Compaction of RNA Duplexes in the Cell

Abstract The structure and flexibility of RNA depends sensitively on the microenvironment. Using pulsed electron‐electron double‐resonance (PELDOR)/double electron‐electron resonance (DEER) spectroscopy combined with advanced labeling techniques, we show that the structure of double‐stranded RNA (dsRNA) changes upon internalization into Xenopus lævis oocytes. Compared to dilute solution, the dsRNA A‐helix is more compact in cells. We recapitulate this compaction in a densely crowded protein solution. Atomic‐resolution molecular dynamics simulations of dsRNA semi‐quantitatively capture the compaction, and identify non‐specific electrostatic interactions between proteins and dsRNA as a possible driver of this effect.


General materials and methods
All commercially available reagents were purchased from Sigma-Aldrich, Acros Organics or Fluka and used without further purification. 2′-O-methyluridine was purchased from Rasayan Inc. USA. All moisture-and air-sensitive reactions were performed in flame-dried glassware under a positive pressure of nitrogen. Solvents were distilled prior to use and stored over activated molecular sieves (4 Å) under nitrogen. Water was purified by a Milli-Q water purification system. Analytical thin layer chromatography (TLC) was performed on silica gel glass plates (Silicycle, ultra-pure silica gel, 60Å, F254); TLC visualization was performed with UV light or by staining with p-anisaldehyde. Flash-column chromatography was performed on silica gel (Silicycle, 230-400 mesh, 60Å). 1 H-spectra were recorded using deuterated solvents as internal standards on a Bruker Advance 400 spectrometer; chemical shifts are reported in ppm. Residual proton signals from the deuterated solvents were used as references [d6-DMSO (2.50 ppm), CDCl3 (7.26 ppm)], for 1 H spectra. 13 C NMR chemical shifts are reported in reference to undeuterated residual solvent (CDCl3 (77.0 ppm), d6-DMSO (39.43 ppm). 31 P NMR chemical shift is reported relative to 85% H3PO4 as an external standard. Commercial grade CDCl3 was passed over basic alumina shortly before use, to record the 1 H-NMR and 31 P-NMR of phosphoramidite 8. Mass spectrometric analyses of all organic compounds were performed on an ESI-HRMS (Bruker, microTOF-Q) in positive or negative ion mode. RNA oligonucleotides were synthesized on an automated ASM800 DNA/RNA synthesizer (Biosset, Novosibirsk, Russia) by using a trityl-off protocol and phosphoramidites with standard protecting groups on a 1.0 mmol scale, using 1000 Å CPG columns. All commercial phosphoramidites, CPG columns, and solutions were purchased from ChemGenes Corporation (Wilmington, MA). Upon completion of synthesis, the oligoribonucleotides were cleaved from the solid support and the nucleobases and phosphodiesters deprotected in a 1:1 mixture of conc. aqueous NH3 and 8 M MeNH2 in EtOH (2 mL) at 65 °C for 40 min. The supernatant was collected, the bead was washed three times with a mixture of EtOH:H2O (1:1, 300 µL), and the combined washings were dried. The 2′-O-TBDMS groups were removed by treatment with a mixture of Et3N•3HF:DMF (3:1, 800 µL) at 55 °C for 1.5 h, followed by addition of H2O (200 µL). This mixture was transferred to a 50 mL Falcon tube and n-butanol (40 mL) was added and stored at -20 °C for 12 h, centrifuged and the solvent decanted from the RNA pellet. The crude RNA was subsequently purified by 20% denaturing polyacrylamide gel electrophoresis (DPAGE). The RNA oligonucleotide bands were visualized under UV light, excised from the gel, crushed, and eluted from the gel with a Tris buffer (2×10 mL; Tris (10 mM, pH 7.5), NaCl (250 mM), Na2EDTA (1 mM)). The RNA elutions were filtered through a 0.45 mm cellulose acetate membrane (Whatman) and desalted using a Sep-Pak cartridge (Waters Corporation). The dried oligoribonucleotides were dissolved in sterile H2O (400 µL) and their final concentrations were calculated according to Beer's law based on UV absorbance of oligoribonucleotides at 260 nm. Extinction coefficients were determined by using the UV WinLab oligoribonucleotide calculator (V2.85.04; Perkin Elmer). Molecular weights of oligoribonucleotides were determined by MALDI-TOF analysis (Bruker, Autoflex III) after calibration with an external standard. UV/vis spectra were recorded on a PerkinElmer Lambda 25 UV/vis spectrometer.
Preparation of RNA containing the E Im Um nitroxide E Im Um was incorporated into RNA oligonucleotides by solid phase synthesis using the previously reported protocol, [1] with a slight modification. The activator 5-(benzylthio)-1H-tetrazole, usually used for RNA synthesis was not suitable for E Im Um incorporation in oligoribonucleotides, as we observed a coupling efficiency of less than 5% (as judged by the color during trityl deprotection after coupling of E Im Um phosphoramidite). Therefore, 5-(ethylthio)-1H-tetrazole, the activator generally used in our laboratory for DNA synthesis, was used instead. The spin-labeled phosphoramidite was incorporated manually into the oligoribonucleotides by pausing the synthesizer program after completion of the prior cycle, removing the column from the synthesizer, and running the standard activator solution (200 μL) and a solution of the spin-labeled phosphoramidite (0.05 M, 200 μL) prepared in 1,2-dichloroethane back and forth through the column for ~10-12 min. After manual coupling, the column was remounted on the synthesizer and the synthesis cycle was completed.

Preparation of RNA containing the E-TU nitroxide
The procedure was essentially the same as previously described, [2] which entailed reacting the 2′-amino uridine-modified RNA with the isothiocyanate in borate buffer (100 mM, pH 8.6) containing DMF (50%) at 37 °C for 8 h. After EtOH precipitation, the spin-labeled product was purified by 20% DPAGE followed by another round of EtOH precipitation to obtain the E-TU-labeled RNA oligonucleotide.
1 H NMR spectrum of 2 13 C NMR spectrum of 2 1,1,3,3-tetraethylisoindoline-5,6-diamine (3). To a solution of compound 2 (400.0 mg, 1.37 mmol) in MeOH (16 mL) was added 10% Pd/C (40.0 mg) and the mixture was hydrogenated at 55 psi hydrogen pressure at 23 °C for 2 h. The reaction mixture was filtered through a pad of celite and the filtrate was concentrated in vacuo to yield 1,1,3,3-tetraethylisoindoline-5,6-diamine (358.0 mg, 100%), which was used in the next reaction without further purification. were subsequently added. The reaction mixture was stirred at 23 °C for 16 h, diluted with CH2Cl2 (10 mL) and washed successively with saturated aq. NaHCO3 (3 x 10 mL) and saturated aq. NaCl (2 x 10 mL). The combined organic layers were dried over anhydrous Na2SO4, filtered and concentrated in vacuo. The phosphoramidite was purified by dissolving the residue in CH2Cl2 (0.5 mL) and precipitation by addition of petroleum ether (50 mL). The solvent was decanted and the operation repeated three times to furnish phosphoramidite 8 as a yellowish solid (60 mg, 48%), which was used for RNA synthesis without further purification.

MALDI-TOF MS analyses of oligoribonucleotides
The incorporation of E Im Um into oligoribonucleotides was confirmed by MALDI-TOF MS analysis. The instrument was calibrated with an external standard prior to measurements. The calculated and observed monoisotopic masses of oligoribonucleotides are listed below in Table S1.

Thermal denaturation experiments
To determine if the E Im Um affected the stability of RNA duplexes, the thermal denaturation curves of both unmodified and spin-labeled oligomers were determined and compared. The RNA oligonucleotides (4.0 nmol of each strand) were dissolved in a phosphate buffer (100 µL; 10 mM phosphate, 100 mM NaCl, 0.1 mM Na2EDTA, pH 7.0), annealed and diluted to 1.0 mL with the phosphate buffer and degassed with argon. The samples were heated up from 20 °C to 90 °C (1.0 °C/min) and absorbance at 260 nm was recorded at 1.0 °C intervals ( Figure S1). Incorporation of two E Im Um modifications into the 20-mer oligoribonucleotide stabilized the duplex by ca. 4.0 °C. Figure S1. Thermal denaturation curves for E Im Um-containing oligoribonucleotides. Black trace: unmodified duplex; red trace: E Im Um-modified duplex.

List of the RNA duplexes
In this work the following RNA duplexes were used:

Preparation of the samples for in vitro EPR spectroscopy
In vitro samples for pulsed EPR spectroscopy were prepared at a 100 µM concentration of the doubly spin-labeled RNA duplexes in PNE buffer (10 mM phosphate buffer pH 7.0, 100 mM NaCl, 0.1 mM EDTA) containing 20% v:v glycerol as glass-forming agent. For the samples in cytoplasmic extract or in concentrated protein solutions (200 mg/mL lysozyme or bovine serum albumin prepared in water/glycerol 8:2 v:v), a lyophilized aliquot of a 100 µM stock solutions of the duplexes in PNE buffer was re-dissolved at a final duplex concentration of 100 µM; in the case of the cytoplasmic extract (vide infra), glycerol was added to a final concentration of 20% v:v. All the aforementioned solutions were prepared using DEPC-treated water.
The samples for Q-and X-band EPR spectroscopy were loaded in 1.1 mm ID × 1.6 mm OD or 1.75 mm ID × 2.80 mm OD Suprasil capillaries, respectively, and flash frozen in liquid nitrogen; typical volumes were 5 µL for Q-band samples and 20 µL for X-band samples.
A sample for room-temperature continuous-wave X-band EPR spectroscopy of RNA II was prepared at a 100 µM concentration of the duplex in PNE buffer and loaded in a 20 µL BLAUBRAND ® intraMark micropipette.

Preparation of the cytoplasmic extract
The cytoplasmic extract was prepared according to Serber et al. [6] Briefly, approximately 500 oocytes were transferred to XB buffer (100 mM KCl, 0.1 mM CaCl2, 1 mM MgCl2, 10 mM K + -HEPES, 50 mM sucrose) supplemented with 2% wt/vol cysteine and swirled for approximately 30 minutes. The oocytes were subsequently washed first with ORi buffer and then with CSF-XB buffer (XB buffer containing 2 mM MgCl2 and 5 mM EGTA). After having transferred the oocytes into a 2 mL polypropylene microcentrifuge tube, the excess buffer was removed by pipetting; cytochalasin B, leupeptin, pepstatin A and chymostatin were added at a final concentration of 10 µg/mL. The oocytes were subsequently packed by centrifugation at 400 g and 4°C for 1 minute. The excess buffer was removed, and the cells were crushed by centrifugation at 16100 g and 4°C for 1 hour. The crude interphase extract was eventually removed by piercing the side of the tube with a needle, aliquoted, shock-frozen in liquid nitrogen and stored at -80°C. Estimation of the protein concentration in the extract was performed using a colorimetric assay (Sigma Aldrich's Advanced Protein Assay Reagent, #57697). The determined values lie around 20 mg/mL; in comparison, the total protein concentration in Xenopus laevis oocytes is expected to be about 125 mg/mL assuming a cellular volume of 1 µL. [7] Internalization of the RNA duplexes inside Xenopus laevis oocytes and preparation of the in-cell EPR samples For the preparation of the samples for the in-cell study, the doubly spin-labeled RNA duplexes were microinjected into mature, defolliculated Xenopus laevis oocytes at the stage V/VI. The oocytes were injected with 50.6 nL of a concentrated (0.4 mM -1.0 mM) stock solution of the RNAs in aqueous buffer. The microinjection always took place starting from the animal hemisphere slightly above the vegetal hemisphere and pointing the needle towards this latter; the exogenous RNA is therefore expected to be delivered to the cytoplasm. The amount of microinjected oocytes was typically 10 for Q-band EPR samples and 50 for pulsed X-band EPR samples. After incubation of the injected oocytes at 16°C in ORi buffer (5 mM HEPES pH = 7.6, 110 mM NaCl, 5 mM KCl, 2 mM CaCl2, 1 mM MgCl2) for a period of one hour, the cells were loaded into an EPR quartz capillary (1.4 mm ID × 1.6 mm OD for Q-band EPR samples, 1.75 mm ID × 2.80 mm OD for pulsed X-band EPR samples). Immediately after preparation, the samples were shock-frozen in liquid nitrogen. The in-cell EPR samples were afterwards stored in liquid nitrogen. The number of oocytes inside the active volume of the used EPR resonators (vide infra) is approximately 3 for Q-band samples and 6 for X-band samples.
The sample for room-temperature continuous-wave X-band EPR spectroscopy was prepared in an analogous way by microinjecting 50 oocytes with 50.6 nL of a 1.0 mM solution of RNA II; the cells were loaded into a 1.4 mm ID × 1.6 mm OD EPR quartz capillary and the sample was measured immediately after its preparation.

X-band room-temperature CW-EPR spectroscopy
The in vitro X-band CW-EPR spectrum of RNA II (Figure S2A, black trace) was recorded at room temperature on a Bruker Elexsys E500 spectrometer equipped with a Bruker ER4122SHQE resonator carrying a quartz Dewar insert. The 20 µL micropipette was inserted for the purpose of the measurement inside a 2 mm ID × 3 mm OD quartz capillary. Relevant acquisition parameters were: microwave power = 2 mW, modulation amplitude = 1. The measurements confirm the stability of the shielded nitroxide in the intracellular environment. The comparison between the spectra plotted in Figure S2A shows furthermore that no significant strand dissociation takes place inside cells.

Echo-detected field-sweep and 2-pulse echo decay experiments at X-and Q-band
In vitro-and in-cell samples of RNA I (Table S2) were characterized by X-and Q-band pulsed EPR spectroscopy. Echo-detected field sweep (EDFS-EPR) experiments (Figures S3A, S3C) were performed by recording the amplitude of the spin echo generated by a π/2 -τ -π -τecho sequence while sweeping the magnetic field. The length of the π/2 and π pulses was set to 16 ns and 32 ns, respectively, and the inter-pulse delay τ was set to 200 ns. The spin echo was integrated over its whole width, using a 100 ns-wide range. A two-step phase cycle [8] (x)x was applied. 2-pulse echo decay experiments (Figures S3B, S3D) were performed at the maximum of the EPR spectrum with the same sequence used for the echo-detected field sweep experiments and varying the length of the inter-pulse delay τ starting from a minimum value of 200 ns. We also collected the background signal coming from the oocytes (blue traces in Figure S3). The background consists of a broad component, originating from the Mn 2+ endogenously present in the cells [9] and visible especially at Q-band frequencies owing to the narrowing of the mS = -½  +½ transitions with respect to X-band frequencies, and of a narrow component in the g ~ 2 region, which according to its nutation behavioris assigned to a S = ½ spin species. Because the phase memory time of this latter (blue traces in Figures S3B, S3D) is much shorter compared to the typical delays used for the PELDOR experiments (vide infra), no contribution to the echo of the observed spins is expected. As the overlap between the excitation profile of the pump pulse (FWHM = 67 MHz, corresponding to approximately 24 G, at both X-and Q-band frequencies) and the two components of the background signal is not significant compared to the overlap with the spectral profile of the nitroxide, we also rule out an involvement of the endogenous paramagnetic centers in the observed decrease of the inter-spin distance upon internalization of the RNA duplexes inside the cells.

Q-band PELDOR spectroscopy of RNA I
Q-band PELDOR experiments on RNA I (Table S2) were performed using the 4-pulse sequence [10] π/2 -τ1 -π -τ1+t -πp -τ2-t -π -τ2echo whereby eleven traces with τ1 increments of 2 ns were added, typically starting from a τ1 value of 300 ns. A two-step phase cycling [8] (x)yy was applied to the detection sequence. For the pump inversion pulse πp a 12-ns rectangular pulse generated from a second, incoherent microwave source was used; the position of this pulse was set to match the maximum of the EPR spectrum. The remaining three pulses were also rectangular, all with a length of 22 ns, and had a frequency offset of -80 MHz with respect to the pump pulse ( Figure S3A). The PELDOR traces were analyzed using the model-free approach implemented in the MATLAB toolbox DeerAnalysis 2019. [11]  For the PELDOR traces shown in Figure 2A a validation of the analysis procedure was performed in order to confirm the statistical significance of the reduction of the inter-spin distance upon internalization of RNA I inside cells; the results are shown in Figure S5. The validation of the results of the model-free analysis (Figure S5D) was performed using the built-in validation tool of DeerAnalysis. The region for the fit of the intermolecular contribution, whose end always coincides with the end of the PELDOR traces, was initially set to the full length of the non-oscillatory part of the raw data and systematically decreased over 50 steps till a minimum length of 750 ns was reached. The model-based analysis of the data (Figure S5E) was performed with a home-written MATLAB-based program that uses a downhill simplex approach to minimize the root-mean-square deviation to the experimental raw data of a calculated time trace V(t) = Vintra(t)·Vinter(t) corresponding to a two-Gaussian distance probability distribution and an exponentially decaying intramolecular signal; the inclusion of this latter in the simulated data overcomes the issues generated by an incorrect a priori processing of the experimental data. For each trace 10000 optimization runs were performed using randomly generated starting points; the 50 runs giving the best fit of the experimental traces were used to estimate the range of variation of the distance probability distribution, displayed as colored areas in Figure S5E, and of the fitting parameters. Additional components in the distance probability distributions in the region around 4.5 -5.0 nm could be observed for all the tested conditions with both the model-free and model-based analyses of the PELDOR traces (Figures S4C, S5E). The blunt ends in the RNA could have caused some end-to-end stacking of the duplexes, [12] and this could be related to the aforementioned long distances. However, a more definitive assignation is not possible because distances above 4.6 nm cannot be determined accurately with a maximum extension of the dipolar evolution window of 3 µs, [13] X-band PELDOR spectroscopy of RNA I X-band PELDOR experiments on RNA I (Table S2) were performed using the 4-pulse sequence [10] π/2 -τ1 -π -τ1+t -πp -τ2-t -π -τ2echo whereby τ1 was incremented in 8-ns steps from 300 ns to 364 ns to average out proton nuclear modulation. All the pulses were generated from the same microwave source; an eight-step phase cycling [8] (x)y[xp]ȳ was applied. [14]  For the pump inversion pulse πp a Gaussian pulse [15] with a FWHM of 8.78 ns and a length of 28 ns was used; the position of this pulse was set to match the maximum of the EPR spectrum. The remaining three pulses were also Gaussian, all with a FWHM of 23.4 ns and a length of 72 ns, and had a frequency offset with respect to the pump pulse between 40 MHz (Figure S6, red traces) and 90 MHz ( Figure S6, blue traces); for the samples in vitro and in lysozyme 6 traces were collected with a 10 MHz step, whereas for the in-cell sample 3 traces were collected with a 25 MHz step. The observed dependence of the intramolecular part of the PELDOR signal on the frequency offset between the pump and observe pulses for the in vitro sample (Figure S6B) reflects the presence of angular correlations between the spin labels, confirming their restricted mobility with respect to the RNA backbone. [16] This effect is substantially washed out both in lysozyme ( Figure S6D) and in cells (Figure S6F), indicating a more heterogeneous conformational ensemble of the RNA duplex in these environments. In order to extract the distance probability distributions, the traces measured at different frequency offsets were summed [17] (Figure S6, black traces) and the resulting traces were analyzed by Tikhonov regularization using the MATLAB toolbox DeerAnalysis 2019 ( Figure  S7). [11]  The very good comparison between the distance probability distributions obtained from the averaged X-band PELDOR data ( Figure  S7, thick colored lines) and the Q-band PELDOR data ( Figure S7, thin grey lines) confirms the absence of angular correlation effects in these latter.

Q-band PELDOR spectroscopy of the E-TU-labeled RNA duplexes
Q-band PELDOR experiments were performed on the 24mer duplexes RNA III and RNA IV ( Figure S8B) in vitro, in cell and in 200 mg/mL lysozyme to rule out the possibility that the spin label and/or the sequence of the duplex plays an explicit role on the detected change of the inter-spin distance upon internalization inside cells. The used RNA sequences carry furthermore 5'-U overhangs to prevent the end-to-end stacking of the duplexes. [12]  The duplex are labeled with the tetraethyl-shielded, thiourea-based spin label E-TU [2] (Figure S8A). The PELDOR experiments ( Figure  S9) were performed using the same settings described for RNA I.

Native polyacrylamide gel electrophoresis (PAGE)
Native polyacrylamide gel electrophoresis (PAGE) of dsRNA samples having different NaCl/lysozyme ratios was performed to test the role of electrostatic interactions in the binding of lysozyme to RNA. The 14mer RNA V (Table S2) was used for this experiment. A 0.75 mm-thick gel was prepared by polymerizing a 18% 19:1 acrylamide/bis-acrylamide solution (BioRad, Inc.) containing 1× TBE buffer (100 mM TRIS, 100 mM boric acid, 2 mM EDTA) with ammonium persulfate and tetramethylethylenediamine (TEMED); for the preparation of the gel, DEPC-treated water was used. The samples, each containing 2 µg of dsRNA, were loaded using a glycerol-containing buffer. A Bio-Rad Mini-PROTEAN Tetra Cell equipment was used to perform the electrophoresis.

Molecular dynamics simulations
As starting structure for the molecular dynamics simulations, an unlabeled double-stranded RNA A-helix with the sequence shown in Figure 1B of the main paper (RNA I; see also Table S2) was generated using the make-na server (structure.usc.edu/make-na/server.html). An NMR solution structure (PDB: 1E8L [18] ) was used as starting structure for hen egg-white lysozyme. The protein structure was subsequently protonated at pH = 7.0 using the propKa implementation of the Maestro Schroedinger software package (https://www.schrodinger.com/maestro). For simulations without lysozyme [-LYZ], the RNA duplex was inserted into a simulation box, solvated with the TIP3P [19] or TIP4P-D water model [20] and Na + and Clions [21] at 150 mM concentration. For simulations with lysozyme [+LYZ], 20 lysozyme proteins were inserted randomly into the box in non-overlapping configurations and solvated. Molecular dynamics simulations were performed using the Gromacs simulation package [v. 2016.5] [22] with the leap-frog algorithm and an integration step of 2 fs. For the RNA duplex, the ff99bsc0χOL3 [23] and the DESRES force field [24] were used. Lysozyme was described by the Amber99SB*-ILDN-Q protein force field. [25] Table S3 provides a detailed list of all simulation setups performed. After energy minimization up to local convergence, all simulations were equilibrated in an NVT ensemble with the velocity rescaling thermostat ( = 1 ps) [26] for 100 ps before equilibration in an NPT ensemble with the Parrinello-Rahman barostat ( = 5 ps) [27] for 5 ns. Production runs were performed with the same parameters as the NPT equilibration. Non-bonded electrostatic interactions were computed with the Particle Mesh Ewald method. [28] Lennard-Jones and real-space electrostatic interactions were calculated up to a cutoff of 1.0 nm. A shift function was used for the Lennard-Jones interactions. The lengths of atomic bonds to hydrogens were constrained with LINCS. [29] We tested that the different combinations of RNA, protein, and water force fields capture the dsRNA structure. As shown earlier, the combination of TIP4P-D for water and Amber99SB*-ILDN-Q for proteins results in an accurate description of crowded solutions of lysozyme and other proteins. [30] The description of an isolated RNA helix with OL3 and TIP4P-D is on par with well-established force fields combinations (OL3 RNA+TIP3P and DESRES+TIP4P-D), as judged by the root-mean-square distance (RMSD) of nucleobase heavy atoms to an idealized A-helix ( Figure S12). The dsRNA structure is stable over time, as demonstrated by the time traces in Figure S12 and the structural similarity to the mean structure of the RNA free in solution. The A-type structure of dsRNA is also stable in concentrated lysozyme solution ( Figure S13). We conclude that the different combinations of RNA, protein, and water force fields preserve the RNA A-helix structure. A systematic comparison between the RNA structures from the molecular dynamics simulations in dilute solution and in lysozyme is performed by means of the analysis of the rigid-body structural parameters [31,32] (Figure S14). The analysis confirms that the RNA helix structure is preserved in the presence of lysozyme. The MD simulations in lysozyme show diverse lysozyme-RNA binding modes, reflecting the nonspecific character of this interaction ( Figure S15). Differences between the runs (R1-R6) arise from limited sampling of the RNA-protein interactions in the simulation. Figure S15. Interactions of lysozyme with strands A and B in the six molecular dynamics simulations (R1-R6) of dsRNA with lysozyme. n̄contacts is defined as the mean number of contacts per simulation snapshot between RNA residues and lysozyme residues. Contacts are determined according to a distance cutoff <1 nm between the RNA residue center-of-mass and protein alpha-carbons. Also shown is the mean across the six runs.