a purine to pyrimidine switch or vice versa). They are the most numerous variant in the human genome, comprising ~80% of the differences between two individuals, and occur at a rate of ~1 × 10–9–1× 10–8 mutations per base pair per generation [6,7]. SNV mutational hotspots include germline variants in CpG-rich loci [6,8] and somatic mutations caused by activation-induced cytidine deaminase (AID) or apolipoprotein B mRNA editing enzyme, catalytic polypeptide (APOBEC)-mediated deamination [9,10].

Figure 1.

Types of Genetic Variation.

Single nucleotide variants (SNVs) and indels are changes that affect between one and 50 base pairs in a single event. (A) Example of a C:T SNV, and a two base pair deletion and a three base pair insertion indel. Examples of events over ≥50 base pairs that constitute SVs are shown in (B); these events include deletions, duplications, inversions, insertions, translocations, and complex combinations of these basic variant types. In each example, the top chromosome is the reference, and the variation is highlighted and displayed beneath.

An insertion or deletion (indel) is the gain or loss of one or more nucleotides spanning <50 base pairs. An analysis of 33 families found a de novo indel rate of ~9 × 10–10 mutations per nucleotide per generation [7]. Approximately 45% of indels are concentrated in ~4% of the genome and most can be explained by polymerase slippage [11] (Figure 2A, Key Figure). Sequences prone to polymerase slippage, such as homopolymeric runs and tandem repeat regions, harbor70% of indels [11]. Coding region indels are under stronger negative selection than are SNVs [12]. Based on the dbSNP database, SNVs in the coding regions outnumber short indels by sevenfold, while the density of SNVs and indels is almost the same in intronic regions [12].

Figure 2.

(A) The mechanism(s) of variant formation can be inferred through sequence context as well as the resultant variant. For example, spontaneous deamination of methylated cytosines result in C:T transitions at the GC-rich loci. Apolipoprotein B mRNA editing enzyme, catalytic polypeptide (APOBEC) cytidine deaminases act upon tCw motifs changing them to tTw or tGw (IUPAC code K). This pattern is observed in kataegis. Finally, repetitive DNA regions are subject to indels when the polymerase dissociates from the template and pairs with the incorrect region upon reassociation, resulting in expansions or contractions. (B) Nonallelic homologous recombination can lead to rearrangements via incorrect repair between repeats. This example highlights a 1.5-Mb region in chromosome 17p11-p12. Two segmental duplications (SDs), denoted as SD A and SD B, represent the proximal and distal ~24-Kb CMT1A-REP repeats. These SDs ectopically pair during meiosis, and repair by nonallelic homologous recombination (NAHR), leading to duplication or deletion of the dosage-sensitive gene PMP22, resulting in Charcot-Marie-Tooth disease 1A (CMT1A) or hereditary neuropathy with liability to pressure palsies (HNPP) respectively.

SVs are large-scale changes of the genome (>50 base pairs), including deletions and duplications (copy number variations), inversions, insertions, and translocations. The estimated SV mutation rate is ~0.29 per generation [13]. Although SVs occur less frequently than SNVs and indels, they account for the variation in a larger number of nucleotides when comparing any two human genomes [14]. SVs are hallmarks of many cancers and can involve complex mutations, such as those observed in chromothripsis and whole-genome duplication events early on in cancer lineages [15,16].

Inherited variants of all of these types in the human population have been subject to selective pressures, leading to a change in frequency; this process can lead to variation hotspots (Box 1) that can be distinct from mutation hotspots. Specific mechanisms and environmental influences have been associated with distinct somatic mutational signatures identified in cancer, and are mentioned here in the sections corresponding to their mechanisms of formation [17]. In some instances, more than one mechanism can contribute to a mutational hotspot (Box 2).

DNA Sequences Prone to Mutation

GC-Rich Regions

Two mechanisms promote mutagenesis of GC-rich regions: spontaneous deamination (cancer mutational signature SBS1 [17]) of methylated cytosines and GC-biased gene conversion [8,18]. Cytosines have a germline mutation rate that is approximately ten times higher than other nucleotides and these changes are most prevalent at CpG dinucleotides [19]. CpG islands are a hotspot for mutation, are commonly located within and upstream of genes, and can regulate gene expression via their methylation status [20]. These loci comprise ~1.5% of our genome, and ~10.6% of all CpG dinucleotides are within islands [21]. Despite spontaneous 5-methylcytosine deamination resulting in C to U lesions, SNVs in CpG islands tend to be approximately seven times less frequent than in other CpGs [21]. Genome-wide methylation analyses have shown that the lower mutation rate of CpG islands is driven primarily by strong negative selection due to their role in gene expression regulation [22]. Although methylated CpGs are a mutation hotspot, their regulatory role means that C to U lesions are under negative selection within CpG islands.

GC-biased gene conversion is an example of a mutational mechanism occuring at recombination hotspots and is a major driver of base composition heterogeneity [23]. In this process, meiotic recombination favors GC-rich alleles over AT-rich alleles and facilitates local GC-content increases [24,25]. During the pairing step of gene conversion, there may be G:T or C:A mismatches resolved in favor of the G or C alleles [24]. After analysis of autosomes in >1000 sequenced genomes, AT to GC changes were enriched compared with GC to AT [24]. This simultaneously increases local GC-content as well as SNVs.

Microsatellites

Microsatellites comprise ~3% of the human genome [26]. These repetitive sequences are prone to polymerase slippage and indel formation, and therefore, are hotspots of mutation in both somatic and germline events. Indels in microsatellites can give rise to the inherited trinucleotide repeat expansions and contractions seen in Huntington’s disease, Fragile X syndrome, and X-linked spinal muscular atrophy [4,27,28]. There are 43 human genes that contain microsatellites in their coding sequence and may act as mutation hotspots for other diseases [11]. Some cancers display somatic microsatellite instability that results from functionally impaired mismatch repair enzymes (cancer mutational signatures SBS15, SBS21, SBS44, DBS7, DBS9, ID1, and ID2 [17]). This phenotype is present in ~15–30% of certain tumor types, resulting in an estimated 200–300-fold increase in mutation rate per cell division as measured in mismatch repair-deficient yeast and mice [29,30].

Meiotic Recombination Hotspots and Nonallelic Homologous Recombination

Meiotic recombination occurs at specific GC-rich loci that contain binding motifs for PR domain zinc finger protein 9 (PRDM9) [31,32]. PRDM9 binds its 12-base pair zinc finger motif, leaves euchromatic H3K4me3 and H3K36me3 marks, and recruits SPO11 to cut the DNA at one of two homologous chromosomes [33]. Holliday junctions form between the homologs, with DNA repair machinery facilitating crossover and repair with the intact homolog [31].

PRDM9 motifs function as both recombination and mutation hotspots [32]. As the target sequence is cleaved and recombined, mutations accumulate due to imperfect repair and these mismatches can inhibit future binding of PRDM9 [34]. Alleles lacking PRDM9 motifs are under positive selection, because the lack of recombination allows them to be inherited more frequently [35,36]. The mutational processes at play lead to both rapid evolution of the PRDM9 zinc finger binding domain and meiotic recombination hotspots across the genome.

Nonallelic homologous recombination (NAHR) can result in SV hotspots and occurs by ectopic recombination at or near repetitive sequences in the genome during meiosis [37]. For example, chromosome 17p contains 1.7 kb segmental duplications flanking the dosage sensitive PMP22 gene. Homologous recombination between the incorrect (98.7% identical) segmental duplications during meiosis results in either duplication or deletion of PMP22, leading to Charcot-Marie-Tooth (CMT1A) or Hereditary Neuropathy with Liability to Pressure Palsies (HNPP), respectively (Figure 2B) [38]. The rate of NAHR can be significantly higher than that of SNVs, with CMT1A duplication and HNPP deletion events occurring at a frequency of ~2–4× 10−5 per every male meiosis [39]. In addition to segmental duplications, transposable elements may facilitate NAHR in some regions of the genome, including the VHL and SPAST loci [40].

Centromeric Rearrangements

Centromeric DNA can be a hotspot for rearrangements because it largely comprises satellite repeat DNAs and transposable elements [41]. Heterochromatin marks on centromeres are important for maintaining centromere integrity and preventing rearrangements [42]. Centromeric rearrangements can involve entire chromosome arms, affecting the dosage of many genes, and can act as a precursor to aneuploidy in cancer [43,44]. Translocations occurring at the centromere are common across almost all tumor types due to the high amount of homology between centromeric loci of many human chromosomes [45,46]. Chromosome mis-segregation and errors in DNA replication can lead to chromosome breakage and result in unbalanced events, such as those occurring in squamous cell carcinomas [47]. The large scale of centromeric mutations can result in the many gene dosage alterations that are typically seen in cancer.

Telomeres and Subtelomeric Regions

Telomeres have an important role in protecting chromosome ends from erosion over many cellular divisions. Shortened telomeres trigger cell cycle arrest and apoptosis. Apoptosis prevents the potential joining of chromosomal ends (chromosome fusion) in senescent cells or aberrantly proliferating pretumor cells [48]. Cells that do not initiate apoptosis enter a state of telomere crisis, which is characterized by frequent telomere fusions, large SVs, chromothripsis, kataegis, and tetraploidy [49,50]. Telomere crisis is seen in many cancers, including chronic lymphocytic leukemia, breast cancer, colorectal adenomas, and gliomas.

Subtelomeric regions are gene-rich sites of frequent meiotic recombination [51]. Subtelomeric sequences are highly polymorphic in copy number, and their rearrangement contributes significantly to intellectual disability, autism, and birth defects [51,52]. The olfactory receptor gene family resides within subtelomeric loci, and variants due to recombination and mutation are common in human olfactory receptor genes [53].

Replication Timing

The first analysis of DNA replication timing established three timepoints of S phase: early; middle; and late [54]. Actively transcribed gene-rich domains in the nuclear interior replicate early, whereas heterochromatin domains comprising centromeric and telomeric repeats, which are largely located in the nuclear periphery, replicate later [55]. Increased mutation rates are correlated with these late-replicating regions [56]. For an in-depth review of the mutational landscape of replication timing, see [57].

The rate of point mutations along chromosomes correlates tightly with replication timing [58]. Late-replicating regions contain a two- and sixfold increase in transition and transversion SNV mutations, respectively, compared with early-replicating loci [58]. Factors contributing to increased mutations in late-replicating regions include replication fork stress and induction of DNA damage response pathways [57]. Other mutation types are correlated with differences in replication timing. For example, SVs mediated by NHEJ are twice more enriched in late-replicating regions, whereas SVs driven by NAHR are four times more enriched in early replicating regions (Figure 3A) [59].

Figure 3.

Mutations during DNA Replication.

(A) Regions replicated during early S phase generally comprise euchromatic DNA and are localized towards the center of the nucleus. Regions replicated during late S phase are primarily heterochromatic and localized near the nuclear periphery. Replication timing is correlated with mutation type. Structural variants (SVs) mediated by nonallelic homologous recombination (NAHR) primarily occur during early S phase. Structural variants mediated by nonhomologous end-joining (NHEJ) as well as single-base transitions and transversions occur more frequently during late S phase. (B) Transcriptionally active G nucleotide-rich repeat tracts can result in secondary structure-forming G-quadruplexes and R-loops. The stabilization of transcription forks at these loci poses a threat to genomic integrity in two ways. First, single-stranded DNA at R-loops is exposed to apolipoprotein B mRNA editing enzyme catalytic polypeptide-like (APOBEC) family enzymes, allowing C:U mutations on the free strand (indicated as red asterisks). Second, if R-loops are not resolved before replication, the transcription and replication forks may collide, causing replication fork collapse and potentially triggering SV formation. Abbreviations: DNAP, DNA polymerase; RNAP, RNA polymerase.

In cancer cells, late-replicating regions have a two to three times higher point mutation rate [60] and a loss of DNA methylation [61]. In prostate and breast cancer, late-replicating regions are correlated with cis-chromosomal rearrangements, whereas early replicating regions contain more trans-chromosomal rearrangements and tandem duplications [61,62]. Collectively, these results exhibit a relationship between replication timing, point mutations, and structural variants, which can contribute to differences in mutation rate across the genome.

Common Fragile Sites

Common fragile sites (CFSs) were first described in 1984 as sites that formed visible gaps and breaks on human metaphase chromosomes following inhibition of DNA synthesis [63]. After sequencing of the human genome, CFSs were found to contain long stretches of AT microsatellite sequences that are prone to secondary structure and double-strand breaks (DSBs), making them susceptible to SV formation [64]. CFSs are late replicating, enriched in large genes, have a low concentration of replication origins, and display cell-type specificity [65,66].

CFSs are prone to mutation through a variety of mechanisms. Through recurrent DSBs, CFSs can lead to structural variants in nearby loci and within topologically associated domains [67,68]. Additionally, instability of CFS sequences can lead to the deletion of oncogenes by an NHEJ-independent mechanism [69]. A model for CFS instability and SV hotspots involves transcription-dependent double fork failure [66]. In this scenario, late-replicating genes interfere with actively transcribed regions. The replication fork and transcription machinery collide, create DSBs, and may be improperly repaired resulting in SVs [66]. This phenomenon is more common in genes with large transcription units that are >500 kb in length, such as LSAMP and AUTS2 [66].

DNA Structures Prone to Variation

AT-Rich and Hairpin-Forming Palindromes

Palindromic sequences can lead to unusual DNA conformations and genomic instability [70,71]. Palindromic AT-rich repeats (PATRRs) form opposing hairpins in DNA, creating a four-way Holliday junction that resembles a cruciform. Upon cleavage, the resolution of the cruciform structures can result in translocations, some of which are recurrent [72]. One such PATRR-mediated translocation results in a supernumerary chromosome containing genetic material from chromosomes 11 and 22, and can lead to Emanuel syndrome, which is characterized by developmental delay and hypotonia in infants [73]. Another translocation between PATRRs on chromosomes 17 and 22 disrupts the NF1 gene and manifests as neurofibromatosis type 1, characterized by patches of darkened skin and benign tumors [74].

Hairpin-forming palindromes throughout the genome can also be nucleation sites for APOBEC-mediated mutations [75]. Some recurrent hotspots of APOBEC mutation occur outside of hairpins in oncogenes or tumor suppressors, and appear to be under positive selection for cellular growth advantages; these are important hotspots for understanding tumor evolution [75]. Other documented APOBEC hotspots may be functionally insignificant and simply occur because they are located within hairpin-forming palindromes that serve as optimal APOBEC substrates [75,76]. Therefore, mutation hotspots in cancer due to the action of APOBEC family proteins may not reflect tumor evolutionary processes.

G-Quadruplexes and R Loops

G-quadruplexes are long stretches of G nucleotide-rich sequences that can self-stack and act as common sites of replication stress and fork stalling. G-quadruplexes are enriched in promoter regions and have been shown to act as docking sites for transcription factors [77,78]. Given their involvement in both active transcription and replication stress, G-quadruplexes serve as sites where transcription and replication forks may collide [79,80].

R-loops are structures where the RNA remains hybridized with the transcribed template DNA strand, exposing the bare nontemplate DNA strand; these three-stranded structures are associated with nontemplate strand G-quadruplexes [81]. R-loops provide a link between late-replicating large transcription units and SV formation due to CFSs [66]. Stabilized R-loops located in late-replicating regions can promote convergence of transcription and replication forks, leading to DSBs and SV formation [66] (Figure 3B). Interestingly, due to cell- and tissue-specific transcriptional programs, these SV-prone fragile sites are likely to differ among species as well as tissues.

Chromatin and Mutational Processes

The 3D organization of DNA has an important and emerging role in the repair of DSBs via HR, and may influence NAHR [82]. DSBs are preferentially repaired using a homologous DNA template in close special proximity, with similar replication timing, and within overlapping chromosome territories [83]. One key example is the BCR-ABL fusion protein (Box 1); whereby the BCR and ABL genes are replicated at the same time, and are located near each other in the nucleus [84]. Both yeast and human studies show that proximal genomic regions recombine more efficiently than distal genomic regions, and this can have a role in either facilitating or preventing NAHR [83,85]. Chromosome loop anchors are responsible for bringing two loci into close spatial proximity [86]. Anchor binding sites may be cut by topoisomerases to rearrange DNA topology, and mutations at topoisomerase binding sites may result in ectopic recombination between chromosome loop anchors and chromosomal rearrangements [87]. These studies highlight how DNA proximity can affect homologous recombination and may prevent or facilitate NAHR.

Focal Mutation Hotspots

Chromothripsis

The first documented chromothriptic event showed that a single complex genomic rearrangement can drive cancer progression by simultaneously deleting tumor suppressors and duplicating oncogenes [15]. This event resulted in 42 somatic rearrangements all on chromosome 4q, but a scarcity of rearrangements elsewhere. Chromothripsis encompasses complex clusters of rearrangements, including duplications, deletions, and inversions, that are proposed to occur in a single chromosome-shattering event. Chromothripsis is highly prevalent in some cancers, occurring in >50% of liposarcomas, osteosarcomas, and glioblastomas [88]. The mechanisms that result in chromothripsis are only now becoming clear. Micronuclei formation is a precursor to chromothripsis, and can result from telomere fusion, chromosomal bridges, unrepaired DSBs, and nuclear envelope collapse [89-91]. The breakage and improper repair of chromosomal bridges can also lead to chromothripsis and has been attributed to the cytoplasmic exonuclease TREX1 [91,92]. It is difficult to distinguish which structural variants from a chromothriptic event are oncogenic. As mentioned earlier, micronuclei harboring amplified oncogenes are one potential contributor to positive selection in tumorigenesis (Box 1). To further understand the functional consequences of chromothripsis, a micronuclei-based model system can now be used [93].

Kataegis

In addition to large-scale chromosomal rearrangements, tumors also harbor clusters of point mutations termed mutation showers or kataegis [94]. Some of these SNVs cluster at TpC dinucleotides that are adjacent to somatic rearrangements, and can be hundreds to thousands of times more common than the background mutation rate [95]. Overexpression of APOBEC enzymes is associated with mutation, genomic damage, and cancer progression (cancer signatures SBS13 and possibly DBS11 [17]) [10]. This family of enzymes mutates TpC dinucleotides, suggesting that they have a role in kataegic foci formation [96,97]. Kataegic foci were first observed in cancer genomes as clustered mutations commonly found at sites of DSBs, DNA repair, and structural variation [95,97-99].

Structural variation and kataegis can co-occur focally in human cancer, suggesting that they are mechanistically linked [96,97,100,101]. Break-induced replication (BIR), a process that repairs single-ended DNA breaks by copying homologous templates, can give rise to kataegic foci [98,101]. During this repair process, single-stranded DNA is made vulnerable to APOBEC cytosine deamination, creating many de novo SNV lesions in the replicated strand (Figure 4A) [102]. Another example involves telomere crisis and chromothripsis [90]. Telomere crisis may induce chromothriptic events surrounded by APOBEC3B-mediated kataegic foci [91].

Figure 4.

Replication-Based Repair Mechanisms Are Associated with Clustered Mutations.

(A) Repair pathways that expose single-stranded DNA are subject to apolipoprotein B mRNA editing enzyme, catalytic polypeptide (APOBEC)-mediated hypermutation. Following replication fork collapse, mutations are primarily focused on exposed single-stranded DNA. In this break-induced replication repair model, the leading strand from the damaged chromosome invades the homologous region of its mate and forms a displacement loop (gray dashed-line box). This single-stranded DNA is exposed to lesions from APOBEC-mediated deamination. These lesions are retained as the lagging strand of the broken chromosome is synthesized. (B) Replication-based repair mechanisms use the homologous chromosome as a template for repair. This creates localized regions of conservative replication leading to accumulation of variants in the homologous region. The displacement loop created by strand invasion can proceed for regions up to ~1 MB, and this process introduces errors or mutations that are not efficiently repaired.

Long Tracts of Mutations Surround BIR-Mediated SVs

De novo, nonrecurrent SVs have been shown to harbor clusters of indels and polymerase slippage events close to their breakpoint junctions [103], but mutations can also extend up to 1 Mb away from the breakpoint [104]. Replication-based repair mechanisms could explain these mutation clusters because the displacement loop and synthesis results in conservative inheritance of new DNA (as opposed to traditional semiconservative inheritance), and can lead to a 1000-fold increase in mutations compared with normal replication (Figure 4B) [104,105]. Indels around de novo, nonrecurrent SVs are likely the result of polymerase slippage and occur within poly(A) tracts and tandem duplications [103]. These findings point to multiple mechanisms leading to higher mutation rates accompanying SV formation.

Concluding Remarks

Mutation hotspots are highly dependent on DNA sequence and structure and are subject to selection. Mutations can occur due to cellular processes, such as DNA replication and repair, meiotic recombination, and immunoglobulin specification. Mutations are not always under positive selective pressure, because some are simply passenger mutations resulting from a multimutational event. Here, we reviewed many of the known and recently identified mutational hotspots and the mechanisms of their formation. However, the mechanistic relationships of certain mutational signatures found in cancers currently remain unclear [17]. The investigation of mutation hotspots in cancer has also resulted in novel therapeutic approaches (see Box 1). As technology advances and more large-scale sequencing studies are completed, we will gain further insight into hotspots of genetic variation and their frequency across the human population and across diverse tissues, as well as how these processes contribute to disease (see Outstanding Questions).