Insights into common fragile site instability: DNA replication challenges at DNA repeat sequences Open Access

Common fragile sites (CFSs) are specific genomic regions within the normal chromosomal structure, that exhibit an inherent sensitivity to perturb DNA replication. Under replication stress conditions, CFS regions suffer from a significant delay in replication progression and fail to complete their replication before mitotic entry (reviewed in [1]). These under-replicated regions fail to properly condensate and can be visualized cytogenetically as gaps and constrictions in metaphase chromosomes (‘CFS expression'). Failure to complete replication and repair results in aberrant chromatid segregation and transmission of the damage to the daughter cells [1,2]. One of the biological consequences of under-replicated DNA at CFSs is the formation of copy number variants (CNVs) (reviewed in [1]). CNVs result from genomic deletions or duplications and are a form of structural chromosome alterations related to many human genomic disorders and cancers. The majority of CNVs occur randomly at different genomic locations, however, a subset of events are formed in focal clusters or hotspots which overlap CFS loci spanning late-replicating, large active transcription units [3]. Characterization of CNV patterns caused by aphidicolin-induced replication stress, identified large gains and losses at CFSs in close proximity to late replicating large expressed genes [4]. In line with this, CFSs were shown to be preferentially unstable in pre-cancerous lesions and during cancer development as a result of the associated replication stress present under these conditions [5,6]. Indeed, pan-cancer analyses of the landscape of homozygous deletions (HDs) in cancer genomes found that HDs preferentially occur either in tumor suppressor genes or over fragile sites [7–9].

The sensitivity of CFSs to perturbed replication dynamics cannot be attributed to a single mechanism but rather to a combination of features contributing together to their instability (Figure 1). Initially, these included difficult-to-replicate sequences, among them AT-rich repeats that upon DNA double strand unwinding during replication are able to readily fold into stable secondary structures which impede replication fork progression (reviewed in [10]). Other suggested features include late replication timing [11,12] and insufficient initiation events [13] leading to a progressive delay in the replication completion; co-localization with large genes [12,14,15]; non-accessible chromatin structures reducing DNA replication and/or repair efficiency [16]; and defective condensin loading which can lead to chromatin-folding defects and impaired repair processes [17].

Another widely studied factor affecting CFS instability is transcription along large genes, however, the underlying mechanism remains a subject of discussion. Transcription was suggested to challenge replication completion by inducing origin displacement from gene bodies, resulting in long traveling forks along large genes [3,18,19]. Additional suggested transcription-dependent mechanism, involved in perturbing DNA replication progression in CFSs, is conflicts between the replisome and transcription machineries along large genes transcribed during S phase. Helmrich et al. [12] suggested that RNA:DNA hybrids (R-loops) form at sites of transcription–replication collisions and provoke CFS instability, an outcome that RNase H1 was able to suppress. However, other recent studies could not find a direct evidence for the involvement of R-loops in CFS instability, using altered RNase H1 expression and DRIP-seq analyses [19], transcription inhibition during S phase [18] or DRIP-PCR (our unpublished data). As these studies did not support the role of R-loops in fragility, the contradicting results leave an uncertainty regarding the role of R-loops in aphidicolin-induced fragile sites.

As only one strand is usually being transcribed, transcription induced replication perturbation can lead to uncoupling of DNA synthesis between the leading and lagging strands [20]. Indeed, Koyanagi et al. [20] observed an increased polymerase uncoupling along large-gene bodies, implying impediments for leading strand DNA synthesis due to transcription–replication conflicts. Moreover, in three regions a significant delay in polymerase progression was observed, two of these within the highly fragile FRA16D and FRA3B CFSs. Although FRA16D and FRA3B are frequently fragile in the analyzed cells (HCT116 cells) [21,22], the expression level of the FRA16D-related WWOX gene is very low [22], suggesting the involvement of other factors, as secondary structure formation at AT-rich repeats which present along this regions, in the observed uncoupling. In contrast with the above models, which rely on active transcription for CFS instability, Blin et al. [23] reported that inducing abnormally high levels of FHIT transcription (∼20 fold increase), led to an advanced replication timing and reduced FRA3B chromosomal fragility in human HCT116 cells. However, the molecular basis for the transcription-dependent regulation of replication timing and genome stability requires further studies.

CFSs were also found to be sensitive to limiting concentrations of licensing factors as reflected by an increased breakage at FRA3B, FRA16D and FRA16C CFSs following knockdown of the mini chromosome maintenance 3 (MCM3) licensing protein [24]. In line with this, limited licensing along CFSs was also observed in a genome-wide analysis of chromatin immunoprecipitation and sequencing of MCM7, showing that large genes overlapping CFSs, among them FRA3B and FRA16D, are depleted of replication origins compared with non-CFS long genes [25].

More recently, our group suggested a fragility signature, comprised of a highly transcribed large gene with delayed replication timing spanning a topologically associated domain (TAD) boundary [26]. TADs are functional genomic units essential for coordinated replication, transcription and DNA-damage repair [27–29]. It was shown that DNA damage repair is preferentially confined to intra-TAD regions, as reflected by a tendency of γH2AX spreading to stop at TAD boundaries [29,30]. Our investigation of the potential role of TAD organization in regulating CFS instability found that the vast majority of CFSs span TAD boundaries harboring transcribed large genes with aphidicolin-induced delayed replication timing (Figure 1) [26]. Moreover, the TAD boundaries were at the core of the delayed region within the expressed large genes. These results show that intra-TAD regions tend to be protected from deleterious replication stress, whereas inter-TAD regions in combination with high transcription and replication timing delay are susceptible to breakage.

The contribution of difficult-to-replicate AT-rich repeats to genomic instability at CFSs was addressed by us and others for almost three decades. The tendency of these sequences to fold into stable secondary structures during DNA replication and to impede by this the replication fork progression, predisposes certain genomic regions to breakage. Indeed, genome-wide analyses of structural variation in cancer genomes found a significant enrichment of [AT/TA] repeats at translocation endpoints and [A/T] repeats at deletion endpoints, supporting the notion that these repeats are an intrinsic risk factor for genomic rearrangements in cancer genomes [31,32].

In this review, we will discuss the present knowledge regarding the contribution of repeat sequences to CFSs instability. We will also highlight the cellular mechanisms involved in their maintenance during duplication of the genome.

The role of specific DNA sequence elements in the predisposition of genomic regions to recurrent chromosomal instability is one of the research pathways aimed to characterize the features underlying CFSs instability. One of the difficulties in the establishment of a causal link between specific sequences and recurrent chromosomal instability is that most CFSs are mapped cytogenetically to large genomic regions (hundreds to thousands kilobases) while the core fragility regions have not yet been identified. Thus, the identification of shared sequences specific for CFSs is challenging. The understanding that AT-rich repeats play a key role in chromosomal instability was originally proposed in early studies designed to identify sequence features that might provide insights into the molecular basis of fragility [33,34]. Computational sequence analyses revealed that several CFSs (FRA3B, FRA7H, FRA7G, FRA7E, FRA7I, FRA16D and FRA10B) have high A/T content and are enriched with interrupted runs of AT/TA dinucleotide repeats (flexibility peaks) [33–37]. These AT-rich repeats are predicted to confer high DNA helix flexibility and low structural (helix) stability [33]. Further In silico analyses revealed that AT-rich repeats longer than 200 bp may readily form stable secondary structures [34], which can potentially impair replication fork progression and may cause fork collapse and increased chromosomal breakage [38]. These secondary structures are formed during DNA replication in vitro especially following treatment with aphidicolin, the DNA polymerase α, δ and ε inhibitor. Aphidicolin leads to uncoupling between the polymerase and the helicase [39], resulting in long stretches of single-stranded DNA (ssDNA) that can form stable secondary structures such as hairpins or cruciform (reviewed in [10]). As these unusual structures are significantly more stable than the ssDNA state, AT-rich repeats tend to adopt this type of structure folding, which may perturb replication fork progression.

Several lines of evidence suggested a direct link between repetitive DNA sequences and recurrent chromosomal fragility. Durkin et al. showed that deleting genomic sequences harboring AT-dinucleotide rich flexibility peaks from the FHIT gene, significantly reduced FRA3B expression level under aphidicolin treatment [36]. This supports a central contribution of AT-rich repeats to the genome instability along this CFS and/or reflecting successful replication completion due to the reduced gene length following deletion. AT-rich repeats were shown to be difficult-to-replicate sequences, able to impede the progression of the replicative Pol δ holoenzyme (reviewed in [40]). As such, other specialized polymerases which are more efficient in synthesis repeat sequences, including AT-rich repeats, are required. Indeed, specialized Pols η, κ and ζ have been implicated in maintaining CFS stability during both normal and stress conditions, facilitating complete genome duplication and DNA repair [41,42].

One of the most studied regions in this aspect is the FRA16D CFS (16q23), frequently expressed in human cells, under replication stress conditions. The sequence spanning FRA16D contains six AT-rich repeats, named Flex 1–6 [35]. Flex1, which is the most flexible sequence, is ∼300 bp long and contains a polymorphic perfect AT-repeat that ranges from 11 to 88 copies among humans. It is important to note that mapping of FRA16D HD endpoints across various cancer cell lines revealed an overlap with Flex 1, implicating that this sequence is a preferred target for instability in vivo [35,43]. A clear evidence for the significant contribution of Flex1 to FRA16D instability came from the Freudenreich lab which found that the Flex1 sequence cloned into a Yeast artificial chromosome (YAC) causes chromosomal fragility during replication in S. cerevisiae [38,44]. The fragility level was dependent on the AT-repeat length. In this system, Flex1 deletion led to a decrease in the breakage frequency [44], suggesting its significant contribution to the overall breakage of FRA16D in human cells. Another in vitro system, using the S1 nuclease cleavage assay, demonstrated cruciform structure formation along Flex1 harboring long AT-repeats [44]. Furthermore, an in vitro primer extension assay showed that ssDNA containing the Flex1 sequence was able to cause human polymerase δ holoenzyme stalling and replication termination [44].

Another study aimed to reveal the molecular basis for the specific instability of CFSs, focused on FRA16C CFS (16q21–16q22.1), harboring several long (>400 bp) AT-rich repeats, among them a very long repeat region of ∼2250 bp [24]. Analyzing the replication dynamics along the FRA16C region, utilizing the high-resolution DNA combing approach, revealed a slower replication fork progression already under normal growth conditions, compared with the entire genome. Analysis of blocks in the replication fork progression revealed a preferential clustering of stalled forks near the AT-rich repeats. Mild replication stress conditions led to an additional reduction in fork rate and increased number of stalled forks, however, no activation of additional rescue origins was observed, leaving this region incapable to complete its replication. These results suggest that the underlying basis of FRA16C instability is fork arrest at AT-rich repeats together with a regional inability to activate additional origins in response to fork stalling, to rescue the replication completion. It is important to note that an expanded long AT-rich repeat (921 bp) derived from this genomic region was previously shown to form stable secondary structures using the re-duplexing assay and to promote fork reversal and polymerase stalling in HEK293 cells transfected with replication constructs carrying this fragment [45]. Altogether, these results support the concept that the formation of highly stable secondary structures at AT-rich repeats is a general mechanism contributing to fragility during replication.

In line with these results, replication dynamic analysis of another CFS, FRA6E, which overlaps with the large Parkinson protein 2 (PARK2) gene and also harbors various relatively long AT-rich repeats (400–1400 bp, according to TwistFlex analysis performed by us), showed decreased replication rate, shorter distance between origins together with increased frequency of fork arrests compared with the whole genome [46]. Interestingly, it was found that the breakpoint-clustering region along PARK2, in patients with autosomal-recessive juvenile Parkinsonism (AR-JP) as well as in cancer cell lines, was flanked by sequences which are highly enriched with AT-dinucleotide, further suggesting the contribution of AT-rich repeats to the vulnerability for rearrangements [47].

The question whether an AT-rich repeat by itself is able to drive fragility was further addressed by targeting a very long AT-rich repeat (3.4 kb) derived from the endogenous FRA16C CFS into a chromosomally non-fragile ectopic region on the X chromosome [48]. Analysis of >1300 X chromosomes, harboring the integrated long AT-rich repeat, revealed recurrent gaps and breaks at the integration site, suggesting that a long AT-rich repeat is indeed able and sufficient to recapitulate fragile site instability. Additional analyses using the re-duplexing assay showed that DNA fragments derived from the integrated AT-rich repeat have an increased tendency to form stable secondary structures under single-stranded state, further supporting their ability to perturb DNA replication [48]. The breakage frequency of the newly generated fragile site was significantly lower than the frequency at the endogenous FRA16C, indicating the involvement of additional factors as other AT-rich repeats, different replication timing program or regional initiation profile along the FRA16C site.

Previous studies showed that the repertoire of CFS is tissue-specific [21,22,49,50], corresponding to the tissue-specific replication initiation signature, namely late replication timing combined with paucity of initiation events [13,49]. This particular replication program was observed along the most expressed CFSs in lymphocytes, FRA3B and FRA16D, and in fibroblasts, FRA3L and FRA1L, confirming a correlation between replication-timing profile and instability. Still, FRA3B and FRA16D are recurrently unstable in fibroblasts, although with a reduced frequency, even though initiation events are evenly distributed along the core regions. The inherent instability of this regions is suggested to stem from perturbed replication along AT-rich repeats present at their core fragility regions. Interestingly, the core fragility of FRA3L and FRA1L CFSs was also found to harbor AT-rich repeats [49], further indicating an involvement of difficult-to-replicate sequences in the regional instability. We thus suggest that CFSs plasticity is dictated by both genetic and epigenetic factors which set the breakage repertoire and expression frequency.

The ability of repetitive sequences to hamper DNA replication was observed also along early replicating fragile sites (ERFSs), another class of fragile sites characterized by early replication timing and high number of origins as compared with CFSs [51]. Tubbs et al. [52] has found that, in murine B lymphocytes treated with Hydroxyurea (HU), homopolymeric (dA/dT) tracts (>20 bp) in the vicinity of replication origins are preferential sites of replication fork stalling and collapse, due to their susceptibility to form non-B DNA structures. Interestingly, it was found that almost 60% of the top HU-induced double strand breaks (DSBs) show recurrent CC/GG dinucleotide sequence interruptions of the poly(dA:dT) tract. This CC/GG signature is indicative of translesion polymerase kappa errors (which involve dG or dC residue insertions) [53], suggesting that preferential sites of fork stalling and collapse in ERFSs are replicated by polymerase kappa, leaving its unique signature at poly(dA:dT) repeats. In another recent study, it was found that the extremities of the BRCA1 gene, identified as an ERFS, are enriched with the repetitive DNA sequences poly(T)n and (ATTATT)n, able to form secondary DNA structures that stall replication fork progression [54]. Under conditions of BRCA1 insufficiency in MCF10A cells, there was an increased usage of error-prone DNA repair pathways to mend these stalled forks, which led to increased levels of mutations, genomic rearrangements, and loss of heterozygosity.

Altogether, numerous studies in different experimental systems, focusing on various fragile sites, have demonstrated that AT-rich sequences can challenge the DNA replication, leading to replication fork arrest and DNA breakage. The ability of these sequences to perturb replication dynamics may lead to incomplete replication and predisposition to genomic instability.

As discussed above, many CFSs are enriched with AT-rich repeats able to form secondary structures that play a role in CFS instability. Yet, other studies found additional sequence elements which tend to fold into stable secondary structures at CFSs. Analysis of chromosome 10 found that aphidicolin-induced CFSs are significantly more favorable of forming stable DNA secondary structures compared with chromosomes 10 non-fragile regions [55]. Interestingly, the fragile regions were generally GC-rich with only few AT-rich, implying diversity in the sequence elements folding into stable secondary structures leading to fragility [55].

Processing of structural impediments emerging during the replication along CFSs in one of the mechanisms involved in CFS maintenance. The following proteins have a role in CFS stability:

One of the key regulators of fragile site stability, is the replication checkpoint kinase ATR (ATM- and Rad3-related protein kinase) that is able to recognize and respond to stalled forks or to incomplete replication along structure-forming repeats. It was previously shown that ATR act already during unchallenged cellular replication and its deficiency led to a significant increase in aphidicolin-induced FRA3B and FRA16D expression in both HCT116 and HeLa cells [56]. Furthermore, a genome wide screen of replication fork collapse sites induced by ATR inhibition combined with aphidicolin in human breast epithelial cells showed that the sequence elements most commonly associated with fork collapse were repetitive DNA that exhibited high structure-forming potential [57]. Notably, hairpin-forming AT-rich repeats were among the most ATR-dependent sequences for their stability. Using the high resolution Breaks Identifies by TdT Labeling (BrITL) method to identify sites of DSB formation, it was found that ATR-dependent break sites overlap with CFSs, notably, in 18/35 observed CFSs, the break occurred at AT-rich repeats [57]. These results were reinforced recently by Shaikh et al. [4] showing that the same AT-rich repetitive motifs are associated with a CNV class of Mb scale genomic losses following aphidicolin treatment.

Altogether, these findings emphasize the vulnerability of structure-forming repetitive sequences to genome instability and their high dependency on ATR function for stability. Importantly, other repetitive sequences not expected to form hairpin structures, such as poly(A)n; poly(T)n, LINE, SINE, LTR, and Alu elements, were not enriched in BrITL after ATR inhibition.

Various nucleases and helicases were implicated in CFS stability, among them structure-specific endonucleases (SSEs), which target and resolve branched DNA intermediates that occur at stalled forks and during recombination [58]. The SSE complex MUS81–EME1 is recruited to chromatin following mitotic onset [59], possibly as a result of the cellular attempt to compact under-replicated loci [2]. The ability of MUS81–EME1 to cleave stalled replication forks [60] allows regions of un-replicated DNA to unravel before sister chromatids segregation [61]. It was further found by the Hickson lab that the nuclease activity of MUS81 can promote POLD3-dependent DNA repair synthesis, in a novel process termed Mitotic DNA Synthesis (MiDAS), aims to minimize chromosome mis-segregation and non-disjunction during nuclear division [2]. If MUS81–EME1 fails to cleave the stalled forks in early mitosis, the sister chromatids would not be disjoined properly, and the region of un-replicated DNA would form an anaphase bridge [61]. The SSEs MUS81–EME1 and XPF–ERCC1 were shown to promote the cytogenetic appearance of FRA16D expression in mitotic cells that had experienced replication stress, indicating that CFS breakage is an active, MUS81–EME1-dependent process, promoting correct sister chromatid segregation [61,62]. Consistently, MUS81 depletion resulted in a significant decreased fragility at the FRA16D–Flex1 sequence, supporting the view that the requirement for MUS81 for FRA16D cytogenetic manifestation is due to structure-mediated cleavage [44]. Moreover, the fragility along FRA16D–Flex1 was shown to be dependent on other SSEs, including Slx1–Slx4 and Rad1–Rad10, all functioning together to cleave the sequence-induced structure. Kaushal et al. [44] further showed that the presence of multiple secondary structures along repetitive sequences can cause inefficient repair of the resected ends left after nuclease cleavage. Altogether, the instability at FRA16D–Flex1 is suggested to stem from a combination of increased DNA breaks and their impaired repair. The reduced healing efficiency along this region may explain its high deletion occurrence in various cancer cell lines [43]. The carboxy-terminal binding protein (CtBP)-interacting protein (CtIP) was also shown to play a role in processing DSBs occurring at CFS-derived AT-rich repeats [63]. Using EGFP-based homologous recombination (HR) substrates harboring hairpin-forming AT-rich repeats, it was demonstrated that Flex1–FRA16D sequence is genetically unstable and induces HR-mediated mitotic recombination. Further analyses revealed that the Mre11 complex and CtIP are specifically required for processing DSBs generated at the vicinity of the Flex1 sequence. It was further shown that in order to promote HR, CtIP process the DSBs ends using an endonuclease activity to cleave the 5′ ssDNA at the base of hairpins or Y shaped branched structures.

Another protein involved in the resolution of DNA secondary structures is the Werner helicase (WRN) [64]. WRN deficiency was shown to enhance chromosomal instability in human fibroblasts as reflected by increased level of total gaps and breaks under normal conditions and aphidicolin treatment [65]. Analyzing the breakage frequency of FRA3B, FRA7H, and FRA16D CFSs revealed highly increased expression levels upon WRN depletion. A direct indication for the importance of WRN for allowing bypass of secondary structures formed at the AT-rich repeats was shown by the Eckert lab. Using an in vitro replication assay, the addition of WRN relieved the polymerase stalling near AT-rich fragments derived from FRA16D CFS [66]. Interestingly, WRN was recently proposed as a target for cancer therapy, as cancers with microsatellite instability (MSI) were shown to be dependent on its function [37]. MSI results from mutations in core components of the mismatch repair (MMR) machinery and is associated with increased DNA damage and high mutational burden. Importantly, WRN was shown to protect cancer cells with MSI from massive DSBs, by alleviating non-B DNA secondary structures at AT-rich repeats which are highly unstable in MSI cells and undergo large-scale expansions [37,67]. During DNA replication, these expanded AT-rich repeats form cruciform structures that stall replication forks, activate the ATR checkpoint kinase, and require unwinding by the WRN helicase, in order to complete DNA replication. In the absence of WRN, these expanded AT-rich repeats will not be resolved before mitosis, when they are recognized and cleaved in an unscheduled manner by MUS81 to generate massive DSBs and consequently cellular lethality. In line with this, WRN depletion in MSI cell lines induced DNA breakage precisely at (AT)n repeats within FRA16D, FRA3B, FRA10B and FRA7I CFSs and at palindromic AT-rich repeats [37]. Nearly all breaks associated with WRN deficiency in both KM12 and HCT116 cells occurred at AT-rich repeats, characterized by high AT content, longer uninterrupted (AT)n sequences and the likelihood to replicate in late S-phase. A more recent study by Mengoli et al. [68] provided additional insights into the mechanisms underlying the synthetic lethality in MSI cells upon WRN inhibition. Using biochemical assays, Mengoli et al. demonstrated that AT repeats are particularly prone to form cruciform structures. This tendency was related to the low melting temperature of (AT)n sequences compared with DNA with inverted repeats of a random sequence with the same length. Furthermore, Mengoli et al. showed that the WRN helicase can efficiently and directly unfold cruciform structures, using its helicase activity and ATP hydrolysis-driven motor activity. It was further demonstrated that the MMR complexes (MutSα, MutSβ and MutLα) can exploit the highly dynamic structural transitions of cruciform DNA, disrupting the secondary structure in a different mechanism than WRN. Collectively, these results demonstrate that the WRN helicase and MMR complexes independently and synergistically govern the toxicity of the cruciform DNA structures at AT repeats in MSI and WRN deficient cells. Another helicase implicated in the stability of structure-forming AT-rich repeats at CFS is the Bloom helicase (BLM), the helicase deficient in Bloom syndrome. Bloom syndrome is a rare autosomal recessive disorder associated with constitutionally increased chromosome instability and a tendency to develop malignancies (reviewed in [69]). Cytogenetic analysis of Bloom syndrome cells showed spontaneous chromosome aberrations correlating with bands harboring fragile sites, oncogenes, and breakpoints involved in cancer rearrangements [70]. The BLM helicase, unwinds a variety of DNA substrates including Holliday junctions, forked duplexes and G-quadruplex DNA and is involved in multiple pathways contributing to maintenance of genome stability [64]. Wang et al. [71] identified a new role of BLM in preventing DSB formation and mitotic recombination at AT-rich repeats derived from CFSs, dependent on its helicase activity and ATR-mediated phosphorylation. Using HR-Flex1 reporter, it was shown that BLM is recruited to Flex1 sequences to suppress Flex1-induced mitotic recombination caused by DSBs. This protective role of BLM was also observed when AT-rich repeats derived from FRA16C were inserted into the reporter, suggesting a general mechanism to avoid instability at AT-rich repeats [71]. BLM inactivation under replication stress conditions caused γH2AX enrichment along AT-rich repeats at the Flex1 plasmid as well as along the endogenous FRA3B locus. Altogether, these results support the model that BLM resolves DNA secondary structure formed by AT-rich repeats at CFSs, thereby preventing fork collapse and damaged DNA.

The Fanconi anemia (FA) family of proteins was also shown to play an important role in CFS protection [72]. FA is a genetic disorder associated with a wide range of abnormalities including severe genome instability and cancer predisposition (reviewed in [73]). FANCM, one of the components of the FA core complex, was shown to be recruited to FRA16D-derived structure-prone Flex1 sequence, where it was able to suppress DSB formation and mitotic recombination [74]. In addition, FANCM depletion under replication stress conditions caused a significant increase in FRA16D and FRA3B expression in HCT116 cells. Another study, aimed to uncover the mechanism by which the FA family member FANCD2 regulates CFS stability, showed that FANCD2 facilitates replication through the AT-rich fragility core of both FRA16D and FRA6E CFSs by resolving DNA:RNA hybrid accumulation and by ensuring optimal firing of dormant origins [75]. Using single-molecule analysis of replicated DNA (SMARD) it was shown that under FANCD2 deficiency an increase in replication fork pausing was observed, accompanied with a prominent activation of dormant origin, suggesting that FANCD2 is a central regulator of replication along these CFSs [75].

Altogether, many cellular specialized proteins are required for resolving replication impediments at AT-rich repeats and thus are crucial for maintaining CFS stability prior to chromosome segregation.

In this mini review, we discuss the complex nature of CFSs, focusing on the role of difficult-to-replicate sequences, mainly AT-rich repeats, in the vulnerability of genomic regions to perturbed DNA replication. The deleterious consequences of AT-rich repeats are supported by the fact that many CFSs harbor AT-rich sequences and by various analyses showing that fragile regions are enriched in clusters of flexibility peaks relative to non-fragile regions [34,48]. Still, the human genome contains numerous AT-rich sequences in non-fragile regions, and furthermore, not all CFSs are enriched with AT-rich repeats, emphasizing that the genomic instability results from an interplay between genetic and epigenetic factors. The preferred sensitivity of CFSs to replication stress stems from various mechanisms and the interaction between them (Figure 1), including replication fork arrest at AT-rich repeats, origin paucity along large genomic regions, failure in activation of dormant origins, delayed replication timing and the 3D genome architecture, all leading to incomplete replication of the CFS region, resulting in chromosomal instability.

• CFSs are specific genomic regions preferentially unstable under replication stress conditions. CFSs show genomic instability during early stages of cancer development, as a result of the associated replication stress present under this condition.

• Many CFSs are enriched with AT-rich repeats which have the potential to fold into stable secondary structures upon unwinding the double helix during DNA replication. These stable structures can potentially perturb DNA replication progression, leading to genomic instability. AT-rich sequences within CFSs are genomic hotspots of instability, overlapping with recurrent cancer breakpoints.

• Proteins involved in the resolution of secondary structure impediments arising during DNA replication are essential for the maintenance of genome stability, specifically along repetitive sequence elements which tend to adopt alternative DNA structures.

• The basis underlying CFSs instability is complex (Figure 1), involving an interplay between delayed replication timing, large transcriptional units and the 3D genome architecture. The involvement of AT-rich repeats in CFSs instability was established in a subgroup of CFSs, however, further studies may enable linking such complex sequences to the fragility signature of other regions.

The authors declare that there are no competing interests associated with the manuscript.

Open access for this article was enabled by the participation of Hebrew University of Jerusalem in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with MALMAD.

This research was supported by grants from the Israel Science Foundation (grant no. 176/11 and 1284/18) and by the ISF-NSFC joint program (grant no. 2535/16) to B.K. The authors thank the members of the Kerem lab for thoughtful discussions and advice.

BLM

Bloom helicase

BrITL

breaks identifies by TdT labeling

CFS

common fragile sites

CNVs

copy number variants

DSBs

double strand breaks

ERFSs

early replicating fragile sites

FA

Fanconi anemia

HDs

homozygous deletions

HR

homologous recombination

HU

Hydroxyurea

MMR

mismatch repair

MSI

microsatellite instability

PARK2

Parkinson protein 2

SSEs

structure-specific endonucleases

TAD

topologically associated domain

WRN

Werner helicase

YAC

yeast artificial chromosome