The Yin and Yang of Repair Mechanisms in DNA Structure-induced Genetic Instability

Abstract

DNA can adopt a variety of secondary structures that deviate from the canonical Watson-Crick B-DNA form. More than 10 types of non-canonical or non-B DNA secondary structures have been characterized, and the sequences that have the capacity to adopt such structures are very abundant in the human genome. Non-B DNA structures have been implicated in many important biological processes and can serve as sources of genetic instability, implicating them in disease and evolution. Non-B DNA conformations interact with a wide variety of proteins involved in replication, transcription, DNA repair, and chromatin architectural regulation. In this review, we will focus on the interactions of DNA repair proteins with non-B DNA and their roles in genetic instability, as the proteins and DNA involved in such interactions may represent plausible targets for selective therapeutic intervention.

Repetitive sequences that do not code for proteins account for more than half of the total DNA in the human genome, which includes transposon-derived repeats, (e.g., Alu and LINE elements), pseudo-genes or duplications, and tandem simple repeats [1]. In contrast to its former nickname, “junk DNA”, repetitive sequences are now known to play important roles in regulating the genome, including putative involvement in shaping chromatin structure, regulating gene expression, and stimulating genomic rearrangement. Of particular interest, some simple repeats, which represent approximately 3% of the total human genome, have been found to form different types of non-canonical structures that differ from the Watson-Crick B-form (i.e., non-B DNA structures). Extensively studied microsatellite repeats, such as triplet repeats can contain tens, hundreds, or even thousands of trinucleotide repeat units [2, 3]. Expanded triplet repeats (such as CTG and CGG) have been implicated in more than 20 genetic diseases, including Fragile × syndrome, Huntington's disease, Friedreich's ataxia, and myotonic dystrophy [4, 5]. The role of microsatellites in disease is not limited to triplet repeats, as evidenced by Myotonic dystrophy type 2 (DM2), which is caused by the extreme expansion of CCTG tetranucleotide repeats from <30 repeats in normal individuals to up to 11,000 in some patients [6], and spinocerebellar ataxia type 10 patients, in which the number of pentanucleotide ATTCT repeats can be expanded from approximately 14 in normal individuals to up to 4,500 [7].

In cells, the majority of genomic DNA exits in the B-conformation at any give time (Figure 1A), but the conformational state of DNA is dynamic and is generally negatively supercoiled corresponding to a high-energy state. The DNA is subjected to manipulation via replication, transcription, and DNA repair proteins. These processes unwind the DNA from the histones, generating more negative supercoiling stress and can open the double helix, giving rise to non-B structures at sequences (e.g., repetitive sequences) conducive to their formation. To date, more than 10 different types of non-B DNA conformations have been identified and characterized [8]. Examples include, slipped structures formed at direct repeats, hairpin or cruciform structures formed at inverted repeats, intra-molecular triplex DNA structures (H-DNA) formed at polypurine-polypyrimidine elements with mirror repeat symmetry, four-stranded G-quartet structures formed at sequences comprising 4 guanine tracks of at least 3 continuous guanosine residues separated by 1–7 nucleotides, and left-handed Z-DNA structures formed at alternating pyrimidine-purine sequences (see Figure 1; for review, see [9]). Notably, under certain conditions, imperfect repeat sequences can also form non-B conformations. For example, short CG repeats (2–4 repeats) that are separated by 3-bp interruptions can adopt Z-DNA structures and form Z-Z junctions on plasmids in vitro, as assessed by two-dimensional gel electrophoresis and chemical and enzymatic probing [10]. Hairpin structure formation typically requires an inverted repeat symmetry, but can also form at CNG triplet repeats, which contain a mismatch in every two C-G [11, 12]. Thus, sequences with the potential to form non-B DNA structures (or non-B DNA-forming sequences) are very abundant in the human genome.

Figure 1. Schematic diagram of DNA conformations.

(A) Canonical B-DNA; (B) intra-molecular triplex , H-DNA; (C) left-handed Z-DNA; (D) cruciform; (E) hairpin (left) and slippage loop (right); (F) G-quadruplex.

1. Non-B DNA structure and genetic instability

Neurological disorders related to triplet repeat expansion have been studied extensively (for reviews see [13, 14]); however, many types of non-B DNA conformations in addition to loop structures due to slippage and misalignment (e.g., slipped DNA) of repetitive units or hairpins/cruciform formed at triplet repeat sequences have been implicated in genetic instability associated with a variety of human disorders [15][16–18], as we briefly outline below.

1.1. Intra-molecular triplex DNA (H-DNA) and genetic instability

H-DNA can form at polypurine-polypyrimidine sequences with mirror repeat symmetry, where half of the symmetry, when single-stranded, binds to the purine-rich strand of the duplex containing the other half of the symmetry via Hoogsteen hydrogen bonding through the major groove, forming a three-stranded DNA structure (Figure 1B) [19]. Genes containing long polypurine-polypyrimidine sequences show higher frequencies of alternative splicing and chromosomal translocations [20], and H-DNA-forming sequences frequently co-localize with chromosomal breakpoint hotspots and genomic rearrangements in disease-related human genes. For example, several polypurine-polypyrimidine tracks can be found near (within hundreds of bps) the major breakage hotspots in the c-MYC gene found in leukemias and lymphomas [21–28], and in the BCL-2 gene major breakpoint region (Mbr) in follicular lymphomas. In vitro experiments demonstrated that these sequences were capable of forming non-B DNA structures, including H-DNA [29]. Furthermore, disruption of the H-DNA conformation by disruption of the mirror symmetry reduced the frequency of translocation events in the BCL-2 Mbr [30]. The 21st intron of the human PKD1 gene, which contains a 2.5 kb polypyrimidine tract with 23 mirror repeats [31] that can adopt H-DNA structures in vitro [32], has been implicated in the high mutation rate in this region in both germ line and somatic cells from autosomal dominant polycystic kidney disease patients (ADPKD) [33]. The 2.5-kb polypyrimidine sequence in a plasmid induced DNA double-strand breaks (DSBs) and resulted in large-scale deletions in E. coli [34]. In our previous work, we discovered that both an endogenous H-DNA-forming sequence from the human c-MYC promoter and model H-DNA-forming sequences, when cloned into a mutation reporter shuttle plasmid, induced DSBs near the H-DNA locus and induced mutation frequencies ~20-fold above spontaneous background levels in mammalian COS-7 cells. Greater than 80% of the mutants exhibited large-scale deletions and/or rearrangements [35]. Moreover, when this H-DNA-forming sequence from the human c-MYC promoter was integrated into chromosomes of mutation-reporter mice, genetic instability was detected in ~8% of the offspring [36], indicating that H-DNA structures can be mutagenic in a chromosomal context in vivo.

1.2. Left-handed Z-DNA and genetic instability

Purine bases can adopt syn conformations in sequences of alternating purines and pyrimidines (e.g., GT, GC), which can facilitate the formation of a left-handed helix with a zigzag-shape backbone (i.e., Z-DNA, Figure 1C) that differs from the canonical Watson-Crick B-DNA helix [37, 38]. Among the alternating purine-pyrimidine sequences, GC repeats have the greatest capacity to form Z-DNA structures, followed by GT repeats, while AT repeats are more likely to form hairpin or cruciform structures rather than Z-DNA [39]. Although long GC repeats are rare in mammalian genomes, GT repeats are the most frequent simple repeating unit in the human genome, accounting for ~0.25% of the total DNA [40]. In addition to pure alternating purine-pyrimidine sequences, purine-pyrimidine repeats with interruptions have been shown to adopt Z-DNA in vitro under certain conditions [41, 42].

Algorithms designed to identify such sequences in the human genome have revealed co-localization of Z-DNA-forming sequences and genetic instability hotspots, particularly chromosome breakage and translocation sites in genes involved in human diseases such as cancer. For example, breakage hotspots in the c-MYC gene, e.g., the P1 promoter and 3' downstream breakpoint region, contain multiple mixed GT and GC repeats capable of forming Z-DNA [43, 44]. Translocation breakpoints also map within hundreds of bps surrounding Z-DNA-forming sequences in lymphoid tumors [45]. Sequencing analysis of a 1.7 kb DNA breakage hotspot cluster region from a subgroup of B-cell precursor acute lymphoblastic leukemia (BCP-ALL) patients identified a region of Z-DNA-forming GT repeats to be involved in amplification and rearrangement events [46]. This study also implicated a chi-like heptamer, CCTCAGC, in the rearrangements, suggesting that both DNA secondary structures and recombinogenic mechanisms may contribute to BCP-ALL etiology [46].

Plasmids carrying Z-DNA-forming sequences are difficult to maintain in E. coli, suggesting that they are unstable in bacteria. For example, CG repeats, which can form stable Z-DNA structures, induce contractions of repeat units in bacteria, likely generated by misalignment during replication [4, 47–49]. Other simple repetitive sequences such as hairpin-forming triplet repeats have been implicated in replication-induced slippage events. When the mutagenic potential of a hairpin-forming AT(12) repeat was compared to that of a CG(12) repeat, which may form a hairpin or Z-DNA, it was found that the CG(12) repeat was more mutagenic than the AT(12) repeat [48], implicating Z-DNA, rather than a hairpin structure in CG repeat-induced mutation in bacteria. Moreover, in mammalian cells model Z-DNA-forming CG(14) repeats induced DSBs within a 300-bp region surrounding the Z-DNA-forming sequences and induced mutation frequencies ~20-fold above spontaneous background levels, largely in the form of deletion and/or rearrangement events [50]. These same Z-DNA-forming repeats resulted in even greater levels of genetic instability on chromosomes of transgenic mutation-reporter mice with ~7% of the population undergoing large deletion and/or rearrangement events in the vicinity of (within 300 bp) the non-B region [36]. On the contrary, this same sequence resulted exclusively in contractions or expansions of repeat units in bacteria [50]. Thus, there are at least two distinct mechanisms of mutagenesis stimulated by Z-DNA-forming sequences in various organisms.

1.3. Hairpins, cruciforms, and G4 DNA in genetic instability

Inverted repeat sequences (or palindromes) can adopt hairpin conformations via intra-strand base pairing, where half of the single-stranded symmetry folds back and forms intra-strand Watson-Crick base pairs with the other half on the same strand (Figure 1D) [51, 52]. Palindromes as short as 14 bp are sufficient for formation of a stable hairpins in vivo [53]. Long inverted repeats are capable of forming large and complex structures. For example, a long inverted repeat on the human X chromosome was reported to form a large double-cruciform structure involving multiple Holliday junctions which allowed for both intra-strand base pairing within the stems and interstrand base pairing of the spacer region bringing the hairpin stems together [54]. Inverted repeats are involved in generating deletions, gene amplification, and translocations, and can serve as substrates for recombination in bacteria [55], yeast [56, 57], and mammalian cells [54, 58–61]. The centers of the AT-rich long inverted repeats on human chromosome 11q23 and 22q11 co-localize with the breakpoints for t(11;22) translocations [62–64], and references therein]. Imperfect inverted repeat sequences, such as triplet repeats, can also adopt slippage loops or hairpin/cruciform structures (Figure 1E) with mismatches in the stems [13–15, 65], which can undergo expansion and/or deletion events in vivo, with a bias toward expansions in humans and deletions in bacteria, yeast, and mice [4, 5, 66, 67]. Further, long inverted repeats, and presumably the hairpin/cruciform structures formed, represent preferred integration sites for avian sarcoma virus and human immunodeficiency virus-1, at least in vitro [68]. The viruses tend to integrate within the stems of cruciform structures on plasmid DNA, adjacent to the loops in the cruciforms [68]. Long inverted repeats are also associated with chromosomal translocations in the human genome [69]. Taken together, these results provide strong evidence that cruciform structures formed at long inverted repeat sequences contribute to genetic instability and evolution.

G-quadruplexes or G4 DNA, four-stranded Hoogsteen-hydrogen bonded structures formed at G-rich sequences (Figure 1F), have been the subject of many studies related to the function of alternatively structured DNA (for reviews see [70–72]). G-quadruplex-forming sequences are found in telomeres, and are enriched in promoters and rDNA, and are even more abundant in mitochondrial DNA than in nuclear DNA in yeast [73, 74]. G4 DNA motifs were found to be more conserved than expected by chance [74], consistent with the suggested biological roles of G-quadruplexes in transcription, replication, and telomere maintenance [70]. G-quadruplex-forming sequences have also been identified in regions within 500 bp of mitotic and meiotic DSB sites [74, 75], and at hotspots of genetic instability in human disease [70]. For example, G-quadruplex structures can interfere with DNA replication leading to stalled and/or collapsed replication forks, which can result in DNA breakage and genetic instability [76–78]. It has also been demonstrated that when G4-forming sequences serve as the template for leading strand replication, G4 structure formation can result in deletions and complex rearrangement events in the absence of the G-quadruplex unwinding helicase Pif1 [78, 79]. The G-quadruplex-induced rearrangements are dependent on the homologous recombination proteins Rad51–Rad52 [79, 80]. Moreover, G4 DNA plays an important role in telomeric repeat stability [81]; on the one hand the single-stranded G-rich telomeric repeats are capable of forming G-quadruplex structures, which may protect the DNA ends from degradation, on the other hand, G-quadruplex structures must be resolved to allow for efficient replication of telomeres [81].

2. DNA repair-related proteins that associate with non-B DNA structures

The mutagenic potential of non-B DNA-forming sequences has been demonstrated in many species, implicating non-B DNA structures in disease etiology and genetic evolution [82]. It is thought that the non-B DNA structure itself, and not the sequence (in B-form DNA) is involved in this instability. Thus, structure formation and stability are key regulators of the subsequent mutagenesis, as are the enzymes that process these structures. A variety of proteins can recognize and bind to non-B DNA structures to stabilize or resolve the conformation, recruit other factors to the structures, or directly process the non-B DNA regions. A family of winged-helix proteins with a general mechanism of Z-DNA recognition has been described suggesting the presence of a common “Z-DNA-binding motif” [83]. For example, the Z-alpha motif of the adenosine deaminase acting on RNA 1 (ADAR1) protein binds with high affinity to Z-DNA [84], and facilitates a B-DNA to Z-DNA structural transition even at the short TA(3) repeat, which is unable to adopt Z-DNA in the absence of ADAR1 [84]. Five 14-3-3 family isoforms (beta, gamma, epsilon, zeta, and sigma), which have been implicated in signal transduction and cell cycle control, can bind to cruciform structures [85–87]. Many architectural and regulatory proteins, such as histones H1 and H5, HMG proteins, and p53 preferentially bind to cruciform structures, or can facilitate the formation of cruciform structures upon binding (for review see [88]). In addition to cruciforms, some of these architectural proteins can bind other non-B DNA structures such as triplex DNA [89].

Non-B DNA structures may provide signals for structure-specific binding proteins involved in DNA damage recognition and repair mechanisms. Nucleotide excision repair (NER) proteins are thought to recognize DNA damage via chemical and physical alterations in the Watson-Crick B-DNA helix. It has been reported that the efficiency of lesion recognition and processing by NER is dependent on the extent of distortion induced in the DNA helix [90–92]; i.e., lesions that induce greater distortions to the DNA helix are repaired more efficiently than less distorting lesions. Because non-B DNA deviates from the canonical B-DNA helix, these structures may be recognized as “damage” by the NER damage/distortion recognition proteins. This may also apply to other DNA repair proteins, as many non-B DNA-structure binding and/or processing proteins are involved in DNA repair, such as NER proteins, mismatch repair (MMR) proteins, helicases, and replication-related proteins, as discussed in detail below. Many other proteins are involved in processing the intermediates of non-B DNA-induced mutations, such as DSB repair proteins; here, we will focus on the enzymes that directly process non-B DNA structures, and not their intermediates, as the DSB repair pathways are beyond the scope of this review.

2.1. Nucleotide excision repair proteins

NER removes bulky DNA adducts by dual incision on either side of the damaged base(s) and removes the damaged nucleotide(s) together with the adjacent region. NER is a critical pathway for maintaining the integrity of the genome and is highly conserved in prokaryotes and eukaryotes [93, 94]. Because non-B DNA conformations result in helical distortions it is not surprising that some types of non-B structures can be recognized and processed by NER factors.

Inter-molecular triplexes, formed when a triplex-forming oligonucleotide (TFO) binds duplex DNA via Hoogsteen hydrogen bonding [95], have been shown to induce NER-dependent mutagenesis in mammalian cells [96]. In HeLa nuclear extracts, the NER recognition/verification proteins, xeroderma pigmentosum group A (XPA), and group C (XPC) were reported to be involved in processing inter-molecular triplexes [97]. Consistent with these results, in vitro binding studies have confirmed that the XPA-RPA complex [98] and the XPC-RAD23B complex [99] recognize covalent inter-molecular triplexes with high affinity and specificity, and can interact with each other on these structures [99]. The bacterial NER equivalent, the UvrABC endonuclease, has the capacity to process inter-molecular triplex structures in vitro [100], suggesting that triplex structures are substrates for NER across species.

Studies on the role of NER in hairpin/cruciform structure-induced genetic instability have largely focused on CNG triplet repeats. The E. coli NER protein UvrA can bind CAG repeats on a supercoiled plasmid that supports a hairpin conformation at a much higher affinity than to linear CAG repeats, and the deletion of CAG repeats was reduced in a UvrA-deficient E. coli strain [101]. Deficiencies in the endonuclease (UvrC) of the UvrABC complex or the UvrD helicase resulted in an enhancement of deletion events at CAG repeats [102], while deficiency of UvrB, which cleaves phosphodiester bonds downstream of the damaged sites, reduced the deletion events on the CAG tracts [102, 103]. In another study, the absence of UvrA in bacteria increased the deletion events at CAG repeats on plasmid DNA [103]. The authors speculated that UvrA could bind to the bubble and loop regions formed at the repeats [104, 105], and thereby prevent bypass synthesis and deletion of the repeats [103]. However, the stability of CAG repeats integrated in the bacterial genome was not affected by UvrA deficiency [102], while XPA depletion in human cells significantly reduced the transcription-induced deletion events on CAG repeats [106]. In Drosophila, CAG repeats exhibited length-dependent instability with a bias toward expansions [107]. Transcription through the repeats dramatically increased both contraction and expansion events, and deficiency of Mus201 (an ortholog of human XPG) suppressed repeat instability, suggesting an important role of transcription-coupled NER (TC-NER) in CAG repeat stability [107]. This was also the case in human cells where transcription through CAG repeats led to repeat contraction [106], and siRNA depletion of CSB, a protein required for TC-NER, and the NER endonucleases ERCC1 and XPG, suppressed CAG repeat contractions [108]. Together, these studies reveal complex effects of NER proteins on the stability of hairpin-forming triplet repeats. The role of NER in processing cruciform structures formed at perfect-inverted repeats is still not clear. Further work is required to better define the roles of NER proteins in triplet repeat instability in various organisms.

2.2. Mismatch repair proteins

Cells deficient in MMR are hypermutable and predisposed to tumorigenesis as evidenced by hereditary non-polyposis colon cancer in humans [109, 110]. MMR is responsible for the correction of base-pairing errors such as mismatches and insertion-deletion loops caused by spontaneous and induced base deamination, oxidation, methylation, reactive oxygen species, recombination intermediates, and replication errors. In this capacity, MMR proteins interact with components of other repair pathways including NER to prevent spontaneous mutation and microsatellite instability [111]. Single-stranded looped regions are found in many types of non-B DNA conformations, such as hairpins, slipped structures [82], and B-Z or Z-Z junctions [112, 113], which may represent substrates for MMR. For example, MMR has been shown to be required for removing a small cruciform structure (12 bp) in Chinese hamster ovary cells [114] and in yeast [53]; yet neither MSH2 nor MLH are required for repairing large loops (e.g., 140 bp) in mammalian cells [57, 115, 116]. The MSH2–MSH3 complex binds to hairpin structures containing mismatches with higher affinities than hairpin structures formed at perfect inverted repeats [117]. So although MMR is efficient in removing small loops within hairpin or cruciform structures, it does not appear to process longer cruciforms that contain perfect base-pairing in the stem region. This is consistent with the well-known function of MMR in post-replicative repair of mismatches and small loops [118]. While MMR deficiency results in instability at microsatellite sequences in humans, the absence of the Msh2 in yeast does not significantly affect the large deletions (>12 repeat units) at CAG(50) repeats [119] or large expansions at CAG(25) (>13 repeat units) [120]. Interestingly, in mouse models of Huntington's disease or myotonic dystrophy (carrying long CNG repeats), deficiencies of Msh2 [121], Msh3 [122], or Pms2 [123] reduced repeat expansion events, or shifted the mutation spectra from expansions to deletions [124]. However, in human cells, MSH2 and MSH3 were responsible for the contraction events at CAG(95) repeats [106]. The reasons for these differences are not clear, but suggest that MMR proteins are not the only proteins that can process hairpin structures formed at CNG repeats, thus subtle differences in intracellular environment may alter the outcomes of CNG repeat processing. The MSH2–MSH6 heterodimer can bind to G-rich repetitive sequences in activated immunoglobulin switch regions in B-cells, which can form G-quadruplexes and RNA/DNA hybrids during transcription [125, 126]. Further, MSH2 or MSH6 deficiency in B-cells can reduce switch recombination and heterogeneity of switch junctions [76–78]. MMR proteins have also been implicated in the recognition and processing of TFO-directed inter-molecular triplexes [127–129]. MSH2–MSH3 binds inter-molecular triplex structures with high affinity and specificity in vitro and in vivo, and MSH2–MSH3 can interact with XPA-RPA or XPC-RAD23B in recognizing these structures [129]. In E. coli, MutS and MutL are required for the large deletions induced by H-DNA-forming GAA repeats from the human PKD1 gene [34]. Thus, MMR proteins clearly play a role in non-B DNA structure-induced genetic instability in a number of species.

2.3. Base excision repair proteins

Base excision repair (BER) removes lesions formed by alkylating agents and oxidative DNA damage, including non-bulky base modifications and abasic sites. Defects in BER have been association with hereditary neurodegenerative diseases, immunodeficiency, cancer, and aging (For a review, see [130]).

BER proteins have been implicated in triplet repeat instability; for example, cells deficient in DNA polymerase beta and the BER cofactor, HMGB1, did not support CAG expansions [131]. DNA polymerase beta was found to be enriched at expanded CAG repeats in the striatum of mice representing a model for Huntington's disease (HD), where the tissue shows high somatic CAG instability, but not in the cerebellum where the CAG repeats are more stable [132]. On the contrary, the activities of Flap endonuclease-1 (Fen-1), Hmgb1, and Ligase 1 were significantly lower in the striatum of these mice [132, 133]. The authors speculated that a lower Fen-1/polymerase beta ratio in the striatum could lead to an accumulation of 5'-flap structures at CAG repeats generated by polymerase beta strand displacement during long patch BER of oxidative DNA lesions. If not repaired, the 5'-flap-containing CAG repeats can form hairpin structures and lead to expansion events [132]. Fen-1 can cleave flap structures and suppress hairpin-induced repeat expansion events [134], and the loss of Rad27 (Fen-1 homolog in yeast) in S. cerevisiae increased CNG repeat expansions and deletions several hundred-fold above background levels [135, 136]. In contrast, stable hairpins formed at longer repeats (e.g., >20 CTG/CAG repeats) were resistant to the activity of Fen-1 [137, 138]. In fact, when the activity of Fen-1 was uncoupled from polymerase beta, it was found to promote expansions by facilitating ligation of the slipped repeats in the hairpins to the single-stranded DNA in the gap, resulting in expansions during the next cycle of replication [131].

8-oxoguanine glycosylase (OGG1) is a DNA glycosylase that is involved in the processing of 7,8-dihydro-8-oxoguanine (8-oxoG) by BER, a common lesion generated by exposure to reactive oxygen species. It was found that repair processing by OGG-1 can stimulate non-B DNA formation and result in repeat expansion in mice [139, 140], as described in the “toxic oxidation cycle” [141]. This model will be discussed in detail in section 3.3. “DNA damage and repair processes can alter DNA structure” below and in [142]. Interestingly, Ogg1 deficiency in yeast mitochondria resulted in an ~six-fold increase in GT repeat deletions as compared to the wild-type, and overexpression of Ogg1 completely suppressed GT repeat deletion events, suggesting that the activity of Ogg-1 could reduce repeat instability mitochondria [143]. In human cells, depletion of OGG-1 by siRNA to 30% of the endogenous levels did not affect the transcription-related CAG repeat instability [108]. Although the differences among these reports are not clear, it appears that Ogg-1 is involved in oxidative damage-related instabilities within repetitive sequences.

2.4. DNA helicases

DNA helicases are critical for unwinding duplex DNA, which is required for DNA metabolic processes including replication, transcription, and repair. Many DNA and RNA helicases have been characterized and it is estimated that as much as 1% of known open reading frames in the human genome code for predicted DNA and RNA helicases [144]. Helicases play a key role in genome maintenance, such that deficiencies in some helicases can results in premature aging and/or increased susceptibility to cancer [145]. For example, deficiencies in BLM, WRN, or RECQL4 are associated with human Bloom's, Werner's, and Rothmund-Thomson syndromes (RTS), respectively. Helicases play important roles in DNA repair and interact with many DNA repair proteins such as BRCA1 [146] and Topo III [145, 147, 148]. Some helicases are more efficient at resolving non-B DNA conformations than B-DNA duplexes [149], and see below], suggesting that they may protect the genome from mutagenic non-B DNA structures.

Non-B DNA structures can stall replicative DNA polymerases resulting in replication fork collapse and subsequent DNA breakage and deletions. Helicases are actively involved in preventing replication stalling, and can assist in replication fork restart, in part by resolving non-B DNA structures [144]. The RecQ helicase family, a superfamily 1 helicase group, has gained increasing attention in studies related to their activity on non-B DNA structures. E. coli and some lower eukaryotic species encode only one RecQ helicase (e.g., RecQ in E. coli and Sgs1 in budding yeast and S. cerevisiae), while higher organisms contain multiple RECQ family members. RecQ helicases bind ssDNA or dsDNA with 3'-overhangs better than blunt substrates or those with 5'-overhangs, and unwind DNA in the 3' to 5' direction [150].

In BLM-deficient mice, cruciform-forming inverted repeats were found at deletion junctions, while deletions in BLM-proficient cells were associated with direct repeats, suggesting that the BLM helicase plays an important role in preventing cruciform-induced genetic instability [151]. WRN-deficient mice experienced an 8-fold increase in mutations at a long inverted repeat sequence integrated at a Line-1 element [152], demonstrating the importance of helicases in maintaining genome stability at non-B DNA structure-forming sequences. RecQ helicases bind Holliday junction (HJ) structures, which are conformationally similar to cruciform structures, with higher affinities than duplex DNA, and promote branch migration in an ATP-dependent manner [153, 154]. In fact, it has been reported that mammalian RECQ helicases, such as WRN, BLM [155–158], and RECQL1 [159] have the capacity to unwind HJ over thousands of bps.

The yeast Sgs1 helicase, in a complex with Top3 and Rmi1, resolves aberrant recombination intermediates [160], disassembles branched DNA structures, and plays an integral role in meiotic recombination [161]. Sgs1-deficient yeast cells accumulate cruciform-shaped replication intermediates at damaged forks [162], and these unprocessed structures can be resolved by overexpression of the E. coli RusA or human GEN1 HJ resolvases [163], or the endogenous Mus81-Mms4 endonuclease [164].

Purified bacterial RecQ protein has been reported to melt triplex structures and facilitate primer extension by T7 DNA polymerase through polypurine-polypyrimidine tracts [165]. Purified BLM and WRN helicases can also unwind the third strand from TFO-induced inter-molecular triplex containing a 3' overhang [166]. An inter-molecular triplex containing a 3' tail might mimic the structural features of a stalled replication fork. WRN was shown to resolve H-DNA and Z-DNA structures in vitro, however, depletion of WRN protein in human cells resulted in increased mutation frequencies in H-DNA-forming, Z-DNA-forming, and B-DNA-forming sequences, suggesting that the increased genetic instability was independent of DNA structure [167]. Hairpin and G-quadruplex structures formed at expanded CGG repeats can stall the eukaryotic replicative DNA polymerases alpha, delta, and epsilon; and the WRN helicase can interact with DNA polymerase delta to facilitate DNA synthesis through the CGG repeats [168].

The RQC domain, a conserved G-quadruplex DNA binding domain located at the C-terminus of the RecQ family helicases, is responsible for high affinity interactions of helicases with G-quadruplex structures [169]. Human telomeres contain long 3' single-stranded TTAGGG repeats and can adopt non-B DNA conformations including G-quadruplexes [170]. In fact, the first crystal structure of a G-quadruplex was obtained using human telomeric repeat sequences [171]. The 3'-end provides a substrate for RecQ helicase loading, and RecQ family helicases are important for maintaining G-rich sequences including the telomeres and immunoglobulin switch regions [145, 147, 169, 172].

In addition to the RecQ family helicases, the FANCJ helicase can unwind non-B DNA conformations such as G-quadruplexes [173] and is involved in maintaining genomic stability. FANCJ, also known as BRIP1 or BACH1, is a member of the superfamily 2 helicases. FANCJ was originally identified through its interaction with BRCA1 and its role in DNA repair and breast cancer suppression [146]. Fanconi anemia patients deficient in FANCJ are hypersensitive to DNA crosslinking agents [174], and consistent with this phenotype, FANCJ has been implicated in crosslink repair via its interaction with MLH1 [175]. Different from the RECQ helicases, FANCJ requires a 5' ssDNA tail for loading and it unwinds duplex DNA substrates in a 5' to 3' direction with respect to the helicase-loaded strand [176].

The C. elegans Dog-1 (deletion of guanine-rich regions) gene, the FANCJ homolog in worms, is required for suppressing deletion events that occur at G-rich regions capable of forming G-quadruplexes [177]. In Dog-1-deficient cells, mutation frequencies at G-rich repeats correlated with the tract lengths and G-richness, i.e., the ability to form a G-quadruplexes, suggesting a critical role of Dog-1 in suppressing G-quadruplex-induced genetic instability [178]. Moreover, Dog-1 and Him-6 (a BLM homolog) double mutants exhibited higher instability at G-quadruplex-forming regions than in Dog-1 mutants [178], suggesting that there are multiple helicases involved in maintaining the stability of G-quadruplex-forming regions in worms. Similar to the functions of Dog-1 in C. elegans, human FANCJ can also resolve G-quadruplex structures in vitro in an ATP-dependent manner, and FANCJ-deficient human cells have deletions at G-quadruplex-containing regions [173].

Another 5'–3' helicase that plays an important role in resolving non-B DNA is the Pif1 protein, a member of helicase superfamily 1, which is found in most eukaryotes [179]. Pif1 helicase is involved in the maintenance of mitochondrial DNA and genomic DNA, and has a particularly important role in telomeric DNA replication in eukaryotic species from yeast to human [179]. In yeast, Pif1 deficiency can result in replication stalling and DSBs at G-quadruplex regions [180]. Consistent with this finding, Pif1 was found to be enriched at G-quadruplex-forming sequences in vivo [78]. Human PIF1 also binds to a G-quadruplexes in vitro with higher affinity than to ssDNA or duplex DNA [181]. Pif1 can unwind G-quadruplex DNA and this reaction requires a 5' ssDNA tail [181], similar to the FANCJ helicase.

DHX9 belongs to the helicase superfamily 2, and is also known as nuclear DNA helicase II (NDH II) or RNA helicase A (RHA) and is essential for efficient transcription and DNA repair. DHX9 unwinds DNA-DNA, RNA-RNA or DNA-RNA duplexes in a 3' to 5' direction, although it unwinds RNA hybrids more efficiently than the corresponding DNA-DNA substrates [182]. DHX9 has been shown to interact with WRN and co-localizes with γ-H2AX at centrosomes in human cells [183]. Following DNA damage, the pre-assembled DHX9-γ-H2AX complex disassociated from the centrosome and was recruited to sites of damaged DNA to assist in their repair [183]. Interestingly, DHX9 helicase co-precipitates with inter-molecular triplex structures in mammalian cells [184], and can resolve these structures in vitro with greater efficiency than that of forked structures. The activity of DHX9 on triplex structures was found to be dependent on ATP hydrolysis and a 3'-free overhang on the third strand TFO [184]. DHX9 has the capacity to unwind intra-molecular H-DNA and Z-DNA structures on plasmids in human cells [167] and is enriched at these structure-forming sequences in cells (Jain et al., unpublished results), suggesting that either a 3' tail is generated during the processing of non-B DNA or that a 3' tail is not necessary for resolving these structures by DHX9 in vivo. The generation of a 3' tail may lead to structure resolution independent of DHX9 because a nick or break would lead to the loss of negative supercoiling. Another possibility is that DHX9 is involved in processing the DSBs induced by non-B DNA as DHX9 has been shown to interact with non-homologous end-joining proteins [185]. In contrast to results from studies using WRN-depleted cells [167], DHX9 depletion resulted in an H-DNA-specific increase in mutagenesis in human cells, suggesting differences in structure-specific functions of helicases (Jain et al., unpublished results). DHX9 helicase has also been shown to unwind RNA and DNA G-quadruplexes in vitro [182], implicating this helicase in the maintenance of genome stability at regions of alternatively structured DNA.

There are many other helicases that are capable of unwinding non-B DNA conformations, such as the SV40 large T antigen, which has been shown to unwind the third strand of triplex structures [186], and RTEL1 that acts on G4 DNA to assist in maintaining telomere integrity [187]. Another superfamily 1 helicase member, the ATP-dependent Srs2 DNA helicase can function with DNA polymerase delta via interactions with PCNA to unwind hairpin structures formed at CGG repeats (but not G4T4 repeats that form G-quadruplex structures), and facilitate replication through the repeats, thus preventing chromosome fragility in yeast [188]. Compared to wild-type cells, Srs2 deficient yeast cells showed an ~33-fold increase in CTG repeats expansions and contractions, ~4-fold more chromosomal breakage on yeast artificial chromosomes [189], and up to a 40-fold increase in expansion rates at CTG, CAG, and CGG repeats on episomal plasmids [190]. As summarized above, in most cases, helicase activity facilitates the unwinding of the mutagenic non-B DNA conformation and thus prevents DNA breakage and mutations at those sites. In addition to this direct effect, helicases are also involved in maintaining epigenetic stability during replication, which is thought to occur via coupling of replication with histone recycling. Sarkies et al. (2010) found that in REV-1-deficient cells, G4 DNA can inhibit processive replication, which uncouples DNA synthesis from histone recycling, resulting in loss of repressive chromatin marks [191]. Thus, in addition to utilizing the RECQ helicase to unwind non-B DNA structures and restart replication, cells can resume replication via REV1-dependent translesion synthesis (see below), and FANCJ appears to coordinate these pathways at G-quadruplex motifs [191].

In contrast, helicases do not always function to maintain genomic stability. For example, deficiency of yeast Sgs1 did not affect hairpin-forming CAG repeat expansions [192], while depletion of Sgs1 reduced the expansion of H-DNA-forming GAA repeats [193]. This is very similar to the roles of DNA topoisomerases on non-B DNA-induced genetic instability; Topo I and II can relax the negative supercoiling on duplex DNA [194], and thereby reduce the formation of non-B DNA and subsequent genetic instability, while the breakage resulting from topoisomerase cleavage on non-B DNA structures [195, 196] can initiate genetic instability. In summary, the effects of helicases on non-B DNA metabolism appear to be structure-specific and are more complicated than simply resolving mutagenic structures to maintain genome stability.

2.5. DNA/RNA polymerases

Repetitive DNA sequences can present a challenge to DNA and RNA polymerases during replication or transcription, resulting in stalling of the polymerases, which can ultimately lead to genetic instability at these sites. For example, T7 RNA polymerase was partially blocked within and downstream of the H-DNA-forming sequence from the human c-MYC promoter region, and within a CG(14) Z-DNA-forming sequence on negatively supercoiled plasmids. It has been speculated that non-B DNA-induced RNA polymerase pausing may stimulate the recruitment and activation of transcription-coupled NER factors at the non-B DNA region [197, 198]. Replicative DNA polymerases delta and epsilon generated small insertions or deletions within CA repeats at frequencies that were 1000-fold higher than in non-repetitive sequences [199]. Replication of single-stranded GGC repeats by DNA polymerases alpha, beta, HIV reverse transcriptase, Taq DNA polymerase, or the Klenow fragment of DNA polymerase I caused expansions of repeat units likely due to misalignment of the repetitive nascent strands and the templates [200]. Some translesion synthesis (TLS) DNA polymerases can replace the replicative polymerases delta and epsilon at forks stalled by non-B DNA structures, and can synthesize through the non-B regions, thus suppressing the replication fork stalling and DNA breakage. For example, depletion of the TLS polymerases eta or kappa in HeLa cells engineered to contain multiple tandem copies of genomic non-B DNA-forming sequences resulted in more DSBs compared to wild-type cells, implicating these TLS polymerases in efficient DNA synthesis through these structure-forming sequences [201]. Rev1 is a member of the Y-family DNA polymerases and functions with polymerase zeta (a B-family DNA polymerase) in the error-prone processing of UV-induced DNA damage [202]. Yeast rev1 mutants were found to have higher rates of both expansions and deletions of CTG repeats, while rev1 deficiency had no impact on the stability of triplet repeats unable to adopt non-B DNA structures [203]. Interestingly, this effect on CTG triplet repeat stability did not require polymerase zeta, an essential Rev1 partner in translesion synthesis of UV-damaged templates, suggesting that Rev1 might interact with other proteins on non-B DNA structure-forming sequences [203].

In summary, it is important to note that the activities of repair-related proteins on non-B DNA structures may represent a double-edged sword in the context of DNA structure-induced genetic stability. Formation of a non-B DNA structure, per se, does not directly lead to breakage and instability, but can impair DNA metabolic processes, e.g., blocking RNA or DNA polymerase progression. On the other hand, some secondary structures are required for efficient transcription, replication, and repair, demonstrating the “yin and yang” of repair and other DNA metabolic processes on DNA structure-induced genetic instability. Helicases can remove the non-B DNA-related structural obstacles to allow for efficient DNA metabolism and thereby reduce non-B DNA-induced genetic instability. Alternatively, the processing and removal of non-B DNA structures by repair proteins (or other enzymes) could lead to DNA nicks or DSBs, giving rise to an increased risk of genetic instability.

3. DNA damage/repair processes and non-B DNA structures

Common features of non-B DNA structures include single-stranded regions and alterations in the Watson-Crick double helix, such that non-B DNA may be hypersusceptible to DNA damage from endogenous and exogenous sources. Moreover, the contributions of repair protein binding to the formation or stabilization of non-B DNA structures in vivo are not clearly understood. Thus, the crosstalk between DNA repair proteins and non-B DNA may not be limited to the processing of the structured DNA by the repair proteins. Because DNA damage and repair processes, and the formation of non-B DNA structures can alter the landscape of the affected DNA in chromosomes, they can influence each other in several ways as we outline below.

3.1. Effect of non-B DNA on DNA damage patterns

DNA damage and mutation are not random processes, but are influenced by many factors, including sequence context, structural features of the DNA, and higher-order chromatin organization. For example, major groove interactions are altered within a triplex structure because the third strand occupies the major groove and forms Hoogsteen hydrogen bonds with the purine-rich strand of the underlying duplex. Moreover, the yields of radiation-induced DNA damage in triplex structures are different from those in duplex B-DNA structures [204], and triplex-duplex junctions are hyperreactive to a variety of agents, including DNA intercalators [205, 206]. Studies to determine the influence of DNA structure on radiation-induced strand breakage demonstrated that the local structure (e.g., a narrow minor groove that gives rise to a low accessibility of H4' and H5'2 and a low probability of H-atom abstraction by OH radicals) affected the likelihood of breakage events [207]. Single-stranded DNA regions are generally more accessible to damaging agents than duplex DNA because the bases are not protected inside the sugar-phosphate backbone. As mentioned above, exposure of single-stranded regions is a common feature of most non-B DNA conformations, and thus they may be preferred targets for DNA damage.

Base extrusion is a general feature of B-DNA to Z-DNA junctions and Z-Z junctions [112, 113, 208, 209]. B-DNA to Z-DNA junctions are hyperreactive to DNA interactive agents such as 4-NQO [210], hydroxylamine, and osmium tetroxide [211]. Guanine bases within Z-DNA are in a syn position and are located over the sugar without protection, and are therefore more sensitive to dimethylsulfate and diethylsulfate modification [211, 212]. In Z-DNA the guanines are more sensitive, and the cytosines less sensitive, to ionizing radiolytic attack than in B-DNA [213].

A guanine in the loop of the hairpin formed at a (CAG)10 repeat was shown to be hypersensitive to modification by peroxynitrite when compared to a guanine in a CAG(10) duplex, and hOGG1 removed 8-oxoG from the loop of hairpin structure at an efficiency that was 700-fold lower than on duplex DNA [214]. G-quadruplex motifs are targets for oxidative damage during hypoxia-induced transcription, resulting in the recruitment of BER enzymes and the formation of transient strand breaks [215]. Ndlebe & Schuster (2006) demonstrated that while charge migration from radical cation reactions with guanines occurred vertically along a DNA chain, it occurred horizontally within a G-quartet [216]. In addition, oxidizing radicals can become trapped within folded quadruplex structures and attack the external tetrads of the quadruplexes with higher efficiency than the center tetrads [217]. These examples emphasize the roles of DNA secondary structures on DNA damage patterns in vitro, suggesting that non-B DNA structure-forming sequences should be considered when analyzing DNA damage patterns in cells.

3.2. Effect of non-B DNA conformations on DNA repair

Genomic DNA is constantly under attack by both endogenous and exogenous DNA damaging agents. Because cell survival depends on genomic stability and the ability to repair DNA damage, exquisitely sensitive DNA repair mechanisms have evolved to process DNA damage. Defects in these repair systems can lead to severe disorders such as xeroderma pigmentosum, Cockayne syndrome, ataxia telangiectasia, Fanconi anemia, Bloom syndrome, and trichothiodystrophy, and an enhanced predisposition to cancer [218]. The consequences of genomic instability are also manifest in several hereditary human neurological disorders that involve the expansion of triplet repeat sequences [14].

The formation of non-B DNA structures introduces helical distortions into the DNA and alters Watson-Crick B-DNA helical parameters, which can influence DNA repair processes. It has been shown that bubble and loop regions in duplex DNA can serve as entry sites for the E. coli NER complex UvrAB, prior to its translocation on the DNA [104, 105]. And the human NER nucleases, ERCC1-XPF and XPG are structure specific in that they incise DNA at pre-initiation bubble structures containing bulky adducts. As discussed above, DNA damage/distortion recognition proteins can bind non-B DNA, but the extent to which non-B DNA (in the absence of DNA damage per se) is subsequently processed by repair proteins is not well understood. The accessibility of repair proteins to damaged sites within non-B DNA regions may be altered compared to those in B-DNA. For example, the removal of N7-methylguanine by DNA glycosylase was shown to be less efficient in Z-DNA than in B-DNA [219, 220], and the repair efficiency of O6-methylguanine in a poly(dG-m5dC).poly(dG-m5dC) sequence in the ZDNA conformation was only 10% of that measured in B-DNA [221]. The hOGG1 protein showed reduced affinity to 8-oxoG when it was located within a hairpin structure, compared to B-DNA, and the repair efficiency was significantly lower [141]. In the striatal tissue of HD mice where CAG repeats are unstable, abasic sites located near the 5' ends of the CAG repeats were repaired with lower efficiency than other locations, such as the 3' ends of the repeats, or within non-repetitive sequences [133]. Further, the interactions of repair proteins with non-B DNA structures may result in altered activity and fidelity of repair. For example, the ATPase activity of the MSH2-MSH3 complex was inhibited, and the MSH2-MSH3-ADP/ATP affinity was altered upon binding to hairpin structures formed at CAG repeats [117, 222].

3.3. DNA damage and repair processes can alter DNA structure

Lesions in the DNA can change the energetics of conformational transitions, which may facilitate or disfavor the formation of non-B DNA structures. For example, psoralen photocrosslinking has been used to map DNA secondary structures (e.g., hairpins and loops) of single-stranded genomes [223], due to its ability to covalently crosslink duplex regions of DNA. Replacement of a guanine with an abasic site in a B-DNA substrate destabilized the duplex, while an abasic site located within the loop of a CAG repeat hairpin structure did not substantially alter the thermodynamic stability of the looped region, and eventually increased the population of looped structures [224]. On the other hand, one can speculate that an interstrand crosslink could inhibit the formation of local secondary DNA structures in double-stranded genomes due to the requirement for strand separation in the generation of most non-B DNA conformations.

Non-B DNA structures exist in high-energy states, and their formation is favored under conditions that generate negative supercoiling and destabilize the B-DNA structure (e.g., DNA replication, transcription, repair). While genomic DNA is largely in the B-form at any given time, certain regions with the capacity to adopt non-B DNA dynamically fluctuate between B-DNA and non-B conformations. In the human genome, DNA is organized and compacted into chromosomes, and is wrapped around histone cores. Interaction of the positively charged arginines on the histone core particle surface with the negatively charged DNA favors a B-DNA conformation, where the arginines match the shape of the minor groove of B-DNA [225]. The left-handed wrapping renders the negative supercoiling torsional stress anonymous, thus favoring a B-DNA conformation in chromatin. DNA repair is coupled to chromatin remodeling to unwrap the DNA from the nucleosome for repair processing, and thereby generates negative supercoiling that can destabilize the duplex structure. The resulting single-stranded DNA is susceptible to conformational transitions, such that DNA repair processing may facilitate the formation of non-B DNA. For example, more expansions occur at CAG repeats when BER processes lesions that are located near the 5' end of the repeats compared to the 3' end of the repeats [133]. We have described this process as “repair stimulated DNA structure formation” model [142].

4. Non-B DNA conformations affect chromosomal DNA stability at a distance

Recently, our group and others have found that non-B DNA conformations can induce mutations (including point mutations, small deletions, or insertions) hundreds of base pairs from the non-B DNA site, while the non-B DNA sequence itself may or may not contain mutations [167, 193]. The locations and spectra of the mutations suggested that they were not directly initiated at the non-B DNA region itself, although they were dependent on the non-B DNA-forming sequences. The mechanisms involved are not known, however, one plausible explanation is that the formation of a non-B DNA structure alters the chromatin structure and/or histone modifications in a large region of adjacent DNA, rendering it more susceptible to damage and mutation. In support of this notion, Ruan & Wang (2008) found that GAA repeats were refractory to nucleosome assembly in vitro when in a triplex structure relative to the same repeats in a duplex conformation [226]. Inter-molecular triplexes have been found to interrupt histone-DNA contacts flanking the triplex structure, and can serve as nucleosome barriers [227]. Triplexes have been reported to form on nucleosome-bound poly(dA).poly(dT) tracts, but they typically form at the ends of nucleosomal DNA fragments [228] with lower stability [229], accompanied by changes in nucleosome structure [230]. Thus, nucleosome formation and triplex formation are thought to be mutually inhibitory processes with the tendency for triplex exclusion from the nucleosome [231].

Despite the inherent CG-richness of the nucleosome core, the Z-DNA-forming CG(9) repeat was shown to be excluded from nucleosomes in S. cerevisiae [232], and methylation of CpG repeats capable of Z-DNA formation in the promoter region of the chicken adult beta-globin gene, prevented DNA from binding to the histone octamer in vitro [233]. In contrast, cytosine methylation of CpG repeats unable to adopt Z-DNA did not affect nucleosome positioning, suggesting that Z-DNA formation was required for nucleosome exclusion of methylated CpG repeats [234]. We have recently discovered that cytosine methylation can induce/stabilize Z-DNA structures at sequences that do not adopt Z-DNA conformations when unmethylated in vitro. And when methylated ZDNA-forming sequences were introduced into mammalian cells, they were excluded from nucleosome assembly in vivo (Wang et al., unpublished data). Interestingly, a short unmethylated CGG repeat that represented an unfavorable site for histone octamer binding converted to a strong nucleosome positioning sequence upon methylation [235]. However, the expanded CGG repeat that was capable of forming a hairpin or G-quadruplex structure was resistant to nucleosome assembly regardless of the methylation status [235], suggesting a direct relationship between DNA secondary structure and nucleosome assembly.

A computer-based analysis of the association between nucleosomes and putative G-quadruplex-forming sequences revealed that they were coincident with nucleosome depleted regions in both human cells and in C. elegans beyond that expected by chance alone [236]. It is not known whether G-quadruplex formation displaces nucleosomes or whether these structure-forming sequences evolved to occupy nucleosome-free regions, but in either case, these findings suggest that these sequences serve a functional role in the genome.

The central AT-rich region of the Ars insulator sequence, a boundary element that can block enhancer-promoter communications, can adopt non-B DNA conformations as evidenced by its sensitivity to diethylpyrocarbonate modification [237]. Moreover, this central AT-rich region was found to be excluded from nucleosomes in DNA isolated from sea urchin embryos [237], implicating non-B DNA in altering nucleosome formation.

Alternatively, some triplet repeat sequences that can adopt hairpin structures have been found to be strong nucleosome positioning elements. Long CTG repeats (75 or 130 repeats) formed stable nucleosomes with calf thymus histone octamers, with nearly 10-fold greater propensity than the 5S RNA gene, a strong natural nucleosome-positioning element [238, 239]. CTG repeats at the 3' noncoding region of the DMPK gene at the Myotonic dystrophy (DM1) locus were found to be centrally located within nucleosomal regions [240]. Moreover, the sequences adjacent to the expanded CTG repeats (over a thousand repeats) from DM1 patients were less accessible for nuclease digestion, suggesting that they were protected in the nucleosome structure [241]. Thus, some non-B DNA conformations may represent strong nucleosome assembly elements and facilitate nucleosome remodeling at adjacent regions.

DNA conformation can also affect epigenetic modifications, such as cytosine methylation and histone acetylation. For example, cytosines within Z-DNA structures are known to be resistant to methyltransferases, resulting in substantially lower levels of methylation compared to the same sequences in B-form DNA [220, 242]. Using a bioinformatics approach, Bock et al. (2006) found that certain genomic characteristics such as DNA sequence and structure, and specific DNA repeats, were highly correlated with CpG island methylation on Chromosome 21 in human lymphocytes. Hence, cytosine methylation in CpG islands tends to occur at the regions that are predicted to form non-B DNA structures [243]. It was also reported that an H-DNA-forming AG repeat increased the level of histone acetylation in the smooth muscle myosin light chain kinase gene promoter [244]. It is not clear that the increased acetylation was caused by a direct interaction between the H-DNA structure and histones (or with histone acetyltransferases or histone deacetylases) or if it was an indirect affect of H-DNA-induced chromatin remodeling.

5. Summary

Non-B DNA conformations interact with a wide variety of DNA processing proteins in cells. The complex interplay of DNA structure and repair can affect each other in both positive and negative ways, demonstrating the “yin and yang” of DNA repair mechanisms in DNA structure-induced genetic instability. For example, the formation of alternatively structured DNA can alter the patterns of DNA damage and the efficiency of repair within that region, while DNA damage and repair can either facilitate or prohibit the formation of non-B DNA conformations. Further, some DNA processing proteins can suppress non-B DNA structure-induced genetic instability, while other proteins in the same pathway might enhance DNA structure-induced mutagenesis. A simplified model of the interactions between DNA damage, repair, and non-B DNA structure-induced mutagenesis is summarized in Figure 2. Based on our current understanding, it is reasonable to postulate that non-B DNA-forming regions may induce genetic instability in part by modulation of epigenetic factors, resulting in inefficient nucleosome assembly surrounding the non-B DNA regions, leaving the DNA more susceptible to damage and mutagenesis. Many questions remain unanswered, for example, what is the range (i.e., how far upstream and/or downstream from a non-B DNA-forming sequence) of influence of non-B DNA on nucleosome assembly or genetic instability? How do non-B DNA-induced histone or DNA modifications influence genetic instability? What proteins recognize and process non-B DNA structures? What is the susceptibility of different non-B DNA-forming sequences to damage induced by various DNA damaging agents, and how efficiently are they repaired? Clearly, more studies are warranted to gain better mechanistic insights into the role of non-B DNA in chromosomal organization and genetic instability both locally and at a distance.

Figure 2. DNA damage/repair and non-B DNA conformation-induced genetic instability.

A repetitive sequence in the B-form is unwound from the histone core during DNA metabolism, e.g., DNA replication, transcription, or repair (step 1), and an altered DNA secondary structure is formed at the repetitive sequence (here shown as a Z-DNA structure). The non-B conformation can alter nucleosome reassembly, and may result in chromatin remodeling in nearby regions (step 2). The altered DNA structure can affect DNA damage and repair activities (step 3), and interfere with DNA replication (step 4). DNA repair proteins can recognize, bind to, and process the non-B DNA conformation, resulting in either melting of the DNA conformation (step 5) or to the formation DSBs eventually leading to deletion and/or translocation events (step 6).