Structures and conformational dynamics of DNA minidumbbells in pyrimidine-rich repeats associated with neurodegenerative diseases

Abstract

Expansions of short tandem repeats (STRs) are associated with approximately 50 human neurodegenerative diseases. These pathogenic STRs are prone to form non-B DNA structure, which has been considered as one of the causative factors for repeat expansions. Minidumbbell (MDB) is a relatively new type of non-B DNA structure formed by pyrimidine-rich STRs. An MDB is composed of two tetraloops or pentaloops, exhibiting a highly compact conformation with extensive loop-loop interactions. The MDB structures have been found to form in CCTG tetranucleotide repeats associated with myotonic dystrophy type 2, ATTCT pentanucleotide repeats associated with spinocerebellar ataxia type 10, and the recently discovered ATTTT/ATTTC repeats associated with spinocerebellar ataxia type 37 and familial adult myoclonic epilepsy. In this review, we first introduce the structures and conformational dynamics of MDBs with a focus on the high-resolution structural information determined by nuclear magnetic resonance spectroscopy. Then we discuss the effects of sequence context, chemical environment, and nucleobase modification on the structure and thermostability of MDBs. Finally, we provide perspectives on further explorations of sequence criteria and biological functions of MDBs.

1. Introduction

Short tandem repeats (STRs) are DNA sequences consisting of 1–6-bp repeating units connected in a head-to-tail manner [1]. They are widely distributed in both the coding and non-coding regions and constitute ∼3% of the human genomic DNA [2]. STRs were originally thought as junk DNAs until their biological roles were recognized in early 1990 s [3], [4]. In particular, genetic instability of STRs has been implicated in cellular dysfunctions associated with human diseases. Approximately 13 types of STRs, including CAG, CTG, CGG and GAA trinucleotide repeats, CCTG tetranucleotide repeats, TGGAA, ATTCT, TTTTA, TTTCA and AAGGG pentanucleotide repeats, and GGCCTG, CCCTCT and GGGGCC hexanucleotide repeats, have been well documented to exhibit dynamic mutations that change their sequence lengths during successive generations or individual’s life [5], [6], [7]. These STRs interfere with cellular processes like DNA replication, transcription and translation when the number of repeats reaches a certain threshold, leading to some neurodegenerative disorders which are also called as repeat expansion diseases [8]. Nearly 50 repeat expansion diseases have been reported so far, such as the fragile X syndrome [9], Huntington’s disease [10] and amyotrophic lateral sclerosis and frontotemporal dementia (ALS/FTD) [11]. For some of these diseases, a larger number of repeats in the mutated gene was found to correlate with a higher disease severity and an earlier disease onset [12], [13], [14].

Pathogenic mechanisms of repeat expansion diseases are very complex at various molecular levels including DNA, RNA and protein [15]. For repeat expansions occurring in the coding region, RNA transcripts harbouring expanded repeats are translated into extralong proteins which lose normal functions [15]. There are nine polyglutamine (poly-Q) diseases caused by CAG repeat expansions in the coding region of respective genes, and the translated proteins containing long tracts of glutamines form aggregates that cause cellular toxicity [16]. For the remaining diseases associated with non-coding repeat expansions, toxic RNA gain-of-function is involved in pathogenic mechanisms. Expanded RNA repeats form secondary structures and sequester various RNA-binding proteins to form nuclear foci, causing nuclear stress and mRNA splicing errors [17], [18]. Besides, some expanded RNA repeats such as rGGCCUG and rGGGGCC are reported to produce polypeptide repeats through a non-AUG translation pathway [19], [20].

Despite a complexity in the pathogenic pathways, expansions of STRs in the genomic DNA are origins of these neurodegenerative diseases. Therefore, a thorough understanding on the molecular mechanisms of repeat expansions lays the foundation for developing therapeutic approaches, especially those targeting at a DNA level. In 1995, Gacy et al. first reported the formation of hairpin structures in CAG and CTG repeats, and provided a strand slippage model to explain the occurrence of repeat expansions [21]. Since then, accumulating evidences have revealed the roles of non-B DNAs in causing repeat expansions and cellular dysfunctions [22]. The structures of STRs have been widely studied in the past three decades, and it is found that the propensity of STRs to adopt non-B DNA structures depends their sequence contexts. For instances, the CAG/CTG repeats formed hairpins [23], the guanine-rich CGG and GGGGCC repeats formed G-quadruplexes [24], [25], and the cytosine-rich CCG and CCCCGG formed i-motifs [26] (Fig. 1).

Fig. 1.

Schematics of the hairpin, G-quadruplex and i-motif structures formed by STRs.

Expansions of pyrimidine-rich CCTG, ATTCT and ATTTT/ATTTC repeats are associated with myotonic dystrophy type 2 (DM2) [27], spinocerebellar ataxia type 10 (SCA10) [28], and SCA37 and six subtypes of familial adult myoclonic epilepsy (FAME) [29], respectively. An earlier chemical/enzymatic probing study showed the hairpin formation propensity of CCTG repeats [30], and a subsequent NMR study revealed that CCTG repeats formed hairpin and dumbbell structures [31]. ATTCT repeats have been reported to form unwound secondary structures which may cause repeat expansions via replication re-initiations [32]. However, knowledge of these pyrimidine-rich STRs rests on their secondary structures, with a lacking of high-resolution structural information. Recently, the discovery of a new type of non-B DNA structure, namely the minidumbbell (MDB), provides new structural insights into these pyrimidine-rich sequences. In 2015, Guo and Lam first reported the MDB structures formed in TTTA and CCTG repeats [33], [34], and recently they found a similar type of MDB structure formed in ATTCT repeats [35]. These findings help to decipher mechanisms of repeat expansions in these pyrimidine-rich STRs. In this review, we first introduce the MDBs formed in natural STRs with a focus on their high-resolution structures revealed by solution nuclear magnetic resonance (NMR) spectroscopy, then discuss the effects of sequence context, chemical environment, and modified nucleobase on structure and thermostability of MDBs, and finally present perspectives on exploring the sequence criteria and biological functions of MDBs.

2. MDB structures formed in natural pyrimidine-rich repeats

2.1. Type I and type II tetraloops

There are two well-defined folding geometries of DNA tetraloops, namely, the type I and type II loops (Fig. 2 A) [36]. To facilitate discussions, here we named the first, second, third and fourth loop residues as L1, L2, L3 and L4, respectively. In a type I loop, L1 and L4 form a loop-closing base pair, and L1 to L3 continuously stack at the 5′-end of the loop. In a type II loop, L1 and L4 form a loop-closing base pair, L2 folds into the minor groove, and L3 stacks on L1-L4. Type I and type II loops are found in non-B DNAs such as hairpins and three-way junctions [37], [38], [39], [40], [41]. With reference to the NMR features of B-DNA, both type I and type II loops exhibit unusually upfield proton (¹H) and phosphorous (³¹P) NMR signals of L3 [36]. For the type II loop, the ¹H and ³¹P NMR signals of L2 are unusually downfield shifted [36].

Fig. 2.

(A) Schematics of type I and type II tetraloops. (B) Solution NMR structure and schematic of the TTTA MDB (PDB ID: 5GWQ). (C) Solution NMR structure and schematic of the CCTG MDB (PDB ID: 5GWL). Base-base stacking, hydrogen bond and hydrophobic interactions are represented by black, red and blue dashed lines, respectively. (D) The 3′-strand slippage model shows that an MDB forming in the nascent strand during DNA replication can cause a two-repeat expansion. (E) In a segment of three TTTA or CCTG repeats, two adjacent repeats can form competing MDBs which undergo conformational exchange.

2.2. MDB structures in TTTA and CCTG tetranucleotide repeats

Two MDB-like structures formed by cyclic octamers d(pCATTCATT) and d(pTGCTCGCT) were reported by Escaja et al. in 2003 [42]. In each octamer, the 5′ and 3′ terminal residues were covalently linked via a cyclization reaction. Solution NMR structure of d(pCATTCATT) revealed an MDB-like conformer composed of two TTCA type II tetraloops. In 2015, Guo and Lam discovered the MDB structure formed in natural TTTA tetranucleotide repeats [33] which were located in the coding region of the intercellular adhesin C (icaC) gene of Staphylococcus aureus. This MDB contained two TTTA repeats, which folded into two type II tetraloops, respectively. Herein, this structure was called as the TTTA MDB. The solution NMR structure of this MDB showed that T1-A4 and T5-A8 formed Watson-Crick loop-closing base pairs, T2 and T6 folded into the minor groove, whereas T3 and T7 stacked on T1-A4 and T5-A8, respectively (Fig. 2B, left) [43]. The TTTA MDB was highly compact and exhibited multiple loop-loop interactions, including base-base stackings between T1-A4 and T5-A8, base-base stacking and hydrogen bonds between T2 and T6, and hydrogen bonds between T2 and T5/T7. Besides, there were hydrophobic interactions between T3 and T1/T2, and T7 and T5/T6 (Fig. 2B, right).

CCTG tetranucleotide repeats are located in intron 1 of the human zinc finger protein 9 (ZNF9) gene. Expansion from a normal range of ∼11–30 repeats to a pathogenic range of ∼75–11000 repeats is associated with DM2 [27]. A subsequent solution NMR study demonstrated the formation of MDB structure in CCTG repeats [34]. The CCTG MDB was composed of two repeats, which folded into two type II tetraloops, respectively. In this MDB, C1-G4 and C5-G8 formed two Watson-Crick loop-closing base pairs, C2 and C6 folded into the minor groove, whereas the T3 and T7 stacked on the loop-closing base pairs (Fig. 2 C, left). The overall structure of the CCTG MDB was similar to that of the TTTA MDB, but it had the two minor groove residues forming a C·C mismatch instead of stacking with each other [43]. The CCTG MDB also exhibited extensive stabilizing interactions among loop residues, including base-base stackings, hydrogen bonds and hydrophobic interactions (Fig. 2 C, right).

The energy landscapes of the TTTA and CCTG MDBs were further investigated by Wales’s group using molecular dynamic (MD) simulations [44]. Their results showed that the CCTG MDB was the most stable structure while the TTTA MDB was not the lowest-energy structure. Some TTTA MDB-like structures with T1-A4 or T5-A8 Hoogsteen base pair were found to have the lowest energies. Therefore, they suggested that the energy landscapes of the TTTA and CCTG MDBs were mainly determined by the stability of L1-L4 and L1′-L4′ base pairs. This was consistent with the result of an experimental study on how sequence context affected MDB thermostability as discussed in Section 3.1.

Based on the discovery of MDB structures, a 3′-strand slippage model has been proposed to explain TTTA and CCTG repeat expansions in the icaC gene of Staphylococcus aureus variants and the ZNF9 gene of DM2 patients, respectively (Fig. 2D) [33], [34], [45]. During DNA replication, an MDB structure can form in the nascent strand upon a strand slippage at the 3′-end. If the MDB escapes from the proofreading by high-fidelity DNA polymerase during replication, and further escapes from mismatch repair (MMR) proteins during post-replication repair, it will be embedded into the nascent strand and cause a small-scale expansion of two repeats. Multiple small-scale expansions may eventually lead to a large-scale expansion [46].

2.3. Conformational dynamics of MDBs in long tracts of TTTA and CCTG repeats

High-fidelity DNA polymerase and MMR proteins recognize replication errors, such as the mismatch and non-B DNA, using their specific DNA binding sites for the mismatched or unpaired nucleotides [47], [48]. Conformational dynamics of non-B DNA may interfere with the recognition by DNA polymerase and MMR proteins, thus affecting the proofreading and repair efficiency, respectively [49], [50]. Therefore, elucidating the conformational dynamics of non-B DNA in STRs helps to improve our understanding on the molecular mechanisms of repeat expansions and facilitates to develop therapeutic strategies [51].

For a long tract of TTTA repeats, Guo et al. reported an interesting phenomenon that two TTTA MDB structures were preferentially formed in the last three repeats at the 3′-end [45]. The two MDBs underwent a conformational exchange and thus they were called as competing MDBs (Fig. 2E). The competing MDBs were also demonstrated to form in a long tract of CCTG repeats [34]. Occurrence of competing MDBs is possible because of the relatively low thermostability of TTTA and CCTG MDBs (melting temperature T_m of ∼20 °C) [33], [34]. Such conformational dynamics may facilitate the MDB to escape from being repaired by DNA polymerase and MMR proteins.

2.4. MDB structure in ATTCT pentanucleotide repeats

ATTCT pentanucleotide repeat expansions in intron 9 of the ATXN10 gene are associated with SCA10. The numbers of repeats in normal individuals and SCA10 patients are ∼10–22 and ∼850–4500, respectively [28]. For a long time, ATTCT repeats had been considered as DNA unwinding elements which could not form stable secondary structures, and thus ATTCT repeat expansions were explained by replication re-initiations [52] and template switching models [53]. In a recent study, Guo et al. determined the solution NMR structure of two ATTCT repeats which folded into an MDB [35]. This MDB was composed of a TTCTA regular pentaloop and a TTCT/A quasi pentaloop. As defined by Wijmenga et al., the quasi loop has a similar folding geometry as a regular loop but with a backbone discontinuous site (represented by a symbol of “/”) between two adjacent loop residues [39]. In the regular loop formed by T2, T3, C4, T5 and A6, T2-A6 adopted a Watson-Crick base pair, T3 folded into the minor groove, C4 stacked on T2-A6, and T5 was located in the major groove. In the quasi loop formed by T7, T8, C9, T10 and A1, T7-A1 adopted a Watson-Crick base pair, T8 folded into the minor groove, C9 stacked on T7-A1, and T10 stacked well with A1 (Fig. 3 A). The stabilizing forces, including base-base stackings, hydrogen bonds and hydrophobic interactions, in the ATTCT MDB are shown in Fig. 3B.

Fig. 3.

(A) Solution NMR structure and (B) schematic of ATTCT MDB (PDB ID: 6IY5). In the schematic, base-base stacking, hydrogen bond and hydrophobic interactions are represented by black, red and blue dashed lines, respectively. (C) The MDB was preferentially formed at the 5′-end in a long tract of ATTCT repeats.

Intriguingly, the MDB was preferentially formed at the 5′-end in a long tract of ATTCT repeats (Fig. 3 C) [35]. The MDB could also be sustained in the presence of flanking sequences, such as a 5′-T, -CT, -TCT, -TTCT or a 3′-A, -AT, -ATT and -ATTC. The 3′-flanking sequences showed less destabilizing effect on the MDB than 5′-flanking sequences, which explains the preferential formation of an MDB at the 5′-termini in longer ATTCT repeats. The capability of ATTCT MDB escaping from DNA polymerase’s proofreading was investigated by in vitro primer extension assay [35], revealing that the MDB formed at 5-bp or further away from the priming site could escape from the proofreading by DNA polymerase, and eventually retained in the primer and caused a two-repeat expansion. When the MDB was formed at 3- or 4-bp away from the priming sites, it could stall the polymerase for primer extensions. The discovery of ATTCT MDB provides an alternative pathway for repeat expansions in SCA10 apart from the replication re-initiation and template switching models [52], [53]. Very recently, Li et al., have reported that the ATTTT/ATTTC repeats associated with SCA37 and FAMEs formed MDB structures [54] that are similar to the ATTCT MDB, further raising the potential biological significance of MDB structures.

2.5. Biological implications of MDB structures in repeat expansions

The model shown in Fig. 2D elucidates a possible pathway for a small-scale repeat expansion via forming MDB during DNA replication. Multiple small-scale expansions may eventually lead to a large-scale expansion [46]. There are also other possible pathways such as replication re-initiations and template switching to explain large-scale expansions during DNA replication [52], [53]. In addition, repeat expansions in post-mitotic neurons are suggested to occur via MMR-dependent pathways, such as the base excision repair [55]. It will be interesting to explore the roles of MDB in repeat expansions in non-dividing cells.

As the formation of an MDB simply requires two repeats, an interesting question is raised that why short repeats in normal individuals do not expand? For autosomal dominant repeat expansion diseases, the normal allele usually contains short repeats. Once the number of repeats reaches a certain threshold, repeats become unstable and prone to expand. For instance, the threshold numbers of CCTG repeats in the ZNF9 gene associated with DM2 and ATTCT repeats in the ATXN10 gene associated with SCA10 are around 75 and 750 repeats, respectively [27], [28]. Although exact mechanisms of the threshold-dependent genetic instability remain elusive so far, Lee and McMurray have proposed a DNA-mediated model [56]. Upon a strand slippage during DNA replication, the formation of complementary duplex by slipped nascent strand and template, and the formation of non-B DNA by slipped nascent strand will compete. If the rate of forming complementary duplex is slower than the rate of folding a non-B DNA, the non-B DNA will be retained in nascent strand and subsequently lead to repeat expansions. When the number of repeats exceeds a threshold, the rate of forming complementary duplex become slow, raising the frequency of repeat expansions. The MDB structure is small and its folding is expected to be much faster than complementary duplex in longer repeats, so that to cause repeat expansions especially when the number of repeats exceeds a threshold.

3. Effects of sequence context on structure and thermostability of MDBs

Structure and thermostability of non-B DNA are known to depend on sequence context. Based on the originally identified MDB-forming sequences, i.e. TTTA, CCTG, and ATTCT repeats, a series of studies have been conducted to systematically reveal the effects of sequence context on structure and thermostability of MDBs containing (i) two regular tetraloops [57], [58], [59], (ii) two regular pentaloops [60], (iii) a regular and a quasi tetraloops, and (iv) a regular and a quasi pentaloops [60], [61].

3.1. Effects of sequence context on MDBs containing two regular tetraloops

The expandable CCTG repeats and the TTTA repeats share the same pyrimidine-rich sequence context YYYR (Y and R indicate pyrimidine and purine, respectively). This inspired a comprehensive study on 64 sequences of 5′-CYYG CYYG-3′, 5′-CYYG TYYA-3′, 5′-TYYA CYYG-3′ and 5′-TYYA TYYA-3′ (Y = C or T) which would potentially form MDBs containing two regular type II tetraloops [58], [59]. The results have revealed effects of sequence context on the MDB thermostability at neutral pH: (i) a L1-L4 or L1′-L4′ C-G was more stabilizing than a T-A base pair, (ii) a L2 or L2′ thymine was more stabilizing than a cytosine residue, and (iii) a L3 or L3′ thymine was more stabilizing than a cytosine residue. As a result, the 5′-CTTG CTTG-3′ sequence formed the most stable MDB at neutral pH (T_m of ∼35 °C). Comparing to the CCTG MDB, the higher thermostability of the CTTG MDB was attributed to a T2·T6 base pair with two hydrogen bonds, and hydrophobic interactions involving the methyl groups of T2 and T6 (Fig. 4 A) [62].

Fig. 4.

(A) Solution NMR structure and schematic of the CTTG MDB (PDB ID: 6J37). (B) Solution NMR structure of 5′-CTTG CATG-3′ MDB (PDB ID: 6M0C). (C) Solution NMR structure and schematic of CTTTG MDB (PDB ID: 7VM9). In the schematics of MDBs, base-base stacking, hydrogen bond and hydrophobic interactions are represented by black, red and blue dashed lines, respectively. (D) Schematics of 3′-regular and quasi pentaloops. A 3′-quasi pentaloop is more stabilizing than a 3′-regular pentaloop. The backbone discontinuous sites are highlighted by red arrows. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

In hairpins and three-way junctions, type II loops were all found to be YYYR and YYYY sequences [37], [38], [39], while type I loops were exclusively favored by YRYR and YRRR sequences [40], [41]. This is because a pyrimidine L2 of smaller size can be favorably accommodated in the minor groove, whereas a purine L2 of larger size prefers base-base stackings with L1 and L3. Ngai et al. for the first time demonstrated that a type II loop could be formed by YRYR sequence in the context of MDB structure [57]. They systematically investigated eight sequences of 5′-CTTG CRTG-3′, 5′-CRTG CTTG-3′ and 5′-CRTG CRTG-3′ (R = A or G), and revealed that 5′-CGTG CTTG-3′, 5′-CTTG CGTG-3′, and 5′-CTTG CATG-3′ formed MDBs with T_m of ∼25, ∼24, and ∼32 °C, respectively. All these three MDBs contained a CTTG and a CGTG or CATG type II tetraloops. Solution NMR structure of the MDB formed by 5′-CTTG CATG-3′ shows that A6 was driven into the minor groove via base-base stacking with T2 (Fig. 4B). In addition, a L2′ adenine stacked better than a L2′ guanine to bring a higher thermostability.

3.2. Effects of sequence context on MDBs containing two regular pentaloops

A recent study showed that an MDB can also be formed with two regular pentaloops, e.g. 5′-CTTTG CTTTG-3′ [60]. The solution NMR structure of the MDB containing two CTTTG repeats reveals asymmetric loop folding patterns (Fig. 4 C, left). In the 5′-loop, C1-G5 formed a Watson-Crick base pair, T2 folded into the minor groove, T3 stacked on C1-G5, and T4 was located in the major groove. In the 3′-loop, C6-G10 adopted a Watson-Crick base pair, T7 folded into the minor groove, T9 stacked on C6-G10, and T8 stacked on T9. The effect of sequence context on MDBs composed of two regular pentaloops was examined using 5′-CTTTG CTTTG-3′ as a model sequence, revealing that a L2/L2′, L3/L3′ or L4/L4′ thymine was more stabilizing than a cytosine residue [60], [63], and this was attributed to the extensive hydrophobic interactions involving the methyl groups of thymine residues (Fig. 4 C, right). In addition, MDBs could sustain a L2′, L3′ or L4′ purine residue (A or G) with T_m of ∼31–36 °C. A L4′ purine was more stabilizing as it stacked better with L1′-L5′ base pair than a L4′ pyrimidine [60].

3.3. Effects of sequence context on MDBs containing a regular and a quasi tetraloops

Quasi loops provided structural flexibility to DNA three-way junctions [64], [65]. The effect of a quasi tetraloop on MDBs was systematically examined by creating different backbone discontinuous sites on the CTTG MDB [61]. NMR results showed that backbone discontinuous site between L3′ and L4′ resulted in an MDB comprising a regular CTTG and a quasi CTT/G tetraloops, demonstrating the possibility of including a quasi tetraloop into the MDB.

Recently, the effect of a quasi pentaloop on MDBs has also been studied via introducing backbone discontinuous sites on the CTTTG MDB [60]. A backbone discontinuous site between L3′ and L4′, or between L4′ and L5′ allowed to form stable MDBs containing a regular CTTTG loop and a quasi CTT/TG or CTTT/G loop, and the T_m values of these two MDBs were ∼42 and 34 °C, respectively. These values were higher than that of the CTTTG MDB (T_m of ∼32 °C), and this was attributed to a better stacking between L3′ or L4′ with L1′-L5′ base pair in the quasi loop (Fig. 4D).

4. Effects of chemical environment and nucleobase modification on MDBs

4.1. Effect of pH on structure and thermostability of the CCTG MDB

The structure and stability of non-B DNA structures can be affected by pH due to the protonation of adenines and cytosines in an acidic microenvironment [66], [67]. For the CCTG MDB, its structure and thermostability changed drastically under different pH conditions [68]. At pH 7, the CCTG MDB contained C1-G4 and C5-G8 Watson-Crick base pairs, and a C2·C6 mispair with predominantly a single hydrogen bond [43]. At pH 5, protonation of C1 and C2 lead to the formation of C1⁺-G4 Hoogsteen base pair and a three-hydrogen-bond C2⁺·C6 mispair (Fig. 5 A). The terminal residue C1⁺ showed cation-π interaction with G8, whereas the C2⁺·C6 mispair contained three hydrogen bonds which undoubtedly provided a substantial stabilization to the MDB. The structural changes of CCTG MDB from pH 7 to pH 5 lead to an increase of T_m by ∼24 °C. While the CCTG MDB has been proposed to a structural intermediate that may cause CCTG repeat expansions in DM2 [34], the genetic instability of CCTG repeats in DM2 patients might be influenced by pH variation in the chemical microenvironment.

Fig. 5.

(A) Solution NMR structures of CCTG MDB at neutral pH (top left), at pH 5 (PDB ID: 7D0Z) (bottom left), with abasic site modification (PDB ID: 6M6K) (top right) and with cytosine methylations (PDB ID: 7D0Y) (bottom right). (B) Solution NMR structures of TTTA MDB without modification (left) and with adenine methylation (PDB ID: 7E4E) (right). Closeup views of the loop-closing base pairs are highlighted by grey-dashed boxes whereas those of the minor groove mispair are highlighted by grey boxes.

4.2. Effect of an abasic site on structure and thermostability of MDBs

Abasic site, which lacks the nucleobase comparing to natural nucleotides, is a frequent DNA lesion occurring at ∼10 000 lesions per human cell per day [69]. It appears to be mutagenic by causing base misincorporation, frameshift and repeat expansions [70], [71]. The abasic site has been shown to destabilize DNA duplexes and G-quadruplexes [72], [73]. When an abasic site was introduced between two tetraloops of an MDB, it did not disturb the overall folding geometry, but significantly enhanced the thermostability of the MDB [74]. Specifically, the T_m of the 5′-CTTGXCTTG-3′ (X represents an abasic site), 5′-CCTGXCCTG-3′ and 5′-TTTAXTTTA-3′ MDBs significantly increased by ∼14, 15 and 19 °C than those of the CTTG, CCTG, and TTTA MDBs, respectively. The stabilization was attributed to an increased backbone flexibility upon introducing an abasic site. For instance, as revealed from the solution NMR structure of the 5′-CCTGXCCTG-3′ MDB (Fig. 5 A), the backbone flexibility allowed the MDB to form a C·C mispair with an additional hydrogen bond and Na⁺-mediated electrostatic interactions. Moreover, the abasic site formed additional hydrophobic contacts with neighboring nucleotides using its 2′-methylene groups. In general, abasic site plays a role in the stabilization of MDBs.

4.3. Effect of cytosine methylation on structure and thermostability of MDBs

DNA methylation is an epigenetic mark that regulates gene expression [75]. 5-methylcytosine (mC) is the most abundant DNA methylation in mammalian cells and accounts for ∼3% of cytosines in genomic DNA [69]. It has been reported that mC stabilizes a variety of non-B DNA structures including hairpins [76], triplexes [77], i-motifs [78], and G-quadruplexes [79].

Effects of mC at the L1 and L1′ positions was revealed from the 5′-mCTTGXmCTTG-3′ and 5′-CTTGXCTTG-3′ MDBs at pH 7 [80]. The former had a higher T_m by ∼11 °C, which was because mC1 and mC5 brought about favorable hydrophobic interactions and more stable loop-closing base pairs. Effect of mC at the L2 and L2′ positions was examined using 5′-CmCTGCmCTG-3′and 5′-CCTGCCTG-3′ at pH 5 [80]. The former had a higher T_m by ∼17 °C, which was attributed to the mC2⁺·mC6 mispair with additional hydrophobic interactions with loop-closing base pairs, as well as favorable electrostatic interactions and improved hydrogen bond geometries (Fig. 5 A). At present, the L1/L1′/L2/L2′-methylated 5′-mCmCTGmCmCTG-3′ MDB at pH 5 remains the most stable MDB ever reported (T_m of ∼69 °C). In short, 5mC enhanced the thermostability of MDBs by enabling more stable base pairs, hydrophobic contacts and improved base-base stackings. The stabilization effect might affect the propensity of CCTG repeat expansions in DM2.

4.4. Effect of adenine methylation on structure and thermostability of MDBs

N⁶-methyladenine (m⁶A) is the most prevalent methylated base in prokaryotes with gene regulatory functions, though its abundance in eukaryotes remains a debate [81]. The occurrence of m⁶A has been reported to destabilize duplexes [82] and G-quadruplexes [83]. N¹-methyladenine (m¹A) is another methylated adenine caused by alkylating agents, which can block DNA replication if not repaired [84]. It has been reported that the substitution of an adenine by m¹A in the duplex turns the T-A Watson-Crick base pair to T·m¹A Hoogsteen base pair and destabilizes the duplex [85].

The effect of adenine methylation on MDBs was investigated by substituting A4 in the 5′-TTTATTTA-3′ MDB with m⁶A or m¹A [86]. The m⁶A-substituted sequence 5′-TTTm⁶ATTTA-3′ was found to form two MDB conformers. The major MDB conformer had a T1·m⁶A4 Hoogsteen base pair while the minor MDB conformer exhibited a T1-m⁶A4 Watson-Crick base pair. The T_m of the major MDB conformer was ∼9 °C lower than that of the unmodified TTTA MDB. For the m¹A-substituted sequence 5′-TTTm¹ATTTA-3′, it adopted a single MDB conformer with a T1·m¹A Hoogsteen base pair (Fig. 5B), and its T_m was ∼4 °C higher than that of the unmodified TTTA MDB. This represents the first report of the stabilization effect of m¹A on a DNA secondary structure.

5. Summary and outlook

The discovery of MDB structure has provided new insights into the genetic instability of pyrimidine-rich repeats associated neurodegenerative diseases such as DM2 and SCA10. Efforts have also been made to understand the effects of sequence context, nucleobase modification and chemical environment on the structure and thermostability of MDBs. However, our understanding on this relatively new type of non-B DNA is still far from mature. First, the sequence criteria for MDB have not been well established. For instance, as all the reported MDBs have C-G or T-A loop-closing base pairs, it is curious if a mismatch can serve as the loop-closing base pair. Second, mapping the MDB-forming hotspots in the human genome and studying their correlations with biological functions need to be considered in the future work. Finally, small-molecule ligands designed for probing or visualizing MDB structure in vivo will definitely facilitate us to explore the biological significance of MDB. Although the biological significance of this new type of non-B DNA remains elusive at the current stage, an illustration of its structures and conformational dynamics may provide useful information for therapeutic development in the future.

CRediT authorship contribution statement

Yuan Liu: Writing – original draft, Review & editing, Funding acquisition. Liqi Wan: Writing – original draft, Review & editing. Cheuk Kit Ngai: Writing – original draft, Review & editing. Yang Wang: Writing – review & editing. Sik Lok Lam: Conceptualization, Supervision. Pei Guo: Conceptualization, Supervision, Resources, Funding acquisition, Writing – review & editing.

Declaration of Competing Interest

The authors declare no competing financial interests.