Divalent cations and molecular crowding buffers stabilize G-triplex at physiologically relevant temperatures

G-triplexes are non-canonical DNA structures formed by G-rich sequences with three G-tracts. Putative G-triplex-forming sequences are expected to be more prevalent than putative G-quadruplex-forming sequences. However, the research on G-triplexes is rare. In this work, the effects of molecular crowding and several physiologically important metal ions on the formation and stability of G-triplexes were examined using a combination of circular dichroism, thermodynamics, optical tweezers and calorimetry techniques. We determined that molecular crowding conditions and cations, such as Na+, K+, Mg2+ and Ca2+, promote the formation of G-triplexes and stabilize these structures. Of these four metal cations, Ca2+ has the strongest stabilizing effect, followed by K+, Mg2+, and Na+ in a decreasing order. The binding of K+ to G-triplexes is accompanied by exothermic heats, and the binding of Ca2+ with G-triplexes is characterized by endothermic heats. G-triplexes formed from two G-triad layers are not stable at physiological temperatures; however, G-triplexes formed from three G-triads exhibit melting temperatures higher than 37°C, especially under the molecular crowding conditions and in the presence of K+ or Ca2+. These observations imply that stable G-triplexes may be formed under physiological conditions.

particularly Ca 21 , promote the formation of G-triplexes. This finding is different from that for G-quadruplexes, whose stabilities are almost unaffected by these two bivalent cations.

Results and Discussion
CD spectroscopy in dilute solutions. Six G-rich oligonucleotides were used to investigate the factors that influence G-triplex formation (Table 1). Three oligonucleotides (TBA, Hum21 and T 2 T 2 T 3 ) with four G-tracts have the potential to form Gquadruplexes. The 3'-most G-tract of each of these oligomers was truncated to generate the TBA11, Hum15 and T 2 T 2 fragments, which potentially form G-triplexes. As G-triplet and G-quartet share similar stacking and loop geometry, it is reasonable to assume that circular dichroism (CD) signals reflecting the strand orientation in the G-quadruplex may also apply to the G-triplex. We therefore used CD spectroscopy to examine the secondary structures of these six oligonucleotides in the presence of different metal ions, as well as under dilute and molecular crowding conditions. CD spectroscopy was first conducted under dilute conditions. The thrombin-binding aptamer, TBA, is a well-known Gquadruplex-forming G-rich oligonucleotide that has been studied extensively. The truncated form of TBA, TBA11, has been shown to form a G-triplex structure by Limogelli and co-workers previously 7,8 . The CD spectra of TBA in a dilute solution containing K 1 or Na 1 had a positive peak at approximately 292 nm and a negative peak at approximately 269 nm ( Figure 2). These results are characteristic of antiparallel G-quadruplexes 24 , indicating that K 1 and Na 1 promote antiparallel G-quadruplex formation of TBA. Under the same conditions, TBA11 in presence of K 1 revealed a similar CD spectrum profile with a positive peak around 288 and a negative peak around 264 nm, indicating the formation of antiparallel G-triplex by TBA11. These features are identical to those reported by Limogelli et al in which G-triplex formation was confirmed by NMR investigations 7,8 . However, it should be noted that the CD signal intensity of the TBA11 G-triplex was much lower than that of the TBA Gquadruplex, suggesting the probability of G-triplex formation is less than that of G-quadruplex. Under these dilute conditions, neither TBA nor TBA11 showed obvious CD signals in the presence of Ca 21 or Mg 21 .
Next, we investigated human telomeric sequences, Hum21 and Hum15 ( Figure 2). Under the dilute conditions without metal ions, the positive peak at around 294 nm and the negative peak at around 266 nm in the CD spectrum of Hum21 indicated partial formation of antiparallel G-quadruplexes. The presence of K 1 or Na 1 further promoted the formation of G-quadruplex structures. In contrast to  TBA, Hum21 exhibited antiparallel G-quadruplex structure in the presence of Na 1 and parallel/antiparallel-mixed hybrid Gquadruplex structure in the presence of K 1 ion 25 . Both K 1 and Na 1 promoted G-triplex formation by Hum15 and the effect of K 1 was notably stronger than that of Na 1 . In the presence of K 1 or Na 1 , Hum15 appeared to form a parallel/antiparallel hybrid G-triplex, as indicated by a positive CD peak at approximately 290 nm, a shoulder peak at approximately 265 nm and a negative CD peak at approximately 235 nm. The features of CD spectrum in the Ca 21 buffer are rather different from those in K 1 and Na 1 buffers. This may be due to different strand orientations in the G-triplex. In the presence of Ca 21 , the   CD spectrum of Hum15 had a strong positive peak at approximately 262 nm and a negative peak at 236 nm, which are typical for parallel strand arrangement 24 , Under the same conditions, the CD spectrum of Hum21 in the presence of Ca 21 had a negative peak at approximately 236 nm and two weak positive peaks at approximately 268 nm and 297 nm. It is possible that a mixture of two Gquadruplex structures was present in the solution; however, the possibility of a mixture of G-triplexes and G-quadruplexes cannot be excluded. Mg 21 did not notably promote G-quadruplex formation of Hum21. However, it seemed that Mg 21 promoted the formation of G-triplexes, as evidenced by a positive peak at approximately 263 nm in the Hum15 CD spectrum.
Under dilute conditions, T 2 T 2 T 3 formed parallel or antiparallel Gquadruplex structures in the presence of K 1 or Na 1 , respectively ( Figure 2). However, the presence of either K 1 or Na 1 promoted the formation of hybrid G-triplex structures of T 2 T 2 . The hybrid structure differed from those formed by Hum15 in the presence of K 1 , as CD peaks for T 2 T 2 in the presence of K 1 or Na 1 were stronger at 266 nm than at 294 nm. In contrast, CD peaks for Hum15 in the presence of K 1 or Na 1 were stronger at 294 nm than at 266 nm. In the presence of Ca 21 or Mg 21 , T 2 T 2 showed similar CD spectrum profiles to Hum15, suggesting they share similar G-triplex conformations.
Overall, the presence of Na 1 , K 1 , Ca 21 or Mg 21 under dilute conditions promoted the formation of G-triplexes to some degree ( Figure 3). However, the effects of K 1 and Ca 21 were much stronger than those of Na 1 and Mg 21 . In addition, K 1 and Ca 21 affected Gtriplex formation differently. It seems that G-rich sequences tend to form parallel G-triplexes in the presence of Ca 21 . However, G-triplexes formed in the presence of K 1 may have different sequencedependent conformations. It is possible that G-triplexes formed under this condition may have a strand orientation similar to corresponding G-quadruplexes (Table S1). In contrast to the well-known promotion of the formation and stabilization of Gquadruplexes by K 1 and Na 1 , the presence of Ca 21 or Mg 21 promotes G-quadruplex formation weakly at best. Importantly, Ca 21 and Mg 21 selectively promote the formation of G-triplexes.
CD spectroscopy under molecular crowding conditions. It has been reported that molecular crowding can affect the confor- In the presence of 40% (v/v) PEG 200, K 1 promoted the formation of G-triplexes in the TBA11, while Na 1 , K 1 , Ca 21 and Mg 21 promoted G-triplex formation in the Hum15 and T 2 T 2 fragments. As observed for dilute conditions, Gtriplexes formed in the Hum15 and T 2 T 2 may assume parallel strand orientation in the presence of Ca 21 or Mg 21 under molecular crowding conditions. The G-triplexes formed in the presence of K 1 or Na 1 exhibited similar strand orientations to the G-quadruplexes formed in the corresponding longer oligonucleotides. Similar to Gquadruplex, we found molecular crowding promoted parallel Gtriplex formation for some oligonucleotides. For example, in the presence of K 1 , T 2 T 2 showed a potential G-triplex structure with strand orientations similar to the hybrid G-quadruplex in a dilute buffer. However, the structure more likely assumes a parallel strand orientation under the molecular crowding conditions ( Figure 4).
CD spectra of four mutants lacking the potential to form Gtriplexes. To support that observed CD spectra reflect the formation of G-triplexes, four mutants, Hum15-1, Hum15-2, T 2 T 2 -1 and T 2 T 2 -2 (Table 1), were designed by replacing G residues essential for the G-triplex formation with nucleotide substitutions (C or T). As expected, the four mutant oligonucleotides exhibited the same CD spectra under either diluted or crowded buffers, regardless of the presence of metal ions (Na 1 , K 1 , Mg 21 or Ca 21 ) ( Figure 5 and Figure 6). The observed positive peaks at approximately 280 nm were characteristic of single-stranded or double-stranded DNA 27 . These results support that the ion-or molecular-crowding-induced changes in CD signals shown in Figures 2 and 4 were not caused by simple interactions between the metal ions and the nucleotide bases. Instead, they could be the result of G-triplex formation.
Melting temperature (T m ) assay was used to demonstrate the formation of G-triplexes. G-rich sequences with fewer than four G-tracts might possibly also form intermolecular G-quadruplexes 28,29 . Thus, the DNA secondary structures formed by TBA11, Hum15 and T 2 T 2 might be intermolecular G-quadruplexes rather than Gtriplexes. It has been reported that the stability will increase with DNA concentration for intermolecular structures but not for intramolecular structures 30,31 . To demonstrate the intramolecular G-triplexes formation by these G-rich oligonucleotides, DNA-concentration dependent stability change of the proposed G-triplexes was investigated. The stability change can be reflected by the melting temperature (T m ) of the DNA secondary structures. Similar to Gquadruplexes 32 , DNA secondary structures formed in the TBA11, Hum15, or T 2 T 2 exhibited decreased UV absorption at 295 nm with increasing temperature ( Figure S1). In contrast, the oligonucleotides without G-quadruplex and G-triplex-forming potential did not exhibit similar behaviour ( Figure S2). These experiments support that stable structures, likely those with a stack of G-triads as suggested from CD experiments, exist in TBA11, Hum15, or T 2 T 2 fragments. These results also imply that, as with G-quadruplexes, the temperature-dependent changes in 295 nm UV absorbance can be used to determine the T m of the G-triplex structures.
To exclude the possibility of intermolecular G-quadruplex formation, T m changes of the three G-rich oligonucleotides (TBA, Hum15, T 2 T 2 ) were determined with different DNA concentrations (7-15 mM). The results indicate that the T m values were independent of DNA concentration in the presence of individual metal ions and under both diluted and crowded buffers ( Figure S3-S5, Table S2), confirming that these three oligonucleotides form intramolecular structures, i.e., G-triplexes.
The G-rich Hum9 oligonucleotide (Table 1), a truncated human telomeric sequence, was used to further exclude the possibility of intermolecular G-quadruplex formation. Because it has only two GGG repeats, it can form only intermolecular G-quadruplex structures. The result of CD spectroscopy indicated that only the CD spectra under K 1 condition differed from the control without ions ( Figure S6), indicating that only K 1 promoted the formation of intermolecular G-quadruplexes (i.e., Na 1 , Mg 21 , or Ca 21 did not). This result supports that the DNA secondary structures formed by Hum15 and T 2 T 2 , especially in the presence of Na 1 , Mg 21 or Ca 21 , were intramolecular G-triplexes and not intermolecular G-quadruplexes. Melting temperature analysis indicated that the intermolecular G-quadruplexes formed by Hum9 were less stable than the proposed G-triplexes formed by Hum15 in the presence of K 1 , thus suggesting that Hum15 may preferentially form more stable intramolecular G-triplexes in the presence of K 1 .
Stabilization of G-quadruplexes and G-triplexes by metal ions and molecular crowding conditions. Having demonstrated the promoting effect of Na 1 , K 1 , Ca 21 and Mg 21 on the formation of Gtriplexes under both dilute and molecular crowding conditions, the effects of these ions on G-triplex stability were compared ( Figure S1). As summarized in Table 2, under both diluted and crowded conditions, the triplex-stabilizing effect of K 1 was stronger than Na 1 . This result corresponds well with their effects on G-quadruplex stability. G-quadruplexes formed in the presence of Ca 21 exhibited lower T m values than those formed in the presence of K 1 , indicating that the G-quadruplex-stabilizing effect of K 1 is better than that of Ca 21 ion 32 . In contrast, Ca 21 exhibited stronger G-triplex-stabilizing effects than Na 1 , K 1 or Mg 21 . The strength of G-triplex-stabilizing effects was ranked as Ca 21 .K 1 .Mg 21 .Na 1 . In the presence of Na 1 , K 1 , Mg 21 or Ca 21 , G-triplexes formed in crowded buffers exhibited higher T m values than those formed in dilute solutions, indicating that molecular crowding stabilizes G-triplexes. Similar results were obtained for G-quadruplex studies 19 .  In the presence of Na 1 or K 1 , G-triplexes exhibited lower T m values than corresponding G-quadruplexes, suggesting that Gquadruplexes are more stable than the corresponding G-triplexes.
Although the G-quadruplex consisting of two G-tetrads (TBA) was relatively stable in the presence of K 1 (T m of 51.1uC under the molecular crowding condition), the G-triplex consisting of two Gtriads (TBA11) was much less stable (T m of 34.3uC in the presence of Ca 21 and under the molecular crowding condition). This observation may indicate that G-triplexes with two G-triads might not be stable at physiological temperatures. In contrast, the G-triplexes with three Gtriads are much more stable, especially in the presence of K 1 or Ca 21 . Under the molecular crowding conditions, the T m values of Ca 21stabilized G-triplexes (Hum15 and T 2 T 2 ) were 52.0uC and 64.5uC, respectively. With K 1 or Ca 21 stabilization, G-triplex structures with three G-triads might be stable at physiological temperatures.
Thermodynamic profiles for the formation of G-triplexes and Gquadruplexes. The T m provides a rough measurement for the stability of a structure at a specific temperature. To quantify thermodynamic stabilities, we used a two-state transition model to evaluate the free energy change at 37uC (DG h 37 ) of these G-rich sequences by analyzing the melting curves ( Figure 7, Table S3) 30 . Under both dilute and molecular crowding conditions, TBA11 gave positive DG h 37 values in the presence of either metal ion, confirming that G-triplexes containing two G-triads cannot be formed at physiological temperatures. However, negative DG h 37 values were found in Hum15 and T 2 T 2 , especially in the presence of K 1 or Ca 21 , indicating that stable G-triplexes can be formed at physiological temperatures in the presence of these two ions. The favourable DG h 37 is a result of the characteristic compensation of a favourable enthalpy term with an unfavourable entropy term (Table  S3) 33 . In the presence of K 1 , the DG h 37 values of G-quadruplexes are more favourable than those of corresponding G-triplexes, and the Gquadruplexes stabilized by K 1 generate more favourable DG h 37 values compared to those stabilized by Ca 21 . These results demonstrate that K 1 is a better G-quadruplex stabilizer than Ca 21 .
In the presence of Ca 21 , G-triplexes yield DG h 37 values comparable to or even more favourable than G-quadruplexes. In addition, the Gtriplexes stabilized by Ca 21 have more favourable DG h 37 values than those stabilized by K 1 . These results suggest that Ca 21 is a better Gtriplex stabilizer than K 1 , which is opposite to the scenario in Gquadruplex.
Binding affinities of K 1 or Ca 21 to G-quadruplexes or G-triplexes. Isothermal titration calorimetry (ITC) is a quantitative technique that can determine the binding affinity (K a ) of the interaction between two or more molecules in solution. This technique has been used to study the binding constant between G-quadruplexes and metal ions or ligands 34,35 . Herein, the interactions of K 1 and Ca 21 with G-quadruplexes or G-triplexes were also investigated using ITC assays. As shown in Figure 8, the binding of K 1 to Hum21 and Hum15 is accompanied by exothermic heats. K 1 binds the Gtriplex Hum15 with a K a value of 735 6 65 M 21 , which is ,17fold weaker than that of K 1 and the G-quadruplex Hum21 (K a 5 (1.24 6 0.12) 3 10 4 M 21 ). This is consistent with the above observation that K 1 has higher stabilizing ability to Gquadruplexes than to G-triplexes. The binding of Ca 21 with Gquadruplexes or G-triplexes is characterized by endothermic heats. In contrast to K 1 , Ca 21 shows a higher binding affinity to G-triplex (K a 5 (2.94 6 0.41) 3 10 4 M 21 ) than to G-quadruplex (K a 5 (8.86 6 0.92) 3 10 3 M 21 ). And as expected, Ca 21 shows a higher binding affinity to G-triplex than K 1 . These results are also consistent those of CD and melting assays, thus further demonstrating that Ca 21 is a better G-triplex stabilizer than K 1 .
Ca 21 concentration-dependent T m change of Hum15. As Ca 21 exerted the strongest G-triplex-promoting and stabilizing effects, the effect of Ca 21 concentration on the T m of the Hum15 G-triplex was investigated. As shown in Figure 9, under the dilute condition, temperature-dependent absorption signal change was not observed in the absence of Ca 21 , indicating that Hum15 cannot form Gtriplexes without Ca 21 . G-triplex formation was promoted by the addition of 2 mM Ca 21 , and the resulting G-triplex exhibited a T m of 35.0uC. The T m of the G-triplex increased as the Ca 21 concentration was increased from 2 mM to 100 mM (Table S4). Under the molecular crowding conditions, even in the absence of Ca 21 , Hum15 can form some G-triplexes which have a T m of 25.9uC. Increment of T m with Ca 21 concentration was also observed in crowded buffers. These results indicate that Ca 21 indeed stabilizes G-triplexes under both diluted and molecularly crowded conditions. Inside cells, K 1 has a concentration of ,140 mM, Na 1 has a concentration of ,10 mM, whereas both Ca 21 and Mg 21 have significant concentrations of mM-mM 36,37 , especially under certain biological activities. Therefore, it is of high physiological significance that these cations would promote the formation of the G-triplexes and stabilize these structures. With the finding that molecular crowding also favours the G-triplex formation, it is possible that like G-quadruplex, G-triplex may exist in vivo for potential biological functions.
The promotion effect of Ca 21 to G-triplex formation is confirmed at the single-molecular level. Optical tweezers represent a unique approach to investigate the mechanical stability of folded nucleic acid structures at the single-molecule level 14,[38][39][40][41] . This technique has been used to demonstrate the formation of G-triplex structures formed by human telomeric sequences containing three GGG repeats under dilute conditions and in the presence of Na 1 ion 14 . To further demonstrate the effect of Ca 21 on the G-triplex formation, optical tweezers assay was conducted to unfold structures in the Hum15 under three conditions (100 mM K 1 , 100 mM Ca 21 , and 100 mM K 1 1 2 mM Ca 21 ). When single-stranded Hum15 tethered between two duplex handles was stretched, a characteristic unfolding event was observed from the force-extension curve, indicating that folded structure was formed in the Hum15 (Figure 10 and Figure S7-S10).
In the presence of 100 mM Ca 21 , this structure could be disrupted at the rupture force of 31 6 3 pN, with the change in contour length (DL) of 5.1 6 0.5 nm. This DL value matches very well with that of Gtriplex 14 , thus strongly supporting that Hum15 folds into G-triplex structure. The possibility of G-triplex formation increased from 21 6 2% under 100 mM K 1 to 48 6 3% under 100 mM Ca 21 , indicating that the capability of Ca 21 to promote G-triplex formation is better than that of K 1 . The G-triplex-promoting ability of Ca 21 could also be observed when Ca 21 has a low concentration. For example, in the presence of 100 mM K 1 1 2 mM Ca 21 , 32 6 2% Hum15 could fold into G-triplexes.
Since stable optical trapping requires a significant difference in the refractive index of a trapped particle and solvent, we used 40% dimethyl sulfoxide (DMSO) to simulate molecular crowding condi- www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 9255 | DOI: 10.1038/srep09255 tions 23 . The mechanical unfolding results showed that molecular crowding conditions indeed could promote the formation of Gtriplexes. In the presence of 40% DMSO, the possibility of G-triplex formation increased from 21 6 2% to 29 6 2% under 100 mM K 1 , and from 32 6 2% to 35 6 3% in the presence of 100 mM K 1 1 2 mM Ca 21 .
Interestingly, the rupture forces under these conditions are similarly located at 31-35 pN (Figure 10, Figure S7-S10). These values are higher than the stall force of many RNA polymerase 42 , which suggests that the G-triplexes may serve as a mechanical block for transcription processes.

Conclusion
In summary, we demonstrated that G-triplex structures can be formed by G-rich sequences containing only three G-tracts. In addition, molecular crowding and the presence of physiological important metal ions, such as Na 1 , K 1 , Ca 21 and Mg 21 , promote and stabilize G-triplex formation. The strength of the G-triplexstabilizing effects of these ions is found to be: Ca 21 .K 1 .Mg 21 .Na 1 . Ca 21 promotes the formation of parallel G-triplexes under either dilute or molecular crowding conditions. G-triplexes formed in the presence of K 1 often have strand orientations similar to those of Gquadruplexes formed in the corresponding oligonucleotides containing four G-tracts. The melting temperatures of the G-triplexes consisting of two G-triads are lower than physiological temperature, even under the molecular crowding condition and in the presence of Ca 21 . However, the G-triplexes containing three G-triads have much higher melting temperatures, especially under the molecular crowding condition and in the presence of K 1 or Ca 21 , implying that stable G-triplexes may be formed at physiological temperatures. The binding of K 1 to G-triplexes is accompanied by exothermic heats, and the binding of Ca 21 with G-triplexes is characterized by endothermic heats. The G-triplex-promoting and stabilizing effect of Ca 21 is noteworthy, as Ca 21 is a physiologically important metal ion and it exhibits rather weak or undetectable effect on the Gquadruplex structures. As G-quadruplexes are believed to have important biological function and the number of putative Gtriplex-forming sequences might be more prevalent in the human genome than putative G-quadruplex-forming sequences, our study may pave a way for further studies on the G-triplexes, such as whether G-triplexes form in vivo and whether they have biological functions.  Methods Materials and reagents. The oligonucleotides listed in Table 1 were purchased from Sangon Biotech. Co. Ltd (Shanghai, China). The concentrations of the oligonucleotides were represented as single-stranded concentrations. Single-stranded concentration was determined by measuring the absorbance at 260 nm. Molar extinction coefficient was determined using a nearest neighbour approximation (http://www.idtdna.com/ analyzer/Applications/OligoAnalyzer). Na 2 EDTA (Disodium ethylenediamine tetraacetic acid), Tris [tris(hydroxymethyl)aminomethane], KCl, NaCl, MgCl 2 , CaCl 2 , PEG 200, HCl were obtained from Sigma. Deionized and sterilized water (resistance . 18 MV/cm) was used throughout the experiments.
Circular dichroism (CD) spectroscopy. Under dilute condition, 3 mL reaction mixture was prepared in 10 mM Tris-HCl buffer (pH 5 7.0) containing 2.5 mM individual DNA oligonucleotides, 0.5 mM Na 2 EDTA, and 100 mM individual metal ions. Under molecular crowding conditions, 3 mL reaction mixture was prepared in 10 mM Tris-HCl buffer (pH 5 7.0) containing 2.5 mM individual DNA oligonucleotides, 0.5 mM Na 2 EDTA, 400 mL/L PEG 200, and 100 mM individual metal ions. The mixture was heated at 95uC for 5 min, cooled to 25uC and then incubated at 4uC overnight. CD spectrum of the mixture was recorded between 230 and 320 nm in 1-mm path length cuvettes on a Jasco J-715 spectropolarimeter. Spectra were averaged from three scans, which were recorded at 100 nm/min with a response time of 1 s and a bandwith of 1.0 nm.
Melting temperature detection of G-quadruplexes or G-triplexes. Melting temperature detection of G-quadruplexes or G-triplexes was carried out on a Cary-60 UV-vis spectrophotometer equipped with a single cell peltier temperature control accessory. Under dilute condition, the G-quadruplexes (10 mM) or G-triplexes (10 mM) solution were prepared in 10 mM Tris-HCl buffer (pH 5 7.0) containing 0.5 mM Na 2 EDTA, 100 mM Na 1 or 100 mM K 1 or 100 mM Mg 21 or 100 mM Ca 21 . Under molecular crowding conditions, additional 400 mL/L PEG 200 was added. The solution was heated to 95uC for 5 min, then cooled rapidly to 25uC and was allowed to incubate at 25uC for 30 min and overnight incubation at 4uC. After a sufficient mixing, the absorption signal at 295 nm (400 nm as control wavelength) was recorded at about 10uC. When the absorption signal became constant, the temperature was increased in steps of 1uC. At each temperature, the mixture was left to equilibrate for at least 1 min. Absorption signal was recorded when the signal did not change any more.
Isothermal titration calorimetry. Isothermal titration calorimetry (ITC) measurements were performed using a MicroCalTM isothermal titration calorimeter iTC200 (GE Healthcare). DNA (50 mM) and Ca 21 (1 mM) solutions were both prepared in 10 mM Tris-HCl buffer (pH 5 7.0). K 1 (1 mM) solution was prepared in 10 mM Tris-HCl buffer (pH 5 7.0) containing 0.5 mM Na 2 EDTA. All of the solutions were heated to 95uC for 5 min, then cooled rapidly to 25uC and were allowed to incubate at 25uC for 30 min and overnight incubation at 4uC. Then the ion (K 1 or Ca 21 , 1 mM) solution was titrated into the corresponding DNA (50 mM) solution. The titration included an initial injection of 0.4 mL ion solution followed by 19 injections of 2 mL ion solution every 120 s with stirring at 750 rpm at 16uC. To define the baseline, the ion was titrated into the same buffer without DNA under the same conditions. The titration data and binding plots after the baseline were subtracted were analyzed using MicroCal Origin software with the one-site binding model.
Optical tweezers single-molecule assay. The DNA construct for single-molecule assay was prepared by sandwiching Hum15 (Table 1) between two double-stranded DNA (dsDNA) handles. One DNA handle (2028 bp) was labelled with biotin at the 59-end, the other DNA handle (2690 bp) was labelled at the 39-end by digoxigenin (Dig). The prepared DNA construct was incubated with anti-Dig antibody-coated polystyrene beads (diameter: 2.17 mm) in 10 mM Tris-HCl buffer (pH 7.0) for 1 h at room temperature to attach the DNA on the beads via the Dig/anti-Dig linkage. The beads attached with DNA construct and the beads coated with streptavidin (diameter: 1.87 mm) were injected into the top and bottom channels of a three-channel microfluidic chamber, respectively. The two types of beads were trapped by laser tweezers in the middle channel 43 , to which 10 mM Tris-HCl buffer (pH 7.0) containing different metal ions was injected. The two trapped beads were brought close to each other to tether the other end of the DNA construct to the streptavidincoated bead via biotin/streptavidin linkage. Then, the steerable mirror of the laser tweezers that controls the anti-Dig-coated bead was moved away from the streptavidin-coated bead with a load rate of ,5.5 pN/s, and the force-extension (F-X) curves were recorded at 1000 Hz using a LabView program (National Instruments Corporation, Austin, TX). The secondary structure formed in the DNA molecule was unfolded when tension inside the tether was increased to a particular level. These raw data were filtered with a Savitzky-Golay function with a time constant of 10 ms using a Matlab program (The Math Works, Natick, MA). The change in contour length (DL) at a particular force (F) was calculated as the extension difference between the stretching and the relaxing traces at that force 44 .