The availability of isotope labeled RNA oligonucleotides (Batey et al. 1992; Nikonowicz et al. 1992; Quant et al. 1994; Batey et al. 1995) turned NMR spectroscopy into a powerful tool to determine structure and dynamics of RNAs up to a size of 100 nucleotides (Varani and Tinoco 1991; Varani et al. 1996; Wijmenga 1998; Cromsigt et al. 2001; Fürtig et al. 2003). Heteronuclear NMR experiments allow resonance assignment of the NMR active 1H, 13C, 15N and 31P nuclei which is a prerequisite for further structural and dynamical studies on RNA molecules. In general, most of the experiments use large heteronuclear couplings in order to create a through-bond correlation of distinct nuclei. However, in some cases these couplings are too small and transverse relaxation especially in the nucleobases is too fast so that NOE-based strategies complement the set of suitable resonance assignment experiments. Usually resonance assignment starts with an NOE-based correlation of the imino resonances, since imino resonances show high dispersion especially in the 1H dimension. The number of peaks is comparably small because only those imino groups give rise to an NMR signal that participate in a hydrogen bridge. From the imino signal, the assignment is extended to the non-exchangeable aromatic protons and then to the H1′ resonances of the ribose moiety (Varani et al. 1996; Wijmenga 1998; Fürtig et al. 2003).

The assignment of the adenine and cytosine bases, which do not possess imino groups, is more difficult. It relies on correlations across the base pair, starting from the imino resonances of guanine and uracil nucleobases. For the adenine nucleobases, the assignment of H2 protons in 13C, 15N-labeled RNA is usually obtained from the strong NOE between the uracil H3 and the adenine H2 in an AU base pair and assignment of the adenine N1 is usually obtained via the HNN-COSY experiment in which the uracil H3-N3 signal is correlated to the N1 resonance of the adenine (Dingley and Grzesiek 1998). Then, it is possible to assign the H2-N1 resonance pair in the adenine nucleobase in the 2J-15N-HSQC unambiguously. The adenine through-base connectivity is most suitably obtained via the TROSY relayed HCCH-COSY experiment (Simon et al. 2001) which allows the assignment of the H2-C2 and the H8-C8 signals to the same nucleobase via the correlation to the C4, C5 and C6 resonances. Base connectivity can also be obtained using a 2D 1H, 13C-HMBC experiment (van Dongen et al. 1996) which is recommended for smaller RNA or DNA molecules where spectral overlap in the proton dimension only rarely occurs.

However, there is the need to correlate the H2-N1 resonances of the 2J-15N-HSQC to the C2 resonances. Such correlation can be accomplished with the 13C-HSQC as long as there is no signal overlap in the 1H dimension. However, to resolve signal overlap, it is necessary to correlate signals observed in the 2J-15N-HSQC to the C2 resonances. Such correlation cannot be achieved in an HCN experiment (Sklenar et al. 1993), since the 1J(C2, N1(3))-coupling is smaller than 1 Hz (Fig. 1) (Fiala et al. 2004). Therefore, we developed a novel pulse sequence that correlates the three nuclei H2, C2 and N1(N3) in the adenine nucleobase by utilizing the strong 2J(H2, N1(3)) and 1J(H2, C2) couplings (as depicted in Fig. 1) (Fiala et al. 2004).

Fig. 1
figure 1

Schematic representation of the scalar couplings involved in the coherence transfer pathway of the 3D-HNHC experiment in adenine (Fiala et al. 2004). The 1J(C, N) couplings that are used for the HCN experiment are shown in gray (Fiala et al. 2004). The schematic coherence transfer pathways of the 3D-HNHC pulse sequence are also indicated

Full size image

The 3D-(15N, 13C, 1H)-HNHC experiment (Fig. 2) is a combination of a 13C-HSQC and a 2J-15N-HSQC. It is designed to transfer the magnetization in a double out and back manner from the H2 to the N1(3) and the C2 using single quantum coherences (Fig. 1). Besides the H2–C2–N1(3) correlation, the experiment also provides the H8–C8–N7 correlation in guanine and adenine bases (Fig. 1) and an adjustment of the 15N carrier frequency from 221 to 170 ppm gives also the correlation of the H8–C8–N9 resonances. However, the common HCN experiment (Sklenar et al. 1993) is able to produce the H8–C8–N9 correlation as well. So we decided only to use hard pulses on the 15N channel (Fig. 2), since these pulses are not short enough to excite the N9 resonances as long as the 15N carrier frequency is set to 221 ppm. Therefore, the 15N-spectral-width in the ω1-dimension can be reduced to 24 ppm and the experiment can be run with a higher resolution without further hassle.

Fig. 2
figure 2

Pulse sequence of the 3D-HNHC experiment in the soft Watergate water suppression (Piotto et al. 1992) implementation. Narrow and wide filled bars correspond to rectangular 90° and 180° pulses applied with RF field strengths of 25.5 kHz (1H), 19.5 kHz (13C), 6.95 kHz (15N), respectively. Gradients (G 1–7) are indicated as black filled semi-ellipses. Bipolar gradients (BG1, BG2) (Sklenar 1995) are indicated as unfilled black rectangles. The default pulse phase is x. The pulse sequence was optimized on a Bruker spectrometer with the Bruker typical phase settings (Roehrl et al. 2005). The wide filled bar indicated with cw is an optional presaturation pulse on the water resonance. Fixed delays are adjusted as follows: \( \Updelta = 1.25\,{\text{ms}}\left( {1/\left( {2*^{1} {\text{J}}_{\text{H2C2}} } \right)} \right), \) \( \tau = 16\,{\text{ms}}\left( {1/\left( {4*^{2} {\text{J}}_{\text{H2N1(3)}} } \right)} \right), \) η = 20.4 μs (length of the 180° pulse on the 1H channel) + t 1(0) = 26.4 μs. The proton carrier frequency is centered at the water frequency (4.7 ppm). The carbon carrier frequency is set to 144 ppm (middle between C2 and C8) and the nitrogen carrier frequency is set to 221 ppm (middle between N1, N3 and N7). At 700 MHz, band selective pulses are set as follows: 90° square pulse (small gray semi-ellipse, water flipback) 1.5 ms, 90° square pulse (small black semi-ellipse, soft Watergate water suppression) 0.85 ms. Asynchronous GARP decoupling (Shaka et al. 1985) on the carbon and the nitrogen channel is used to suppress heteronuclear scalar couplings during acquisition. The pulse field gradients G 1–G 7 have sine bell shaped amplitudes. G 1–G 6 are 1 ms and G 7 is 200 μs in length. All gradients are applied along the z-axis and have the following strengths: G 1: 41%, G 2: 7%, G 3: 11%, G 4: 53%, G 5: 23%, G 6: 29%, G 7: 85, 100% of gradient strength corresponds to 55 Gauss/cm. The bipolar pulse field gradients BG1 and BG2 (Sklenar 1995) are applied along the z-axis, have rectangular shaped amplitude and are t 1/2 (BG1) and t 2/2 (BG2) in length. The bipolar gradients have the following strengths: BG1: 2%, BG2: 3%. Phase cycling: φ1 = x; φ2 = x, −x; φ3 = 2(x), 2(−x); φ4 = x; φ5 = 4(x), 4(−x); φ6 = x; φrec = x, −x, x, −x, −x, x, −x, x. φ1 and φ2 are incremented in a States-TPPI (Marion et al. 1989) manner to achieve quadrature detection in the 15N (ω1) dimension and φ4 and φ5 are incremented according to States-TPPI (Marion et al. 1989) to achieve quadrature detection in the 13C (ω2) dimension

Full size image

In the 3D-(15N, 13C, 1H)-HNHC pulse sequence (Fig. 2), the coherence 2H2 z N1(3) z at time point a is created via an INEPT step utilizing the strong 2J(H2, N1(N3)) coupling (15 Hz). Between a and b the coherence is labeled with the 15N chemical shift. In the refocusing INEPT, evolution of the 2J(H2, N1(N3)) coupling is concatenated with the evolution of the 1J(H2, C2) coupling (200 Hz), leading to the coherence 2H2 z C2 z exp(iωN1(3) t 1) at time point c. C2 chemical shift evolution occurs between c and d. Finally the 1J(H2, C2) coupling (200 Hz) is refocused resulting in H2 x exp(iωN1(3) t 1)exp(iωC2 t 2) inphase magnetization in e. Neglecting relaxation effects, optimal sensitivity is achieved by setting the delay τ to 16 ms (1/(4*2JH2N1(3))). However, for larger RNA molecules showing high R2 rates of the H2 protons, the delay τ has to be optimized towards shorter delays in order to obtain optimal sensitivity. The pulse sequence illustrated in Fig. 2 contains a soft Watergate water suppression element (Piotto et al. 1992), which makes the pulse sequence applicable to H2O-samples. Although the pulse sequence can be applied to a D2O sample, we optimized gradients and water flip back pulses on an H2O sample so that the important experiments such as 2D NOESY, HNN-COSY, HNHC and TROSY relayed HCCH-COSY can be performed on the same sample. Additionally to this implementation, an echo/antiecho sensitivity enhanced (Fig. S1) (Kay et al. 1992) and an echo/antiecho TROSY (Fig. S2) version (Nietlispach 2005) of the 3D-(15N, 13C, 1H)-HNHC experiment have been developed. The pulse sequences are part of the supplementary material. We found that the soft Watergate version has a slightly higher sensitivity than the echo/antiecho sensitivity enhanced version. For the 36mer test RNA, the TROSY implementation is less sensitive by a factor of 2, but the signals have smaller line widths especially in the 13C dimension. All the soft pulses, applied on the water frequency (small gray semi-ellipses, Fig. 2), are optimized to reduce the intensity of the water signal. They are applied in a way that the water magnetization is stored along z. Gradients are applied to eliminate artifacts and suppress radiation damping of the water signal during the time periods in which the water magnetization is transverse or along −z. During the chemical shift evolution periods the water magnetization is along −z for time periods of t 1/2 and t 2/2, respectively. Therefore, bipolar gradients (Sklenar 1995) are used to suppress radiation damping during t 1 and t 2.

We also investigated the alternative option to obtain the required correlations by exciting double- and zero-quantum (DQ/ZQ) coherence such as 4H2 z N1(3)+/−C2+/− and 4H2 z N1(3)+/−C2−/+ and evolve sums and differences of the chemical shifts of the N1(3) and C2 nuclei in a concatenated manner. Such an experiment would be shorter for one CH-INEPT transfer period than the SQ version we propose. However, the DQ/ZQ experiment requires post-acquisition processing and a more elaborate phase cycle to separate DQ and ZQ coherences.

The new pulse sequence was applied to a 0.7 mM 13C, 15N adenosine-, cytidine-labeled 36mer RNA oligonucleotide (Fig. 3b) sample in D2O. Water suppression was optimized on a different sample. Slices taken from the spectrum of the 3D-(15N, 13C, 1H)-HNHC experiment are shown in Fig. 3a. Each slice shows two peaks, representing the H2-C2-N1 and H2-C2-N3 correlations. Figure 3c shows a 2J-15N-HSQC recorded under the same conditions as the 3D-(15N, 13C, 1H)-HNHC experiment. A comparison of Fig. 3a and Fig. 3c indicates that all the signals present in the 2J-15N-HSQC are also present in the 3D-(15N, 13C, 1H)-HNHC experiment. Consequently, our data show that the experiment is applicable to mid-size RNA oligonucleotides. However, given the high S/N-ratio of the experiment it should be readily applicable also to larger RNAs. Furthermore, Fig. 3 shows that signals overlapping in the proton dimension, especially A22/A16 and A29/A7, can be resolved in the 3D-(15N, 13C, 1H)-HNHC experiment and the assignment ambiguities can be removed successfully.

Fig. 3
figure 3

a Spectrum of the 3D-HNHC experiment in the strip plot representation with the 13C (ω2) dimension orthogonal to the paper plane. The spectrum was recorded at T = 20°C for 41 h on a Bruker 700 MHz spectrometer equipped with a 5 mM TXI cryogenic probe. Spectra were recorded with 88 (15N) * 148 (13C) increments, each with 8 scans. Spectral widths in the three dimensions were set to 24 ppm (15N), 18 ppm (13C) and 16 ppm (1H). A selectively adenosine, cytidine 13C, 15N-labeled sample with a concentration of 0.7 mM in a D2O buffer was used. b Secondary structure of the 36mer RNA hairpin 2 derived from the transcribed 5′-UTR (untranslated region) of the Salmonella agsA gene (Waldminghaus et al. 2007). c Section of a 2D-2J-15N-HSQC spectrum from the RNA molecule depicted in (b) recorded for 1.8 h under the same conditions than the 3D-HNHC experiment. 512 increments were recorded with 8 scans each. Spectral widths were set to 140 ppm (15N) and 16 ppm (1H). The 15N and 13C carrier frequencies were set to 210 ppm (15N), 144 ppm (13C) and 4.7 ppm (1H). GARP decoupling (Shaka et al. 1985) was applied to the carbon channel during acquisition

Full size image

In conclusion, we have introduced a novel method to correlate the H2-C2-N1 and H2-C2-N3 resonances of the adenine nucleobases in 13C, 15N labeled oligonucleotides. The new method simplifies the assignment of the adenine nucleobases since it reduces assignment ambiguities occurring especially in larger RNA molecules due to signal overlap. By correlating the H2-N1 resonances with the C2 resonances, the new method provides the missing link for correlating the imino H3 resonances of the uracil nucleobases across the AU base pairs to the H2 resonances and subsequently to the H8 resonances of the adenine nucleobases.

Supplementary material

The pulse schemes for an echo/antiecho sensitivity enhanced version and an echo/antiecho TROSY version of the 3D-(15N, 13C, 1H)-HNHC experiment are part of the supplementary material.