NMR scalar couplings across Watson–Crick base pair hydrogen bonds in DNA observed by transverse relaxation-optimized spectroscopy

Abstract

This paper describes the NMR observation of ¹⁵N—¹⁵N and ¹H—¹⁵N scalar couplings across the hydrogen bonds in Watson–Crick base pairs in a DNA duplex, ^hJ_NN and ^hJ_HN. These couplings represent new parameters of interest for both structural studies of DNA and theoretical investigations into the nature of the hydrogen bonds. Two dimensional [¹⁵N,¹H]-transverse relaxation-optimized spectroscopy (TROSY) with a ¹⁵N-labeled 14-mer DNA duplex was used to measure ^hJ_NN, which is in the range 6–7 Hz, and the two-dimensional ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY experiment was used to correlate the chemical shifts of pairs of hydrogen bond-related ¹⁵N spins and to observe, for the first time, ^hJ_HN scalar couplings, with values in the range 2–3.6 Hz. TROSY-based studies of scalar couplings across hydrogen bonds should be applicable for large molecular sizes, including protein-bound nucleic acids.

Hydrogen bonds in biological macromolecules can usually only be inferred, rather than directly evidenced by experimental techniques (1), including NMR spectroscopy (2). Here we describe the observation by two-dimensional (2D) [¹⁵N,¹H]-transverse relaxation-optimized spectroscopy (TROSY) (3–5) of scalar couplings across the Watson–Crick base pairs in isotope-labeled DNA, which affords direct observation of the hydrogen bonds in these structures. Scalar couplings across hydrogen bonds have been previously reported for organic-synthetic compounds (6, 7), RNA fragments (8), and a metalloprotein (9, 10). The variability of such couplings observed so far indicates that they may become sensitive new parameters for detection of hydrogen bond formation and associated subtle conformational changes. Furthermore, in conjunction with quantum-chemical calculations, precise measurements of scalar couplings across hydrogen bonds can be expected to provide novel insights into the nature of hydrogen bonds in chemicals and in biological macromolecules.

MATERIALS AND METHODS

Fully and partially ¹³C,¹⁵N-doubly labeled DNA oligomers were synthesized on a DNA synthesizer (Applied Biosystems model 392–28) by the solid-phase phosphoroamidite method, by using isotope-labeled monomer units that had been synthesized according to a previously described strategy (11). Approximately 1 μmol of oligomer was obtained from 5 μmol of nucleoside bound to the resin. NMR samples of the DNA duplex at a concentration of ≈2 mM were prepared in 90% H₂O/10% D₂O containing 50 mM potassium phosphate and 20 mM KCl at pH 6.0. NMR measurements were performed at 15°C on Bruker DRX500 and DRX750 spectrometers equipped with ¹H-{¹³C,¹⁵N} triple-resonance probeheads.

For the present study, an estimate of the line widths to be expected from the use of TROSY was of critical interest. To estimate the reduction of the ¹⁵N and ¹H relaxation rates in TROSY when compared with conventional spectroscopy, one needs information on the principal values and orientations of the chemical shift tensors of ¹⁵N and ¹H, as well as on the ¹⁵N—¹H distance (3–5). Here, we collected this information from the following sources: An estimate of the chemical shift anisotropy tensor of the imino ¹⁵N spin was obtained by use of the solid-state NMR data (12) on ¹⁵N₃ in uracil, which has principal values of δ₁₁ = 200 ppm, δ₂₂ = 131 ppm, and δ₃₃ = 79 ppm, with the largest component oriented at 9° relative to the ¹⁵N₃–¹H₃ bond and the smallest component perpendicular to the ring plane. For the imino proton, we used Δσ = 8 ppm, based on the measured anisotropy, Δσ = δ₁₁ - 0.5(δ₂₂ + δ₃₃), of ≈6 ppm in 2′,3′,5′-tri-O-benzoyl(3-¹⁵N)uridine (13) and the available data on hydrogen-bonded amino protons (14–18). For the imino ¹⁵N—¹H bond length, the solid-state NMR value of 0.11 nm for G and T in a hydrated DNA duplex (19) was used. Relaxation of the imino proton due to dipole–dipole (DD) coupling with remote protons in the DNA duplex was represented as follows (2): in the Watson–Crick A⩵T pair by an adenosine amino proton at a distance of 0.24 nm and the adenosine C2 proton at 0.3 nm; in G C by a guanosine amino proton at 0.22 nm and a cytosine amino proton at 0.25 nm. For both base pairs, two imino protons in sequentially stacked bases at 0.4 nm also were considered. Following the calculations outlined in refs. 3–5, the use of TROSY at a polarizing magnetic field B_o = 17.6 T is expected to yield 65% and 30% reductions of the ¹⁵N and ¹H linewidth, respectively, for A⩵T base pairs and 55% and 20% reductions for G C base pairs. If the contributions from dipolar interactions with remote protons are neglected, the calculations predict reductions of 85% and 75% for ¹⁵N and ¹H in both A⩵T and G C base pairs, which may perhaps be in part exploited when working with H₂O/D₂O mixed solvents.

RESULTS AND DISCUSSION

The theoretical considerations presented in the preceding section predict that compared with the corresponding conventional NMR experiments, [¹⁵N,¹H]-TROSY (3–5) yields ≈70% and 30% reductions of the ¹⁵N and ¹H linewidths, respectively, when used to record signals for the guanosine ¹⁵N₁—¹H₁ and thymidine ¹⁵N₃—¹H₃ imino groups in a ¹⁵N-labeled DNA duplex. The reduced TROSY line widths then allow direct measurements of ¹⁵N—¹⁵N and ¹H—¹⁵N scalar spin–spin couplings across hydrogen bonds, ^hJ_NN and ^hJ_HN, respectively. Fig. 1 A and B, shows regions of the 2D [¹⁵N,¹H]-TROSY spectra measured with two differently ¹⁵N-labeled 14-bp DNA duplexes. For the uniformly ¹⁵N-labeled duplex, an in-phase splitting along ω₁(¹⁵N) is observed for all imino ¹⁵N spins involved in Watson–Crick base pairs (Fig. 1C). A first indication that these splittings are due to spin–spin coupling came from the observation that the measured ^hJ_NN values are independent of the strength of the polarizing magnetic field, B_o, as evidenced by experiments at ¹H frequencies of 500 and 750 MHz. To unambiguously identify the origin of this splitting, an identical spectrum was recorded for a partially ¹⁵N-labeled duplex, in which either only one or both nucleotides in the individual Watson–Crick base pairs are ¹⁵N-labeled (20, 21). For example, the [¹⁵N₁,¹H₁]-correlation peak of G¹² shows a doublet fine structure, whereas the cross-peak of G⁵ appears as a singlet, since C²⁴ is not ¹⁵N-labeled (Fig. 1D Upper). Similarly, T²¹ shows doublet and singlet fine structures, respectively, in the two differently labeled duplexes (Fig. 1C), confirming that the splitting observed in the uniformly labeled duplex is due to ¹⁵N—¹⁵N coupling.

Figure 1.

NMR observation of scalar ¹⁵N—¹⁵N couplings across hydrogen bonds in DNA, ^hJ_NN. (A Top and Middle) Contour plots of the region containing signals of A⩵T base pairs from [¹⁵N,¹H]-TROSY spectra (3, 5) of the partially and uniformly ¹³C,¹⁵N-labeled DNA duplex, respectively. (A Bottom) Contour plot showing the relayed cross-peaks obtained with 2D ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY, where the broken contours indicate negative spectral intensities. The direct [¹⁵N₃(T),¹H₃(T)]-correlations (Top and Middle) and the relayed [¹⁵N₁(A),¹H₃(T)] cross-peaks (Bottom) of the A⩵T base pairs are shown. Chemical shifts are relative to 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt (DSS). The DNA sequence is shown in the Top, where the ¹⁵N- and ¹³C-containing nucleotides in the partially ¹³C,¹⁵N-labeled DNA (see text) are underlined. (Insert) Definition of the scalar couplings ^hJ_NN and ^hJ_HN in Watson–Crick base pairs. (B) Same presentation as A for the G C base pairs. (C) Cross sections along ω₁(¹⁵N) through the individual cross peaks in the spectra (A). (C Lower) Direct correlation cross-peaks in the uniformly ¹³C,¹⁵N-labeled DNA duplex. (C Upper) Direct correlation cross peaks in the partially labeled duplex. (D) Same presentation as C for cross sections along ω₁(¹⁵N) taken through the individual cross peaks in the spectra (B).

In addition to the observation of the resolved in-phase splittings in the 2D [¹⁵N,¹H]- TROSY spectra (Fig. 1 C and D), precise values for ^hJ_NN were obtained by inverse Fourier transformation of the in-phase multiplets (22). Table 1 shows that there are sizeable variations in the ^hJ_NN couplings among the A⩵T base pairs, e.g., T¹⁰ and T¹⁶ (Fig. 1C), as well as among the G C base pairs (Fig. 1D), which probably reflect local differences in Watson–Crick base pairing geometry and/or dynamic processes such as fraying of the chain ends.

Table 1.

Values and SDs of ¹⁵N—¹⁵N and ¹H—¹⁵N scalar spin–spin couplings across hydrogen bonds, ^hJ_NN and ^hJ_HN, in Watson–Crick base pairs in DNA (see Fig. 1A).

Residue	hJNN, Hz	hJHN, Hz
T9	6.7  ±  0.1	3.6  ±  0.1
T10	7.0  ±  0.1	3.2  ±  0.3
T16	6.5  ±  0.1	2.0  ±  0.2
T18	6.8  ±  0.1	2.7  ±  0.2
T21	6.5  ±  0.1	3.7  ±  0.3
T25	7.0  ±  0.1	2.7  ±  0.2
T26	6.9  ±  0.1	2.7  ±  0.1
T27	7.0  ±  0.1	1.8  ±  0.2
G5	6.0  ±  0.1	2.8  ±  0.3
G12	6.3  ±  0.1	2.4  ±  0.5
G22	6.5  ±  0.1	2.9  ±  0.6
G23	6.4  ±  0.1	3.6  ±  0.5

The new 2D ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY experiment (Fig. 2) uses the slowly relaxing component of the imino ¹⁵N doublet to relay magnetization via ^hJ_NN across the hydrogen bond to the tertiary ¹⁵N position of the second base in the Watson–Crick base pair. The sensitivity of the experiment is further enhanced by use of both the ¹H and ¹⁵N steady–state magnetizations (3–5, 23). The correlation of the chemical shifts of pairs of hydrogen bond-related nucleotides should be of particular interest for unambiguous identification of base pairs in DNA or RNA molecules exhibiting a distinct tertiary structure. The inherent high sensitivity of [¹⁵N,¹H]-TROSY for studies of large molecular sizes (3–5) should enable the use of this approach with large nucleic acid fragments, and with nucleic acids in protein complexes.

Figure 2.

Experimental scheme for the 2D ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY experiment used to correlate the chemical shifts of pairs of hydrogen bond-related ¹⁵N spins and to measure ^hJ_HN. On the lines marked ¹H, ¹⁵N, and ¹³C, narrow and wide bars stand for nonselective 90° and 180° radio frequency pulses, respectively, with the carrier frequencies at 12, 190, and 145 ppm, respectively. The delays are τ₁ = 2.7 ms, τ₂ = 18 ms, and τ₃ = τ₂ − τ₁. The line marked PFG indicates the pulsed magnetic field gradients applied along the z axis: G₁, amplitude 30 G/cm, duration 1 ms; G₂, 42 G/cm, 1 ms; G₃, 45 G/cm, 1 ms; G₄, 40 G/cm, 1 ms; and G₅, 48 G/cm, 1 ms. Two data sets, I and II, are measured in the interleaved manner with the following phase cycling schemes: (I), φ₁ = {−x, x}; φ₂ = {2(−x),2x}; ψ₁ = {4y,4(−y)}; ψ₂(receiver) = {y,−y,−y, y,−y, y,y,−y}; (II), φ₁ = {−y, y}; φ₂ = {2(−y),2y}; ψ₁ = {4(−x),4x}; ψ₂(receiver) = {−x, x,x,−x, x,−x,−x, x}; x on all other pulses. For both data sets, quadrature detection in the t₁ dimension is obtained by using States-TPPI (38) by simultaneously incrementing each of the phases φ₂ and ψ₂ by 90°. The transfer of ¹⁵N magnetization to ¹H for detection of single transition-to-single transition polarization transfer (5) was used. Fourier transformation of the sum and difference of the two data sets results in two 2D ^hJ_NN- correlation-[¹⁵N,¹H]-TROSY spectra in which the relative cross-peak positions are determined in E.COSY manner by the chemical shifts and the scalar-coupling constants (27–30). For example, both the direct and the relayed correlation cross-peaks that appear in the two spectra are shifted by ¹J_HN along ω₂(¹H). Along ω₁(¹⁵N), the direct peaks also are shifted by ¹J_HN, but the separation of the relayed cross-peaks is determined exclusively by ^hJ_HN. To enhance the signals of rapidly exchanging imino protons, water saturation is minimized by keeping the water magnetization along +z during the entire experiment, using the water-selective 90° pulses indicated by curved shapes on the line ¹H.

The coupling constants ^hJ_NN appear to be smaller for G C than for A⩵T base pairs (Fig. 1 C and D), which possibly reflects the longer ¹⁵N—¹⁵N distance in G C when compared with A⩵T base pairs (1, 8). In principle, the ^hJ_NN values could be estimated from the ratio of the direct and relayed cross-peak volumes (Fig. 1), provided that the ¹⁵N—¹⁵N antiphase magnetization does not relax much faster than ¹⁵N in-phase magnetization (24–26). However, even for small nucleic acid fragments with rotational correlation times of ≈10 ns, 2D [¹⁵N,¹H]-TROSY provides an ≈fivefold improved sensitivity when compared with 2D ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY (Fig. 2), so that it is preferable to use inverse Fourier transformation of the in-phase peaks in 2D [¹⁵N,¹H]-TROSY to determine the ^hJ_NN couplings (22).

When the TROSY-type detection scheme ST2-PT (5) was combined with exclusive correlation spectroscopy (E.COSY) (27–30), the 2D ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY experiment of Fig. 2 allowed, for the first time, observation and quantification of ^hJ_HN couplings between tertiary ¹⁵N atoms and the imino proton across the Watson–Crick hydrogen bond in DNA (Fig. 3). In the E.COSY patterns for covalently linked ¹⁵N—¹H moieties and the hydrogen-bonded ¹⁵N… ¹H combinations (Fig. 3), a positive sign of ^hJ_HN can be inferred from the negative sign of ¹J_HN (31). A significantly larger relative variation is observed for the ^hJ_HN couplings in the individual base pairs than for ^hJ_NN, with values ranging from 2 Hz to 3.6 Hz (Table 1). Notably, there is a significant correlation between the magnitudes of corresponding ^hJ_HN and ^hJ_NN couplings within each type of nucleotides (Table 1), with the only exceptions of G⁵ and T¹⁶.

Figure 3.

NMR observation of scalar ¹H—¹⁵N couplings across hydrogen bonds in DNA, ^hJ_HN. The plots show comparisons of the sum and difference 2D ^hJ_NN-correlation-[¹⁵N,¹H]-TROSY spectra recorded with the uniformly ¹³C,¹⁵N-labeled DNA duplex (Fig. 1A) on a Bruker DRX-750 spectrometer, using the experimental scheme of Fig. 2. (A and B) Contour plots with signals from G C base pairs and A⩵T base pairs, respectively. To facilitate comparison, one spectrum was shifted relative to the other by ≈90 Hz along ω₂(¹H), so that corresponding peaks overlap along this dimension. The relayed cross-peaks from each of the two spectra are represented by two contour lines. For selected base pairs, superpositions of cross sections along ω₁(¹⁵N) taken through the original, unshifted peaks are shown above and below the two panels, where the two peaks to be compared have been normalized to the same intensity. The shift between the two peaks provides an E.COSY-type representation of ^hJ_HN (see text).

The nature of the ^hJ_HN and ^hJ_NN interactions as scalar couplings due to electron-coupled interactions between the related nuclear spins (32) is clearly evidenced: (i) the observed in-phase splittings (Fig. 1) are field-independent; and (ii) the observed relative magnitudes of corresponding ^hJ_HN and ^hJ_NN pairs could not be explained by residual DD coupling (33–35). It has been suggested that the Fermi-contact term, which usually dominates scalar couplings through covalent bonds (32), is also effective for scalar couplings through hydrogen bonds (36, 37). Since both the donor imino group and the acceptor nitrogen are embedded in π-conjugated systems, π-bond polarization upon “cyclic” hydrogen bond formation could give rise to π-cooperativity and thus enhance the hydrogen bond stability and the covalent character of the hydrogen bonds (1). In fact, this view seems to be supported by measurements of scalar ^hJ_HH couplings of 1–2 Hz between hydroxyl and formyl protons in O-H… O⩵CH moieties attached to π-conjugated 1,6-dioxapyrene derivatives, where the ^hJ_HH couplings could only be observed in the presence of a substituent that donates additional π-bond polarization (6).

ABBREVIATIONS

TROSY

transverse relaxation-optimized spectroscopy

E.COSY

exclusive correlation spectroscopy

DD

dipole–dipole coupling

2D

two-dimensional