Potential in vivo roles of nucleic acid triple-helices

Abstract

The ability of double-stranded DNA to form a triple-helical structure by hydrogen bonding with a third strand is well established, but the biological functions of these structures remain largely unknown. There is considerable albeit circumstantial evidence for the existence of nucleic triplexes in vivo and their potential participation in a variety of biological processes including chromatin organization, DNA repair, transcriptional regulation and RNA processing has been investigated in a number of studies to date. There is also a range of possible mechanisms to regulate triplex formation through differential expression of triplex-forming RNAs, alteration of chromatin accessibility, sequence unwinding and nucleotide modifications. With the advent of next generation sequencing technology combined with targeted approaches to isolate triplexes, it is now possible to survey triplex formation with respect to their genomic context, abundance and dynamical changes during differentiation and development, which may open up new vistas in understanding genome biology and gene regulation.

Introduction

Nucleic acid triple-stranded structures, also called triplexes, are complexes of three oligonucleotide strands made of either RNA or DNA.^1–3 Over the last decades triplexes have been implicated in a range of cellular functions such as transcriptional regulation, post-transcriptional RNA processing, modification of chromatin and DNA repair. Evidence underpinning these implications is, however, primarily based on in vitro experiments leaving the in vivo existence and function of triplexes a mystery yet to be revealed. Important pieces to the puzzle have been added by recent work on new methods for triplex detection,⁴ the discovery of new triplex-unwinding helicases,⁵ as well as the increasingly apparent biological roles played by non-coding RNAs⁶ that suggest the involvement of triple-helices.^7–10 With the advent of high-throughput sequencing of RNA and DNA we now have the technology to map the genomic locations of RNA-DNA triplexes and to begin unraveling their functions.

Here, we critically review the body of evidence that suggests endogenous triple-helices form in vivo and implicates them in diverse biological processes. In the final section, we suggest how modern sequencing technology can be coupled with computational approaches to unravel the prevalence and function of triplex structures in living cells.

Triplex Formation and Pairing Rules

Triplex-helices are formed by sequence-specific binding rules that are distinct although conceptually akin to the familiar Watson-Crick base-pairing rules. The triplex-formation involves a double-stranded nucleic acid such as duplex DNA and a single-stranded nucleic acid such as RNA. The ‘triplex-forming oligonucleotide’ (TFO) of the single-stranded nucleic acid binds in the major groove of the targeted duplex through sequence-specific recognition of a polypurine•polypyrimidine sequence. Variations that bind in the minor groove have been reported¹¹ but insights into these three-stranded variants are very limited. We therefore focus on major groove triple-helical structures unless stated otherwise.

TFOs that bind in the major groove of the duplex form Hoogsteen hydrogen bonds with the purine-rich strand. Two configurations are possible: (1) the Hoogsteen configuration promotes binding of the third strand in a parallel orientation to the polypurine strand of the duplex, while (2) anti-parallel orientation of the TFO is established by reverse Hoogsteen bonds (Fig. 1).

Figure 1.

Parallel and anti-parallel oriented triple-helix motifs. There are three basic triple-helix motifs: [T,C], [G,A] and [G,T]. The [G,T] motif can bind in both orientations to the duplex while [T,C] binds parallel and [G,A] binds anti-parallel to the purine tract of the duplex. The inset shows a cartoon of a triple-helical structure derived from a NMR-structure contained in the protein database (PDB-id:1BWG).

The stability of triple-helices depends heavily on the availability of hydrogen acceptor and donor groups from the third base, as well as its steric features. There are basically three “motifs” for stable triple-helix formation, all of which permit two stabilizing hydrogen bonds between the base in the third strand and the purine in the duplex. These motifs are referred to by the bases of the third strand that participate in the formation: [T,C] (pyrimidine motif), [G,A] (purine motif) and [G,T] (see Appendix). Of course, “T” refers to uracil when the third strand is made of RNA.

In principle, there are eight possible ways to form a triple-helix when each of the three strands is either made of RNA or DNA. Thermodynamic in vitro studies have shown that the RNA-DNA•DNA and DNA-DNA•DNA triple-helical formations are most stable (where “-” refers to Hoogsteen/reverse Hoogsteen and “•” refers to Watson-Crick binding),^12–16 although the nucleic acid backbones were found to have a profound effect on the formation efficiency for different triplex motifs, and RNA may not be able to participate in reverse Hoogsteen structures (reviewed in refs. ¹⁷ and ¹⁸). In addition, triplexes are distinguished according to whether the third strand is tethered to the duplex (intramolecular triplex) or an independent molecule (intermolecular triplex).

Thus, triplex formation represents an additional mechanism for the cell to target nucleic acid sequences besides duplex hybridization and sequence-specific protein binding. Importantly, it permits the single-stranded nucleic acid (presumably mainly RNA) to bind the targeted duplex structure (RNA or DNA) without requiring it to unwind first. This unique ability has considerable biotechnological potential and has been extensively studied for use in such applications as modulation of transcription^19–23 and site-directed recombination as well as mutagen delivery.^24–26

Evidence for the Existence of Triple-Helical Structures in vivo

Triplex-specific antibodies and dyes bind nuclear triple-helical structures.

The existence of triplex structures in the nucleus of insects, nematodes and mammals has been suggested by immunofluorescence using triple-helix specific monoclonal antibodies.^27–34 The antibodies recognize distinct triplex epitopes as indicated by different staining patterns on polytene chromosomes of Drosophila melanogaster.^30,34 Among the most compelling evidence for the existence of triplexes in vivo is immunodetection by triple-helix specific antibodies in human cell nuclei,³¹ which detected an abundance of triplex signatures (Fig. 2).

Figure 2.

Immunofluorescent staining of cell nuclei using triplex-specific monoclonal antibodies. (Figure reproduced from ref. 31 with kind permission from the Springer Science + Business Media).

Although the antibodies used in all but one of the above studies were prepared against DNA-DNA•DNA triplexes,^27,28 it has subsequently been shown that these antibodies have an even stronger affinity to RNA-DNA•DNA triplexes.^9,29,35 The signals observed in vivo may hence largely reflect RNA-DNA•DNA triplexes.

Complementary evidence has recently been put forward using Thiazole Orange (TO), a dye that preferentially detects triplexes under in vivo conditions. TO treated U2OS cells reveal distinct staining patterns in cell nuclei, which are almost opposite to that of the duplex DNA binding dye DAPI,⁴ but may merely reflect non-specific staining of the nucleolus by TO. However, TO staining signatures also extend to the cytoplasm. A preliminary study showed no evidence of the dye interacting with proteins promoting the idea that triplexes may also occur in the cytoplasm possible in the form of RNA structural motifs.⁴ Additional investigations into the binding specificity of these indicators are required to resolve all doubts concerning the origin of in vivo staining patterns from non-triplex-specific binding.

Proteins specifically recognize triple-helical structures.

The existence of endogenous proteins that recognize and act specifically on triplexes further supports their existence in the cell. Table 1 lists proteins that have been observed to bind to triplexes in vitro or that have a functional role specific to triplexes.

Table 1.

Proteins detecting or binding nucleic triple-helices

Protein (origin)	Comments	Nuclear localization	Evidence	Affinity to	Ref.
GAGA factor (fruitfly)	The GAGA factor (GAF) in Drosophila appears to perform different functional roles in distinct genomic locations (reviewed in ref. 196). in fact, its functional role seems to be subject to the structure of the targeted heterochromatin. The target motifs of GAF, (GA)n repeats, exist in several promoters and are able form triple-helical structures. interestingly, GAF shows affinity to bind triple-helical structures in vitro with a similar specificity and affinity as duplex DNA. Thus triplex formation may account for the different functional roles GAF can adopt, which includes gene regulation but also alteration of chromatin structure.197,198	Yes	EMSA analysis of interaction between triplex probe and protein.	pyrimidine-motif triplexes forming at (GA)n repeats	47
Stm1p (yeast)	Stm1p preferentially associates with ribosomal RNA but can also interact with DNA, especially subtelomeric sequences199 and is thought to be involved in mitosis. its first 113 amino acids comprise the minimal domain that is capable of binding purine triplexes but unable to bind duplex DNA.200	Not reported	EMSA of yeast whole-cell extract using purine motif probe	purine-motif triplexes	48
CDP1p (yeast)	CDP1p thought to be involved in chromosome segregation and histone displacement.201	Yes	Southwestern library screening with subsequent EMSA analysis	purine-motif triplexes	49
Tn7-encoded protein (Bacterial transposon)	Tn7 detect triplex DNA, which leads to specific insertion of the transposon adjacent to both intra- and intermolecular pyrimidine motif triplexes	Transposon	Chemical footprinting	intramolecular triplexes at mirror repeat sequences	202
IF proteins: vimentin, GFAB, desmin (mouse)	Vimentin, GFAB and desmin, recognize and bind to repetitive DNA that preferably adopts intramolecular triplex H-DNA structure. Vimentin, which is localized in the cytosol as well as in the nuclear matrix,113,203 is possibly identical to a protein with a molecular mass of ∼55 kDa found by earlier studies.204,205	Not reported	Vector plasmid constructs containing IF-purified DNA fragments form triplexes as shown in band shift assays and prevention of nucleolytic cleavage and chemical modification.	intramolecular triplexes at mirror repeats (GA)n	44
Alternative loricrin reading frame (human)	Loricrin is a major keratinocyte cell envelope protein.	Yes	South-Western screening of a human keratinocyte cDNA expression library using artificial triplex probe.	purine-motif triplexes and pyrimidine-motif triplexes (weaker)	45
MBP-LOR3ARF (human)	DNA-binding protein initially found by South-Western screening of a human keratinocyte cDNA expression library.45		EMSA analysis	pyrimidine-motif triplexes	206
Orc4 subunit 4 (human)	Orc4 subunit 4 prefers triplex DNA to duplex or single stranded DNA suggesting that triple-stranded structures are important for origin organization and activity during DNA replication.	Yes	EMSA analysis using purified HsOrc4	(T-A•T)n triplets	46
55 kDa (human)	No binding to the corresponding single stranded DNA (TFO) or the duplex DNA	Yes	Affinity purification from HeLa nuclear extracts using artificial triplexes	(T-A•T)n triplets, pyrimidine and (weaker) purine triplexes	204, 205
100 kDa, 60 kDa, 15 kDa (human)	No recognition of the corresponding duplex DNA or pyrimidine triplex by the proteins.	Yes	Affinity purification from HeLa nuclear extracts using artificial triplexes (GT-motif) with subsequent EMSA and Southwestern blotting.	purine-motif triplexes	207
hnRNP (K, L, A2/B1, e1, I) (human)	Detected proteins show affinity to triple-helical structures but not duplex or single-stranded DNA. Possibly identical to the proteins found by Musso et al.207	Yes	2D-electrophoresis and mass spectrometry using artificial triplex probes.	various triplexes	43
Zinc-dependent protein (human)	The zinc-dependent protein present in nuclear extracts from different cell types was found to bind to DNA sequence prone to adopt triple-helical structure and to recognise structure-specific rather than sequence-specific features.	Yes	EMSA and DNase I study using the (GAA)n repeat from the first intron of frataxin alleles	(GAA)n repeats	208
High mobility group (HMG) proteins (mouse)	HMG box proteins promote and stabilize triplex formation as shown by time course PAGE, though an unusual purine-motif triplex was suggested. in fact, the small tetra peptide PKRW containing the conserved sequences of the HMGB1 DNA-binding domain were found to be able to recognize triple-helical structures.	Yes	Recombinant proteins incubated with triplex probe analyzed by time course PAGE and DNase I protection assays.	purine-motif triplexes forming at (GGA)11 repeats	209–211
RecQ helicases	DHX9 and SV40 large antigen and helicase are two specimen that can melt triplex in a 3′-5′ polarity with respect to the displaced strand requiring a 3′ overhang on the third strand and ATP.	Yes	Strand displacement assay using recombinant helicase incubated with DNA substrates and detected using PAGE and dimethyl sulfate footprinting.	Melting of various forms of triplexes	5, 36, 38, 212
Nucleotide excision repair and mismatch repair proteins (human)	XPA-RPA repair complex participates in the nucleotide excision repair resolving DNA damage. XPA was found to bind psoralen-crosslinked TFO-induced triplexes only in conjunction with RPA but contributes to specificity in detection of triplex-induced helical distortions via RPA.	Yes	EMSA studies using proteins expressed in E. coli and ChIP.	Anti-parallel motif triplexes	213–215

Included among these are helicases of the RecQ family, which actively unwind triplexes in a 3′→5′ direction.^5,36–39 In the case of SV40 large T-antigen helicase, as well as of DHX9 helicase a flanking 3′ end on the triplex is required. Otherwise, the triplex inhibits even the unwinding of the duplex by the helicase, which has implications for DNA replication.⁴⁰ In contrast, the triplex-melting property of the single-strand DNA affine human replication protein A (RPA) could be due to competition for binding to the third (single) strand under equilibrium conditions.⁴¹ The fact that a comparable single-stranded DNA-binding protein poorly destabilizes triplex structures, however, suggests that RPA specifically detects and processes triplexes as previously suggested.⁴² This has been further substantiated by immunodetection of triplex signatures in HeLa cells after siRNAs-induced suppression of RPA.⁴¹

A range of other triplex-binding proteins have been found in human,⁴³ other mammals,^44–46 flies⁴⁷ and yeast^48,49 (see Table 1). They belong to chromatin-associating protein families such as heterogeneous ribonucleoproteins (hnRNP), cytoplasmic type III intermediate filament (IF) proteins, transcription factors (TFs), high mobility group (HMG) box proteins as well as proteins involved in the cell cycle and DNA repair. While this suggests that cells throughout the eukaryotic kingdoms of life possess proteins able to recognize and process nucleic triple-helices it may as well simply reflect the preference of some (especially cationic) proteins for the higher negative charge densities of such multi-stranded nucleic acids structures.

Abundance of putative triplex target sites in genomes.

Polypurine regions spanning more than 15 nucleotides, which constitute putative ‘triplex target sites’ (TTSs), are overrepresented in both prokaryotic and eukaryotic genomes.^50–55

A special class of polypurine stretches that have the characteristics of mirror repeats are enriched in the yeast as well as the human genome and common in many other genomes.^55,56 Mirror repeats are especially interesting because they can engage in inter- and intramolecular triplex formation. This special type of intermolecular triplex could be formed by an RNA being transcribed from the repeat and binding back to the duplex, creating a perfectly matched triplex (Fig. 3).⁵⁷ Intramolecular DNA-triplexes (H-DNA), on the other hand, consist of three strands of DNA where one strand folds back onto the duplex leaving an unpaired single strand (reviewed in ref. 58). Evidence that such tracts indeed form triple-helices was provided by in vitro assays using S1-nuclease, which preferentially targets single-stranded DNA, involving however a rather low pH.⁵⁹ Using RNA-FISH, a recent study by Zheng et al. show a heterogeneous population of (GAA)_n repeat containing ncRNA to localize in distinct subnuclear domains, which likely resemble genomic (GAA)•(TTC) repeats, possibly forming a triplex at these repeats or targeting the single-strand of an H-DNA triplex already in place. Mirror repeats capable of forming H-DNA occur in promoters as well as coding regions of several disease-involved genes (reviewed in ref. 60). For Friedreich ataxia it has been demonstrated that the formation of such triple-helical structures are induced by the expansion of (GAA)_n repeats, which causes the transcriptional silencing of the frataxin gene.⁶¹

Figure 3.

Putative triple-helical structures formed by palindrome RNA or DNA mirror repeats. The upper part illustrates how an RNA transcript from mirror repeat DNA can potentially result in a feedback signal via formation of a triplex at the same site. The lower part illustrates how mirror repeat DNA can form H-DNA. In both cases simple repeats easily comply with the respective constraints applied.

Triple-helical structures in RNA.

Intramolecular triple-helices in RNA are known to contribute to RNA folding and tertiary structure stability.^62,63 These pseudoknot structures are also crucial for enzymatic function such as catalytic activity^64,65 or ribosomal frameshifting during translation.⁶⁶ In fact, the HIV virus shows absolute reliance on an ribosomal frameshift event during translation,⁶⁷ and mutation analysis of the pseudoknot sequence found that intramolecular triple-helix formation in both the minor as well as the major groove of mRNA is essential in −1 ribosomal frame-shifting.^68,69 These examples indicate that the cell employs intramolecular triplex-formation in RNA, which likely extends to intermolecular RNA-triplexes and triplexes made of both RNA and DNA.

Evidence for the Function of Triple-Helical Structures

Triplex formation regulates gene transcription.

The possibility of triplex formation in vivo raises the question of their functional roles. Computational studies show that putative triplex target sites are enriched upstream of eukaryotic genes,^54,70 suggesting a possible role of triplexes in transcriptional regulation. Supporting evidence for this functional role is provided by several biological studies investigating triplex formation at specific gene loci as reviewed by Dyke.²²

One of the most compelling pieces of evidence for a regulatory role for TFO-induced triplex formation in vivo is the regulation of the dihydrofolate reductase (DHFR) gene in U2OS cells.⁷ Here, a non-protein-coding transcript (ncRNA) from a minor promoter interacts in a sequence-specific way to form a triplex within the downstream major promoter, which is responsible for 99% of the transcription of the DHFR gene. The transcript also interacts with TFIIB and causes dissociation of the pre-initiation complex from the major promoter resulting in promoter-specific transcriptional repression of the DHFR gene. While triplex formation was supported by accompanying in vitro assays, the transcriptional inhibition observed in vivo might also be attributed to formation of stable DNA-RNA duplexes¹² possibly at the G-rich sequences in the DHFR promoter. G-rich sequences on the sense strand have been shown to stall transcription in vitro⁷¹ and in vivo⁷² via interactions of the nascent transcript with the DNA template. The means of stable RNA•DNA duplex formation involving a transcript of an upstream promoter remains to be investigated.

If triplex-formation is indeed the principle cause of the silencing of DHFR in the above example, then a plausible mechanism is that triplex formation shields duplex DNA from duplex-targeting proteins such as transcription factors.^73–75 Such a mechanism was shown to downregulate the Ets² gene in vitro⁷⁶ as well as in prostate cancer cells.⁷⁷ Here, specifically designed TFOs overlap the DNA binding site of the transcription factor Sp1 in the Ets2 promoter, restricting Sp1 from binding when the triplex is formed. Another example is the bcr/abl locus, where TFO-induced triplex formation was demonstrated by EMSA assays and its downregulating effect on bcr/abl transcription as well as cell growth was shown in in vivo cell studies.¹⁹ Similarly, increased mortality of human breast cancer cells was shown using specifically designed polypurine hairpins that form intermolecular triplexes at polypyrimidine target DNA sites.²³ Aside from occluding DNA-binding sites triplexes could also affect the function of DNA-binding proteins. There is evidence that the transcription factor GAGA adopts different functional roles depending on the conformation of the targeted DNA, which overlaps with putative TTSs and may hence form triplexes (see Table 1).

Interestingly, several TFs have been found to bind RNA as well as DNA.⁷⁸ This suggests that an RNA-bound TF could home in on its target loci guided by triplex-specific formation between the RNA and the targeted duplex DNA. This concept has been exploited in biomedical applications for the site-specific delivery of mutagen agents by means of intermolecular triplex formation between a TFO and the targeted duplex DNA.^24,79 It should be noted that in these biomedical applications the “TFO” is often made of DNA or of an analogous molecule such as a peptide nucleic acid.²⁶ Similarly, RNA-bound molecules could facilitate triplex-formation as an anchor point to remain and act in the spatial proximity of a specific genomic context. Proof-of-principle was obtained using an artificial TFO made of DNA that is physically linked to the SRE enhancer, which accommodates binding of the TFs SRF and ELK-1.⁸⁰ Here, the SRE-TFO successfully recruits the TFs via the linked enhancer and forms a triplex at its target locus, which otherwise shows no affinity for SRF and ELK-1. Hence, these proteins might be able to perform their function in the vicinity of the targeted locus such as modifying the epigenetic landscape or altering gene transcription. The fact that many proteins are known to interact with RNA^81,82 suggests that the cell may employ such a cargo-delivery mechanism.

Target specificity.

Any loci specific targeting mechanism requires a unique, non-ambiguous address. Computational studies indicate that large numbers of putative TTSs occur exactly once in the human as well as in the mouse genome⁸³ suggesting that these sequences generally fulfil the requirement of site-specificity.

Mirror repeats able to form H-DNA are also non-randomly distributed in the genome and are associated with genes involved in cell trafficking and communication.⁸⁴ Indeed, formation of H-DNA on polypurine•polypyrimidine sequences was shown by S1 nuclease assays implicating such sequences as cis-acting transcriptional regulators (reviewed in ref. 59). The proposed formation of H-DNA on (GC)_n repeats and a resulting change in gene expression has been reported to strongly depend on the length of the repeat and the induction of local superhelical strain.⁷⁴ The applied assays do, however, neither rule out the possible formation of G-quadruplexes at these repeats, which have been linked to a similar phenotypic behavior^85,86 nor the formation of RNA-DNA hybrids and R-loops.⁷¹ In fact, in vitro studies indicated that neither H-DNA nor G-quadruple formation is a pre-condition for the transcriptional arrest of T7 RNA polymerase at G-tracts.⁷¹ (GAA)_n repeats, which do not comply to G-quadruple formation rules, have been shown to stall replication in vivo⁸⁷ as well as cause transcriptional arrest.^88,89 So-called “suicidal” mirror-repeats were shown to cause permanent or temporary transcriptional arrest in vitro. Here, the non-template strand displaced by the polymerase during DNA synthesis folds back to the duplex prompting the formation of H-DNA.⁹⁰ This transcriptional interference has been well studied in the genomic locus of the c-myc oncogene (reviewed in ref. 91). c-Myc participates in several pathways determining the cellular fate, and its abnormal expression has been associated with cancer.⁹² The H-DNA-forming sequences at this locus map to translocation breakpoints, which can activate or disrupt specific genes in its proximity bypassing or undermining cellular control mechanisms.^93,94

Endogenous triplex-forming microRNAs interfere with viral replication.

Triplex formation may not be limited to the cell nucleus. A recent study by Kanak et al. implicates cytosolic triplex formation in viral defence of eukaryotic cells. The authors showed that HeLa-CD4⁺ as well as T-lymphotic cells are able to resist HIV-1 infection when expressing specific endogenous microRNAs that are able to form stable triplexes with DNA motifs from HIV-1, whereas cells expressing control microRNAs lacking triplex-formation capability show no resistance to HIV-1. Importantly, the proposed binding of the investigated microRNAs incorporates a substantial portion of triplets lacking at least one of the two triplex-stabilizing hydrogen bonds, which is contrary to models obtained from thermodynamic in vitro studies.⁹⁵ More research is clearly needed to substantiate cytosolic triplex formation in vivo and reveal the cytosolic determinants involved in this process.

Involvement of triple-helices in RNA processing.

Triple-helical structures may also be involved in post-transcriptional processing of RNAs. Many proteins regulating translation and post-transcriptional events such as RNA splicing and editing associate with double- or single-stranded RNAs through RNA recognition motifs.^81,96 The formation of pure RNA triple-helical structures⁹⁷ provides another RNA motif to be considered as either the target of proteins or as a decoy. Some circumstantial evidence suggests that triplex formation prevents binding and activation of RNA-dependent protein kinase (PKR) and possibly other nucleotide-dependent proteins in vitro.⁹⁸ Active PKR inhibits protein synthesis in higher cells and leads to growth inhibition in response to viral infection,⁹⁹ and binding to double-stranded RNA activates PKR in vitro.¹⁰⁰ The formation of a triple-helix between duplex RNA and a TFO can efficiently inhibit the binding of PKR and, subsequently, its activation.⁹⁸ In the eukaryotic cell, endogenous RNA may hence adopt triplex rather than duplex form to avoid triggering cellular responses involving proteins that monitor foreign (viral) duplex RNA. It also suggests that intermolecular triplex-formation may play a role in RNA-RNA interactions in other contexts, which may be responsible for the cytosolic staining patterns obtained immuno- and fluorescence-detection studies.^4,31

HnRNP proteins, of which five members show specificity for pyrimidine-rich DNA, also bind to triple-helices⁴³ implicating triplex formation in mRNA splicing. HnRNP A2/B1 and L have been linked to pre-mRNA processing and have been found in the splicosome.¹⁰¹ HnRNP I in particular has been found to bind to polypyrimidine tracts of introns.^102,103 HnRNP K and E1 bind predominately polyC single-stranded DNA and may hence target the unbound strand of H-DNA or G-quadruplexes as well as be responsible for the single-strand binding observed in the study.¹⁰⁴ Polypyrimidine tracts have been associated with alternative splicing events¹⁰⁵ raising the question to what extent TFOs able to bind to their antisense polypurine may be involved in protein recruitment and alternative splicing mechanisms. In fact, it has been demonstrated in vitro that a polypurine stretch with alternating A and G can stimulate the splicing of the surrounding intron(s) by modulating the binding of polypyrimidine tract-binding protein.¹⁰⁶ Furthermore, recent work indicates that self-splicing of precursor RNA in group I and II introns involves enzymatic triplexes (reviewed in ref. 64). U2 and U6 RNAs, which form a ribozyme complex that is also part of the spliceosome, and which are related to the self-splicing precursor RNA, can carry out RNA-based catalysis.¹⁰⁷ Importantly, sequences involved in the first splicing step seem to be recognized by the snRNAs without the aid of additional proteins.¹⁰⁸ It remains to be shown if U2 and U6 RNAs also rely on catalytic triplex formation.

Triplex-mediated chromatin structure, organization and epigenetics.

Chromosome organization and architecture plays a central role in the regulation of differentiation and development, and has strong effects on DNA replication,¹⁰⁹ transcription and splicing.¹¹⁰ The formation of H-DNA provides contact points that may participate in chromatin organization possibly through ncRNAs, which in turn may interact with the nuclear matrix or nuclear matrix-associated proteins.¹⁰ Repeat regions in particular may be capable of forming links between different regions of DNA via triplex formation and in vitro studies on plasmid DNA substantiate the formation of such DNA loops.¹¹¹

Proteins specifically binding triple-helices may orchestrate chromatin organization. Intermediate filament (IF) proteins are an important structural feature in eukaryote cells that can bind to triplexes and also to other higher-order structures such as G-quadruplexes (reviewed in ref. 112). One IF protein, Vimentin, which is localized in the cytosol as well as in the nuclear matrix,¹¹³ is known to attach to several cytosolic organelles potentially positioning them in fixed locations and may perform similar functions in the nucleus. Triplexes may serve as anchor points or be used as landmarks to recruit chromatin regions to nuclear compartments.

A recent study links a ncRNA to epigenetic modifications in gene promoters of non-coding genes.⁸ Here, a TFO-directed triplex has been suggested to mediate the recruitment of DNA methyltransferases, and in particular the de novo methyltransferase Dmnt3b, to gene promoters and hence regulate the methylation status of DNA. DNA methylation is known to alter the expression of nearby genes¹¹⁴ and plays an essential part in gene imprinting and cell differentiation.¹¹⁵ While RNase H, EMSA and psoralen-based assays demonstrate that the binding of the ncRNA to the promoter is neither via formation of RNA-DNA hybrids nor due to mediating proteins, further investigations into the proposed triple-helical formation are required. In particular, the observed deviation from the common triplex model needs to be addressed by characterising the triplex regarding its binding motif (see Appendix) or by evaluating alternative models such as R-triplexes, where the third strand binds parallel to an identical nucleotide sequence in the duplex.¹¹⁶ The authors also report the formation of heterochromatin in promoters targeted by the ncRNA subsequently leading to the silencing of downstream genes. The recruitment of chromatin remodelling complexes is due to a second motif in the ncRNA,⁸ demonstrating how a non-coding transcript can guide enzymes to act at specific genomic loci. The timed availability of a TFO-containing transcript together with the site-specificity of TFO-induced triplex formation may indeed provide the very means for the dynamic regulation of the epigenome. Similarly, (GAA)_n repeats and other polypurine•polypyrimidine tracts prone to form H-DNA were found to cause regional epigenetic changes in cell models^117,118 especially hypoacetylation and hypermethylation.¹¹⁹

Triplex-helices also influence chromatin condensation in a more direct way. The accommodation of the third strand in the major groove changes the rigidity of duplex DNA. Triplexed DNA is less flexible and cannot as easily wrap around a histone complex, which subsequently affects nucleosome reconstitution.^117,120,121 Conversely, studies based on DNaseI cleavage patterns showed that TFOs cannot bind to DNA that is tightly associated with the nucleosome core (reviewed in ref. 122).

Taken together, triple-helix formation at targeted double-stranded DNA may participate in retaining open chromatin, assist in accurate nucleosome positioning as well as modification of histone complexes and DNA itself.

Triple-helices induce mutagenesis, recombination and DNA repair.

Triplex formation causes duplex DNA to adopt non-B conformations, which promotes genetic instability, mutation and recombination leading to repeat extension or genomic rearrangement.¹²³ Both inter- and intramolecular triplexes are mutagenic and induce recombination as well as DNA repair (reviewed in ref. 124 and 125). Transgenic mice models indicate that H-DNA can induce genetic instability in the mammalian genome.¹²⁶ Similarly, a purine-motif TFO induces mutagenesis in mammalian cells fivefold above that of a control oligonucleotide with scrambled sequence.¹²⁷ TFO-directed mutations were introduced in somatic cells of mice, validating triplex formation in vivo as well as its mutagenic character in several tissues.¹²⁸ It appears that the TFO-induced helical alteration of the duplex provokes the mutagenesis¹²⁹ and possibly contributes to common chromosomal translocation in cancer (reviewed in ref. 79).

A more exotic triplex, where the third strand binds parallel to an identical nucleotide sequence in the duplex, was proposed to function as a primer inducing DNA replication and promoting DNA rearrangements.¹³⁰ Phage and bacterial DNA polymerases could accommodate three DNA strands in their enzymatic centre and are particularly prone to elongate G-rich primers.

Regulation of the Regulator?

To be useful to the cell as a biological signaling mechanism, triplex formation must be reversible in a controllable way. As indicated above such a negative control system exists in the form of helicases, which actively unwind triplexes. Since triplex structures can interfere with polymerase activity it is crucial for the cell to melt such obstacles efficiently before or during DNA replication. Helicases may also interact with DNA-binding proteins such as TFs to resolve triplexes in a locus-specific manner although this is purely speculative.

Another potential mechanism for regulating triplex formation is sequence modification. RNA editing converts cytosine into uracil or adenosine into inosine in double-stranded RNA¹³¹ and occurs in transcripts of many coding and non-coding sequences.¹³² It has been shown that replacement of certain nucleotides by inosine in a triple-helix alters its stability.^133,134

Methylation of cytosine to 5′-methyl-cytosine is another indigenous nucleotide modification that occurs in DNA and RNA (reviewed in refs. ¹³⁵ and ¹³⁶), which is particularly beneficial for the formation and stability of pyrimidine-motif triplexes due to its elevated pKa value.^137,138 Only non-CpG methylations, like CpC, CpT and TpC, are relevant to pyrimidine-motif triplexes. Such modifications have been reported in plants,¹³⁹ flies¹⁴⁰ and mammals^141,142 and seem to be mediated by the protein Dnmt2,^140,142 and Dnmt3b.¹⁴³ Dnmt2 appears in fact to be an RNA methyltransferase^144,145 that plays an important role in developmental processes¹⁴⁶ while Dmnt3b has been associated with hypermethylation of non-CpG cytosines at the PCG-1alpha promoter in human diabetic patients¹⁴³ and has recently been suggested to associate with TFO-induced triplexes.⁸ Furthermore, methylation of CpT has been linked to mammalian embryonic stem cells but not to somatic cells, indicating that this reversible modification may be involved in cell differentiation.^147,148 Non-CpG-methylation was also found to be reinstated at each generation in spite of the fact that the majority of the sperm genomes contained next to no methylated non-CpG sites.¹⁴⁹ The full extent of such modifications in higher organism is yet to be revealed. Besides these natural modifications, several artificial modifications to TFOs have also been shown to increase triplex stability (reviewed in ref. 150).

Future Prospects

Given the conceptual simplicity of nucleic triple-helix formation and the availability of all required components, it would be surprising if triplex formation were not utilized functionally by the cell.

Challenging aspects.

While there is now considerable circumstantial evidence for triplex formation and function in vivo, categorical proof is still missing. Incisive new experimental approaches are needed that can unambiguously identify the location, composition and function of naturally occurring triplexes. New triplex-capture approaches and further characterization of triplex formation rules are required in order to determine the true range of roles triplexes may play in vivo. Below, we discuss how to proceed to gain such a definitive picture.

The vast majority of the genome in higher organisms is transcribed into ncRNA in complex, developmentally controlled and tissue-dependent patterns, which strongly suggests that ncRNAs play a central role in the regulation of development and cellular processes where TFO-induced triplex formation may play an important part. Consistent with this, it is now clear that ncRNA is involved in the regulation of many genetic and epigenetic processes including transcription, translation and chromatin modification, and numerous mechanisms have been identified (reviewed in ref. 6 and 151).

The challenges are to determine the extent of the involvement of triplex formation in gene regulation, and to detail the specific regulatory mechanisms and the sources and targets of the triplex-forming ncRNA transcripts involved. One component of this task comprises genome-wide mapping of TFOs and TTSs, along with identifying their dynamic complex formation, in a time- and tissue-dependent manner. This may be achievable using a combination of existing technologies (such as next-generation sequencing) and novel approaches, several of which we describe below. Important facets include developing more detailed TFO-TTS binding rules, creating tissue-specific catalogues of expressed putative TFOs, and occupancy measurement of TTSs and associating TFOs with their target TTSs to build models of regulatory (sub-)networks.

Refining triplex formation rules.

While the abundance of previous biochemical experiments provides a good overview on how triplexes form, their different setups and protocols make it infeasible to infer an accurate thermodynamic binding model. Instead, the screening of the vast sequence space with high throughput methods is required. Furthermore, it would facilitate the development of more accurate in silico methods for TFO and TTS prediction by filling in gaps in our knowledge about the minimum requirements on length and purine-content as well as tolerable “mismatches” for triplex formation.

Identification of triplex-forming sequences in genomic DNA.

Empirical determination of the precise sites of triplex formation in vivo, and the dynamic changes in their position during cell differentiation, should be possible by deep sequencing of both the DNA and RNA content of triplexes isolated by chromatin immunoprecipitation with specific anti-triplex antibodies or by nuclease digestion targeted either at triplexes (analogous to that used to determine DNaseI hypersensitivity sites,¹⁵² e.g., using nuclease S1) or designed to remove duplex DNA (e.g., by DNaseI digestion). Indeed a significant gap in the large-scale analyses to define genome architecture and biology, such as the ENCODE project,¹⁵³ is the lack of information on the dynamical occurrence of non-canonical structures in the genome, such as triplexes, Z-DNA tracts and G-quartets, all of which are known to occur and would be expected to play important roles in gene regulation and genome biology, and which should be amenable to targeted high-throughput sequence analysis. The data obtained will also enable comparison with the known or expected composition of such sequences (see below), with an expected validation of both, as well as the identity of the sequences forming the third (Hoogsteen) strand, with a concomitant refinement of the rules of engagement and the precise identity of the sequences involved.

Mining deep sequencing data for putative triple-helices.

To identify expressed TFOs, short ncRNA sequences from deep-sequencing experiments can also be filtered computationally for their propensity to function as TFOs using the general triplex formation rules reviewed here, or applying more detailed rules derived from empirical deep sequencing data, possibly in combination with RNA-folding predictions. Applying this filtering to sequenced ncRNA from a particular cell or tissue under given conditions would result in a tissue-specific catalogue of putative, expressed TFOs. Repeating this experiment under multiple conditions and intersecting the sets of predicted, expressed TFOs can uncover sets of differentially expressed TFOs. Further in silico filtering could be applied by mapping the TFOs to potential target sites in the genome, again using triplex formation rules and comparison to empirical data. Given that most promoters contain a predicted, often unique TTS,⁸³ gene expression data can be leveraged by correlating changes in the expression of genes containing a (putative) TTS with changes in expression of a matching (putative) TFO.

Refining gene regulatory networks.

Determining the existence and extent of triplex-based gene regulatory networks will enable the construction and validation of network models. The critical steps in building such a model are identifying TFOs and their cognate TTSs. This can be accomplished using the approaches described above for mapping TTSs, cataloguing TFOs and associating TFO with cognate TTS using binding rules and inference based on coordinate expression of the TFO and genes associated with its predicted TTS targets, perhaps incorporating adjacency filters such as suggested by the example of the transcriptional regulation of DHFR.⁷ Such inferred network models will of course need to be verified to establish the direct effect of triplex formation on the epigenetic status and/or transcription of target genes, which can procede using RNA-based approaches analogous to those currently used for testing protein-based regulation models.

Abbreviations

ncRNA

non-protein-coding RNA

TTS

triplex target site

TFO

triplex-forming oligonucleotide

TF

transcription factor

Appendix

Triplex motifs.

In the [T,C]-motif, also referred to as the pyrimidine motif, thymine binds adenosine (T-A•T) and cytosine binds guanine (C-G•C) in the polypurine strand of the duplex. Due to the Hoogsteen configuration, the third strand is oriented parallel to the polypurine strand. The cytosine in the third strand must be protonated to be able to form the second Hoogsteen hydrogen bond, which makes cytosine-containing triple-helices pH-dependent, and hence limits their stability at physiological pH.¹⁵⁴ The positive charge provided by the protonated cytosine, however, compensates for unfavorable charge repulsions of the polyanionic oligonucleotide backbones. Contiguous protonated cytosines, on the other hand, result in unfavorable charge repulsion between them.¹⁵⁵ Such a cytosine-tract can be replaced by guanine in the TFO in some circumstances.¹⁵⁶

In the [G,A]-motif, also referred to as the purine motif, (G-G•C) and (A-A•T) triplets are formed in reverse Hoogsteen configuration, thus resulting in anti-parallel triple-helices.¹⁵⁷ While guanine-rich TFOs show good DNA affinity under physiological pH,¹⁵⁸ triplex formation by long tracts of guanine competes with quadruplex formation,¹⁵⁹ which may limit their in vivo effectiveness.¹⁶⁰ [G,A]-TFOs can tolerate a small fraction of thymines, which form (T-A•T) triplets.¹⁶¹,¹⁶²

The steric properties of guanine and thymine allow for the remaining [G,T] motif, where (G-G•C) and (T-A•T) triplets are formed, which prefers one orientation over the other depending on its underlying sequence.¹⁶³

By contrast, in the parallel (recombination) “R-triplex”, the third nucleotide strand orientates in parallel to the identical nucleotide sequence in the duplex.¹¹⁶ The R-triplex is considerably less stable compared to above-mentioned triplex configurations especially for intermolecular formation.¹⁶⁴

Triple-helix formation and sequence specificity.

Triple-helical formation comes in two flavors: intramolecular and intermolecular. While in the first case the third strand is physically tethered to the duplex molecule providing the target,¹⁶⁵ in the second case the TFO is an independent molecule. Intramolecular DNA triple-helices, often also referred to as H-DNA, were first observed in plasmid DNA at homopurine•homopyrimidine mirror repeats (reviewed in ref. 166). There are four possible configurations in which three of the four DNA strands form H-DNA in the pyrimidine and the purine motif, respectively (see ref. 59 and 167 for detailed reviews on the biological relevance of H-DNA). Intramolecular triple-helices also form in RNA where they may contribute to RNA folding and tertiary structure stability.⁶²,¹⁶⁸

Sequence specificity and binding affinity of the third strand to its target duplex is a crucial factor for functional triplex formation in vivo. For a given overall base composition, the actual sequence of the participating strands has a strong impact on stability of the triple-helix (reviewed in ref. 95). Long A-tract duplex DNA, for example, seems to be unfavorable for triplex formation due to its higher propeller twist and rigidity.¹⁶⁹,¹⁷⁰ Long TFOs generally tend to form more stable complexes than shorter ones, although some high-affinity structures have been demonstrated with TFOs as short as nine nucleotides.¹⁷¹ However, TFOs of length greater than 17 nt may exhibit substantial affinity for a secondary (shorter) target site by tolerating mismatches, loops or other structures¹⁷²–¹⁷⁴ or by alternating the target strand in the duplex (as described in the next paragraph).¹⁶³ Triplexes in principle tolerate mismatches between the strands,¹⁷⁵,¹⁷⁶ however these have a strong destabilizing effect that increases with the number of contiguous mismatches.¹⁷⁷ The destabilizing effect depends furthermore on the nature of the mismatch and its positional location, i.e., a terminal mismatch causes less disruption to the triplex than mismatch in the centre of the triple-helix.¹⁷⁸

A single TFO can bind to a target duplex in which the polypurine tract switches strands.¹⁶³ In this situation, different segments of the TFO bind to different strands in the duplex. The alternating target strand preference is accompanied by a switch in the binding configuration (Hoogsteen/reverse Hoogsteen) of the corresponding binding segment in the TFO.¹⁶³ All six possible junctions for combining two short TFO-segments have shown to be functional in vitro.¹⁷⁹–¹⁸¹

The cellular environment likely imposes specific constraints on the potential for triple-helix formation. High concentrations of multivalent cations promote triplex stability by compensating for the unfavorable electrostatic repulsion of the three negatively charged oligonucleotide backbones.¹⁵⁸,¹⁸²,¹⁸³ It is well established that metal cations are naturally present in DNA in vivo.¹⁸⁴ However, the stabilizing effect of any multivalent cation is limited by its competition with available monovalent counterfeits.¹⁸⁵,¹⁸⁶ Physiological concentrations of potassium, for example, have been shown to reduce the efficiency of triplex formation,¹⁸⁵ most likely in favor of quadruplex formation.¹⁸⁷ Polyamines also promote triplex formation under physiological pH,¹⁸⁸ as do some charge-neutralizing basic polypeptides.¹⁸⁹ In addition to the effects of cations and pH,¹⁹⁰,¹⁹¹ triplex formation is also highly dependent on temperature, organic solvents²,⁷³,¹⁹²–¹⁹⁴ and chromatin accessibility,¹²²,¹⁹⁵ all of which complicate in silico assessments of the potential of a given TFO and target sequence to form triplexes in vivo. So far, little is known about how local microenvironments and the involvement of other factors such as proteins might favor the formation and/or increase the stability of particular types of triplexes in vivo.