Recognition of Non-standard Base Pairs by Triplex-Forming Oligonucleotides Containing an Expanded Genetic Alphabet

TL;DR

Scientists expanded the genetic alphabet to create new DNA-targeting molecules that work under normal conditions and can detect DNA damage or synthetic bases.

Contribution

The study introduces 12 new modular base triplets for triplex DNA recognition at neutral pH using an expanded genetic alphabet.

Findings

New triplets enable nanomolar-affinity DNA targeting at neutral pH.

The method detects oxidative lesions and synthetic base pairs in DNA.

Triplex-forming oligonucleotides are synthesized both chemically and enzymatically.

Abstract

The sequence-specific recognition of double-stranded DNA by biocompatible molecules is fundamental to molecular medicine and synthetic biology. Triplex-forming oligonucleotides (TFOs) enable programmable major-groove recognition via Hoogsteen base pairing; however, the limited repertoire of natural nucleobases imposes strict constraints on target sequences and requires acidic conditions for stability. Here, we have expanded the triplex recognition space using nucleobases from an artificially expanded genetic information system (AEGIS). Through a systematic evaluation of 120 base triplet combinations, we identify at least 12 new modular triplets that can be combined interchangeably to target duplex DNA containing standard, damaged, or synthetic base pairs with nanomolar affinity at neutral pH. We further demonstrate the versatility of this expanded recognition code by detecting oxidative lesions or AEGIS base pairs in enzymatically assembled duplex constructs using both chemically and enzymatically synthesized TFOs. This generalized framework provides a robust platform for precision gene-targeting, molecular sensing, and nucleic acid nanotechnology.

Figures

b](#F1)).^[13](#R13),[14](#R14)^ [Here, X:R-Y denotes a triplet, where the third strand base X interacts with the duplex base pair R-Y, by bonding to base R.] Parallel triplexes are inherently more stable than antiparallel triplexes because of the isostructural nature of their triplets; however, they rely on cytosine protonation and therefore low pH (<6.0) conditions for assembly. Purine-rich oligonucleotides also readily form non-canonical secondary structures that compete with triplex formation.^[15](#R15),[16](#R16)^ For both motifs, stability relies on relatively high, yet physiologically

a](#F2)). Each nucleobase was assessed against 20 different base pair combinations presenting distinct Hoogsteen hydrogen bonding patterns in the major groove ([Figure 2b](#F2) and S1). These included the standard bases (A-T, G-C, C-G, T-A, 5meC-G, U-A), 8-derivatives of purines often found in damaged DNA (8amA-T, 8amG-C, 8oxoA-T, 8oxoG-C), 5-derivatives of pyrimidines (ΨU-A, ΨC-G, 5ohU-A, 5ohC-G), and AEGIS base pair themselves (2amA-T, B-S^c^, B^c^-S^c^, P-Z, Z-P). In doing so, we identified more than 12 non-standard base triplets that, owing to their modular nature, can be used interchangea

a](#F3)), nine additional sequences (T2-T10) were prepared (Table S4). Constructs T2-T5 probed duplexes bearing multiple purine modifications *via* the formation of T:8amA-T, Z:8amG-C, S^c^:8oxA-T, and K^n^:8oxG-C triplets, respectively. Construct T6 evaluated the recognition of a mixed sequence target containing ~33% pyrimidines through the incorporation of T:ΨU-A and C:ΨC-G triplets. While sequences T6-T10 assess the recognition of duplexes harboring AEGIS base pairs; construct T7 evaluated the formation of multiple T:2amA-T triplets, while constructs T8-T10 assessed recognition of Hachimoji

a](#F5)). Melting profiles for the matched and mismatched base pairs in the absence of TFO showed that, as expected, G-C and 8oxoG-C duplexes were the most thermally stable, followed by 8oxoG-A, with the remaining mismatches displaying significantly reduced stabilities (Figure S12b). Analysis of the corresponding triplex melting profiles revealed that the TFO bound all duplex variants; however, the extent of binding and stabilization was highly triplet dependent. Notably, the K^n^:8oxoG-C triplet was approximately 10 °C more stable than the next most stable triplet, K^n^:G-C or K^n^:8oxoG-A ([

b](#F6)). To this end, we assembled the TFO-T oligonucleotide designed to generate T:B-S^c^ and T:P-Z triplets, using a modified primer-extension protocol previously developed for the synthesis of ZT-containing TFOs by exploiting Z-G pairing under high pH conditions.^[19](#R19)^ The protocol was adapted to enable release of the completed TFO after synthesis through incorporation of an inosine-containing cleavage sequence compatible with EndoV digestion.^[46](#R46)^ Successful assembly and release of the oligonucleotide was confirmed by first demonstrating triplex formation with a standard A-T

Full text

A central goal of molecular medicine and synthetic biology is to create biologically compatible molecules that can selectively bind unique double-stranded sequences in natural or synthetic DNA.^1^ Such recognition agents should distinguish the Watson-Crick base pairs from the major or minor grooves based on their molecular shape, electrostatics, and/or hydrogen bonding capacity. Ideally, they should consist of modular recognition elements that can be strung together and used interchangeably to target sequences with varying base pair compositions and lengths. Moreover, their compatibility with enzymatic synthesis would not only simplify assembly, but also enable their integration within biological systems, for example, as synthetic modulators of gene expression.

Triplex-forming oligonucleotides (TFOs) are promising candidate molecules owing to their unique ability to bind specific DNA sequences in the major groove through programmable Hoogsteen base-pairing interactions, forming stable triplex structures (Figure 1a).^2^ Synthetic TFOs have been used in vitro to detect and/or functionalize natural or synthetic DNA constructs.^3–5^ They have shown potential as gene-targeting tools, modulating gene expression in vitro,^6–8^ in cell culture, and in animal models.^9,10^ Growing evidence also suggests that endogenous RNA modulates gene expression through triplex-based mechanisms.^11^

However, the standard nucleobases impose strict sequence, structural, and ionic constraints on triplex formation. Binding is typically limited to oligopurine-oligopyrimidine tracts, and pyrimidine-purine inversions are destabilizing. Purine-rich TFOs bind antiparallel to the purine strand of the duplex, forming A:A-T and G:G-C triplets.^12^ Pyrimidine-rich TFOs bind parallel, forming T:A-T and C^+^:G-C triplets (Figure 1b).^13,14^ [Here, X:R-Y denotes a triplet, where the third strand base X interacts with the duplex base pair R-Y, by bonding to base R.] Parallel triplexes are inherently more stable than antiparallel triplexes because of the isostructural nature of their triplets; however, they rely on cytosine protonation and therefore low pH (<6.0) conditions for assembly. Purine-rich oligonucleotides also readily form non-canonical secondary structures that compete with triplex formation.^15,16^ For both motifs, stability relies on relatively high, yet physiologically relevant, cation concentrations to offset electrostatic repulsion between the three negatively charged strands.^17^

Much effort has focused on developing synthetic nucleotides to overcome the inherent limitations of triplex formation.^2,18^ However, their incorporation into oligonucleotides often relies on bespoke phosphoramidite chemistry, which restricts their widespread use. An alternative strategy, pioneered by us and others, is to assemble TFOs from modified nucleotide triphosphates via templated or non-templated polymerase synthesis.^19,20^ However, the structural complexity of existing nucleotides, often involving modifications to the base, sugar or phosphate, can hinder compatibility with polymerases. Nonetheless, such strategies have enabled gram-scale synthesis of modified oligonucleotides for antisense applications.^21^

A potential route to overcome this bottleneck is to utilize nucleobases from an artificially expanded genetic information system (AEGIS).^22^ Their nucleotide triphosphates are compatible with natural or engineered polymerases, and the sequences can be amplified,^23–28^ sequenced,^29,30^ and even transcribed.^31^ Moreover, their rearranged hydrogen bonding patterns create orthogonal duplex pairings distinct from those in natural DNA^32^ and could provide a new set of possible Hoogsteen triplet pairing geometries. Perhaps the best-characterized of these duplexes pair the pyrimidines T, Z, S, K, and V (Figure 2) with the purines D, P, B, X, and J, respectively. Four of these base pairs have been used alongside their natural counterparts to generate eight-letter “Hachimoji” DNA.^31^ We have already established that the Z nucleobase is an effective mimic of protonated cytosine (Figure 1b) and can be incorporated into TFOs via an adapted primer extension strategy.^19^

Here, we characterized the triplex-forming properties of the natural (C, T) and all second-generation AEGIS pyrimidines (Z, S^c^, K^n^, and V) within a pH-independent parallel triplex containing Z:G-C triplets (Figure 2a). Each nucleobase was assessed against 20 different base pair combinations presenting distinct Hoogsteen hydrogen bonding patterns in the major groove (Figure 2b and S1). These included the standard bases (A-T, G-C, C-G, T-A, 5meC-G, U-A), 8-derivatives of purines often found in damaged DNA (8amA-T, 8amG-C, 8oxoA-T, 8oxoG-C), 5-derivatives of pyrimidines (ΨU-A, ΨC-G, 5ohU-A, 5ohC-G), and AEGIS base pair themselves (2amA-T, B-S^c^, B^c^-S^c^, P-Z, Z-P). In doing so, we identified more than 12 non-standard base triplets that, owing to their modular nature, can be used interchangeably to recognize DNA sequences containing multiple standard and synthetic base pair combinations with nanomolar affinity at neutral pH. We further highlight the practical utility of this new triplex recognition code by establishing that enzymatically assembled duplex constructs harboring oxidative lesions or AEGIS base pairs can be recognized by modified TFOs containing this expanded genetic alphabet.

RESUJTS

Triplex selectivity of an expanded genetic alphabet

Here, we utilized a model triplex sequence from our previous work, in which a pyrimidine-rich 13-nucleotide TFO binds in a parallel orientation to a 13-base pair oligopurine-oligopyrimidine duplex containing four isolated G-C base pairs (Figure 3a, Table S2). Targeting of such a sequence using unmodified CT-containing TFOs is strongly pH dependent but alleviated by substituting cytosine with the AEGIS nucleobase Z (pKa = 7.8).^33,34^ Such ZT-containing TFOs form highly stable complexes, generating triplexes that are more stable than the underlying duplex at a neutral pH.^19,35^

Using this pH-independent scaffold, we determined the interactions of the six pyrimidine nucleobases with 20 standard and non-standard base pairs, each presenting distinct Hoogsteen donor and acceptor arrangements in the major groove (Figure 2 and S1). By systematically positioning each nucleobase opposite each base pair in turn, we evaluated 120 distinct X:R-Y triplet combinations (Figure 3a). Duplexes were first annealed at pH 7.0 in a sodium cacodylate buffer containing magnesium and subsequently incubated with the appropriate TFO before examining the samples. Because most triplex applications operate in buffers containing 5–60 mM magnesium, we selected 10 mM as a useful reference point and as an approximation of physiological conditions.

Triplex formation was first assessed using an electrophoretic mobility shift assay (EMSA) at neutral pH (Figure 3b). To maximize the sensitivity to differences in complex stability, electrophoresis was performed in a tris-acetate running buffer lacking magnesium.^19,35^ Initial experiments with a CT-containing TFO (con TFO-T: Table S2) confirmed that, as expected, the unmodified oligonucleotide did not bind to its duplex target at neutral pH (Figure 3b). Only when the samples were prepared at pH 5.0 did the anticipated triplex—positioning thymine opposite an A-T base pair—form a stable complex, as evidenced by the reduced mobility of the sample (Figure S3). In contrast, all ZT-containing TFOs generated shifted species at neutral pH, with binding influenced by the identity of the third strand base and the base pair positioned opposite.

Analysis of interactions with standard base pairs revealed the behavior of the nucleobases largely reflected the different hydrogen bonding patterns presented in the major groove (Figure 3b). A-T and G-C present donor-acceptor and acceptor-acceptor configurations, respectively; C-G and U-A present donor-donor and donor-acceptor arrangements when C–H…N/O interactions are considered;^36^ whereas T-A contributes only a single acceptor together with a bulky methyl group that is known to destabilize triplex formation.^34^ Consistent with these patterns, TFOs containing T, Z, and C recognized A-T and G-C base pairs, forming established T:A-T, Z:G-C, and C:G-C triplets, respectively (e.g. Figure 1a). While TFOs containing S^c^ or K^n^ showed no detectable binding to any standard base pair, reflecting the incompatibility of their Hoogsteen pairing geometries.

An exception, however, was observed for the V nucleobase, which showed no interaction with the anticipated A-T base pair and instead displayed weak but reproducible binding to G-C. This was surprising because V mirrors the hydrogen bonding capacity of thymine at the 3- and 4-positions (Figure S4a). To test if selectivity was affected by sequence context, we investigated a second triplex motif positioning the base between Z:G-C triplets, but again, the same preference for binding to G-C was observed by EMSA analysis (Figure S4b and Table S3). One plausible explanation is that the presence of the 2-amino group, which is left uncompensated in a V:A-T triplet, promotes the formation of a V:G-C triplet by shifting the base further into the major groove (Figure S2c). The formation of such a non-conventional triplet is expected to be destabilizing because it is not isostructural with T:A-T and Z:G-C triplets (Figure S4a). Consistent with this interpretation, the incorporation of multiple V:G-C triplets into the triplex sequence substantially reduced the complex stability (Figure S4c).

Further analysis of the gels revealed that none of the nucleobases recognized any of the pyrimidine-purine inversions under the experimental conditions. This is consistent with the reduced donor/acceptor complementarity presented by pyrimidines and/or steric interference from bulky substituents at the 5-position of the pyrimidine ring. Experiments conducted at higher TFO concentrations and in the presence of additional magnesium revealed only modest recognition of a subset of these base pairs, confirming that such interactions are intrinsically disfavored (Figure S5b).

We next evaluated binding of the nucleobases to purines bearing 8-amino and 8-oxo substituents, which introduce additional hydrogen-bonding functionality in the major groove (Figure 3b). The incorporation of an amino group at the 8-position of the purines was not expected to disrupt their recognition by T, C, or Z and could, in principle, form an additional hydrogen bond with the 2-keto group of the bases.^37,38^ Consistent with this expectation, EMSA analysis revealed the formation of stable T:8amA-T, C:8amG-C, and Z:8amG-C triplets. In contrast, the incorporation of a carbonyl substituent at the 8-position alters the Hoogsteen face of the purine, converting N7 from a hydrogen bond acceptor to a donor, thereby reversing the polarity of the Hoogsteen edge. Consequently, T and Z were not expected to recognize these modified purines, whereas C, S^c^, and K^n^, which each present a potential acceptor at N3, were predicted to bind through acceptor–donor complementarity. Again, formation of these triplets might also be stabilized by a third hydrogen bond between the 8-oxo group and the 2-amino group of the bases. Indeed, EMSA analysis confirmed the expected behavior, with S^c^:8oxA-T, C:8oxG-C, and K^n^:8oxG-C triplets producing stable complexes on the gels.

Similarly, we investigated whether non-natural pyrimidines bearing altered 5-position substituents influenced their triplex-forming capacity (Figure 3b). The C-nucleosides ΨU and ΨC were of particular interest,^39^ as both introduce an additional N-H donor into the major groove without perturbing their W-C pairings; although the recognition of ΨC will depend on its tautomeric equilibrium.^40^ EMSA analysis revealed that T, Z, C, and K^n^ each formed detectable triplets opposite a ΨU-A base pair, whereas only C could recognize ΨC-G. The formation of these triplets is consistent with both conventional (T:ΨU-A, Z:ΨU-A, and C:ΨC-G) and non-conventional hydrogen bonding geometries (C:ΨU-A, and K^n^:ΨU-A) (Figure S2). We also assessed interactions with pyrimidines bearing bulky 5-position moieties and, irrespective of the potential to form hydrogen bonds, their presence strongly destabilized triplex formation; no interactions were observed with either 5-methyl (T and 5meC) or 5-hydroxy modified (5ohU and 5ohC) base pairs.

Finally, we assessed TFO binding to duplex sequences containing non-standard AEGIS base pairs (Figure 3b). From the Hoogsteen face of the purine, both 2amA-T and B-S^c^ present the same acceptor-donor pattern as an A-T base pair; accordingly, thymine was expected to form stable T:2amA-T and T:B-S^c^ triplets. EMSA analysis confirmed this prediction, revealing clear formation of both triplets. In contrast, the P nucleobase displays a Hoogsteen edge not found in standard purines owing to replacement of the N7 acceptor with a C-H moiety. Such a substituent might form a non-conventional C-H…O/N hydrogen bond with the nucleobases, for example, that is mediated by an N3 acceptor.^36^ Interestingly, T, C, Z, and K^n^ all showed stable interactions with this base pair, reflecting the formation of both standard (C:P-Z and K^n^:P-Z) and non-standard (T:P-Z and Z:P-Z) triplet geometries (Figure S2). We also assessed binding to the second generation B^c^-S^c^ base pair but, owing to the removal of N7 and altered Hoogsteen face, triplex formation was not detected with any of the bases.

Additional control experiments confirmed that the formation of the modified triplets requires hydrogen bonding, as none of the nucleobases interacted with the universal analogue 5-nitroindole, which lacks hydrogen bonding capacity (Figure 3b). They also confirmed that the shifted species observed on the gels correspond to the same triplex architecture, as running the complexes side-by-side gave identical mobilities (Figure S5c). In addition, the selectivity of the nucleobases was not altered at either pH 5.0 or pH 9.0, despite the potential for protonation of C, S^c^, and K^n^ at their N3 positions (Figure S6). The only exception was with K^n^, which exhibited a weak interaction with 8amG-C at low pH, suggesting that the nucleobase might be partially protonated at pH 5.0.

To quantify the thermal stability of the triplets identified by EMSA analysis, we then performed fluorescence melting using SYBR Green I and a qPCR machine (Figure 3c, S7). SYBR green I fluoresces upon binding to duplex and triplex DNA, and the melting of these complexes results in a decrease in fluorescence intensity at 522 nm.^19^ Duplexes were first annealed at pH 7.0 in sodium cacodylate buffer containing magnesium and subsequently incubated with the appropriate TFO before melting analysis. Triplex formation was identified by an upward shift in the melting profile relative to that of the corresponding duplex-only control. Melting temperatures (Tm) were determined from the first derivatives of the melting curves and used to evaluate the relative stability of each triplet.

Examination of the melting profiles of the duplex sequences in the absence of TFO revealed modest differences in duplex stability, depending on the nature of the base pair under study (Figure S7a). As expected, the most stable base pairs (ca. 57°C) were capable of forming three hydrogen bonds, except for B^c^-S^c^, which exhibited stability comparable to those containing two (ca. 53°C). Melting profiles produced in the presence of all TFOs exhibited clear shifts in Tm values upon binding, with the extent of stabilization dictated by the identity of the base triplet and distributed across a window of approximately 10 °C (Figure 3c and S7b). The relative stabilities of the most stable triplets used later in the study followed the order Z:8amG-C = Z:G-C > T:8amA-T = T:2amA-T = T:A-T = T:B-S^c^ > C:ΨC-G > T:ΨU-A = K^n^:8oxG-C > C:P-Z = S^c^:8oxA-T > K^n^:P-Z >> T:P-Z.

Surprisingly, under these conditions, T:8amA-T or Z:8amG-C triplets were of a similar stability as their parent T:A-T and Z:G-C triplets, respectively, despite the capacity to form three hydrogen bonds. The K^n^:8oxG-C and S^c^:8oxA-T triplets also exhibited comparatively lower stability, despite the potential to form additional hydrogen bonding. However, previous work demonstrated that triplex stability is influenced by the stability of the underlying duplex and, consequently, thermal melting can underestimate stabilization afforded by nucleobase modifications.^41^ Nevertheless, these experiments have established a new triplex recognition code that enables the rational targeting of duplex sequences containing standard, damaged, and non-standard base pairs.

Modular recognition of different base pair combinations

To further assess the modularity and stability of this new triplet recognition code, we designed a panel of oligonucleotide sequences to evaluate the formation of triplexes comprised of multiple modified triplets (Figure 4). Starting from the parent construct comprising only T:A-T and Z:G-C triplets (T1, Figure 3a), nine additional sequences (T2-T10) were prepared (Table S4). Constructs T2-T5 probed duplexes bearing multiple purine modifications via the formation of T:8amA-T, Z:8amG-C, S^c^:8oxA-T, and K^n^:8oxG-C triplets, respectively. Construct T6 evaluated the recognition of a mixed sequence target containing ~33% pyrimidines through the incorporation of T:ΨU-A and C:ΨC-G triplets. While sequences T6-T10 assess the recognition of duplexes harboring AEGIS base pairs; construct T7 evaluated the formation of multiple T:2amA-T triplets, while constructs T8-T10 assessed recognition of Hachimoji DNA via formation of T:B-S^c^ triplets and either T:P-Z, C:P-Z, or K^n^:P-Z triplets, respectively.

EMSA analysis at pH 7.0 confirmed that all TFO constructs successfully assembled into discrete, slower-migrating species in a concentration-dependent manner, despite the significant chemical deviation from natural DNA (Figure S8). The most stable complexes—those containing T:A-T and Z:G-C (T1), their 8-amino (T2 and T3) or 2-amino derivatives (T7), as well as K^n^:8oxG-C triplets (T4)—remained intact in tris–acetate running buffer lacking magnesium, indicating particularly robust Hoogsteen complementarity. Successful binding was also observed for sequences harboring S^c^:8oxA-T (T5) triplets or the combination of T:ΨU-A and C:ΨC-G triplets (T6) but required magnesium ions to maintain structural integrity during electrophoresis. Similarly, triplex formation was also evident for all sequences containing Hachimoji base pairs, combining T:B-S with either T:P-Z, C:P-Z, or K^n^:P-Z triplets. Circular dichroism spectroscopy confirmed that the shifted species corresponded to triplex structures, with all complexes displaying the characteristic enhancement of the negative band near 210 nm, consistent with the adoption of an A-like triplex conformation upon TFO binding (Figure S9).^42^

UV melting analysis was performed to assess the thermal stability of the resultant complexes (Figure S10, Table 1). Six of the ten triplexes exhibited cooperative melting transitions that shifted to higher temperatures relative to the corresponding duplex-only controls. Notably, the Tm values for the triplexes incorporating T:8amA-T, Z:8amG-C, K^n^:8oxG-C, and T:2amA-T triplets were indistinguishable, clustering at 68 °C. In contrast, the S^c^:8oxA-T triplet was less stable, displaying a substantially reduced Tm of 47 °C. The remaining four triplexes exhibited a single melting transition that overlapped with the underlying duplex. This was most prominent for Hachimoji constructs containing multiple B-S^c^ and P-Z base pairs, reflecting the high intrinsic stability of the underlying duplex. Importantly, no additional low-temperature transitions were observed for any of these constructs, indicating the dissociation of a single, well-defined complex.

The thermodynamic parameters for triplex formation were then quantified using isothermal titration calorimetry (ITC) (Figure 4 and Table 1). As a reference, the unmodified triplex sequence containing T:A-T and C^+^:G-C triplets (T0) exhibited weak affinity (KD = 3180 ± 1900 nM) and poorly defined binding, as expected at neutral pH (ΔG = −7.50 ± 0.36 kcal mol^−1^) (Figure S11a). Although a large apparent enthalpy was observed, this was almost completely offset by a substantial entropic penalty, consistent with the non-productive or transient association of the oligonucleotide. In stark contrast, replacing the four C^+^:G-C triplets in this sequence with Z:G-C (T1) led to roughly a 300-fold enhancement in TFO affinity (KD = 11 ± 1.8 nM) at this pH, with a larger ΔG (−10.8 ± 0.1 kcal mol^−1^) driven by a favorable enthalpic contribution (ΔH = −57.7 ± 0.6 kcal mol^−1^).

Across the remainder of the panel, triplex formation was uniformly exergonic, with ΔG values spanning −10.3 to −11.7 kcal mol^−1^ and KD values in the 3–30 nanomolar range (Figure 4).

Triplets incorporating 8-amino purines (T2 and T3) displayed the most favorable enthalpic profiles (ΔH ≈ −65 ± 1.0 kcal mol^−1^), consistent with additional or strengthened hydrogen bonding interactions in the major groove. However, they exhibited reduced apparent stoichiometries, suggesting partial duplex destabilization arising from the presence of uncompensated amino substituents, consistent with the reduced duplex stability observed by UV melting. Triplets incorporating oxidized purines retained productive binding but with attenuated enthalpic driving forces: S^c^:8oxA-T (T4) exhibited the weakest driving force (ΔH = −27.0 ± 1.1 kcal mol^−1^), while K^n^:8oxG-C (T5) showed intermediate strength (ΔH = −39.1 ± 0.6 kcal mol^−1^). The triplex containing T:ΨU-A and C:ΨC-G triplets (T6) also displayed comparable thermodynamic stability to these complexes (ΔG = −10.4 ± 0.1 kcal mol^−1^; ΔH = −36.2 ± 0.7 kcal mol^−1^) despite the presence of ~33% pyrimidines in the sequence.

Interestingly, the heat profile obtained for the T:2amA-T containing triplex (T7) revealed biphasic binding, indicative of two thermodynamically distinct association events. Increasing the number of injections during the experiment allowed us to resolve these contributions (Figure S11b). The dominant event was the strongest interaction measured in this study (KD = 2.9 ± 0.3 nM; ΔG = −11.7 ± 0.1 kcal mol^−1^) and driven by a large favorable enthalpy (ΔH = −58.8 ± 0.9 kcal mol^−1^), consistent with efficient Hoogsteen recognition of duplexes preorganized by 2-amino adenine. The weaker event (KD = 734 ± 28.0 nM; ΔG = −8.37 ± 0.02 kcal mol^−1^) likely reflects duplex heterogeneity of the sample, where the presence of the 2-amino on adenine promotes alternative duplex conformations that support a less efficient triplex association.

Finally, targeting Hachimoji DNA through the formation of T:B-S^c^ triplets and either T:P-Z, C:P-Z, or K^n^:P-Z generated stable complexes with KD values of 12–23 nM and ΔG values clustered between −10.4 and −10.8 kcal mol^−1^. These interactions were consistently enthalpy-driven, although with smaller ΔH values than those observed for the standard or amino purines, suggesting that the steric and/or electronic features of the triplets were less optimized.

Collectively, these data establish that modular combinations of modified triplets enable stable, well-defined triplex formation across chemically diverse duplex targets. Triplex stability is primarily governed by favorable enthalpic contributions arising from optimized Hoogsteen interactions. These thermodynamic trends correlate closely with the relative stabilities inferred from melting analysis, providing a mechanistic framework for future base triplet designs.

Recognition of 8oxG is enzymatically assembled constructs

8oxG is one of the most prevalent DNA lesions in genomic DNA and is highly mutagenic, commonly pairing with adenine during replication to generate GC→TA transversions.^43^ The ability to distinguish 8oxoG-C from the mispair 8oxoG-A or any other W-C base pair, within long DNA constructs, is of strong desire. We therefore investigated the interactions of K^n^ with base pair mismatches that might be introduced through replication errors by DNA polymerase, as well as with 8oxG lesions introduced into plasmids or PCR products. Experiments were undertaken at pH 7.4 to better reflect physiological conditions.

The selectivity of K^n^ across relevant base-pairing contexts was investigated using fluorescence melting analysis. Triplexes were assembled with a single K^n^:G-Y or K^n^:8oxG-Y triplet with a systematic variation of the base at position Y (Figure 5a). Melting profiles for the matched and mismatched base pairs in the absence of TFO showed that, as expected, G-C and 8oxoG-C duplexes were the most thermally stable, followed by 8oxoG-A, with the remaining mismatches displaying significantly reduced stabilities (Figure S12b). Analysis of the corresponding triplex melting profiles revealed that the TFO bound all duplex variants; however, the extent of binding and stabilization was highly triplet dependent. Notably, the K^n^:8oxoG-C triplet was approximately 10 °C more stable than the next most stable triplet, K^n^:G-C or K^n^:8oxoG-A (Figure 5b). We also assessed the ability of K^n^ to detect the extent of oxidation in duplex constructs by assaying its interaction with samples containing different ratios of 8oxoG-C to G-C base pairs by EMSA analysis (Figure 5c). As expected, as the percentage of 8oxoG-C increases, the extent of triplex formation increases.

We next assessed whether TFO-K^n^ could target a triplex-forming sequence containing a single 8-oxoG lesion embedded within longer DNA fragments assembled by primer extension. These experiments used a modified pUC18 plasmid containing the same triplex target sequence used above (T1).^19^ A 31-mer mutagenic primer was designed to substitute the central A-T base pair with an 8oxoG–C pair upon extension by Taq polymerase, thereby recreating the same local sequence context as the synthetic 8oxoG-C duplex also used above (Table S5). Following thermal cycling, the parental plasmid was digested with DpnI, and the resulting linear product (Figure 5d(i)) was probed with 5′-radiolabelled TFO-K^n^. Separation by PAGE revealed concentration-dependent interactions; however, binding was only detectable at the highest plasmid concentration tested, owing to the low yield of plasmid generated under these non-PCR conditions (Figure 5d(ii)). Equivalent experiments were undertaken with a non-mutagenic primer and probing with TFO-K^n^ revealed, as expected, no interaction of the oligonucleotide (Figure S13b).

To improve the sensitivity of detection, a PCR-based amplification strategy was then implemented using the same mutagenic primer in combination with a second primer located approximately 250 bp upstream in the plasmid. The resulting product was assayed as a function of PCR cycle by probing with 5′-radiolabelled TFO-K^n^ (Figure 5d). Under these exponential amplification conditions, binding was clearly observable after only four PCR cycles and increased progressively with further amplification of the target construct. Equivalent experiments were undertaken with a non-mutagenic primer and probing with TFO-K^n^ revealed only minimal interaction of the oligonucleotide after 32 cycles (Figure S13c).

Taken together, the data demonstrates that TFOs containing the nucleobase K^n^ can selectively assay the presence of 8-oxoG-C base pairs in both short and long DNA fragments assembled enzymatically, establishing a practical route for the detection of this mutagenic lesion in biologically relevant constructs.

Recognition of AEGIS base pairs in enzymatically assembled constructs

AEGIS base pairs B-S^c^ and P-Z form part of an eight-letter (“Hachimoji”) genetic alphabet, which is both replicable and transcribable.^31^ The ability to recognize assembled duplex sequences containing these base pairs would enable the direct detection and/or functionalization of this synthetic genetic information. Earlier in this study, we established that T allows stable recognition of B-S, while T, C, or K^n^ can each be used to recognize P-Z through stable Hoogsteen interactions. We therefore sought to determine if these modified TFOs could target a Hachimoji DNA sequence assembled enzymatically (Figure 6).

To assemble the modified construct, we utilized a simple primer extension assay to generate the same Hachimoji duplex containing B-S^c^ and P-Z base pairs used above (T8-T10) (Figure 6a). The template was designed to encode the modified pyrimidine strand, with its purine complement assembled from a mixture of standard and AEGIS nucleotide triphosphates (Table S6). Primer and template strands were annealed and subjected to polymerase extension at 72 °C in the presence of the four natural dNTPs together with dBTP and dPTP. Because incorporation of dPTP opposite Z is known to be influenced by pH,^44,45^ these reactions were performed at pH 7.8 and with an excess of dPTP to promote efficient incorporation into the extended duplex.

Extension reactions were carried out using either Taq polymerase, which exhibits limited tolerance for unnatural nucleotides, or KlenTaq polymerase, which lacks 3′–5′ exonuclease proofreading activity and has previously been shown to be more permissive towards modified substrates. Reaction products were analyzed by PAGE (Figure 6a(i)). In both cases, polymerase extension was evident, yielding fragments longer than the primer-template duplex with reduced electrophoretic mobility, consistent with successful incorporation of AEGIS nucleotides. However, reactions performed with Taq polymerase also produced a population of shorter, aborted extension products, indicative of reduced processivity on the modified template. Subsequent experiments were therefore conducted exclusively using KlenTaq to maximize yield and homogeneity of the AEGIS-containing duplex substrates used for triplex targeting studies.

Having demonstrated successful assembly of the Hachimoji duplex, we then investigated whether it could be targeted by the synthetic TFOs that position T opposite B-S^c^ and either T, C, or K opposite P-Z (Figure 6a(ii)). In all cases, triplex formation was observed, as evidenced by the formation of slower migrating species that ran at an identical position on the gel as an equivalent synthetic A-T control duplex targeted by TFO-T.

We next asked whether an enzymatically assembled TFO could target the synthetic AEGIS-containing duplex substrate studied above (T8–T10; Figure 6b). To this end, we assembled the TFO-T oligonucleotide designed to generate T:B-S^c^ and T:P-Z triplets, using a modified primer-extension protocol previously developed for the synthesis of ZT-containing TFOs by exploiting Z-G pairing under high pH conditions.^19^ The protocol was adapted to enable release of the completed TFO after synthesis through incorporation of an inosine-containing cleavage sequence compatible with EndoV digestion.^46^ Successful assembly and release of the oligonucleotide was confirmed by first demonstrating triplex formation with a standard A-T duplex target, yielding the expected shifted species by PAGE (Figure 6b(i), lane 7). Using this validated approach, the enzymatically assembled TFO was then incubated with the synthetic Hachimoji duplex and analyzed by EMSA, revealing clear, concentration-dependent binding to the AEGIS-containing target. (Figure 6b(ii)).

Finally, we examined whether an entirely enzymatic workflow could be implemented by combining the enzymatically assembled TFO with the enzymatically assembled AEGIS duplex construct. Both components were prepared independently using the corresponding polymerase-driven protocols and subsequently incubated together. Once again, a distinct concentration-dependent triplex species was observed by PAGE analysis, demonstrating successful triplex formation (Figure 6). Collectively, these experiments establish that both the recognition element and the target duplex can be generated enzymatically while retaining full triplex-binding functionality, thereby enabling a fully polymerase-compatible interface for the assembly and recognition of triplexes containing an expanded genetic alphabet.

DISCUSSION

DNA recognition by parallel triplex formation is governed by specific Hoogsteen hydrogen bonding patterns within the duplex major groove.^13,14^ By systematically probing and reprogramming these interactions using nucleobases from an artificially expanded genetic information system, we have established a second, orthogonal triplet recognition code that operates efficiently under physiologically relevant conditions. The modular nature of this code allows the nucleobases to be strung together and used interchangeably to recognize sequences containing diverse arrangements of standard, damaged, and synthetic base pairs, maintaining nanomolar affinity across highly modified targets.

Interactions of the nucleobases largely reflected what was anticipated based on shape and hydrogen bonding complementarity (Figure S2). Recognition of purines typically involves bonding between the 3- and 4-positions of the third strand nucleobase and the 7- and 6-positions of the purine within the base pair.^13,14^ Conversely, recognition of pyrimidines generally involves the 2- and 3-positions of the third strand base and the 5- and 4-positions of the pyrimidine.^47,48^ Formation of the Z:G-C, T:8amA-T, Z:8amG-C, S^c^:8oxA-T, K^n^:8oxG-C, C:ΨC-G, T:ΨU-A, T:2amA-T, T:B-S^c^, C:P-Z, and K^n^:P-Z triplets all appear to adhere to these established geometries, with putative triplet structures shown in Figure 7.

Several deviations from the standard triplet geometries were observed that likely rely on bonding between the 2- and 3-positions of the third strand base and the purine partner. Perhaps most notably, T recognized P-Z (Figure 7), while V, which was expected to bind to A-T, recognized G-C in its place (Figure S4a). The preference of V to bind G-C is particularly intriguing and might be explained by a requirement to compensate the 2-amino group and reduce electrostatic repulsion or unfavorable desolvation penalties.^32^ However, S^c^ and K^n^ also possess 2-amino groups and did not form equivalent triplet geometries with A-T base pairs in this configuration. This hypothesis could be tested using 5-nitro uridine which more closely mirrors T while retaining the nitro group. A second possibility is that the 5-nitro substituent on V may perturb the triplet’s electronic structure in a way that specifically disfavors A-T pairing. In particular, steric interactions with the adjacent 4-carbonyl may force the nitro group out of plane, decreasing the acidity and positioning of the ring N-H group, weakening hydrogen-bond donation. This contrasts with Z and K, which also contain 5-nitro groups but possess adjacent 4-amino substituents capable of intramolecular hydrogen bonding that likely enforce planarity and enhance hydrogen-bonding interactions (e.g. Figure S14).

Across the panel, no simple correlation between hydrogen-bond number and triplet stability was observed, highlighting the additional roles of geometric complementarity, electronic and tautomeric effects, π-stacking, and steric constraints within the major groove. EMSA and melting analysis showed that the most stable triplets generated with purines involved Z opposite G-C or 8amG-C, followed closely by T opposite A-T, 8amA-T, 2amA-T, or B-S^c^. Intermediate stability was observed for K^n^ opposite 8oxG-C and C opposite P-Z, with further reductions in stability for S^c^ opposite 8oxA-T and K^n^ opposite P-Z. ITC measurements corroborated this hierarchy and highlighted that the Z:8amG-C, T:8amA-T, and T:2amA-T triplets, each capable of additional hydrogen-bonding interactions on either the Hoogsteen or Watson–Crick face, were the most stable of the triplets studied. Previous work by Orozco and co-workers demonstrated the stabilizing effects of C^+^:8amG-C and T:8amA-T triplets within pH-dependent triplex motifs;^37,49^ here, we extend this strategy to neutral pH through the use of pH-independent Z:8amG-C and/or Z:G-C triplets.

The most stable triplet generated with pyrimidines arose from interactions of C, K^n^, Z, and T opposite ΨU-A, and of C opposite ΨC-G. Combining T:ΨU-A and C:ΨC-G enabled triplex recognition of a mixed target sequence containing 33% pyrimidines. Although related strategies have been reported previously,^39^ combining these triplets with the Z:G-C triplet markedly enhances triplex stability at neutral pH and, in principle, provides a means to regulate the expression of synthetic genes incorporating these base pairs.^8^

The ability of K^n^ to detect 8oxG lesions is particularly significant, representing the first demonstration of stable triplex recognition of an oxidative lesion under neutral pH conditions. K^n^ was prioritized over its non-nitro counterpart to minimize the likelihood of N3 protonation at neutral pH.^50^ Consistent with this, experiments performed at pH 5.0 revealed only partial protonation of the nucleobase, as evidenced by the formation of a K^n^:8amGC triplet. Moreover, melting analysis showed that a single K^n^:8oxG-C triplet is approximately 10 °C more stable than the next most stable triplets formed with G-C or the 8oxG-A base pair mismatch. Notably, selectivity of the nucleobase is retained with long, enzymatically assembled DNA fragments, establishing a non-destructive strategy for the direct detection of oxidative damage in biologically relevant constructs.

One of our original motivations for integrating AEGIS nucleobases into triplex DNA—either in the third strand or within the duplex—is their inherent compatibility with polymerase-mediated assembly. As a proof-of-principle, we show that Hachimoji duplex sequences generated from modified nucleotide triphosphates can be targeted by synthetic TFOs as well as by TFOs produced enzymatically. This capability opens several avenues. First, because AEGIS-containing sequences do not cross-pair with natural DNA, they prevent unintended priming in complex genomic samples, enabling fully orthogonal amplification architectures in which TFO-based molecular beacons could act as selective readers of this amplified information. Second, oligonucleotides containing AEGIS nucleobases have been used extensively to construct aptamers and non-standard DNA architectures^51^ and TFOs provide a means to fold or functionalize such structures.^4^ Third, the biological compatibility of AEGIS triplexes creates a feasible path toward coupling TFO assembly with the regulation of natural or artificial genes constructed from an expanded genetic alphabet.

However, two challenges remain. The stability of triplets formed with P-Z are currently limited by suboptimal hydrogen bonding, likely arising from weak C-H…N interactions between the 3-position of the third stand base and 7-position of P.^36^ Derivatives of P that address this imbalance are therefore desirable. Targeting mixed pyrimidine/purine sequences also remains difficult because T, S^c^, and Z introduce bulky substituents into the major groove that destabilize triplex formation. Potential solutions include replacing T with pseudouridine, as shown in this study; substituting S^c^ with isocytosine or pseudocytosine analogs,^52^ to remove the methyl group from the major groove; and/or developing alternative variants of Z that retain pH-independent G-C recognition while reducing steric effects.^45,53^

The next phase of this work will focus on structural characterization of the modified base triplets to define their precise hydrogen-bonding and stacking geometries. These insights will establish a mechanistic framework for the rational design of next-generation triplex-forming nucleotide analogues, enabling the exploitation of triplex DNA as a new programmable recognition and regulatory platform in molecular medicine and synthetic biology.

MATERIAJS & METHODS

Oligonucleotides:

The sequence of each of the unmodified and modified oligonucleotides used in this study are shown in Tables S2–6. Unmodified oligonucleotides were purchased from Sigma Aldrich (UK), oligonucleotides containing the synthetic nucleobases 8oxA, 8oxG, 8amA, 8amG, ΨC, ΨU, 5ohC, 5ohU, 2amA, and 5NI were purchased from ATDbio (UK), while oligonucleotides containing the synthetic nucleobases Z, S^c^, K^n^, V, B, B^c^, and P were provided by Firebird Biomolecular Sciences (US).

Nucleotide triphosphates:

Unmodified deoxyribonucleotide triphosphates were purchased from Promega (UK), whilst modified deoxyribonucleotide triphosphates of Z, B, and P were provided by Firebird Biomolecular Sciences (US).

Electrophoretic mobility shift assay (EMSA):

Interactions of the TFOs with duplexes containing standard and non-standard base pairs were determined by native EMSA at different pH. Oligonucleotides were prepared in 10 mM sodium cacodylate containing 10 mM magnesium chloride at either pH 5.0, 7.0, or 9.0 as indicated by the experiment. The final duplex concentration was 1 μM and the final TFO concentration varied between 30 μM and 0.01 μM in a total volume of 20 μL. Duplex strands were annealed before addition of the TFOs by heating the strands at 95 °C for 5 minutes and slowly cooling the sample to room temperature over 2 hours. Samples were then left to equilibrate for greater than 16 hours at 4 °C. Samples were subjected to non-denaturing polyacrylamide gel electrophoresis at 100 V for ~2 ½ hours in 40 mM tris-acetate running buffer (Sigma Aldrich) at either pH 5.0, 8.3, or 9.0 in the presence or absence of 10 mM magnesium acetate as indicated by the experiment. Gels were post-stained with GelRed (Biotium) and visualised using a Gel Doc XR+ Imaging System (BioRad).

Fluorescence melting:

Thermal melting profiles for the duplexes and triplexes were determined using SYBR Green I (Thermofisher) and either a MyGo Pro or Roche LightCycler.^54^ Oligonucleotides were prepared in 10 mM sodium cacodylate containing 10 mM magnesium chloride at pH 7.0. The final duplex concentration was 1 μM and final TFO concentration was 2 μM in a total volume of 20 μL. Duplex strands were annealed before addition of the TFOs by heating the strands at 95 °C for 5 minutes and slowly cooling the samples to room temperature over 2 hours. Samples were then left to equilibrate for greater than 16 hours at 4°C, before adding SYBR green I dye at a final working concentration. Fluorescence melting profiles were obtained by recording the emission of SYBR Green I at 522 nm, after an excitation at 488 nm, at a slow ramp rate of 0.2 °C/min between 30 °C and 80 °C. [Continuous temperature change of the LightCycler caps at 0.1°C/s, however slower melting profiles were obtained by increasing the temperature by 1°C/step, leaving the samples to equilibrate over a set period of time.] Melting temperatures (Tm) were determined from the first derivatives of each profile using the analysis software provided by the machines and varied by less than 0.5 °C between experiments.

Ultraviolet (UV) melting:

Thermal melting profiles for the duplexes and triplexes were determined using a Shimadzu UV-3600 UV-Vis-NIR spectrophotometer. Oligonucleotides were prepared in 10 mM sodium cacodylate containing 10 mM magnesium chloride at pH 7.0. The final duplex concentration was 5 μM and final TFO concentration was 10 μM in a total volume of 100 μL. Duplex strands were annealed before addition of the TFOs by heating the strands at 95 °C for 5 minutes and slowly cooling the complexes to room temperature over 2 hours. Samples were then left to equilibrate for greater than 16 hours at 4°C. UV melting profiles were recorded at 260 nm, 1 s response time, in a Shimadzu 8 series micro multi-cell with a 10 mm pathlength between 30 °C and 80 °C at a ramp rate of 0.2 °C/min. Melting temperatures (Tm) were determined from the first derivatives of each profile using the analysis software provided by the machine, and varied by less than 1 °C between experiments.

Circular dichroism (CD):

CD spectra for the duplexes and triplexes were determined using either an Applied Photophysics PiStar-180 or a Chirascan VX spectrophotometer. Oligonucleotides were prepared in 10 mM sodium cacodylate containing 10 mM magnesium chloride at pH 7.0. The final duplex concentration was 5 μM and final TFO concentration was 10 μM in a total volume of 300 μL. Duplex strands were annealed before addition of the TFOs by heating the strands at 95°C for 5 minutes and slowly cooling the complexes to room temperature over 2 hours. Samples were then left to equilibrate for greater than 16 hours at 4°C. Spectra were collected between 320–200 nm, at 100 nm/min, 1 s response time, 1 nm bandwidth in Hellma synthetic quartz cuvettes with a 1 mm pathlength. Each spectrum was accumulated five times and averaged to smooth.

Isothermal titration calorimetry (ITC):

Thermodynamic parameters for each of the triplexes were determined using a MicroCal PEAQ-ITC Auto (Malvern Panalytical). Oligonucleotides were prepared in 10 mM sodium cacodylate containing 10 mM magnesium chloride at pH 7.0. The calorimeter cell held the duplex target at 5 μM in a total volume of 200 μL, while the syringe contained the TFO binding partner at a concentration of 60 μM. Titration experiments were performed at 25°C at a reference power of ~9.5 μcal/s. Unless otherwise stated, nineteen aliquots of 4 μL were injected from the syringe at 2.5 min intervals with syringe screw-paddle agitation at 750 rpm. The binding isotherms processed and the thermodynamic parameters obtained using the PEAQ-ITC analysis software v1.52. Residual heat of dilution produced from injecting the TFO was automatically offset by the software and validated by TFO-into-buffer controls.

Site-directed mutagenesis (SDM):

An 8-oxoG lesion was introduced into a triplex target sequence previously cloned in pUC18 using a modified version of the Agilent^®^ QuickChange site-directed mutagenesis protocol. The 31-mer mutagenic forward primer carried a single 8oxoG residue that, upon polymerase extension, replaced a single adenine in the extension product. The overlapping reverse primer contained thymine at the same position to introduce a different substitution on the opposing strand. Reactions contained 20 ng of plasmid DNA, 125 ng of each primer, and 10 mM dNTPs, and were carried out using Taq polymerase (New England Biolabs) in 1X Thermopol^®^ buffer in a final volume of 50 μL. Samples were cycled for 12 rounds in a BioRad^®^ T100 thermal cycler with an initial denaturation at 95 °C for 30 s, followed by annealing at 58 °C and extension at 72 °C. Following assembly, reaction products were digested with 20 units of DpnI for 1 hour 30 min at 37 °C to remove parental (methylated) plasmid DNA. Samples were then subjected to a further round of extension to produce the final linearised constructs.^55^ A control construct was also prepared using a non-mutagenic forward primer. For triplex binding experiments, the assembled constructs were used directly and diluted 4-, 12-, 40-, 100-, and 400-fold in the assay.

Polymerase chain reaction (PCR):

A single 8-oxoG lesion was introduced into a triplex target sequence previously cloned in pUC18 using PCR to generate a base-modified 251-bp fragment. The 31-mer mutagenic forward primer carried a single 8oxG residue that, upon polymerase extension, replaced the corresponding adenine in the extension product. A second 31-mer reverse primer was positioned upstream of the sequence to generate the final fragment. The reaction contained 5 ng plasmid, 1 μM forward and reverse primers, 10 mM dNTPs and was performed by Taq polymerase (1.25 units, New England Biolabs) in 1X Thermopol^®^ buffer in a total volume of 50 μL. Samples were cycled for either 2, 4, 8, 16, and 32 rounds in a BioRad^®^ T100 thermal cycler with an initial denaturation at 95 °C for 30 s, followed by annealing at 56 °C and extension at 68 °C. A control construct was also prepared using a non-mutagenic forward primer. For triplex binding experiments, the assembled constructs were used directly.

Oligonucleotide radiolabelling:

To assay TFO binding to enzymatically assembled constructs containing 8oxG, the oligonucleotide carrying a single K^n^ modification (TFO-K^n^) was radiolabelled at its 5′-end using T4 polynucleotide kinase (New England Biolabs) and 1 μL [γ−^32^P]ATP (Revvity) in 1X polynucleotide kinase buffer in a total volume of 20 μL for 1 hour. The enzyme was then deactivated by heating the sample at 65 °C for 30 minutes and the oligonucleotide isolated using a G-25 spin column (Cytiva) to remove excess ATP according to the manufacturer’s instructions. After elution, the oligonucleotide was at a final concentration of ca. 0.5 μM in ddH_2_0, yielding approximately 10 c.p.s./μL, as determined by a handheld Geiger counter.

Duplex assembly by primer extension:

The modified duplex sequence containing AEGIS base pairs (B-S^c^/P-Z) was assembled by primer extension of a template containing the appropriate modified nucleobases. Primer and template strands were dissolved in ThermoPol^®^ reaction buffer (New England Biolabs) at pH 7.8 at a final concentration of 1 μM in a total volume of 20 μL. The strands were annealed by heating at 95°C for 5 minutes and slowly cooled to room temperature over 2 hrs. Deoxyribonucleotide triphosphates (i.e., dATP/dTTP/dBTP/dPTP) were then added at a final concentration of 500 μM, except for dPTP which was added at a 5-fold excess to promote its incorporation opposite Z.^25^ The reaction mixture was pre-incubated at 72 °C for 30 seconds before addition of 2 units of either Taq or KlenTaq polymerase (New England Biolabs).^56^ Samples were left >2 hour to allow the extension reaction to reach completion. For triplex binding experiments the assembled duplex was used directly.

TFO assembly by primer extension:

The modified ZT-containing TFO (TFO-T) was assembled by primer extension on a primer-template containing an inosine residue adjacent to the TFO sequence to enable its release after assembly.^19,46^ Primer and template strands were dissolved in ThermoPol^®^ reaction buffer (New England Biolabs) at pH 8.8 at a final concentration of 1 μM in a total volume of 20 μL. Deoxyribonucleotide triphosphates (e.g., dCTP/dTTP or dTTP/dZTP) were then added at a final concentration of 500 μM. The reaction mixture was pre-incubated at 72 °C for 30 seconds before addition of 2 units of Therminator^®^ DNA polymerase (New England Biolabs). Samples were left >2 hour to allow the extension reaction to reach completion. EndoV (New England Biolabs) was then used to release the assembled TFO from the extension product by addition of the enzyme (10 units) at 37 °C for 2 hours before its inactivation at 80 °C for 10 minutes. For triplex binding experiments the assembled TFO was used directly.

Supplementary Material

Supplementary Files

This is a list of supplementary files associated with this preprint. Click to download.

• NatCommSI.docx

Contains oligonucleotide sequences, and additional PAGE, CD, fluorescence melting, UV melting, and ITC data.