Dynamic alternative DNA structures in biology and disease

Abstract

Repetitive elements in the human genome, once considered ‘junk DNA’, are now known to adopt more than a dozen alternative (that is, non-B) DNA structures, such as self-annealed hairpins, left-handed Z-DNA, three-stranded triplexes (H-DNA) or four-stranded guanine quadruplex structures (G4 DNA). These dynamic conformations can act as functional genomic elements involved in DNA replication and transcription, chromatin organization and genome stability. In addition, recent studies have revealed a role for these alternative structures in triggering error-generating DNA repair processes, thereby actively enabling genome plasticity. As a driving force for genetic variation, non-B DNA structures thus contribute to both disease aetiology and evolution.

Introduction

Sequencing of the human genome revealed that more than 50% is composed of repetitive elements¹. Initially thought of as mere by-products in genetic evolutionary trajectories, we now know that many repetitive sequences have important biological functions, such as the regulation of chromatin structure, gene expression, DNA replication and genomic rearrangement^2,3. A crucial feature of some repetitive sequences is the potential to fold into alternative, non-canonical DNA structures^4,5 that differ from the right-handed DNA double helix, referred to as the canonical B-form or B-DNA structure, described by Watson, Crick, Wilkins and Franklin in 1953. Since then, more than 15 types of DNA structure that differ from canonical B-DNA have been reported^6,7, with an estimated 13% of the human genome containing sequences that support such structures⁸. In addition to the primary sequence, the formation of non-B DNA structures is dictated by many cellular factors such as chromatin structure, DNA negative supercoiling stress and DNA binding proteins. Thus, depending on the conditions, rapid transitions from B-DNA to non-B DNA can occur, making this a highly dynamic process⁹. As a consequence, non-B DNA structures range from small single-stranded loop-outs of a few nucleotides formed by simple tandem repeats¹⁰ to more complex structures such as hairpin or cruciform DNA, Z-DNA, H-DNA and G quadruplexes (G4 DNA), which can contain hundreds of nucleotides (Fig. 1).

Fig. 1 |. Schematic of non-B DNA structures.

a, Canonical B-form DNA. b, Z-DNA forms at alternating purine–pyrimidine sequences, where the syn-formation purines and anti-conformation pyrimidines twist the backbone into a zigzag shape³¹⁰. c, H-DNA forms at polypurine or polypyrimidine sequences that contain a mirror repeat, where half of the repeat in single-stranded form folds back into the major groove of the DNA duplex to form a triplex structure via Hoogsteen hydrogen bonding^311,312. H-DNA can exist in various isomers depending on strand orientation and whether the purine-rich or pyrimidine-rich strand is used as the third strand. d, G quadruplexes form at sequences containing four runs of three or more guanines. Four guanine bases associate through Hoogsteen hydrogen bonding (guanine tetrad), and three continuous guanine tetrads stack to form a G quadruplex (G4 DNA)^88,313. e, Cruciform or hairpin structures form at inverted-repeat sequences^167,314,315, whereby two symmetrical arms self-anneal to form a duplex stem. f, R-loops contain a nascent RNA strand annealed to the DNA template strand³¹⁶, leaving the non-template strand unpaired, which can adopt a stable structure, such as a hairpin or G4 DNA. The red/blue letters in the sequences represent the bases involved in the non-B conformation. RNAP, RNA polymerase; ssDNA, single-stranded DNA.

The co-localization of non-B DNA structures with functional genomic loci and genetic instability hotspots has suggested a role for non-B DNA in important physiological and pathophysiological events, including the regulation of transcription, DNA replication, DNA recombination and genome integrity. For example, non-B DNA structures can not only regulate the initiation of transcription^11–13 and replication^14–16 but also act as impediments to the transcription and replication complexes, leading to replication stalling, template slippage and/or replication fork collapse and DNA breakage¹⁷. Indeed, the dysregulation of DNA replication at non-B DNA structures is a major driving force of repeat expansion events¹⁸, which occur at different stages of development in different cell types and have been associated with human disease¹⁹. Since the initial discoveries connecting expansions of trinucleotide CGG repeats with fragile X syndrome^20,21 and CAG repeat expansion with spinal and bulbar muscular atrophy²² more than 30 years ago, expansions of non-B DNA structure-forming mono-, di-, tri-, tetra-, penta- and hexa-nucleotide repeats have been implicated in more than 50 neurodevelopmental, neuromuscular and neurodegenerative disorders, among many other diseases^23–29. Although the detailed mechanisms by which different repeats (different in both repeat unit and number) are involved in disease-related gene dysregulation and/or genetic instability may differ, expanded long repeats tend to form more stable structures and thus increase the risk of further instability events, a feature referred to as ‘dynamic mutation’³⁰. An important recent discovery in the field is that non-B DNA structures can be recognized by DNA repair proteins, triggering error-generating repair processes, resulting in replication-independent genetic instability and variation^31,32. This structure-specific repair processing mechanism can contribute to the DNA repeat-related mutations that occur in many diseases^33–36.

In this Review, we discuss the key types of non-B DNA structure with a focus on their roles in genetic instability and disease aetiology. We begin by highlighting the dynamic nature of non-B DNA structures and the conditions that favour non-B DNA formation, before reviewing how non-B DNA structures influence cellular processes such as transcription, replication, recombination and DNA damage and repair. We discuss replication-dependent and replication-independent mechanisms of non-B DNA-induced mutagenesis, before concluding with a brief discussion of non-B DNA sequences in human disease.

Dynamic non-B DNA in the human genome

The nucleotide sequence dictates the potential for the formation and stability of a particular non-B DNA structure (Fig. 1). Hence, many sequence-based computer algorithms are now available to search for potential non-B DNA-forming sequences in genomes. Some examples include palindrome, detectIR³⁷, QGRS Mapper³⁸, G4Hunter³⁹ and DNA Structure Search⁴⁰ (see Related links) and others that search for inverted repeats, G4 DNA-forming, H-DNA-forming and Z-DNA-forming sequences^41–44. Recently, deep learning or feature representative machine learning approaches that use large datasets to identify sequence features have been used to search for potential Z-DNA and G4 DNA-forming sequences^45–47. Combining DNA sequence features with other biological factors could provide more accurate prediction power. However, many challenges remain in detecting and characterizing non-B DNA structures in genomes of living cells and organisms. Taking into consideration primary sequence analysis, RNA polymerase II (RNAPII) binding sites, and permanganate and S1 nuclease footprinting data, putative non-B DNA sites have been mapped at high resolution in mammalian genomes. The results showed that the promoter regions of oncogenes contained significantly more non-B DNA structures than other regions, even after excluding the non-B motifs that overlapped with transcription bubbles (RNAPII chromatin immunoprecipitation followed by sequencing (ChIP–seq) peaks)⁴⁸. Although the reason for such enrichment is not clear, plausible explanations include the altered activities at these genes and/or that non-B DNA-induced genetic instability facilitates the formation of oncogenes from proto-oncogenes.

Genomic DNA is nevertheless largely maintained in the B-form, as this is the most energetically stable structure even when the sequences meet the pattern requirements for non-B DNA formation (Fig. 1). The transition from the canonical B-form DNA to a non-B DNA structure requires energy⁴⁹. Therefore, non-B DNA formation is dependent not only on primary sequence features, but also on conditions induced by genomic activities. For example, when a DNA duplex is separated into single strands during transcription, replication or DNA repair processes, the B-DNA to non-B DNA transition can be facilitated by the negative supercoiling and open chromosome structure that occur during these processes (Fig. 2). Thus, the potential for repetitive elements to adopt non-B DNA structures, the type of non-B DNA formed and the location are determined by several factors, including negative supercoiling⁵⁰, the presence of specific binding proteins^51,52, open nucleosome and chromatin conformations⁵³, and intracellular microenvironments such as pH⁵⁴ and salt concentration⁵⁵. For Z-DNA formation in the mouse prefrontal cortex, it was found that negative supercoiling levels and the presence of the Z-DNA-specific binding protein ADAR1 were the most important factors⁹.

Fig. 2 |. Dynamic non-B DNA structure induced by transcription.

The inverted-repeat sequence (blue) is maintained in B-DNA form on histones and is unwrapped during transcription. The progressing transcription machinery unwinds DNA from the nucleosome structure and creates positive supercoiling in front (removed by topoisomerases) and negative supercoiling behind, which facilitates non-B DNA structure formation (shown in the schematic as a cruciform). RNAP, RNA polymerase; ssDNA, single-stranded DNA.

Depending on the nuclear environment, the same sequence has the potential to adopt different structures. For example, the purine-rich strand from an H-DNA-forming sequence can fold back to form a triplex structure at neutral pH in the presence of bivalent cations such as Mg²⁺ (Fig. 1c). However, under acidic conditions, the cytosines can be protonated, and the pyrimidine strand can serve as the third strand in the triplex structure^56,57. As another example, high concentrations of aluminium maltolate can convert a CCG(12) repeat sequence in the FMR1 gene — involved in fragile X syndrome — into Z-DNA, as evidenced by circular dichroism spectra analyses⁵⁸. Molecular dynamics simulations also predicted a stable Z-DNA structure at CCG repeats with alternately extruded Gs that favour syn conformations followed by symmetrically extruded junctions between adjacent Z-DNA formations⁵⁹ (Fig. 1b). However, a similar CCG repeat was found to adopt a G-quadruplex structure in the presence of high concentrations of sodium chloride⁶⁰ (Fig. 1d). Moreover, a recent single-molecule study found that a TG(11) duplex opened an unpaired bubble under low stretching tension and unwinding torsion. However, with increased negative supercoiling tension, the TG(11) repeat formed a Z-DNA structure⁶¹. These examples show that different environments can support the formation of different non-B DNA structures at the same or similar sequences. Thus, cellular and genomic metabolic conditions should be considered when studying DNA structure-induced cellular activities.

Chromatin conformation

DNA–histone interactions maintain B-DNA formation, such that an important initiating step for non-B DNA formation is nucleosome disassembly during DNA metabolic processes. A study using Fourier transform infrared spectroscopy found that histone acetylation could lead to open chromatin structures and a concomitant increase in Z-DNA formation in trichostatin A-treated HeLa cells⁶². Using the same technology, and confirmed by ChIP, the authors found that Z-DNA formation was increased in cells that overexpressed BRG1 or BRM ATPases, key components of the mammalian SWI/SNF chromatin remodelling complex. As expected, Z-DNA formation was reduced when BRG1 and BRM were depleted by short interfering RNA⁶³.

Negative supercoiling

Genomic DNA is wrapped on histone cores in a left-handed toroidal manner, and the helical tension in the linker regions is released by topoisomerases. Thus, unless the DNA has been unwrapped from histones to allow for negative supercoiling, the energy to facilitate non-B DNA structure formation may not be available. A recent study mapped DNA supercoiling regions and non-B DNA structures in the genome of Caenorhabditis elegans embryos and found that 400 bp regions around transcription start sites had significantly increased negative supercoiling and non-B formation (Z-DNA and cruciform structures)⁶⁴. Supercoiling induced by transcription through the MYC gene can stimulate Z-DNA formation in the promoter region, as shown by Z-DNA-specific antibody binding in permeabilized mammalian cell nuclei^65–67. Approximately 1.5 kb upstream of the MYC promoter P1, a region known as the far upstream element (FUSE), lies a well-characterized supercoiling responsive region that can adopt non-B DNA structures when the gene is expressed⁶⁸. Using an immunofluorescence labelling method, cruciform structures that had strong levels of negative supercoiling induced by active transcription were detected in growing mouse oocytes, but not in fully matured oocytes where transcription was not active⁶⁹. Furthermore, the inhibition of transcription with α-amanitin treatment in growing oocytes significantly reduced cruciform DNA foci. When mouse genomic DNA was fragmented, circularized by ligase and negatively supercoiled at a near-physiological level, the single-stranded DNA (ssDNA) regions exposed by non-B DNA structures on the artificially supercoiled naked circular DNA resembled the ssDNA regions detected in vivo, demonstrating the contribution of negative supercoiling to non-B DNA conformation in vivo⁴⁸.

Non-B DNA binding proteins

Various proteins have been identified that bind to non-B DNA and alter their stability. A family of Z-DNA binding proteins that share a common Z-DNA binding domain (ZDBD) have been described⁷⁰, such as Z-DNA binding protein 1 (ZBP1), a pathogen-sensing protein that regulates cell death and inflammation^71,72; PKZ, a PKR-like protein kinase that has a role in host responses to viruses⁷³; and poxvirus virulence factor E3L and ORF112, which are crucial for viral pathogenesis^74,75. The ZDBD of the ADAR1 protein binds to Z-DNA with high affinity and can convert even a short TA(3) repeat into a Z-DNA structure, which cannot form in the absence of ADAR1 (ref.⁵²). Many chromosomal architectural proteins such as histones H1 and H5 and the high mobility group (HMG) proteins bind preferentially to cruciform structures (reviewed in ref.⁷⁶), and the HMGB1 protein has high affinity for triplex DNA structures in collaboration with the nucleotide excision repair (NER) protein complex XPA–RPA⁷⁷. Many proteins are known to interact with G4 DNA, such as POT1, RPA and BRCA1 (reviewed in ref.⁷⁸). A recent study that used a cell-permeable G4 DNA ligand and crosslinking of the G4 DNA-interacting proteins identified hundreds of putative G4 DNA-associated proteins with known functions in transcription regulation, mRNA processing, cell cycle regulation and DNA damage and repair processes⁵³. The presence and the local concentrations of these proteins in vivo could modulate the formation and activities of non-B DNA structures.

DNA helicases

DNA helicases are responsible for unwinding duplex DNA during replication, transcription and repair and have a key role in genome maintenance⁷⁹. Helicases have gained increasing attention recently regarding their roles in non-B DNA and disease (Table 1). In general, helicases have the capacity to unwind non-B DNA structures, such that their deficiency can increase the risk of non-B DNA-related diseases, but this conclusion should not be oversimplified. For example, deficiency of yeast Sgs1 helicase, a homologue of the bacterial RecQ helicase, resulted in an accumulation of cruciform-shaped replication intermediates⁸⁰. However, Sgs1 deficiency reduced the expansion of GAA repeats in yeast⁸¹. It was hypothesized that Sgs1 unwinds nascent strands from their templates when replication forks are stalled by non-B DNA structures, with the shorter Okazaki fragments annealing to the longer nascent leading strand, resulting in extra repeat units (that is, repeat expansion)⁸¹. Thus, the effects of helicases on non-B DNA processing seem to be more complicated than simply resolving structures to maintain genome stability.

Table 1 |.

DNA helicases and non-B DNA structures

Helicase	Cellular function	Reported types of non-B DNA processed	Results from helicase deficiency	Ref.
Superfamily 1 DNA helicases
PIF1	Unwinding replication barriers, assisting fork progression	G4	Genetic instability, increased risk of cancer	289
RRM3	Unwinding replication barriers, assisting fork progression	G4	Genetic instability, increased risk of cancer	290
DNA2	Telomere maintenance; helicase and G4 nuclease	G4	Cell senescence, telomere replication defects, genetic instability, increased risk of cancer	148
Srs2	Post-replication repair to UV, ionizing radiation or MMS lesions	Cruciform at inverted repeats, hairpin at triplet repeats	UV sensitivity, genetic instability	291
UvrD	Nucleotide excision repair, mismatch repair, HR	G4, cruciform/Holliday junctions, triplet repeats	UV sensitivity, genetic instability	292
Rep	DNA replication	G4	Slow progression of chromosomal replication forks	293
RecBCD	Helicase and nuclease activities, DSB repair by HR	Cruciform at inverted repeat, hairpin at triplet repeat	Genetic instability	294
Superfamily 2 DNA helicases
RECQ1 (also known as RECQL and RECQL1)	DNA repair, cell cycle and growth, telomere maintenance, transcription	Cruciform/Holliday junctions, very weak on G4	Genetic instability and increased risk of cancer	295
RECQ2 (also known as BLM)	DNA replication, immunoglobulin class-switch recombination	G4, triplex H-DNA, direct repeats including triplet repeats	Hereditary Bloom syndrome: primordial dwarfism, genetic instability, increased risk of cancer	296
Him-6	BLM homologue in Caenorhabditis elegans	G4	Genetic instability	297
RECQ3 (also known as WRN)	Telomere maintenance	G4, hairpin at triplet repeat, triplex H-DNA, Z-DNA	Hereditary Werner syndrome, premature ageing, increased risk of cancer, cell senescence	298
Sgs1	Yeast homologue of human BLM and WRN. Forms a complex with Top3 and Rmi1. DNA replication, regulating HR	Cruciform/palindrome, G4, hairpin at triplet repeat	Genetic instability	291
RECQ4 (also known as RTS)	ATPase activity and single-strand annealing activity, replisome assembly	Holliday junctions, G4	Hereditary Rothmund–Thomson, RAPADILINO and Baller–Gerold syndromes: skin, hair, skeletal and dental abnormalities, increased risk of cancer	299
RECQ5	DNA replication, transcription, repair, suppressing sister chromatid exchanges during HR	G4, an order of magnitude weaker than BLM and WRN	Genetic instability	300
RTEL1	Telomere maintenance and HR regulation	G4, hairpin at triplet repeat	Hoyeraal–Hreidarsson syndrome, pulmonary fibrosis and/or bone marrow failure, telomere-related 3 (PFBMFT3), dyskeratosis congenita, autosomal recessive 5 (DKCB5)	301
DHX36	RNA and DNA helicase activity, transcription and translation regulation, genetic stability, telomere maintenance	G4	Genetic instability and increased risk of cancer	302
DHX9	DNA replication, transcription, translation, microRNA biogenesis, genetic stability	Triplex H-DNA, Z-DNA	Genetic instability and increased risk of cancer	303
FANCJ	DNA repair, HR, replication fork progress during replication stress	G4, hairpin at triplet repeat	Hereditary breast and ovarian cancer, Fanconi anaemia, bone marrow failure	304
Dog-1	FANCJ homologue in C. elegans	G4	Genetic instability	305
XPB/XPD	TFIIH components with functions in nucleotide excision repair	G4, triplex H-DNA	Xeroderma pigmentosum (UV sensitivity, cancer); trichothiodystrophy, Cockayne syndrome (development)	306
DDX5	DNA and RNA helicase activity, transcriptional regulation, splicing	G4	Aberrantly expressed in many tumours	307
DDX11 (also known as CHLR1), CHL1 in yeast	Chromosome segregation, cell cycle progression, sister chromatid cohesion, putative RNA helicase, translation initiation, splicing	G4, triplex H-DNA	Warsaw breakage syndrome	308
ATRX	Chromosome alignment and meiotic spindle organization, recombination pathway selection	G4	α-Thalassaemia with mental retardation	309

DSB, double-strand break; HR, homologous recombination; MMS, methyl methanesulfonate.

Non-B DNA induced by DNA damage or repair

DNA damage and repair can affect non-B DNA structure formation by altering local topological conditions. Some types of DNA lesion or repair intermediate can alter the energetics of structural transitions of DNA, affect protein–DNA interactions and modulate nucleosome and chromosome conformations, eventually leading to non-B DNA structure formation. For example, DNA double-strand break (DSB) processing near short inverted repeats can stimulate the formation of hairpin structures, likely by creating ssDNA that enables self-annealing⁸².

Abasic (AP) sites generated during the repair of 8-oxo-7,8-dihydroguanine (8-oxo-G) by OGG1 can destabilize duplex DNA and provide thermodynamic energy for the transition from G-rich duplex DNA to a more stable G4 DNA structure⁸³, or a cruciform structure when the processing occurs within an inverted-repeat region⁸⁴. The distribution of AP sites, OGG1 and AP endonuclease 1 (APE1) binding sites in lung cancer genomic DNA determined by ChIP–seq exhibited genome-wide correlation with G4-forming motifs, particularly in promoter and gene regulatory regions⁸⁵. Furthermore, binding of APE1 to AP sites within G4-forming motifs in the MYC promoter stimulated the formation of G4 DNA in vitro⁸⁵. Another study reported that an AP site located in the centre spacer region between two symmetrical arms of an inverted repeat could destabilize B-DNA formation and increase the formation of hairpin structures⁸⁶.

A tetrahydrofuran abasic site analogue within a GAA repeat was processed by base excision repair, during which the repeats on the template strand could form a loop of approximately eight TTC repeat units in vitro, which recruited DNA polymerase-β (Polβ) for bypass, resulting in large deletions⁸⁷.

In summary, non-B DNA conformations are highly dynamic in living cells. Although the primary sequence is crucial for structural transitions, other cellular activities such as transcription, replication and DNA repair can have an impact on non-B DNA formation via the modulation of nucleosome or chromatin structures, alterations in DNA supercoiling levels and/or DNA binding proteins.

Biological functions of non-B DNA

Repetitive sequences capable of adopting non-B DNA structures are enriched at highly conserved regions with biological functions, such as promoters and replication origins. Although co-localization does not necessarily relate to function of non-B DNA in these processes, numerous studies have suggested that non-B DNA structures can contribute to several important biological functions^88,89. Non-B DNA structure formation can change the local topology of genomic regions, thereby influencing interactions of DNA metabolic processes, protein binding and chromatin structures. The important biological roles of non-B DNA-forming sequences, perhaps leading to positive selection pressure during evolution, may explain the abundance and conservation of these unstable elements across genomes (Box 1).

Box 1.

Non-B DNA as a driver of evolution

Non-B DNA-forming sequences have been found to co-localize with evolutionarily active regions^3,318. Transposable elements comprise a large fraction of many eukaryotic genomes and most contain terminal inverted repeats³¹⁹ and direct repeats that can stimulate double-strand breaks (DSBs). A comparative autosomal map covering >90% of the mouse and human genomes revealed that the breakpoint regions of intrachromosomal rearrangements contained a high density of repetitive sequences³²⁰. Although it is not easy to distinguish which occurred first, a mutagenic non-B DNA-forming sequence or an active evolution hotspot over time, unstable non-B DNA is considered as a driving force for genetic variations and evolution^321–323.

Insects contain a xenobiotic-metabolizing P450 gene that can detoxify xenobiotics, and its expression is regulated by a G4 motif in the promoter region that is thought to be acquired from an HzIS1–3 transposon³²⁴. Bacterial transposon Tn7, which encodes a TnsC protein that can bind to triple-helical DNA, created selective insertion of Tn7 adjacent to a H-DNA-forming sequence in an in vitro transposition assay³²⁵. It will be interesting to see whether similar H-DNA-directed transposon mobility and genome evolution also occurs in mammalian genomes.

The genome of muntjac deer has undergone drastic evolutionary changes with a dramatic reduction in the number of chromosomes from 2n = 70 in the ancestral karyotype to 2n = 6 in female and 7 in male Muntiacus muntjak vaginalis. Analysis of the fusion sites revealed repetitive elements that may have stimulated DSB formation and mediated recurrent fusions between different chromosomes³²⁶. Non-B DNA structures formed in simple sequence repeats (SSRs), including microsatellites, also stimulate replication slippage, crossover and/or gene conversion events³²⁷. SSRs are very abundant in penaeid shrimp and the distribution is highly associated with transposable element expansion and intrachromosomal rearrangements³²⁸. Compared with Fenneropenaeus chinensis, which lives exclusively in salt water, Litopenaeus vannamei, which is capable of surviving in a large range of salinities, showed significant alterations of SSRs within introns or untranslated regions (UTRs) of differentially expressed genes related to amino acid and lipid metabolism involved in osmoregulation, suggesting a regulatory role of these repetitive elements in adaptive evolution in these species³²⁸.

A more direct connection between non-B DNA structure and evolution comes from comparing marine stickleback fish, which have developed a robust pelvic apparatus, and many independently derived freshwater populations that have adaptively lost pelvic hind fins over the past ~15,000 years. This repeated pelvic loss maps to recurrent deletions of a pelvic enhancer (Pel) of the homeodomain transcription factor gene Pitx1 (ref.³²⁹). The Pel sequence from marine populations contains a long Z-DNA-forming GT repeat that was shown to stimulate the formation of DSBs and large deletions (>100 bp) in a repeat length- and orientation-dependent fashion in yeast and on mutation reporters in mammalian cells mirroring the situation in stickleback fish. Similar repeats in human genomes were also mapped with aphidicolin-sensitive breakage sites, suggesting that non-B DNA structure-induced genetic instability is a common process that has contributed to genetic evolution³³⁰.

The male-specific region of the human Y chromosome occupies ~95% of the chromosome, and eight nearly identical palindromic sequences, the result of duplication events, contain many of the testis-specific genes. The variations on palindromic sequences in existing human populations suggest frequent recurrent arm-to-arm gene conversion events in testis gene families³³¹.

Notably, non-B DNA-forming sequences are not always deleterious. Many non-B DNA-forming sequences are associated with distinct genomic features that are evolutionarily conserved, such as regulatory elements in promoters^5,88. Motif-containing elements for the formation of G4 DNA, triplexes and hairpins increased rapidly in eumetazoan genomes during evolution and seemed to be under positive selective pressure, suggesting that the conservation of non-B DNA-forming sequences may be beneficial during evolution⁸⁹.

Non-B DNA in chromatin organization

Eukaryotic DNA is packaged into nucleosomes and then higher tertiary DNA structures in vivo. When B-DNA is wrapped around histone cores, the minor groove of the helix aligns and interacts with the positively charged arginines on the histones⁹⁰. These electrostatic interactions are important for maintaining nucleosome structure and B-form DNA. Non-B DNA conformations change the orientation of DNA strands and the shape of the grooves and interrupt the DNA–histone arrangement and therefore alter nucleosome structures.

Non-B DNA affects local chromatin organization.

Some repetitive satellites serve as signalling sequences for nucleosome assembly, whereas other repeats, such as Z-DNA-forming CG or CGG repeats⁵⁹ or H-DNA-forming GA repeats, are resistant to placement within the nucleosome structure⁹¹ (reviewed in ref.⁹²). GAA repeats were more refractory to nucleosome assembly in supercoiled plasmid DNA when H-DNA formation was supported, yet the same repeats can be packaged into nucleosomes when in B-form DNA⁹³. In Saccharomyces cerevisiae, most short inverted-repeat sequences were found in regions with low nucleosome occupancy⁹⁴. G4 DNA sequences are enriched in nucleosome-depleted regions in both human cells and C. elegans⁹⁵. Furthermore, a G4 DNA-stabilizing ligand created open nucleosome structures for RNAPII binding even in compacted chromatin regions⁹⁶.

There are reports that suggest that non-B DNA-forming repeats facilitate the formation of nucleosome structures. CTG repeats (as short as six repeat units) are enough to facilitate nucleosome assembly, although expansion to 62 repeats — considered to be more prone to forming a hairpin structure — did not further affect the assembly of histone octamers⁹⁷. Note that most studies were performed in vitro and details on the DNA conformation adopted at repetitive non-B-forming sequences are lacking owing to the difficulties in determining nucleosome and non-B DNA structures simultaneously.

Non-B DNA may affect distal chromatin organization.

Disrupting the formation of a key nucleosome in the β-globin gene by altering the underlying positioning sequence was shown to affect adjacent nucleosomes⁹⁸. Although this has not been verified as a universal mechanism throughout the entire genome in all species, this discovery suggests the interesting possibility that changing the position of one nucleosome by non-B DNA structure formation could affect distant regions.

G4 DNA motifs are significantly enriched at distal interchromosomal interaction sites⁹⁹ and can recruit the chromosomal architectural protein RIF1, which brings multiple G4 DNA-forming sequences together at different regions to create local chromosomal compartments via chromatin looping at the nuclear lamina¹⁰⁰. The ssDNAs from the tips of hairpins can interact with each other to form ‘kissing’ complexes, similar to the NMR solution structure of a kissing complex formed between deoxyoligoribonucleotides corresponding to the dimerization initiation site SL1 of HIV-1Lai RNA¹⁰¹. This loop–loop interaction is important for tertiary and topology structure maintenance and provides a basis for molecular recognition¹⁰². Some long potential G4 DNA-forming sequences that contain multiple G4 elements were identified in antibody switch regions, where the single-stranded loops of neighbouring G4 structures were frequently complementary and base paired with each other, perhaps contributing to chromosomal rearrangements in cancer¹⁰³. In addition, ssDNA regions from two H-DNA structures formed at long GAA repeats on plasmids were shown to interact to form a dumbbell-shaped complex referred to as ‘sticky’ DNA in bacterial cells^104,105. Thus, non-B DNA-mediated interactions may bring distal elements together, contributing to 3D genome organization and stimulating crosstalk between chromosomal territories and DNA elements that regulate gene or chromosomal functions.

Taken together, the formation of both non-B DNA structures and nucleosome structures is dynamic, and often competitive. DNA in nucleosomes is typically maintained in the B-form structure by DNA–histone interactions, and non-B DNA structures, once formed, are often more refractory to nucleosome assembly than B-DNA. The impact of non-B DNA on chromosomal structure, particularly on long-range chromosomal architecture, may have important biological and pathological functions that remain to be discovered.

The impact of non-B DNA on transcription

Non-B DNA affects transcription initiation.

A recent bioinformatics study investigated the distribution of non-B DNA-forming sequences in 15 species and found that promoter regions contain a unique pattern of non-B DNA positioning: G4 DNA and Z-DNA are the most enriched of the non-B types and are frequently found in core promoter regions in nearly all species¹⁰⁶. Direct repeats are enriched in the immediate (50–100 bp) upstream region of core promoters, and mirror repeats are often located (100–300 bp) upstream of the core. It will be interesting to determine whether these repeat patterns underlie as yet unknown mechanisms of transcription regulation.

In the yeast genome, G4 DNA-forming sequences are enriched at promoters approximately sixfold over random distribution¹⁰⁷. Inverted repeats, which can form cruciform structures (Fig. 1), are substantially enriched in regions adjacent to stop codons, at the end of genes, near start codons, 5′-untranslated regions (UTRs) and promoter regions¹⁰⁸. Their conservation and enrichment within or surrounding these key elements suggests a role for non-B DNA in transcription regulation. In promoters, non-B DNA formation could provide an open chromatin structure for transcription initiation complex formation or block or recruit transcription factors and thus affect transcription initiation¹⁰⁹ (Fig. 3).

Fig. 3 |. Biological functions of non-B DNA.

a, Non-B DNA can facilitate the initiation of transcription and replication. Non-B DNA formation (shown in the schematic as G quadruplexes (G4 DNA)) unwinds DNA from nucleosomes and creates an open structure that facilitates the assembly of transcription (left) and replication (right) complexes. b, Non-B DNA can stimulate homologous recombination (HR). There are multiple pathways by which non-B DNA can directly or indirectly stimulate HR. Shown in the schematic is a unique structural alteration between two H-DNA isomers containing complementary single-stranded DNA (ssDNA) regions. With the presence of a nick on either strand or with the assistance of a topoisomerase, the two strands could wind around each other to form Watson–Crick base pairs. Owing to the dynamic nature of H-DNA in vivo, the third strand in both H-DNA structures could disassociate from the duplex and anneal to each other to form a double Holliday junction structure and thereby stimulate HR³¹⁷. RNAP, RNA polymerase; TBP, TATA-box-binding protein.

Although Z-DNA-forming sequences are essential for transcription initiation in viruses, a Z-DNA-forming sequence in the promoter region of the rat Ncl gene, which encodes nucleolin, has been shown to inhibit promoter activity, such that its deletion increased transcription by ~50%²⁵. Interestingly, in this study the effect of Z-DNA on promoter activity was neither location nor orientation dependent; relocating the repeat 458 bp from the promoter or cloning it in the opposite direction did not change its inhibitory effects²⁵. A plausible explanation is that local negative supercoiling stress is essential for transcription initiation because it not only initiates melting of the DNA duplex, which is an energy-consuming step^110,111, but also facilitates the interaction of transcription factors with promoters¹¹². However, formation of non-B DNA also requires negative supercoiling and relaxes the local supercoiling level once formed. One left-handed helical turn of Z-DNA can relieve 1.8 helical turns of negative superhelical twisting on the B-DNA helix¹¹³. Therefore, non-B DNA formation could affect gene expression by altering the local DNA topological tension, independently of its (short-range) location and orientation.

G4 DNA has been shown to either enhance or suppress gene expression (reviewed in ref.¹¹⁴). Many of the published reports on this topic were based on the overlap of computational predictions of G4 DNA-forming sequences and gene functional analyses. A recent study that used an antibody-based G4 ChIP–seq approach identified ~10,000 G4 DNA structures in human chromatin, predominantly in the promoters and 5′-UTRs of highly transcribed genes¹¹⁵. However, these data reveal only a correlation between non-B DNA and transcription regulation rather than providing direct evidence for cause and effect. In another study, G4 DNA-stabilizing ligands were used to observe transcriptional alterations associated with increased G4 DNA structures in human cells. A database of transcriptome alterations induced by seven distinct G4 DNA ligands, including 25,228 genes, was recently published¹¹⁶. Again, although such correlations are informative, the alterations in transcription may be due to other cellular responses induced by the ligands and not necessarily a direct consequence of G4 DNA formation. Thus, studies that demonstrate direct evidence of DNA structure-associated transcription regulation are warranted. However, given the dynamic nature of non-B DNA structures, it is challenging to determine DNA structure in living cells in real time; studies that combine bioinformatic approaches, small-molecule compounds that regulate non-B DNA conformation, and genetic and molecular studies with engineered non-B DNA conformations in the same model systems can provide convincing evidence for the effects of non-B DNA in vivo.

Non-B DNA affects transcription elongation.

Transcription complexes use DNA helicases and transcription elongation factors to unwind or bypass barriers during elongation, including those imposed by non-B DNA. However, if a non-B DNA conformation is stable enough to resist helicase activity, or is stabilized by structure-specific proteins, it can act as a barrier to transcription and reduce the processivity and fidelity of RNA polymerases^117–119. In yeast, Spt4/5, individually or cooperatively with Elf1, interacted with RNAPII to facilitate transcription elongation and increased the run-off transcripts through CTG(40) repeats in the B-form in an in vitro assay; however, when the CTG repeat formed a stem–loop structure, the presence of Spt4/5 in fact enhanced transcription pausing in front of the stem–loop¹²⁰.

Interestingly, RNA polymerases not only pause in front of non-B DNA conformations during extension but can also stall after passing through the non-B DNA-forming sequences. For example, transcription by T7 RNA polymerase or RNAPII was paused at H-DNA-, Z-DNA- and G4 DNA-forming sequences within and downstream of the non-B DNA sequences in a length- and supercoiling-dependent manner in in vitro multiple-round transcription assays^117,118,121. It is plausible that the negative supercoiling generated behind the progressing polymerase, the non-template ssDNA and/or the nascent RNA stimulated a structural complex in this area that impeded the progression of the RNA polymerase complexes¹²¹.

In contrast to H-DNA, Z-DNA or cruciform structures, in which both strands are involved in the conformation, G4 DNA forms on the G-rich strand of a duplex. Therefore, the location of G4 DNA on the template versus the non-template strand during transcription results in different effects. For example, G4 DNA in the template strand upstream of the start codon in the Renilla luciferase gene substantially inhibited transcription, but showed no effect when located in the non-template strand¹²².

Non-B DNA affects RNA splicing.

A genome-wide screen and statistical analyses suggested strong associations between non-B DNA structures, including G4 DNA, cruciform DNA, triplex DNA, slipped DNA and Z-DNA, and exon skipping in both human and mouse genomes¹²³. Although the mechanisms are still unclear, it is possible that non-B DNA-induced stalling of RNA polymerase complexes and the delay in elongation could facilitate the recruitment of splicing factors and the recognition of splice sites. It has also been proposed that non-B DNA on the template strand enables transcription initiation from non-contiguous regions, producing alternative RNA isomers^124,125. However, to date, the experimental evidence is scarce; therefore, further studies on the roles of non-B DNA structures in regulating alternative splicing are warranted and would provide timely and important advances for the field.

Although many details still remain to be elucidated, it is clear that the effects of non-B DNA on transcription are associated with functional genomic regions and higher-order chromosome conformations, such that they cannot be considered simply as ‘activators’ or ‘repressors’¹²⁶. With such a wide variety of effects, the specific manipulation of DNA structure formation could serve as a unique type of ‘epigenetic’ regulation of gene expression and subsequent cellular activities¹²⁷.

The impact of non-B DNA on replication

Non-B DNA affects replication initiation.

Non-B DNA-forming sequences have emerged as key controllers of genome replication at the level of both replication origin firing and replication fork progression. Genome-wide studies have revealed significant enrichment of non-B DNA-forming sequences, including G4 DNA, H-DNA, Z-DNA and cruciform- or hairpin-forming inverted repeats, at replication origins^14–16. These sequences are important for prokaryotic DNA replication¹²⁸, viral replication in eukaryotic cells^129,130 and the replication of eukaryotic genomes^131,132. For example, it was recently found that direct, inverted and mirror repeats as well as Z-DNA- and G4 DNA-forming sequences are associated with origins of replication that are consistent with the position and firing of the origins. However, the regulation of replication origins is complex and involves many different factors. Interestingly, many of these factors, such as base composition, DNA accessibility and chromatin structure, are often overlain by or cluster with non-B DNA-forming sequences in human genomes¹³³. Formation of non-B DNA structures at or near replication origins can alter local topological conditions that affect melting of the DNA duplex and loading of replication factors, and can recruit structure-specific binding proteins for replication machinery assembly^134,135 (Fig. 3). A G4 DNA-forming sequence from the β^A origin in the chicken DT40 cell line was able to initiate replication when it was cloned into a region that lacked strong initiation sites, and a G>A point mutation that destabilized the G4 DNA structure reduced origin efficiency¹³⁶. Cruciform-binding proteins that belong to the 14-3-3 protein family form dimers and bind to cruciform structures at the four-way junctions. Deleting the DNA binding domain reduced the cruciform-binding activity and suppressed the replication of plasmids that contained a yeast replication origin in yeast cells¹³⁷.

The human origin recognition complex (ORC) binds preferentially to replication origins at G-rich ssDNA that can form G quadruplex structures to facilitate the initiation of DNA replication in eukaryotic cells¹³⁸. A recent study directly explored the functions of a G4 DNA-forming sequence from an origin G-rich repeated element (OGRE) on different types of replication origin¹³⁹. Deletion of the OGRE-G4 sequence substantially decreased the corresponding origin activity, whereas adding the OGRE-G4 element created a new replication origin. G4 DNA stabilizer binding to G4 DNA in intergenic regions tended to activate new origins or enhance existing origin activities. By contrast, G4 DNA ligand binding reduced firing efficiency of origins that were clustered and located in active promoters, likely owing to the G4 DNA-induced reduction in transcription, thereby attenuating the stimulating effects of transcription on replication origin firing¹³⁹. Thus, non-B DNA can affect replication initiation even when located hundreds of base pairs from the initiation sites.

Non-B DNA affects replication elongation.

After initiation and priming, replication forks progress, and DNA polymerases act on both the leading and lagging strands. Non-B DNA on the template strand can impose a barrier for many DNA polymerases^{35,128,140,141}, which can reduce their fidelity, stall replication and cause replication fork collapse, resulting in DNA strand breaks^142,143. Topoisomerases and helicases are actively involved in replication and can unwind some non-B DNA structures in front of replication forks^144–146. The template DNA enters and is pulled through a tunnel formed at the zinc finger region of the N-tier ring and C-tier motor domain of human replicative helicase complexes. DNA in a non-B conformation is generally much bulkier than the tunnel and cannot pass through¹⁴⁷. Although the detailed thermal energy characteristics of helicase activities in the context of non-B DNA structures are not fully determined, at least some types of non-B DNA can resist helicase unwinding^148,149.

Direct evidence of replication fork stalling at non-B DNA structures in vivo is challenging because both the non-B DNA structure and replication stalling are transient, and a single paused replication fork at a specific non-B DNA region is difficult to distinguish from a normal progressing fork. 2D gel electrophoresis has been used to successfully determine replication stalling at specific non-B regions^40,150. However, it requires a substantial number of forks to stall at the same location simultaneously, making this technology more appropriate for bacterial and/or yeast genomes. DNA fibre analysis has also been used to monitor replication rates; however, this technique detects large regions and therefore must be combined with a targeted technique, such as fluorescence in situ hybridization (FISH), to identify non-B DNA regions in the genome¹⁵¹.

The impact of non-B DNA on recombination

Non-B DNA-forming sequences are enriched at recombination hotspots, implying a link between non-B DNA and homologous recombination. For example, a 1,000 bp motif in the I–B and I–J subregions within the mouse E_β gene in the major histocompatibility complex (MHC) contains several types of repeat such as AGGC and GC/GT-rich repeats that can adopt non-B DNA structures, including Z-DNA. This short region has been estimated to account for ~2% of the recombination events in the entire genome¹⁵². Unequal sister chromatid exchange between the Cγ2a and Cγ2b heavy chain genes in MPC-11 mouse myeloma cells occurs at a GA repeat that can adopt an H-DNA structure, followed by a GT repeat that can form Z-DNA¹⁵³. G4 DNA is abundant in immunoglobulin switch (S) regions and may contribute to class-switch recombination (CSR) by stalling transcription, leading to the nicking of DNA strands¹⁵⁴. In human bladder carcinoma EJ cells, a Z-DNA-forming GT(30) repeat was found to increase recombination between non-replicating plasmids up to 20-fold. Both gene conversion and reciprocal exchange events were found 237–1,269 bp from the Z-DNA-forming sequence¹⁵⁵.

Certainly, there are different mechanisms involved in non-B DNA-induced recombination. Some non-B DNA structures are composed of self-folding formations on one strand and create long ssDNA on the complementary strand, such as H-DNA and G4 DNA structures. The exposed ssDNA could potentially invade a homologous duplex and form a structure similar to a D-loop, which is known to induce homologous recombination¹⁵⁶ (Fig. 3). Guanosines in Z-DNA are in the syn position and are exposed, and the N7 and C8 of guanosines in Z-DNA are ‘stickier’ and can interact with other DNA molecules¹⁵⁷. The left-handed helix of Z-DNA has been shown to facilitate the formation of paranemic joints during synapsis between two topo domains containing homologous sequences¹⁵⁸. In addition, non-B DNA structures can stall DNA replication forks and generate DNA nicks and breaks that could stimulate recombination.

A distinct mechanism of non-B DNA-stimulated recombination in immune cells includes the AID protein, which belongs to the APOBEC cytidine deaminase family, and involves G4 DNA structures on the non-template DNA strand of the Sμ and Sγ regions. AID is expressed in B cells and can deaminate deoxycytidine, converting it into deoxyuridine within single-stranded regions¹⁵⁹. As AID targets ssDNA, many different types of non-B DNA structure that expose ssDNA regions could potentially serve as targets. In mice, MSH2–MSH6 can bind to both G4 DNA formed within transcribed S regions and G–U mismatches to facilitate DNA synapsis and recombination¹⁶⁰.

Interestingly, when G4 DNA-forming sequences in human or mouse S regions were replaced with chicken or Xenopus laevis Sμ sequences, which are rich in palindromic and stem–loop structures, CSR was still functional in murine B lymphoma cells although it was less efficient. The Xenopus Sμ sequence is in fact AT-rich, rather than G-rich, and it supported CSR in an orientation-independent manner. By contrast, a non-palindromic G-rich sequence was not able to activate CSR¹⁶¹. These results suggest that common features of non-B DNA structures, such as exposure of ssDNA or transcription modulation, rather than the primary sequences, are important for modulating recombination and immune reactions, providing a potential therapeutic target to manipulate CSR and immunoglobulin secretion levels in autoimmune or immunoallergic disorders.

Non-B DNA-induced mutations

Many non-B DNA-forming sequences have been shown to stimulate genetic instability in various organisms^162,163. DNA replication has long been considered a major process implicated in non-B DNA-induced mutagenesis^10,28. Hairpin-forming triplet repeats are often more mutagenic in highly proliferative tissues and rapidly dividing cells than in differentiated non-replicating cells^18,164,165. Consistent with this finding, the signature mutations stimulated by triplet repeats are expansions or contractions of repeat units, which are likely the result of slippage errors during DNA replication^18,166 (Fig. 4). Using a forward mutagenesis assay, different forms of non-B DNA structure, including H-DNA, Z-DNA and cruciform DNA, were found to induce point mutations, DNA single-strand breaks and DSBs, and large-scale deletions in replication-deficient HeLa cell extracts^167–169. Therefore, there are evidently multiple mechanisms involved in the mutagenic processing of non-B DNA structures that depend on several factors, including the type of DNA structure, topological conditions and genomic processes such as transcription, replication and DNA repair. In this section, we summarize the replication-dependent and replication-independent mechanisms of non-B DNA structure-induced mutagenesis.

Fig. 4 |. Replication-associated genetic instability induced by non-B DNA.

A, Non-B DNA formed at a progressing replication fork. A progressing DNA replication fork is depicted on the top. Aa, A non-B DNA structure (shown in the schematic as H-DNA) in front of a replication fork slows or stalls replication, which gives rise to further structural alterations on the replication complex. Ab, A hairpin structure formed on the template of a lagging strand can lead to replication stalling or repeat contraction (repeat template skipping). Ac, Ad, Hairpin structures formed on the nascent strands on the leading and lagging strands can lead to repeat expansion (via nascent strand self-folding and misalignment). B, Non-B DNA-induced transcription and replication collisions. Ba, Transcription and replication forks in the same direction. Bb, Non-B DNA (shown as a cruciform structure) slows or stalls transcription elongation and leads to a co-directional collision. Bc, Non-B DNA slows or stalls replication or transcription and disrupts the coordination, leading to headon collisions. Collisions in either direction can lead to replication stress and genetic instability. RNAP, RNA polymerase; Pol, DNA polymerase.

Replication-dependent mutations

Both the stability and topological features of non-B DNA on the template strand can affect the processivity and fidelity of DNA polymerases^170,171. New advances in nucleic acid sequencing, such as single-molecule real-time (SMRT) sequencing technology have made it possible to determine the processivity and fidelity of DNA polymerases at the nucleotide level⁸. Studies using such techniques have revealed that many types of non-B DNA motif, such as G4 DNA-forming sequences, Z-DNA-forming GC repeats and hairpin/cruciform-forming AT and CAG repeats, increased sequencing errors that were positively associated with the reduced kinetics of the DNA polymerases⁸. In a high-throughput primer extension assay at 20,000 different sequences, T7 DNA polymerase was found to be significantly stalled at G4 DNA, hairpins and loop structures, even after controlling for GC context. Its fidelity was also reduced, supporting a plausible mechanism for non-B DNA-induced DNA polymerization difficulties and errors that lead to genetic instability¹⁷².

Hairpin or G4 DNA structures formed at CGG repeats in the FMR1 gene impeded all three eukaryotic replicative B-family DNA polymerases; Polα, Polδ and Polε¹⁷³. An AT(24) repeat and a run of 19–28 As from the common fragile site FRA16D were capable of stalling both Polα and Polδ¹⁷⁴. In addition, H-DNA-forming GA or GGAA repeats resulted in Polα pausing, which was more pronounced when the polypurine sequence served as the template¹⁴⁰. Polδ can be stalled on the G-rich template of telomeric TTAGGG repeats even in the presence of proliferating cell nuclear antigen (PCNA) and replication factor C (RFC). In addition, a G4 DNA stabilizer, BRACO-19, further inhibited Polδ stalling within G-rich regions¹⁷⁵. The proofreading activity of the B-family polymerases helps to reduce misalignment-based replicative errors^176,177, and deficiency of Pols α, δ and ε enhanced expansion of the GAA triplet repeat in yeast¹⁷⁸. Still, the B-family polymerases created ~1,000-fold more misalignment-based insertion or deletion events on GT(10) or CA(10) repeat templates than in adjacent non-repetitive sequences¹⁷⁹, demonstrating the impact of non-B DNA structure-forming repeats in this process.

If the replicative Polδ and Polε are stalled for an extended period of time, DNA polymerases from other families with lower fidelity can be recruited to take over the synthesis through non-B DNA regions to complete genome replication. Such DNA polymerases include the DNA repair X-family members Polβ¹⁸⁰ and Polλ¹⁸¹, the Y-family translesion synthesis polymerases Polκ, Polη^182–188 and REV1 (refs.^189–191), and the A-family translesion polymerase PolQ (Polθ)^192,193. Depletion of Polκ and Polη sensitized human cells to the G4 DNA stabilizer telomestatin and led to more DSBs in transgenic HeLa cells that harboured multiple copies of G-rich sequences from the human MYC promoter, which contains multiple G4 DNA, Z-DNA and H-DNA motifs. Furthermore, there were more DSBs in transgenic HeLa cells that contained either the BCL2 gene major break region (Mbr), which contains an H-DNA-forming sequence, or the H-DNA-forming sequences from Kaposi’s sarcoma-associated herpesvirus (KSHV)¹⁸³. Polκ and Polη facilitated replication through a CTG(100) repeat or a polypurine–polypyrimidine sequence from the PKD1 gene that can adopt H-DNA or G4 DNA structures and attenuated the formation of DSBs¹⁸⁵. These results suggest that the less stringent repair and translesion bypass DNA polymerases can facilitate DNA synthesis through non-B DNA-forming regions. However, these lower-fidelity polymerases can lead to base misincorporations and misalignments, resulting in various mutations¹⁹⁴. Replication stalling at G4 DNA caused DSBs, and their repair required PolQ, which led to small deletions, in a mechanism that differs from non-homologous end-joining or homologous recombination¹⁹². Thus, recruitment of error-prone polymerases to bypass non-B DNA-induced impediments to replication seems to be a double-edged sword for the maintenance of genomic integrity and stability.

Replication fork collapse and DSB formation can occur if the non-B DNA-induced impediment is not unwound or bypassed^35,128 (Fig. 4). Using a unique exogenous G-rich sequence with two distinct G4 DNA structure folding possibilities that stall replication forks at different positions, a study in C. elegans revealed that G4 DNA was stable enough to be maintained at the same location during proliferation and stimulated deletions in daughter cells similar to those in parental cells¹⁹⁵. Whether or not this persistency occurs in other organisms and/or is unique to G4 DNA remains to be determined.

Expansion of simple repeats in genomic DNA occurs at various stages of development in different cell types and is associated with more than 30 hereditary human diseases. The dysregulation of DNA replication at non-B DNA formed within these repeats is a major driving force of expansion events^10,28. Large expansions of CTG/CAG triplet repeats occurred more frequently when the CAG repeats were used as the template during lagging strand replication¹⁹⁶, with the nascent CTG strands more likely to form stable hairpin structures than the CAG repeats^10,28. A CAG repeat at the 3′ end of an Okazaki initiation zone resulted in expansion, yet it caused contraction events when located at the 5′ end¹⁹⁷. In the yeast URA3 reporter gene, long GAA repeats tended to gain a relatively narrow range of 44–63 extra triplets within the length of an Okazaki fragment¹⁷⁸. Together, these data demonstrate replication-dependent mechanisms of non-B DNA-induced mutagenesis.

However, a CTG(55) repeat in a transgenic mouse genome surrounded by its native 45 kb genomic segment of the human DMPK gene, which encodes dystrophia myotonica protein kinase, showed expansion with age in the heart, gastrocnemius, liver, pancreas and kidney, with no obvious relationship to cellular proliferation rates¹⁹⁸. CTG repeats are also highly unstable in tissues with low levels of proliferation, such as the basal ganglia, cerebral cortex and frontal cortex^199,200. Cells derived from various tissues of transgenic mice carrying long CTG(162) repeats exhibited different levels of expansion events, with the highest levels found in the kidney and lower levels in the lung; thus, there was no simple correlation between repeat instability and cell proliferation¹⁹. In addition, several cell types in the cerebellum and hippocampus, such as Purkinje cells, showed high levels of CTG repeat expansion, and granule cells were relatively more stable in spinocerebellar ataxias²⁰⁰, suggesting a genomic DNA replication-independent mechanism of DNA structure-induced mutagenesis (see below).

Non-B induced replication–transcription collision

Replication and transcription can occur on the same DNA strand simultaneously in both prokaryotic and eukaryotic genomes. In prokaryotic genomes, the transcriptional templates of highly expressed genes are predominantly on the leading strand during replication²⁰¹, such that replication and transcription move in the same direction. Eukaryotic genomes are more complicated, as there are tens of thousands of replication origins²⁰², with no substantial preference of positioning genes on replication leading versus lagging strands. As a result, both transcription and replication can initiate at multiple sites and move in different directions on the same chromosome, which increases the risk of ‘head-on’ collisions. Formation of non-B DNA on the templates that are shared by both transcription and replication complexes can enhance the potential of such collisions¹⁷ (Fig. 4).

Non-B DNA-forming sequences such as CGG repeats, which can adopt Z-DNA, hairpin or cruciform structures, or cruciform-forming AT-rich palindromes are enriched in fragile sites²⁰³. These sites are generally replicated more slowly and at later stages compared with other regions, and are associated with chromosome breakage and disease development^204–206. Many common fragile sites nestle in or overlap with large genes that are actively transcribed^207,208. Both the large size of the genes and non-B DNA-forming sequences can drastically slow the RNA polymerase complexes, and transcription can take longer than a cell cycle²⁰⁹. Therefore, active transcription and replication must occur together within these fragile sites, increasing the risk of collision. Replication–transcription collision in the genome of actively dividing Bacillus subtilis bacterial cells resulted in duplications, deletions and base substitutions not only at the sites of collision but also in adjacent areas²¹⁰. Because transcription requires NTPs for RNA synthesis and the progressing transcription complex recruits UTP, thus increasing the local UTP concentration near the stalled DNA polymerases, misincorporation-incorporation of UTP at the template of A could occur when the collision is resolved and replication is restarted²¹¹.

Notably, it is common that cancer cells have replication stresses that lead to slower replication fork progression and dysregulated firing of replication origins²¹². As a result, the inappropriately timed and prolonged replication and overactivated transcription could result in an increased risk of collisions in cancer cells, particularly at non-B DNA-containing regions, which may contribute to oncogene-induced DNA damage and genomic instability in cancer.

In addition to increasing the risk of replication–transcription collision, non-B DNA such as G4 DNA or hairpin structures could also form within the transcription bubble on the non-template ssDNA region. It is reasonable to speculate that, together with an RNA–DNA hybrid formed on the template strand, known as an R-loop (Fig. 1f), non-B DNA formation on the non-template strand could render the stalled R-loop more difficult to resolve before restarting of the replication fork.

Structure-specific repair cleavage models

Some DNA repair proteins screen genomes by sensing distortions of the DNA double helix induced by lesions, which is the initial signal to recognize DNA damage²¹³, followed by recruitment of repair enzymes (for example, nucleases) to the sites of damage to remove the lesions and restore genome integrity. Non-B DNA structures induce distortions in the DNA, affecting protein binding and chromosome organization, similar to some types of DNA damage. As a result, non-B DNA can stimulate DNA damage responses and may be recognized and cleaved by structure-specific DNA repair proteins (Fig. 5). For example, long tracts of CAG(175) repeats on plasmids stimulated expression of the sfiA (sulA) gene, an inhibitor of septum formation induced early in the SOS response²¹⁴. In comparison with an SOS-defective strain, cells with activated SOS responses increased the supercoiling density of the plasmid, which in turn stimulated non-B DNA formation and deletion frequencies at the CAG repeats²¹⁴. Long CAG repeats in a yeast artificial chromosome (YAC) also stimulated a DNA damage checkpoint response^34,215. A plasmid containing a 2.5 kb H-DNA-forming polypurine–polypyrimidine tract from intron 21 of the unstable human PKD1 gene was recognized by the NER proteins UvrB and UvrC, induced an SOS response in bacteria and delayed growth of the transformed cells²¹⁶. These findings suggest that non-B DNA structures may be recognized as ‘damage’ to stimulate cellular DNA damage responses. If the DNA structure-specific repair cleavage is subsequently processed in an error-free fashion, the primary sequence will be retained, which could result again in non-B DNA formation, triggering further recognition and cleavage until mutations occur to remove or prevent non-B DNA formation. Thus, structure-specific DNA repair cleavage models provide a plausible explanation for replication-independent non-B DNA-induced mutagenesis in cells, particularly in those with low proliferation rates.

Fig. 5 |. Structure-specific cleavage modulates non-B DNA structure-induced genetic instability.

a, Structure-specific cleavage of non-B DNA leads to genetic instability. A non-B DNA structure (shown in the schematic as H-DNA) causes helical distortions and creates an open structure for recruiting DNA repair nucleases. DNA structure-specific cleavage generates breaks within or surrounding the non-B DNA structure, followed by error-free or error-generating repair. This ‘structure forming–repair’ cycle can occur repeatedly until a mutation interrupts the formation of non-B DNA or a deletion removes the non-B DNA-forming sequence. b, A non-B DNA structure (shown in the schematic as H-DNA) is formed in front of a progressing replication fork and stalls DNA replication, increasing the chance for fork collapse and double-strand break (DSB) formation. Structure-specific cleavage of the non-B DNA structure creates a nick or DSB, which unwinds the non-B DNA conformation and reduces structure-induced genetic instability by allowing continuous replication.

Mismatch repair proteins.

Mismatch repair (MMR) proteins, particularly the MSH2–MSH3 complex (MutSβ), sense heteroduplex DNA that contains small loops generated by slippage events at microsatellite repeats (reviewed in ref.³³). Single-strand loop-outs are common in many types of non-B DNA structure, such as the simple repeat-induced slippage loops, the unpaired tips of hairpin structures and the unpaired junctions of B-DNA to non-B DNA transitions. Thus, it is reasonable to speculate that MMR proteins recognize and process non-B DNA structures that contain small loops. Indeed, MSH2–MSH3 was found to process short loops within CTG trinucleotide repeats, resulting in repeat unit number alterations, but was not required for processing larger CTG trinucleotide repeat loops (more than five CTG repeats)^217–219. Consistent with this result, long CTG(20) loop-outs were processed in cell extracts from multiple different human cells, including neuronal cells, independently of MSH2, MSH3, MSH6, MLH1, PMS2 or PMS1 (ref.²²⁰), despite the potential A–A or T–T mismatches formed in the loop-outs. Interestingly, the MSH2–MSH3 complex has a higher binding affinity for small loop-outs than for mismatched base pairs²²¹, and A–A mismatches in fact reduced the binding affinity of MSH2–MSH3 for the hairpins formed at CTG or CAG repeats. The ATPase activity of MSH2–MSH3 was also reduced on the small CA(4) loop compared with perfect base-paired hairpins of the same length²²²; therefore, the mismatches do not seem to be the main recruiter of MSH2–MSH3 in this case. Similarly, hairpins with 3–17 bp stems and 6–8 nt tips²¹⁹ or longer perfect inverted repeats^217,218 were typically processed independently of MMR proteins. A reasonable hypothesis is that the MSH2–MSH3 complex binds to the junctions of hairpins and can also interact with the tips; yet the tips of longer hairpins are too distant from the junctions and the binding repair complex. However, direct evidence is still lacking. Note that a long CAG repeat could form either a large hairpin or multiple small loops, and the effects of MSH2–MSH3 could be very different depending on the size of each loop. This may explain the different effects of MSH2–MSH3 on CAG repeats in different species and under different experimental conditions. As an example, deficiency of MSH2 (ref.²²³) or MSH3 (ref.²²⁴) reduced CAG expansions and resulted in more contractions in the genomes of mice²²⁵ but reduced contractions at long repeats in genomic DNA from human cells²²⁶. Therefore, although there is strong evidence to support a role for MMR in the mutagenic processing of hairpin structures, factors such as nucleosome and chromosome structures, transcription and replication activities, and the presence of DNA binding proteins could affect the stability and/or MMR-associated processing of non-B DNA structures.

The MSH2–MSH3 complex has been shown to bind to intermolecular triplex DNA structures with high affinity together with a NER damage or distortion recognition complex, XPA–RPA or XPC–RAD23B²²⁷. GAA repeats from the FXN gene, involved in Friedreich’s ataxia, can form loop-outs owing to slippage events or H-DNA structures²²⁸, and GAA(120–340) repeats were shown to impede DNA replication forks, resulting in chromosomal breakage and gross chromosomal rearrangements in yeast²²⁹. Deficiency of the MMR proteins MSH2, MSH3, MLH1 or PMS1, but not MSH6, suppressed DSBs and reduced large deletions in yeast, suggesting a role for MMR in creating DSBs at GAA repeats, although the MMR deficiency increased the small deletions within these repeats, consistent with the canonical MMR activity²²⁹. In addition, the MMR proteins MSH2, MSH3, MLH1 and PMS1 have been shown to be involved in stimulating the formation of DSBs at H-DNA formed at GAA(100) repeats in non-dividing yeast cells, where the DSBs were processed by Exo1 and re-joined by non-homologous end-joining activity, leading to large deletions²³⁰.

The human MSH2–MSH6 complex was found to bind to G4 DNA as visualized by electron microscopy and could bind a G4 DNA-forming oligonucleotide as assessed by slower migration in gel mobility-shift assays¹⁶⁰. Surprisingly, bacterial MutS binds to G4 DNA with a higher affinity than to G–T mismatches, a canonical MMR substrate. However, adding ATP to the reaction failed to release MutS from G4 DNA as it does on duplex DNA, suggesting a G4 DNA-specific interaction²³¹. In addition, when MutS and MutL were bound to G4 DNA, the hydrolysis activity of MutH was increased by about threefold over that of MutH alone. Interestingly, the binding of MutS to G4 DNA did not seem to require its mismatch discrimination function, as the specific binding was maintained after a highly conserved crucial residue for heteroduplex recognition and mismatch correction was mutated²³². Thus, it is possible that MutS has a unique G4-DNA binding motif, leading to repair activity that differs from canonical mismatch-direct MMR. The MSH2–MSH3 complex in conjunction with Polβ facilitated synthesis through GAA and CAG repeats containing abasic sites in vitro. The interaction of MSH2–MSH3 with Polβ increased the potential for flap formation and repeat expansion, rather than contraction events that occurred when Polβ acted on the repeats alone²³³. Consistent with MMR proteins functioning outside of their canonical roles on non-B DNA, we recently discovered that the MSH2–MSH3 complex, in conjunction with the NER complex Rad10–Rad1 (ERCC1–XPF), was required for Z-DNA-induced genetic instability in yeast and human cells. MSH2–MSH3 associated with Z-DNA, as evidenced by ChIP assays. However, instead of recruiting downstream MMR proteins, the NER complex ERCC1–XPF was recruited to the MSH2–MSH3-bound Z-DNA-containing region. Cleavage of Z-DNA by ERCC1–XPF led to DSBs and genetic instability in yeast and mammalian cells^31,234.

Nucleotide excision repair proteins.

Interestingly, another non-B DNA structure, H-DNA, is also cleaved by ERCC1–XPF³², similarly to Z-DNA. However, in contrast to Z-DNA, the functional NER pathway was involved in a distinct mechanism of H-DNA-induced genetic instability, independently of MMR proteins. Deficiency of the NER nucleases ERCC1–XPF and XPG, or the central NER scaffold molecule XPA, reduced H-DNA-induced mutations in yeast and human cells, independently of the DNA replication status. Both ERCC1–XPF and XPG were able to cleave H-DNA in vitro, and ERCC1–XPF binding to H-DNA was reduced in XPA-deficient cells³². These results suggest that NER is responsible for a replication-independent, structure-specific cleavage model of H-DNA-induced mutagenesis. NER proteins monitor helical distortions in the DNA helix induced by bulky DNA adducts²³⁵; therefore, H-DNA might be recognized as ‘damage’ owing to its associated helical distortions. In fact, the NER mechanism was also required for intermolecular triplex structure-induced mutagenesis in mammalian cells²³⁶, and purified human recombinant XPA–RPA²³⁷ and XPC–RAD23B²³⁸ were shown to bind to intermolecular triplex DNA structures in vitro with high affinity and specificity.

Purified UvrA binds to supercoiled CAG repeats with higher affinity (by two orders of magnitude) than to linear CAG repeats, suggesting a role for DNA structure in protein–DNA binding interactions. Moreover, deficiency of UvrA reduced the deletion events at CAG repeats in Escherichia coli²³⁹, as did its damage recognition partner UvrB^240,241. UvrD helicase deficiency increased deletion events at long CAG repeats as expected, perhaps owing to increased or stabilized non-B DNA formation in the helicase-deficient bacterial cells and/or its role in MMR²⁴². Surprisingly, deficiency in the endonuclease UvrC also enhanced CAG-repeat-induced deletions, perhaps because other enzymes can cleave the UvrAB-bound CAG repeats²⁴⁰. The NER proteins XPA, XPC, ERCC1 and XPG, and the MMR proteins MSH2 and MSH3 have also been implicated in contraction of CAG repeats in human cells in a transcription-dependent fashion^226,243. Why so many proteins are required for CAG instability and how they coordinate with each other in recognizing and processing these structures remains to be fully elucidated.

Other DNA repair proteins.

In addition to MMR and NER proteins, many other enzymes have been found to have activities on non-B DNA. For example, in yeast, Mre11 can bind to long (>160 bp) palindromic DNA sequences^244,245. Purified Mre11 exhibited DNA structure-specific endonuclease activity at hairpin and cruciform structures, and cleaved the DNA at the 5′-junction at the loop of a hairpin, and the junction of a 3′-end flap structure²⁴⁴. The MRN (Mre11–Rad50–Nbs1) complex also interacts with BRCA1, which contains a four-way branched DNA structure binding domain²⁴⁶ and can facilitate the recruitment of the MRN complex to cruciforms²⁴⁷. Sae2 functions together with the MRX complex in yeast to initiate DNA end resection for DSB repair and to process hairpin or cruciform structures²⁴⁸. The Mre11 protein can bind to G4 DNA with higher affinity than to B-DNA and cleave it in a Mn²⁺-dependent manner²⁴⁹. MRX generates DSBs at cruciform structures at an early stage during pre-meiotic replication, and the meiotic recombination protein Rec12 creates DSBs at a later stage²⁵⁰. Meiosis-specific endonuclease Spo11 also cleaved hairpin structures and was involved in CAG repeat expansions and deletions in yeast²⁵¹.

Cruciform DNA shares some structural similarities with Holliday junctions and, as expected, the Yen1/GEN1 (refs.^252,253) resolvases, SLX1/SLX4 (refs.^254,255) and Mus81–Eme1 (Mms4 in budding yeast) resolvases²⁵⁶ were all reported to cleave cruciform structures^257,258. These enzymes were also recruited to and cleaved many common fragile site sequences that contained repetitive sequences, leading to DSBs and subsequent genetic instability in mammalian cells²⁵⁹. Recently, a YAC reporter system was used to investigate the genes involved in processing long inverted repeats ranging from 320 bp to 2.7 kb, and identified many endonucleases including the MRX–Sae2 complex, Mus81–Mms4 resolvase, and replication and DSB repairrelated proteins such as Rfa2 (a subunit of heterotrimeric RPA) to be involved. Surprisingly, MUS81, YEN1, SLX4, RAD1 and MLH1 did not significantly affect the DSBs and gross chromosomal rearrangements at these long inverted repeats²⁶⁰. Whether these discrepancies are due to the different reporter systems and/or the different hairpin or cruciform substrates used remains to be determined.

Repair proteins that suppress non-B DNA-induced mutation.

Notably, not all nucleases increase non-B DNA-induced genetic instability. For example, the FANCI-associated nuclease 1 (FAN1), involved in DNA inter-strand crosslink repair, has been shown to bind to MLH1 and inhibit its interaction with MSH3, thereby reducing MMR-promoted CAG repeat expansion in human cells²⁶¹. FAN1 can dimerize and bind to trinucleotide repeats and cleave slipped CAG or CTG repeats near the junctions²⁶². In addition to its endonuclease activity, FAN1 shows 5′ to 3′ exonuclease activity on hairpin loops containing A–A and T–T mismatches, but is paused by perfectly paired hairpins. Thus, it is plausible that the nuclease activities of FAN1 on CAG repeats represent a counterforce against repeat expansion. In support of this idea, reduced exonuclease activity of FAN1 in individuals with autism was found to be associated with CGG expansions²⁶².

Interestingly, we found that flap structure-specific endonuclease 1 (FEN1) can cleave H-DNA structures in vitro and attenuates H-DNA-induced mutagenesis in eukaryotic cells^31,32. FEN1 deficiency increased H-DNA-induced mutagenesis approximately fivefold only in replicationcompetent human and yeast cells. A possible explanation for this finding is that FEN1, as a replication assistant, cleaves H-DNA and diminishes the structural impediment ahead of replication forks, allowing for continuous replication to maintain genetic stability^31,32 (Fig. 5).

This section summarizes mechanisms of non-B DNA-induced mutagenesis, and the roles of non-B DNA on replication, transcription–replication collisions, stimulation of structure-specific cleavage and the impact of alternative structures on DNA damage and repair. Of note, although many types of non-B DNA share some structural and functional features, and many enzymes show similar activities towards different types of non-B DNA conformation, each structure has distinct features in terms of how it is recognized and processed. For example, the hairpin structure formed at CAG repeats affects the function of MSH2–MSH3 on DNA mismatches²²², whereas G4 DNA does not affect MMR activity on a G–T mismatch in close proximity²³². Instead, G4 DNA has been shown to prevent the recognition and excision of 8-oxoG by NEIL1, NEIL3 and OGG1 (ref.²⁶³). We also present evidence for the distinct processing of Z-DNA and H-DNA, with both being cleaved by ERCC1–XPF, albeit via different mechanisms^31,32. Therefore, individual types of non-B DNA structure should be investigated separately for their distinct mechanisms of mutagenic processing and associated biological outcomes.

Non-B DNA and human disease

DNA is no longer considered passive with regard to incurring damage or mutagenesis. Instead, the DNA itself, in the presence or absence of exogenous damage, can act as a causative factor for genetic instability, leading to DSBs, deletions, recombination and/or large genomic rearrangements that are associated with disease aetiology. Expansions of non-B DNA structure-forming repeats have been implicated in many diseases via various mechanisms, depending on the sequence, length and/or the location of repeats in the affected genes (reviewed in refs.^26,28). For example, expansion of CAG repeats within coding regions of genes leads to stretches of polyglutamine (polyQ) in the resulting proteins, contributing to more than 20 neurodegenerative and neuromuscular diseases, including Huntington disease, spinal and bulbar muscular atrophy, dentatorubral-pallidoluysian atrophy and several types of spinocerebellar ataxia (reviewed in ref.²⁶). CGG expansions in the non-coding 5′-UTR of FMR1 inhibit transcription and contribute to fragile X syndrome²⁶⁴. The repeats can also stall replication²⁶⁵ and recruit structure-specific ‘repair’ cleavage enzymes, leading to chromosomal abnormalities²⁶⁶. The CGG repeats in the FMR1 RNA can also affect the translation of other RNAs within the same RNA granules²⁶⁷. Expansion of CTG repeats in the 3′-UTR of the DMPK gene leads to long CUG runs in the RNA that can interfere with the developmentally regulated alternative splicing of pre-mRNAs, resulting in toxicity associated with myotonic dystrophy type 1 (ref.²⁶⁸).

In addition to disorders caused by repetitive DNA itself, many types of non-B DNA structure are involved in human diseases associated with genetic instability, such as cancer (reviewed in ref.³²). Using a bioinformatic approach, it was found that ~33% of missense mutations and ~37% of microdeletions in the Human Gene Mutation Database (HGMD)²⁶⁹ occurred within non-B DNA-forming repeats, which is significantly higher than by chance alone²⁷⁰. Potential non-B DNA-forming sequences were also found to have higher substitution frequencies in the UCSC Genome Browser and the Simons Genome Diversity datasets¹⁶². From an analysis of 1,809 whole-genome sequences from ten cancer types, more mutations were found near or within non-B DNA regions in general, particularly at the single-stranded spacer regions between symmetrical arms of direct and inverted repeats or H-DNA mirror repeats¹⁶³. Non-B DNA-forming sequences are also enriched at chromosomal breakpoints identified in many diseases, including translocation-related cancers, such as lymphomas and leukaemias^271,272. For example, the human MYC promoter region contains multiple overlapping non-B DNA-forming sequences with the capacity to adopt H-DNA, G4 DNA and Z-DNA within a small region (~400 bp) surrounding the P1 promoter, which is one of the major breakpoint cluster regions in MYC-related translocations in lymphomas⁴⁰. H-DNA has been implicated in translocation events within the BCL2 major breakpoint region in follicular lymphomas²⁷³. G4 DNA is associated with mutations in ataxias and fragile X syndrome²⁷⁴. In a dataset from the International Cancer Genome Consortium Data Portal, comprising 2,234 samples from ten cancer types, stem–loop structures fit the blood, brain, liver and prostate cancer breakpoint hotspot profiles, and G4 DNA-forming sequences co-localized with mutation hotspots in bone, breast, ovarian, pancreatic and skin cancers at levels much higher than by chance alone²⁷⁵. In 200–400 bp windows flanking breakpoints, characterized in 19,947 translocations in human cancer genomes, non-B DNA-forming sequences were significantly enriched, including simple AT repeats, GAA or GAAA repeats and other repetitive sequences that have the propensity to adopt H-DNA, Z-DNA, G4 DNA and cruciform or hairpin structures^32,272. A very recent study analysed ~630,000 cancer breakpoints using machine learning tools to search for genetic features associated with genome breakage. It was found that the cancer-associated breakpoint hotspots were predictable, and that transcription and the presence of non-B DNA motifs were predominant factors responsible for the development of the hotspots²⁷⁶.

Although non-B DNA-forming sequences have been implicated in the development of many diseases, it is likely that many of the non-B DNA-induced mutations are silent and do not exhibit an obvious phenotype. For example, long palindromic AT-rich repeats on chromosomes 11q23 and 22q11 have been shown to stimulate a high frequency (~10^–5) of de novo t(11;22) translocations in sperm cells from healthy men²⁷⁷. However, most cells that carry these mutations do not show a phenotype and/or are lost in screening systems, whereas other cells with non-B DNA-induced mutations in crucial genomic regions could trigger apoptosis and would no longer be detectable in the population. Therefore, non-B DNA-induced mutations in human genomes are likely much more frequent than estimated.

Conclusions and perspectives

Although our knowledge of non-B DNA structures, their functions in the genome and contributions to evolution and disease has advanced substantially over the past few decades, there is still much to be discovered. Thus, many challenges remain to be addressed in future studies, some of which are listed below.

To date, it remains a challenge to directly detect non-B DNA structures in genomes of living cells, in part because their formation in vivo is dynamic and transient. The methods to detect non-B DNA in cells thus far depend on indirect measurements and/or the inclusion of steps that can alter non-B DNA formation, such as cell fixation, protein removal, changes in pH and salt concentration or the addition of probes or antibodies that can induce structure formation, and/or stabilize or destabilize existing non-B DNA structures. Some recent progress has been made in this respect. For example, a small-molecule fluorophore was conjugated with a G4 DNA ligand, pyridostatin, to form a fluorescent G4 DNA probe²⁷⁸. The small molecular weight and polarity of the probe molecule enabled penetration of the cell membrane, and the low cell toxicity fluorescent probe was used for real-time detection of single G4 DNA structures in living cells. This small-molecule probebased procedure is less toxic to host cells than other methods and is compatible with studies in living cells, and as such better reflects the physiological situation in vivo. Note that owing to the dynamic nature of non-B DNA conformations, using a DNA structure-specific probe will inevitably affect the balance of the B-DNA to non-B DNA structure formation transition and equilibrium. Thus, more sensitive and less invasive methods that can directly detect non-B DNA without interfering with DNA structural transitions in cells would substantially advance the field.

Our understanding of free energy alterations, the kinetics of B-DNA to non-B DNA transitions, the interactions and bonds between the atoms involved in non-B DNA formation and stabilization are still limited, particularly measurements under physiological conditions. As a result, although many computer programs for predicting non-B DNA conformations in genomes are available, they are empirically derived. Thus, a better understanding of these most fundamental and basic factors in non-B DNA structure formation is crucial for future progress in the field.

On the basis of decades of research, we now know that non-B DNA-forming sequences have important roles in various biological processes and can stimulate genetic instability, implicating them in evolution and disease aetiology. Therefore, strategies to modulate these structures using small molecules or probes are warranted owing to their ability to provide spatial and temporal tools to interrogate DNA structures and their potential as therapeutic agents^279,280. Much progress has been made towards this goal by employing in silico simulation and high-throughput methods to identify small-molecule ligands to modulate non-B DNA formation and/or stability. However, small ligands can interact with genomic and mitochondrial DNA and affect DNA structure throughout the genome and may have off-target effects, potentially limiting their therapeutic efficacy. Thus, the development of approaches to increase specificity by combining ligands with sequence-specific targeting is warranted. Such approaches may include the use of non-B DNA-specific probes, antisense oligonucleotides, intermolecular triplex-forming oligonucleotides or peptide nucleic acids²⁸¹.

In a recent study, the structure-specific DNA ligand naphthyridine-azaquinolone (NA) was used to treat cells from patients with Huntington disease or a Huntington disease mouse model that harbours a transgene of HTT (huntingtin) exon 1 containing CAG repeats²⁸². NA, which binds specifically to slipped CAG DNA intermediates and thus affects expanded CAG repeats, induced repeat contractions in both cultured human cells and medium spiny neurons of the mouse striatum. NA injection reduced mutant HTT protein aggregates in mice, a biomarker of Huntington disease pathogenesis, suggesting a promising approach to reduce the pathogenic repeat length with non-B DNA structure-specific DNA ligands²⁸². This type of compound that can specifically interact with pathology-related non-B DNA conformations is needed to advance the effort towards developing DNA-targeted therapeutics.

Using the G4 DNA-specific binding domain of RNA helicase associated with AU-rich element (RHAU, also known as DHX36) fused with a cleavage domain of the Fok1 nuclease, a G-quadruplex-specific DNA endonuclease was constructed to cleave double-stranded DNA adjacent to G4 DNA-forming sequences²⁸³. Such enzymes represent powerful tools for DNA structure research. If they can be delivered in vivo and their expression controlled, they may provide another approach for targeted genome modification. Although not necessarily a DNA structure-specific approach, CRISPR–Cas9 techniques have also been applied to delete core G4 DNA sequences within a microRNA cluster in rat, which led to increased microRNA levels in the heart, contributing to cardiac contractile dysfunction²⁸⁴. In future studies, similar strategies could be employed to modulate the structure and function of other non-B DNA-forming sequences in vivo.

Taking advantage of sequence-specific DNA triplex formation, short triplex-forming oligonucleotides have been developed that can form intermolecular triplexes that share structural similarities with intramolecular H-DNA. Binding of triplex-forming oligonucleotides to target duplexes to form stable triplex structures can inhibit DNA replication and gene expression, and stimulate site-specific mutations in genomes²⁸¹. Chemotherapeutic DNA-damaging agents can also be used in conjunction with or covalently linked to triplex-forming oligonucleotides to direct site-specific damage in genomes to enhance the efficacy of cancer chemotherapy^285,286.

In summary, because non-B DNA structures have important roles in various biological and pathological processes and provide unique structural features for recognition and binding, they represent potential druggable targets for use as both research tools and therapeutics. Although limitations still exist in this type of approach, such as specificity for the targeted non-B DNA versus B-DNA regions, potential off-target effects and efficacy based on targeting multiple similar structures, progress in this area is promising. For example, a small-molecule G4 DNA-specific ligand, CX-5461, is currently being evaluated in an open-label, multi-centre phase Ib study as a potential targeted cancer therapeutic for BRCA1/BRCA2 homologous recombination-deficient tumours^287,288. Nonetheless, until direct and conclusive methods are available to detect and specifically target non-B DNA in vivo, caution is warranted when studying such complex and dynamic non-B DNA structures in the genome.

Glossary

Circular dichroism

Absorption spectroscopy method to detect the differential absorption of leftand right-handed light spectra for rapid evaluation of the secondary structures of macromolecules such as protein and DNA

DNA helicases

A class of motor proteins that move along DNA and transiently separate duplexes into two single strands using energy from ATP hydrolysis

Fourier transform infrared spectroscopy

A spectroscopy method that simultaneously collects the absorption, emission and photoconductivity of a wide spectral range at high resolution to measure the intensity and wavelength of light required to vibrate molecules in a sample

Holliday junctions

Branched DNA structures containing four arms covalently linked together that serve as key intermediates in many meiotic and mitotic homologous recombination events

Negative supercoiling

A segment of underwound DNA in which the two strands wind around the helical axis less than 360° every 10.5 bp and retain twist strain (free energy)

Okazaki fragments

Short fragments of DNA produced by discontinuous replication on the lagging strand during DNA replication. Because the template for lagging strand synthesis is exposed in the 5′–3′ direction at the progressing replication fork, the nascent strand is composed of sequential Okazaki fragments created by DNA polymerase working backwards from the replication fork

Satellites

A subfraction of genomic DNA consisting of short repetitive nucleotide sequences that are repeated a large number of times. These non-coding repeats are important for centromere and heterochromatin construction and separate from the rest of the genomic DNA on a density gradient because of their higher content of AT base pairs

SOS response

A complex global response to DNA damage identified in bacteria that includes activation of multiple factors, leading to the stalling of cell division and alteration of DNA replication, recombination and repair to promote genome integrity and cell survival, at the cost of increased mutagenesis

Stretching tension

When both ends of a segment of DNA are anchored (for example, by proteins) and the DNA is pulled mechanically, it carries stretching tension coupled with twisting torsion along the helix and can be elongated by up to 70% without disrupting base pairs

Topoisomerases

A class of enzymes that are able to cleave one or both strands of DNA to release topological stress on DNA duplex, and to link or unlink, knot or unknot associated DNA molecules

Translesion synthesis polymerases

Polymerases that can catalyse DNA polymerization at damaged templates during replication and/or repair, although often with lower fidelty than replicative polymerases

Unequal sister chromatid exchange

A mitotic crossover event that leads to the exchange of genetic material between homologous chromosomes and is also a major repair pathway for double-strand breaks

Related links

detectIR: https://sourceforge.net/projects/detectir/

DNA Structure Search: http://utw10685.utweb.utexas.edu/nonbdna/

G4Hunter: http://bioinformatics.ibp.cz

palindrome: http://emboss.bioinformatics.nl/cgi-bin/emboss/palindrome

QGRS Mapper: https://bioinformatics.ramapo.edu/QGRS/index.php