Complex human chromosomal and genomic rearrangements

Abstract

Copy number variation (CNV) is a major source of genetic variation among humans. In addition to existing as benign polymorphisms, CNVs can also convey clinical phenotypes, including genomic disorders, sporadic diseases and complex human traits. CNV results from genomic rearrangements that can represent simple deletion or duplication of a genomic segment, or be more complex. Complex chromosomal rearrangements (CCRs) have been known for some time but their mechanisms have remained elusive. Recent technology advances and high-resolution human genome analyses have revealed that complex genomic rearrangements can account for a large fraction of non-recurrent rearrangements at a given locus. Various mechanisms, most of which are DNA-replication-based, for example fork stalling and template switching (FoSTeS) and microhomology-mediated break-induced replication (MMBIR), have been proposed for generating such complex genomic rearrangements and are probably responsible for CCR.

The nature of copy number variation and the link with human disease

In human and medical genetics we are often faced with a molecular observation or phenomenon that we seek to explain. Such phenomena often consist of a genetic alteration in the context of either a Mendelian or sporadic trait; usually a disease or birth defect. One major limitation of human genetics is the inability to perform the ‘test cross’ that is so experimentally useful in other genetic model organisms. It is a fact of life that one is often unable to execute either preferred or even desired matings in humans. Nevertheless, positive aspects of human genetics include a wealth of subjects with a myriad of biological variation, tremendous reagents and resources, ready ascertainment by virtue of the conveyed phenotype eliciting a visit to a physician and, perhaps most importantly, the study of phenomena directly relevant to the human organism. Whereas in other model organisms we might be able to experimentally manipulate them to surmise mechanism, the relevance to humans is by inference or analogy alone, not by direct observations. Here, we review the recent experimental observations revealing complex rearrangements of the human genome in association with different disease traits. We present the evidence for their existence and potential explanations (i.e. mechanisms) for their cause.

Whereas Watson–Crick base-pair changes have long been appreciated as a cause of mutation, and millions of single nucleotide polymorphisms (SNPs) are known to exist in human individuals and populations [1,2], it was not until recently that copy number variation (CNV; see Glossary) has been found to represent a major source for human genetic variation and genome diversity. In the past few years, a large number of CNVs have been identified by various genome-wide technologies, including array comparative genomic hybridization (aCGH), SNP genotyping arrays, and next-generation sequencing using ‘paired-end’ methods [3–13]. According to the latest statistics (March 2009), >38 000 entries of structural variations (SVs) have been included in the Database of Genomic Variants (DGV; http://projects.tcag.ca/variation/).

Many of the CNVs in DGV might represent benign polymorphisms; however, many CNVs can convey clinical phenotypes by various molecular mechanisms [14,15]. For example, deletion CNVs that disrupt functional genes can cause loss-of-function mutations and lead to human diseases such as color blindness [16], whereas CNVs that alter the copy number of dosage-sensitive genes can also result in disorders, for example, Charcot-Marie-Tooth disease type 1A (CMT1A; Mendelian Inheritance in Man [MIM] 118220; http://www.ncbi.nlm.nih.gov/omim) and hereditary neuropathy with liability to pressure palsies (HNPP; MIM 162500) [17]. Both disorders are peripheral neuropathies caused by CNV of the peripheral myelin protein 22 gene (PMP22). To date, many disease-associated CNVs have been reported [14,18]. The clinical phenotypes resulting from genomic structural changes owing to genomic instability and arising because of susceptible genome architecture have been referred to as genomic disorders [19,20]. In addition to being associated with Mendelian traits and sporadic diseases, CNVs have also been recently shown to cause human complex traits, such as susceptibility to HIV infection [21], autism [22,23], schizophrenia [24–26] and mental retardation [27–29]; for a review, see Ref. [30]. Thus, uncovering the mechanisms underlying CNV formation has tremendous implications for the diagnostics of CNV-associated diseases. Molecular mechanisms for simple rearrangements (e.g. simple deletion or duplication of a locus), either recurrent or non-recurrent, have been well studied (for recent reviews, see Refs [30,31]).

However, the complexity of selected human genomic rearrangements is becoming increasingly appreciated. In this article, we review the evidence for such complexity at the chromosomal and genomic level and discuss the potential mechanisms to explain these phenomena.

Complexity is not uncommon for human genome rearrangements

Complex chromosomal rearrangements

Complex chromosomal rearrangements (CCRs) describe genome structural rearrangements that involve at least three cytogenetically visible breakpoints and the exchange of genetic material between two or more chromosomes [32] (Table 1). These are rare, although clinically important to recognize because carriers can have phenotypes that include normal individuals, recurrent miscarriages in females, infertile males, mental retardation and congenital abnormalities [33–38]. The alterations can be de novo or familial; familial CCRs tend to involve fewer chromosomes and fewer breakpoints than de novo CCRs [39]. A survey of 269 371 prenatal studies on amniotic fluid, chorionic villus and fetal blood samples reported a total of 246 apparently cytogenetically balanced anomalies; among them, 3% were de novo presumably balanced CCRs [38]. This study also showed a sex ratio bias for the occurrence of CCRs, females are carriers ~2 times more frequently than males, confirming previous data on familial cases [39]. Interestingly, the origin of de novo CCR and reciprocal translocation cases are frequently reported as paternal [39–41], although 77% of the prenatal cases with balanced rearrangements analysed by Giardino et al. [38] were referred owing to advanced maternal age.

Table 1.

Frequency of breakpoints detected for CCRs in 226 patients

Number of breakpoints	Frequency for CCRs
3	70
4	67
5	36
6	22
7	17
8	10
9	8
10	2
11	0
12	2
13	0
14	1
15	1

The occurrence of de novo CCR is associated with an increased risk of mental retardation and the degree of severity correlates with a higher number of breakpoints [33,42]. The development of the clinical phenotype is believed to be caused by the alteration of the regulation of genes located nearby (i.e. a position effect [43]), by disruption of a gene at or near the breakpoint or by imbalances at the breakpoints or elsewhere in the genome [41]. Despite the importance of refining the multiple rearrangement breakpoints at the sequence level in CCR cases, virtually no breakpoints have been sequenced and no molecular mechanisms have been proposed for how they might occur. To date, most of the breakpoints have been mapped using conventional cytogenetic G-banded karyo-typing, multi-subtelomeric fluorescence in situ hybridization (FISH), whole chromosome painting FISH, M-FISH or SKY or multicolor banding (MCB) [34,35,44–46].

More recent studies have used CGH to uncover cryptic rearrangements [40,47–49], but the resolution of the breakpoints is to the level of hundreds or thousands of base pairs. Deletions at the breakpoint regions are a common finding, but duplications or inversions are also detected [35,36,40,50]. Importantly, when the resolution of the analysis methods used to analyse CCR increases, the initially identified number of breakpoints tends to increase [40,48,50–54]. This observation suggests that many, and possibly the majority of, CCRs detected to date might actually be more complex than initially thought. In fact, De Gregori et al. [40] reported that 40% of patients reported as ‘balanced translocations’ were unbalanced and, remarkably, 18% of the reciprocal translocations were, instead, complex rearrangements. Interestingly, some patients present alterations not contiguous to the translocation breakpoint [52]; sometimes they are even located in different chromosomes [41,50,53].

The non-random occurrences of CCR breakpoints have been shown in different studies; however, other studies did not find an association with specific chromosomes or cytogenetic bands or genomic regions [40]. Giardino et al. [38] reported the preferential involvement of chromosomes 22, 7, 21, 3 and 9 in balanced chromosomal anomalies of all types (including reciprocal translocations, inversions and CCRs) in prenatal cases. Non-random distribution of the breakpoints was also noted in the same study, with most breakpoints co-localized with fragile sites. Vermueulen et al. [36] reviewed the literature and showed that 30% of the CCRs had breakpoints on chromosome 7, which was shown to be a segmental-duplication-enriched chromosome (7% according to Ref. [55]). Vissers et al. [47] reported two patients in which 17 total breaks were found scattered throughout chromosomes 17 and 21; four of them mapped within low copy repeats (LCRs) usually associated with different recurrent chromosome aberrations, including those causing CMT1A, HNPP, Smith–Magenis syndrome (SMS; MIM 182290) and Potocki–Lupski syndrome (PTLS; MIM 610883). Batanian and Eswara [34] reviewed 100 CCR cases and noted that band 18q21 has an assigned breakpoint in 11/14 cases where the chromosome 18 was involved. Giardino et al. [51] detected the preferential involvement of chromosomes 3, 4, 6 and 10 in 20 prenatally ascertained CCRs; particularly, they reported a patient in which the breakpoints observed on chromosomes 4 and 10 were coincident with bands containing members of the olfactory receptor family. According to them, such preferential involvement is not observed in simple translocations. In fact, after reviewing 226 CCRs reported in the literature to date, it is possible to observe a clear chromosome preference in CCR events (Figure 1). Chromosomes 2, 3, 4, 7 and 11 are the most frequently involved with similar frequencies of approximately 10–12% and chromosomes 17, 22, X and Y are rarely involved, with a frequency of <1% of the cases. The reason for this preference is not obvious.

Figure 1.

Frequency of specific chromosomes involved in CCR. The x-axis lists specific human chromosomes and the y-axis shows the percentage of times a particular chromosome is involved in a complex rearrangement. Data complied on the basis of 226 CCRs.

The mechanism(s) underlying CCR formation remains elusive; some studies propose models based upon the principle of parsimony and the minimum amount of breaks required for the formation of the CCR [35,47]. Li et al. [56] propose a translocation–deletion–inversion process involving multiple breakage–fusion events to explain a patient carrying nine breakpoints. A similar model was used by Lindstrand et al. [57] to explain a patient with 14 breakpoints along chromosome 1. Patsalis [58] also cites the ‘major’ catastrophe within the gamete (spermatogenesis) hypothesis and remarks that some reports found an association of parental exposure to potential mutagens and de novo CCRs. Thus, the phenomena of CCRs with breakpoints that involve many different regions of the human genome can occur. However, because the breakpoint sequences of CCRs have not yet been experimentally determined, the potential association between genomic architecture and the formation of the CCRs, along with the ability to infer or surmise underlying mechanism for formation from rearrangement products, is not possible.

Complex genomic rearrangements

Accumulating data have revealed frequent complex genomic rearrangements of smaller scale at many different loci that are not resolved by conventional G-banded karyotyping and FISH. Actually, complex genomic rearrangements can be revealed at different levels of genome resolution (i.e. resolved at levels of BAC aCGH, oligonucleotide aCGH, or DNA sequencing; Figure 2) and often multiple molecular methods must be used to visualize the complexity.

Figure 2.

Resolutions of genomic complexities. Resolving complex genomic rearrangements can require different genome analysis tools. (a) Complex rearrangements of the short arm of human chromosome 17 revealed by FISH analysis (adapted, with permission, from Ref. [47]). (b) Complex rearrangement at 1q revealed by BAC aCGH. A BAC array is shown on the left with the chromosome numbers depicted above. The y-axis shows log ratios of patient versus control BAC aCGH with green dots showing gain and red dots showing loss. To the right is a depiction of the complex rearrangement. (c) The additional small duplication of a potential complex rearrangement (duplication–normal–duplication) at 17p11.2 was confirmed by oligonucleotide aCGH. Copy number gains were shown in red. See Ref. [67] for an alternative explanation of these data. (d) FoSTeS-mediated complex PLP1 duplication was revealed by breakpoint sequencing in the apparent ‘simple’ rearrangement cases without detectable complexity by oligonucleotide aCGH (adapted, with permission, from Ref. [59]). Abbreviations: del, deletion; DistRef, distal reference sequence; dup, duplication; nml, normal; ProxRef, proximal reference sequence; 1264F, forward sequence of Subject 1264.

Employing both oligonucleotide aCGH and breakpoint sequence analyses enabled higher genome resolution for mechanistic studies. Lee et al. [59] studied the large non-recurrent genomic duplications of the proteolipid protein 1 (PLP1) locus mapped at Xq22 associated with Pelizaeus– Merzbacher disease (PMD; MIM 312080). Interestingly, the PMD patients with three or more copies of PLP1, as detected by the locus-specific multiplex ligation-dependent probe amplification (MLPA) assay and suggestive of either triplication or more complex rearrangements, were reported to have a more severe phenotype [60]. The approach by Lee et al. [59] combined methods of oligonucleotide aCGH (average resolution of 0.5 kb) and breakpoint sequencing analysis (resolution of 1 bp) and revealed an unanticipated high frequency of complexity, that is, ~65% of non-recurrent PLP1 duplications were shown to have interspersed stretches of DNA of normal copy number or triplication and, thus, were shown not to be simple tandem duplications [59] (Table 3). The sequencing data from successfully amplified breakpoint intervals also identified 2–5 bp microhomologies at breakpoint junctions or join points, suggesting a microhomology-mediated rearrangement process [59]. Interestingly, the sequencing of the breakpoint junctions in an aCGH-based ‘simple’ duplication revealed more sequence complexity with microhomologies of GC, AGT and TT at each breakpoint [59] (Figure 2d). Such complex rearrangements are neither rare instances nor limited to certain loci, but can be frequent and occur throughout the human genome (Tables 2,3).

Table 3.

Complex genomic rearrangements identified by aCGH and/or breakpoint sequencing

Disease	Gene	% complex in non-recurrent rearrangementsa	Sequenced breakpoints	Microhomology	Size (bp) of microhomology	Refs
PMD [MIM 312080]	PLP1	65%	6	100%	2–5	[59,109]
dup MECP2	MECP2	27%	2	100%	2	[62]
del(9)(q34.3)	EHMT1	7%	1	100%	3	[110]
dup LIS1	PAFAH1B1	29%	4	100%	2–21	[66]
PTLS [MIM 610883]	RAI1	57%	7	86%	2–31	[67]
CMT1A [MIM 118220] or HNPP [MIM 162500]	PMP22	29%	3	100%	3–5	[67]

^aPercentages represent the lower estimates because in the case of each locus specific breakpoint sequencing was not achieved for several of the rearrangements.

Table 2.

Examples of complex genomic rearrangements

Chromosome or gene involved	Eventsa	Sequenced breakpoint	Microhomology	Size (bp) of microhomology	SINE or LINE	Refs
1p36.12/PINK1, DDOST	del + ins	2	100%	3	–	[102]
2p21/MSH2	dup + inv + del	3	100%	2	–	[103]
3p22.2/MLH1	del + ins + nml + ins + del	2	50%	1	–	[104]
8q13.3/EYA1	del + ins	2	0%	–	–	[105]
	dup + inv + del	2	100%	1–8
15q21/CYP19	del + inv + del	2	50%	1–3	–	[106]
	del + inv	2	100%
	del + inv + del + ins	2	50%
	ins + inv	2	–
16p13.3/α-globin cluster	del + inv	2	100%	4–8	–	[107]
16q24.3/SPG7	del + ins + del	2	50%	10	1 Alu–Alu	[108]
Xp21.2/DMD	inv + dup + del	2	50%	2	–	[65]
Xq28/EMD, FLN1	del + dup + inv	1	100%	2	–	[63]
Xq28/F8, MPP1	inv + del	1	0%	–	–	[64]
Xq28/MECP2, L1CAM, CLIC2	dup + ins	3	100%	2–39	3 Alu–Alu	[61]
	dup + ins + inv	3

^aAbbreviations: del, deletion; dup, duplication; ins, insertion; inv, inversion; nml, normal.

Bauters et al. [61] reported two complex duplications out of 16 non-recurrent rearrangements at the methyl CpG-binding protein 2 (MECP2) locus on Xq28 (Table 3). Carvalho et al. [62] also identified structural complexity in eight out of 30 genomic duplications of the MECP2 locus (Table 3), of which two show a pattern of two distinct duplication intervals (duplication–normal–duplication) and six are triplications embedded within duplications (duplication–triplication–duplication or triplication–duplication). In addition, complex genomic rearrangements have also been found at other X-linked loci [63–65] (Table 2).

Similar findings also come from the observations of disease-associated de novo genomic rearrangements on autosomes (Tables 2,3). For example, complex duplications were identified that involved the platelet-activating factor acetylhydrolase, isoform Ib, subunit 1 (PAFAH1B1) gene encoding the LIS1 protein in chromosome 17p13.3, deletions and point mutations of which are associated with Miller–Dieker syndrome (MDS; MIM 247200) and isolated lissencephaly. Bi et al. [66] identified three complex rearrangements out of seven cases with copy number changes of PAFAH1B1, two of which were de novo complex genomic rearrangements, one duplication–normal–duplication and one duplication–triplication–duplication. Interestingly, an additional PAFAH1B1 duplication was an insertional translocation [66]. However, its breakpoint sequence was not obtained and further complexity could not be evaluated. Complex genomic rearrangements have been identified in other 17p loci associated with PTLS (57% of non-recurrent duplications) and non-recurrent rearrangements of both CMT1A and HNPP [67]. Notably, a complex exonic rearrangement of PMP22 has also been identified [67]. This latter complex rearrangement recurred in affected siblings of unaffected parents; the mother of whom was mosaic in her blood for the complex rearrangement, documenting a mitotic event as predicted for a DNA replication mechanism [67].

Examples of complex genomic rearrangements with sequenced breakpoints are listed in Tables 2 and 3. The majority of these breakpoints reveal microhomology of 1–10 bp, whereas some have longer microhomologies (21–39 bp) shared between Alu elements. These observations suggest that these complexities are caused by serial microhomology-mediated processes. Similar microhomology-mediated complexities have also been identified in polymorphic CNVs among humans. Perry and colleagues sequenced breakpoint junctions of 23 CNVs uncovered in a study using HapMap individuals; five out of 23 CNVs (22%) showed sequence complexities [12]. Interestingly, microhomologies of 2–6 bp were identified at all the breakpoints of these five complex CNVs. Furthermore, sequence complexities have also been shown at many human–gibbon (Nomascus leucogenys) breakpoints of synteny [68].

Replication-based mechanisms for complex rearrangements

FoSTeS model

The observed human complex chromosomal and genomic rearrangements cannot be generated by a single recombination event (e.g. non-allelic homologous recombination [NAHR] or non-homologous end-joining [NHEJ]). Thus, a mechanism to more parsimoniously explain the multiple breakpoints involved in the complex rearrangements, the often large linear distance between them and the microhomology observed at the breakpoints must be invoked for their formation. To parsimoniously explain the sequence complexities (e.g. duplication–triplication–duplication) and microhomologies at breakpoints, a novel replication-based FoSTeS model was proposed [59].

During DNA replication, it is proposed that the DNA replication fork can stall, the lagging strand disengages from the original template and invades another replication fork and restarts DNA synthesis on the new fork by priming it via the microhomology between the switched template site and the original fork [59]. The new template strand is not necessarily adjacent to the original replication fork in primary sequence, but probably in 3D physical proximity. It could be on the same chromosome, the homologous chromosome or a non-homologous different chromosome; the proximity in 3D space potentially reflecting a replication factory [69]. Upon annealing, the transferred strand primes its own template-driven extension at the transferred fork. Depending on the direction of the fork progression and if the lagging or leading strand in the new fork was invaded and copied, the erroneously incorporated fragment from the new replication fork could be in direct or inverted orientation with respect to its original position. Furthermore, depending on if the new fork is located downstream or upstream of the original fork, the template switching results in either a deletion or a duplication. This procedure of disengaging, invading and synthesizing can occur multiple times in series (FoSTeS×2, FoSTeS×3, etc.), potentially reflecting the poor processivity of the DNA polymerase used and resulting in the observed complex rearrangements.

According to the FoSTeS model [59], two ‘hallmarks’ can be identified for FoSTeS-mediated rearrangements: (i) sequence complexity that results from serial template-driven juxtaposition of sequences from different genomic locations; and (ii) DNA sequence microhomologies at the join points. From the phenomenology observed in human disease-associated complex rearrangements, one could not determine if a single-stranded or double-stranded DNA was the intermediate involved in the invasion during template switching. Nor could it be determined if a stalled fork or broken (collapsed) fork was the initiating event.

A recent experimental study of stress-induced CNVs in normal human cells also revealed structural complexities to the DNA rearrangements and microhomologies at most sequenced breakpoints in the CNVs induced by treatment with aphidicolin, which can directly inhibit DNA polymerase and cause replication fork stalling [70]. This suggests the potential involvement of FoSTeS in CNV formation in mitotic human cells.

Based on the aforementioned experimental evidence from human subjects, the FoSTeS mechanism not only underlies complex genomic rearrangements involving large segments of megabases in size, but also can result in small rearrangements involving only one gene or even just a single exon. This implicates FoSTeS in gene duplication and potentially in exon shuffling, both important to human gene and genome evolution [67,71,72].

Related replication-based rearrangement mechanisms

Complex rearrangements are not limited to human subjects. Actually, the FoSTeS mechanism was proposed based on the observations of the stress-induced gene amplification by template-switching that occurs during DNA replication in E. coli [73].

Break-induced replication (BIR), a double-strand break (DSB) repair model of replication restarting at broken forks that has been developed from experimental observations in yeast [74–76], might also be able to cause human complex rearrangement associated with hemophilia A [77]. The classic BIR model is a homologous recombination-based mechanism, which is dependent upon recombination proteins and requires long homology for template invasion [74–76] (Table 4). The microhomologies identified at the breakpoints of human complex rearrangements are not of sufficient length and cannot be employed in classic BIR. Recently, an investigation into the formation of segmental duplications in yeast reveals a genetic requirement for Pol32, a replicative polymerase subunit, whereas the recombination proteins tested therein (Rad52, Dnl4, Rad1) are not required [78]. A template-switching model of microhomology or microsatellite-induced replication (MMIR) was proposed for DSB repair in this yeast study [78] (Table 4), supporting the contention that a replication-based mechanism can be involved in CNV formation.

Table 4.

Several mechanisms potentially involved in complex rearrangement formation

Mechanism	Cause	DSB	Process	Breakpoint feature	Refs
FoSTeS	Stalled fork	Not required	Template switching between forks	Microhomology	[59]
MMBIR	Collapsed or broken fork	One-ended break	Template switching between forks	Microhomology	[79]
BIR	Broken fork	One-ended break	Homologous recombination-based template invasion between forks	Homology	[74–76]
MMIR	Broken fork	One-ended break	Rad52-independent template switching between forks	Microhomology or microsatellite	[78]
SRS	Slipped strand mispairing	Not required	Serial replication slippage within forks	Microhomology	[80,83]
Multiple NHEJ	DSB	>1 break	Multistep recombination repair of DSB ends	Homology not required	[84,85]

Based upon the evidence regarding genomic rearrangements in human, bacteria and yeast, a generalized, replicative, template-switch model has been proposed to explain structural variations from all domains of life; the microhomology-mediated break-induced replication (MMBIR) model [79]. Both FoSTeS and MMBIR are consistent with the features of complexity and microhomology at the junctions that are identified in most human complex rearrangements and alternative to some extent; although MMBIR is break-induced (i.e. generated by a collapsed fork), FoSTeS is initiated by replication fork stalling (i.e. no DSB is required). However, there is currently no experimental data that determines whether the complex rearrangements causing human diseases are triggered by DSB, a collapsed replication fork, or a single-strand DNA transfer. FoSTeS was proposed based upon the observed phenomena in humans, whereas MMBIR provides mechanistic details from experimental observations in model organisms [59,79]. We propose that FoSTeS and/or MMBIR can be responsible for human complex rearrangements.

Other related replication-based models, for example ‘serial replication slippage (SRS)’ [80] and ‘strand misalignment–realignment’ [81], have also been proposed for complex rearrangements. In the SRS model, the primer strand can serially disassociate with template strand, reanneal by misalignment and restart DNA replication to cause human complex rearrangements [80,82,83]. Because the hypothesized slippages take place within replication forks, SRS might not be able to cause large genomic rearrangements involving hundreds of kb or longer.

The replication-based mechanisms can readily explain complex rearrangements and the microhomologies at breakpoints. However, the breakpoint sequence analyses of complex rearrangements also show absence of microhomology in some cases, which could be interpreted as NHEJ-specific nucleotide insertion or deletion [84]. Such observations suggest that the involvement of NHEJ in one rearrangement step of complex rearrangements or the entire complexity cannot be excluded [85]. However, because the NHEJ-specific nucleotide insertion at breakpoints are short (1–4 nt [86]), longer nucleotide insertions at breakpoints might not be caused by NHEJ, but by template-switching from other genomic loci via replication mechanisms.

Simple rearrangement that can be explained by FoSTeS and/or MMBIR

During the replication-based FoSTeS and/or MMBIR mechanism, template switching can occur in different ways and cause diverse chromosomal structural changes [79]. In contrast to the complex rearrangements, a simple rearrangement can be generated if the template switching occurs only once during DNA replication. Such FoSTeS and/or MMBIR-mediated simple rearrangements share some features (simple rearrangement pattern and microhomology at breakpoint) with those that are generated by NHEJ or microhomology-mediated end-joining (MMEJ) but do not have an ‘information scar’ [59,86,87]. Therefore, these rearrangements can be either explained by NHEJ, MMEJ or a single template-switching event. However, it is presently impossible to distinguish which exact mechanism, recombination-based NHEJ or MMEJ, or a replication-based FoSTeS or MMBIR, actually acts in each of these cases. In deletion cases, the analysis can be even more complicated because of the inability to assay the genomic regions lost during the rearrangement process. For example, there is still another interpretation for the simple rearrangement; that is, multiple FoSTeS or MMBIR events can also lead to a simple deletion CNV if the last template switching occurs to a template ahead of the fork and removes all evidence for sequence complexity.

In cases in which Alu elements were identified at both of the breakpoint ends, the formation mechanism has been interpreted as an Alu–Alu-mediated NAHR [87–91]. However, for NAHR to take place, there must be segments of a minimal length sharing extremely high similarity or identity between the homologous segments, termed minimal efficient processing segments (MEPS) [92]. In fact, the breakpoint sequencing of complex rearrangements reveals that many of the Alu–Alu rearrangements do not provide the MEPS requirements for homologous recombination via NAHR, especially those involving Alu from different subfamilies. Thus, if they were involved as substrates in a homologous recombination reaction, any Holliday structure that formed might be broken down by the mismatch repair apparatus. Bi et al. [66] reported two AluSg elements of 79% identity at the breakpoint junctions of one patient carrying a non-recurrent PAFAH1B1 tandem duplication. Matejas et al. [93] reported a non-recurrent deletion involving PMP22 at 17p11.2 region on which an AluYb and an AluSq with 86% of similarity were found at the breakpoint junctions. The microhomology was restricted to 27 and 29 bp in each case.

Interestingly, Bauters et al. [61] reported sequencing of three MECP2 duplicated breakpoint junctions in three patients, two of which were complex rearrangements and presented Alu–Alu at the juxtaposition regions; one patient had an AluY–AluY with 39-bp microhomology and another AluJo–AluSg with 23-bp microhomologies. In the aggregate, these data indicate that NAHR might not be the mechanism underlying many of the complex rearrangements showing Alu–Alu switching at one of the breakpoints. Instead, the long microhomology (~20–40 bp) shared between Alu elements can potentially facilitate template switching and prime DNA replication to a new fork during DNA replication. Therefore, we propose that template switching between Alu elements can also cause simple rearrangements. It is noteworthy that a prevalence of Alu elements has been observed at the ends of LCRs [88,94] and at breakpoints of the subunits of complex LCRs [95]; therefore, FoSTeS and/or MMBIR involving Alu elements might also be responsible for LCR or SD formation [79].

In summary, in spite of simple CNVs that can be caused by Alu-mediated NAHR and NHEJ or MMEJ, the potential involvement of a replication-based template-switching mechanism in these cases cannot be excluded.

Breakpoint grouping and the influence of genomic architecture on complex rearrangement

FoSTeS and/or MMBIR-mediated complex rearrangements at the loci of PLP1 [59] and MECP2 [62] preferentially occur within or adjacent to complex genomic architecture that undergoes frequent polymorphic variation. Obvious breakpoint grouping was revealed at both loci (Figure 3). Monte Carlo simulation, which was performed to evaluate the breakpoint distribution of the MECP2 rearrangements, strongly supports the contention that the segments of MECP2 rearrangements are non-randomly distributed and that breakpoints are grouping around the shifted average breakpoint 153.0888 Mb, coincident with the LCR-laden region [62] (Figure 3b).

Figure 3.

Breakpoint grouping of the genomic rearrangements at the PLP1 and MECP2 loci. (a) Proposed FoSTeS-associated breakpoints of PLP1 rearrangements are grouped near LCRs (A–E). Adapted, with permission, from Ref. [59]. (b) Proposed FoSTeS-associated breakpoints of MECP2 rearrangements are grouped near LCRs (L, J and K). Top: alignment of the join points of patients carrying complex rearrangements, including those with triplication, to the genomic location based on the data reported in Ref. [62]. Note that the distal breakpoint of the patients carrying triplication is mapped within the same region in all six patients. This region is the LCR K (it can be either K1 or K2 because there is a high frequency of polymorphic inversion involving both within the population) that spans 11.3 kb at Xq28. Bottom: alignments of the join points obtained by DNA sequencing for the patients carrying MECP2 duplications. Adapted, with permission, from Ref. [62].

It is possible that the complex architecture flanking both MECP2 and PLP1 genes can form cruciforms or other non-B DNA structures [96,97], which can potentially cause fork stalling. Moreover, such structures could lead to single-stranded DNA regions that could produce collapsed DNA replication forks and generate one-ended double-stranded DNA, the presumed substrates for MMBIR [79]. Alternatively, in some cases, such structures can approximate participating DNA replication forks that are otherwise separated by sizeable linear distances and provide single-stranded regions for template switching.

Concluding remarks

Various replication-based molecular mechanisms, including FoSTeS and/or MMBIR, can be involved in complex genomic rearrangements, most of which can be consistent with the features of complexity and microhomology at breakpoints. CCR might also occur by replication-based mechanisms given their genomic complexities and preferential paternal origin; that is, there are many additional mitoses in sperm versus egg formation. However, it remains a technical challenge to determine which exact mechanism is responsible for selected disease-associated rearrangements because it is not feasible to distinguish between replication-based and recombination-based or between DSB-induced and fork-stalling-caused processes in some human subjects. Notably, accumulating evidence reveals that many CNVs represent complex rearrangements and, thus, can be more parsimoniously explained by replication-based microhomology-mediated mechanisms, including FoSTeS and/or MMBIR. The complex nature of many LCRs or SDs and the conveyed clinical phenotypes for de novo complex rearrangements implicate replication-based mechanism in human genome and gene evolution and, thus, in human health and disease.

Glossary

Glossary

Copy number variation (CNV)

a DNA segment that is present at a variable copy number in comparison with a reference genome, for example, deletion and duplication [98]

Non-recurrent rearrangement

occurs at a single locus, can vary in size and have different breakpoints. This can be caused by several mechanisms, including NHEJ [84], retrotransposition [7,10,99] and NAHR between repetitive sequences such as L1 [100] and Alu elements [88,90,91]

Recurrent rearrangement

occurs in multiple unrelated individuals, shares a common rearrangement interval and size, and has breakpoints that cluster. This is caused by NAHR between relatively large (>10 kb) DNA repeats, that is, LCRs or segmental duplications (SDs) [31,101]

Structural variation (SV)

includes CNV and other balanced variations such as inversion and translocation. Structural variations are caused by genome rearrangements, which can be categorized into two major groups:: recurrent or non-recurrent rearrangements