Metal-responsive structural transformation between artificial DNA duplexes and three-way junctions† †Electronic supplementary information (ESI) available: Full experimental procedures, thermal denaturation experiments, ESI-MS and NMR spectra, and other experimental results. See DOI: 10.1039/c6sc00383d

Metal-responsive structural transformation between DNA duplexes and three-way junction structures was demonstrated utilizing artificial oligonucleotides modified with a 2,2’-bipyridine ligand.


Introduction
DNA branched structures have been proven to be an essential structural motif for developing DNA nanoarchitectures, whose structures are programmed based on the sequence-specicity of DNA hybridization. 1 Since Seeman published a milestone paper demonstrating the construction of immobile DNA junctions, 2 DNA branched structures, especially three-way junctions (3WJs), have been widely utilized as building blocks of two-and three-dimensional nanoscale structures 1 as well as so materials. 3 In addition, 3WJs have been employed as scaffolds for molecular assembly 4 and as reaction spaces. 5 In light of the functional versatility of DNA 3WJ structures, it would be highly advantageous to be able to stabilize and induce these motifs. 6 This would endow DNA-based materials with stimuli-responsiveness.
Metal-ligand coordination is one of the most exploited molecular interactions to develop stimuli-responsive supermolecules and materials. 7 This is also the case for DNA architectures, in which a variety of interstrand metal complexes have been covalently incorporated. 8 Notable examples include the introduction of articial metal-mediated base pairs into DNA duplexes. 9 This approach has yielded a wide range of metal-responsive functional DNA, whose structure, 10 catalytic activity, 11 as well as electrical conductivity 12 can be regulated. In contrast, conjugation of metal complexes with other DNA structural motifs, including triplexes, 13 quadruplexes, 14 and junctions, 15,16 has not been widely explored for the purpose so far. We have previously developed an articial metallo-DNA 3WJ, 15 which was composed of three oligonucleotides containing an unnatural bipyridine-modied nucleoside. Upon addition of Ni II ions, the 3WJ was thermally stabilized by the formation of a tris-(bipyridine) metal complex that crosslinked the three strands. A similar metallo-DNA 3WJ was thereaer exploited for constructing higher-order structures by others, 16 suggesting its potential usefulness as a component of DNAbased materials.
In this study, we have investigated the structural transformation between DNA duplexes and 3WJ structures in response to metal coordination (Scheme 1). Such a structural reorganization involving metallo-DNA conjugates is of great use because the structural motif would be readily embedded into higher-order DNA architectures. The transformation was demonstrated with six DNA strands, whose sequences were designed so as to form both duplexes and 3WJs (Table 1). Three strands (L1, L2, and L3) have a bpy ligand at the middle, and the others (S4, S5, and S6) are complementary to the bpy-modied strands. We expected that the addition of transition metal ions would induce a structural transformation to 3WJs through the formation of a tris(bipyridine) metal complex at the junction core.

Results and discussion
We synthesized a novel bpy-modied nucleoside, which possesses a bpy ligand at the 2 0 -a position (U bpy ). The novel U bpy nucleoside was efficiently incorporated into DNA strands to provide bpy-modied oligonucleotides L1, L2, and L3 (see the ESI †). As described later, a DNA 3WJ containing three 2 0 -a-modied U bpy nucleotides (i.e. L1L2L3) showed a higher thermal stabilization upon metal coordination compared to a 3WJ with 2 0 -b-modied nucleotides used in the previous study. 15 We began our investigation by determining the thermal stability of the hybridization products (Tables 2 and S3 †). Bpy-modied 3WJ L1L2L3 showed a sigmoidal melting curve with a T m ¼ 51.7 C in the absence of metal ions (Fig. 1). When one equiv. of Ni II ion was added, the T m of L1L2L3 was increased substantially to 70.5 C (DT m ¼ +18.8 C). Such a signicant metal-dependent stabilization was not observed for other 3WJs containing fewer than three bpy ligands, i.e. S1S2S3 (DT m ¼ +0.5 C), S1S2L3 (+0.7 C), or S1L2L3 (+5.8 C) ( Fig. S3 and Table  S2 †). Thus, the stabilization was attributed to the formation of an interstrand Ni II (bpy) 3 complex at the core of the 3WJ, similarly to the previously reported metallo-3WJ. 15 The formation of the metallo-3WJ L1L2L3$Ni II was further conrmed by mass spectrometry (Fig. S4 †). In contrast, T m values of the duplexes (L1S4, L2S5, and L3S6), containing one bpy ligand, were not affected by the addition of Ni II ions (Table S3 †), precisely because of the inability to form an interstrand bipyridine-metal complex.
Comparison of the melting temperatures demonstrated that the duplexes (L1S4, L2S5, and L3S6, T m ¼ 62.0, 60.3, and 65.1 C, respectively) were more stable than the 3WJs (L1L2L3 and S4S5S6, T m ¼ 51.7 and 43.3 C, respectively) under metalfree conditions. This suggests that duplex formation is favored in the absence of metal ions. Conversely, addition of one equiv. of Ni II ions made the bpy-modied 3WJ the most stable structure (L1L2L3$Ni II ). This raised the possibility that formation of 3WJs can be induced by metal coordination.
Subsequently, melting proles of a mixture of all six strands (L1, L2, L3, S4, S5, and S6) were analyzed (Fig. 2). In the absence of metal ions, the melting curve showed a one-step transition with a T m of 62.5 C, which is almost an average of the T m values of the duplexes, L1S4, L2S5, and L3S6. This indicated that the duplexes were predominantly formed from the six DNA strands. When one equiv. of Ni II ions was added to the DNA mixture, its melting curve seemed to change to a two-step transition. The T m value of each transition (43.7 C and 69.3 C) is in good agreement with those of the unmodied 3WJ S4S5S6 (43.0 C) and the metallo-3WJ L1L2L3$Ni II (70.5 C), respectively. This melting prole implied preferential formation of the 3WJs (L1L2L3$Ni II and S4S5S6) in the presence of Ni II ions. 17 Notably, Scheme 1 Schematic representation of the metal-responsive structural transformation between artificial DNA duplexes and three-way junctions. Table 1 Sequences of DNA strands used in this study a

DNA
Sequences GTT CCA CGC TAC ACT CGC AG a See the ESI for other strands. b X ¼ U bpy for L1, L2, and L3, X ¼ T for S1, S2, and S3.  the melting curve of the mixture of six natural strands (S1, S2, S3, S4, S5, and S6) was not changed upon Ni II addition, indicating that Ni II (bpy) 3 complexation altered the hybridization behavior of the DNA mixture. The hybridization products were then evaluated by native polyacrylamide gel electrophoresis (PAGE). Fig. 3a shows an image of the gel stained with SYBR Gold dye. The DNA mixture without metal ions showed nearly a single band corresponding to the DNA duplexes (lane 1). Upon addition of one equiv. of Ni II ions, two new bands with lower mobility appeared (lane 2). These bands were ascribable to the 3WJs, S4S5S6 and L1L2L3$Ni II , indicating that Ni II addition induced the formation of the two 3WJs. To quantify the amount of the 3WJs formed, native PAGE analysis was conducted with a FAMlabeled S4 strand (FAM-S4), and then the band intensities of the FAM-labeled products were compared (Fig. 3b). The results showed that the DNA duplexes were exclusively formed in the absence of the metal ions (lane 6) while 3WJs were formed in ca. 60% yield 17 in the presence of Ni II ions (lane 7). Such 3WJ formation was not observed when natural DNA strands were used in place of some of the bpy-modied strands (e.g. a mixture of S1, L2, L3, S4, S5, and S6; Fig. S7 and S8 †). This reinforces the conclusion that the 3WJ formation was induced by the interstrand Ni II (bpy) 3 complexation.
The yield of the 3WJs varied in response to the identity and the amount of metal ions added ( Fig. S9 and S10 †). The yield dropped to ca. 20% when Co II ions were substituted for Ni II , while other ions showed almost no 3WJ induction effects (Fig. 4a). This presumably correlates with the large formation constant of the Ni II (bpy) 3 complex (log b 3 ¼ 20.2). 15,18 The efficiency of the Ni II -mediated 3WJ formation increased in proportion to the Ni II concentration in the range of 0 to 1.2 equiv. and began to decline with excess additions (Fig. 4b). This stoichiometric behavior is almost consistent with a 3 : 1 complexation of the bpy ligand and Ni II ions. These results represent a coordination-driven feature of the structural transformation using the bpy-modied DNA strands.
With the goal of promoting a metal-responsive duplex-to-3WJ transformation, we redesigned the natural counter strands M4, M5, and M6, such that the two nucleobases in the middle were mutated from TA to AT (Fig. 5 and Table S1 †). The Fig. 2 Melting curves of a mixture of the six DNA strands in the absence (white) and presence of Ni II ions (black); (circles) L1, L2, L3, S4, S5, and S6; (squares) S1, S2, S3, S4, S5, and S6. The conditions were the same as for Fig. 1.   resulting duplexes showed lower thermal stability (T m ¼ 55.3, 54.8, and 59.6 C for L1M4, L2M5, and L3M6, respectively) than the original duplexes (L1S4 etc.) (Table S4 †) owing to the existence of mismatch pairs (U bpy -T and A-A). In contrast, the 3WJ M4M5M6 contains no mismatches and thus showed a similar thermal stability (T m ¼ 45.1 C) to the original 3WJ S4S5S6 (43.3 C). Native PAGE analysis (Fig. 5c) revealed that alteration of the relative thermal stabilities led to more efficient 3WJ induction formation (ca. 90%). 19 In addition, subsequent treatment with a chelating agent (EDTA) followed by annealing, to remove the Ni II ions, led to the complete regeneration of the metal-free duplexes (lane 3). These results indicate that the Ni IIresponsive structural transformation under these conditions is both nearly quantitative and reversible.

Conclusions
In conclusion, we have demonstrated a metal-dependent structural transformation between DNA duplexes and 3WJs using bpy-modied articial oligonucleotides. While the mixture of bpy-modied strands and complementary strands exclusively forms duplexes under metal-free conditions, Ni II addition induces the formation of 3WJs. This transformation occurs as a result of the formation of an interstrand Ni II (bpy) 3 complex. In addition, the removal of Ni II ions by EDTA regenerates the duplexes, conrming the metal-ion responsiveness of the structural rearrangement. Stimuli-responsive switching of 3WJ structures has gained much attention as a versatile tool to functionalize DNA materials. 20 Therefore, we believe that the metal-responsive 3WJ transformation presented here will provide an exciting advance in DNA nanotechnology and create new opportunities in DNA-based materials science.

Oligonucleotide synthesis
Oligodeoxynucleotides were synthesized on an Applied Biosystems 394 DNA synthesizer by standard phosphoramidite chemistry. The synthesis of the phosphoramidite derivative of U bpy is presented in the ESI. † The DNA synthesis was carried out on a 1 mmol scale in DMTr-on mode with standard reagents purchased from Glen Research. The coupling time of the nucleosides was extended to 15 min. The products were deprotected in 25% NH 3 solution at 55 C for 8 h. The oligomers were rstly puried and detritylated using a PolyPak II cartridge (Glen Research) and further puried by reverse-phase HPLC (Waters XBridge C18 column, 0.1 M TEAA (pH 7.0)/MeCN gradient, 60 C) (Fig. S1 †). All DNA strands were identied by MALDI-TOF mass spectrometry (see ESI †). The amount of the oligomers was determined based on the UV absorbance at 260 nm. The molar extinction coefficients (3 260 ) of the bpy-modied DNA strands (L1, L2, and L3) were estimated 15 by the sum of the 3 260 value of the bpy group and that of corresponding unmod-ied oligonucleotides calculated by the nearest-neighbor method. Some of the unmodied oligonucleotides puried by HPLC were purchased from Japan Bio Services and used without further purication.

Melting analysis
All samples were prepared by mixing the DNA strands (1.0 mM) in 10 mM MOPS buffer (pH 7.0) containing 100 mM NaCl. Aer addition of NiSO 4 $7H 2 O (Soekawa) or EDTA, the solutions were heated to 85 C and cooled slowly to 5 C at the rate of 1.0 C min À1 . Absorbance at 260 nm was monitored by a UV-1700 spectrophotometer (Shimadzu) equipped with a TMSPC-8 temperature controller while the temperature was raised from 5 C to 85 C at the rate of 0.2 C min À1 . A drop of mineral oil was laid on the sample to prevent evaporation. Normalized absorbance shown in the gures was calculated as follows: The melting temperature (T m ) was determined as an inection point of a melting curve using the LabSolutions T m analysis soware (Shimadzu) with a 17-point adaptive smoothing program. Average T m values of at least 3 independent runs are shown in Tables S2-S4 in the ESI. †

PAGE analysis
General procedure. Samples were prepared by mixing the DNA strands (1.0 mM) in 10 mM MOPS buffer (pH 7.0) containing 100 mM NaCl. Aer addition of metal sulfates (Soekawa), the solutions were heated to 85 C and cooled slowly to 4 C at the rate of 1.0 C min À1 . The gels were prepared using TAMg buffer (40 mM Tris, 76 mM MgCl 2 , 14 mM acetic acid, pH 8.0). The sample was mixed with 6Â loading buffer (not containing urea or EDTA, 1 mL) and applied on an 18% gel (19 : 1). Aer running at 120 V for 3 h in the cool incubator (4 C), the gels were observed using an Alpha imager mini (LMS) with a blue-LED transilluminator (Optocode). For unlabelled samples, the gels were stained with SYBR Gold (Invitrogen). Quantication of each product was accomplished by comparing the band intensities of a 3WJ S4S5S6 (or M4M5M6) with that of a duplex L1S4 (or L1M4), in which S4 (or M4) was labeled with FAM. Averages of at least three independent experiments are shown in gures.
Successive transformation between duplexes and 3WJs. Six DNA strands (20 mM, 1.5 mL each) were combined in 10 mM MOPS buffer (pH 7.0) containing 100 mM NaCl to prepare the sample solutions (27 mL in total). The solutions were heated up to 85 C and cooled slowly to 4 C at the rate of 1.0 C min À1 . One third of the sample solutions (9 mL) were pipetted out and stored at À20 C. To the residual solutions (18 mL), one equiv. of Ni II ions (20 mM, 1 mL) was added. Aer the annealing, half of the solutions (9.5 mL) were taken out and stored at À20 C. Subsequently, 10 equiv. of EDTA (200 mM, 0.5 mL) was added to the rest of the samples (9.5 mL), which were then annealed. All the samples were subjected to native PAGE following the general procedure outlined above ( Fig. 5 and S11 †).