Site-specific regulation of RNA editing with ribose-modified nucleoside analogs in ADAR guide strands (2024)

,

Article history

Abstract

Introduction

Adenosine (A) to inosine (I) is one of the most prevalent RNA modifications in mammals (1,2). This conversion is catalyzed by a group of enzymes called adenosine deaminases acting on RNA (ADARs) and takes place exclusively in duplex RNA. In humans, ADAR1 and ADAR2 are active adenosine deaminases targeting RNA. Importantly, the A to I reaction can produce codon changes in mRNA because I is read as G during translation (3,4). In addition, since ADARs require duplex structure for activity, their reaction can be directed to specific adenosines in different transcripts using complementary guide strands that form duplexes at the target sites (5). This approach is currently being pursued to develop a new class of therapeutic oligonucleotides that recruit ADARs to correct disease-causing mutations in RNA (6). Chemical modifications are required for these ADAR-guiding oligonucleotides to facilitate delivery, increase metabolic stability, and increase the efficiency and selectivity of the ADAR reaction (7–11). Identifying and characterizing ADAR guide strand modification patterns that lead to highly efficient and selective editing at different therapeutically relevant target sites is an ongoing effort by several laboratories in academics and industry (7–10). In earlier work, our group showed that placing a single locked nucleic acid (LNA) monomer at a position opposite an editing site's 5′ nearest neighbor nucleotide resulted in dramatically reduced editing at the targeted adenosine (10). Here, we probe the effect of LNA modification at other positions in ADAR guide strands. Furthermore, we describe efforts to define the mechanism by which ADARs are inhibited by LNA modification at specific positions in the guide strand. These efforts were facilitated by the synthesis of RNA oligonucleotides bearing 4′-C-methylated nucleoside analogs designed to probe the steric effect of a small 4′-alkyl substituent without the conformational restriction of the 4′-C, 2′-oxymethylene linkage found in the LNA monomers. Importantly, while LNA modification potently inhibited target site editing at two adjacent sites in the ADAR guide strand (-1 and -2), this effect is seen for the 4′-C Me modification only at the −2 position. The ability of a 4′-C-methylated nucleoside to retain editing at the -1 site but not at the −2 site provides a novel method to block bystander editing in 5′-AA target sequences. Finally, these results are rationalized considering the different effects of steric and conformational flexibility at different positions in ADAR guide strands.

Materials and methods

Synthesis of 4′-C-methyladenosine (4′-C-MeA) phosphoramidite

N⁶-Benzoyl-2′-O-acetyl-3′-O-benzyl-4′-C-methyl-5′-O-tert-butyldiphenylsilyladenosine, 2. To a suspension of N⁶-benzoyladenine (2.07 g, 8.66 mmol, 2 eq.) and an anomeric mixture of the protected and methylated sugar 1 (2.5 g, 4.33 mmol) in anhydrous CH₂Cl₂ (31 ml) at 0°C was added tin tetrachloride in one portion (0.5 ml, 4.33 mmol, 1 eq.) under an argon atmosphere. After 16 h of stirring at room temperature, the reaction was quenched with 2 ml of saturated NaHCO₃ dropwise at 0°C. The organic layer was diluted with 200 ml of CH₂Cl₂ and extracted twice with 100 ml of a saturated solution of NaHCO₃. The combined organics were dried over Na₂SO₄, filtered, and reduced under pressure to afford a white and oily precipitate. The crude mixture was chromatographed with silica gel with 0.3% MeOH in CHCl₃ (R_f=0.22, 3% MeOH in CH₂Cl₂) to afford 2 as a light colorless foam in 63% yield (2.05 g). ¹H NMR (400 MHz, CDCl₃) δ 9.11 (s, 1H), 8.70 (s, 1H), 8.20 (s, 1H), 8.04 (d, J=7.1 Hz, 2H), 7.67–7.31 (m, 20H), 6.28 (d, J=5.1 Hz, 1H), 5.98 (t, J=5.5 Hz, 1H), 4.69 (d, J=5.7 Hz, 1H), 4.62 (q, J=38.9 Hz, 2H), 3.88 (d, J=11.0 Hz, 1H), 3.59 (d, J=11.0 Hz, 1H), 2.09 (s, 3H), 1.38 (s, 3H), 1.08 (s, 9H). ¹³C NMR (101 MHz, CDCl₃) δ 206.95, 169.90, 152.72, 151.55, 149.56, 141.79, 137.55, 135.65, 135.53, 133.73, 132.77, 132.71, 132.60, 129.96, 129.92, 128.86, 128.48, 128.04, 127.92, 127.88, 127.84, 123.37, 87.58, 86.01, 75.01, 74.25, 68.06, 30.94, 26.95, 20.72, 19.24, 18.53. Orbitrap-HRMS calculated for C₄₃H₄₅N₅O₆Si [M+H]⁺ 756.31 m/z, found 756.3234m/z.

N⁶-Benzoyl-3′-O-benzyl-4′-C-methyl-5′-O-(tert-butyldiphenylsilyl)adenosine, 3. To a solution of 2 (0.214 g, 0.283 mmol) in 11 ml of MeOH was added K₂CO₃ (0.473 g, 3.42 mmol, 12 eq.) at 0°C and stirred at this temperature for 1.5 h. After completion of the reaction, the mixture was neutralized with 10 ml of aqueous 10% HCl (v/v). The aqueous phase was separated and extracted three times with 10 ml of CH₂Cl₂. The combined organics were dried over Na₂SO_4, filtered, and reduced under pressure to afford a colorless syrup. The crude mixture was purified by silica gel chromatography with a gradient of 1.75 to 2.25% MeOH in dichloromethane (R_f=0.43, 4% MeOH in CH₂Cl₂) to afford 3 as a colorless foam in 88% yield (178 mg). ¹H NMR (300 MHz, CDCl₃) δ 9.18 (s, 1H), 8.68 (s, 1H), 8.08 (s, 1H), 8.03 (d, J=6.9 Hz, 1H), 7.71–7.25 (m, 20H), 6.01 (d, J=5.8 Hz, 1H), 4.92 (t, J=5.8 Hz, 1H), 4.74 (q, 2H), 4.41 (d, J=5.8 Hz, 1H), 3.86 (d, J=10.8 Hz, 1H), 3.88 (s, 1H), 3.58 (d, J=10.8 Hz, 1H), 1.44 (s, 3H), 1.08 (s, 9H). ¹³C NMR (76 MHz, CDCl₃) δ 152.47, 151.36, 149.56, 141.86, 137.13, 135.63, 135.53, 133.70, 132.77, 132.71, 132.60, 130.03, 128.84, 128.73, 128.40, 128.15, 127.94, 127.90, 127.88, 123.27, 89.00, 87.52, 79.34, 77.51, 77.08, 76.66, 74.68, 74.51, 68.46, 26.95, 19.27, 19.01. Orbitrap-HRMS calculated for C₄₁H₄₃N₅O₅Si [M−H]⁻ 712.29 m/z, found 712.2974m/z.

N⁶-Benzoyl-3′-O-benzyl-4′-C-methyladenosine, 4. Compound 3 (0.178 g, 0.249 mmol) was co-evaporated twice with anhydrous THF. Subsequently, the silyl-protected nucleoside was dissolved in an anhydrous 1M TBAF solution in THF (0.5 ml) and stirred for 24 h at room temperature under an argon atmosphere. After complete conversion of the starting material, the mixture was reduced to an orange-colored oil and chromatographed with silica gel with a 2–3% MeOH in CH₂Cl₂ eluent to afford 4 as a white solid in 79% (93 mg)(R_f=0.35, 5% MeOH in CH₂Cl₂). ¹H NMR (300 MHz, CDCl₃) δ 9.43 (s, 1H), 8.67 (s, 1H), 7.96 (d, 2H), 7.75 (s, 1H), 7.58–7.25 (m, 8H), 6.14 (s, 1H), 5.73 (d, J=7.6 Hz, 1H), 5.13 (t, J=6.6 Hz, 1H), 4.75 (q, 2H), 4.30 (d, J=5.6 Hz, 2H), 3.89–3.69 (m, 2H), 3.58 (d, J=12.5 Hz, 1H), 1.86 (q, 1H), 1.36 (s, 3H). ¹³C NMR (76 MHz, CDCl₃) δ 164.59, 151.94, 150.12, 143.19, 137.27, 133.22, 132.84, 128.75, 128.70, 128.41, 128.25, 127.86, 123.92, 91.14, 89.11, 80.78, 75.18, 74.29, 69.02, 67.99, 25.62, 18.67. Orbitrap-HRMS calculated for C₂₅H₂₅N₅O₅ [M+H]⁺ 476.19 m/z, found 476.1935m/z.

N⁶-Benzoyl-4′-C-methyladenosine, 5. Debenzylation procedure was adapted from Zhou etal. (11) 4 (0.230 g, 0.484 mmol) was co-evaporated 3× with anhydrous dichloromethane and dried overnight under high vacuum with a stir bar. The 3′ benzylated nucleoside was then dissolved in 20 ml of anhydrous dichloromethane and stirred 5 min at −78°C under an argon atmosphere, followed by the dropwise addition of 1 M BCl₃ in CH₂Cl₂ (3.6 ml, 3.6 mmol, 7.5 eq.) over 10 min. After stirring for 4 h at the same temperature, the reaction was quenched with 16 ml of MeOH followed by 3.2 g of NaHCO₃. The mixture was left stirring for 4 h at room temperature. The insoluble solid was then filtered through a silica gel pad (2.8 g) and washed with 49 ml of 2:1 DCM:MeOH. The filtrate was evaporated to dryness and chromatographed with silica gel using a 3–8% gradient MeOH in CH₂Cl₂ to afford 5 as a white solid in 41% yield (75 mg) R_f = 0.15 (10% MeOH in CH₂Cl₂). ¹H NMR (400 MHz, DMSO) δ 11.10 (s, 1H), 8.72 (d, J = 14.0 Hz, 2H), 8.09–8.03 (m, 2H), 7.65 (t, J = 7.3 Hz, 1H), 7.55 (t, J = 7.6 Hz, 2H), 6.02 (d, J = 7.2 Hz, 1H), 5.46 (d, J = 6.3 Hz, 1H), 5.30 (t, J = 5.8 Hz, 1H), 5.21 (d, J = 4.7 Hz, 1H), 4.91 (d, J = 6.5 Hz, 1H), 4.11 (t, J = 4.6 Hz, 1H), 3.57 (dd, J = 11.6, 4.8 Hz, 1H), 3.44–3.35 (m, 3H), 1.20 (s, 3H). ¹³C NMR (101 MHz, DMSO) δ 166.39, 152.72, 152.00, 143.75, 134.17, 132.80, 128.98, 128.90, 126.40, 87.94, 87.89, 87.36, 74.44, 72.27, 67.39, 40.60, 40.39, 40.18, 39.97, 39.76, 39.55, 39.34, 19.30. Orbitrap-HRMS calculated for C₁₈H₁₉N₅O₅ [M + H]⁺ 386.14 m/z, found 386.1475 m/z.

5′-O-(4,4′-Dimethoxytrityl)-N⁶-benzoyl-4′-C-methyladenosine, 6.After co-evaporating 5 (0.118 g, 0.300 mmol) 3×with anhydrous pyridine (3×2 ml) and drying it overnight under high vacuum, the nucleoside was dissolved with distilled and anhydrous pyridine (2.0 ml). DMTr-Cl (0.124 g, 0.367 mmol, 1.2 eq.) was added in one portion to the solution, followed by the catalytic addition of 4-dimethylaminopyridine (DMAP) (1.7 mg, 0.014 mmol, 0.046 eq.). The solution was stirred for 24 h under argon and at room temperature and quenched with 1 ml of MeOH. All the volatiles were evaporated and the resulting crude was diluted with 30 ml of CH₂Cl₂, washed 1×with 5% aqueous NaHCO₃ (10 ml), 1×with brine (10 ml), dried over Na₂SO₄ and filtered. The crude was purified via silica gel chromatography (base washed with 1% TEA) with a gradient of 0–2.75% MeOH in CH₂Cl₂ to give 6 as a light yellow foam in 70% yield (144 mg). R_f=0.18 (4% MeOH in CH₂Cl₂). ¹H NMR (300 MHz, CDCl₃) δ 9.18 (s, 1H), 8.71 (s, 1H), 8.22 (s, 1H), 8.03 (d, J=7.5 Hz, 2H), 7.58 (dt, J=28.4, 7.4 Hz, 3H), 7.37–7.09 (m, 10H), 6.74 (d, J=8.3 Hz, 4H), 6.04 (d, J=6.2 Hz, 1H), 5.93 (s, 1H), 5.01 (t, J=5.8 Hz, 1H), 4.34 (d, J=5.2 Hz, 1H), 3.76 (s, 6H), 3.28 (d, J=10.0 Hz, 1H), 3.14 (d, J=10.0 Hz, 1H), 1.43 (s, 3H), 0.89 (d, J=9.2 Hz, 1H). ¹³C NMR (76 MHz, CDCl₃) δ 158.51, 152.10, 150.93, 149.61, 144.33, 141.55, 135.38, 135.28, 133.50, 132.96, 130.00, 129.16, 128.94, 127.97, 127.90, 127.84, 126.92, 123.11, 113.13, 90.46, 89.13, 86.71, 77.47, 77.05, 76.63, 76.18, 73.71, 68.47, 55.23, 29.73, 18.98. Orbitrap-HRMS calculated for C₃₉H₃₇N₅O₇ [M+H]⁺ 688.27 m/z, found 688.2744 m/z.

5′-O-(4,4′-Dimethoxytrityl)-2′-O-(tert-butyldimethylsilyl)-N⁶-benzoyl-4′-C-methyladenosine,7. Prior to silylation, 6 was co-evaporated 3x with anhydrous pyridine and dried under high vacuum overnight. To a solution of 6 (0.124 g, 0.180 mmol) in anhydrous pyridine and THF (2.25 ml, 1:1) was added silver nitrate (0.035 g, 0.200 mmol, 1.16 eq.) under an argon atmosphere at 0°C. After stirring for 5 min, TBS-Cl (0.030 g, 0.200 mmol, 1.16 eq.) was added in one portion at the same temperature. The reaction was allowed to stir overnight at room temperature. After the starting material was consumed, the reaction was diluted with 10 ml of ethyl acetate and filtered. The filtrate was collected and evaporated to a yellow oil, diluted with 10 ml of CH₂Cl_2, and washed once with 5% aqueous NaHCO₃. After drying and filtering the organic phase with Na₂SO_4, the crude was chromatographed with a base-washed (0.5% TEA) silica gel column with a gradient of 0–40% ethyl acetate in n-hexanes (R_f=0.41, 8:3 n-hexane:EtOAc). Purification afforded 7 as a colorless foam in 64% yield (92 mg). ¹H NMR (300 MHz, CD₂Cl₂) δ 9.16 (s, 1H), 8.68 (s, 1H), 8.18 (s, 1H), 8.03 (d, J=7.6 Hz, 2H), 7.73–7.15 (m, 12H), 6.98–6.79 (m, 4H), 6.09 (d, J=5.8 Hz, 1H), 5.19 (t, J=5.5 Hz, 1H), 4.42 (t, J=3.6 Hz, 1H), 3.82 (s, 6H), 3.45 (d, J=9.9 Hz, 1H), 3.25 (d, J=9.9 Hz, 1H), 2.87 (d, J=3.6 Hz, 1H), 2.04 (s, 1H), 1.44 (s, 3H), 0.88 (s, 9H), 0.01 (s, 3H), -0.17 (s, 3H). ²⁹Si NMR (60 MHz, CD₂Cl₂) δ 24.61. ¹³C NMR (76 MHz, CD₂Cl₂) δ 164.42, 158.70, 152.43, 151.61, 149.65, 144.82, 142.14, 135.65, 135.60, 134.06, 132.63, 130.14, 130.11, 128.82, 128.12, 127.90, 127.76, 126.89, 123.49, 113.15, 88.16, 86.83, 86.58, 75.80, 72.63, 68.01, 55.21, 29.69, 25.31, 18.86, 17.73, -5.33, -5.44. Orbitrap-HRMS calculated for C₄₅H₅₁N₅O₇Si [M+H]⁺ 802.36 m/z, found 802.3658 m/z.

5′-O-(4,4′-Dimethoxytrityl)-3′-O-[(2-cyanoethoxy)(N,N-diisopropylamino)phosphino]-2′-O-(tert-butyldimethylsilyl)-N⁶-benzoyl-4′-C-methyladenosine,8. Compound 7 was co-evaporated 3×with anhydrous CH₂Cl₂ (8 ml) and dried overnight under P₂O₅ and high vacuum. To a solution of 7 (0.092 g, 0.114 mmol), DMAP (2.7 mg, 0.0228 mmol, 0.2 eq.), distilled over CaH₂N,N-diisopropylethylamine (60 μl, 0.342 mmol, 3 eq.), and 0.45 ml of anhydrous CH₂Cl₂ was added 2-cyanoethyl N,N-diisopropylchlorophosphoramidite (50 μl, 0.230 mmol, 2 eq.) under an argon atmosphere. The reaction was allowed to stir overnight at room temperature. The reaction was quenched with 0.1 ml of MeOH and further diluted with dichloromethane. The organic phase was then washed with 5% aqueous NaHCO₃ once, dried over Na₂SO₄ and filtered to afford a yellow-orange oil. The residue was purified over a 0.5% TEA-washed silica gel column with a gradient of 30–45% ethyl acetate in n-hexanes to furnish the phosphoramidite 8 as a colorless foam in 77% yield (88 mg)(diastereomers of 8, R_f=0.64, 0.52 in 54% ethyl acetate in n-hexanes). ³¹P NMR (122 MHz, CD₂Cl₂) δ 150.19, 149.87. Orbitrap-HRMS calculated for C₅₄H₆₈N₇O₈PSi [M+H]⁺ 1002.47 m/z, found 1002.4713 m/z.

Solid-phase oligonucleotide synthesis and purification

All oligonucleotides were synthesized in 0.2 μmol scale based on the phosphoramidite chemistry and standard DNA/RNA reagents and cycles. DMTr-Off and 2′-O-TBDMSi protecting strategy was employed for all RNA oligonucleotide synthesis. The coupling time for all phosphoramidites was between 3 and10 min. For incorporation of the 4′-C-MeA phosphoramidite, an extended coupling time of 45 min was used and was dissolved with a 3:1 mixture of MeCN:DCM to 0.1 M(16). Before its incorporation, the purified phosphoramidite was co-evaporated 3x with anhydrous DCM, filtered using syringe filter and was dried under high vacuum overnight over P₂O₅. Coupling efficiency of the 4′-C-MeA was measured by means of the conductivity released by the trityl cation during detritylation of the growing strand with 3% trichloroacetic acid in CH₂Cl₂ and was comparable with commercially available phosphoramidites detritylations. Post synthesis, columns were dried for 2 h under high vacuum. Cleavage from the solid support and deprotection of the nucleobases were accomplished using a fresh aliquot of ammonium hydroxide in ethanol at 55°C for 12 h (1 ml of 28–30% NH₄OH/EtOH in 0.5 ml EtOH). After cleavage, the solid support was allowed to cool to room temperature and centrifuged for 5 min at 13000 ×g. The supernatants were collected and lyophilized. 2′-O-silyl ether deprotection was done using 250 μl of anhydrous 1M tetrabutylammonium fluoride (TBAF) in THF for 24 h at room temperature. Crude oligonucleotides were desalted by precipitation with the addition of 1 ml of 1-butanol and 50 μl of 3 M NaOAc and cooled for 2 h at −70°C. The solution was centrifuged at 13000×g for 30 min, supernatants were removed, and the pellets were washed twice with cold 95% ethanol.

The oligonucleotides were purified by loading a 1:1 mixture of 10–20 nmol of crude RNA to 80% formamide in 1×Tris-borate EDTA (TBE) and 1mM EDTA into a denaturing polyacrylamide gel. Depending on their size, oligonucleotides were loaded into different acrylamide matrices with different percentages and electrophoresed in 1×TBE. After visualization by UV shadowing on glass-backed silica gel plates, gel bands were excised, crushed, and soaked overnight at 4°C in 0.5 ml of 0.5 M NH₄OAc and 0.1 mM EDTA. Gel particles were removed from the extracted RNA by filtering the suspension for 5 min, 13000 ×g on 0.2 μm cellulose sterile membrane filter tubes (Corning). The oligonucleotides were precipitated from a solution of 75% EtOH at−70°C for 2 h. The oligonucleotides were then pelleted from the solution by centrifugation (13000×g for 30 min), with the supernatant being discarded. The pellets were lyophilized to dryness, resuspended in nuclease-free water, and quantified by absorbance at 260 nm. Oligonucleotide mass was confirmed by MALDI-TOF (Supplementary Information, Table S2).

Target RNA synthesis

The gene fragment of the corresponding human MECP2, mouse IDUA, human SRC, modified IDUA and MECP2, and human PPP2R5D were purchased from Integrated DNA Technologies. These were amplified with New England BioLabs (NEB) Q5 Hot Start high-fidelity DNA polymerase and purified by extraction from 1% agarose gels. Target RNA was synthesized with NEB’s HiScribe T7 RNA polymerase according to the protocol. The target RNAs were purified by 8% denaturing polyacrylamide gel electrophoresis.

Hybridization of guide and target RNA

The target and guide RNA were combined in a 1:10 ratio of target:guide to a final concentration of 180 nM of hybrids in 1×TE and 100 mM NaCl. This solution was heated to 95°C for 5 min then it was allowed to cool to room temperature for 2 h.

ADAR1 p110 overexpression and purification

Human ADAR1 p110 (UniProtKB P55265-5) with a C-terminal intein fusion protein with chitin binding domain (CBD) was overexpressed in S. cerevisiae BCY123 as previously described (12). Human ADAR1 p110 was purified by lysing cells in 20 mM Tris−HCl pH 8.0, 5% (v/v) glycerol, 750 mM NaCl, 50 mM imidazole, 1 mM tris(2-carboxyethyl)phosphine-HCl (TCEP-HCl), 50 μM ZnCl₂ and EDTA-free Protease Inhibitor using a microfluidizer. The lysate was centrifuged at 39 000 × g for 45 min at 4°C. The clarified lysate was passed over an equilibrated chitin binding column using gravity flow and was incubated for 2 h at 4°C for binding. The chitin column was then washed with 10 column volumes (CV) of lysis buffer and subsequently, the column was quickly washed with 5 bed volumes of the cleavage buffer (20 mM Tris−HCl pH 8.0, 5% (v/v) glycerol, 350 mM NaCl, 50 mM imidazole, 50 mM DTT and 50 μM ZnCl₂. To ensure cleavage of the CBD domain, the resin was incubated in 2 CVs of cleavage buffer overnight at room temperature for 18 h. The protein was eluted by washing columns with an additional 5 CVs of cleavage buffer. Finally, the eluted target protein was pooled, concentrated to 0.3 mg/ml, and then dialyzed against a storage buffer containing 20 mM Tris−HCl pH 8.0, 20% (v/v) glycerol, 350 mM NaCl, 50 mM imidazole, 1 mM TCEP-HCl. Protein concentration was determined by running the sample alongside bovine serum albumin (BSA) standards in a sodium dodecyl sulfate-PAGE (SDS-PAGE) gel followed by SYPRO Orange (Invitrogen) staining. The protein was stored at −80°C in aliquots for further use.

ADAR2 wild-type (WT) and ADAR2 E488Q overexpression and purification

ADAR2 wild-type and E488Q mutant overexpression and purification were performed as previously described (13)

In vitro deamination reactions and Sanger sequencing

For ADAR2 WT and ADAR2 E488Q deaminations, hybridized target and guide RNA were diluted to 5 nM in 1×ADAR2 reaction buffer (15 mM Tris–HCl pH 7.5, 3% glycerol, 60 mM KCl, 1.5 mM EDTA, 0.003% Nonidet P-40and 3 mM MgCl₂), 0.5 mM dithiothreitol (DTT), 160 units/ml RNAse inhibitor, 1 μg/ml yeast tRNAand 15 nM ADAR2 protein. The reactions were conducted at 30°C. For two time points reactions analysis, as well as for kinetic analysis (1, 3, 5, 10, 15, 30, 60 min), aliquots of 8 μl were quenched with 190 μl of 95°C heated nuclease-free water for 5 min. Single turnover conditions were achieved by reacting duplex RNA with excess enzyme (three-fold or greater excess, depending on ADAR isoform). For ADAR1 p110 deaminations, hybridized target and guide RNA were diluted to 15 nM in 1×ADAR1p110 reaction buffer (15 mM Tris–HCl pH 7.5, 3% glycerol, 26 mM KCl, 40 mM potassium glutamate, 1.5 mM EDTA, 0.003% Nonidet P-40and 4% glycerol), 0.5 mM DTT, 160 units/ml RNAse inhibitor, 1 μg/ml yeast tRNAand 150 nM ADAR1 p110. Due to the lower enzymatic activity of overexpressed and purified ADAR1 p110, a higher concentration of molar ratio of ADAR1 than ADAR2 was used to optimize deamination reactions. The reactions were conducted the same way as ADAR2. A 5 μl aliquot of each time point quench solution was reverse transcribed with Access Reverse Transcription (RT) Polymerase Chain Reaction (RT-PCR) system purchased from Promega. DNA was cleaned and concentrated using Zymo's purification kit and protocol. For Sanger sequencing with Azenta Life Sciences, purified DNA was diluted to 0.66 ng/μl in 1.6 μM of forward or reverse sequencing primer. Sanger sequencing traces were observed with SnapGene Viewer and statistical analysis along with nonlinear fits were conducted with Microsoft Excel and GraphPad Prism. The editing level for the corresponding zero time point was subtracted from each data point as background subtraction.

Oligonucleotide metabolic stability assay

A 60 μM stock solution of oligonucleotide was made in IDT Nuclease-Free Buffer (30 mM HEPES, pH 7.5; 100 mM potassium acetate). In a 50 μl reaction volume, the oligonucleotide sample was diluted to a final concentration of 8 μM with Dulbecco's Phosphate Buffer Saline (DPBS, Gibco 14190144) in 80% human serum (Millipore Sigma H4522). Each reaction was incubated at 37 °C and 5 uL aliquots were taken out at 0, 2, 4, 6, 8, 10, 12and 16 hand combined with 5 ul of 80% formamide in 1X TBE and 1 mM EDTA, vortexed and immediately stored at −80°C. Each aliquoted time point containing 40 pmol of oligonucleotide was loaded into a 15% denaturing Novex TBE Gels (Invitrogen, EC62752BOX or EC63152BOX) and electrophoresed at 200 V for 60 min in 1×TBE. The gel was then stained in SYBR Gold Nucleic Acid Gel Stain (Invitrogen S11494) for 30 min and imaged on the GelDoc Go Imaging System (Bio-Rad) under SYBR Gold settings on a UV/Stain-free tray. The densitometry of RNA bands was determined using Image Lab and plotted by a function of time using GraphPad Prism.

Oligonucleotide melting temperature assay

Melting temperature analysis of oligonucleotides was performed as previously described (14).

Directed editing on endogenous β-actin transcript in HEK293T cells

Procedure was performed as previously described (15).

Results

Inhibition of the ADAR reaction by LNA substitution is highly position-dependent

Previously, we showed that an LNA monomer located at the −1 position on an ADAR guide strand inhibited target editing and that strategic positioning of LNA could block unwanted bystander editing when the off-target was placed at least 6 nucleotides away from the target(14). Since LNA modifications in RNA therapeutics could be useful for enhancing target RNA binding affinity and resistance to nuclease degradation, we tested how ADAR responded to LNA modification at other positions in the guide RNA (16). We allowed ADAR2 to react with an RNA duplex bearing the sequence present in the c-Src mRNA. c-Src is a protein kinase frequently overexpressed in cancers (17). Specifically, we targeted for RNA editing the K295 codon (5′-AAA-3′). Editing of the central adenosine will convert this to an arginine codon resulting in a K295R mutation that renders the oncogenic protein kinase inactive (18). The sequence surrounding the target site is 5′-AUCAAAA-3′, where the target adenosine is directly flanked by two potential bystander editing sites (Figure 1A) with two additional bystander sites nearby. We first determined the percent editing by human ADAR2 at 60 min for the desired target and the four bystander sites employing an unmodified guide RNA as a control for comparison with the LNA-containing guide RNAs (Figure 1B). We then modified each sugar independently starting from the −4 position to the +3 position of the guide relative to the target adenosine and measured the percent editing for the target and bystander sites (Figure 1C-I). A schematic of the numbering of positions relative to the target A in the guide RNA is shown in Figure 1A, including the ‘orphan’ position, which is the base on the guide opposite to the target adenosine. Without LNA modifications, we observed significant editing of bystander adenosines four nucleotides and one nucleotide 5′ and one nucleotide 3′ of the target A. However, certain patterns of inhibition were observed depending on whether the LNA monomer was placed in the guide 5′ or 3′ of the orphan position. For instance, LNA placement 3′ of the orphan position at positions -4 and -3 in the guide RNA reduced by > 50% or blocked editing entirely at the bystander sites, while also reducing editing at the target site by 43% and 46%, respectively (Figure 1C and1F). Moreover, when the LNA monomers were placed closer to the target site in positions -2 and -1, editing at the target site and bystander sites decreased by > 95% for each site (Figure 1D-E). A very different editing profile was observed when LNA monomers were positioned 5′ of the orphan nucleotide. When LNA monomers were placed +1, +2 and + 3 in the guide RNA, either no inhibition (+2 and + 3) or less than two-fold reduction (+1) was observed. Yet, sites opposite the 3′ base flanking the LNA monomer or directly across it showed reduced or inhibited editing (Figure 1I).

High-resolution structures reveal possible origin of LNA inhibition of the ADAR reaction

Intrigued by the potent inhibitory effect of LNA monomers at the −1 and −2 positions of a guide RNA, we sought to understand the mechanistic origin of this phenomenon. The available high-resolution structures of ADAR2 bound to duplex RNA show that the ribose 4′ carbon (4′-C) at the −1 position is in proximity (4.4 Å) to the γ and δ carbons of isoleucine 491 (I491) (Figure 2) (19). This residue lies on a loop involved in stabilizing the flipped-out conformation required for adenosine deamination by ADARs (Figure 2C) (20,21). Furthermore, these structures also show this nucleotide adopts an unusual conformation with a DNA-like C2’-endo sugar pucker, a high anti glycosidic bond angle, and a disruption of base stacking (Figure 2E) (10). Likewise, these structures also reveal that the 4′-C of the −2 nucleotide in the guide RNA approaches even more closely (3.7 Å) the side chain of proline 492, a highly conserved amino acid in the ADAR family of enzymes while retaining an RNA-like C3’-endo sugar pucker (Figure 2D andF) (22).

These observations suggested that inhibition by the substitution of the −1 and −2 nucleotides in the guide RNA with an LNA monomer could result from steric or conformational effects (or both). In the case of each of the −1 and −2 LNA nucleotides, one can predict a steric clash between the side chains of I491 and P492 with the 4′-C-2′-oxymethylene linkage present in the minor groove at an LNA sugar. Considering that the −1 nucleotide undergoes an RNA-to-DNA sugar pucker conversion, inhibition when LNAs are placed at this position could arise from the restricted/locked conformation. To distinguish the possible origins of the inhibitory effect of LNA substitution in ADAR guide strands (i.e. restricted conformation or steric clash), we compared LNA substitution to two nucleoside analogs with different sugar methylation sites. We envisioned these analogs arising from cleaving either the bond between 2′-O and 6′-C to afford 4′-C-methyladenosine (4′-C-MeA) or cleavage of the 4′-C and 6′-C bond to furnish 2′-O-methyladenosine (2′-O-MeA) (Figure 3). Each of these analogs retain the atoms present in LNA A, but lack the conformational constraint imposed by the 4′-C to 2′-O bridge. Thus, if the steric demand of the additional methylene present in LNA A (compared to rA) was responsible for the inhibitory effect, at least one of these two analogs would be expected to inhibit the ADAR reaction when placed at the corresponding position in the guide strand. However, if each were well tolerated by ADAR, the LNA effect on conformational flexibility would be the more likely origin of inhibition. Since ADAR’s preferred 5′ nearest neighbor nucleotide is U or A (i.e. 5′-UA or 5′-AA), the −1 nucleotide in a guide RNA targeting these sequences is uridine or adenosine. Yet, in this project, we tested sugar analogs of adenosine for studies of mostly 5′-UA targets. In addition to the 2′-O-MeA, LNA A and 4′-C-MeA, we also tested the unlocked nucleic acid version of adenosine (UNA A). Since UNA A is highly flexible, results with this analog could determine to what extent flexibility is beneficial at the -1 position (23).

4′-C-Methyladenosine phosphoramidite and RNA synthesis

To incorporate each of these modifications into RNA, their corresponding phosphoramidites were needed. However, the 4′-C-MeA phosphoramidite had not been previously reported. Waga etal. developed a synthetic strategy to incorporate 4′-C-alkyl substituents into free ribonucleosides to test for effects on HIV replication (24,25). Furthermore, Koizumi and Kano synthesized 4′-C-aminoalkyl uridine modified RNA and tested for effects on siRNA delivery in RNA interference (26,27). Lastly, Guo et al. synthesized a 4′-C-methyluridine phosphoramidite and modified RNA to understand the effects of 4′-C-alkylation on ribose conformation during base-catalyzed RNA cleavage via internal transesterification (28). Here, we used a seven-step synthesis to furnish the 4′-C-MeA phosphoramidite starting from the previously synthesized 4′-α-C-methylribose 1 in reasonable yields (Scheme 1) (29,30). Initially, regioselective N⁹ ribosylation with N⁶-benzoyladenine and tin tetrachloride (SnCl₄) afforded the protected adenosine analog 2. The desired regioisomer was identified by heteronuclear multiple bond correlation (HMBC), heteronuclear single bond correlation (HSQC), and hom*onuclear correlation (COSY) NMR experiments (Supplementary Information, Figure S1A). In these experiments, the N⁷ analog displayed an interaction between the glycosyl anomeric hydrogen and C⁵ and C⁸ of the purine, while for the N⁹ analog, the anomeric hydrogen interacts with carbons C⁴ and C⁸ (31). In addition, the selective formation of the β-ribonucleoside is supported by Höfler and Vorbrüggen's mechanistic explanation of glycosylations employing Friedel-Craft catalysts, such as SnCl₄ (32). Subsequent deacylation of the 2′-OH without the removal of the N⁶-benzoyl group was accomplished with potassium carbonate (K₂CO₃) and methanol in high yields, followed by the desilylation of the 5′-OH with 1M tetrabutylammonium fluoride (TBAF). Debenzylation of the 3′-OH was achieved by employing 1M boron trichloride (BCl₃) in CH₂Cl₂ to afford 5 in a form ready for advancement to the phosphoramidite. Hence, tritylation of the 5′-OH with 4′4’-dimethoxytrityl chloride (DMTr-Cl) and pyridine, followed by the regioselective silylation of the 2′-OH in the presence of silver nitrate (AgNO₃), pyridine, and tert-butyldimethylsilyl chloride (TBS-Cl) provided compound 7 in good yields. The selective silylation of the 2′-OH in ribonucleoside phosphoramidites often produces low yields due to the poor selectivity between the 2′ and 3′ hydroxyl groups (33) However, here we propose that the 2′-silylation reaction furnishes higher than usual yields and selectivity due to the inhibitory/steric effect of the 4′-C methyl group for the 3′-O-silylation. The desired isomer was identified by NMR hom*onuclear correlation of the 1′-2′ protons and one distinct signal from the ²⁹Si HMBC correlation of the 2′-OSiTBDM–2′-H (Supplementary Information, Figure S1B). Lastly, phosphitylation of the 3′-hydroxyl provided phosphoramidite 8, suitable for automated RNA synthesis.

Guide RNAs containing 4′-C-MeA at the −1 position do not inhibit ADAR

To test the effect of the sugar modifications at the −1 position, we hybridized chemically modified guide strands to an RNA transcript bearing the sequence from the human methyl CpG binding protein 2 (MECP2) containing the W104X site caused by a G-to-A mutation that causes Rett syndrome (34). The resulting duplexes were then allowed to react with purified human ADAR2. The effect of each analog was measured by determining the percent editing at the target site at 60 min (Figure 4A). In addition, the effect of 4′-C-MeA was evaluated by measuring deamination rates under single turnover conditions and compared to duplexes with canonical adenosine (rA) (Figure 4A). We also determined the effect of these analogs in an ADAR1 reaction using a target RNA bearing the mouse IDUA W392X nonsense mutation and purified human ADAR1 p110 (Figure 4B). Mutations within the human IDUA gene, which encodes α-l-iduronidase, can cause Hurler syndrome (35). The mouse IDUA W392X mutation mimics the Hurler-causing human W402X mutation that can be reversed by ADAR-mediated editing and is known to be a good substrate for ADAR1 p110 (20). The effects of having 2′-O-MeA, 4′-C-MeA, LNA A, and UNA A at the -1 position for ADAR2 and ADAR1 p110 editing are shown in Figure 4. Interestingly, 2′-O-MeA and 4′-C-MeA behaved similarly to rA and do not show any significant difference in the level of ADAR2 editing at 60 min. Yet, as previously shown by Brinkman et. al, having LNA A at the -1 position inhibited the ADAR2 at the MECP2 W104X site (10). However, UNA A does show a significant difference in the level of editing compared to rA. Importantly, we determined the ADAR2 deamination rate constant for the 4′-C-MeA-containing guide and found it to be indistinguishable from the adenosine-containing guide (Figure 4A).

Guide RNAs containing 4′-C-MeA at the −2 position inhibit the ADAR reaction

RNA editing experiments analogous to those described above were carried out to examine the impact of 4′-C-MeA on ADAR reactivity when it is placed at the −2 position of the gRNA. First, a modified MECP2 W104X* transcript was designed and generated where the −2 base 5′ of the target was changed from a cytidine to uracil to base pair with 4′-C-MeA. Then, guide RNAs containing rA, LNA A, 2′-O-MeA and 4′-C-MeA at the −2 position were synthesized. In vitro deaminations with purified human ADAR2 were carried out and the effect of each analog was measured by determining the percent editing of the target site at the reaction endpoint at 60 min (Figure 5A). Notably, no editing was detected when 4′-C-MeA was positioned −2 relative to the target, suggesting that the methyl group of the 4′-C-Me in the −2 sugar does indeed clash with P492 (Figure 5C). Likewise, when LNA A was placed at the −2 position, editing was reduced by 86% when compared to the control (-2 rA), displaying similar editing results to those seen with the c-Src target (Figure 1D). No significant difference was observed when 2′-O-MeA was positioned −2 relative to the target A, suggesting that this sugar modification is well tolerated at this position for ADAR editing.

Effect on ADAR editing of 4′-C-MeA at various positions of the gRNA

Our previous experiments focused on highlighting the effect of 4′-C-MeA when it is placed at the −1 and −2 positions of the gRNA for the MeCP2 and IDUA targets. However, we sought to define the effect of this analog in another sequence context while also evaluating the effect of modification at other positions relative to an editing site. The A-rich c-Src provides a useful framework to achieve this. For this, we placed 4′-C-MeA in a gRNA at a position that would be −2 relative to a 5′ adjacent off-target site (blue), −3 relative to the target (red) and −4 and + 1 relative to other bystander sites (grey) (Figure 6A). In vitro deaminations show that 4′-C-MeA is still highly inhibitory when placed −2 to the 5′ bystander site, to the point of no detectable editing when compared to a non-modified gRNA (Figure 6A). Interestingly, 4′-C-MeA is significantly inhibitory (3-fold reduction in editing level) when placed −3 to the target site. This phenomenon can be attributed to a potential steric clash with ADAR’s flipping loop amino acid cluster R481, I456 and F457 and the proximal 4′-C substituent at the −3 nucleotide in the gRNA (Figure 6B). Yet, no significant difference in editing was observed when this analog was placed +1 and −4 relative to other bystander sites.

4′-C-Me-modified guide RNAs can selectively block adjacent bystander sites while retaining on-target editing efficiency

Since it appeared that 4′-C-methyl modification was tolerated at the −1 position but not at the −2 position of an ADAR guide strand, we decided to test the impact of the modification on a substrate with two adjacent adenosines. Directing ADARs to edit one of two adjacent adenosines can be challenging, particularly if the desired edit is at an adenosine with a non-optimal nearest neighbor nucleotide (36,37). Therefore, we generated a substrate with the target sequence 5′-UAAG-3′, with the desired edit to take place at the first of the two adjacent adenosines and in a non-optimal sequence 5′-UAA-3′ with the adjacent bystander site in the near-optimal 5′-AAG-3′ sequence context. ADAR guide strands were synthesized that vary the nucleotide at the -1 position and orphan position relative to the target adenosine. The duplexes formed with these guide strands were then incubated with human ADAR2 E488Q, a hyperactive mutant of the enzyme or wild-type human ADAR1 p110 (38). Interestingly, with a guide strand of unmodified RNA and uridine at the orphan position (Figure 6AB, X = rA and Y = rU) we find nearly equal editing levels at the two adjacent adenosines with both proteins. With unmodified RNA and cytidine at the orphan position (Figure 6AB, X = rA, Y = rC), the reaction becomes selective for the target A but with significant editing for both ADAR2 E488Q (24 ± 1%) and ADAR1 p110 (21 ± 3%) at the bystander A at the 120 min time point, respectively. However, with both proteins, the guide strand with 4′-C-methyladenosine at the −1 position relative to the target site (i.e. −2 position relative to the bystander site) reduces bystander editing to an undetectable level while maintaining efficient editing at the target adenosine (Figure 6AB, X = 4′-C-MeA, Y = rC).

To further evaluate the site-specific inhibitory effect of 4′-modified sugars, we targeted a transcript with a potential therapeutic editing site adjacent to a bystander adenosine (5′-UAAG). Point mutations in the human protein phosphatase 2 regulatory subunit B’delta (PPP2R5D) gene can lead to Jordan syndrome, a neurodevelopmental disorder (39). Mutation of the E200 codon of PPP2R5D to a codon for lysine (E200K) is one of the mutations responsible for pathogenic phenotypes in Jordan syndrome patients (40). Reversion of this lysine codon (5′-UAA) to glutamic acid (5′-UGA) is expected to restore protein function. Because this site is a 5′-UAAG site, where the target adenosine has a 5′-U neighbor and a 3′-adjacent potential bystander, we generated guide RNAs to stimulate efficient editing at the target site (underlined A), while also blocking undesired editing at the proximal bystander site. To achieve this, we synthesized guide oligonucleotides with A-C mismatches at the target site while placing 4′-C-methyladenosine or LNA A at the −2 nucleotide position relative to the bystander site and performed in vitro deaminations with three ADAR isoforms (ADAR1p110, ADAR2 and ADAR2 E488Q) (Figure 8). Similar to previous results, we observed that placement of −1 4′-C-MeA strongly inhibits the 3′-adjacent bystander site, while retaining substantial on-target editing with all ADAR proteins. Yet, as predicted, LNA A placement −1 to the target inhibits both sites for all ADAR isoforms.

4′-C-MeA-modified oligonucleotides selectively direct editing in human cells with ADAR2

So far, we have described how selective positioning of 4′-C-MeA and LNA A monomers in guide RNAs can block off-targets in biochemical assays with purified ADAR proteins and target RNAs. To determine if the inhibitory effects of these sugars are observed in directed editing experiments in mammalian cells, we synthesized metabolically stable guide oligonucleotides to direct editing by full-length ADAR2 to an endogenous target found in the 3′-UTR of β-actin mRNA (Table S1) (15,41). The guide oligonucleotide was designed to target the 5′-UUAG-3′ sequence where the -1 (X) and -2 (Y) nucleotides in the guide are adenosine analogs. Hence, we tested the effect of having either X or Y = 4′-C-MeA or LNA A, in addition to a control guide with unmodified adenosine as these positions (X, Y = rA) (Figure 9A). Additionally, we stabilized the guide oligonucleotides by replacing most riboses with 2′-O-methylriboses and positioning phosphorothioate linkages near the 3′ and 5′ ends, as described previously (Figure 9B) (20,42). Expression of wild-type human ADAR2 in HEK293T cells transfected with the positive control guide (X, Y = rA) led to 30 ± 4% editing at the target site, while the negative control sample (no guide RNA transfection) showed no editing at this site (Figure S4). In the presence of a −1 4′-C-MeA guide, a very similar level of 26 ± 1% editing was observed. Yet, when cells were transfected with the −1 LNA A guide, no editing was detected at the target site, indicating complete inhibition by the LNA modification at -1. Inhibition was also observed, albeit to a lesser extent, when the −2 position of the guide was LNA A (3-fold reduction in editing level). Importantly, the guide RNA bearing the 4′-C-MeA monomer at the −2 position led to a 4.4-fold reduction in editing level in these experiments, consistent with the inhibition patterns observed in the biochemical assays described above.

The effects of 4′-C-Me A on duplex stability and nuclease sensitivity are similar to adenosine

Lastly, we determined the melting temperature (T_M) and ribonuclease sensitivity of RNAs bearing 4′-C-MeA. We conducted thermal denaturation assays to elucidate the impact of 4′-C-MeA on duplex stability and base pairing selectivity. For this purpose, we synthesized 12 nt RNAs containing 4′-C-MeA or adenosine and hybridized them with complementary 12 nt RNAs for T_M measurements (Supplementary Information, Figure S2). The nuclease sensitivity of 4′-C-MeA-containing RNA was assessed using a 15 nt long oligonucleotide with this analog at the 7th position flanked by all 2′-O-methyl nucleotides and subjecting it to 80% human serum over a 24 h time course. For comparison, a control oligonucleotide was designed the same, but with rA at the 7^th position (Supplementary Information, Figure S3A, B). The T_M analysis showed that 4′-C-MeA has an almost identical profile to that of adenosine, suggesting that 4′-methylation of the sugar does not influence the thermal stability of an RNA duplex or pairing selectivity of the nucleobase. However, compared to rA, a slight decrease in serum stability was observed for the 4′-C-MeA-containing oligonucleotide.

Discussion

Oligonucleotide therapeutics is a maturing field with over 15 FDA-approved drugs at the time of this writing (43,44). ADAR-catalyzed therapeutic RNA editing, guided to specific locations in the transcriptome by oligonucleotides, is a novel mechanism of action for oligonucleotide therapeutics (6). Safe and effective drugs that work by this mechanism are currently under development with clinical trials expected to start in 2024 (45,46). ADAR-guiding oligonucleotides for therapeutic RNA editing require extensive modification of the parent RNA to facilitate metabolic stability, cell uptake, and deamination efficiency and selectivity. Several groups have reported efforts in tailoring guide RNAs with chemical modifications (2′-O-methyl, 2′-fluoro, phosphorothioate with stereopure backbones, LNA, etc.) for this purpose (7,9,47,48). In this study, we demonstrate the effect of specific sugar-modified nucleotides at different positions in ADAR guide strands varying the proximity to the editing site. Importantly, the effects of different sugar modifications were found to be highly dependent on the position modified and rationalized by analysis of existing ADAR2-RNA crystal structures.

Generally, LNA placement in the guide strand 3′ of the orphan position inhibits editing at the target A. The degree of ADAR inhibition is directly proportional to the proximity of LNA to the editing site. As shown in Figure 1C-E, when LNA is placed between the -4 and -1 positions, the inhibition significantly increases as it approaches the −1 nucleotide. Furthermore, editing of specific adenosines can be entirely inhibited when LNA is placed at either the −1 or −2 positions of the gRNA. In contrast, placing LNA monomers 5′ of the orphan position in the gRNA has substantially less impact on the ADAR reaction. Our previous structural studies showed that the RNA conformation deviates significantly from A-form at the editing site and the adjacent nucleotide on the 5′ side (3′ direction on the guide strand) whereas the RNA conformation 3′ to the editing site is A-form (5′ direction on the guide strand) (21). Thus, the greater severity of ADAR inhibition by LNA substitution on the 3′ side of the guide strand can be explained by the asymmetry of the RNA conformational changes that take place during the ADAR reaction, in addition to the innate rigidity of LNA monomers and their influence on neighboring nucleotides (23,49,50). Lastly, it was observed that LNAs placed at the orphan position render complete inhibition (Figure 1F). This observation is also explained by the inspection of ADAR2-RNA crystal structures. The orphan nucleotide hydrogen bonds with glutamic acid 488 in ADAR2’s flipping loop to stabilize the base-flipped conformation while also adopting an unusual sugar pucker that falls in between C2’-endo and C3’-endo conformations (20). An LNA monomer is likely unable to adopt this conformation and inhibits the reaction at this position. Also, if an LNA monomer is located at the position immediately on the 5′ side of the orphan (Figure 1I, +1 LNA), it is likely that the orphan nucleotide will resist the formation of alternate sugar puckers due to the previously reported influence of the vicinal LNA effect (51).

While LNA substitution potently inhibited the ADAR reaction at the −1 and −2 positions of the guide, we found that 4′-C-MeA inhibited the ADAR reaction only when placed at position -2. One possible explanation for the lack of inhibition with 4′-C-MeA at −1 is that ADAR’s I491 side chain may adopt a different conformation to accommodate the 4′-C-methyl group emanating from the RNA’s minor groove. Interestingly too, based on previous work on the characterization of 4′-C-methyluridine monomers in RNA, this type of aliphatic modification promotes a DNA-like C2’-endo sugar pucker (28). This structural attribute of the sugar in the 4′-C-MeA monomer can be beneficial for ADAR editing when placed at the -1 position based on previous literature that reveals that analogs that favor C2’-endo usually promote more efficient deamination than C3’-endo nucleotides at this position (10). Based on the biochemical and structural data presented herein, we conclude that LNA monomers positioned at -1 on a gRNA inhibit ADAR by hindering the sugar re-puckering required during the ADAR reaction. Finally, it can be inferred that the presence of 4′-C-MeA or LNA A at the -2 position leads to inhibition of the ADAR enzyme due to steric clashes, considering that the sugar pucker of the -2 nucleotide does not change upon deamination (Figure 5C). Unlike isoleucine's flexible side chain, proline lacks the conformational flexibility to accommodate the 4′-C-Me or methylene bridge emanating from the −2 nucleotide's minor groove of the 4′-C-MeA or LNA, respectively. When 4′-C-MeA is placed -3 relative to an editing site in the gRNA inhibition is also observed, albeit to a lesser extent (Figure 6A). This can be rationalized by the crystal structures where the −3 nucleotide 4′-carbon is close to a cluster of amino acid side chains (4.6Å–Arg481, 3.7Å–Ile456 and 3.8Å–Phe457).

Considering that the field of therapeutic RNA editing is emergent, there is still a need to regulate bystander editing. The strategic positioning of 4′-C-Me monomers in ADAR-guiding oligonucleotides, as described here, is a novel way to achieve this goal. While LNA monomers can block off-target editing at sites that are distal to the target site (i.e. approximately six nucleotides away surrounding the target), using LNA modification to block adjacent bystander editing in a 5′-AA target sequence will likely cause a substantial loss of efficiency at the target site. Nevertheless, as we have shown here for two different RNA targets, placement of a 4′-C-Me monomer at the -1 position relative to the target adenosine in a 5′-AA sequence can eliminate editing at the adjacent bystander adenosine while still maintaining high editing efficiency at the target site (Figures 7–8). Finally, we show that the same patterns of inhibition observed in our biochemical assays are seen in a directed editing experiment using metabolically stabilized guide oligonucleotides and cultured human cells.

These structural observations along with biochemical results hint at further chemical manipulations that can be furnished at the minor groove of −1 and −2 nucleotides in the gRNA for the optimization of more selective site-directed editing systems with ADARs. For instance, bulkier 4′-C aliphatic nucleotides can be synthesized, as well as engineered ADARs with I491X and P492X mutants to generate new combinations of shape complementarity (i.e. bump-hole method) between the interface of ADAR’s flipping loop and the RNA’s minor groove. Such combinations could add to the growing list of directed editing systems capable of highly selective corrections of disease-causing mutations at the RNA level (3,20,34,47,52–54).

In summary, we describe a synthesis of 4′-C-MeA phosphoramidite and its use to generate modified ADAR guide strands with novel properties. We found that 4′-C-MeA behaves similarly to adenosine with respect to its thermal stability in oligoribonucleotides and is slightly more labile to nucleases. 4′-C-MeA-modified guides facilitated identification of the mechanistic origin of inhibition of ADAR using site-specific LNA modification, while also shedding additional light on ADAR’s base-flipping mechanism (i.e. illustrating the importance of conformational changes happening at the −1, −2and orphan nucleotides). ADAR inhibition arises for different reasons depending on the placement of the LNA in the guide oligonucleotides relative to the target. An LNA monomer at the −1 position relative to the target renders no editing due to the inhibition of the re-puckering that the -1 sugar must undertake to stabilize the base-flipped complex. Whereas LNA and 4′-C-methyl modification at −2 inhibits ADAR due to steric clash with P492. In conclusion, we have provided a framework that describes guide oligonucleotides with different patterns of chemical modifications containing LNAs and 4′-C-Me monomers that can be used to achieve efficient ADAR editing while also blocking editing at bystander sites, including within challenging 5′-AA sequences.

Data availability

The data underlying thisarticle are available in the article and in its online supplementary material.

Supplementary data

Supplementary Data are available at NAR Online.

Acknowledgements

We thank William T. Jewell from the Campus Mass Spectrometer Facilities for troubleshooting and aiding in acquiring MALDI-TOF and HF Orbitrap ESI mass spectra. We thank Derrick C. Kaseman for aiding in developing 2D silicon 29 and phosphorus 31 NMR experiments.

Funding

National Science Foundation Graduate Research Fellowship Program [165004]; Rett Syndrome Research Trust; National Institute of Health Sciences [R35GM141907]. Funding for open access charge: P.A.B. acknowledges financial support from the National Institutes of Health [R35GM141907]; Rett Syndrome Research Trust.

Conflict of interest statement. P.A.B. is a consultant, S.A.B. member and holds equity in ProQR Therapeutics and holds equity in Beam Therapeutics, companies developing therapeutic editing technologies.

References

© The Author(s) 2024. Published by Oxford University Press on behalf of Nucleic Acids Research.

Issue Section:

Chemical Biology and Nucleic Acid Chemistry

Comments

0 Comments

I agree to the terms and conditions. You must accept the terms and conditions.

Email alerts

Citing articles via

More from Oxford Academic