Consolidated 3‑Fold Isotopic Lens for Probing RNAs

Abstract

Undesired scalar and dipolar couplings are two major interactions that complicate structural and dynamic studies in solution nuclear magnetic resonance (NMR) spectroscopy. Recent developments in site-specific isotopic labeling technologies have gone a long way toward alleviating these problems. While some nuclei have intrinsic properties that make them suitable for specific NMR experiments, these same properties render them inefficient in other experiments. Site-specific isotopic labeling facilitates the controlled incorporation of isotopes to enable facile analysis of RNAs. Here, we describe the synthesis and incorporation of [1′-¹³C, 2-¹⁹F, 7-¹⁵N] adenosine 5′-triphosphate into the 9 kDa Escherichia coli rRNA, thus expanding the applications of previously synthesized [2-¹³C, 7-¹⁵N]-adenosine 5′-triphosphate, with the added benefit of ¹⁹F incorporation. We utilized these ¹³C and ¹⁵N probes to characterize the structural dynamic features within this RNA, and ¹⁹F was used to monitor binding interactions. Finally, we leveraged the chemical shielding anisotropy-dominated relaxation of ¹⁵N7-adenosine for straightforward analysis of R ₁ and R _1ρ rates.

Introduction

Among the several biophysical techniques for the structural determination of RNAs, nuclear magnetic resonance (NMR) spectroscopy remains an attractive tool, as it provides structural information at the atomic level. − Nonetheless, some inherent problems are associated with NMR. These include line broadening, signal overlap, and spectral overcrowding as molecular weight increases. − To limit these problems, spectroscopists have developed several stable isotope labeling technologies (e.g., ²H, ¹³C, ¹⁵N, and ¹⁹F). − They choose the isotope to probe by NMR spectroscopy based on practical and efficiency considerations. For example, the low gyromagnetic ratio of ¹⁵N nucleus translates into relatively sharper lines at high molecular weights. , But again, the low gyromagnetic ratio makes ¹⁵N less sensitive to external magnetic fields, resulting in weaker NMR signals. In contrast, ¹⁹F not only has a high gyromagnetic ratio that makes it very sensitive to such fields but also a high chemical shift dispersion (at least six times that of ¹H) that potentially mitigates spectral crowding. Nonetheless, the large chemical shielding anisotropy (CSA) of ¹⁹F leads to extensive line broadening in ¹⁹F NMR at high molecular weights and high magnetic field strengths. This is a significant limitation to the use of ¹⁹F NMR for studying larger biomolecules. Conversely, the very narrow line widths of the ¹⁵N nucleus make it an attractive candidate for use in large RNAs, regardless of its inherently low sensitivity. In addition to ¹⁵N and ¹⁹F, ¹³C has a gyromagnetic ratio 1/4th that of ¹H, also translating to sharp lines, albeit broader than ¹⁵N. Nonetheless, the line width of ¹³C in the ribose ring is relatively sharper due to its small CSA values. This makes ribose ¹³C desirable for experiments that require relatively high magnetic field strengths. Given that some of the limitations of ¹⁵N, ¹³C, and ¹⁹F are offset by each other, consolidating the properties of all three into a single probe would likely enhance its overall utility in studying RNAs by NMR spectroscopy.

For instance, we recently incorporated ¹⁵N and ¹³C into the same nucleotide. We synthesized and incorporated [2-¹³C, 7-¹⁵N]-adenosine 5′-triphosphate into ∼20 kDa human hepatitis B ε RNA viral element. With ¹H–¹⁵N heteronuclear single quantum coherence (HSQC) experiments, we obtained well-resolved peaks, with all expected signals accounted for. We further characterized this RNA with ¹⁵N-edited NOESY HSQC and ¹³C spin-relaxation experiments. ,, In another previous work, researchers have shown that the substitution of adenosine-H2 with ¹⁹F does not significantly perturb RNA’s physical or structural properties. It is unclear if we can ascribe this observation to two opposing effects: Is ¹⁹F’s ability to destabilize Watson–Crick base pairing counteracted by its stable base π-stacking? , Another noteworthy observation is that because ¹⁹F polarizes uridine nucleotides, researchers can readily identify different RNA secondary structural motifs. A similar observation was documented by Sochor et al. Imino protons of uridines involved in canonical Watson–Crick base pairing to 2-fluoroadenosine showed significant chemical shift perturbations (CSPs) from the regular A-U base pair. Additionally, the presence of fluorine at the H2 position makes it very sensitive to its chemical environment. , With such a dual ¹³C–¹⁹F label, Kreutz and co-workers characterized RNA-small molecule interactions and detected low-populated dynamic processes in RNA on the microsecond time scale, otherwise invisible in a ¹H spectrum, and different RNA secondary structural motifs.

In this work, we therefore extend those dual labels to a triple ¹⁵N–¹³C-¹⁹F site-specific label. This site-specific triple label exploits the favorable spectroscopic properties of ¹⁹F and thus may enhance its versatility. We synthesized and applied [1′-¹³C, 2-¹⁹F, 7-¹⁵N] adenosine 5′-triphosphate to a 9 kDa E. coli rRNA. This new isotopic labeling pattern enables us to qualitatively tease out the effects of ¹⁹F substitution on nuclei even up to five bonds away and analyze the structural dynamics and interactions of RNA with ligands.

Materials and Methods

Theoretical Simulations

The chemical shielding tensors (CST) for ¹H8-¹⁵N7 and H1′–C1′ spin pairs were calculated using density functional theory (DFT) methods, as described in previous work (Tables S1 and S2). − Using these CST parameters, we computed the R ₂ relaxation rates using eqs and for the ¹H8–¹⁵N7 and ¹H1–¹³C1′ spin pair in A (Figure A) assuming isotropic tumbling. We assumed an isolated N7–H8 and H1′–C1′ spin pair, where dipolar contributions from neighboring nuclei were ignored. For bond distances between H8–N7 and H1′–C1′ in A, we used 2.1 and 1.1 Å, respectively, based on the helical region of the PDB structure 6VAR, the human hepatitis B encapsidation signal ε RNA. We repeated these calculations using SPINACH software to consider every significant relaxation, cross-relaxation, and cross-correlation process in a rigid system with two spin-1/2 nuclei.

$$R_2=\left(\frac{A}{2}\right){[4J(0)+3J(\omega }_{\mathrm{N}})+{6J(\omega }_{\mathrm{H}8}+{\omega }_{\mathrm{N}})+{6J(\omega }_{\mathrm{H}8})+{J(\omega }_{\mathrm{H}8}-{\omega }_{\mathrm{N}})]+\left(\frac{B}{6}\right)[4J(0)+{3J(\omega }_{\mathrm{N}})]+R_{\mathrm{ex}}$$ 1

$$A={\left(\frac{{\mu }_0}{4\pi }\right)}^2{\left(\frac{{\gamma }_{\mathrm{N}}{\gamma }_{\mathrm{H}}\hslash }{{2r^3}_{\mathrm{NH}}}\right)}^2,B=\frac{({{\mathrm{\Delta }\sigma }_{\mathrm{N}}}^2{B_0}^2{{\gamma }_{\mathrm{N}}}^2)}{3},J(\omega )=\frac{{\tau }_{\mathrm{c}}}{5(1+{{(\omega \tau }_{\mathrm{c}})}^2)}$$ 2

$${\mathrm{\Delta }\sigma }_{\mathrm{N}}=\sqrt{({\sigma }_X^2+{\sigma }_Y^2-{\sigma }_X{\sigma }_Y)},{\sigma }_X={\sigma }_{\mathit{ZZ}}-{\sigma }_{\mathit{XX}},{\sigma }_Y={\sigma }_{\mathit{ZZ}}-{\sigma }_{\mathit{YY}}$$ 3

6.

(A) Theoretical simulation for ¹⁵N7-A and ¹³C1′-A line widths at correlation time (τ _c) of 5 ns and varying magnetic field strengths using SPINACH software. ¹⁵N7 transverse relaxation rates (R ₂) increase quadratically with increasing magnetic field strength. (B) Experimental R _1ρ for ¹³C1′-A_F/H in fluorinated RNA (magenta) and nonfluorinated RNA (orange). R _1ρ for ¹⁵N7-A_F in fluorinated RNA (green). Magenta, green, and orange dashed lines represent the mean for ¹³C1′-A_F, ¹⁵N7-A_F, and ¹³C1′-A_H, respectively. (C) Longitudinal relaxation rates (R ₁) for ¹⁵N7-A_F. All relaxation experiments were conducted at 298 K. (D) R ₂/R ₁ values for ¹⁵N7-A_F. The dashed black lines in (C) and (D) represent the average relaxation rates.

σ _XX , σ _YY , and σ _ZZ are the principal components of the traceless chemical shielding tensor in the Haeberlen notation, , ordered as follows: | σ _ZZ -σ _iso | ≥ | σ _XX -σ _iso | ≥ | σ _YY -σ _iso |; ${\sigma }_{\mathrm{iso}}=\frac{1}{3}({\sigma }_{\mathit{XX}}+{\sigma }_{\mathit{YY}}+{\sigma }_{\mathit{ZZ}})$ . Here, μ₀ is the permeability of the vacuum, γ_N and γ_H are the gyromagnetic ratios of ¹⁵N and ¹H respectively, ℏ is Planck’s constant divided by 2π, r _NH is the internuclear distance between ¹⁵N7 and ¹H8, Δσ _N (eq ) is the chemical shift anisotropy of ¹⁵N, B _o is the static magnetic field strength, J(ω) is the spectral density function at a given frequency assuming isotropic tumbling, τ _c is the rotational correlation time, ω is the angular frequency of the nucleus, and R _ex is the chemical exchange rate constant. The internuclear distance and magnetic field strength were adjusted accordingly when simulating relaxation rates for C1′ in A.

DFT Calculations

Quantum chemistry calculations were carried out using ORCA , and the automated fragmentation (AFNMR) package , as described earlier. −

General Procedures

A detailed description of the synthesis of [2-¹³C, 7-¹⁵N]-adenine and [7-¹⁵N]-guanine was described previously. , Commercially available reagents were used for the syntheses of the nucleobases without further purification. Isotopically labeled [¹⁵N]-sodium nitrite was purchased from Sigma-Aldrich (St. Louis, MO).

Reagents

[1-¹³C]-ribose was purchased from Omicron Biochemicals (IN). The enzymes used in the one-pot coupling reaction of [1-¹³C]-ribose to an isotopically labeled nucleobase were ribokinase (RK), phosphoribosyl pyrophosphate synthetase (PRPPS), and adenine phosphoribosyl transferase (APRT). These enzymes were purified and expressed by methods we have previously described. ,− Thermostable pyrophosphatase was purchased from New England Biolabs (Ipswich, MA). Bovine serum albumin, creatine kinase (CK), and myokinase (MK) were purchased from ThermoFisher Scientific (Waltham, MA). A boronate column with Affi-Gel Boronate Gel (BioRad) was used in the purification of isotopically labeled nucleoside triphosphates.

Nucleoside Triphosphate Synthesis

The nucleoside triphosphates used in the enzymatic transcription of both RNA constructs (fluorinated and nonfluorinated) were synthesized as described previously. , All reactions were carried out at 37 °C in synthesis buffer (100 mM creatine phosphate, 50 mM HEPES pH 8, 10 mM magnesium chloride, 10 mM dithiothreitol, 100 mM potassium chloride, and 0.1 mg/mL bovine serum albumin) and a dATP regeneration system (0.5 mM dATP, 100 mM creatine phosphate, 0.05 mg/mL creatine kinase, and 10 mU/μL myokinase). The reaction was initiated with 10 μU/μL RK, 10 μU/μL adenine phosphoribosyl transferase, 10 μU/μL PRPPS, 0.1 mg/mL thermal-stable inorganic pyrophosphatase, 8 mM fluorinated and nonfluorinated adenine, and 7 mM [1-¹³C]-ribose. On completion in 24 h, all of the reactions were purified with a boronate resin column at 4 °C. ,,

RNA Preparation

The DNA template used for transcription is 5′-mGmGC GAC TTC ACC CGA AGG TGT GAC G C CTA TAG TGA GTC GTA TTA G-3′. The T7 RNA polymerase promoter sequence was 5′-CTA TAG TGA GTC GTA TTA G-3′. DNA template was purchased from Integrated DNA Technologies (IDT) (Coralville, IA). 10 mL in vitro transcription reaction was carried out at 37 °C using a transcription buffer (40 mM Tris–HCl pH 8.0, 1 mM spermidine and 0.01% Triton X-100), 0.3 μM single strand DNA template, 80 mg/mL PEG, 1 mM DTT, 2 U/μL thermostable inorganic pyrophosphatase, 10 mg/mL T7 RNA polymerase, a total of 10 mM rNTPs (2.5 mM each of unlabeled GTP, CTP, UTP, and either 18 or 19), and 10 mM MgCl₂. The concentrations of rNTPs and MgCl₂ were chosen following optimization in small (50 μL) reactions. After 3 h, the reaction was quenched by adding 1 mL of 0.5 mM EDTA (pH 8). 10 mL of phenol–chloroform was added to the aqueous transcription reaction. The suspension was spun in a centrifuge at 10,500 rpm for 15 min. Upon completion, the aqueous RNA layer (top layer) was collected into a clean falcon tube with a pipet, followed by the addition of 1 mL of 3 M NaOAc and 30 mL of ice-chilled absolute ethanol. This solution was left in the refrigerator overnight at −20 °C. The solid precipitate formed was isolated from the solution by centrifugation at 10,500 rpm for 1 h. The aqueous solution was decanted, leaving behind a white solid crude RNA residue that was resuspended in a 1× TBE buffer. The crude RNA solution was purified via preparative polyacrylamide (10%) denaturing gel electrophoresis, followed by extraction of RNA from the gel using the crush and soak method. The collected RNA was exchanged with deionized water and lyophilized. The RNA pellets were dissolved in 270 μL of D₂O and 30 μL of 10× NMR buffer (200 mM Na₃PO₄ pH 6.5, 1 M NaCl, 5 mM DTT, 2 mM EDTA pH 8, 0.27 mM NaN₃, 0.1 mM DSS). The RNA solutions were heated to 95 °C for about 3 min and rapidly cooled on ice for 5 min. Concentrations of RNA were calculated using a molar absorption coefficient for the E. coli rRNA of 281 mM^–1 cm^–1. The concentrations of the purified unbound E. coli rRNA samples were ≈0.6 mM. The concentrations of the purified paromomycin-bound E. coli rRNA samples were 0.3 mM.

NMR Spectroscopy

All NMR experiments were conducted at 14.4 T (600 MHz ¹H frequency) on a Bruker Avance III HD spectrometer equipped with a triple-resonance shielded z-gradient probe. Two-dimensional two-bond (² J _H8N9) ¹H–¹⁵N HSQC data for fluorinated and nonfluorinated RNAs were collected in D₂O at 277 and 298 K. Two-dimensional one-bond (¹ J _H1′C1) ¹H–¹³C HSQC data for fluorinated and nonfluorinated RNAs were collected in D₂O at 277 and 298 K. ¹³C R _1ρ relaxation data on fluorinated E. coli rRNA were collected in D₂O at 298 K using standard Bruker pulse programs. For the ¹³C R _1ρ experiment, relaxation delays of 0.004 (x2), 0.016, 0.02, 0,024, 0.036, and 0.064 s were used. ¹⁵N R ₁ and R _1ρ relaxation data on fluorinated E. coli rRNA were also collected in D₂O at 298 K using standard Bruker pulse programs. For the ¹⁵N R ₁ experiment, relaxation delays of 0.02 (x2), 0.04, 0.06, 0.08, 0.12, 0.16, 0.2, 0.24, and 0.3 s were used. For the ¹⁵N R _1ρ experiment, relaxation delays of 0.004 (x2), 0.016, 0.02, 0.024, 0.036, and 0.064 s were used. ¹⁵N R _1ρ relaxation data on nonfluorinated E. coli rRNA were also collected in D₂O at 298 K using standard Bruker pulse programs. Relaxation delays of 0.004 (x2), 0.016, 0.02, 0.024, 0.036, and 0.064 s were used. Spin-lock strength of 1.9 kHz was applied for all R _1ρ measurements. Two-dimensional (2D) ¹H–¹H NOESY was conducted on the fluorinated RNA with 300 ms mixing time, 64 scans, and 2048 × 1024 complex data points in the ¹H dimensions at 298 K. 2D imino ¹H–¹H NOESY was conducted on both fluorinated and nonfluorinated RNAs with 250 ms mixing time, 64 scans, and 2048 × 400 complex data points in the ¹H dimensions at 283 K. 2D ¹³C- and ¹⁵N-edited NOESY HSQC data were collected on the fluorinated RNA with a 250 ms mixing time at 298 K. 2D ¹⁵N-edited NOESY HSQC experiment was carried out with 40 scans per increment with spectral widths of 14 ppm (2048 complex data points) in the direct ¹H-dimension and 14 ppm (200 complex data points) in the indirect ¹H-dimension. ¹³C-edited NOESY HSQC experiment was carried out with 64 scans per increment with spectral widths of 14 ppm (2048 complex data points) in the direct ¹H-dimension and 14 ppm (200 complex data points) in the indirect ¹H-dimension. ¹H–¹H TOCSY was collected for the fluorinated RNA with a mixing time of 120 ms at 298 K. This experiment was carried out with 128 scans per increment with spectral widths of 12 ppm (2048 complex data points) in the direct ¹H-dimension and 10 ppm (234 complex data points) in the indirect ¹H-dimension. 2D ¹³C-edited TOCSY HSQC experiment was performed with 64 scans per increment, with spectral widths of 16 ppm (2048 complex data points) in the direct ¹H-dimension and 16 ppm (200 complex data points) in the indirect ¹H-dimension. All NMR data were processed and analyzed by Bruker Topspin Software version 4.1.4.

UV Melting Experiment

1 μM solutions of 18 and 19-labeled RNA were prepared in a buffer containing 15 mM sodium cacodylate (pH 6.4), 25 mM NaCl, and 0.1 mM EDTA (pH 8), such that their absorbances were between 0.1 and 1. Samples were then transferred into clean cuvettes and placed in a ultraviolet (UV) spectrophotometer with the temperature change set at 1 °C per minute. The temperature range was set from 4 to 95 °C. The absorbances were recorded as a function of varying temperature. The melting curves were fit by using the Marquardt nonlinear least-squares method. Melting point and thermodynamic parameters such as ΔG ^298K were extracted using the van’t Hoff expression and nonlinear curve fitting previously described for unimolecular hairpin formation by Luteran et al. and Siegfried and Bevilacqua. The heat-up (4 to 95 °C) and cool-down (95 to 4 °C) absorbances were used as duplicates for data analysis.

Mass Spectrometry

Mass spectrometry was conducted using a JEOL AccuTOF-CS electrospray instrument (ESI) with positive mode ionization for nucleobase and negative mode ionization for nucleoside 5′-triphosphate.

Synthesis of [1′-¹³C, 2-¹⁹F, 7-¹⁵N] Adenosine 5′-Triphosphate

The synthesis of the atom-specifically labeled [2-¹⁹F, 7-¹⁵N] adenine probe was carried out in 10 chemical steps, starting with the sodium ethoxide (C₂H₅ONa)-mediated cyclization of guanidine 1 and ethylcyanoacetate 2 to yield diaminopyrimidinone 3 (Scheme ). Nitrosylation of 3 with commercially available ¹⁵N-sodium nitrite (Na¹⁵NO₂) then installed ¹⁵N at the 5-position to afford compound 4.

1.

The nitroso group in 4 was reduced by sodium dithionite (Na₂S₂O₄) to form triaminopyrimidinone 5, which was then cyclized by reacting with formamide (CH₃NO) and formic acid (CH₃COOH) to provide guanine 6. Chlorination of 6 was achieved using the Vilsmeier–Haack reaction to yield intermediate 7 followed by subsequent reactions with CH₃COOH and sodium hydroxide (NaOH) to yield 8 and 9. The chloro group in 9 was converted to an azido group by the addition of sodium azide (NaN₃) to afford 10. In the second-to-last step, a diazotization reaction with tetrafluoroboric acid (HBF₄) provided fluorinated 11. Finally, the azido group in 11 was reduced by sodium borohydride (NaBH₄) to yield the desired [7-¹⁵N]-2-fluoroadenine 12. Taken together, our synthetic route provided 12 in 10 chemical steps with an overall yield of 5% (cf 12% for previously synthesized phosphoramidites and 5% nonhydrolyzable 2FATP) and without the need for chromatographic purification (Scheme ).

We then adapted our established one-pot enzymatic reaction ,,,, to couple the newly synthesized [7-¹⁵N]-2-fluoroadenine 12 with [1-¹³C]-d-ribose followed by subsequent phosphorylation to the corresponding adenosine 5′-triphosphate (ATP) (Scheme ). In brief, ribokinase (RK, E.C. 2.7.1.15) phosphorylates [1-¹³C]-d-ribose 13 at its position O5 to yield 14. Then, phosphoribosyl pyrophosphate synthetase (PRPPS, E.C. 2.7.6.1) pyrophosphorylates 14 at its O1′ site to form 5-phospho-d-ribosyl-α-1-pyrophosphate 15. Adenine phosphoribosyl transferase (APRT, E.C. 2.4.2.7) then facilitates the nucleophilic attack of the C1′ position of 15 by N9 of 12 to afford the 5′-monophosphate 16. Finally, myokinase (MK, E.C. 2.7.4.3) phosphorylates 16 to yield 5′-diphosphate 17, which is phosphorylated a final time by creatine kinase (CK, E.C. 2.7.3.2) to form the desired [1′-¹³C, 2-¹⁹F, 7-¹⁵N] ATP 18 in 95% yield (Scheme ). The characterization of compounds 9, 10, 11, 12, and 18 is shown in the Supporting Information (Figures S1–S11).

2.

Results

NMR Spectroscopy of Atom-Specifically Labeled RNA

We synthesized building block 18 for NMR analysis of RNA. To this end, we used our atom-specifically labeled 18 along with unlabeled guanosine (G), cytidine (C), and uridine (U) 5′-triphosphates to make a 27-nucleotide (nt) fluorinated RNA construct by DNA-templated RNA polymerase-based in vitro transcription (Figure A). This RNA is derived from the aminoacyl site of the rRNA in E. coli and is therefore critical for protein synthesis and bacterial survival. We will refer to this RNA as the A-site throughout the text. A one-dimensional (1D) ¹⁹F NMR spectrum of 18-labeled RNA showed all four fluorine signals corresponding to each of the modified adenosine nucleotides although two peaks downfield overlap with each other (Figure B). As a control (Figure A), we made a nonfluorinated (i.e., unmodified) RNA construct with our recently reported [2-¹³C, 7-¹⁵N]-adenosine into [1′,2-¹³C₂, 7-¹⁵N]-ATP 19.

1.

Validation of atom-specifically labeled RNA constructs. (A) Secondary structure of the 27-nt A-site RNA made from in vitro transcription with either 18 (left) or 19 (right) incorporated. Nucleotides labeled with 18 or 19 are numbered and shown in green or black, respectively. (B) 1D ¹⁹F NMR spectra at 298 K of the 18-labeled A-site with (magenta) or without (green) paromomycin. (C) Optical melting data from the 18- and 19-labeled A-site with the extracted melting temperature (T _m) and Gibbs free energy of folding at 298 K (ΔG^298K). Comparison of the imino region of ¹H NMR for (D) 19-labeled A-site RNA and (E) 18-labeled A-site RNA with and without paromomycin at 298 K. ¹H–¹H NOESY NMR at 283 K showing imino–imino ¹H cross peaks in (F) 19-labeled A-site RNA and (G) 18-labeled A-site RNA. Asterisks (*) indicate proton signal with the largest chemical shift perturbation between 18- and 19-labeled A-site RNA as confirmed by imino–imino NOE contacts in F and G.

The H2 position in purines is sensitive to changes in chemical environment such as drug binding due to its direct involvement in π-stacking interactions in an RNA. Given that fluorine is also extremely sensitive to changes in its environment owing to the uneven distribution of electrons in its valence p-orbitals, the substitution of ¹H with ¹⁹F at the 2-position should be an effective probe to validate binding of cognate ligands. Not surprisingly, we observed significant chemical shift perturbations in the ¹⁹F NMR spectrum upon addition of the well-characterized bulge-binding A-site ligand paromomycin (PAR) to 18-labeled RNA (Figure B).

Optical melting experiments with these two RNAs confirmed that ¹⁹F labeling did not significantly perturb A-site thermal stability or folding thermodynamics (Figure C), in agreement with previous reports. Optical melting experiments provided overall thermodynamics comparison between the fluorinated and nonfluorinated RNAs. To ascertain local site-specific differences between both RNAs, we monitored changes in the H-bonded imino proton signals upon fluorination of the RNA. The imino proton of U18 (U18 H3), which is base-paired to A9 within the regular A-helix, appeared to be more shielded in the fluorinated RNA (Figure D,E). This is likely due to the electrostatic repulsions between the fluorine atom of fluorinated adenosine (F2) and oxygen (O2) of uridine. This observation agrees with previous reports on HIV TAR RNA. While we observed more imino ¹H signals in the fluorinated RNA, the NMR imino melt experiment revealed no differences in the melting profiles for both fluorinated and nonfluorinated RNAs (Figure S12).

Upon titration of PAR, both RNA constructs undergo CSPs (Figure D,E), indicating that the fluorinated RNA retained the binding properties of the nonfluorinated RNA (Figure B). Again, both RNA samples had nearly similar imino–imino proton NOE contacts (Figure F,G). For example, even though U18 H3 was more shielded in the fluorinated RNA, the pattern of its NOE cross peak with the preceding G17 H1 was comparable to that of the nonfluorinated RNA (Figure F,G).

As a second application, we implemented two-dimensional (2D) HSQC experiments using one (¹ J _H1′C1′)- and two (² J _H8N7)-bond couplings for coherence transfer in both fluorinated (i.e., 18-labeled) and nonfluorinated (i.e., 19-labeled) RNA. As expected, all four H8–N7 (Figure A) and H1′–C1′ (Figure B) resonances were observed. In the ¹H–¹⁵N experiment, fluorine incorporation deshielded the ¹⁵N–N7 atoms but shielded the ¹H atoms (Figure A). Nucleotides A20 and A21 exhibited the largest CSPs between the fluorinated and nonfluorinated RNAs in both ¹⁵N and ¹³C HQSCs (Figure C,D).

2.

Carbon and Nitrogen NMR parameters of fluorinated and nonfluorinated RNA. (A) 2D ¹H–¹⁵N HSQC experiments using the two-bond (² J _H8N7) coupling for coherence transfer (inset) to show all H8–N7 resonances. (B) 2D ¹H–¹³C HSQC experiments using the one-bond (¹ J _H1′C1′) coupling for coherence transfer (inset) to show all four H1′–C1′ resonances. Both HSQC experiments were conducted at 298 K. Chemical shift differences were calculated as Δδ = $\sqrt{{(\mathrm{\Delta }{\delta }_{\mathrm{H}})}^2+{(\alpha \mathrm{\Delta }{\delta }_{\mathrm{N/C}})}^2}$ between 18- and 19-labeled RNAs in (C) ¹H–¹⁵N and (D) ¹H–¹³C HSQCs, where Δδ_H and Δδ_N/C are the changes in proton and nitrogen/carbon chemical shifts, respectively, and α is the ratio of the gyromagnetic ratio between nitrogen/carbon and proton. All NMR spectra are annotated with resonance assignments.

To qualitatively assess the effect of fluorine on the spectroscopic properties of nearby nuclei in more detail, we repeated our HSQC experiments after titration with PAR and at a reduced temperature. Any ¹⁹F-induced effect on relaxation will be amplified when bound to PAR or at lower temperatures. Indeed, titration with PAR resulted in the disappearance of two H8–N7 resonances in the nonfluorinated RNA spectrum (Figure A) while only a single signal was lost in the fluorinated construct (Figure B). In the case of H1′–C1′ resonances, only one nonfluorinated RNA signal was lost upon PAR titration (Figure C), while the fluorinated RNA retained all four resonances (Figure D). However, in all cases, we observed attenuation of the signal intensity of A20 or the complete disappearance of its signal. This agrees with previous reports that have shown nucleotide A20 to be a dynamic hotspot for aminoglycoside binding. At a lower temperature of 10 °C, only one WT H8–N7 resonance disappeared (Figure S13A), while all four signals were retained in the fluorinated construct (Figure S13B). The H1′–C1′ resonances, on the other hand, were mostly unaffected (Figure S13C,D).

3.

Ligand titrations in fluorinated and nonfluorinated RNA. 2D ¹H–¹⁵N HSQC experiments with PAR titrated against (A) 19-labeled and (B) 18-labeled A-site. 2D ¹H–¹³C HSQC experiments with PAR titrated against (C) 19-labeled and (D) 18-labeled A-site. All HSQC experiments were conducted at 298 K. All NMR spectra are annotated with resonance assignments.

We also performed a 2D ¹H–¹H total correlation spectroscopy (TOCSY) experiment to reveal all uridine and cytidine H5/H6 cross peaks. Given our previous findings that fluorination of adenosine at the 2-position alters the chemical environment of H8–N7 and H1′–C1′ resonances, we hypothesized that fluorine may perturb additional nearby resonances relative to the unfluorinated RNA. Given that all adenosine nucleotides are in or near the bulge, any ¹⁹F-induced chemical shift perturbations (CSPs) would likely affect uridine and cytidine nucleotides that are near this structural motif. Indeed, comparison of fluorinated and nonfluorinated spectra shows CSPs for H5/H6 nuclei in bulge nucleotides C6 and C8, as well as the adjacent but overlapping U5/C10/U23 (Figure A). However, CSPs were also observed for nucleotides C10–U12 that extend into the top of the upper helix and into the tetraloop.

4.

TOCSY experiments in fluorinated and nonfluorinated RNA. (A) 2D ¹H–¹H TOCSY experiments in the 18-labeled and 19-labeled A-site. (B) 2D ¹H–¹H TOCSY experiments with PAR titrated against the 19-labeled A-site. (C) 2D ¹H–¹H TOCSY experiments with PAR titrated against the 18-labeled A-site. All TOCSY experiments were conducted at 298 K. All NMR spectra are annotated with resonance assignments. CSPs mapped onto the secondary structure of the 27-nt A-site RNA are shown to the right of each respective spectrum.

To further explore this observation that ¹⁹F-induced CSPs localize mainly to bulge nucleotides, we titrated PAR against WT RNA. Again, paromomycin binds to the A-rich bulge region of the nonfluorinated E. coli rRNA. Here, we observed a set of CSPs similar to those of bulge nucleotides C6, C8, and U5/C10/U23 (Figure B). As before, CSPs to upper helix nucleotides C10 and C11 were also observed, as were CSPs to lower helix nucleotides C3, C24, and C26 (Figure B). As with the WT RNA, a very similar set of CSPs was observed for the fluorinated RNA upon titration of PAR (Figure C). Taken together, these data establish a set of shared ¹⁹F- and PAR-induced CSPs that correspond to H5/H6 nuclei from uridines and cytidines in and around the bulge region (Figure ). Given that only adenosines were fluorinated (18-labeled RNA), we expected the most significant CSPs for nucleotides near adenosines. Similarly, the binding pocket for PAR in the nonfluorinated RNA (19-labeled RNA) is the A-rich bulge. Hence, we also expected major CSPs upon PAR binding to be localized to the nucleotides near adenosines.

Chemical Shift Assignments and Structure of 18-Labeled RNA

Peak patterns observed in both ¹H–¹³C and ¹H–¹⁵N HSQCs (Figure A,B) were offset but identical for both RNA constructs (i.e., 18-labeled and 19-labeled RNAs). Hence, we performed a series of experiments to assess our preliminary assignments for the fluorinated RNA, which was solely based on the peak patterns for the WT RNA.

We performed a ¹H–¹H NOESY experiment on the fluorinated RNA to obtain intranucleotide adenosine H8–H1′ correlations (Figure A). However, given severe signal overlap among aromatic, ribose, and H5 of pyrimidine hydrogen atoms, we employed a series of ¹³C- and ¹⁵N-edited experiments to simplify data analysis. ¹⁵N7-edited NOESY provides both intranucleotide H8–H1′ cross peaks and internucleotide cross peaks between H8 and neighboring protons in the 3′ to 5′ direction. Similarly, ¹³C1′ edited NOESY produces intranucleotide H1′–H8 cross peaks and internucleotide cross peaks between H1′ and neighboring aromatic protons in the 5′ to 3′ direction. A set of shared cross peaks between both ¹³C1′- and ¹⁵N7-edited NOESY experiments unambiguously represents intranucleotide H1′–H8 correlation (Figure S14). In addition, ¹³C1′- and ¹⁵N7-edited NOESY experiments are complementary techniques in that they permit analysis of internucleotide correlations in either 5′ to 3′ or 3′ to 5′ directions, respectively.

5.

(A) Regular ¹H–¹H NOESY spectrum of 18-labeled RNA with a mixing time of 250 ms. Red and blue shaded regions in A are edited by ¹³C and ¹⁵N-nuclei, respectively, for facile analyses in (B–D). (B) ¹³C1′-edited NOESY HSQC showing both intramolecular H8–H1′ contacts in adenosine and some intermolecular contacts to adenosine H8. (C) ¹⁵N7-edited NOESY-HSQC showing both intramolecular H8–H1′ contacts in adenosine, and some intermolecular contacts to adenosine H8. Asterisks (*) indicate shared cross peaks between ¹⁵N7- and ¹³C1′-edited NOESY experiments. (D) ¹³C1^′ TOCSY-HQSC showing H1′–H2′ correlations within adenosines with the 2′-endo sugar pucker conformation. All experiments were conducted at 298 K.

First, we performed both ¹³C1′- and ¹⁵N7-edited NOESY HSQC experiments that enabled facile correlation of ¹³C1′ to their corresponding intranucleotide ¹⁵N7 between their shared proton resonances, even in the absence of chemical shift assignments (Figure S15). These correlations (Figure S15) corroborate our preliminary assignments for the fluorinated RNA (Figure A,B). Furthermore, we observed the expected set of shared cross peaks between ¹⁵N7- and ¹³C1′-edited NOESY experiments (highlighted by the magenta asterisks*), indicating intranucleotide H1′–H8 correlations within the adenosines (Figure B,C). Additionally, we observed internucleotide cross peaks in both ¹⁵N7- and ¹³C1′-edited NOESY experiments (Figure B,C).

Second, we performed an H1′–H2′ TOCSY experiment that provides structural information about sugar puckering in a nucleotide. For nucleotides in 2′-endo (B-DNA like) conformation, H1′ and H2′ have sizable three-bond scalar coupling (³ J _H1′H2′ ≈8 Hz), while those in 3′-endo (RNA like) have small scalar coupling (³ J _H1′H2′ < 2 Hz). Hence, nucleotides that show H1′–H2′ TOCSY cross peaks in RNA are more likely to be in nonhelical regions of RNA. Again, this experiment was limited to only adenosine using ¹³C1′ editing. Out of the four adenosines in the fluorinated RNA, we observed only two H1′–H2′ TOCSY cross peaks (Figure D). These correspond to nucleotides A21 and A20, located in the A-rich bulge. H1′–H2′ dihedral angles measured for A21 and A20 in the deposited PDB structure (PDB ID 1PBR) were 155.2° and 101.2°, respectively. Interestingly, these were the nucleotides that exhibited the largest CSPs in Figure C,D. This raises the following question: Is there a link between CSPs observed in Figure C,D and nucleotide flexibility?

¹³C and ¹⁵N NMR Spin-Relaxation Experiments

Given the anticipated flexibility of the bulge, we used our newly synthesized probe 18 to monitor the RNA dynamics. Such RNA motions at different time scales are implicated in its various cellular processes such as ligand sensing and signaling, viral genome packaging, and catalysis. Longitudinal (R ₁) and transverse (R ₂) relaxation rates measure the rates at which magnetization decays along the longitudinal and transverse planes, respectively. In addition, R ₂ is directly proportional to the line width of an NMR signal and the spin-relaxation rate in a rotating frame (R _1ρ).

While ¹H and ¹⁹F are very sensitive to external magnetic fields, they become challenging to work with at high magnetic fields and molecular weights due to their fast relaxation rates from dipolar interactions and CSA, respectively. Nonetheless, extensive relaxation experiments have been conducted on ¹H and ¹⁹F for small protein and RNA constructs. ,

The ¹H8–¹⁵N7 spin pair is isolated from other active nuclei in our site-specific label, making it an attractive system to probe. Our simulations for an RNA with a correlation time of 5 ns at 600 MHz showed that an isolated ¹H1′–¹³C1′ spin pair exhibits ∼2-fold broader line widths for ¹³C1′ nuclei compared with ¹⁵N7 nuclei in an isolated ¹H8–¹⁵N7 spin pair (Figure A). In agreement with our simulations, the average experimental ¹⁵N line width was ∼2× narrower than ¹³C (Figure B). Additionally, we observed attenuated R _1ρ values for A21 and A20 in both ¹³C and ¹⁵N R _1ρ experiments, indicating increased internal motions within both nucleotides. We observed the same trends for R _1ρ in both the ¹⁵N7 and ¹³C1′ relaxation rates (Figure B). The measured ¹³C1′ R _1ρ values were identical for corresponding adenosines in the fluorinated and nonfluorinated RNAs (Figure B), except for A21.

Interestingly, there was no significant difference between measured R ₁ values (Figure C). Like R ₂, the attenuated R ₂/R ₁ ratios indicate fast local motions, with the added benefit of removing variations in local CSAs and dipolar interactions. Again, both A20 and A21 had the smallest R₂/R ₁ values, indicating fast motions within those nucleotides (Figure D).

Discussion

¹H NMR is routinely used for structural studies due to its high sensitivity to magnetic fields. However, the use of ¹H NMR in biomolecular systems is limited owing to low chemical shift dispersion and line broadening. ¹⁹F, on the other hand, has better chemical shift dispersion in addition to its high sensitivity. Unfortunately, the high CSA of ¹⁹F also limits its application in larger biomolecules at high magnetic fields due to large line broadening. Conversely, ¹³C and ¹⁵N have lower gyromagnetic ratios, which render them less sensitive to external magnetic fields, yet possess sharper NMR line widths. A chemical probe that can establish a “middle ground” by leveraging the benefits of all three nuclei shows promise to be versatile for studying both small and large RNAs.

First, we demonstrated in a series of NMR experiments that the incorporation of ¹⁹F into RNA is not completely innocuous. While we do not observe significant changes in thermodynamic properties between 18- and 19-labeled RNA, we observe these changes in the magnetic properties of nuclei proximal and distal to the site of fluorination. Furthermore, these changes are not limited to the fluorinated nucleotide but are also seen in neighboring nuclei, as evidenced by the ¹H–¹H TOCSY experiment. Additionally, ¹⁹F can simplify an otherwise highly overlapped spectrum. Consequently, we showcase the ability of ¹⁹F to detect the binding of cognate ligands to RNA by monitoring the CSPs. It is worth noting that the application of ¹⁹F is not limited to only small RNAs, as recent studies have shown that installing ¹³C–¹⁹F pairs in RNA nucleobases reduces the directly attached ¹³C line width by a factor of 2–5, allowing its application to large RNAs. , A suite of elegant experiments has also been designed using ¹⁹F as a probe to measure various protein dynamics properties that can be implemented in RNA studies.

Second, we demonstrated the application of the site-specific label to simplify an otherwise overlapped ¹H–¹H NOESY spectrum. We observed all adenosine intranucleotide H8–H1′ cross peaks in both ¹³C- and ¹⁵N-edited NOESY experiments that were subsequently used to confirm preliminary chemical shift assignments. In addition, we observed TOCSY cross peaks for H1′- H2′ correlations in the ¹³C-edited TOCSY HSQC experiment. The presence/absence of these correlations provides structural information about the ribose sugar pucker. Taken together, our labeling scheme enables complementary probing of structural features within the RNA in the 5′ to 3′ or 3′ to 5′ direction.

Finally, we performed spin-relaxation experiments on the 18-labeled RNA. We obtained ∼2-fold sharper lines for ¹⁵N7 compared with ¹³C1′ from our simulations for an RNA tumbling at 5 ns at a magnetic field strength of 600 MHz. Our experimental results corroborated this prediction, where we measured ∼2-fold increase in the average line widths of ¹³C1′ compared with ¹⁵N7, under the same experimental conditions. We also show that there is no significant difference in the transverse relaxation rates for the fluorinated and nonfluorinated RNAs, except for nucleotide A21.

Taken together, we have demonstrated the utility of the novel triple labeling pattern to obtain invaluable structural and dynamics information on a 27-mer RNA. Our data also suggest that there may be a link between chemical shift changes induced by ¹⁹F incorporation and nucleotide flexibility for the RNA system we investigated. While we did not fully leverage the capability of ¹⁹F in this study, we intend to do so by enriching the C2 position with ¹³C, as reported recently. This, in turn, should produce very sharp lines in ¹³C as the two relaxation mechanisms of dipole–dipole relaxation and chemical shielding anisotropy almost cancel out.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsbiomedchemau.5c00075.

• Detailed description of chemical synthesis and characterization data; additional NMR data for RNA; additional resonance assignment strategies; chemical shift tensors and atomic coordinates for theoretical simulations (PDF)

S.K.A.: Conceptualization, investigation, writing. L.T.O.: Conceptualization, investigation, and writing. F.P.S.: Investigation and writing. T.K.D.: Conceptualization, investigation, writing, and supervision.

We acknowledge support from NIH (U54 AI17660-TKD) and NSF (DBI1040158-TKD).

The authors declare no competing financial interest.

Published as part of ACS Bio & Med Chem Au special issue “Juneteenth 2025”.