Introduction

Transitions between A- and B-form duplexes were discovered by Franklin and Gosling when conducting X-ray fibre diffraction analyses under various humidities1. In solution, protein binding reactions can also partially dehydrate duplex DNA, giving global A-form viral genomes2, as well as local A-form perturbations at specific binding sites3. Small molecules such as polyamines4, aminoglycosides5,6, hexaamminecobalt (III)7, and cisplatin8,9 can induce local B → A transitions via mechanisms independent of global changes in hydration and water activity. The binding of metal ions to discreet coordination sites10 in nucleic acids can be coupled to the (re)folding of DNA and RNA molecules11,12,13,14,15,16 that activate DNAzymes17,18, ribozymes19,20, riboswitches21,22,23, and DNA-based materials24,25,26,27,28,29,30,31. Previous studies mostly focused on characterizing the changes in structure and function of metal-bound versus metal-free (apo) nucleic acids. Here we report the structures and dynamics of two interconverting structures of the same metallo duplex, where local changes in metal-nucleobase ionization and isomerization are directly coupled to a global conformational exchange. This observation, together with the structural differences between metal-bound and apo duplex structures, enabled our design and identification of a duplex DNA containing 15 C-T mismatches that exhibits a global B → A helical transition upon adding HgII.

Transition metal ions participate in reversible binding interactions between opposing nucleobases32,33,34,35,36,37. T-HgII-T provided the first such example of an all-natural metal-mediated base pair composed of a pyrimidine-pyrimidine mismatch and a transition metal ion38,39,40. T-HgII-T base pairs exhibit very high kinetic and thermodynamic stabilities41,42,43, and can serve as functional mimics of T-A by causing enzymatic misincorporation of dTTP across from thymidine to give T-HgII-T in vitro44 and possibly in vivo44,45. NMR studies confirmed early predictions that HgII binds to T-T mismatches via N3 coordination of two deprotonated thymidine residues46. Structurally analogous C-AgI-C base pairs have also been reported47, and in both these cases, little or no impact on the global structure of the B-form duplex was reported47,48,49.

C-HgII-T is a newly discovered, all-natural metallo base pair for which relatively little information is available32,50,51. Using fluorescent nucleobase analogues and 1H NMR spectroscopy, we recently reported stoichiometric, high affinity binding of HgII to DNA duplexes containing C-T mismatches51. Conducted in parallel, crystal screening of various oligonucleotides and metal ions produced an X-ray structure of a short (8-mer), A-form DNA sequence containing two C-HgII-T base pairs with an unexpected metal binding mode involving the exocyclic amine (N4) of a deprotonated cytosine “C” residue and (N3) of thymine “T” (Fig. 1)50. This coordination mode was in contrast to a preliminary proposal for (N3)T-HgII-(N3)C coordination based on structural homology with T-HgII-T and small increases in thermal stabilities of duplexes containing C-T mismatches after adding HgII (Fig. 1)32. The global A-form structure observed in the crystal structure was inconsistent with circular dichroism (CD) data of slightly longer, 14–21-mer duplexes containing one or two C-HgII-T base pairs51. The CD spectra suggested B-form helices, and little-to-no changes in their global conformation upon adding HgII. The metal binding mode(s) and global structural characteristics of duplex DNA containing C-HgII-T base pairs in solution were therefore unclear.

Fig. 1
figure 1

Proposed C-HgII-T binding modes. a, (N3)T-HgII-(N3)C coordination based on structural homology with T-HgII-T32 and b, (N3)T-HgII-(N4)C metal binding mode found in a short, A-form duplex DNA50

Full size image

Here we report a detailed NMR study using 15N-labelled DNA and 199Hg enriched mercury salts to determine the solution structures and dynamics of C-HgII-T base pairs in duplex DNA. Unlike previous examples of metal-mediated base pairs, C-HgII-T exhibits two types of covalent connectivities that are dynamically coupled via a global conformational change in helical structure. A palindromic, 14-mer duplex with two C-HgII-T sites separated by six canonical base pairs (ODN1, Fig. 2a, Supplementary Table 1, and Supplementary Figs. 1, 2) exists as a 3:1 mixture of well-defined duplexes in dynamic equilibrium. Both structures exhibit groove and rise dimensions intermediate between ideal A- and B-form helices. The most abundant duplex contains (N3)T-HgII-(N3)C connectivity and mostly B-form helical characteristics, whereas the minor species contains (N3)T-HgII-(N4)C base pairs and more A-form characteristics. No indication of a third duplex containing one of each type of metallo base pair is evident, consistent with long-range conformational coupling between the two metal centres. Furthermore, the rate constants for nucleobase-metal-nucleobase isomerization (kforward = 3.5 s−1 and kreversed = 7.7 s−1) measured using [15N,1H]-HSQC experiments are nearly identical to those of the global conformational exchange of duplex structures measured using [1H,1H]-NOESY experiments (kforward = 4.3 ± 0.6 s−1, kreversed = 8.8 ± 0.9 s−1). These results therefore support the coupling of metal-ligand isomerization reactions over long distances (> 20 Å) via a global conformational change of the double helix. Taken together with the greater A-form characteristics upon metal binding, these results suggest that placing numerous C-T mismatches throughout a repetitive duplex sequence can facilitate a global B → A helical transition upon adding HgII. To test this possibility, we prepare and analyse a small library of hairpin duplex DNAs (n = 10) and identify a duplex sequence that adopts a global A-form structure upon adding stoichiometric HgII. This helical transition is rapid (< 30 s) and fully reversible upon addition of N-acetylcysteine in a cycle that can be repeated more than 10 times on the same DNA. In addition to its broad implications in structural biology and biochemistry, such demand control of B → A and A → B helical switching may be utilized in the future development of advanced DNA-based materials and devices.

Fig. 2
figure 2

Solution structure of metal-free duplex ODN1 “C-T”. a DNA sequence. Blue and green bases indicate C-T mismatches. b Overlay of 20 lowest energy structures. The structures were aligned to a representative model (#1) containing fewest outliers in the geometric quality criteria. c Facial recognition of non-planar, C-T pair in model #1 (PDB 6RLS)

Full size image

Results

Solution structure of HgII-free (apo) duplex ODN1 “C-T”

Our preliminary 1H NMR and CD studies utilized a 14-mer C2 symmetric, self-complementary sequence containing two C-T mismatches (ODN1 “C-T”, Fig. 2a)51. This sequence exhibits a thermodynamically predictable52, two-state transition between single strand and duplex (Supplementary Fig. 2), as well as a 1H NMR exchangeable imino region that is well resolved at pH = 7 (Supplementary Fig. 3). Here the full assignment of proton resonances was conducted by sequential walking along H1′ and aromatic protons (H1′n → H6/H8n+1 → H1′n+1) (Supplementary Fig. 14). As cross references for our assignments, H2′/aromatic, H2″/aromatic (H2′/H2″n → H6/H8n+1 → H2′/H2″n+1), and aromatic/aromatic (H6/H8n → H6/H8n+1 → H6/H8n+2) regions in the [1H,1H]-NOESY, [1H,1H]-TOCSY (H1′ → H2′/H2″) and [13C,1H]-HSQC spectra (aliphatic and aromatic regions) were used. All signals of the duplex were well resolved and the sequential walk could be followed through the entire sequence. Models for the duplex were constructed based on 958 conformationally restrictive nuclear Overhauser effect (NOE) distance restraints (Table 1). The models did not contain any artificial constraints of co-planarity or hydrogen bonding for the C-T mismatches. Superimposition of the 20 lowest energy structures from 200 computed structures gave an overall root mean square deviation (r.m.s.d) of all heavy atoms of 0.74 ± 0.26 Å, and 0.54 ± 0.21 Å for the C-T mismatch (Table 1, Fig. 2b). The central region forms a canonical B-form duplex, yet local perturbations about the non-coplanar C-T mismatches cause a 19 ± 3° bend in the helical axis at each mismatch. Given the high dynamics of this system and limitations of the modelling used, it is difficult to ascertain the exact pattern(s) of hydrogen bonding present in C-T mismatches. Consistent with an early NMR model, the mismatched pyrimidines are stacked inside the duplex in a co-facial orientation (Fig. 2c)53. Our structure suggests C-T mismatches containing only one or two very weak hydrogen bonds. Evidence for weak hydrogen bonding is observed in the C-T imino proton resonances at 10.9 ppm that are broader than the imino resonances of an analogous duplex ODN2 “T-T” containing T-T mismatches, and much broader than those of ODN3 “G-T” containing G-T wobble base pairs (Supplementary Fig. 3). These observations correlated very well with the thermal stabilities of these duplexes (Tm ODN1 “C-T” = 35 °C), (Tm ODN2 “T-T” = 38 °C), and (Tm ODN3 “G-T” = 47 °C)51, as well as other reported duplexes containing C-T, T-T, and G-T52,54,55. A weak interaction between C and T is further supported by the 10-fold faster HgII binding of C-T versus T-T mismatches51. Taken together with the axial bending in our structure, these results are consistent with the fact that C-T mismatches are among the most thermodynamically destabilizing mismatches known in duplex DNA52,54,55.

Table 1 NMR restraints and statistics.a
Full size table

Nucleobase-metal-nucleobase connectivity of C-HgII-T

Adding three equivalents of HgII (1.5 : 1.0 with respect to the number of mismatches) to duplex ODN1 “C-T” caused disappearance of the mismatched imino resonance at 10.9 ppm (Fig. 3a and Supplementary Fig. 3)51. Similar results were obtained for the duplex ODN2 “T-T” containing two T-T mismatches (Supplementary Fig. 3). These results reflect specific binding reactions, since the addition of HgII to an analogous duplex containing two G-T wobble base pairs (ODN3 “G-T”) caused no such deprotonation (Supplementary Fig. 3). [1H,1H]-NOESY cross peaks between NH of thymidine and guanine residues, and between thymidine NH and H2 of adenine residues enabled assignment of all imino proton signals of ODN1 “C-T” in both the presence and absence of HgII (Supplementary Figs. 4, 5). The imino proton resonances of the thymidine residues flanking the C-T mismatch exhibited the largest changes in chemical shifts upon HgII addition, giving a final spectrum similar to that of ODN2 “T-T” containing the widely studied T-HgII-T base pairs (Supplementary Fig. 3, Supplementary Table 2)41,42,46,48,49.

Fig. 3
figure 3

1H and 15N NMR spectra of duplex DNA upon formation of C-HgII-T base pairs. a Imino proton region of 1H NMR spectra of ODN1* “C*-T*” in the absence and presence of 1.5 equiv of HgII (where blue C* = 15N-labelled C, and green T* = 15N-labelled T). b 15N NMR spectra of ODN1* “C*-T*” before and after adding HgII. c 1J 1H,15N coupling of cytosine (N4)NH2major and of cytosine (N4)NHminor according to [15N,1H]-HSQC. For full spectrum see Supplementary Fig. 9b. d Assignment of N1major and N1minor of thymidine and cytosine by 3J- and 2J-coupling between N1 and H5 and/or H6 in the long-range [15N,1H]-HSQC spectrum. For full spectra see Supplementary Fig. 10a. e Major- and minor coordination modes of HgII bound to C-T mismatches. DNA samples contained 1.0 mM b or 0.5 mM a, c, and d of duplex DNA and 1.5 equiv of HgII (relative to the mismatches present) in aqueous buffer (200 mM NaClO4, 50 mM cacodylic acid in H2O / D2O (9:1) at pH = 7.8). Blue labels indicate 15N resonances of C* and green labels indicate 15N resonances of T*

Full size image

To characterize the structure(s) of C-HgII-T base pairs, we synthesized a 15N-labelled ODN1* “C*-T*” by synthetic incorporation of 15N-labelled C and T residues at positions 4 and 11 in an otherwise unlabelled duplex. The splitting of the 1H resonance at 10.9 ppm by 15N (88 Hz), and its disappearance upon adding HgII (Fig. 3a) further confirmed its assignment as the mismatched thymidine NH-resonance. The five 15N-resonances in the absence of mercury were assigned by proton-coupled and proton-decoupled 15N NMR spectra, 1J 1H,15N coupling of thymidine N3-H and cytosine NH2 by heteronuclear single quantum coherence (HSQC), and 3J 1H,15N- and 2J 1H,15N coupling between N1 and H5 and/or H6 by long-range [15N,1H]-HSQC (Supplementary Figs. 6, 7). After adding three equiv of HgII (1.5 : 1.0 with respect to the number of mismatches), two sets of 15N-resonances were observed, corresponding to a “major” and a “minor” species in a 3:1 ratio (Fig. 3b). When adding only two equiv of HgII, these same signals were observed, in addition to those of unbound DNA, confirming that the minor species was not a result of any excess of HgII (Supplementary Fig. 8). [15N,1H]-HSQC spectra allowed assignment of all 15N signals for both the major and minor complexes (Fig. 3c, d, Supplementary Figs. 9, 10. Disappearance of the N3-H cross peak in the 1J [15N,1H]-HSQC spectrum confirmed deprotonation of thymidine N3 upon addition of 1.5 equiv of HgII (Supplementary Figs. 7, 8). The downfield shift of thymidine N3major (Δppm = + 29) and N3minor (Δppm = + 28)46,56 suggested that direct (N3)T-HgII coordination was present in both binding modes (Fig. 3b, Supplementary Tables 3, 4). 3J 1H,15N- and 2J 1H,15N coupling of N1 to H5 and/or H6 observed in long-range [15N,1H]-HSQC spectra allowed for assignment of N1major and N1minor resonances, as well as H5 and H6 protons for cytosine and H6 of thymidine (Fig. 3d and Supplementary Fig. 10). The appearance of a single cross peak in the direct [15N,1H]-HSQC spectrum at 124/6.39 ppm indicated deprotonation of cytosine (N4)NHminor in the minor binding mode (Fig. 3c). The large downfield shift of cytosine (N4)NHminor (Δppm = + 28, Fig. 3b, Supplementary Tables 3, 4) and a doublet observed in the [15N,1H]-proton-coupled HSQC (J = 86 Hz, Supplementary Fig. 9c) further suggested displacement of one (N4)NH2 proton by HgII and direct HgII coordination to N4 in the minor species. An NOE-cross peak between (N4)NHminor to cytosine H5 (6.52/5.48 ppm at 25 °C) confirmed this 15N-resonance assignment (124 ppm) as being the deprotonated exocyclic amine of cytosine (Supplementary Fig. 11). 3J 1H,15N coupling of the 15N-signal at 207 ppm to (N4)NHminor, as well as to (N4)NH2major was observed in a band-selective, long-range [15N,1H]-HSQC (Supplementary Fig. 10b). This allowed assignment of the overlapping 15N-resonances of the N3major and N3minor of cytosine. It was still unclear, however, if HgII was directly coordinated to cytosine N3major.

To unambiguously determine metal-nucleobase connectivities57, we monitored changes in the 15N NMR spectrum of ODN1* “C*-T*” upon addition of 199Hg-isotopically enriched (79 %) Hg(ClO4)2 (Fig. 4a). In the major binding mode, the N3-resonances for both cytosine and thymidine appeared as doublets, thereby revealing their direct coordination to HgII (Fig. 4a). The large 1J 15N,199Hg coupling constant of 1095 Hz for thymidine HgII-N3major is consistent with T(N3)-HgII binding reported for a T-HgII-T dinucleoside complex measured in d6-DMSO57. Cytosine N3major-HgII exhibited a much smaller coupling constant 1J 15N,199Hg = 114 Hz (Fig. 4a), consistent with a longer, weaker bond. A doublet with a coupling constant 1J 15N,199Hg = 1052–1063 Hz confirmed direct HgII coordination to cytosine (N4)NHminor (Fig. 4a, b). This larger coupling constant is consistent with a stronger, shorter bond for HgII-C(N4) versus HgII-C(N3). Further support of our assignments was observed in 1H,199Hg through-bond coupling with 2J 1H,199Hg = 168 Hz in the [15N,1H]-HSQC (Fig. 4b), a value that was similar to reported 1H,199Hg coupling across nitrogen in unrelated systems measured in d6-DMSO or CDCl358,59,60. Splitting of the correlation between cytosine N3major and (N4)NH2major in the band-selective, long-range [15N,1H]-HSQC further confirmed our assignment of N3major of cytosine (Fig. 4c). These data provide 1J 15N,199Hg coupling constants in aqueous solutions, as well as cytosine N3,HgII and cytosine N4,HgII 1J 15N,199Hg coupling constants. Taken together, these results revealed that HgII binds to cytosine via two distinct coordination modes in solution. The major species contained two identical and unambiguous (N3)T-HgII-(N3)C base pairs, and the minor species two identical and unambiguous (N3)T-HgII-(N4)C base pairs. No evidence for a lower symmetry duplex containing one of each type of base pair was observed in any NMR experiment.

Fig. 4
figure 4

Nucleobase-metal-nucleobase connectivity of C-HgII-T base pairs. a 15N NMR of ODN1* “C*-T*” in the presence of 1.5 equiv of 199Hg-enriched Hg(ClO4)2 (79 % enriched see Supplementary Fig. 13). b 1J 15N,199Hg and 2J 1H,199Hg coupling according to [15N,1H]-HSQC. For full spectra see Supplementary Fig. 12b. c 1J 15N,199Hg coupling between cytosine N3major and 199Hg observed in the band-selective, long-range [15N,1H]-HSQC spectrum. For full spectrum see Supplementary Fig. 12c. d Summary of major- and minor coordination complexes in duplex DNA and the observed couplings. DNA samples contained 1 mM duplex DNA and 3 mM 199Hg-enriched Hg(ClO4)2. Blue labels indicate 15N resonances of C* and green labels indicate 15N resonances of T*

Full size image

Global structures of duplexes containing C-HgII-T

Consistent with 15N NMR spectra (Figs. 3, 4), 1H NMR experiments confirmed the presence of two C2 symmetrical species (Supplementary Fig. 14). Proton resonance assignments and modelling for each duplex were conducted exactly the same as for the metal-free duplex. Aside from the first and last residues of the duplexes (C1 and G14), all signals of the major and minor metallo duplexes were well resolved and the sequential walk proceeded through the entire sequence (Supplementary Figs. 14, 15). Models for major and minor duplex were constructed based on 646 and 640 NOE-derived conformationally restrictive distance restraints, respectively (Table 1). Superimposition of the 20 lowest energy structures of 200 computed structures gave an overall root mean square deviation (r.m.s.d) of all heavy atoms of 1.21 ± 0.44 Å for the major duplex and 1.09 ± 0.36 Å for the minor duplex (Fig. 5a, Table 1).

Fig. 5
figure 5

Solution structures of ODN1 containing two C-HgII-T base pairs and comparisons with the metal-free duplex. a Side-view overlay of the 20 lowest energy conformations for major- (grey) and minor (blue) metallo duplex structures. The 20 lowest energy structures were aligned to one representative model of each duplex containing zero outliers in the geometric quality criteria: for the major duplex, #1 of PDB 6FY6, and for the minor duplex #6 of 6FY7. HgII ions are depicted in gold. b Overlay comparison of these representative models of the major (grey) and minor (blue) duplexes. The duplexes were aligned by the end of each helix (C1, G2, T3, A26, C27, G28). The (N3)T-HgII-(N4)C connectivity in the minor duplex makes this metallo-base pair more solvent accessible than (N3)T-HgII-(N3)C, with a difference in solvent accessible surface area (ΔSASA) = 41 Å2 as calculated by PyMOL.85 c Overlay comparison of representative minor metallo duplex (blue, model #6) as compared to metal-free duplex (green, model #1, PDB 6RLS). The duplexes were aligned by the center of each helix (A7, T8, A21, T22)

Full size image

As compared to the major duplex, the minor duplex exhibits more axial bending and a deeper, narrower major groove (Fig. 5b). The differences between the metal-bound and metal-free structures are also the greatest for the minor duplex (Fig. 5c). Inspection of the 20 lowest energy models for all three structures (apo, major metallo, and minor metallo) revealed a high frequency (65–100 %) of an unusual, O4′-endo sugar pucker at the cytosine residue of the C-T and C-HgII-T base pairs (Supplementary Table 6). Direct support for this was observed in the 3J H1′,H2′ coupling constant for the HgII-coordinated cytosine residue (3J H1′,H2′ = 6.5 Hz) in the major structure (Supplementary Table 5, Supplementary Fig. 17). Using the Karplus equation (Supplementary Equation 1)61 a dihedral bond-angle of Φ1ʹ2ʹ = 142° was calculated, which is in excellent agreement with dihedral angles observed in the solution structure models (Φ1ʹ2ʹ = 131°) (Supplementary Fig. 18, Supplementary Equation 1)61,62. However, the apparent O4′-endo sugar puckers present in all three structures likely reflect the averaged conformations of rapidly interconverting C2′- and C3′-endo sugar puckers that occur much faster than the time scale of these NMR measurements63,64,65,66,67. Support for this conclusion can be found in the analyses of global parameters, where the sugar pucker amplitudes throughout all three duplexes were much smaller than both A-form and B-form duplexes (Fig. 6a). In contrast, nearly all other structural parameters including groove dimensions, twist and rise gave values intermediate between A- and B-form duplexes (Fig. 6b–f, and Supplementary Figs. 19–22)68.

Fig. 6
figure 6

Selected base pair and base pair step parameters of apo-, HgII-bound major-, and HgII-bound minor duplex structures. Parameters of apo (green diamonds, PDB 6RLS), HgII-bound major (black circles, PDB 6FY6) and HgII-bound minor (grey squares, PDB 6FY7) duplex structures were calculated using Curves+.86 a Sugar pucker amplitude. b Major groove width. c Minor groove width. d Major groove depth. e Helical rise. f Inclination. The reference values for standard A- (red) and B-form (blue) duplex DNA were taken from (ref 86), which analyzed crystal structures of A-DNA (PDB 1d1388) and B-DNA (PDB 1bna66) dodecamers. Reported values represent the mean and standard deviation of the 20 lowest energy conformations for each structure. For additional analyses of base pair parameters see Supplementary Figs. 19 –22. Source data are provided as a Source Data file

Full size image

Dynamics of major—minor metallo duplex interconversion

The presence of two individual sets of NMR signals, together with exchange cross peaks and exchange-mediated cross peaks between them (Supplementary Figs. 23–26 and Supplementary Figs. 36–38) indicated the presence of conformational changes with rates suitable for determination by standard NMR methods. To investigate the dynamic changes in local metal ion coordination, z-z exchange [15N,1H]-HSQC spectra were measured69. With increasing delay time (tm), new exchange signals appeared (Fig. 7a, Supplementary Fig. 23). Global fitting of integrated peak volumes versus exchange delay times (Supplementary Equations 6–9) furnished rate constants of k1 = 3.5 s−1 and k−1 = 7.7 s−1 for the forward and reverse reactions of nucleobase-metal-nucleobase isomerization, respectively (Fig. 7a).

Fig. 7
figure 7

Dynamic isomerization of duplex DNA containing two C-HgII-T base pairs. a Dynamic changes in local metal ion coordination. Rate constants of interconversion were determined by plotting changes in cross peaks (NH2 and NH) and exchange cross peaks (NH2' and NH') observed in z-z exchange [15N,1H]-HSQC spectrum as a function of different delay times. Rate constants of interconversion were determined by global fitting of integrated peak volumes versus exchange delay times using Supplementary Equations 6-9, where R1, k1, and k-1 were constrained as being equal for all four curves with the assumption that all proton- and nitrogen atoms have the same relaxation rate (R1) and that the NH proton is converted equally to both NH2 protons during the exchange. b Representative example (G9H1') of duplex interconversion rate constants as determined by monitoring changes in exchange cross peaks and exchange-mediated NOE cross peaks in [1H,1H]-NOESY spectra with various mixing times (tm). At longer mixing times, a decrease in signal intensity was observed due to auto relaxation (R1). The sum of interconversion rate constants (k1 + k-1) was determined by fitting changes in cross peak intensity vs mixing time (tm) (Supplementary Equation 12). The individual rate constants of interconversion (k1 and k-1) were determined by fitting (Supplementary Equation 14) cross peak intensities normalized by the intensities of diagonal peaks versus mixing times. See Supplementary Figs. 37, 38 for [1H,1H]-NOESY spectra, and Supplementary Figs. 39–42 for the same types of analyses at other positions. c Summary of major- and minor duplex structures, structures of the metallo base pairs they contain, and the rate constants of their exchange. Rate constants values are given as mean ± standard deviation from 5 exchange cross peaks and exchange-mediated NOE cross peaks. The HgII-ligand complexes present in the major species are net positive (+1), whereas those of the minor species are neutral. The combined contributions of more favourable electrostatic interactions and near-canonical B-form helical structure present in the major species likely outweigh the stronger ligand–metal–ligand binding interactions present in the minor duplex

Full size image

Single exchange signals in ROESY spectra were observed for the HgII-bound nucleosides as well as for sugar- and aromatic proton signals of various residues throughout the duplex (Supplementary Fig. 24). The same exchange signals were found in the [1H,1H]-TOCSY spectrum in the H1′/H5 region (~5.0–6.2 ppm, Supplementary Fig. 25). Within that region, no protons belonged to the same spin system. The observed cross peaks therefore occurred by exchange rather than through-bond coupling. Exchange-mediated cross peaks were observed in the aromatic → H2′/H2″ region of the [1H,1H]-NOESY spectrum (Supplementary Fig. 26), confirming the global nature of the conformational change. To evaluate the potential impact of variable ionic strength on the exchange rates, samples of the metallo duplex were prepared in the presence of 50 mM, 200 mM and 500 mM NaClO4. The overall exchange rates of interconversion between the two structures decreased with increasing ionic strength, but the ratio of the two structures remained the same (Supplementary Figs. 27–31). Likewise, pH-dependent measurements revealed slower overall exchange rates with increasing pH from pH 6–9 while maintaining the same ratio (Supplementary Figs. 32–35). The lower rates of exchange correlate with the increasing persistence length and therefore rigidity of the duplex with increasing pH and ionic strength70,71.

To determine rate constants of global interconversion of the two duplexes, we measured [1H,1H]-NOESY spectra with various mixing times (tm) using samples prepared in the presence of 200 mM NaClO4 at pH = 7.8. Selected exchange cross peaks (‘Aa’ and ‘aA’, Supplementary Fig. 37) and exchange-mediated NOE cross peaks (‘Ab’ and ‘aB’, Supplementary Fig. 38) at residues throughout the duplex were integrated, normalized to signal intensity at mixing time = 200 ms, and plotted as a function of mixing time (tm) (Supplementary Table 7, Supplementary Figs. 39–42). For definitions of peak labels, see Supplementary Fig. 36. Exchange cross peaks were fit to Supplementary Equation 12 to determine the sum of rate constants k1 + k−1 (Fig. 7b, Supplementary Fig. 39 and Supplementary Table 7). For signals having sufficient resolution of the diagonal peaks (‘AA’ and ‘aa’) for integration purposes, exchange cross peaks were normalized by aA (tm) / (AA (tm) + aA (tm)) and fit to Supplementary Equation 14 to determine individual rate constants k1 and k−1 (Fig. 7b and Supplementary Fig. 41). Dividing exchange-mediated cross peaks (‘Ab’ and ‘aB’) by the sum with their corresponding NOE cross peak (‘AB’ and ‘ab’) allowed the determination of rate constants k1 and k−1 by fitting to Supplementary Equations 18 and 19 (Supplementary Fig. 42). The sum of rate constants k1  + k−1 determined for the global conformational change according to exchange cross peaks (10.5–15.0 s−1), as well as the individual rate constants k1 (3.8–5.2 s−1) and k−1 (7.5–9.7 s−1) independently determined from exchange cross peaks and exchange-mediated cross peaks, were all in excellent agreement with rate constants determined for changes in metallo base pair structure (k1 = 3.5 s−1, k−1 = 7.7 s−1, Fig. 7a, Supplementary Table 7). Together with the absence of a third, lower symmetry duplex containing one of each type of metallo-base pair, these results revealed that dynamic changes in local metal-ligand isomerization were coupled to the global interconversion of the two duplex structures (Fig. 7c).

HgII-induced, conformational switching from B- to A-form

The long-range coupling (>20 Å) between the metal centres as well as the increased A-form characteristics in the centre region of the duplex upon metal binding, suggested that placing numerous C-T mismatches throughout a repetitive duplex sequence could facilitate a global B → A helical transition upon adding HgII. To test this possibility, we introduced C-T mismatches into (G4C4)n type DNA sequences that are known to exhibit partial A-form characteristics50,72,73,74. To supress formation of intramolecular G-quadruplex structures that would otherwise interfere with intermolecular duplex formation of such repetitive sequences, we designed a small library of 120-mer DNA hairpins (Supplementary Table 8) containing the tetraloop sequence cGCTAg that is known to stabilize both RNA and DNA hairpins75. To fold the hairpins, dilute solutions of DNA (1 μM) were heated (95 °C, 5 min) and rapidly cooled on ice at 0 °C. Samples were then incubated with 0.0 or 1.5 equiv of HgII (relative to number of C-T mismatches present) at 25 °C for 3 h prior to their analysis. Gel electrophoresis revealed clean, intramolecular hairpin formation for most sequences in both the presence and absence of HgII, including our hit ODN13 (Supplementary Fig. 43). To screen for the induction of A-form structures by HgII, we used a fluorescent aminoglycoside binding assay74. Aminoglycoside antibiotics exhibit a general selectivity for binding A-form over B-form helices6. Changes in the fluorescence anisotropy of a 40 nM solution of a Neomycin-BODIPY conjugate “Neo-BODIPY”76 were therefore measured in the presence and absence of each DNA (600 nM) pre-treated with HgII (0.0 or 1.5 equiv per C-T mismatch). No changes in anisotropy were observed upon addition of all hairpins in the absence of HgII, however, the pre-incubation of ODN13 with HgII caused a 3.5-fold increase in fluorescence anisotropy of Neo-BODIPY (Table 2). Titration of the ODN13-HgII complex into solutions of Neo-BODIPY revealed an apparent dissociation constant (Kd) = 1.4 ± 0.7 µM (Fig. 8a). This value is similar to the values reported for binding of neomycin to A-form, duplex RNA77. The ternary complex formed between Neo-BODIPY and ODN13-HgII was disrupted by the addition of unlabelled neomycin B (Supplementary Fig. 44), as well as N-acetylcysteine that sequesters HgII (Supplementary Fig. 45). These results demonstrate the reversibility of Neo-BODIPY binding, a lack of significant impact by the BODIPY tag, as well as the switch-like (on/off) effect of HgII binding to ODN13. The analogous hairpin containing T-A base pairs (ODN14) in place of T-C mismatches exhibited no such behaviour (Table 2). These observations are confirmed using CD spectroscopy.

Table 2 DNA hairpin repeat sequencesa and fluorescence anisotropy of Neo-BODIPY in the presence of each DNA with and without HgII.b
Full size table
Fig. 8
figure 8

Fluorescence anisotropy and CD spectra reveal global B-form to A-form helical transition upon addition of HgII. a Changes in fluorescence anisotropy (λex = 480 nm, λem = 515 nm) of Neo-BODIPY (40 nM) upon addition of hairpin ODN13 prepared in the presence (light blue) and absence (dark blue) of HgII. The values represent mean and standard deviation of three independent measurements. C-HgII-T base pairs were formed by pre-incubation of the DNA with 1.5 equiv HgII for 3 h prior to addition of Neo-BODIPY. b CD spectra of hairpin ODN4 (1 μM), or, c CD spectra of hairpin ODN13 (1 μM) in the presence and absence of 1.5 equiv of HgII. All samples were prepared in an aqueous buffer containing 200 mM NaClO4 and 50 mM cacodylic acid (pH = 7.8). The number of HgII equiv are relative to the number of mismatches present. Source data are provided as a Source Data file

Full size image

The (G3C3)n-containing hairpins ODN4 and ODN5 exhibited CD spectra consistent with previous publications72, having a double maximum at 260 nm and 280 nm (Fig. 8b and Supplementary Fig. 46) that are thought to reflect a mixture of A-form and B-form-like stacking of the guanine and cytosine nucleobases, respectively. The addition of HgII had little-to-no impact on the CD spectra of hairpins ODN4–12 or ODN14 (Fig. 8b and Supplementary Fig. 46). However, addition of HgII to hairpin ODN13 caused changes in its CD spectrum indicative of a global B- to A-form helical transition (Fig. 8c and Supplementary Fig. 47a)78. This HgII-induced conformational change exhibited a 1:1 stoichiometry between HgII and the number of C-T mismatches present (Supplementary Fig. 47b), was extremely rapid (<30 s to complete) and fully reversible upon addition of N-acetylcysteine (Supplementary Fig. 47c). Given the relatively low G-quadruplex propensity of sequence ODN13, we were successful in preparing the corresponding intermolecular duplex lacking a hairpin turn. This simple duplex “ODN13 ds” also exhibited a reversible, HgII-inducted switching between global B- and A-form helices (Supplementary Fig. 48). By alternating between the addition of HgII and N-acetylcysteine, the helical switching cycle from B- to A-form, and A-form to B-form could be repeated more than 10 times on the same DNA (Supplementary Fig. 49).

Discussion

Here we report solution structures of C-HgII-T base pairs. In addition to providing fundamental insights into C-HgII-T binding and dynamics, these results provide 1J 15N,199Hg and 2J 1H,199Hg coupling constants. Prior to these studies, little or no such information was available for these coupling constants in water46,47,58,79. Large differences in 1J 15N,199Hg coupling constants for complexes containing (N3)C-HgII (114 Hz) versus (N4)C-HgII (1052 Hz) will provide a basis for future computational studies that address the relationships between coupling constants, ligand ionization and metal binding57.

Aside from the gain/loss of a proton from (N4)C, the major/minor duplexes of ODN1 bound to HgII are constitutional isomers with respect to each other. Isomerization-coupled conformational exchange between these structures was relatively slow on the chemical-shift NMR time scale, yet it was fast enough to allow direct characterization by monitoring changes in exchange cross peaks as a function of mixing time (tm). The rate constants of nucleobase-metal-nucleobase isomerisation (k1 = 3.5 s−1 and k−1 = 7.7 s−1) were nearly identical to rate constants independently measured for the global conformational exchange of the two duplex structures (k1 = 4.3 ± 0.6 s−1, k−1 = 8.8 ± 0.9 s−1). No evidence for a third duplex containing one of each type of metal-base pair was observed, giving further support for coupling between the metal centres via global conformational exchange. The forward and reverse rate constants of this process were ~103-fold faster than dissociation of HgII from C-HgII-T base pairs51. The HgII ions therefore remained bound to the DNA during multiple structural exchanges. Given the high quality of the structures and dynamics reported here, this system provides a highly attractive model for the development of molecular dynamics simulations aimed at elucidating the pathways of dynamic conformation exchange processes80.

Consistent with our structure of the minor duplex (Figs. 4, 5), crystals of a short, A-form DNA sequence also contained two C-HgII-T base pairs having (N3)T-HgII-(N4)C connectivity50. However, the global A-form structure observed in the crystal structure was inconsistent with our circular dichroism (CD) data collected in solution using five different, slightly longer (14 to 21-mer) duplexes containing one or two C-HgII-T base pairs51. Together with the observation that (N3)T-HgII-(N3)C binding present in our major structure is associated with more B-form character, and the (N3)T-HgII-(N4)C present in the minor structure is associated with more A-form character, a general picture is suggested. (N3)T-HgII-(N3)C binding and B-form structure is likely dominant in most DNA sequences, in direct analogy with high resolution structures of T-HgII-T base pairs49. However, the 10-fold larger 1J coupling constants (15N,199Hg) observed for (N4)C-HgII (1052 Hz) connectivity in the minor duplex as compared to the (N3)C-HgII (114 Hz) in the major structure indicate that stronger nucleobase-metal-nucleobase binding interactions are present in the minor structure. The minor duplex, however, also exhibits a greater degree of structural perturbation away from a canonical B-form duplex that likely destabilizes the complex as compared to the major form. The lower bonding energy of (N3)T-HgII-(N3)C versus (N3)T-HgII-(N4)C is therefore compensated by the overall higher stability of B-form versus A-form in the major and minor structures, respectively. However, in situations that favour A-form (crystallization, dehydration, G/C-rich DNA sequences, etc.) the stronger (N3)T-HgII-(N4)C coordination mode can dominate. The exact causal relationship between global helical conformation and metallo-base pair connectivity is currently unclear. An X-ray study involving a short A-form DNA containing two (N3)T-HgII-(N4)C base pairs suggested the combined effects of base pair geometry, neighbouring base effects, and a bridging water molecule in the minor groove were invovled50.

For both of our metal-bound structures in solution, the amount of A-form character was greatest in the centre of each helix, suggesting a medium-range influence of C-HgII-T formation. In contrast, the opposite pattern was observed in the structure of the metal-free duplex, having the greatest B-form characteristics in the centre of the helix. Together with the long-range coupling (>20 Å) between the two metallo-base pairs via a global conformational exchange, these results suggested that placing numerous C-T mismatches throughout a repetitive duplex sequence could support a global B-form → A-from helical transition upon addition of stoichiometric HgII. Indeed, we were able to identify one such sequence exhibiting a fully-reversible switching cycle from B- to A-form, and A-form to B-form by tandem additions of HgII and N-acetylcysteine (Supplementary Fig. 49). Both transitions were complete in <30 s, and could be repeated more than 10 times. While numerous examples of local A-form perturbations caused by DNA-protein and DNA-small molecule binding interactions have previously been reported3,4,5,6,7,8,9, the previous examples of global B-form → A-from helical transitions involved global dehydration of the duplex1,2,78. Here the global B- to A-form helical transition was a result of discrete, reversible metal binding. In addition to its broad implications in structural biology and biochemistry, this type of A/B-form helical switching can be potentially be utilized in the development of advanced DNA-based materials and devices24,25,26,27,28,29,30,31.

Methods

Sample preparation

For Fig. 2, duplex DNA (0.3 mM) was prepared by dissolving 0.6 mM of the self-complementary sequence in an aqueous solution of NaClO4 (50 mM, 90:10 H2O/D2O) and the pH was adjusted to pH = 7.8 by addition of an aqueous solution of NaOH. The sample was annealed by heating to 95 °C for 5 min and slow cooling to room temperature over 4 h. For Figs. 3, 4, 7, duplex DNA (0.5–1.0 mM) was prepared by dissolving 1.0 mM or 2.0 mM of the self-complementary sequence in aqueous buffer (200 mM NaClO4, 50 mM cacodylic acid in H2O / D2O (9:1) at pH = 7.0), heating to 95 °C for 5 min, and slowly cooling to room temperature over 4 h. Hg(ClO4)2 was added, and the pH adjusted to pH ≈ 7.8 by addition of an aqueous solution of NaOH. For Figs. 5, 6, duplex DNA (0.4 mM) was prepared by dissolving 0.8 mM of the self-complementary sequence in an aqueous solution of NaClO4 (50 mM, 90:10 H2O/D2O) and the pH was adjusted to pH = 7.8 by addition of an aqueous solution of NaOH. The samples were annealed as described above, mixed with HgII (1.5 equiv with respect to mismatch), the pH re-adjusted to pH ≈ 7.8, and the sample was treated with Chelex-100 (BIO-RAD) for 10 min to remove excess HgII43. Samples measured in D2O were prepared the same way and then lyophilized, dissolved in 99.9% D2O and the pD was adjusted to ~7.4 by addition of a solution of NaOD in 99.9% D2O. Samples measured at 4 °C were equilibrated at 4 °C for 15 min prior to measuring.

NMR spectra measurements

1H NMR spectra were recorded on a Bruker Avance II 500 MHz spectrometer equipped with a TXI z-axis gradient probe head using excitation sculpting for water suppression. Proton chemical shifts were referenced to the water line at 4.70 ppm. The spectra were processed with a line broadening factor of 10 Hz. 1D 15N-NMR spectra were recorded on a Bruker Avance II 500 MHz spectrometer equipped with a BBO z-axis gradient CryoProbe at 4 °C or 25 °C using either inverse gated or no proton decoupling. The spectra were processed with a line-broadening factor of 10 Hz. [15N,1H]-HSQC spectra were recorded at 4 °C or 25 °C on a Bruker Avance II 500 MHz spectrometer equipped with a BBO z-axis gradient CryoProbe, Bruker Avance 600 MHz spectrometer equipped with TCI z-axis gradient CryoProbe or on a Bruker Avance 700 MHz spectrometer equipped with TXI z-axis gradient CryoProbe. The INEPT times were set to select for a 90 Hz coupling for 1J [15N,1H]-HSQC’s, 20 Hz coupling for long-range [15N,1H]-HSQC spectra, and to 25 Hz for band-selective long-range [15N,1H]-HSQCs. Water flip-back pulses together with the WATERGATE method were used for water suppression. Proton chemical shifts were referenced to the water line at 4.70 ppm at 4 °C and 15N chemical shifts were indirectly referenced against 1H using Ξ = 0.10132911881. For solution structure determination, non-exchangeable resonances were assigned from [1H,1H]-NOESY spectra measured in D2O (4 °C and 25 °C and mixing times of 60 ms and 250 ms). Exchangeable protons were assigned from [1H,1H]-NOESY spectra measured in H2O/D2O (90:10) (4 °C, 150 ms mixing time). Spectra were recorded on a Bruker Avance 600 MHz spectrometer equipped with a TCI z-axis gradient CryoProbe or on a Bruker Avance 700 MHz spectrometer equipped with TXI z-axis gradient CryoProbe. For additional information see supporting information.

NMR solution structure calculations

The integrated peak volumes from a representative [1H,1H]-NOESY spectrum (mixing time = 250 ms) measured at 25 °C were calibrated to distances using CALIBA macro in DYANA82. The NOE signals were grouped into four categories: (i) strong (1.8–3.0 Å), (ii) medium (1.8–4.5 Å), (iii) weak (3.0–6.0 Å), and very weak (4.0–7.0 Å). Structure calculations were performed with XPLOR-NIH 2.46 using standard implemented force field parameters83. For introduced restraints for the calculations see supporting information. Starting from a strand generated based on the sequence of nucleoside residues, 2000 structures were calculated based on NOE-, dihedral-, planarity-, and H-bond distance restraints using simulated annealing. The 20 lowest energy structures were selected and used for further refinement using additional RAMA and ORIE database terms. 200 refined structures were calculated and the 20 lowest energy structures were visualized and analyzed. Root mean square deviation (r.m.s.d.) were calculated using MOLMOL84 and duplexes were visualized using PyMOL85. Base-pair parameters were determined using Curves +86 or 3DNA87.