Remarkable Metal Counterion Effect on the Internucleotide J

Jul 4, 2008 - The effects of metal ion binding on the 2hJNN-coupling and δ(1H)/Δδ(15N) chemical shifts of N−H···N H-bond units in internucleot...
0 downloads 0 Views 1MB Size
9174

J. Phys. Chem. B 2008, 112, 9174–9181

Remarkable Metal Counterion Effect on the Internucleotide J-Couplings and Chemical Shifts of the N-H · · · N Hydrogen Bonds in the W-C Base Pairs Huifang Li,† Robert I. Cukier,‡ and Yuxiang Bu*,†,‡ The Center for Modeling & Simulation Chemistry, Institute of Theoretical Chemistry, Shandong UniVersity, Jinan 250100, People’s Republic of China, and Department of Chemistry, Michigan State UniVersity, East Lansing, Michigan 48824 ReceiVed: April 8, 2008; ReVised Manuscript ReceiVed: May 12, 2008

The effects of metal ion binding on the 2hJNN-coupling and δ(1H)/∆δ(15N) chemical shifts of N-H · · · N H-bond units in internucleotide base pairs were explored by a combination of density functional theory calculations and molecular dynamics (MD) simulations. Results indicate that the NMR parameters vary considerably upon cation binding to the natural GC or AT base pairs, and thus can be used to identify the status of the base pairs, if cation-perturbed. The basic trend is that cation perturbation causes 2hJNN to increase, ∆δ(15N) to decrease, and δ(1H) to shift upfield for GC, and in the opposite directions for AT. The magnitudes of variation are closely related to the Lewis acidity of the metal ions. For both base pair series (Mz+GC and Mz+AT), these NMR parameters are linearly correlated among themselves. Their values depend strongly on the energy gaps (∆ELPfσ*) and the second-order interaction energies (E(2)) between the donor N lone pair (LPN) and the acceptor σ*N-H localized NBO orbitals. In addition, the 2hJNN changes are also sensitive to the amount of σ charge transfer from LPN to σ*N-H NBOs or from the purine to the pyrimidine moieties. The different trends are a consequence of the different H-bond patterns combined with the polarization effect of the metal ions in the cationized Mz+AT series, Mz+ r A f T, and the cationized GC series, Mz+ r G r C. The predicted cation-induced systematic trends of 2hJNN and δ(15N,1H) in N-H · · · N H-bond units may provide a new approach to the determination of H-bond structure and strength in Watson-Crick base pairs, and provide an alternative probe of the heterogeneity of DNA sequences. Introduction Hydrogen bonds between imino proton donors and acceptor nitrogens (N-H · · · N) in Watson-Crick (W-C) base pairs of nucleic acids have a partially covalent character that gives rise to the existence of an internucleotide 2hJNN scalar coupling across the H-bond (Figure 1).1–3 Recently, introduction of heteronuclear 1H/15N/15N correlation spectroscopy (HNN-COSY)4–6 and transverse relaxation optimized spectroscopy (TROSY)7 allows for a direct identification of internucleotide N-H · · · N H-bonds in nucleic acid base pairs, as monitored by internucleotide 2hJNN couplings.1–8 These rapid and sensitive pulse schemes greatly facilitate NMR investigations of nucleic acid structures by revealing the presence, localization, and even the lengths of H-bonds involved in base pairs.6 Successful application of these correlation spectroscopy methods has been demonstrated in the natural canonical GC and AT/AU,4,7 and the noncanonical GA and reverse Hoogsteen AU RNA base pairs.3,6,8 Because coupling constants are very sensitive to the electronic features of H-bonds,9 it can be proposed that internucleotide 2hJNN scalar couplings must be influenced by the counterion environment. For example, an enlarged 2hJNN coupling (about 1 Hz) and a δ(1H) upfield shift (about 2-3 ppm) for the N-H · · · N H-bond were observed in the C+ (protonated cytidine) cationized GC pair, compared with those in the natural canonical GC base pair.4–6 Subsequent calculations for the scalar J-couplings and the magnetic shieldings involved in the H-bonds of the C+ · G-C * To whom the correspondence should be addressed. E-mail: byx@ sdu.edu.cn. † Shandong University. ‡ Michigan State University.

Figure 1. H-bond patterns in two natural canonical W-C base pairs: Guanosine-Cytidine (GC) and Adenosine-Thymine (AT) and their metal ion (Mz+)-bound derivatives (Mz+GC and Mz+AT). The digits 1 or 2 in parentheses beside Mz+ denote the minor or major groove binding modes, and the corresponding cation-bound base pairs are denoted by Mz+GC(1), Mz+GC(2), Mz+AT(1), and Mz+AT(2), respectively. The N-H · · · N H-bond units mainly considered here are highlighted by the solid-line boxes, while other H-bonds (N-H · · · O) are dot-line boxed. In addition, for the GC series, the digits 1 or 2 in parentheses beside the O · · · H-N H-bond units denote their numbering.

triplet also confirmed these NMR parameter changes, and suggest that inter-residue coupling constants can be used as excellent indicators of H-bond interactions.9 Environmental factors, including the presence of metal cations, affect the H-bond structure and strength of the W-C base pairs.10 As have been revealed by recent data obtained by both experimental and theoretical methods,11–25 metal ions preferentially bind to the nucleic bases in the major or minor grooves, disrupting H-bonds in the base pairs, and further affecting their stability and properties. While the influence of metal ions on the structural stability and biochemical properties of DNA has been extensively investigated, little is known about the effect of Mz+-DNA

10.1021/jp8030545 CCC: $40.75  2008 American Chemical Society Published on Web 07/04/2008

N-H · · · N Hydrogen Bonds in the W-C Base Pairs interactions (where Mz+ denotes a bioactive metal ion) on the NMR parameters, 2hJNN and δ(1H,15N), and their possible use for elucidating structural and functional aspects of DNA. Thus, it is of interest to explore if there are systematic correlations between the nature of the cation binding to DNA and the cationinduced J/δ changes. It may be that metal ion coupling to the natural canonical base pairs provides versatility in DNA functions, important to transcription, replication, and even impact DNA damage/lesion processes. The sensitivity of J-couplings to H-bond structure and environmental effects implies that their accurate determination is inherently difficult due to the structural fluctuations induced by base-pair dynamics. Sattler et al.26 found that agreement of calculated and experimentally determined J-couplings required, in the calculations, an average over the geometric configurations obtained from the naturally occurring statistical ensemble. They demonstrated that the J-coupling values determined, via DFT calculations at the UB3LYP/6-311G** level, from the MDaveraged H-bond geometries (the interresidue 3hJNC′ scalar couplings in R-helices and β-sheets in proteins) are in excellent agreement with experimental data.26,27 In the present work, we first combine density functional theory (DFT) calculations and molecular dynamics (MD) simulations to explore the influence of Na+ on the 2hJNN and δ(1H,15N) NMR parameters in the natural canonical W-C base pairs, and show that the NMR parameters are correlated with the H-bond geometry and metal binding. In addition, we find that the NMR parameters vary considerably upon binding cations of different Lewis acidity to the natural GC or AT base pairs, and this observation warrants a more extensive investigation of NMR parameters for these cations using only the optimized geometries. The trends in the NMR parameters upon binding different cations to GC and AT can be rationalized with use of a natural bond orbital analysis of the amount of charge transfer in orbitals contributing to the N-H · · · N H-bond unit. The predicted cation-induced systematic trends of 2hJNN and δ(1H,15N) in N-H · · · N H-bond units may provide a new approach to the determination of H-bond structure and strength in Watson-Crick base pairs, and lead to a new probe of the heterogeneity of DNA sequences. Results and Discussions 1. Considerable Variations of J/δ Parameters Subject to the Cation Bindings. The effects of metal ion binding on the 2hJNN-coupling and δ(1H)/∆δ(15N) chemical shifts of N-H · · · N H-bond units in internucleotide base pairs were explored by a combination of density functional theory calculations and molecular dynamics (MD) simulations. Two schemes were employed to determine the relevant NMR parameters: (i) with snapshot geometries in the presence and absence of a counterion (Na+), extracted from the MD trajectories on two decamer duplexes, d(AGGAAACCAA) and d(AGGAGACCAA), and (ii) with optimized geometries for a set of isolated metal ion-bound base pairs. The calculational details are given in the Supporting Information. A. ObserWations from MD Simulated Decamer Duplexes. The impact of Na+ on the internucleotide 2hJNN/1hJNH scalar coupling constants and δ(1H)/δ(15N) chemical shifts in the N-H · · · N H-bond units of the W-C base pairs was evaluated for a set of geometries obtained along the MD trajectory. Snapshots saved every 0.5 ps were sorted into two groups depending on the presence or absence of counterions at particular locations around the decamer DNA duplexes. Group 1 consisted of bound states, with a counterion binding in the major groove associated with a given W-C base pair (Na+ · · · N7 distance

J. Phys. Chem. B, Vol. 112, No. 30, 2008 9175

Figure 2. Averaged DNA structures obtained from MD trajectory averaged over a 10-ps MD run when a Na+ is bound with a base or not. MD-AT and MD-Na+AT(2) were obtained from MD simulations on the d(AGGAAACCAA) decamer while the other two duplexes were obtained from MD simulations on d(AGGAGACCAA). The given W-C base pairs associated with a Na+ counterion are highlighted by a ball and stick presentation in duplexes, and also given in the bottom view. The Na+ counterion penetrating into the major groove with direct interaction with W-C pairs is represented as a big purple ball.

∼2-3 Å), and group 2 consisted of unbound states with the counterion having departed from the binding site (Na+ · · · N7 distance ∼9-10 Å).28 The calculated results are summarized in Table S1 in the Supporting Information. The averaged structures shown in Figure 2 show how the bases can be responsible for forming a metal ion-binding site without apparent backbone involvement. The bare Na+ ion penetrates the major groove of the two designed DNA duplexes, adjacent to the N7 site in the A or G bases. Owing to the sequence-specific binding character of the metal counterions, or a lack of sufficient simulation time, we did not find penetration of Na+ ions into the minor groove of the designed DNA. For all snapshot structures of the base pairs saved from the MD simulations including the natural and the counterion-bound W-C (AT and GC) base pairs (Figure 2), the calculated 2hJNN couplings over the MD trajectory range from 3.5 to 7 Hz. They are larger than the corresponding 1hJNH couplings between the donor 1H nuclei and the acceptor 15N nuclei, which range from 1.5 to 3.5 Hz (Table S1, Supporting Information). There is a counterintuitive character of these two quantities. Namely, the one-bond 1hJNH in each base pair is smaller than the corresponding two-bond 2hJNN, revealing that all base pairs considered here exhibit a trans-H-bond coupling effect, as has been observed for the natural canonical base pairs in HNN-COSY experiments.4,5 A comparison among these calculated NMR parameters reveals a significant effect of the Na+ ions on the 2hJNN, 1hJNH, δ(15N), and δ(1H) parameters associated with the N-H · · · N H-bonds of all the W-C base pairs. Our calculations exhibit a good correlation between calculated 2hJNN/δ(1H) parameters over the MD trajectories and the corresponding H-bond distances, as displayed in Figure 3 as well as Figure S1 in the Supporting Information (others are given in Table S1, Supporting Information) in agreement with a recent prediction that 2hJNN and δ(1H) parameters exhibit monotonic changes with the N-H · · · N H-bond distances in nucleic acids, obtained by DFT calculations of NMR parameters in model systems.5,9 It should be noted that although the correlations between 2hJNN and H-bond distance shown in Figure 3 are not as strong as those predicted by Barfield10 for modeled systems extracted from a C+GC triplet,

9176 J. Phys. Chem. B, Vol. 112, No. 30, 2008

Figure 3. The dependences of the 2hJNN couplings on the N · · · N distances in the corresponding N-H · · · N H-bonds of GC/Na+GC(2) and AT/Na+AT(2) base pairs in their snapshot geometries extracted from the relevant MD simulations, determined from the DFT/MD calculations.

the overall tendency is consistent between them.29 In contrast to results from the C+GC case,9 in this work the dependences of the NMR parameters on the N · · · N distances were extracted from simulations for both Na+GC and Na+AT, and may be applied to interpreting the natural trends of these parameters associated with the N-H · · · N H-bond units in all base pairs as their surroundings change. The deviations of our data (Figure 3) from the fitted correlation lines should be attributed to the fact that the geometries used for the NMR parameter calculations are directly extracted from the simulation snapshots, instead of from the optimized ones. Indeed, the N-H · · · N angles and the dihedral angles between the purine and pyrimidine moieties (viz other local geometrical parameters) can also affect the Jcouplings and chemical shifts of the N-H · · · N H-bonds. Another interesting phenomenon may be observed from the data in Figure 3. Almost all of the calculated 2hJNN couplings for the Na+-perturbed GC base pairs are distributed with values above those for the corresponding natural canonical GC base pairs, while those for the Na+-perturbed AT base pairs are distributed below those for the corresponding natural canonical AT base pairs. Although no experimental data for the 2hJNN couplings in the metal ion-bound W-C base pairs have been reported, an analogous increase in the 2hJNN (by ca. 1 Hz) and δ(1H) (by 2-3 ppm) NMR parameters in the GC dimeric subunit of the C+GC triplex was observed in an E.COSY experiment, compared with 2hJNN (∼6.0 Hz) and δ(1H) (∼12 ppm) of the free GC,4 where C+ is protonated cytidine and may be viewed as a solvated proton (a weak Lewis acidic cation).5 In addition, the experimentally observed 10.6 ppm of δ(1H) for the U H3 proton in (SL5)Mg2+AU also reflects the effect of metal ion binding on the NMR parameters of the base pairs,30 a considerable downfield shift compared with that (13.6 ppm) of the free AU.4 These experimentally observed changes of δ(1H)/δ(15N) for several metal ion-bound base pairs31 support our findings that are obtained from DFT/MD calculations. B. J-Couplings Variations in Modeled Cation-Bound W-C Base Pairs. To further explore the influence of metal ions on NMR parameter variations as obtained from DFT/MD calculations, a detailed analysis of the 2hJNN and 1hJNH couplings of the W-C base pairs for the isolated natural (unbound) and Mz+bound W-C base pairs (Mz+ ) Na+, K+, Mg2+, Ca2+, and Zn2+) was made on the basis of their optimized geometries at a B3LYP/6-311+G* level of theory. For GC and AT, the calculated 2hJNN/1hJNH values in the N-H · · · N H-bond units are 4.84/2.81 and 6.37/2.58 Hz, respectively. The 2hJNN coupling in AT is larger than that in GC, in agreement with previous studies.2,4,7,32 Clearly, the stronger coupling for AT coincides with a shorter N · · · N distance (AT/GC: 2.82/2.95 Å observed in crystal structures33 and 2.85-2.91/2.94-2.97 Å calculated

Li et al. at the B3LYP/6-311+G* level of theory), indicating the sensitivity of the 2hJNN scalar coupling to the N · · · N distance that has been noted before.9 The calculated NMR parameters are given in Tables S2 and S3 (Supporting Information) for the W-C base pairs bound by Mz+ (Mz+ ) Na+, K+, Mg2+, Ca2+, Zn2+) and by hydrated Mg2+, with corresponding geometries displayed in Figure 1 as well as Figures S2 and S3 in the Supporting Information. It is noteworthy that the above NMR parameters may significantly shift to different extents upon bindings of metal cations, in a manner that depends on the effect of the positive charge or the Lewis acidity of the cation on the base pairs. As known from experimental findings,5,9,30,31 and supported by the above DFT/ MD results, GC and AT behave as two distinctly opposite variations upon cationization (Table S3, Supporting Information, and Figures 4 and S4-S6, Supporting Information). For the GC series, cationization enlarges the 2hJNN values by ca. 1-2 Hz for the monovalent case and ca. 3-5 Hz for the divalent case, and the purine N3 site (the minor groove site) binding yields slightly larger 2hJNN increments than does the purine N7 site (the major groove site) binding. The AT series present an inverse change behavior, that is, the cationization decreases the 2hJNN values by ca. 1-3 Hz for the monovalent case and ca. 3-5 Hz for the divalent case. Moreover, the purine N3 site binding yields a smaller 2hJNN decrement than the purine N7 site binding for the AT series. Although the 2hJNN couplings of the Mz+GC and Mz+AT series present different trends upon cationization of GC and AT, they exhibit a good correlation with the N · · · N distances for all these GC/AT derivatives falling approximately on the same correlation line (Figure S5, Supporting Information). Since, typically, metal ions in biological surroundings are hydrated or partly hydrated,21,22 hydrated metal ion binding results can describe a generic situation. Therefore, the relevant NMR parameters were also determined for (H2O)nMg2+GC(2) (n ) 1-4, where n ) 4 would correspond to the most common case of six coordinate Mg2+, see Figure S3, Supporting Information), and the results are given in Tables S2 and S3 in the Supporting Information. As predicted, the 2hJNN increment decreases along with the increased number of waters in the primary hydration shell. The tendency is that hydration of the cation site decreases its Lewis acidity, by screening its positive charge, thus reducing the increment of 2hJNN compared with the unhydrated case. Similarly, considerable variations were also observed for the 2hJNO couplings of the N-H · · · O H-bonds between purine and pyrimidine moieties, which support the above findings. C. ICSs Variations in Modeled Cation-Bound W-C Base Pairs. In contrast with the NMR spin-spin coupling constants that monitor the mutual interactions of nuclear spins in pairs of nuclei connected by a path of chemical bonds, the NMR ICSs may be used to probe the electronic environment in the vicinity of the nuclei. The predicted HNN-COSY plot displayed in Figure S4 (Supporting Information) also exhibits large changes of these cation-bound base pairs relative to natural canonical GC and AT. The range of δ(1H) for the embedded protons is 8-18.5 ppm. Similar to the variation of 2hJNN couplings, the general rule is that cationization causes δ(1H) to shift upfield from 12.3 (GC) to 13.7-17.9 ppm for the Mz+GC series and downfield from 13.7 (AT) to 12.5-8.5 ppm for the Mz+AT series, and the magnitude variation (the increment for the Mz+GC series or the decrement for the Mz+AT series) depends strongly on the Lewis acidity of the binding metal ions. As shown in Figure S5 (Supporting Information), these δ(1H) shifts

N-H · · · N Hydrogen Bonds in the W-C Base Pairs

J. Phys. Chem. B, Vol. 112, No. 30, 2008 9177

Figure 4. A schematic representation of the NBO (σN-H, σ*N-H of the localized N-H bonds of the G and T moieties and the lone pair (LP) centered on the acceptor N of the A and C moieties) energy levels for the N-H · · · N H-bond units in the GC/AT derivatives. E(2) denotes the second-order interaction energy between the donor N LP and the acceptor N-H σ* NBOs indicated by a bidirectional arrow, while ∆ELPfσ* denotes the energy gap between these two NBOs. The black upward-pointing arrow denotes the E(2) increase, while the downward-pointing arrow denotes the E(2) decrease, due to Mz+ binding. The spacing between LPN and σ*N-H denotes their relative separations (∆ELPfσ*) for different situations.

correlate well with the 2hJNN changes for both Mz+GC/Mz+AT series. These trends agree well with experimental observations.5,30 Another marked change induced by cationization is on the ICS differences, ∆δ(15N), between the δ(15N) for the H-bond acceptor and donor 15N nuclei, in the same N-H · · · N H-bond.34 The calculated ∆δ(15N) values in GC (58.1 ppm) and AT (61.8 ppm) are almost equal to each other for the two natural GC and AT. Both are very close to the 60 ppm obtained by 2hJNN HNN-COSY spectroscopic experiments.6 Although the two δ(15N) values of the N-H · · · N unit in GC are considerably different from those in AT, indicating their sensitivity to the N-H · · · N unit local surroundings,35 ∆δ(15N) is not as sensitive as is evident from the almost equivalent ∆δ(15N) values for the two cases (58.1 vs 61.8 ppm), which implicates the similar H-bond property of the two N-H · · · N units in GC and AT. The cationized base pair ∆δ(15N) values in GC or AT deviate significantly from 60 ppm (Figure S6 and Table S3 in the Supporting Information) because of the effect of the metal ions on the electron cloud distribution around 15N nuclei, as well as on its nearest neighbor 15N in the H-bond. Moreover, compared with ∆δ(15N) in the natural canonical GC/AT W-C base pairs shown in Figures S4 and S6 (Supporting Information), ∆δ(15N) increases for the Mz+AT series, but decreases for the Mz+GC series. They are also well correlated with δ(1H) and2hJNN for all of GC/AT and their cationized derivatives (Figure S6, Supporting Information). 2. Linear Correlations of 2hJNN, δ(1H) and ∆δ(15N) with the N-H · · · N Unit NBO Quantities. Unlike a number of NMR properties that can be understood in terms of pseudoclassical physics, J-coupling is a purely quantum mechanical phenomenon, and is closely associated with the covalent character of H-bonds,36–38 as can be demonstrated from an orbital analysis.2 The covalency of the H-bonds in the base pairs considered here should be attributed to the electronic properties of the N-H · · · N H-bond,9 and hence the nature of these localized orbitals (the lone-pair orbital of the acceptor N (LPN) and the σ* orbital of the donor N-H (σ*N-H)) and their interactions can directly affect the structural properties of the N-H · · · N H-bonds and impact their J-coupling and ICS values, as schematized in Figure 4. The NBO description has been recognized to be a reliable tool for the rationalization of H-bonds. Thus, the different variation trends caused by cation perturbation for these NMR parameters in both Mz+GC/Mz+AT series should be attributed to their different H-bond patterns, and can be explained well from the NBO analysis. NBOs (that correspond closely to the Lewis structure representations of a molecule) and their contributions to

J-couplings originate from three sources: one from pairwise delocalization of spin density between the LPN in the H-bond acceptor and the σ*N-H in the donor, and two Lewis-type contributions from LPN and σN-H NBOs, respectively (Figure 4).2 On the other hand, from the electronic mechanism of indirect J-coupling, the total contribution may be described as a sum of four terms: FC, DSO, PSO, and SD.39 Inspection of NMR data shows that all of the trans-H-bond couplings are dominated by the FC contribution, while the contributions from DSO, PSO, and SD are very small (Table S2 and Figure S8 in the Supporting Information). As found in previous work, our results also confirm that the N · · · N distance9 and the energy gap (∆E(LPNfσ*N-H))10 are responsible for the FC contribution to 2hJNN couplings across the N-H · · · N H-bonds in these metal cation bound base pairs. Similarly, a detailed analysis regarding our NMR results and NBO quantities also reveals good correlations of the N · · · N distance, as well as ∆E(LPNfσ*N-H), with the FC term, and thus with the 2hJNN couplings for the N-H · · · N H-bond units of various cationized base pairs (Figure 5, as well as Table S2 in the Supporting Information). On the other hand, the magnitude of the dominating FC term strongly depends on the electronic contributions, and the second-order interaction energy E(2) between LPN and σ*N-H represents the deviation of the molecular energy from the energy of the Lewis structure. It can thus be considered as a measure of the electronic delocalization (see the Supporting Information) from the electron donor LPN NBO to the acceptor σ*N-H NBO. As shown in Figure 5, there is a strong correlation between 2hJNN and E(2) in both Mz+GC/Mz+AT series. NBO analyses should be able to explain the trends in 2hJNN and ICSs (∆δ(15N) and δ(1H)) that we have calculated. It is known that ionization energies of the purine moieties in the base pairs are lower than those for pyrimidines. Therefore, together with their favorable nucleophilic sites (N3 and N7), the purine moieties behave as the preferred attacking centers for the Lewis acidic metal cations. As a consequence of the metal ion binding, the electron cloud over the purine moiety redistributes with a major movement to the metal ion bound sites (N3 or N7), resulting in a decrease of electron density over the pyrimidine moiety. For the Mz+GC series, the polarization effect of the metal ions increases the ability of the G moiety to withdraw the electron over the N3 site (the H-bond acceptor site) of the C moiety, and thus promotes σ charge transfer from the occupied LPN NBO in the C moiety to the unoccupied σ*N-H NBO in the G moiety. Consequently, their E(2) is increased, and the energy gap ∆E(LPNfσ*N-H) is correspondingly decreased (Figure 4). This influence is larger if the electron-

9178 J. Phys. Chem. B, Vol. 112, No. 30, 2008

Li et al.

Figure 5. (a) Inverse correlation between 2hJNN and the stabilization energy of the N-H · · · N H-bond for all natural canonical and metal ioncoupled base pairs considered in this work. For each series (GC or AT), the correlation is blocked as distinct regions. The second-order interaction energies E(2) between the donor NBO (LPN) and the acceptor NBO (σ*N-H) were calculated in the framework of NBO analysis. (b) The inverse correlation between 2hJNN and ∆ELPfσ* for all natural canonical and metal ion-bound base pairs considered in this work. ∆ELPfσ* values, the energy differences between the LPN and σ*N-H orbitals (in kcal/mol), were calculated in the framework of NBO analysis. (c) The correlation of 2hJNN with the electron occupancy on the N-H σ* bond in the N-H · · · N unit for all natural canonical (Mz+-unbound) as well as Mz+-bound W-C base pairs considered in this work.

withdrawing ability of the introduced metal ion is stronger. That is why the binding of Mz+ can increase the σ*N-H occupancy by 0.01-0.06 electron as displayed in Table S2 (Supporting Information) and the σ*N-H occupancies depend on the electronwithdrawing ability of the introduced Mz+ ions. Furthermore, this LPNfσ*N-H electron cloud migration increases the electron densities between the two N centers and around the H nuclei of the N-H · · · N H-bond, thus elongating N-H bond and shortening the H · · · N and N · · · N bonds. At the same time, this electron cloud redistribution also increases the covalent character of the N · · · N region, which favors stronger J-coupling. As a result, the binding of metal cations at either the N7 or N3 sites of the G moiety increases 2hJNN, shifts δ(1H) upfield, and reduces ∆δ(15N), the δ(15N) difference between two 15N atoms, as is evident in Figure 5 as well as Figures S4-S6 in the Supporting Information. For the Mz+AT series, the trend is inverted. This should be attributed to the different N-H · · · N H-bond pattern of the Mz+AT series with respect to the Mz+GC series. In the N · · · H-N H-bond unit of AT, the A moiety acts as a proton acceptor instead of a proton donor, and the charge transfer direction for LPNfσ*N-H is from the purine (A) to the pyrimidine moiety (T). Introduction of a metal ion at the purine moiety (A) may impede the σ charge transfer from the occupied LPN (A) to the unoccupied σ*N-H (T). Such an impediment may decrease E(2), and increase the energy gap ∆E (LPNfσ*N-H) (Figure 4). In this case, the σ*N-H occupancy at the T moiety is reduced by 0.01-0.05 electron compared with that in the cation-unbound case. As a result, the cation binding to the purine moiety decreases the covalent property of the N-H · · · N H-bond with an elongation of the N · · · N distance, and thus leads to 2hJ 1 15 NN being decreased, δ( H) shifting downfield, and ∆δ( N) 15 enlargement, via upfield-shifting δ( N) of the purine (A) N1 atom (Figure 5, as well as Figures S4-S6 and Table S2 in the Supporting Information). 3. Dependences of the NMR Parameters on Charge Transfer Amounts Between Two Base Moieties. In general, the effect of metal ion binding on the N-H · · · N H-bonds in the base pairs can be understood qualitatively in terms of changes of the N-H · · · N H-bond lengths and redistribution of electron density over the base pair. Although, inherently, the bond length changes are also related to the electron density redistribution, the geometrical changes cannot account for the large changes observed here for the NMR parameters of these metal ion-bound base pairs, as revealed by the shifts of the

correlation lines in Figure 3. Clearly, the upward-downward shifts of the 2hJNN-N · · · N distance correlation lines in the Na+bound case relative to those in the Na+-unbound case should be attributed to the pure electronic factor, instead of the geometrical one. Therefore, the charge transfer between the purine and the pyrimidine moieties is the dominant factor in influencing these NMR parameter variations, and the trans-Hbond couplings and the corresponding chemical shifts in the N-H · · · N H-bond units of the base pairs are charge-transfer controlled. The NBO analysis indicates that the σ*N-H occupancy should be an overall indicator of the charge exchange between the purine N or N-H and the pyrimidine N-H or N in the base pairs, although a part of this migrated charge is directly used to generate J-coupling and chemical shift in the N-H · · · N unit and others are used to change the H-bond geometry, thus indirectly affecting the NMR parameters. Clearly, the σ*N-H occupancy and its variations upon cation binding account well for 2hJNN and the other NMR parameters, as shown in Figure 5c, with a strong correlation between 2hJNN and σ*N-H occupancy. However, the polarization effect of metal ion binding to the purine moiety causes not only the LPNfσ*N-H charge migration in the N-H · · · N H-bond unit, but also LPOfσ*N-H charge migrations in the coexisting N-H · · · O H-bond units (one in AT and two in GC, Figure 1). The latter should undoubtedly be responsible for the geometry and NMR parameters of the corresponding N-H · · · O H-bond units, and its change can account for the NMR parameter variations (see the Supporting Information). It should be noted that, in addition to the σ-type electron exchanges along the H-bond channels, the metal ion binding to the purine moiety can also induce π-type charge exchange due to the large conjugation between two base moieties, although this type of charge migration does not considerably change the π electron distribution over the N-H · · · N H-bond unit and thus does not contribute to the NMR parameter variations. Clearly, the net charge exchange between the purine and the pyrimidine moieties in any of the base pairs (GC/AT or Mz+GC/Mz+AT) is measured by a balance among the LPNfσ*N-H and LPOfσ*N-H σ-type charge migration, and the π-type charge migration. To clarify the dependence of 2hJNN couplings involved in the N-H · · · N H-bond on the amount of charge transfer through just the N-H · · · N channel, a model system was used in which other internucleotide N-H · · · O H-bonds that also act as bridges for charge transfer between the purine and the pyrimidine

N-H · · · N Hydrogen Bonds in the W-C Base Pairs

J. Phys. Chem. B, Vol. 112, No. 30, 2008 9179

Figure 6. The scheme used for discussing the dependence of 2hJNN and δ(1H) on the σ charge transfer amount (∆QCNBO or ∆QTNBO) only through the N-H · · · N H-bond channel (the other channels (through N-H · · · O H-bond(s)) are broken via the twist) in which ∆QCNBO and ∆QTNBO are adjustable subject to the Na+ · · · N contact distances. The corresponding digits in their vicinities are the contact distances for scan calculations in a range of 2-4.5 Å, for the Na+GC(1) and Na+AT(1) cases. The curved arrows denote the electron transfer directions (from the electron donor LPN to the acceptor σ*N-H) and the straight arrows denote those directions from the purine N3 LPs to the Na+ ions (see the Supporting Information for calculational details).

moieties were broken by a rotation (Figure 6). Concomitantly, the π channel for electron transfer also disappears. Consequently, the amount of σ-type charge transfer between the purine and the pyrimidine moieties in a base pair through the N-H · · · N H-bond can be measured by ∆QTNBO or ∆QCNBO (where T and C denote the corresponding pyrimidine bases, and ∆QT/CNBO denotes the NBO charge change over T or C). Moreover, the amount of σ charge transfer between the donor LP(N) NBO and the acceptor σ*(N-H) NBO can be modulated by varying the coupling distance between the metal counterion and the purine moiety (the N3 site examined here) because the electronwithdrawing effect is weakened as their corresponding distance is lengthened (shown in Figure 6). Starting with the optimized Na+AT(1) and Na+GC(1) geometries, J-coupling constants and chemical shifts were determined for Na+AT(1) and Na+GC(1) base pair complexes by varying the Na+ · · · N3(purine) distance, as displayed in Figure 6 (see the Supporting Information for details). The computed 2hJ , δ(1H), and ∆δ(15N) values and the σ charge transfer NN amounts (∆QCNBO, ∆QTNBO) versus different Na+ · · · N3(purine) distances are summarized in Table S4 and Figure S9 in the Supporting Information. A notable character for these NBO charge changes is that ∆QCNBO > 0 in the Na+GC case (electron density increase over C), but ∆QTNBO < 0 in the Na+AT case (electron density decrease) over T. This observation should be attributed to the different H-bond patterns and thus to the different charge transfer directions required for strengthening the N-H · · · N H-bonds in GC and AT base pairs (GfC versus AfT). Further, the size of ∆QTNBO/∆QCNBO reflects the magnitude for the increase/decrease of the electron densities over the T/C moieties due to the charge transfer into/out of the pyrimidine moiety. In general, -∆QTNBO decreases and ∆QCNBO increases as a Na+ ion approaches the N3(purine) site, as shown in Figure S10 (Supporting Information). These variations are also understandable from their different charge transfer directions (Na+rGrC versus Na+rAfT). That is, when a Lewis acidic cation approaches the N3 site of the purine moiety, the polarization effect can pull a part of the electron cloud to the metal ion bound site, and cause a redistribution of electron density over the base pair dimer. As a result, the amount of charge transfer changes between the purine and pyrimidine moieties. That is, the charge transfer amount to T decreases in Na+AT compared with AT, while that from C increases compared with GC, and the closer the Na+ is to the purine, the larger the magnitude of the change, which produces a strong correlation between ∆QNBO and the Na+ · · · purine distance. As expected, for the two series, all NMR parameters exhibit linear correlations with the charge transfer amount (∆QTNBO or ∆QCNBO) associated with the corresponding N-H · · · N H-bond units (Figure 7), indicating again that variations of these NMR parameters are charge transfer controlled. Therefore, it is clear

Figure 7. The strong dependences of 2hJNN, δ(1H), and ∆δ(15N) associated with the N-H · · · N H-bond units in the Na+GC(1) and Na+AT(1) cases, as shown in Figure 6, on the σ charge transfer amount (∆QCNBO or ∆QTNBO).

that the different changes of the 2hJNN, δ(1H), and ∆δ(15N) parameters in the W-C base pairs upon cationization can be attributed to the different electron transfer directions, amounts, and the polarization effect of the metal ions, in the Mz+AT series (Mn+rAfT) versus the Mz+GC series (Mn+rGrC). Conclusions The 2hJNN-coupling and δ(1H)/∆δ(15N) chemical shifts of the N-H · · · N H-bond units in internucleotide base pairs were determined theoretically in this work at a B3LYP/6-311G** level via two schemes: (i) using snapshot geometries with the presence and absence of a counterion (Na+) extracted from molecular dynamics simulations and (ii) using the optimized geometries for a set of the isolated metal ion-bound base pairs. Our main conclusion is that these NMR parameters can vary considerably upon cation bindings with the natural GC or AT base pairs, and thus may be used as important indicators of cation-perturbation of base pairs. In particular, cation-bindinginduced variations are distinctly different from each other for GC and AT, with the basic trends that cation binding can cause 2hJNN to increase, ∆δ(15N) to decrease, and δ(1H) to shift

9180 J. Phys. Chem. B, Vol. 112, No. 30, 2008 upfield for GC, and in opposite directions for AT. These trends have been indirectly seen in experiments by the variations observed for 2hJNN/δ(1H) in a C+GC triplex and the shifts in δ(1H) in other metal ion-bound base pairs. The magnitudes of variation for these quantities depend strongly on the Lewis acidity of the metal ions, thus offering the possibility of differentiating the metal ion properties (e.g., divalent or monovalent). More interesting is that all the quantities, 2hJNN, ∆δ(15N), and δ(1H), for both base pair series (Mz+GC and Mz+AT), are linearly correlated, and their values may be interpreted from NBO analysis. These NMR quantities strongly depend on the energy gaps (∆ELPfσ*) and their second-order interaction energies (E(2)) between the donor N lone pair (LPN) and the acceptor σ*N-H localized NBO orbitals for all base pairs. In particular, the 2hJNN changes are also closely related to the amount of σ charge transfer from the H-bond acceptor N LPN to the donor σ*N-H NBOs (also corresponding to the σ*N-H occupancy), or from the purine to pyrimidine (∆QT/CNBO), and thus these changes are likely charge-transfer-controlled. Their differing trends of variation may be attributed to the different H-bond patterns and thus different charge transfer directions in the Mz+AT series ((purine)N · · · H-N(pyrimidine) combined with polarization effect of the metal ions, Mz+rAfT) and the cationized GC series ((purine)N-H · · · N(pyrimidine), Mz+rGrC). The sensitivity of the NMR parameters to the presence of metal ions should provide an alternative method for probing the heterogeneity of DNA sequences. It should be noted that there is a paucity of experimental data for 2hJNN (and other parameters) in metal ion-bound DNA/ RNA complexes because of the difficulty of their determinations, due in part to the dynamics and fluctuations of DNA/RNA duplex structures. Certainly, further direct evidence is still needed from NMR experiments (e.g., HNN-COSY or 15N-1H TROSY techniques, etc.) for an enhanced understanding of cation binding effects on H-bond NMR properties in DNA/RNA. Acknowledgment. This work was supported by NSFC (20633060 and 20573063 to Y.B.), NIH (Grant No. GM62790 to R.I.C.), NCET (to Y.B.), and Virt. Laboratory for Comput Chem & SCC of CNIC-CAS, MCBILIN at MSU, HPCC at SDU. Supporting Information Available: Detailed calculational methods, NBO remarks, examples about metal ion binding to DNA, analyses about NMR parameters of the N-H · · · O H-bond units, as well as the calculated NMR (J-couplings 2hJ 2hJ 1hJ 15 1 15 NN, ON, NH and chemical shifts δ( N, H), ∆δ( N)) parameters, main geometrical parameters, and additional correlations among NMR quantities, geometrical, and NBO parameters. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Gemmecker, G. Angew. Chem., Int. Ed. 2000, 39, 1224–1226. (2) Wilkens, S. J.; Westler, W. M.; Weinhold, F.; Markley, J. L. J. Am. Chem. Soc. 2002, 124, 1190–1191. (3) Henning, M.; Williamson, J. R. Nucleic Acids Res. 2000, 28, 1585– 1593. (4) Dingley, A. J.; Grzesiek, S. J. Am. Chem. Soc. 1998, 120, 8293– 8297. (5) Dingley, A. J.; Masse, J. E.; Peterson, R. D.; Barfield, M.; Feigon, J.; Grzesiek, S. J. Am. Chem. Soc. 1999, 121, 6019–6027. (6) Wo¨hnert, J.; Dingley, A. J.; Stoldt, M.; Go¨rlach, M.; Grzesiek, S.; Brown, L. R. Nucleic Acids Res. 1999, 27, 3104–3110. (7) (a) Pervushin, K.; Ono, A.; Ferna´ndez, C.; Szyperski, T.; Kainosho, M.; Wu¨thrich, K. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 14147–14151. (b) Pervushin, K.; Riek, R.; Wider, G.; Wu¨thrich, K. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 12366–12371.

Li et al. (8) Zˇ´ıdek, L.; Sˇtefl, R.; Sklena´rˇ, V. Curr. Opin. Struct. Biol. 2001, 11, 275–281. (9) Barfield, M.; Dingley, A. J.; Feigon, J.; Grzesiek, S. J. Am. Chem. Soc. 2001, 123, 4014–4022. (10) Guerra, C. F.; Bickelhaupt, M.; Snijders, J. G.; Baerends, E. J. J. Am. Chem. Soc. 2000, 122, 4117–4128. (11) Denisov, V. P.; Halle, B. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 629. (12) Va´rnai, P.; Zakrzewska, K. Nucleic Acids Res. 2004, 32, 4269– 4280. (13) Ponomarev, S. Y.; Thayer, K. M.; Beveridge, D. L. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 14771–14775. (14) Sines, C. C.; McFail-Isom, L.; Howerton, S. B.; VanDerveer, D.; Williams, L. D. J. Am. Chem. Soc. 2000, 122, 11048–11056. (15) Tereshko, V.; Minasov, G.; Egli, M. J. Am. Chem. Soc. 1999, 121, 470–471. (16) Hud, N. V.; Feigon, J. J. Am. Chem. Soc. 1997, 119, 5756–5757. (17) Moldrheim, E.; Andersen, B.; Frøystein, N. Å.; Sletten, E. Inorg. Chim. 1998, 273, 41–46. (18) Kieft, J. S., Jr. Structure 1997, 5, 713–721. (19) Okamoto, A.; Kanatani, K.; Taiji, T.; Saito, I. J. Am. Chem. Soc. 2003, 125, 1172–1173. (20) (a) Shui, X.; McFail-Isom, L.; Hu, G. G.; Williams, L. D. Biochemistry. 1998, 37, 8341–8355. (b) Shui, X.; Sines, C. C.; McFailIsom, L.; VanDerveer, D.; Williams, L. D. Biochemistry. 1998, 37, 16877– 16887. (21) (a) Draper, D. E. RNA 2004, 10, 335–343. (b) Misra, V. K.; Draper, D. E. J. Mol. Biol. 2002, 317, 507–521. (c) Misra, V. K.; Shiman, R.; Draper, D. E. Biopolymers 2003, 69, 118–136. (22) Juneau, K.; Podell, E.; Harrington, D. J.; Cech, T. R. Structure 2001, 9, 221–231. (23) Wang, G.; Gaffney, B. L.; Jones, R. A. J. Am. Chem. Soc. 2004, 126, 8908–8909. (24) Saito, I.; Nakamura, T.; Nakatani, K. J. Am. Chem. Soc. 2000, 122, 3001–3006. (25) Catte, A.; Cesare-Marincola, F.; Maarel, J. R. C. V. D.; Saba, G.; Lai., A. Biomacromolecules 2004, 5, 1552–1556. (26) Markwick, P. R. L.; Sprangers, R.; Sattler, M. J. Am. Chem. Soc. 2003, 125, 644–645. (27) Sass, H.-J.; Schmid, F. F.-F.; Grzesiek, S. J. Am. Chem. Soc. 2007, 129, 5898–5903. (28) Snapshot geometries with the Na+ · · · N7 contact distances ranging from 3 to 9 Å may be viewed as a transition from the Na+-bound state to the Na+-unbound state, corresponding to a zone with weak interaction. The NMR parameter values fall in between those for the bound and the unbound states, and thus are not considered here. (29) It should noted that the correlation lines for the dependences of several NMR parameters on the N · · · N distance (in a range of 2.7-3.2 Å) of the N-H · · · N H-bond units are basically linear in this work, but for C+GC in ref 9, those for the dependences of various J-couplings (2hJNN, 1J , 1hJ ) and chemical shifts (δ(1H), δ(15N), and ∆δ(15N)) associated NH NH with the N-H · · · N H-bond units on the N · · · N distance were determined to be curved in a large range of 2.6-4.0 Å. These two observations are not in contradiction; in the same N · · · N distance range (2.6-3.2 Å) both are in good agreement with each other, although the correlation lines in these two situations were determined via two completely different methods. (30) For (SL5)Mg2+AU where SL5 may be viewed as a ligand of Mg2+, only the H3 proton δ(1H) of the AU N-H · · · N unit was reported experimentally to be 10.6 ppm (see: Campbell, D. O.; Bouchard, P; Desjardins, G.; Legault, P. Biochemistry 2006, 45, 10591-10605), which is slightly larger than but close to 10.0 ppm of Mg2+AT. Both are smaller than the calculated values (∼13 ppm) and the experiment value (∼13 ppm) of AT and AU. (31) Other examples for the variations of chemical shifts caused by metal ion bindings are as follows: (i) the Cd2+ binding results in a δ(15N7) upfield shift of more than 12 ppm of N7/guanine; see: Tanaka, Y.; Kojima, C.; Morita, E. H.; Kasai, Y.; Yamasaki, K.; Ono, A.; Kainosho, M.; Taira, K. J. Am. Chem. Soc. 2002, 124, 4595–4601. (ii) Zn2+ and Mg2+ ions can cause a 20 ppm upfield shift for the N7/guanosine in DMSO; see: Buchanan, G. W.; Stothers, J. B. Can. J. Chem. 1982, 60, 7787–7791. (iii) Different bindings of different metal ions (Mg2+, Zn2+, and Cd2+) can lead to different N7/guanine upfield shifts ranging from 0.7 to 20 ppm; see: Wang, G.; Gaffney, B. L.; Jones, R. A. J. Am. Chem. Soc. 2004, 126, 8908–8909. (iv) A ∼70 ppm upfield shift occurs for the N1/adenosine upon protonation; see: Markowski, V.; Sullivan, G. R.; Roberts, J. D. J. Am. Chem. Soc. 1977, 99, 714–718. Wang, C.; Gao, H.; Gaffney, B. L.; Jones, R. A. J. Am. Chem. Soc. 1991, 113, 5486–5488. (v) the H-bond to purine N1 or N7 atoms may be viewed as their weak protonation like-Bδ--Hδ+ · · · N1/N7, so smaller upfield changes (typically 2-9 ppm) were experimentally observed upon formation of these specific H-bonds in oligonucleotide duplexes and triplexes; see; Gao, X.; Jones, R. A. J. Am. Chem. Soc. 1987, 109, 3169– 3171. Gaffney, B. L.; Goswami, B.; Jones, R. A. J. Am. Chem. Soc. 1993, 115, 12607–12608. Gaffney, B. L.; Kung, P.-P.; Wang, C.; Jones, R. A.

N-H · · · N Hydrogen Bonds in the W-C Base Pairs J. Am. Chem. Soc. 1995, 117, 12281–12283. All these changes are in agreement with those calculated in this work for the metal ion bound and the unbound base pairs. (32) The good agreement between the theoretical and the experimental values for the considered base pairs implies the biological surrounding effect on 2hJNN/1hJNH to be small. See: Sychrovsky´, V.; Vacek, J.; Hobza, P.; Zidek, L.; Sklenar, V.; Cremer, D. J. Phys. Chem. B 2002, 106, 10242–10250. (33) Saenger, W. Principles of Nucleic Acid Structure; Cantor, C. R., series Ed.; Springer-Verlag: New York, 1984. (34) The same H-bond forms are referred to according to the H-bond donor-acceptor N-H · · · N vector direction. That is, the N-H · · · N H-bond vector is from the purine moiety (G, as a donor) to the pyrimidine moiety (C, as an acceptor) in GC, but is from the pyrimidine moiety (T, as a donor) to the purine moiety (A, as an acceptor) in AT. For the N-H · · · N units in both base pairs, since δ(15N) of the acceptor moieties are always greater

J. Phys. Chem. B, Vol. 112, No. 30, 2008 9181 than those of the donor moieties, we define the δ(15N) difference mentioned in text as ∆δ(15N) ) δ(15N in acceptor)-δ(15N in donor), keeping ∆δ(15N) always positive. (35) Levy, G. C. Nitrogen-15 Nuclear Magnetic Resonance Spectroscopy; John Wiley & Sons: New York, 1979. (36) Cordier, F.; Rogowski, M.; Grzesiek, S.; Bax, A. J. Magn. Reson. 1999, 140, 510–512. (37) Benedict, H.; Shenderovich, I. G.; Malkina, O. L.; Malkin, V. G.; Denisov, G. S.; Golubev, N. S.; Limbach, H. H. J. Am. Chem. Soc. 2000, 122, 1979–1988. (38) Cornilescu, G.; Hu, J.-S.; Bax, A. J. Am. Chem. Soc. 1999, 121, 2949–2950. (39) Ramsey, N. F. Phys. ReV. 1953, 91, 303.

JP8030545