NMR on ¹³C/¹⁵N-labeled RNA

In this chapter, we proceed with the experiments that can be performed on RNA samples that include ¹³C-labeling (either by selective incorporation, uniform ¹³C-labeling and/or double (¹³C¹⁵N) labeling). The typical carbon chemical shift ranges have been presented in the introducing Chapter 1 (Figure 4). The ribose carbon signals (C1’-C5’) generally appear between 60 to 95 ppm and the aromatic nucleobase carbons can be observed from 95 to 170 ppm. The ¹³C chemical shift data harbors valuable structural information as these nuclei are sensitive local probes that can distinguish between base-pairing, stacking interactions and expose the ribose conformation (Ebrahimi et al.; Cherepanov et al.).

Another advantage of ¹³C-labeling is that the associated couplings in the nucleobases (Figure 1) can be exploited for the transfer of magnetization by using advanced NMR experiments that correlate these spin-systems, providing additional structural parameters and assignments. For the ribose moiety, the ¹³C-couplings are more complex (Figure 2) and can provide structural information on the sugar puckering, the glycosidic bond angle and backbone angles (Wijmenga and van Buuren; Fürtig et al.). Overall, ¹³C-labeling provides additional structural information, enables through-bond correlations for assignments, effectively reduces signal overlap, and allows obtaining essential relaxation parameters (dynamics) and RDCs not only for the nucleobase but also for the ribose moiety. However, the size of the RNA has an impact on whether these experiments can yield sufficient signal-to-noise.

For small size RNAs (< ~30 nucleotides) the H1’-H6/H8 sequential walk in the 2D ¹H¹H NOESY (using a unlabeled sample) can usually be accomplished with little peak overlap. If the peak overlap is, however, still complicating the assignment, an additional 3D NOESY-HSQC or 3D NOESY-HMQC (using a ¹³C-labeled sample) may resolve the ambiguities. Moreover, as an alternative to the NOESY based sequential walk, the HCP and HCP-TOCSY experiments allow for a sequential assignment incorporating the ³¹P resonances of the backbone. Unfortunately, low peak dispersion in helical regions in the ³¹P dimension limit this technique to smaller sized RNAs. To validate the assignment of H6/8 and H1’, the 3D HCN and CNC experiments can be used, which correlate the base H6/H8 (of the pyrimidine/purine) to the ribose H1’ of the same nucleotide over the glycosidic N1/N9. Furthermore, the 3D HCCH-TOCSY experiment can complement the assignment of all ribose protons (H1’, H2’, H3’, H4’ and H5’#’).

With increasing RNA size (~ 30-70 nucleotides) the sequential walk generally becomes less feasible. Specific experiments that require ¹³C¹⁵N-labeled samples, are needed to support the assignment. These experiments typically do not rely on NOESY-based experiments, but on through-bond magnetization transfers. The before mentioned 3D HCN, CNC and HCCH-TOCSY experiments are also applicable to RNAs of larger size and support the assignment of the ribose-protons. Another very useful experiment is the 2D or 3D HCCNH-TOCSY, which links the U:C6 with the U:N3 of the same nucleotide, as well as the G:C8 with the G:N1. With the TROSY-HCCH-COSY the A:H2 can be correlated to the A:H8, which can be helpful since the A:H2 is linked to its base-paired U:H3 by a strong cross peak in the NOESY spectrum. Furthermore, for larger RNAs, the 3D NOESY-HSQC, 3D NOESY-SFHMQC and 4D SOFAST-NOESY-HCCH become valuable as they better resolve the cross-peaks in the crowded NOESY spectra.

For even larger RNAs (> 70 nucleotides) the assignment often becomes very challenging due to increasing peak overlap. Here, selectively labeled samples, where either one, two or three types of nucleotides are labeled can be helpful. To reduce the amount of signals in the NOESY spectra, ¹³C-filtered or ¹³C-edited NOESY experiments can be recorded. The same effect obviously helps with every experiment that requires ¹³C¹⁵N-labeling, effectively removing all signals of unlabeled nucleotides (given that the sample allows for a sufficient signal to noise ratio). If the RNA is produced via solid-phase-synthesis, segmental labeling is also an option for the regions of interest. The shorter transverse relaxation times of larger RNAs prevent the efficient usage of experiments that feature TOCSY magnetization transfer. In such cases, a divide and conquer approach, where the RNA is separated in smaller sub-constructs, might help, as long as the natural folded conformation can be preserved. This can be verified by overlaying the corresponding HSQC and NOESY spectra of the shortened and full length constructs.

Proper referencing of RNA ¹³C chemical shifts is critical for reliable assignment and comparison of NMR data (Aeschbacher et al.). The IUPAC defines tetramethylsilane (TMS) as the universal chemical-shift reference standard for NMR spectroscopy (Harris et al.). However, for biomolecular NMR studies in aqueous solution, the water-soluble compound DSS (2,2-dimethylsilapentane-5-sulfonic acid) is recommended as a practical reference (Markley et al.). Because TMS and DSS employ slightly different ¹³C/¹H frequency ratios, the choice of reference convention results in a systematic offset of approximately 2.66 ppm in the ¹³C chemical shifts. Bruker TopSpin uses the TMS ratio by default, whereas the Biological Magnetic Resonance Data Bank (BMRB) advocates referencing according to the DSS convention. Consequently, chemical shift discrepancies observed in publications and the BMRB often reflect differences in referencing conventions rather than genuine variations. In TopSpin, the offset between DSS and the default TMS referencing can be corrected by manually adjusting the spectrum reference frequency (SR value) or automatically using BioTop (btproc biorefonly).

Experiment Setups for ¹³C/¹⁵N-labeled RNA:

Measurement of the ¹H¹³C HSQC requires optimized parameters for the specific correlations of interest, as the RNA is characterized by distinct moieties (e.g. ribose or nucleobase) with different J-couplings, chemical shifts, relaxation behavior and proximity to the interfering solvent signal. One spectrum covering the whole range may not sufficiently resolve and efficiently show all the cross-peaks. Moreover, in order to observe signals close to the solvent (water), additional suppression techniques or a sample in D₂O may be required. In addition to acquiring the whole ¹³C spectrum (HSQC-ALL), selective pulse programs and parameter sets can be used to acquire spectra for the ribose (HSQC-C1’, HSQC-C5’’ and HSQC-ribose) or aromatic nucleobase regions (HSQC-C2C6C8 and HSQC-C5) (Figure 3).

The carbonyl groups of G, C, and U frequently serve as hydrogen-bond acceptors in structured RNA regions, especially in WC-type base-pairs. The ¹³C chemical shift of the carbonyl carbon is sensitive to hydrogen bonding and thus provides useful information regarding the type of base-pair formed. For example, in canonical AU base-pairs, the U:O4 is involved in a hydrogen-bond, whereas the U:O2 is not. In contrast, in GU base-pairs, the U:O2 is usually hydrogen-bonded to G:H1 while the U:O4 is free. Thus, the U:C2 carbon resonance is shifted to higher ppm values in GU base-pairs with respect to AU base-pairs and the U:C4 is shifted to lower ppm values. In G residues, the O6 is a hydrogen-bond acceptor in both GC and GU base-pairs, and C2 is attached to the N2-H2# amino group serving as a hydrogen-bond donor in GC base-pairs. The HNCO experiment (Figure 4) provides a direct readout of the U:C2/C4 and G:C2/C6 chemical shifts by correlating the respective imino proton via 2 INEPT transfer steps first to the imino nitrogen and subsequently to both carbons attached to it: G:H1 » N1 » C2/C6 and U:H3 » N3 » C2/C4.

The H5C5C4 experiment provides an alternative for the HNCO, especially for U residues lacking an observable imino proton, and for C residues. For both pyrimidine nucleotides, the C4 chemical shifts are detected on the non-exchangeable H5 protons (Figure 5), which in turn are usually well resolved in the C5H5 HSQC and in the ¹H¹H TOCSY.

Transfer of the imino proton assignment to G:H8 and U:H6 aromatic protons is achieved by the HCCNH-TOCSY experiment, in which magnetization from the aromatic protons is transferred to the imino protons via scalar H-C, C-C and C-N couplings. The experiment can be recorded as a 2D (HC or HH) or 3D (HNC or HNH) spectrum (Sklenár et al.). The 2D HC version typically yields better spectra compared to the 2D HH version because of its higher magnetization transfer efficiency. Spectra of imino protons are usually recorded at lower temperatures due to slower solvent exchange, whereas ¹³C line shapes are usually sharper at higher temperatures. Therefore, temperature optimization is important and in many cases, it may be necessary to record the HCCNH-TOCSY at two or even more temperatures.

Because imino protons are only present in guanosine and uridine, information obtained from the HCCNH-TOCSY is restricted to these two nucleotides. Correlations are observed between G:C8-H8 and U:C6-H6 in the indirect H or C dimension and G:H1 and U:H3 in the direct dimension. If the experiment is recorded as a 3D, U:N3 and G:N1 resonances are observed in the additional ¹⁵N dimension. Using the ¹³C chemical shift from the HCCNH-TOCSY, the aromatic proton chemical shifts can be assigned in an aromatic ¹H¹³C HSQC. Thereby, assignments for base-paired nucleotides can be transferred from imino to aromatic protons.

Usually, C:H5 and C:H6 resonances of base-paired cytidines are well accessible by complete assignment of 2D ¹H¹H NOESY spectra of unlabeled RNA via G:H1-C:H41/H42-C:H5-C:H6 correlations and confirmation of H5-H6 connectivities in the 2D ¹H¹H TOCSY. Alternatively, the Py H(CC)NN-COSY directly correlates U/C:H5 and U/C:N3 and A/G:N1 resonances for base-paired pyrimidines.

Regarding referencing of the HCCNH-TOCSY spectra it is important to note that there is a TROSY shift, which differs for G and U, because of the different H8-C8 (216 Hz) and H6-C6 (185 Hz) coupling. Generally, the ¹H dimension needs to be corrected by -47 Hz and the ¹³C dimension by 108 Hz for G and 92.5 Hz for U. Therefore, two differently referenced spectra are necessary for assignment of G and U aromatic carbon resonances. In Figure 6 the HCCNH-TOCSY, correctly referenced for U, and aromatic ¹H¹³C HSQC are shown for the 24mer RNA and assignment transfer from imino protons to aromatic carbons is indicated by lines.

RNA imino-protons are usually broadened beyond detectability, when they are not sterically protected or part of a base-pair. Furthermore, RNA amino protons, especially those of guanosine and adenosine, are often exchange-broadened due to the rotation around the C-N bond. To circumvent this phenomenon, heteronuclear ¹³C-detection can be employed and the imino-proton exchange with the solvent can be determined.

Information about the solvent exchange rate of all imino-protons, also those, which exchange fast and are broadened in 1D and 2D experiments, can be obtained by recording a CN-HSQC with an additional ¹H¹⁵N spin echo element. First, magnetization is transferred from carbon to nitrogen during an INEPT element. Next, a ¹H¹⁵N spin echo element, consisting of a delay 2Δ and centered 180° pulses on nitrogen and proton frequencies, is applied. During this spin echo element, scalar ¹H¹⁵N coupling evolves, while the chemical shift on nitrogen is refocused. The spin-echo element modulates signal intensity by an additional cos(π J_N,H 2Δ) term depending on the delay time 2Δ and the imino-proton exchange rate k_ex, if relaxation is neglected. In the limit of slow imino-proton solvent exchange (k_ex ~ 0) and with a delay time of 2Δ = 1 / 2J_N,H the cosine term results in cos(π) = -1. Therefore, the signal intensity of exchange-protected nucleotides is modulated with a negative sign during the spin echo element.

In the limit of fast imino-proton exchange (k_ex >> J_N,H) many exchange events occur during Δ. This prevents the evolution of scalar coupling and results in self-decoupling. The modulation by the cosine term is removed and the signal intensity is not altered by the spin-echo element. Therefore, nitrogens with fast-exchanging imino protons behave similarly to non-protonated nitrogens.

Between these two extreme cases, modulation of signal intensity is dampened as exchange rates increase. At an exchange rate of k_ex = 360s^-1 the additional modulation by the cosine term equals zero, resulting in the loss of the signal (Fürtig et al.).

Typically, the experiment is recorded for at least two different delay times. In the reference experiment recorded with minimal delay time all signals (U:C2-N1, U:C2-N3, G:C2-N1, C:C2-N1) are observed with positive signal intensity. (Figure 7 A) The delay time equals 2Δ = 1 / 2J_N,H = 5.56 ms, considering the imino proton J_N,H coupling of 94 Hz. With this delay, the signal intensity of U:C2-N3 and G:C2-N1, but not U:C2-N1 and C:C2-N1 is modulated depending on the imino proton exchange rate. (Figure 7 B). All base-paired nucleotides are annotated in purple rectangles, all unpaired nucleotides in green rectangles. For uridine resonances the difference between exchange-protected base-paired nucleotides and unpaired nucleotides is most striking. For the two base-paired nucleotides, the U:C2-N3 signal intensity is negative, whereas all unpaired nucleotides still show positive U:C2-N3 signal intensity.

Similar to the lr-HNN-COSY described for ¹⁵N-labeled RNA, the Py H(CC)NN-COSY experiment (Figure 8) exploits detection of non-exchangeable protons for the correlation of imino donor and acceptor nitrogens in more labile base-pairs (Dallmann et al). Here, the C/U:H5 proton of U and C is correlated via three INEPT transfers to C/U:C5, C/U:C4 and C/U:N3, yielding the diagonal peak (C/U:N3H5). In case of base-pairing, the ²J_NN-COSY transfer step yields the corresponding C/U:H5-G/A:N1 cross-peak. Since the C/U:H5 resonances are close to the water (~4.5-6.5 ppm), this experiment should be recorded with D₂O samples. Another issue is the severe sensitivity loss for larger RNAs (>30 nts) due to the long duration of the pulse sequence.

With increasing RNA size, the increasingly severe signal overlap in classical 2D ¹H¹H-NOESY spectra becomes a challenge for unambiguous assignment and for high-confidence distance calibration from NOE intensities. To increase resolution, an additional ¹³C dimension can be recorded, by using HSQC or SFHMQC pulse program blocks after the NOESY transfer (Wijmenga et al.). Thus, NOE cross-peaks are encoded with the individual ¹³C chemical shift of one of their respective diagonal signals. This allows the selective recording of only certain NOE peaks, that are associated with a certain ¹³C chemical shift frequency. This association is either direct via J-coupling or indirect via NOE. By iterating through the ¹³C dimension of the 3D, the NOESY is displayed as slices of 2D planes, that contain only certain NOESY peaks, thereby reducing the overlap of peaks (Figure 9). If the ¹³C as well as the ¹H shifts of two peaks are similar, this separation obviously fails. In this case a 4D NOESY experiment can be considered.

Similar to the 3D ¹H¹H¹⁵N NOESY-HMQC, signal overlap is thus resolved at the cost of sensitivity due to the more complex pulse sequence, which results in relaxation based signal loss. Due to the additional dimension, the required measurement time is long and a ¹³C-labeled sample is obligatory for the NOESY-SFHMQC or NOESY-HSQC.

While the SFHMQC version tends to yield better signal to noise ratios for the aromatic peaks, with a reduced measurement time due to the SOFAST approach (Rossi et al.), the HSQC version offers improved water suppression using the WATERGATE scheme. This facilitates the detection of the ribose peaks around 6 - 5.3 ppm (Figure 10).

If the intention of the measurement is to exploit NOE intensities for distance determination, any acquisition (or processing techniques) that modulate signal intensity in favor of reduced acquisition time (e.g.: BEST, SOFAST, non-uniform sampling, NUS) should be avoided. Therefore the HSQC version is the method of choice in this case, even though the SFHMQC version might yield more peaks.

One severe bottleneck of RNA assignment is the signal overlap in 2D NOESY spectra, which frequently hinders unambiguous resonance assignment. An easy way to reduce spectral crowding is to suppress the signals of certain nucleotide types. For the sequential assignment, selective ¹³C labeling of either one or a combination of two or three nucleotides enables the application of ¹³C-filtered and ¹³C-edited NOESYs in order to distinguish intra- and internucleotide NOEs, identify resonances belonging to a certain nucleotide type, and to unambiguously assign NOEs in crowded spectral regions. As the possible combinations of selective nucleotide labeling and application of isotope-filtering/-editing in each dimension quickly multiply to a confusing number of possible strategies, careful sample and experiment design requires a certain degree of pre-knowledge of the system. The following rules of thumb are, however, generally applicable to selective labeling of RNA: The C/U:C6-H6 cross-peaks are the least well separated aromatic resonances, so unambiguous identification of resonances is achieved by selective labeling of either U or C; C/U:H5-H6 cross-peaks are among the most intense NOE signals in the spectral region important for the “sequential walk”, making them attractive candidates for isotope-filtering – at the cost of also getting rid of the respective H1’-signals; G provides the least number of aromatic CH signals in HSQCs (A: C8-H8 and C2-H2; C/U: C6-H6 and C5-H5; G: only C8-H8), making it the least useful ¹³C-probe in HSQC-based experiments; purines and pyrimidines are easiest to distinguish, so a combination of selective A/U or A/C labeling is usually most helpful for RNAs with a fairly even nucleotide type distribution. Similarly, the combinations of isotope-filtering and editing in the 2D NOESY experiments are not equally useful: In general, ¹²C¹H cross-peaks are superior in terms of linewidths and relaxation behavior, resulting in both better resolution and intensity compared to ¹³C¹H cross-peaks; filtering or editing in the F1 dimension typically results in significant sensitivity loss and thus requires high sample concentrations and sufficient scans; the F2 dimension always provides better resolution than the F1 dimension, but it is also the dimension where effective solvent suppression is crucial. While editing steps usually remove all signals from unlabeled nucleotides, filtering steps typically only reduce the signals of labeled nucleotides to 1-5%. Here, we describe and show (Figure 11) a ¹³C¹⁵N(F1)-filtered and ¹³C¹⁵N(F2)-edited 2D NOESY experiment. In some cases, additional filter options and combinations might be helpful. In the NOESYGPPHXF19 parameterset four combinations of ¹³C-filtered and -edited spectra are recorded: ¹²C(F2)-¹²C(F1), ¹³C(F2)-¹²C(F1), ¹²C(F2)-¹³C(F1), ¹³C(F2)-¹³C(F1). In selectively labeled samples for example, NOEs are observed between unlabeled-unlabeled, labeled-unlabeled, unlabeled-labeled and labeled-labeled residues in the four different spectra.

The combination of 3D NOESY-HMQC element with an isotope-filter in F1 provides distance information exclusively between ¹²C-attached protons and ¹³C-attached protons with high resolution of the ¹³C-labeled species via ¹³C chemical shift encoding in F2. Although also applicable to selectively labeled RNA, this experiment is especially valuable for detection and assignment of intermolecular distances, where the RNA is ¹³C¹⁵N-labeled and the interaction partner is at natural abundance. A ¹³C/¹⁵N filter is applied prior to chemical shift evolution in F1, followed by the NOESY mixing time and ¹³C chemical shift evolution (F2), and finally acquisition (F3) with ¹³C- and ¹⁵N-decoupling. Due to the high similarity of the F1-filtered NOESY-SFHMQC to a normal NOESY-HMQC, please see the descriptions of 3D ¹H¹H¹³C NOESY-HSQC/SFHMQC and X-filtered NOESY for the analysis of the F1-filtered NOESY-SFHMQC.

The TROSY-HCCH-COSY experiments correlate H2 and H8 resonances within the nucleobase in adenosine. Using a COSY building block magnetization is transferred from H2 (F3) via C2 (F2) and C5 to C4 & C6 (F1) in an out-and-back manner. Simultaneously, magnetization is transferred from H8 (F3) via C8 (F2) and C4/C6 to C5 (F1). As a result, both H2 and H8 show correlation to C4, C5 and C6 within the nucleobase (Simon et al.).

Typically, analysis of the TROSY-HCCH-COSY (Figure 12) and thereby assignment of adenosine nucleobase resonances is achieved by looking at the HC (F3-F1) planes. A certain C2 or C8 resonance is selected in F2 and A:H2/H8-C2/C8, A:H2/H8-C4, A:H2/H8-C5 and A:H2/H8-C6 resonances are assigned. Because H2 and H8 are not directly correlated in the TROSY-HCCH-COSY, unambiguous assignment is achieved when C4, C5 and C6 chemical shifts are identical for a H2:C2 and a H8:C8 resonance. The transfer of assignment from A:H2 to A:H8 is another valuable piece of information when progressing from sequential imino-proton assignment to nucleobase and ribose assignment. In helical RNA regions there is an intense cross-peak in the ¹H¹H NOESY between U:H3 and A:H2 within a base-pair. The TROSY-HCCH-COSY then enables unambiguous A:H8 assignment, which subsequently simplifies the sequential walk between H1’-H6/H8.

Sequential assignment along the RNA sequence depends on the identification and distinction of H1’-H6/H8 correlations within a nucleotide and between consecutive nucleotides. The 3D HCN (Figure 13) simplifies the intranucleotide assignment of H1’ and H6/H8 resonances.

The pulse sequence comprises two INEPT steps: 1.) transfer from H to C through ¹J_CH scalar coupling; 2.) transfer from C to N through ¹J_CN scalar coupling. In the first step, transfer is achieved simultaneously from H1’ to C1’ in the ribose and from H6/H8 to C6/C8 in the nucleobase. Subsequently, magnetization is transferred simultaneously from C1’ to N1/N9 and from C6/C8 to N1/N9 in pyrimidines/purines (Sklenář et al.).

For the assignment, the ¹H¹³C planes are analyzed. Purines and pyrimidines are separated in the ¹⁵N (F1) dimension due to the difference in the C/U:N1 (~150 ppm) and G/A:N9 (~ 170 ppm) ¹⁵N chemical shifts. In the ¹H¹³C plane, two peaks are observed for each nucleotide, one originating from the H1’-C1’ and one from the H6/H8-C6/C8. Due to the significant difference in the ¹³C chemical shifts of C1’ (~90 ppm) and C6/C8 (~ 140 ppm) selective pulses are required for ¹³C excitation. Furthermore, the offset on ¹³C (o2p) must be centered between the two resonance ranges at ~115 ppm. As a compromise between limited measurement time and sufficient resolution, the ¹³C spectral width is reduced to ~ 20 ppm. This results in a spectrum in which the H1’-C1’ and H6/H8-C6/C8 signals are folded. Corrected chemical shift values can be determined by adding (C6/C8) or subtracting (C1’) the spectral width to or from the chemical shift values. Subsequently, the H6/H8-C6/C8-N1/N9 and H1’-C1’-N1/N9 signals can be assigned and connected via the N1/N9 resonance within one nucleotide.

The 3D CNC (Figure 14) provides similar information as the 3D HCN, resulting in the intranucleotide connection of ribose and nucleobase resonances. In the CNC experiment magnetization is evolved on C/U:C6 & G/A:C8 in the F1 dimension and subsequently correlated with C/U:N1 & G/A:N9 in the F2 dimension and C1’ in the F3 dimension. Therefore, the 3D CNC only gives rise to a single peak per nucleotide, in contrast to two peaks in the 3D HCN. If there are multiple nucleotides with similar N1/N9 ¹⁵N chemical shift, it can be challenging to unambiguously assign the H6/H8-C6/C8-N1/N9 and H1’-C1’-N1/N9 pairs that belong to the same nucleotide based on the HCN experiment. These ambiguities can often be resolved by recording a 3D CNC experiment. Using the combination of HCN & CNC, H1’/C1’-H6/H8/C6/C8 pairs can be identified and the assignments transferred to NOESY spectra to ease sequential assignment. If an unambiguous assignment is not possible due to significant overlap within purines and pyrimidines, selective isotope‑labeling might be helpful to complete the assignment.

The ¹H¹³C¹³C/¹H HCCH-TOCSY (Figure 15) experiments (H(C)CH- and (H)CCH-type) are the most valuable for ribose assignment in RNA. All ribose protons are excited and the ¹³C or ¹H chemical shift is detected in F1. Subsequently, magnetization is transferred along the ribose carbons using a CC-TOCSY step. Following the TOCSY transfer, the H1′‑C1′ resonances are recorded in F2 and F3. This results in spectra of H1′–C1′–Hx′/Cx′, where x represents all the protons and carbons in the ribose. It is recommended that a total of four experiments are recorded. These include two different TOCSY transfer times, as well as ¹³C and ¹H detection in the F1 dimension. The combination of ¹³C and ¹H detection in the F1 dimension enables assignment, even when ribose resonances overlap in either the ¹³C or ¹H dimension. As the H2’-C2’ and H3’‑C3’ signals are observed at similar chemical shift values, differentiating between them is challenging. To address this, TOCSY spectra are recorded with short (4.7 ms) and long (14.1 ms) mixing times. With the shorter mixing time, magnetization transfer to C1’ is efficient only for C2’, but not for the other ribose carbons, C3’, C4’ and C5’. Only the H1’-C1’-H2’/C2’ peaks are observed with high intensity, whereas the H1’-C1’-H3’/C3’ peaks may exhibit low or no intensity, and the other peaks are usually not observed. During the longer mixing time, efficient transfer is achieved from all ribose carbons to C1'. Consequently, all H1’-C1’-Hx’/Cx’ peaks are observed with similar intensity.

If severe peak overlap in a 3D NOESY-HSQC/HMQC prohibits unambiguous assignment of CH resonances, a 4D NOESY experiment is available for additional resonance dispersion (Stanek et al.). This experiment should, however, not be used for quantitative distance determination from NOEs, as the acquisition requires non-uniform sampling (NUS). It is however helpful in completing the assignment.

The pulse sequence consists of an HMQC transfer of the aromatic H6/H8 signals followed by a NOESY step, where magnetization is transferred to neighboring protons, including the H1’. An additional HMQC transfer allows the ¹J based correlation to the C1’ resonance. The additional ¹³C dimension allows a further selective step based of the ¹H and ¹³C shifts of the aromatic C6-H6 and C8-H8. During analysis it is possible to only display the C1’-H1’-2D-plane (F3-F4) featuring only two residues, based on the ¹³C and ¹H shift of the respective aromatic C6-H6 or C8-H8 to which two of the H1’ protons are connected via the NOE.

It is therefore possible to adjust the H6/8 (F1) and C6/8 (F2) axes of the 4D to sequentially go through all H1’ protons. Figure 16 shows an exemplary walk over five residues for the C1’-H1’ (F3-F4 plane), with a 2D ¹H¹³C HSQC as comparison for the location of the peaks.

Residue 1 typically displays only one peak in its C1’-H1’ plane as there are no further nucleotides in the 5’ direction. While every C1’-H1 plane associated with the other residues displays its own H1’ and the one of its 5’ oriented neighbor. The experiment is not optimized for analysis of the C6/8-H6/8 plane, which features more artifacts than the C1’-H1’ plane.

The 3D triple-resonance proton-carbon-phosphorus (HCP, Figure 17) experiments allow for straightforward correlation of the ribose signals on both sides of the backbone phosphate group. These HCP experiments are robust and especially suitable for the sequential assignment of smaller RNAs (limited by the ribose signal overlap and severe line broadening of ³¹P resonances for RNAs larger than ~30 nt). The correlations observed in the HCP experiment are from the ³¹P phosphate to the ribose C3’_i-H3’_i, C4’_i-H4’_i on one side and the ribose C5’_i+1-H5’#_i+1, C4’_i+1-H4’₊₁ on the other side. The favorable CP coupling constants (e.g. 3-5 Hz to C5’ and C3’ and 8-10 Hz to C4’) usually allow the detection of all cross-peaks.

Moreover, the 3D HCP-TOCSY (Figure 18) triple-resonance experiment extends the 3D HCP experiment with a CC-TOCSY sequence, in which the magnetization is further relayed to the neighboring ribose nuclei through their ¹J(C,C) couplings (typically around 40-45 Hz). In this way, the relevant correlations are resolved via the well-dispersed C1′-H1′ signals. The HCP experiments can thus be used to confirm and complement the assignment. Moreover, it allows one to resolve the resonances in the non-helical RNA regions, which are often infeasible to obtain by other assignment experiments. When the spectral resolution of the ribose protons and phosphorus resonances is sufficient, a 2D ³¹P(¹³C)¹H version of the experiments can be used.

Besides the experiments that were described before to probe the base-pair motions by the imino groups (requires ¹⁵N-labeled RNA), the dynamics of ribose and nucleobase moieties can also be investigated using ¹³C-reporter signals. Here, the ¹³C-T1 and ¹³C-T1rho relaxation rates and the ¹³C{¹H}-hetNOE values provide the information on their dynamics on the ps-ns timescale. Again, this data can be analyzed using the model-free approach to yield parameters that describe the order, internal motion and exchange contributions. Moreover, relaxation dispersion NMR experiments can be used to characterize the RNA dynamics that occur on the microsecond-to-millisecond timescale, allowing the detection of low-populated and otherwise ‘invisible’ and excited states.

The ¹³C{¹H}-hetNOE experiment for ¹³C1’ and aromatic carbons provides information on the fast dynamics (ps-ns timescale) by measuring the ¹H¹³C dipole-dipole interactions and is a key experiment to probe RNA flexibility. The pseudo 3D experiment consists of two 2D ¹H¹³C HSQC-type spectra acquired in an interleaved fashion: one in the presence of proton saturation and one reference spectrum in the absence of saturation. The ¹³C{¹H}-hetNOE value is defined as the ratio of the signal intensities (or volumes): hetNOE = I_saf/I_ref, which typically ranges from 1.0 to 1.2 for rigid regions, while (opposite to the ¹⁵N{¹H}-HetNOE experiment) higher values indicate significant internal mobility.

The ¹³C-T1 experiment measures the spin-lattice (longitudinal) relaxation time. This pseudo 3D experiment involves a series of 2D ¹H¹³C HSQC-type spectra with varying relaxation delay times. Typically, 8 or more spectra (with delays ranging from 100 ms to 1-2 s) are recorded where the intensity (or integral) of the signals decrease exponentially over time (I_t = I₀ × e^-t/T1). The ¹³C-T1 relaxation time is obtained by fitting of the data to the exponential decay function. Generally, a mono-exponential behavior is observed, but in case of strong anisotropic motions non-exponential signal decay behavior can occur. The obtained relaxation times reflect on the fast internal motions which occur on the ps-ns timescale. Here, a shorter T1 time corresponds to a higher relaxation rate (R1 = 1/T1) and indicates higher mobility.

Directly measuring ¹³C-T2 in RNA is notoriously difficult due to severe signal line-broadening caused by large scalar ¹J(C,C) couplings. The unfavorable relaxation pathways can be suppressed by locking the magnetization along the transverse plane. Hence, the spin-spin (transverse) relaxation time is usually measured via the ¹³C-T1rho experiment, where a spin-lock is applied in a rotating frame either on- or off-resonance. This pseudo 3D ¹³C-T1rho experiment also involves a series of 8 or more 2D ¹H¹³C HSQC-alike spectra (or recorded as a pseudo 2D ¹³C-T1rho with 1D alike spectra) with varying relaxation delay times (usually ranging from around 10 to 100-200 ms). Likewise, the ¹³C-T1rho times are obtained by fitting of the data to a mono-exponential decay. The T1rho time contains spin-lock (power and offset) dependent contributions of T1 and T2. Unless an on-resonance method is used, deriving T2 from T1rho thus requires corrections for the effective magnetic field angle using the off-resonance position of the spin-lock, spin-lock field strength, and T1. The T2 time is sensitive to both fast (<ns) and slow (us-ms) motions (exchange) and is reduced by a larger molecular size (slower global tumbling). A longer T2 time corresponds to a lower R2 rate and indicates higher mobility. Moreover, the T1/T2 ratios depend on the reorientational motion of the CH bond and thereby provide an estimate on the (local) rotational correlation times (τ_c). Increased T1/T2 ratios indicate longer τ_c values, which correspond to a slower tumbling and thereby a more rigid RNA region.

Analysis to obtain the rates, values and other dynamic parameters for the hetNOE, T1 and T1rho can be conveniently performed using the Bruker Dynamics Center.

Relaxation-dispersion data can be acquired by measuring ¹³C-T1rho-rates at different spin-lock strengths and offsets. The complete workflow exceeds the scope of this tutorial, but is well documented in Hansen et al. 2007, Hansen et al. 2009 and Xue et al. Basically, the spin-lock effectively suppresses (conformational) exchange effects and its contributions and can be determined by fitting the T1rho values measured at different spin-lock strengths and offsets (Xue et al.). The analysis of the obtainable kinetic parameters requires more advanced equations like Bloch-McConnell (McConnell) as discussed in literature (Xue et al. Rangadurai et al.).

The principles, analysis and the preparation of samples for the measurement of residual dipolar couplings (RDCs) have already been described in the ¹⁵N-labeled part of the tutorial: here we only summarize the most important points (Tjandra et al., Hansen et al.):

• prepare two samples, one with and one without alignment medium
• measure both samples at the same spectrometer
• verify the alignment with D₂O splitting in a 1D ²H spectrum
• measure inphase-antiphase (IPAP) spectra to reduce peak overlap due to coupling based peak splitting
• precisely determine the peak center of the signals by processing with at least 16 k points in F2 and 4 k points in F1
• analyse the RDCs by fitting to a structure using third party software (such as Pales)

The advantage of measuring ¹H¹³C based RDCs in addition to ¹H¹⁵N RDCs is that more signals, also in dynamic regions of the RNA, can be investigated. The imino ¹H¹⁵N RDC are limited to the base-paired regions of the RNA. This can be helpful in structure validation when bulges and junctions cause ambiguous orientations of these RNA regions. However, the orientation and behavior of the nucleobases in the dynamic regions can be determined using ¹H¹³C RDCs. By obtaining the A:C2-H2, C5-H5, C6-H6, C8-H8 and C1’-H1 couplings, also the residues in RNA bulges, junctions or loop regions can be investigated. Yet, one needs to keep in mind that the resulting RDC values may represent population-weighted averages of multiple conformational states that are in fast (ps-µs) exchange.

Due to different decoupling in the ¹³C dimension (F1), separate experiments should be acquired for the specific spectral regions (Figure 20): C1’H1, C5H5 and C2H2/C6H6/C8H8. The regions of the other ribose signals are generally not so well dispersed and due to signal overlap practically not suitable for RDC determination.

< Back to Overview