Probing excited conformational states of nucleic acids by nitrogen CEST NMR spectroscopy

Abstract

Characterizing low-populated and short-lived excited conformational states has become increasingly important for understanding mechanisms of RNA function. Interconversion between RNA ground and excited conformational states often involves base pairing rearrangements that lead to changes in the hydrogen-bond network. Here, we present two ¹⁵N chemical exchange saturation transfer (CEST) NMR experiments that utilize protonated and non-protonated nitrogens, which are key hydrogen-bond donors and acceptors, for characterizing excited conformational states in RNA. We demonstrated these approaches on the B. Cereus fluoride riboswitch, where ¹⁵N CEST profiles complement ¹³C CEST profiles in depicting a potential pathway for ligand-dependent allosteric regulation of the excited conformational state of the fluoride riboswitch.

1. Introduction

Many non-coding RNA functions depend on their intrinsic conformational flexibility to dynamically interconvert between different structures upon recognition of specific cellular cues [1,2]. As a powerful tool for characterizing biomolecular structures and dynamics, NMR has played critical roles in uncovering RNA conformational dynamics, leading towards a deeper mechanistic understanding of RNA functions [3-6]. In particular, recent developments in nucleic-acid NMR, including conventional R_1p relaxation dispersion (RD) [7-9], Carr-Purcell-Meiboom-Gill (CPMG) RD [10-15], low spin-lock field R_1p RD [16-23], and chemical exchange saturation transfer (CEST) spectroscopy [24-28], have opened avenues for characterizing RNA excited states (ES) that are often too sparsely populated and transiently lived to be detected by conventional structural biology approaches. These NMR RD methods have enabled discovery and identification of excited states across diverse non-coding RNAs, reinforcing an emerging view of RNA excited states as a ‘hidden’ layer of regulation [5,6].

Dynamic transition to an RNA excited state often involves formation and/or rearrangement of canonical and non-canonical base pairs at tertiary and secondary structural levels, which are mediated by various distinct hydrogen-bond interactions. The current nucleic-acid NMR RD techniques have provided a variety of probes for characterizing RNA excited states, including protons [9,28], proton-bonded carbons [7,8,13-15,18,26,27], and proton-bonded imino and amino nitrogens [21,22]. However, no current techniques employ non-protonated nitrogens, which serve as key hydrogen-bond acceptors in nucleic acids, to characterize RNA excited states. Hence, it is of interest to utilize these unique probes to complement existing RD measurements and directly characterize ES base pairing interactions in RNA. Developed by Forsen and Hoffman in the early 1960s [29], the saturation transfer type NMR experiment has become a powerful and versatile approach for studying excited states in proteins and nucleic acids [24,30-32]. In particular, by avoiding complications due to homonuclear and heteronuclear scalar couplings [33-35], CEST NMR spectroscopy has demonstrated accurate characterization of excited states using complex spin systems, such as base and sugar carbons from uniformly ¹³C/¹⁵N labeled RNA samples [26,27] that can be difficult to study with other NMR RD techniques. Here, we present two CEST approaches that utilize protonated and non-protonated nitrogens as probes for characterizing RNA excited states. We demonstrated these methods on the B. Cereus fluoride riboswitch [36] and showed that ¹⁵N CEST profiles provide complementary information to previously reported ¹³C CEST profiles [37] in depicting a potential pathway of ligand-dependent allosteric regulation of the excited state of the fluoride riboswitch.

2. Materials and methods

2.1. Sample preparation

Uniformly ¹⁵N-labeled Bacillus cereus fluoride riboswitch samples were prepared as previously described [26]. Briefly, after in vitro transcription, RNA samples were ethanol precipitated, gel purified, electro-eluted with the Elutrap system (Whatman), anion-exchange purified with a Hi-Trap Q column (GE Healthcare), and desalted by exchanging into H₂O. The ligand-free (apo) samples were prepared by exchanging the desalted RNA into a Mg²⁺ saturated condition with a final RNA concentration ~ 0.5 – 1mM with 2mM free Mg²⁺ in NMR buffer (10 mM sodium phosphate with pH 6.4, 50mM KCl, 50μM EDTA). The fluoride-bound (holo) samples were prepared by the addition of 10mM sodium fluoride directly to the apo RNA samples. 5% D₂O (Sigma) was added into all NMR samples.

2.2. NMR spectroscopy

All NMR experiments were carried out on a Bruker Avance III 600 spectrometer equipped with 5mm triple-resonance cryogenic probes at 303K. For protonated imino (G-N1 and U-N3) ¹⁵N CEST experiments on the apo riboswitch, ¹⁵N B₁ fields (ω/2π) of 27.21 Hz and 37.17 Hz were used during the T_EX = 0.4 s period. The ¹⁵N carrier was set to 152.4 ppm with a spectral width of 25 ppm, and the ¹⁵N offsets ranged from −990 to 990 Hz with spacing of 30 Hz. For non-protonated (A-N1/N3/N7 and G-N7) ¹⁵N CEST experiments on the apo riboswitch, ¹⁵N B₁ fields (ω/2π) of 27.21 Hz and 52.1 Hz were used during the T_EX = 0.1 s period. The ¹⁵N carrier was set to 224 ppm with a spectral width of 26 ppm, and the ¹⁵N offsets ranged from −990 to 990 Hz with spacing of 30 Hz. For protonated imino (G-N1 and U-N3) ¹⁵N CEST experiments on the holo riboswitch, a ¹⁵N B₁ field (ω/2π) of 27.21 Hz was used during the T_EX = 0.4 s period. The ¹⁵N carrier was set to 152.4 ppm with a spectral width of 25 ppm, and the ¹⁵N offsets ranged from −990 to 990 Hz with spacing of 30 Hz. For non-protonated (A-N1/N3/N7 and G-N7) ¹⁵N CEST experiments on the holo riboswitch, a ¹⁵N B₁ field (ω/2π) of 27.21 Hz was used during the T_EX = 0.1 s period. The ¹⁵N carrier was set to 224 ppm with a spectral width of 26 ppm, and the ¹⁵N offsets ranged from −1000 to 1000 Hz with spacing of 50 Hz. These spin-lock powers were calibrated according to the 1D approach by Guenneugues et al. [38] as previously described [24,26]. For all CEST measurements, three spectra with T_EX = 0 s were recorded for reference in data fitting and error estimation.

2.3. Data analysis

NMR spectra were processed and analyzed using NMRPipe/NMRDraw [39], NMRView [40], and Sparky 3.110. (University of California, San Francisco, CA). All CEST profiles were obtained by normalizing the peak intensity as a function of spin lock offset Ω to the peak intensity recorded at T_EX = 0 s, where Ω = ω_rf-Ω_obs is the difference between the spin-lock carrier (ω_rf) and the observed peak (Ω_obs) frequencies. Measurement errors were estimated based on both triplicates at T_EX = 0 and the baseline of CEST profiles. The CEST profiles of residues displaying conformational exchange were fit to a two-state exchange between ground (G) and excited (E) states based on the Bloch-McConnell equation [41] that describes magnetization evolution in a coupled two-spin ¹⁵N-¹H system [42,43]. For individual state (i), the evolution of its magnetization (vⁱ) as a coupled two-spin ¹⁵N-¹H system is described by [42],

$$\frac{d}{dt}{\mathbf{v}}^i=-{\mathbf{R}}^i{\mathbf{v}}^i=\left(\begin{array}{cccccc}\hfill R_2^i\hfill & \hfill {\omega }_N^i\hfill & \hfill 0\hfill & \hfill {\eta }_{xy}^i\hfill & \hfill \pi J_{NH}^i\hfill & \hfill 0\hfill \\ \hfill -{\omega }_N^i\hfill & \hfill R_2^i\hfill & \hfill {\omega }_1\hfill & \hfill -\pi J_{NH}^i\hfill & \hfill {\eta }_{xy}^i\hfill & \hfill 0\hfill \\ \hfill 0\hfill & \hfill -{\omega }_1\hfill & \hfill R_1^i\hfill & \hfill 0\hfill & \hfill 0\hfill & \hfill {\eta }_z^i\hfill \\ \hfill {\eta }_{xy}^i\hfill & \hfill \pi J_{NH}^i\hfill & \hfill 0\hfill & \hfill R_{2HN}^i\hfill & \hfill {\omega }_N^i\hfill & \hfill 0\hfill \\ \hfill -\pi J_{NH}^i\hfill & \hfill {\eta }_{xy}^i\hfill & \hfill 0\hfill & \hfill -{\omega }_N^i\hfill & \hfill R_{2HN}^i\hfill & \hfill {\omega }_1\hfill \\ \hfill 0\hfill & \hfill 0\hfill & \hfill {\eta }_z^i\hfill & \hfill 0\hfill & \hfill -{\omega }_1\hfill & \hfill R_{1HN}^i\hfill \end{array}\right)\left(\begin{array}{c}\hfill N_x^i\hfill \\ \hfill N_y^i\hfill \\ \hfill N_z^i\hfill \\ \hfill 2H_z{N}_x^i\hfill \\ \hfill 2H_z{N}_y^i\hfill \\ \hfill 2H_z{N}_z^i\hfill \end{array}\right)$$

where R₁ⁱ is the ¹⁵N longitudinal relaxation rate, R₂ⁱ is the ¹⁵N transverse relaxation, R_1HNⁱ is the ¹⁵N-¹H two-spin order relaxation rate, R_2HNⁱ is the ¹⁵N antiphase relaxation rate, η_zⁱ is the N-H dipolar-dipolar/nitrogen CSA cross-correlated relaxation between the ¹⁵N longitudinal and two-spin order elements, η_xyⁱ is N-H dipolar-dipolar/nitrogen CSA cross-correlated relaxation between ¹⁵N transverse and antiphase magnetizations, ω_Nⁱ is the offset of the applied ¹⁵N B₁ field with a strength of ω₁ from state i, and J_NHⁱ is the ¹⁵N-¹H scalar coupling. The evolution of GS and ES magnetizations in a two-state exchange model can be described by,

$$\frac{d}{dt}\sigma (t)=-\mathbf{L}\cdot \left[\begin{array}{c}\hfill {\mathbf{v}}^G\hfill \\ \hfill {\mathbf{v}}^E\hfill \end{array}\right]=\left(\left[\begin{array}{cc}\hfill {\mathbf{R}}^G\hfill & \hfill 0_6\hfill \\ \hfill 0_6\hfill & \hfill {\mathbf{R}}^E\hfill \end{array}\right]+\left[\begin{array}{cc}\hfill -k_{GE}\hfill & \hfill k_{EG}\hfill \\ \hfill k_{GE}\hfill & \hfill -k_{EG}\hfill \end{array}\right]\otimes 1_6\right)\cdot \left[\begin{array}{c}\hfill {\mathbf{v}}^G\hfill \\ \hfill {\mathbf{v}}^E\hfill \end{array}\right]$$

where v^G/E and R^G/E are magnetization and relaxation matrices for ground and excited states as detailed above, 0₆ and 1₆ are 6×6 null and identity matrices, respectively, and k_GE and k_EG are forward and backward exchange rates as defined by k_GE = p_E k_ex and k_EG = p_G k_ex. Here, k_ex = k_GE + k_EG is the rate of exchange, p_G and p_E are populations of ground and excited states, respectively, and ω^G = Ω_obs and ω^E = ω^G + Δω, where Δω is the chemical shift difference between the ground and excited states. Ground state and excited state magnetizations at the beginning of the T_EX period are along Z and are set to be at populations of p_G and p_E. For G-N1 and U-N3 ¹⁵N CEST profiles obtained using the ¹J_NH-based HSQC CEST pulse sequence, the initial magnetization is N_z magnetization; for A-N1/N3/N7 and G-N7 ¹⁵N CEST profiles obtained using the ²J_NH-based HSQC CEST pulse sequence, the initial magnetization is the two-spin order (2N_zH_z). During analysis of 2N_zH_z CEST profiles, the fitting parameters are Δω, k_ex, p_E, R₂ = R₂^G/E, R_1HN = R_1HN^G/E, and R_2HN = R₂^G/E + R_1HN^G/E - R₁^G/E as described previously [27,34]. To simplify data fitting, η_z^G/E and η_xy^G/E were set to 0 as they have been shown not to affect the extracted Δω, k_ex, and p_E values [35]. In addition, we set R₁^G/E to 0, as the data does not constrain determination of R₁, and varying R₁ (0.0–2.0) s⁻¹ minimally effects extracted Δω, k_ex, and p_E values. Since G-N1 and U-N3 ¹⁵N CEST profiles were measured in the presence of ¹H decoupling, their data analysis can be simplified by setting all two-spin relaxation parameters (R_1HNⁱ, R_2HNⁱ, η_zⁱ, η_xyⁱ and J_NHⁱ) to 0, which essentially turns the individual-state evolution from a 6×6 matrix to a 3×3 matrix. While the non-¹H-decoupled ²J_NH CEST profiles are split into doublets, the small couplings (²J_NH ~ 10–15 Hz) are not resolved with the applied B₁ fields. For residues without conformational exchange, the two-state model was simplified to a one-state model by fixing all exchange parameters (rate of exchange k_ex and population of excited state p_ES) to 0. All profiles were fitted using an in-house MATLAB® program with a Levenberg-Marquardt algorithm.

3. Results and discussion

The pulse sequences for measuring ¹⁵N CEST profiles in nucleic acids are based on the conventional ¹H-¹⁵N HSQC scheme and feature different nitrogen magnetization preparations prior to the saturation transfer period (Fig. 1A,B). These schemes are used to meet the distinct spectroscopic needs of two types of nitrogens in nucleic acids: protonated nitrogen, such as G-N1 and U-N3, and non-protonated nitrogen, such as A-N1/N3/N7 and G-N7 (Fig. 1C). Shown in Fig. 1A is the pulse sequence for probing protonated nitrogen, which is similar to the original CEST scheme by Kay and co-workers [24]. Here, a flip-back scheme is used to provide a simple way of water suppression with sufficient sensitivity for our measurements. The sensitivity enhancement scheme in the original CEST pulse sequence [24] can be further implemented to study larger RNAs. Given ¹J_NH couplings being ~ 90–100Hz, ¹H magnetization can be efficiently converted to pure ¹⁵N magnetization (N_z) via refocused INEPT prior to the saturation period, which we refer to hereinafter as the ¹J_NH-based approach. During the subsequent mixing period with a weak ¹⁵N B1 field, a 2.35 kHz 90_x240_y90_x composite pulse train [44] is used for ¹H decoupling to suppresses N-H cross relaxation, dipolar-dipolar/nitrogen CSA cross-correlated relaxation, and the ¹⁵N multiplet structure in the CEST profile [24].

Fig. 1.

2D ¹⁵N CEST pulse schemes for using (A) protonate nitrogen and (B) non-protonated nitrogen as probes for characterizing slow chemical exchange in nucleic acids. Highlighted in (C) are schematic magnetization transfers from protons to G-N1 and U-N3 via ¹J_NH couplings and to A-N1/N3/N7 and G-N7 via ²J_NH couplings. The sequences are based on conventional ¹H-¹⁵N HSQC pulse scheme with a simple water flip-back approach. Narrow (wide) rectangles are hard 90° (180°) pulses, and open shapes are 1-ms selective 90° pulses on water. All pulses are applied along x-axis unless indicated otherwise. (A) For measuring protonated nitrogens, a 90_x240_y90_x composite pulse [44], as previously described by Kay and co-workers [24], is used for ¹H decoupling to suppress N-H cross relaxation and N-H dipolar-dipolar/nitrogen CSA cross-correlated relaxation during the T_EX period with a weak ¹⁵N B₁ field. The ¹H carrier is kept on water resonance throughout the experiment except during the T_EX period, where it is shifted to the center of the region of interest. The ¹⁵N carrier is also kept on-resonance throughout the experiment and is shifted to a desired offset during the T_EX period. Inter-pulse delays are set to τ_a = 2.4 ms and τ_b = 2.77 ms. The phase cycle used is ϕ₁ = {x, −x}, ϕ₂ = {y}, receiver = {x, −x}. Gradients with smoothed-square shape (SMSQ10.100) profile are applied with the following strength (G/cm)/duration (ms): g₁ = 19.8/1.0, g₂ = 29.7/1.0, g₃ = 6.6/1.0, g₄ = −26.4/1.0, g₅ = 29.7/1.0, g₆ = 33.0/1.0. ϕ₂ and the receiver phase are incremented in a States-TPPI manner. ¹³C and ¹⁵N decoupling during acquisition are achieved using 2.5 kHz GARP and 1.25 kHz WALTZ-16, respectively. (B) For measuring non-protonated nitrogens, only a weak ¹⁵N B₁ field is applied during the T_EX period. The ¹H carrier is kept on water resonance throughout the experiment. The ¹⁵N carrier is kept on-resonance throughout the experiment except being shifted to a desired offset during the T_EX period. Inter-pulse delay is set to τ_c = 9 ms. The phase cycle used is ϕ₁ = {y}, ϕ₂ = {x, −x}, receiver = {x, −x}. Gradients with smoothed-square shape (SMSQ10.100) profile are applied with the following strength (G/cm)/duration (ms): g₁ = 19.8/1.0, g₂ = 29.7/1.0, g₃ = −26.4/1.0, g₄ = 29.7/1.0, g₅ = 33.0/1.0. ϕ₁ and the receiver phase are incremented in a States-TPPI manner. ¹³C and ¹⁵N decoupling during acquisition are achieved using 2.5 kHz GARP and 1.25 kHz WALTZ-16, respectively. To ensure uniform heating for experiments with variable lengths of T_EX, a heat compensation scheme is employed after the acquisition with length of T_MAX - T_EX, where T_MAX is the maximum relaxation delay time, and far off-resonance for both ¹H and ¹⁵N channels.

For probing non-protonated nitrogens that can serve as hydrogen-bond acceptors, we employed a slightly different HSQC-based strategy to obtain ¹⁵N CEST profiles (Fig. 1B) that is similar to recent approaches for measuring protein amide ¹H CEST profiles [35,45]. Here, unlike protonated nitrogens, transferring magnetization of a carbon-bonded proton to pure N_z magnetization of a nitrogen of interest is significantly less efficient due to small ²J_NH scalar couplings (~ 10-15 Hz), which requires a lengthy refocused INEPT module that results in significant signal losses. Instead of generating pure N_z magnetization, the pulse scheme shown in Fig. 1B converts ¹H magnetization of a non-exchangeable proton to longitudinal two-spin order (2N_zH_z) prior to the saturation period, which we refer to hereinafter as the ²J_NH-based approach. However, due to the small ²J_NH scalar couplings, we found that an inter-pulse delay of 1/4J ~ 17–25 ms during INEPT remains too long to obtain adequate sensitivity. Hence, in practice, the length of the inter-pulse delay is varied to optimize sensitivity and achieve an optimal balance between the proton-nitrogen magnetization transfer and the loss of magnetization due to relaxation. For the riboswitch studied here, an inter-pulse delay of τ_c = 9 ms was found to give optimal sensitivity in the 2D ¹H-¹⁵N HSQC spectrum, and any residual inphase transverse proton magnetization is dephased via Z gradient prior to the T_EX period. Applying a selective ¹⁵N pulse together with nucleus-specific τ_c could further improve sensitivity for measuring site-specific CEST profiles, such as for N1 and N3 in As. In contrast to the ¹J_NH-based approach (Fig. 1A), only a weak ¹⁵N B₁ field is applied during the mixing period (Fig. 1B), as any inhomogeneity in the ¹H B₁ field can lead to additional signal losses of the two-spin order. Contributions from in-phase/anti-phase relaxation, longitudinal relaxation and cross-correlated relaxation to CEST profiles can be taken into account during the stage of data analysis with Bloch-McConnell equations (see Material and Methods). It is worth noting that the longitudinal two-spin order (2N_zH_z) can relax faster than pure N_z magnetization. Therefore, a shorter mixing time T_EX is typically used, which, consequentially, also limits the overall sensitivity of the ²J_NH-CEST approach in detecting excited states. While the triple resonance HCN scheme [46] could provide a potential pathway to transfer magnetizations from H8 to N7 in As/Gs and from H2 to N1/N3 in As, its application is severely limited due to extremely small scalar couplings (¹J_CN < 3 Hz) for C8-N7 and C2-N1/N3 spin systems [47]. Alternatively, pure N_z magnetization can be prepared using 1D selective Hartmann–Hahn polarization transfer scheme [48,49], as demonstrated in obtaining residue-specific R_1ρ [17,18] and carbon CEST [26,34] profiles. While it would still require a lengthy period of magnetization transfer (~1/J), the CEST profile of a single nitrogen of interest could be measured with a longer T_EX within a reasonable experimental time. Finally, given the large range of N1/N3/N7 chemical shifts, multi-frequency irradiation techniques, such as DANTE [50,51] (D-CEST) [52] and cosine-modulated (cos-CEST) [53] excitation schemes, can be implemented to expedite data collection.

To demonstrate the two ¹⁵N CEST methods, we carried out measurements on the Bacillus cereus fluoride riboswitch, a transcriptional riboswitch that regulates gene expression of fluoride transporters (Fig. 2A) [36]. Riboswitches are a class of non-coding RNAs that regulate transcription and/or translation via ligand-dependent conformational changes [54]. However, we recently showed that the aptamer domain of the B. cereus fluoride riboswitch adopts essentially identical tertiary structures with and without ligand [37]. Instead of changing its ground-state (GS) structure, the fluoride riboswitch utilizes a novel switching mechanism for riboswitch in which the ligand allosterically regulates dynamic access to a functional excited state (Fig. 2B). In the absence of ligand, the apo aptamer undergoes a GS ↔ ES conformational exchange, where the aptamer transiently unlocks the highly conserved reverse Hoogsteen base pair A37-U45, a linchpin that resides at the interface between the aptamer and expression platform of the riboswitch. This linchpin gating process was shown to provide an efficient path for strand invasion to terminate transcription [37]. By contrast, ligand binding allosterically suppresses this ES transition, resulting in a single stable conformation that outcompetes the terminator during the co-transcriptional event to ensure continued gene transcription.

Fig. 2.

Quantification of the excited state in the B. cereus fluoride riboswitch by ¹J_NH-based 2D ¹⁵N CEST. (A) Sequence and secondary structure of the B. cereus fluoride riboswitch aptamer. (B) Schematic representation of ligand-dependent conformational transitions in the fluoride riboswitch aptamer. The apo aptamer undergoes an exchange between the excited state (ES) and the ground state (GS). Upon ligand binding, the holo aptamer adopts a single stable state. (C) ¹⁵N CEST measurements of protonated nitrogens in the apo state. Shown on the left panel is ¹H-¹⁵N HSQC spectrum of N1–H1 (Gs) and N3–H3 (Us) of the apo riboswitch, where colored in red are residues whose ¹⁵N CEST profiles are shown on the right panel. Solid lines represent the best fits to a two-state exchange process using the Bloch-McConnell equation. (D) ¹⁵N CEST measurements of protonated nitrogens in the holo state. Shown on the left panel is ¹H-¹⁵N HSQC spectrum of N1–H1 (Gs) and N3–H3 (Us) of the holo riboswitch, where colored in red are residues whose CEST profiles are shown on the right panel. Shown are representative profiles with a ¹⁵N B₁ field (ω/2π) of 27.21 Hz for a duration of T_EX = 0.4 s.

We first applied the ¹J_NH-based approach to obtain ¹⁵N CEST profiles for the protonated G-N1 and U-N3 nitrogens of the fluoride riboswitch (Fig. 2C,D). The ¹H-¹⁵N HSQC spectrum of the imino region of the apo riboswitch is well resolved, and we were able to obtain a total of 18 ¹⁵N CEST profiles, including 12 G-N1 profiles and 6 U-N3 profiles (Fig. 2C and Fig. S1 A). Consistent with our previous ¹³C CEST characterizations [37], the imino ¹⁵N CEST profiles also revealed the presence of an excited state in the apo state. While most Gs and Us display single dips in their intensity profile that match peak positions in the ¹H-¹⁵N HSQC spectrum, second and asymmetrically broadened intensity dips that correspond to the apo ES can be clearly seen for U38, G39, and U45 (Fig. 2C and Fig. S1A). Global fitting of these three ¹⁵N CEST profiles to a single two-state model gave an exchange rate (k_ex) of 184 ± 10 s⁻¹ and an ES population (p_ES) of 1.6 ± 0.1%, resulting in an ES lifetime (τ_ES = 1/k_EG) of 5.5 ± 0.5 ms. These ES parameters are similar to values (p_ES = 1.4 ± 0.1% and τ_ES = 3.2 ± 0.3 ms) from our previous ¹³C CEST measurements [37]. The observed discrepancy is largely due to differences in sample conditions, where the ¹⁵N CEST experiments were carried out on ¹⁵N-labeled samples in H₂O and the ¹³C CEST measurements were carried out on ¹³C/¹⁵N-labeled samples in D₂O. As a control, we measured imino ¹⁵N CEST profiles for the holo aptamer (Fig. 2D). Due to spectral overlap between G10 and G31 as well as the appearance of U35 resonance, we were able to obtain a total of 17 ¹⁵N CEST profiles in the holo state, including 10 G-N1 profiles and 7 U-N3 profiles (Fig. 2D and Fig. S2A). The uniform single-intensity dip of these ¹⁵N CEST profiles is consistent with the holo state being in a single stable conformation.

Relative to carbon-based approaches, measurable imino nitrogens can be too spatially clustered around structured regions of RNA to provide a comprehensive set of CEST profiles. This is largely due to imino protons being solvent exchangeable, and only structured or highly protected imino resonances are observable for direct measurement. In contrast to imino nitrogens, non-protonated nitrogens can provide probes distributed across structured and unstructured regions, as their magnetizations are transferred from non-exchangeable, carbon-bonded protons. In addition, while a single ¹H-¹⁵N HSQC spectrum detects one non-protonated nitrogen (N7) of guanine, each adenine has up to three of these probes (N1, N3 and N7), further expanding the number of probes available for characterizing conformational exchange. Shown in Fig. 3 are ¹H-¹⁵N HSQC spectra of these non-protonated nitrogens, and their corresponding ¹⁵N CEST profiles are measured using the ²J_NH-based approach. In the apo state, we were able to obtain a total of 36 ¹⁵N CEST profiles from the non-protonated nitrogens, 11 from G-N1 and 25 from A-N1/N3/N7 (Fig. 3A and Fig. S1B). While most of the non-protonated nitrogen CEST profiles display single intensity dips at corresponding resonance positions in ¹H −¹⁵N HSQC spectrum, 9 of the 36 profiles exhibit second and asymmetrically broadened intensity dips. These non-protonated nitrogen profiles can be globally fitted to extract k_ex = 171 ± 31 s⁻¹ and p_ES = 2.0 ± 0.2%, agreeing well with parameters obtained from imino nitrogen CEST profiles. Hence, we globally fit all ¹⁵N CEST profiles to a single two-state model, and the resulting k_ex = 192 ± 15 s⁻¹, P_ES = 1.7 ± 0.1%, and τ_ES = 5.3 ± 0.5 ms are similar to ES parameters from previous ¹³C CEST characterization of the apo fluoride riboswitch. We also measured ¹⁵N CEST profiles for non-protonated nitrogens in the holo aptamer, in which all 39 profiles uniformly displayed single intensity dips (Fig. 2D and Fig. S2B).

Fig. 3.

Quantification of the excited state in the B. cereus fluoride riboswitch by ²J_NH-based 2D ¹⁵N CEST. (A) ¹⁵N CEST measurements of non-protonated nitrogens in the apo state. Shown on the left panel is ¹H-¹⁵N HSQC spectrum of N7–H8 (As/Gs) and N1/3–H2 (As) of the apo riboswitch, where colored in red are residues whose ¹⁵N CEST profiles are shown on the right panel. Solid lines represent the best fits to a two-state exchange process using the Bloch-McConnell equation. (B) ¹⁵N CEST measurements of non-protonated nitrogens in the holo state. Shown on the left panel is ¹H-¹⁵N HSQC spectrum of N7–H8 (As/Gs) and N1/3–H2 (As) of the apo riboswitch, where colored in red are residues whose ¹⁵N CEST profiles are shown on the right panel. Shown are representative profiles with a ¹⁵N B₁ field (ω/2π) of 27.21 Hz for a duration of T_EX = 0.1 s.

With protonated and non-protonated nitrogen probes, we have obtained a total of 54 and 56 ¹⁵N CEST profiles for the apo and holo fluoride riboswitches, respectively. These data, together with our previously reported 61 apo-state and 65 holo-state ¹³C CEST profiles [37], provide a comprehensive map of the ligand-dependent conformational tuning of the fluoride riboswitch (Fig. 4). Upon ligand binding, the fleeting dynamics to the ES is suppressed, resulting in one kinetically stable conformation as strongly supported by the uniform single-intensity dips across all holo-state ¹³C/¹⁵N CEST profiles. Consistent with our previous results, except A17 and A19 that are stacked on the end of P1, residues experiencing the chemical exchange in the apo state are clustered around the junction of P3, J13, J23, and the 3’ tail (Fig. 4). In particular, the N7 ¹⁵N CEST profile of A40 unveiled the presence of GS ↔ ES transition at this location, which was not detected in our previous ¹³C CEST measurements. The ¹⁵N CEST profiles have also unveiled changes in hydrogen bonds that cannot be directly characterized using ¹³C CEST experiments. For example, the N3 ¹⁵N CEST profile of U45, which base pairs with A37 in the GS, displays an upfield-shifted ES chemical shift (Fig. 2C), indicating that U45-N3H3 may not be base paired in the ES. Interestingly, we could not detect dispersion on the hydrogen-bond receptor A37-N7, which could be due to limited difference between GS and ES chemical shifts of A37-N7. Similar to U45, the N3 ¹⁵N CEST profile of U38 also displays an upfield-shifted ES chemical shift (Fig. S1A), suggesting that the hydrogen bond between U38-N3H3 and C41-O2P may be absent in the ES. Furthermore, the N7 ¹⁵N CEST profiles of G7 and A40 reveal that the G7-N7 to A37-2’OH and A40-N7 to U38-2’OH hydrogen bonds are possibly broken in the ES. Due to the unique N–HO hydrogen bond, N7 chemical shifts of G7 and A40 are substantially upfield-shifted in the GS (Fig. 3A). In contrast, their ES chemical shifts are both downfield-shifted to the chemical shift range of non-hydrogen-bonded N7s in As and Gs (Fig. S1B). Together, ¹⁵N/¹³C CEST profiles of the apo riboswitch depict a trail of spatially continuous residues that bridge the ligand binding pocket and the linchpin gating site, suggesting a potential pathway for how ligand-binding allosterically regulates dynamic access to the functional excited state.

Fig. 4.

Nitrogen and carbon CEST mapping of the excited state in the apo B. cereus fluoride riboswitch. Spheres shown on the apo aptamer structure are the sites where (A) ¹⁵N, (B) ¹³C, and (C) ¹⁵N/¹³C CEST data were measured. Gray spheres are probes fit to a single-state model and red spheres are probes fit to a two-state exchange model.

In summary, we have presented two ¹⁵N CEST experiments to characterize slow chemical exchange in nucleic acids, which add to the growing list of NMR RD methods that have played critical roles in discovering and identifying excited states across diverse non-coding RNAs. The protonated and non-protonated nitrogens employed in these methods not only complement current probes for NMR RD measurements, but further enable direct characterization of hydrogen-bond donors and acceptors, which provide key interactions for canonical and non-canonical base pairs. Recently, it was shown that the network of these base pairs defines the overall topology of RNA tertiary structures [55,56], where even sparsely populated, imino-based NMR constraints can provide sufficient experimental input for computational modeling of high-resolution RNA structures [57]. With the ability to directly probe ES base pairing interactions, the ¹⁵N CEST methods presented here further promise an opportunity for high-resolution structural modeling of functional RNA excited states.

Highlights.

Protonated G-N1 and U-N3 ¹⁵N CEST used to probe chemical exchange in RNA

Non-protonated G-N7 and A-N1/N3/N7 ¹⁵N CEST used to probe chemical exchange in RNA

¹⁵N CEST complements ¹³C CEST for more comprehensive mapping of RNA dynamics