key: cord-0928247-tqx420oa
authors: Goldsmith, Gunaseelan; Rathinavelan, Thenmalarchelvi; Yathindra, Narayanarao
title: Selective Preference of Parallel DNA Triplexes Is Due to the Disruption of Hoogsteen Hydrogen Bonds Caused by the Severe Nonisostericity between the G*GC and T*AT Triplets
date: 2016-03-24
journal: PLoS One
DOI: 10.1371/journal.pone.0152102
sha: 991aed7275084ce0661be421e281a8f47dcc8637
doc_id: 928247
cord_uid: tqx420oa

Implications of DNA, RNA and RNA.DNA hybrid triplexes in diverse biological functions, diseases and therapeutic applications call for a thorough understanding of their structure-function relationships. Despite exhaustive studies mechanistic rationale for the discriminatory preference of parallel DNA triplexes with G(*)GC & T(*)AT triplets still remains elusive. Here, we show that the highest nonisostericity between the G(*)GC & T(*)AT triplets imposes extensive stereochemical rearrangements contributing to context dependent triplex destabilisation through selective disruption of Hoogsteen scheme of hydrogen bonds. MD simulations of nineteen DNA triplexes with an assortment of sequence milieu reveal for the first time fresh insights into the nature and extent of destabilization from a single (non-overlapping), double (overlapping) and multiple pairs of nonisosteric base triplets (NIBTs). It is found that a solitary pair of NIBTs, feasible either at a G(*)GC/T(*)AT or T(*)AT/G(*)GC triplex junction, does not impinge significantly on triplex stability. But two overlapping pairs of NIBTs resulting from either a T(*)AT or a G(*)GC interruption disrupt Hoogsteen pair to a noncanonical mismatch destabilizing the triplex by ~10 to 14 kcal/mol, implying that their frequent incidence in multiples, especially, in short sequences could even hinder triplex formation. The results provide (i) an unambiguous and generalised mechanistic rationale for the discriminatory trait of parallel triplexes, including those studied experimentally (ii) clarity for the prevalence of antiparallel triplexes and (iii) comprehensive perspectives on the sequence dependent influence of nonisosteric base triplets useful in the rational design of TFO’s against potential triplex target sites.

Implications of DNA, RNA and RNA.DNA hybrid triplexes in diverse biological functions, diseases and therapeutic applications call for a thorough understanding of their structure-function relationships. Despite exhaustive studies mechanistic rationale for the discriminatory preference of parallel DNA triplexes with G * GC & T * AT triplets still remains elusive. Here, we show that the highest nonisostericity between the G * GC & T * AT triplets imposes extensive stereochemical rearrangements contributing to context dependent triplex destabilisation through selective disruption of Hoogsteen scheme of hydrogen bonds. MD simulations of nineteen DNA triplexes with an assortment of sequence milieu reveal for the first time fresh insights into the nature and extent of destabilization from a single (non-overlapping), double (overlapping) and multiple pairs of nonisosteric base triplets (NIBTs). It is found that a solitary pair of NIBTs, feasible either at a G * GC/T * AT or T * AT/G * GC triplex junction, does not impinge significantly on triplex stability. But two overlapping pairs of NIBTs resulting from either a T * AT or a G * GC interruption disrupt Hoogsteen pair to a noncanonical mismatch destabilizing the triplex by~10 to 14 kcal/mol, implying that their frequent incidence in multiples, especially, in short sequences could even hinder triplex formation. The results provide (i) an unambiguous and generalised mechanistic rationale for the discriminatory trait of parallel triplexes, including those studied experimentally (ii) clarity for the prevalence of antiparallel triplexes and (iii) comprehensive perspectives on the sequence dependent influence of nonisosteric base triplets useful in the rational design of TFO's against potential triplex target sites.

It is well documented that Watson and Crick paired DNA duplex interacts with sequence specific oligonucleotides to form triple helices stabilised by either Hoogsteen (parallel triplex) or reverse Hoogsteen (antiparallel triplex) pair of hydrogen bonds. Evidences for involvement of triple helices in biological processes come from their participation in regulation of gene expression [1, 2] , DNA damage and repair [3, 4] , RNA processing and folding [5, 6, 7] and chromatin organization [8] . Triplexes are also known to impair DNA polymerization and influence DNA recombination process [9] . Base triple interactions crucial for function are shown to be present in pseudoknots in telomerase RNA [10] and in transcripts during programmed ribosomal frame shifting in viruses like SARS coronavirus [11] . Most recently, intramolecularly folded RNA triplexes are shown to be formed in the highly abundant and up-regulated long noncoding transcripts such as metastasis associated lung adenocarcinoma transcript 1 (MALAT1) and others to evade exonuclease action [12, 13] . Formation of RNA.DNA hybrid triplexes involving noncoding RNA found to repress genes like DHFR, TGF-β and MAT2A has further enhanced the biological importance of nucleic acid triplexes [14, 15, 16] . Existence of triplexes in vivo is supported by the discovery of triplex unwinding helicases [17, 18] , triplex-specific antibodies [19, 20] and other endogenous triplex specific proteins [21] [22] [23] [24] . Association of triple helices with colorectal cancer [25] , neurodegenerative disorder Friedreich's ataxia [26] and a number of inherited as well as acquired human diseases [27] further underscore their profound significance.

DNA triplexes have evoked potential technological applications which include creating high quality DNA vectors for human gene therapy [28] , nanomachines for monitoring intracellular pH gradient [29] , molecular switches [30, 31, 32] , and for developing drug delivery systems [33, 34] . They have also been used as biosensors for the recognition and analysis of toxic metal ions [35] , SNPs [36] , DNA methylation [37] and cancer cells [38] . Quite recently, DNA triplex formation process has been exploited in developing PCR based biosensor for detection of pathogens [39] , high-throughput assays for measuring activity of DNA topoisomerases and other enzymes involved in DNA topology modification [40] , controlled assembly of liposomes [41] and as modular probes for DNA detection [42] . Recognition of abundant putative triplex target sites in both prokaryotic and eukaryotic genomes [43, 44, 45] , together with new experimental approaches for identification of triple helices [46] [47] [48] [49] [50] and development of algorithm/databases for analysis of triplex target sites [51, 52, 53] highlight their increasing relevance in the post genomic era. Several reviews detailing the biological significance [2, [54] [55] [56] [57] [58] [59] [60] and potential therapeutic applications [61] of triplexes have been appearing to underscore their importance. Notwithstanding, structural information on triple helices, specifically those addressing the issue of base triplet nonisostericity and their effects, crucial in the design of Triplex forming Oligonucleotides (TFOs) for targeting duplex is rather inadequate.

It is well recognised that TFOs rich in ( [56, 62] . While T & C rich TFOs favour parallel triplex [63, 64] , those with G & A favour antiparallel triplex [65, 66] . Although TFOs with G & T form both parallel and antiparallel triplexes, they predominantly favour the latter [67] [68] [69] [70] . Interestingly, both intramolecular [67, 71] and intermolecular [62, [72] [73] [74] parallel triplexes containing a limited number of GÃGC & TÃAT triplet pairs are observed. They are shown to be effective in interfering with biological functions in several instances. For instance, GT rich 38-mer and 15-mer TFOs inhibit transcription of HIV-1 in infected human cells [75] and replication in SV40 containing plasmid in COS-1 cells [76] respectively, through formation of parallel triplexes. On the other hand, the 22-mer TFO with  10 TÃAT & GÃGC juxtapositions, and the 36-mer TFO with 21 TÃAT & GÃGC juxtapositions, targeted against the promoter region of human Ki-ras [77] and the human epidermal growth factor receptor gene [78] respectively, are unable to form parallel triplex, but forms only antiparallel triplex. Seemingly then, selective ability to form parallel triplex appears to be determined by the number of GÃGC & TÃAT juxtapositions, their larger incidences, thwarting parallel triplex formation. While this may be notionally attributed to nonisosteric traits between them, their quantitative estimates and precise mechanistic effects on the ability and stability for triplex formation are obscure.

A cursory glance at the superposition of GÃGC and TÃAT triplet pairs (Fig 1A and 1B) , readily point to the existence of an intrinsically large nonisostericity (defined by the residual twist angle, Δt°) in parallel compared to antiparallel orientation. We argue that this might be responsible in imparting differential mechanistic influence in parallel vis-à-vis antiparallel triplexes causing selective preference of the former. To probe this, extensive MD simulations of a variety of parallel and anti-parallel DNA triplexes formed by these triplets, in different sequence contexts (Sequences 1-19 in Table 1 ) have been carried out. Results have yielded clear insights into the role of nonisostericity on triplex stability, stereochemical reasoning for the prevalence of antiparallel triplex and selective preference for parallel triplex. These should aid in comprehensive understanding of sequence dependent structure of nucleic acid triplexes and their implication in biological processes.

Superposition of GÃGC (green) and TÃAT (brown) triplets in parallel (A) and anti-parallel (B) orientation to bring out the nature and source of nonisostericity between triplets. Triplets are depicted as circles formed with the three C1' atoms of the triplet bases lying on the circumference. Appearance of a twist between the adjacent Hoogsteen bases (G H & T H ) or reverse Hoogsteen bases (G RH & T RH ), even prior to the application of formal helical twist (t) constitutes the residual twist Δt°. This together with the radial difference Δr Å between the two superposed triplets, suffice to quantify nonisostericity between them. Note the higher value of Δt°in parallel (A) than in antiparallel (B) triplets. Illustration of the effect of over winding (30°+Δt°) at the G H T H step (C) and under winding (30°−Δt°) at the T H G H step (D) of the Hoogsteen strand consequent to residual twist. Two adjacent GÃGC triplets with the ideal twist angle of (t) 30°(E) is shown to underscore the effect of residual twist.

Manifestation of the highest nonisostericity between the parallel G * GC & T * AT triplets Superposition of the parallel GÃGC & TÃAT base triplets reveals (Fig 1A) misalignment of the C1' atoms of the Hoogsteen strand (see also S1A Fig). This results in the largest value for the residual twist angle Δt°(± 21.6°) between the G. . .G and A. . .T Hoogsteen pairs. It is obvious that the effect of residual twist would be to impart over winding (+Δt°) at the G H T H step (GÃGC preceding TÃAT) ( Fig 1C) and under winding (-Δt°) at the T H G H step (TÃAT preceding GÃGC) in the Hoogsteen strand of the triplex (Fig 1D) . Consequently effective triple helical twist would be 30°(t°) + 21.6°(+Δt°) = 51.6°and 30°(t°)−21.6°(-Δt°) = 8.4°at the GT and TG steps of the Hoogsteen strand respectively. 51.6°is close to nearly twice the value of triple helical twist of 30°(t°), while 8.4°is an exceptionally low value for triple helical twist. Outcome of these creates disconnects of nearly one nucleotide length (5.5 Å to 6.3 Å) in the sugar-phosphate backbone at the successive steps of the Hoogsteen strand (S1B Fig). They also generate severe steric overlap between the adjacent sugars, especially, at the under wound TG step.

Bridging these disconnects and ensuring a steric free conformation, while retaining Hoogsteen hydrogen bonds in the GÃGC & TÃAT triplets, demand considerable conformational rearrangements (see below). Incidentally, the other nonisostericity parameter viz., radial difference (Δr Å) is relatively small (ffi 0.4 Å) for the parallel (Fig 1A) compared to the antiparallel base triplets (ffi 1.1 Å) ( Fig 1B) . Given this, it is expected that Δt°would impact more than Δr Å in a parallel DNA triplex.

Successive incidence of G * GC & T * AT triplets destabilizes parallel DNA triplex by disrupting Hoogsteen hydrogen bonds It might be anticipated that the conformational changes necessitated by the unusually high and low twist angles at the alternating G H T H and T H G H steps might be absorbed by the inherent flexibility of the triplex. In order to examine this, MD simulations (250 ns) have been carried out on a 15-mer parallel triplex comprising alternating GÃGC & TÃAT triplets (Sequence 1 in Table 1 ). The results at once indicate a proclivity for the loss of Hoogsteen hydrogen bonds in both the triplets (Fig 2A) : N6 (A W ) . . . O4 (T H ) and N7 (A W ) . . .N3 (T H ) hydrogen bonds in the TÃAT triplets are absent for over 52% and 49% of the simulation period respectively, while N7 (G W ) . . .N2 (G H ) and O6 (G W ). . . N1 (G H ) hydrogen bonds are absent in the GÃGC triplets for over 64% and 47.7% respectively. This is caused by the movement of T H and G H to overcome the effects of high Δt°. In the process, Hoogsteen hydrogen bonds are disrupted. In striking contrast, the reverse Hoogsteen bonds of the GÃGC & TÃAT triplets in the antiparallel DNA triplex (Sequence 2) are largely preserved ( Fig 2B) . This can be attributed to the much lower value of Δt°(10°) in antiparallel, as opposed to 21.6°in the parallel triplex. Evidently much smaller conformational variations required by the lower residual twists are readily absorbed by the inherent triplex flexibility without the need for the loss of reverse Hoogsteen hydrogen bonds. Pronounced Nonisostericity between G*GC and T*AT Triplets Destabilizes Parallel DNA Triplex

Emergence of noncanonical Hoogsteen schemes. Consequent to the disruption of Hoogsteen hydrogen bonds, alternative schemes of hydrogen bonds emerge for both the TÃAT (NC1 to NC4; Fig 3D-3G ) and GÃGC (NC5 to NC8; Fig 3H-3K ) triplets. These are referred to as noncanonical Hoogsteen schemes. For TÃAT, NC1 scheme ( Fig 3D) is predominant and is formed by way of water mediated interaction involving N3 of T H and N6 & N7 of A W . In NC2 (Fig 3E) , both ion and water interact with T H (O4 & N3) and N6 & N7 of A W and O4 of T H . Interaction via two water molecules describes NC3 ( Fig 3F) . In all of these none of the canonical Hoogsteen hydrogen bonds are retained. But in NC4 (Fig 3G) , canonical N6. . .O4 Hoogsteen hydrogen bond scheme is preserved while a water molecule mediates the interaction between N3 of T H and N7 of A W .

In the case of GÃGC triplet, NC5 scheme ( Fig 3H) is predominant. This is characterised by bifurcated hydrogen bonds involving N2 of G H and O6 & N7 of G W . In the NC6 scheme ( Fig  3I) , a water molecule mediates interaction between G H and WC paired bases. In both these situations, Hoogsteen G H swivels causing the amino group to glide towards the WH groove. But, similar action in the direction of the CH groove results in the NC7 scheme ( Fig 3J) . Here, an ion comes in between O6's of G H and G W , while N7 of G W forms bifurcated hydrogen bonds with N2 & N1 of G H . NC8 scheme is alike NC5, but in addition, is mediated by a water molecule between N7 & N3 atoms of G W and G H (Fig 3K) .

A common feature observed in all of the above is the propensity for reducing the initial large Δt°(21.6°) to lower values (as low as 1.2°in NC3; Fig 3G) to circumvent the large mechanistic effects (steric overlap and disconnect in the sugar-phosphate backbone) discussed above. This can happen only at the expense of losing canonical Hoogsteen hydrogen bonding. Such a loss in a large number of triplets would critically affect triplex stability. This is clearly reflected in the very low binding free energy of only-9.4 kcal/mol for the parallel triplex (Sequence 1) in sharp contrast to the comparatively high value of-70.80 kcal/mol for the antiparallel triplex (Sequence 2). The latter by and large retains all the reverse Hoogsteen hydrogen bonds in all the triplets. Thus, the parallel DNA triplex with consecutive GÃGC and TÃAT triplets is not expected to be stable due to the mechanistic effects of extreme nonisostericity at every base step of the Hoogsteen strand. Hence the discussion about this unstable structure is not pursued further.

In the 15-mer parallel DNA triplex (Sequence 1, Table 1 ) discussed above, each triplet say TÃAT, is flanked on either side by the nonisosteric base triplet GÃGC, giving rise to two pairs of consecutive and overlapping nonisosteric base triplets (NIBTs), GÃGC & TÃAT and TÃAT & GÃGC. The intervening TÃAT triplet, being common to both, overlaps with both the flanking GÃGC triplets, and it therefore in effect experiences nonisostericity effects due to the preceding and the following GÃGC base triplet. Consequently, two Δt°s, viz., +Δt°at the GT step (between the GÃGC & TÃAT triplets) and−Δt°at the TG step (between the TÃAT & GÃGC triplets) ( Fig  1C) operate on the intervening TÃAT triplet. We refer to this scheme as overlapping pair of triplets. In fact, the 15-mer triplex (Sequence 1; Table 1 ) may deem to be comprised of such trimer triplets constituting 14 pairs of overlapping NIBTs. The deleterious effect of which has been discussed above. But in reality a limited number of NIBTs may occur in the target duplex. Moreover, these may be spaced out rather than occurring alternately along the sequences. Such circumstances are in fact investigated [79, 80] . These prompted us to study the effect of isolated and lone occurrence of such NIBTs. This situation can be visualized when a GÃGC or TÃAT triplet interrupts an otherwise homopolymeric triplex (Sequences 4 and 8; Table 1 ). It will be intriguing to examine if the Hoogsteen hydrogen bonding pair is retained here and if not, evaluate its influence on the triplex structure and stability.

Structure of a T Ã AT triplex with a G Ã GC interruption mediated by overlapping pairs of NIBTs (-Δt°followed by +Δt°). It is readily seen that a GÃGC triplet interruption in the parallel homopolymeric TÃAT DNA triplex (Sequence 4; Table 1 ) generates two sets of NIBTs which are overlapping and consecutive. The first pair of NIBT at the TG (T 27 G 28 ) step results in a residual twist of -Δt°=−21.6°and effects under winding. This is followed by a residual twist of +Δt°= +21.6°at the GT (G 28 T 29 ) step to effect over winding. MD simulation of this triplex reveals that the interrupting G 28 Ã G 17 C 6 triplet loses the N1. . .O6 hydrogen bond of the canonical Hoogsteen scheme (Fig 4A and S2B Fig) in favour of the noncanonical NC5 Hoogsteen scheme (Fig 3H) . This clearly demonstrates that even an isolated nonisosteric triplet is unable to retain the canonical Hoogsteen pair and instantly transits to a noncanonical scheme (NC5) which reduces the Δt°values to~−4.1°and + 4.6°at the T 27 G 28 and G 28 T 29 steps respectively from the initial value of 21.6°. As a result, slightly lower (20.2°) and higher (36.6°) helical twist angles are seen at these steps occurring on either sides of the GÃGC interruption ( Fig 5A) . Stacking interactions between the Hoogsteen bases (G 28 and T 29 ) are minimal at the GT step compared to that between T 27 & G 28 Average value for the X-displacement of base pair at the interruption site (-3.6 Å) is slightly higher compared to other places (-2.9 Å). On the other hand, propeller twist of WC base pairs at the interruption site is slightly lower (~6.86°) compared to -9.6°found at other places. Sugar ring of G 28 of the G 28 Ã G 17 C 6 interruption assumes the C1' exo pucker, while thymidines of the Hoogsteen strand favour O4' endo pucker. All the torsion angles lie in their preferred ranges. Average value of minor groove width is lowered to 12.2 Å from the initial value of 14 Å concomitant with an increase in the major groove width by 1.5 Å to a value of 20.1 Å. Increase in the widths of WH and CH groove by 2 and 4 Å to 9.3 Å and 16.1 Å respectively is also seen (S3A Fig). The binding free energy (~-53.5 kcal/mol) of this triplex (Sequence 4) is reduced bỹ 10 kcal/mol compared to~-63.6 kcal/mol obtained for the uninterrupted homopolymeric TÃAT triplex (Sequence 3; S2A Fig) pointing clearly to the destabilizing effect of base triplet interruption which is tolerated as a triplet mismatch. Incidentally, a TÃAT triplex with a GÃGC interruption forms a stable parallel triple helix [62] .

Structure of a G Ã GC triplex with a T Ã AT interruption mediated by overlapping pairs of NIBTs (+Δt°followed by -Δt°). Here, the roles of GÃGC and TÃAT triplets are swapped. A homopolymeric GÃGC triplex is interrupted by the T 28 Ã A 17 T 6 triplet, creating two overlapping and consecutive pairs of NIBTs (Sequence 8; Table 1 ). Contrary to the previous case, here the over winding GT step (+Δt°) precedes the under winding TG step (-Δt°) in the Hoogsteen strand. MD simulations show that the interrupting T 28 Ã A 17 T 6 loses the canonical Hoogsteen N3 (T H ). . .N7 (Aw) hydrogen bond for over 95% of total simulation time ( Fig 4B and S4B  Fig) , while the O4 (T 28 ). . .N6 (A 17 ) hydrogen bond is lost for over 13% of total simulation time. Eventually, a water mediated noncanonical NC1 Hoogsteen scheme (Fig 3D) prevails, as a triplet mismatch, entailing reduced Δt°values of +5.8°and -6.1°at the Hoogsteen GT and TG steps respectively. Akin to the previous case (Sequence 4; Table 1 ), twist angles (Fig 5C) , base stacking pattern (Fig 5D) Pronounced Nonisostericity between G*GC and T*AT Triplets Destabilizes Parallel DNA Triplex

In view of the above findings it is extremely important to ascertain the influence of a single pair vis-à-vis 2 overlapping pairs of NIBTs in retaining or disruption of the canonical Hoogsteen hydrogen bonds and their impact on triplex stability. A single pair of (non-overlapping) NIBTs is harbored uniquely at a triplex junction interface. This can be differentiated by an over winding GT (+Δt°) or an under winding TG (-Δt°) step along the Hoogsteen strand. Results of MD simulations of 5 DNA triplexes differing in the number of triplex junctions and GT/TG step junction interfaces (Sequences 10-14; Table 1) , mimicking a few experimentally studied instances [72, 76, 80, 81] are discussed below.

Structure of the DNA triplex junction with a GT step (+Δt°) interface. Sequence 10 in Table 1 represents a mini 8-mer GÃGC triplex followed by a mini 7-mer TÃAT triplex constituting a triplex junction harboring a lone pair of NIBTs mediated by the over winding G 38 T 39 step (+Δt°). MD simulation of this structure surprisingly shows the retention of canonical Hoogsteen hydrogen bonds (Fig 6A) in both the NIBTs. Effect of overwinding at the GT step (+Δt°) is reflected by way of large X-displacement of WC pairs in the junction neighborhood (S7A Fig) viz., C 8 . . .G 23 exhibiting larger value (-9.23 Å) followed by C 9 . . .G 22 (-4.7 Å) and C 10 . . .G 21 (-3.7 Å). X-displacement is around -3.4 Å and -3.1 Å for the other C. . .G pairs (in Structure of the triplex junction harboring the TG step (-Δt°) interface. Sequence 11 in Table 1 corresponds to a triplex junction with a single pair of NIBTs intervened by the T 37 G 38 step. Here again the results of MD simulation reveal preservation of canonical Hoogsteen hydrogen bonds in the nonisosteric triplet pairs (Fig 6B) . The under winding effect caused by the T 37 G 38 step is reflected in the low X-displacement for the WC pairs, C 8 . . .G 23 (~-1.31 Å) and T 9 . . .A 22 (~-0.65 Å) respectively, proximal to the triplex junction. This is in contrast to large X-displacement seen at the GT junction interface (see above). X-displacement for the other C. . .G and T. . .A pairs correspond to~-3.23 Å and -2.45 Å respectively. Although a small warp is observed at the triplex junction (Fig 7B) it is not as prominent as in GT step junction interface (Sequence 10). Interestingly BII phosphodiester conformation is found at the under wound T 37 G 38 (S9B Fig) while the rest favour BI. As expected, a low twist angle of~15°i s found at the Hoogsteen T 37 G 38 step concomitant with a higher twist at the abutting C 8 T 9 (~38.3°) and C 7 C 8 (~35.1°) steps of WC duplex (S8B Fig). Twist angles at the other CC & TT steps of the WC duplex remain at~30°. As in the previous case, propeller twist for the C. . .G Pronounced Nonisostericity between G*GC and T*AT Triplets Destabilizes Parallel DNA Triplex pair (-4.02°) is lower than T. . .A pair (-13.7°). Average minor groove width is~13.12 Å in the mini GÃGC triplex and it is~11.7 Å in the mini TÃAT triplex. Average major groove width in both the triplexes is~21.1 Å. Likewise, WH and CH groove width lie in the range of~9 Å and 13 Å respectively (S3D Fig). Stacking interaction is minimal at the neighborhood of triplex junction (S8D Fig) involving Hoogsteen bases, G 38 & G 39 and WC duplex pyrimidines, C 7 and C 8 . However partial stacking prevails between the purines, A 22 & G 23 of WC duplex Binding free energy of this triplex corresponds to~-78.4 kcal/mol. These data clearly brings about the distinctions in the influence of overlapping and non-overlapping NIBTs.

Incidence of multiple junctions contributes to instability factor. MD simulations of 15-mer triplexes with the incidence of two (Sequences 12-13; Table 1 ) and four (Sequence 14; Table 1 ) triplex junctions under different sequence contexts are performed to discern the limit on the number of occurrences of non-overlapping pair of NIBTs that might sustain, hinder or completely prevent triplex formation. In sequence 12, the 8-mer GÃGC triplex is flanked by TÃAT triplexes creating a T 34 G 35 step junction interface on the 5'-side and G 42 T 43 step junction interface on the 3'-side of TFO. On the other hand, in sequence 13, the 7 mer TÃAT triplex is flanked by GÃGC triplexes creating a G 34 T 35 junction interface on the 5'-side and a T 41 G 42 step interface on the 3'-side of TFO. Both these triplexes possess two sets of non-overlapping pair of NIBTs spaced apart by at least 7 or more triplets. sequence 14 is designed to contain 4 sets of non-overlapping NIBTs to examine the effect of increased incidence of triplex junctions. Results reveal canonical Hoogsteen hydrogen bonds are retained at the junction NIBTs akin to sequences 10 & 11 with only junction. Features characteristic to the GT or TG step junction interface including conspicuous bend at the over wound Hoogsteen GT (+Δt°) steps (G 42 This clearly indicates that although not as detrimental as the overlapping NIBTs incidence of large non-overlapping NIBTs in a given sequence could affect triplex stability.

Here, we have investigated the impact of 2 interruptions (4 pairs of overlapping NIBTs) and 2 junctions (2 pairs of non-overlapping NIBTs), which together makeup six pairs of NIBTs (~25%) in the 25-mer triplex (Sequence 15; Table 1 ). Results are on the expected lines with disengagement of canonical Hoogsteen hydrogen bonds to noncanonical Hoogsteen sequences in the T 55 Ã A 30 T 21 and G 66 Ã G 41 C 10 interruptions as in the sequences 4 to 9, while they are retained in at the triplex junction interfaces G 60 Ã G 35 C 16 & T 61 Ã A 36 T 15 and T 71 Ã A 46 T 5 & G 72 Ã G 47 C 4 triplets as in the sequences 10-14. Likewise, trend of high (~37°) and low twist (~12.7°) angles at the Hoogsteen G 60 T 61 and T 71 G 72 steps respectively concomitant with the reverse trend at the corresponding steps of WC duplex namely; low (~20.6°) and high (~37.2°) twist angles at the T 15 C 16 and C 4 T 5 steps respectively, are noticed. Also, larger X-displacement (-7.6 Å) is seen for the C 17 . . .G 34 and C 18 . . .G 33 WC base pairs causing a bend in the triplex (Fig 7C) . This triplex with 6 pairs of NIBTs is associated with a binding free energy of~-81.2 kcal/mol.

Ever since the finding of the ability of nucleic acid duplex to accommodate a TFO along its major groove, a variety of biological roles have been demonstrated for DNA, RNA and DNA. RNA hybrid triplexes. Distinctive disposition of donors and acceptors in guanine and thymine of the GT rich TFOs enables them to interact with the purine rich strands of WC duplexes in both parallel and antiparallel orientations forming GÃGC and TÃAT triplets, with unique geometries rendering them to be nonisosteric. Knowledge of the nature and magnitude of nonisostericity between triplet pairs is crucial in the design of TFOs as they determine the degree of influence on triplex conformation as well as stability. Nonisostericity can be effectively and elegantly defined by residual twist and radial difference [82] which can relate directly to structural traits. More assuring is that they have proven to be valuable in assessing the impact of nonisostericity on DNA triplexes [82, 83] .

Surprisingly GÃGC & TÃAT parallel triplets exhibit strikingly large nonisostericity with the highest value of~21.6°for the residual twist (Δt°) making it a potential culprit for the observed selective preference of parallel triplexes [73, 74, 75, 77] as this is expected to strongly perturb triplex. Qualitative arguments based on nonisostericity are unable to offer clarity to explicate the above, leave alone a mechanistic rationale. MD simulations of 14 parallel and 5 antiparallel DNA triplexes (Table 1) , comprising these triplets in an assortment of sequence contexts, are carried out to critically evaluate the decisive influence of single, double and multiple pairs of overlapping and non-overlapping nonisosteric base triplets (NIBTs) on triplex stability. Results have provided new insights for a comprehensive understanding of nonisostericity effects which should aid in the efficient design of TFO to target nucleic acids duplexes to form triplexes.

One of the key revelations from the present study is the disruption of canonical Hoogsteen hydrogen bonds in both the GÃGC and TÃAT parallel triplets when they occur alternately (Sequence 1 ; Fig 2A) (Fig 2B and S5 Fig) . This is directly correlatable to the vastly contrasting nonisostericity between the triplet pairs in parallel and antiparallel orientations, as reflected in the disparate values of Δt°and Δr Å. While Δt°is twice (21.6°) the value in parallel compared to 10.6°in antiparallel triplexes, Δr Å is slightly larger in antiparallel (1.1 Å) than in the parallel (0.4 Å) triplexes. Though both Δt°and Δr Å independently impact in creating backbone disconnect at the successive phosphodiester links, influence of Δr Å seems to be more in the antiparallel triplex although the effects are absorbed by the intrinsic triplex flexibility. On the other hand, mechanistic influence of the large residual twist of 21.6°in parallel triplexes is reflected in lowering the stability sharply through disruption of Hoogsteen hydrogen bonds and associated effects through significant conformational modifications (Figs 2A, 3, 4 & 7) . In contrast, smaller Δt°(~10.6°) merely causes a zig-zag sugar-phosphate-sugar backbone conformation concomitant with minor twist angle variations in the antiparallel triplex and these by no means affect its formation [83] . Therefore, it can be reckoned that the role of Δt°is more dominant than Δr Å in influencing the formation as well as stability of triplexes.

Isosteric base triplets (with inherent Δt°= 0 and Δr Å = 0) endow a 'uniform' DNA triplex. However, presence of residual twist (Δt°), as large as 21.6°between the parallel GÃGC & TÃAT triplets undermines the cannonical Hoogsteen hydrogen bonds in the DNA triplex causing them to assume noncanonical Hoogsteen schemes (NC1 to NC4 for T. . .A and NC5 to NC8 for G. . .G ; Fig 3) . By doing so, nonisosteric effects are offset via reduction in Δt° (Fig 3D-3J ) concomitant with loss of adjacent base stacking (Fig 8) . Thus, loss of canonical Hoogsteen G. . .G and T. . .A hydrogen bonds together with interrupted stacking decrease triplex stability. Absence of these in the antiparallel DNA triplex [82] results in highly favorable binding free energy of -70.8 kcal/mol compared to -9.4 kcal/mol for the parallel triplex. It is evident then that parallel triplexes with frequent juxtaposition of GÃGC and TÃAT triplets would be less favored.

Comparative assessment of overlapping vis-à-vis non-overlapping NIBTs on structure, energetics and stability of triplexes As mentioned earlier, a single base triplet interruption in an otherwise homopolymeric triplex (typified by isosteric base triplets) results in two consecutive and overlapping pairs of NIBTs, with the interrupting NIBT simultaneously bearing the brunt of nonisostericity effects of +Δt (+21.6°) and -Δt (-21.6°) (Sequences 4,8; Table 1 ). This causes the canonical Hoogsteen hydrogen bond to switch to noncanonical scheme (Fig 3D and 3H) , resulting in disrupted base stacking (Fig 5B and 5D ). It is obvious that loss of even one of the canonical hydrogen bonds in the triplet interruption enables the Hoogsteen base to sway in a direction to elude the deleterious effect of large Δt°. The base triplet interruption under these circumstances is reckoned as a triplet mismatch and a limited number of them, spaced apart, could be tolerated in a sufficiently long sequence (Sequence 15). Systematic investigations made here suggests that an interrupting GÃGC or TÃAT triplet decreases the binding energy by nearly 10-11 kcal/mol with reference to the homopolymeric triplex (Sequence 3 vs. Sequence 4 and Sequence 7 vs. Sequence 8). This reduction nearly doubles to~23 kcal/mol with 2 interruptions (Sequence 3 vs. Sequence 5 and Sequence 7 vs. Sequence 9, Table 1 ). Increase in the incidence of interruptions from 1 to 3 (Sequence 4 vs. Sequence 6) further reduces the binding energy by~37 kcal/mol indicating a linear correlation between the base triplet interruptions and the binding energy penalty. Hence, it is obvious that multiple incidence of interruptions ie. overlapping pair of NIBTs leads to less stable nature of parallel triplex leading even to inhibition of triplex formation. On the other hand, in the antiparallel triplex (Sequence 16 vs. Sequence 17), the estimated loss in the binding energy (~3 kcal/mol) caused by an interruption due to retention of canonical reverse Hoogsteen hydrogen bond (S5 Fig) and marginal loss in stacking is not severe or detrimental to antiparallel triplex formation.

On the other hand incidence of non-overlapping pairs of NIBTs harboring at triplex junctions (Sequences 10-14) experiences nonisostericity effects of either +Δt°or -Δt°alone. The over winding influence of +Δt°at the GT step causes the triplex to bend (Fig 7A and 7C and  S10 Fig). Unwinding influence of -Δt°at the TG step merely brings about change from BI to BII phosphodiester conformation (S9 Fig; Sequences 11-14) . Comparison of the binding energies of the five 15-mer triplexes (Sequences 10 to 14) designed out of 8 GÃGCs and 7 TÃATs suggest a reduction in the binding energy with the increase from 2 (Sequences 12 and 13) to 4 (Sequence 14) in the number of non-overlapping NIBTs (triplex junctions). Sequences 12 and 13 have 1 TG and 1 GT junction while in Sequence 14 there are 2 TG and 2 GT junctions. This causes a reduction in binding energy of~11.5 to 13 kcal/mol (Sequence 13 & Sequence 14 and Sequence 12 & Sequence 14) when non-overlapping NIBTs are increased from 2 to 4. Using this, a crude estimate of the binding energy penalty for a non-overlapping pair of NIBTs turns out to be~5.8 to 6.5 kcal/mol. This is nearly one half compared to the binding energy penalty (~11 to 14 kcal/mol) for an interruption entailing 2 overlapping pairs of NIBTs. Antiparallel triplexes with GT or TG junction interfaces (non-overlapping pair of NIBTs) (Sequence 18 & Sequence 19) retain canonical reverse Hoogsteen hydrogen bonds (data not shown) at the junction interface like in parallel triplexes. Their binding energies of -76.4 kcal/ mol (Sequence 18) and -82.3 kcal/mol (Sequence 19) are slightly more favorable by~5 kcal/ mol than their parallel counterparts (-71.6 kcal/mol and -76.4 kcal/mol) due to better stacking at the interface.

It is significant that characteristic bending of~25° (Fig 7A and 7C and S10 Fig) is seen in the GT step triplex junction interfaces (Sequences 10, [12] [13] [14] . Also, it has been shown earlier [83] Fig 8. Incidence of alternating G * GC and T * AT triplets in a parallel triplex disrupts base stacking. Stacking interaction at the GT and TG steps of the Hoogsteen strand in the various noncanonical Hoogsteen schemes seen for the sequence 1 during simulation: NC1 to NC4 for T * AT and NC5 to NC7 for G * GC triplets. Loss of stacking is indicated by dark arrows. C1' atom of the sugar is shown as open circle.

doi:10.1371/journal.pone.0152102.g008 that the GA step triplex junction interface also exhibits bending of~17°. Besides, it is known that a tethered GT rich TFO induces a bend in the triplex formed [84, 85] which regulates the transcription of Luciferase gene [86] . Therefore it might be speculated that bent triplexes could serve as recognition sites for triplex binding proteins. Interestingly, an intramolecular triplex formed by GT rich TFO is demonstrated to bind a few hnRNP family proteins [87] . Likewise an intermolecular triplex formed by a CT rich TFO is shown to bind the GAGA transcription factor from drosophila melanogoster [88] .

Present results convincingly provide a mechanistic argument for the experimental observations concerning the ability or lack of it for formation of parallel DNA triplex with GÃGC & TÃAT triplets and preference of antiparallel over triplex. For instance the 22-mer TFO, GGGTT GTGGGTTGTGGGGGTGG targeted to form parallel triplex with the human Ki-ras promoter [77] , entails 5 interruptions (marked bold) resulting in 8 sets of overlapping pairs of NIBTs, of which 3 occur consecutively twice in the sequence. Apart from this, 2 sets of non-overlapping pairs of NIBTs (mini triplex junctions marked in italics and underlined) occur in the sequence. As a consequence, it is expected that a parallel triplex would be less stable owing to the disruption of canonical Hoogsteen hydrogen bonds (5 out of 22) in the 5 triplet interruptions (3 TÃAT & 2 GÃGC). Indeed, this TFO forms only antiparallel triplex. Similarly, the 31-mer TFO, 5'GTTTTTGGGTGTTGTGGGTGTGTGTGGTTTG, designed to inhibit transcription of HIV-1 in infected human cells [75] could bind target duplex only in anti-parallel and not in parallel orientation due to presence of 12 sets of overlapping pairs of NIBTs due to 8 interruptions, spaced in proximity, effecting disruption of Hoogsteen hydrogen bonding in 8 out of 31 triplets. On the other hand, a GÃGC triplet interruption in 5'TTTTGTTTTG [62] and 5'TTC TTCTTGTTTCT [89] does not inhibit parallel triplex formation due to loss of only one Hoogsteen hydrogen bond out of 9. Likewise the 14-mer TFO, 5' GTGGTGGGTGTGTG can form only antiparallel triplex [73] since in parallel triplex it will result in 5 interruptions (10 overlapping pairs of NIBTs) causing loss of Hoogsteen pairs in 5 out of 14 triplets. Similarly the 36-mer TFO, 5'GGTGGTGGGTTGGGGTGGTGGTGTGGTGGTGGTGTT3' (10 interruptions with 19 overlapping pairs of NIBTs), and the 38-mer TFO, 5' GGGGGGGTGGGGGTGT TTGGGTGGTGTGGTGGGGGTGG 3' (8 interruptions with 15 overlapping pairs of NIBTs) targeted against the promoter region of human epidermal growth factor receptor gene and the transcription start site of mouse insulin receptor gene respectively, form only antiparallel triplexes [78] to effect transcription inhibition. Their inability to form parallel triplex is obvious in view of expected loss of Hoogsteen pairs at all the interruption sites (9 out of 36 in 36-mer TFO and 8 out of 38 in 38-mer TFO) making the triplex less viable due to lower stability.

On the other hand, non-overlapping NIBTs in 5'TTTTCmTTTTGGGGGG (one TG step junction) enables it to form a stable triplex with the oligopurine sequence flanking the SV40 origin of replication containing plasmid in COS-1 cells to stall replication [76] . Likewise, a stable parallel triplex [81] is formed with the TFO 5' GGGGGTTTTCTTTT 3', in spite of GT step triplex junction (a pair of non-overlapping NIBTs) which facilitates retention of Hoogsteen pair. This finding that triplex junctions (non-overlapping NIBTs) do not significantly influence triplex formation or its stability gains further support from the observation that the 5' TTTTTGGTTTTTGG 3' forms a triplex in both parallel and antiparallel orientation with a T m of~46°C [80] . Likewise, parallel and antiparallel triplex formed by 5' GGGGTTTTGGGG 3' with 2 triplex junctions (2 sets of non-overlapping pairs of NIBTs) display a similar T m of 57°C & 58°C [72] . But, T m is lowered by~10°C for the parallel triplex formed by the TFO 5' GGTTTGTTTTGTTT 3' compared to its antiparallel counterpart triplex formed by the TFO 5' TTTGTTTTGTTTGG 3'. This can be readily attributed to the loss of 2 (out of 14) canonical Hoogsteen hydrogen bonds at the two GÃGC interruptions (4 overlapping NIBTs) in the parallel triplex [80] . In the antiparallel triplex, reverse Hoogsteen hydrogen bonds are preserved since minor effects of low residual twist of Δt = 10.6°are absorbed by the inherent triplex flexibility. Thus, there exists a direct bearing between the experimental observations and the number and type of NIBTs with MD simulations clearly bringing out the criticality of the mechanistic attributes of base triplet nonisostericity.

Growing evidence for occurrence of nucleic acid triple helices in vivo and their implication in diverse critical biological processes, disease and therapy, makes their structural study imperative. Unlike DNA and RNA duplexes where isosteric WC base pairs dominate, nucleic acid triplexes encounter mixture of isosteric and nonisosteric base triplets. Nature and magnitude of base triplet nonisostericity is solely determined by neighbouring triplets. This together with the number of nonisosteric base triplets (NIBTs) determines triplex forming ability, parallel or non-parallel, with or without sequence dependent structural variations as well as their stability. Although base triplet nonisostericity between the parallel GÃGC & TÃAT triplets is seemingly obvious [74] quantitative estimates of its nature and magnitude and its mechanistic influence are by no means readily comprehendible. The work presented here fills this critical gap by providing quantitative description of base triplet nonisostericity as well as their mechanistic effects in influencing stability and formation of DNA triplexes. It is found that the extreme base triplet nonisostericity (residual twist) prevalent between the parallel GÃGC & TÃAT triplets adds an element of instability, owing to the inevitability of significant stereochemical rearrangements, to the extent of even precluding triplex formation when their incidences are recurrent. Thus, the concept of residual twist has provided a means not only to assess the source and degree of nonisostericity but also has enabled in providing a mechanistic basis for the experimental observations in relation to the feasibility or the lack of same to form parallel DNA triplexes. Although these results are derived from analyses of DNA triplexes they should hold valid in assessing the role of nonisostericity in RNA and RNA.DNA hybrid triplexes as well thus enabling a comprehensive understanding of sequence dependent structural variation in nucleic acid triplexes in general. In fact, this concept has found an application in developing an algorithm to identify triplex forming sequences [90] . The results presented here should aid in designing more specific and efficient TFOs as gene targeting agents (antigene therapeutics). Interestingly, the concept of residual twist proved immensely useful in predicting mechanistic effects of base pair nonisostericity due to juxtaposition of Watson-Crick and different non-Watson-Crick pairs in RNA duplexes [91] .

Parallel and anti-parallel DNA triplexes (Sequences 1-19 in Table 1 ), were built conforming to a 12-fold helix [92] with stereochemistry of the third strand regularized by constrained-restrained molecular geometry optimization and van der Waals energy minimization using X-PLOR [93] . Generated models were then subjected to steepest descent energy minimization using the Sander module of AMBER 12.0 [94] and were used as starting model for MD simulation.

MD simulation is a technique employed to obtain the dynamical aspects of a protein or nucleic acid structure obtained from X-ray, NMR or modelling investigations. AMBER [94] and CHARMM [95] are the widely used force-fields. Modified version of the Cornell et al. force field [96] as in parmbsc0 and its minor variations in the AMBER [97] [98] [99] [100] has provided not only a decisive stabilization of nucleic acid simulations, but also have been able to reproduce with reasonable accuracy several structural and dynamical features of BDNA and non BDNA structures which includes detailed insights into BDNA$ADNA [101] , B!ZDNA transition [102, 103] , base flipping [104] , DNA bending [105] and deformation of DNA/RNA in the presence of proteins [106] etc. AMBER has also been successfully utilised to study properties like stability, folding pathway, role of ions, interaction with drugs and hydration property of various non-BDNA structures like RNA [91] , DNA.RNA hybrids [97, 107] , t-RNA [108, 109] , imotif [110] , Holliday junction [111] , PNA.DNA [112] , G-quadruplexes [113, 114] etc.

Likewise, MD simulations [115] using AMBER force field bring out the feature intermediate to that of A and B DNA of the duplex in a triplex in accordance with geometrical stipulation [116] and NMR data [117, 118] . Also, it has been competently used to explore conformational flexibility [119] , stability [120] , hydration [121] , protonation and folding property [122] etc. of DNA triplexes. In fact, our own earlier investigations on antiparallel triplexes bought out base triplet nonisosteric effects [82, 83] . Given its established utility, AMBER force field is used in the present study of parallel triple helices to enable comparison with antiparallel triplexes.

All MD simulations were performed using AMBER 12.0 program [94] . Triplexes were solvated in a periodic box of TIP3P waters and net-neutralized with Na+ counter ions using Leap module of AMBER Tools 13.0 [123] . Following this, initial equilibration minimization was carried out using steepest descent (2000 cycles) and conjugate-gradient (500 cycles) algorithms, with a positional restraint of 500 kcal/mol Å 2 on the solute atoms. Further minimization was carried out with positional restraints on the solute reduced in steps of 100 kcal/mol Å 2 . Subsequently, minimization was effected for 1000 cycles without any positional restraints. MD simulation was then carried out in an NVT ensemble, during which the entire system was heated from 0 to 300K over 100 ps. Further equilibration MD was carried out for 200 ps under constant pressure (NPT). Following this, production run MD simulations were initiated for 250 ns for Sequence 1 and 100 ns each for triplex sequences in Sequences 2-19 (Table 1) . Equilibration and production run simulations were carried out using the Sander and PMEMD module (optimized for CUDA) of AMBER 12.0 (ff99+parmbsc0) respectively. Periodic boundary conditions and PME under isothermal isobaric conditions (T = 298 K; P = 1 atm) with an integration time step of 2 fs was employed in the simulation. All bonds involving hydrogen were constrained using SHAKE algorithm (tolerance = 0.0005 Å) [124] . Trajectories were analyzed using Ptraj module of Amber Tools 13.0. Base pair, base step parameters and backbone torsion angles were extracted from the output of 3DNA Ver 2.1 [125] using in-house programs. Helical twist angles were calculated with respect to C1'. . .C1' vector due to the presence of nonisosteric base pairs [82, 83] . Hydrogen bond distances and angles were computed using the criteria of donor. . .acceptor distance of < 3.6 Å and hydrogen bond angle >120°. Figures were prepared either using PyMOL (ww.pymol.org) or graphical interface of insight II [126] .

Residual twist (Δt°) is calculated by measuring the angle between the line joining C1 0 . . .C1 0 atoms of the Hoogsteen/Reverse Hoogsteen pair of the superimposed base triples prior to the application of helical twist angle (t) requisite to generate a triple helical structure. Radial difference (Δr Å) corresponds to half the difference between diameters of the GÃGC (14.5 Å) & TÃAT (15.3 Å) base triplets.

Binding free energy (ΔG Bind ) of the triplex structures is estimated by using MM-PBSA [127, 128] module of AMBER 12.0. It is calculated as the binding free energy difference between the bound (triplex) and the unbound (WC duplex and TFO) states.

DG ðbindÞ ¼ G ðTriplexÞ À ðG ðWC duplexÞ þ G ðTFOÞ Þ G Triplex , G WC duplex & G TFO are individually calculated by post-processing of the simulation trajectories taken at 20 ps interval after removing counter ions and water molecules. Free energy (G) for each component is given by; G = ΔE (gas) −ΔG (sol) −TΔS (gas) . The gas-phase energy ΔE gas is calculated as summation over the bond length, bond angle, dihedral angle, van der Waals and electrostatics energy contributions. ΔG solv is calculated by summing the polar (G PB ) and non-polar (G nonpol , sol ) contributions. G PB is calculated by the Poisson-Boltzmann (PB) equation with a dielectric constant value of 1.0 and 80.0 set for solute and solvent respectively. Non-polar contribution G nonpol'sol = γ Ã SASA, wherein SASA stands for solvent accessible surface area. Value of γ corresponds to 0.0072 kcal/mol. Using normal mode analysis, entropy term (TΔS) is calculated as the sum of translational, rotational and vibrational components. For the normal mode calculation distance-dependent dielectric constant energy minimization is used with ε = 4r and convergence value of 0.5. Terminal triplet at the 5' and 3' ends of the triplexes is not considered due to end-effects. 

Monitoring DNA triplex formation using multicolor fluorescence and application to insulin-like growth factor I promoter downregulation

DNA Triple Helices: biological consequences and therapeutic potential

Repair of DNA lesions associated with triplex-forming oligonucleotides

Triplex-induced DNA damage response

Structural features and stability of an RNA triple helix in solution

PolyA tail recognition by a viral RNA element through assembly of a triple helix

Regulation of the RNA-dependent protein kinase by triple helix formation

Plasmid dimerization mediated by triplex formation between polypyrimidine-polypurine repeats

Repair and recombination induced by triple helix DNA

A triple helix within a pseudoknot is a conserved and essential element of telomerase RNA

An intermolecular RNA triplex provides insight into structural determinants for the pseudoknot stimulator of -1 ribosomal frame shifting

Formation of triple-helical structures by the 3'-end sequences of MALAT1 and MENβ noncoding RNAs

A triple helix stabilizes the 3' ends of long noncoding RNAs that lack poly (A) tails

Repression of the human dihydrofolate reductase gene by a non-coding interfering transcript

MEG3 long noncoding RNA regulates the TGF-β pathway genes through formation of RNA-DNA triplex structures

PARTICLE, a Triplex-Forming Long ncRNA, Regulates Locus-Specific Methylation in Response to Low-Dose Irradiation

Human DHX9 helicase unwinds triple-helical DNA structures

A distinct triplex DNA unwinding activity of ChlR1 helicase

Triplex-forming DNAs in the human interphase nucleus visualized in situ by polypurine/polypyrimidine DNA probes and antitriplex antibodies

Triplex DNA in the nucleus: direct binding of triplex-specific antibodies and their effect on transcription, replication and cell growth

Characterization of purine-motif triplex DNA-binding proteins in HeLa extracts

Selective recognition of pyrimidine motif triplexes by a protein encoded by the bacterial transposon Tn7

Triple helix-interacting proteins and cancer

Human initiation protein Orc4 prefers triple stranded DNA

Triplex DNAbinding proteins are associated with clinical outcomes revealed by proteomic measurements in patients with colorectal cancer

Inhibitory effects of expanded GAA.TTC triplet repeats from intron I of the Friedreich ataxia gene on transcription and replication in vivo

Triplex DNA and human disease

Increasing the minicircle DNA purity using an enhanced triplex DNA technology to eliminate DNA contaminants

Stable DNA Nanomachine Based on Duplex-Triplex Transition for Ratiometric Imaging Instantaneous pH Changes in Living Cells

Ling L Triplex DNA logic gate based upon switching on/off their structure by Ag(+)/cysteine

Programmable pH-triggered DNA nanoswitches

Time-resolved fluorescent detection of Hg2+ in a complex environment by conjugating magnetic nanoparticles with a triple-helix molecular switch

DNA Triplexes That Bind Several Cofactor Molecules. Chemistry

Binding cofactors with triplex-based DNA motifs

Fabricating a reversible and regenerable Ramanactive substrate with a biomolecule-controlled DNA nanomachines

Gold nanoparticle-based colorimetric assay of single-nucleotide polymorphism of triplex DNA

Triplex-mediated analysis of cytosine methylation at CpA sites in DNA

Detection of cancer cells using triplex DNA molecular beacons based on expression of enhanced green fluorescent protein (eGFP)

Detection of Pseudomonas aeruginosa by a triplex polymerase chain reaction assay based on lasI/R and gyrB genes

A high-throughput assay for DNA topoisomerases and other enzymes, based on DNA triplex formation

Assembly of Liposomes Controlled by Triple Helix Formation

Triplex molecular beacons as modular probes for DNA detection

An overabundance of long oligopurine tracts occurs in the genome of simple and complex eukaryotes

Triplex-forming oligonucleotide target sequences in the human genome

Intrastrand triplex DNA repeats in bacteria: a source of genomic instability

Spectroscopic and calorimetric studies on the binding of an indoloquinoline drug to parallel and antiparallel DNA triplexes

Spectroscopic and calorimetric studies on the triplex formation with oligonucleotide-ligand conjugates

Triplex-directed recognition of a DNA nanostructure assembled by crossover strand exchange

High-performance liquid chromatography-electrospray ionizationmass spectrometry ligand fishing assay: a method for screening triplex DNA binders from natural plant extracts

New approaches toward recognition of nucleic acid triple helices

Triplex-Inspector: an analysis tool for triplex-mediated targeting of genomic loci

TTS mapping: integrative WEB tool for analysis of triplex formation target DNA sequences, G-quadruplets and non-protein coding regulatory DNA elements in the human genome

The TTSMI database: a catalog of triplex target DNA sites associated with genes and regulatory elements in the human genome

Oxford Handbook of Nucleic Acid Structure

Triple-Helical Nucleic Acids

Triplex DNA structures

Oligonucleotide directed triple helix formation

The triple helix: 50 years later, the outcome

Potential in vivo roles of nucleic acid triple-helices

RNA triplexes: from structural principles to biological and biotech applications

Human therapeutics based on triple helix technology

Triple-helix formation by oligonucleotides containing the three bases thymine, cytosine, and guanine

Relative stability of triplexes containing different numbers of T.AT and C+.GC triplets

The contribution of cytosine protonation to the stability of parallel DNA triple helices

The high stability of the triple helices formed between short purine oligonucleotides and SIV/HIV-2 vpx genes is determined by the targeted DNA structure

Effect of competing self-structure on triplex formation with purine-rich oligodeoxynucleotides containing GA repeats

Nuclear magnetic resonance structural studies of intramolecular purine.purine.pyrimidine DNA triplexes in solution. Base triple pairing alignments and strand direction

Second structural motif for recognition of DNA by oligonucleotide-directed triplehelix formation

Solution structure of a purine.purine.pyrimidine DNA triplex containing G. GC and T.AT triples

Specificity of antiparallel DNA triple helix formation Biochemistry

Spectroscopic studies on ethidium bromide binding to intramolecular parallel and antiparallel triple helices containing T * A:T and G * G:C triplets

Calorimetric analysis of triple helices targeted to the dG3A4G3.dC3T4C3 duplex

Base changes and triple-helix hybridization properties of GT containing third strands: A systematical study

Extension of the range of recognition sequences for triple helix formation by oligonucleotides containing guanines and thymines

Inhibition of transcription of HIV-1 in infected human cells by oligodeoxynucleotides designed to form DNA triple helices

Site-directed inhibition of DNA replication by triple helix formation

Inhibition of nuclear protein binding to the human Ki-ras promoter by triplex-forming oligonucleotides

Binding of triple helix forming oligonucleotides to sites in gene promoters

Structure and stability of X.G.C mismatches in the third strand of intramolecular triplexes

Alternate strand recognition of double-helical DNA by T,G-containing oligonucleotides in the presence of a triple helix-specific ligand

Optimization of alternate-strand triple helix formation at the 5"-TpA-3" and 5"-ApT-3" junctions

New insights into DNA triplexes: residual twist and radial difference as measures of base triplet nonisomorphism and their implication to sequence-dependent non-uniform DNA triplex

Base triplet nonisomorphism strongly influences DNA triplex conformation: Effect of nonisomorphic G * GC and A * AT triplets and bending of DNA triplexes

Design of artificial sequence-specific DNA bending ligands

The design of an agent to bend DNA

Regulation of Transcription by Synthetic DNA-Bending Agents

Selection and identification of proteins bound to DNA triple-helical structures by combination of 2D-electrophoresis and MALDI-TOF mass spectrometry

The GAGA factor of Drosophila binds triple-stranded DNA

Extension of the range of DNA sequences available for triple helix formation: stabilization of mismatched triplexes by acridine-containing oligonucleotides

A dynamic programming algorithm for identification of triplex-forming sequences

An innate twist between Crick's wobble and Watson-Crick base pairs

Structures for the polynucleotide complexes poly(dA) with poly (dT) and poly(dT) with poly(dA) with poly (dT)

Optimization of the CHARMM additive force field for DNA: Improved treatment of the BI/BII conformational equilibrium

A second generation force field for the simulation of proteins, nucleic acids, and organic molecules

Refinement of the AMBER force field for nucleic acids: improving the description of alpha/gamma conformers

al Reference simulations of noncanonical nucleic acids with different χ variants of the AMBER force field: quadruplex DNA, quadruplex RNA and Z-DNA

Toward Improved Description of DNA Backbone: Revisiting Epsilon and Zeta Torsion Force Field Parameters

Refinement of the Cornell et al. Nucleic Acids Force Field Based on Reference Quantum Chemical Calculations of Glycosidic Torsion Profiles

Theoretical study of large conformational transitions in DNA: the B$A conformational change in water and ethanol/water

Reaction path ensemble of the B-Z-DNA transition: a comprehensive atomistic study

A mismatch in a CAG trinucleotide repeat overexpansion provokes left-handed Z-DNA conformation

Molecular dynamics of a kappaB DNA element: base flipping via crossstrand intercalative stacking in a microsecond-scale simulation

Mechanical Model of DNA Allostery

Studying allosteric regulation in metal sensor proteins using computational methods

Structure, recognition properties, and flexibility of the DNA-RNA hybrid

Molecular dynamics simulations of human tRNA Lys,3 UUU: the role of modified bases in mRNA recognition

Free-energy landscape of reverse tRNA translocation through the ribosome analyzed by electron microscopy density maps and molecular dynamics simulations

Structure and mechanical characterization of DNA i-motif nanowires by molecular dynamics simulation

Molecular dynamics of a DNA Holliday junction: the inverted repeat sequence d(CCGGTACCGG) 4

Molecular dynamics structures of peptide nucleic acid x DNA hybrid in the wild-type and mutated alleles of Ki-ras proto-oncogene-stereochemical rationale for the low affinity of PNA in the presence of an AC mismatch

Extended molecular dynamics of a ckit promoter quadruplex

Molecular dynamics simulations of G-DNA and perspectives on the simulation of nucleic acid structures

Molecular dynamics investigations of DNA triple helical models: unique features of the Watson-Crick duplex

Symmetry and molecular structure of a DNA triple helix: d(T)n.d(A)n.d(T)n

Sugar conformations in intramolecular DNA triplexes determined by couping constants obtained by automated simulation of P.COSY cross peaks

Proton nuclear magnetic resonance assignments and structural characterization of an intramolecular DNA triplex

Molecular dynamics simulation study of DNA triplex formed by mixed sequences in solution

Sugimoto N Comparable stability of Hoogsteen and Watson-Crick base pairs in ionic liquid choline dihydrogen phosphate

Triplex hydration: nanosecond molecular dynamics simulation of the solvated triplex formed by mixed sequences

The Triplex-Hairpin Transition in Cytosine-Rich DNA

Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes

3DNA: a software package for the analysis, rebuilding and visualization of threedimensional nucleic acid structures

Calculating Structures and Free Energies of Complex Molecules: Combining Molecular Mechanics and Continuum Models

Converging free energy estimates: MM-PB GB SA studies on the protein-protein complex Ras-Raf

We thank CDAC, Pune, India for computational resources and R.Srivatsan (IBAB) for his help in creating one of the plots.

Conceived and designed the experiments: NY GG TR. Performed the experiments: GG TR. Analyzed the data: GG NY. Contributed reagents/materials/analysis tools: NY. Wrote the paper: GG TR NY.