1-Deazaguanosine-Modified RNA: The Missing Piece for Functional RNA Atomic Mutagenesis

Atomic mutagenesis is the key to advance our understanding of RNA recognition and RNA catalysis. To this end, deazanucleosides are utilized to evaluate the participation of specific atoms in these processes. One of the remaining challenges is access to RNA-containing 1-deazaguanosine (c1G). Here, we present the synthesis of this nucleoside and its phosphoramidite, allowing first time access to c1G-modified RNA. Thermodynamic analyses revealed the base pairing parameters for c1G-modified RNA. Furthermore, by NMR spectroscopy, a c1G-triggered switch of Watson-Crick into Hoogsteen pairing in HIV-2 TAR RNA was identified. Additionally, using X-ray structure analysis, a guanine–phosphate backbone interaction affecting RNA fold stability was characterized, and finally, the critical impact of an active-site guanine in twister ribozyme on the phosphodiester cleavage was revealed. Taken together, our study lays the synthetic basis for c1G-modified RNA and demonstrates the power of the completed deazanucleoside toolbox for RNA atomic mutagenesis needed to achieve in-depth understanding of RNA recognition and catalysis.


■ INTRODUCTION
Deazanucleosides are needed for atomic mutagenesis studies to explore RNA structure, function, and catalysis. 1−5 The exchange of a specific nitrogen atom in a nucleobase by carbon can critically affect RNA properties because the hydrogen acceptor (imino, N−) or hydrogen donor (amido or amino, −NH−) capabilities are impaired at the specific position. 6−8 This is crucial for base pairing, 8,9 RNA− protein recognition, 3,6,8,9 RNA−small molecule recognition, 10 and RNA-catalyzed chemical reactions. 4,11 −14 In particular, atomic mutagenesis lead to our current mechanistic understanding of ribozymes, 15−18 including the ribosome. 19−22 Thus far, diverse deazanucleosides have been utilized for RNA atomic mutagenesis experiments; these are 3-deazacytidine (c 3 C), 14,23,24 7-deazaadenosine (c 7 A), 4,12−14,16,17 3deazaadenosine (c 3 A), 15,23 1-deazaadenosine (c 1 A), 12,13,15,23 and 7-deazaguanosine (c 7 G). 14,16 Furthermore, an efficient synthesis of 3-deazaguanosine (c 3 G) and the corresponding phosphoramidite has been reported recently and adds to the deazanucleoside tool box. 25,26 The missing piece, however, is 1-deazaguanosine (c 1 G), which is urgently needed for RNA atomic mutagenesis studies to probe the role of active site guanosines in catalysis of diverse ribozymes and for ligand recognition in the binding pockets of many riboswitches. In this work, we present a novel synthetic route toward c 1 G, the corresponding phosphoramidite and its incorporation into oligoribonucleotides by RNA solid-phase synthesis. Furthermore, we describe the impact of c 1 G on the thermodynamic stability of RNA double helices. Moreover, we found evidence for Hoogsteen base pair formation of c 1 G with protonated cytosine in HIV-2 TAR RNA by nuclear magnetic resonance (NMR) spectroscopy. The study is complemented by the crystal structure of a c 1 G-containing RNA hairpin to shed light on a specific guanine N1−phosphate backbone interaction observed in the wild-type RNA, and finally, we evaluate the crucial role of the guanosine N1 atom in catalysis of phosphodiester cleavage by the twister ribozyme.

■ RESULTS AND DISCUSSION
To date, synthetic routes to 1-deazaguanine nucleoside building blocks for oligonucleotide synthesis have been described for DNA only. 27 DNA containing 1-deaza-2′deoxyguanosine (c 1 dG) is unstable toward acids, and this feature has been utilized to generate abasic sites. 28 Access to the naked ribonucleoside 1-deazaguanosine was first reported in the nineteen eighties, 29,30 employing rather harsh nucleosidation reactions involving mercury cyanide and based on O 6 -benzylated 1-deazaguanine, which itself requires laborious multistep synthesis. 31 Later, access to 1-deazaguanosine was demonstrated via 5-amino-1-β-D-ribofuranosylimidazole-4-carboxamide (AICA-riboside) as the key intermediate. 32 We, however, decided to put efforts into a direct, more efficient, and unprecedented route from readily available starting materials.
Synthesis of c 1 G Nucleoside. For c 1 G nucleoside 6 and the phosphoramidite precursor 5 (Scheme 1), we started the synthesis from commercially available 6-chloro-1-deazapurine, which was quantitatively transformed to the corresponding 6iodo derivative 1 by treatment with hydroiodic and phosphorous acids. Then, silyl-Hilbert−Johnson nucleosidation of 6-iodo-1-deazapurine 1 and 1,2,3,5-tetra-O-acetyl-β-Dribofuranose in the presence of N,O-bis(trimethylsilyl)acetamide (BSA) and trimethylsilyl trifluoromethanesulfonate (TMSOTf) provided nucleoside 2 in good yields. The exchange of acetyl to tert-butyldimethylsilyl (TBS) protection of the ribose hydroxyl groups (compound 3) was required to enable efficient copper-catalyzed coupling of the aryl iodide with benzyl alcohol to furnish nucleoside 4, inspired by the work of the Buchwald laboratory. 33 Then, site-specific nitration of the O 6 -benzyloxy-1-deazapurine moiety was accomplished using tetrabutylammonium nitrate (TBAN) and trifluoroacetic anhydride (TFAA) to obtain nucleoside 5, in analogy to work by Koomen. 34−36 Finally, selective reduction of the nitro group was conducted by HSiCl 3 referring to Benaglia and coworkers, 37 followed by the cleavage of the benzyl group by hydrogenation, and cleavage of the silyl ethers to give 1deazaguanosine 6 in 22% overall yield in six steps with five chromatographic purifications; in total, 0.4 g of 6 was obtained in the course of this study.
Synthesis of c 1 G Phosphoramidite. We started the synthesis toward c 1 G building block 12 from precursor 5 with the reduction of the nitro group using HSiCl 3 , followed by the cleavage of the benzyl group via hydrogenation providing nucleoside 7 (Scheme 2). Then, the O 6 functionality was protected with a (2-nitrophenyl)ethyl (NPE) moiety applying Mitsunobu reaction conditions, followed by protection of the exocyclic NH 2 group using trifluoroacetic anhydride (TFAA), resulting in derivative 8. By the cleavage of the silyl ethers with tetrabutylammonium fluoride (TBAF), triol 9 was quantitatively obtained. Next, the 5′ and 3′ OH groups were simultaneously protected using di-tert-butylsilyl bis-(trifluoromethanesulfonate) (tBu 2 Si(OTf) 2 ), 38,39 followed by silylation of the 2′-OH group with tert-butyldimethylsilyl chloride (TBS-Cl) and subsequent removal of the 5′-O and 3′-O protection clamp with a solution of HF in pyridine to give compound 10. The functionalization of the 5′-OH group with 4,4′-dimethoxytrityl chloride was conducted under standard conditions and yielded compound 11. Finally, the phosphoramidite 12 was generated by treatment with 2-cyanoethyl-N,N,N′,N′-tetraisopropylphosphorodiamidite (CEP(NiPr 2 ) 2 ) in the presence of 5-benzylthio-1H-tetrazol (BTT). Starting from precursor 5, the target compound 12 was synthesized in six steps, with six chromatographic purifications and an overall yield of 33%; in total, 1.1 g of 12 was obtained in the course of this study.
Synthesis of c 1 G-Modified RNA. The solid-phase synthesis of RNA with site-specific c 1 G modifications was performed using the new building block 12 together with 2′-O-TBS protected A, C, G U phosphoramidites, or alternatively, with 2′-O-[(triisopropylsilyl)oxy]methyl protected (TOM) amidites. 40,41 The novel building blocks were coupled with yields higher than 98% according to the trityl assay. The cleavage of the oligonucleotides from the solid support and deprotection were conducted using methylamine/ammonia in water (AMA), followed by treatment with tetra-n-butylammonium fluoride (TBAF) in tetrahydrofuran. Salts were removed by size-exclusion chromatography, and RNAs were purified by anion-exchange chromatography under denaturating conditions (60 to 80°C column temperature; Figure 1 and Supporting Table S1). The molecular weights of the purified RNAs were confirmed by liquid chromatography (LC) electrospray-ionization (ESI) mass spectrometry (MS). The sequences of c 1 G containing RNAs synthesized in the course of this study are listed in Supporting Table S1. Notably, HPLC analysis of the crude deprotected c 1 G containing RNAs displayed a second product that was migrating slower, in particular, when TOM amidites were used. Isolation of this product and mass spectrometric analysis using a highresolution Fourier-transform ion cyclotron resonance (FT ICR) spectrometer suggested RNA dimers that were crosslinked between two c 1 G nucleosides by a CH 2 bridge, most likely between their exocyclic amino groups (for details, see Supporting Figure S1). Such a linkage most likely forms during deprotection of the TOM group where formaldehyde emerges as a byproduct.
Base Pairing Stability of c 1 G-Modified RNA. In principle, for the nucleobase of c 1 G, tautomeric forms and distinct rotamers have to be considered. An earlier study reported the energy differences of 9-methyl-1-deazaguanine tautomers and rotamers estimated by ab initio calculations. 27 It was found that the c 1 G tautomer/syn-rotamer that we show in Figure 2A is the most stable one, followed by the anti-rotamer with 6-OH providing H-donor properties at the Watson Crick face, being 4.7 kcal/mol less stable. Importantly, the N3−H pyridone tautomer is 20.4 kcal/mol less stable. 27 c 1 G is expected to impair Watson−Crick pairing because the central N1−H of G is replaced by C−H, thereby depriving the capability for the formation of strong hydrogen bonds ( Figure  2A). The design of the RNA double helices investigated is shown in Figure 2B. The first motif (Type I) represents a bimolecular duplex of nine base pairs with a single c 1 G modification in the center. The second motif (Type II) is a hairpin with a GCAA loop (extrastable GNRA) and c 1 G residing in the center of its short stem. The third RNA motif (Type III and III′) consists of palindromic RNA of eight base pairs and equivalent purine−pyrimidine stacking patterns with the c 1 Gs either directly stacked toward each other, or separated by two standard base pairs. The type III/III′ design is very sensitive for the impact arising from a modification on base pairing. With only two or three regular Watson−Crick base pairs next to the modification, the nucleation of such duplexes can become significantly hindered. 42,43 Thus, these RNA palindromes are anticipated to significantly respond to the c 1 G modification reflected in changes of the thermodynamic pairing parameters (T m , ΔG, ΔH, ΔS).  The thermodynamic data we obtained for the three RNA systems by UV-spectroscopic melting profile measurements are summarized in Table 1 (for the corresponding melting profiles, see the Supporting Figures S2 to S11). 44,45 The native type I RNA I melts at 66.7°C ( Figure 2C). 1-Deazaguanine opposite of cytosine (Ia) destabilizes the duplex by 15.8°C. Destabilization is even more pronounced if U, G, and A are the mismatch partner (−21.8°C for Ib, −23.2°C for Ic, and −26.1°C for Id, respectively). For the hairpin RNA (Type II), c 1 G opposite of C (IIa) also decreases the melting temperature compared to the native hairpin II (by −27.9°C) ( Figure 2D). We note that duplex Ia shows a second melting transition at lower temperature, around 18°C ( Figure 2C). This may arise from a higher order structure (e.g., triplex formation) that we were not able to characterize in detail.
To further elucidate the impact of c 1 G on base pairing, we investigated the short palindromic RNAs that are particularly sensitive to double helix nucleation as mentioned above. 42,43 Indeed, for c 1 G, the destabilization in both palindromic RNAs was large, reflected in −35.9/−36.5°C reduced T m values (IIIa and III′a), accounting for −17.9/−18.3°C destabilization per single modification which is higher compared to the destabilization that we observed for a single c 1 G−C base pair in the bimolecular 9 bp duplex Ia with four regular − and hence nucleation-supportive − Watson−Crick base pairs at both 5′ and 3′ directions to the modification site.
Finally, we mention that the replacement of G in the synG•A Hoogsteen base pair of a GNRA loop in hairpin IIb was tolerated with significantly less decrease in stability ( Figure  2D). This is reasonable because the G-N1-H atom is not involved in H-bonding in the synG•A Hoogsteen pair. Of note, we observe a second low temperature melting transition for IIb, which may arise from competitive formation of a mismatched duplex.
Acid−Base Properties of 1-Deazaguanine. To understand base-pairing and catalytic properties of nucleobases in functional RNA, knowledge about their acid−base properties is crucial. 7,26 To quantify the acid−base properties of c 1 G, we conducted pH-dependent UV-spectroscopic titration experiments. Figure 3 shows an overlay of spectra for the c 1 G nucleobase that were used for pK a determinations. A value of 3.93 ± 0.07 (pK a 1) was obtained, attributed to the protonation of N7 (Supporting Figure S12). The second value of 9.10 ± 0.10 (pK a 2) was attributed to the deprotonation of the 6-OH group. The pK a values are thus comparable to the ones of guanine, which range from 9.2 to 9.6 (pK a 1, N1−H) and 3.2 to 3.3 (pK a 2, N7), respectively. 46 X-Ray Structure Analysis of a c 1 G-Modified RNA. To shed further light on the structural impact of c 1 G in RNA, we  refs 44, 45. Errors for ΔH and ΔS, arising from noninfinite cooperativity of two-state transitions and from the assumption of a temperature-independent enthalpy, are typically 10−15%. Additional error is introduced when free energies are extrapolated far from melting transitions; errors for ΔG are typically 3−5%. We note that for the biphasic profiles of Ia and IIb, the T m values and the errors were calculated for the second melting transition (between 30 and 85°C). Journal of the American Chemical Society pubs.acs.org/JACS Article aimed at a high-resolution X-ray crystallographic analysis. We utilized the 27 nt fragment of the E. coli 23S rRNA sarcin/ricin loop (SRL), which is a frequently applied crystallization scaffold ( Figure 4A). 47,48 For the replacement of G by c 1 G, we tested three different positions, including nucleotide G2669 which forms a Watson−Crick base pair with C2651 in the regular A-form double helical region, G2659 which forms a Hoogsteen pair with A2662 in the loop, and G2655 which interacts with the phosphate of G2664 in a bidentate fashion ( Figure 4B, C). From these three c 1 G-modified RNAs, only c 1 G2655-modified SRL RNA provided crystals that diffracted to atomic resolution (0.9 Å) (Supplementary Table S2). X-ray structure determination demonstrated that the c 1 G nucleobase is well defined in the electron density maps for the c 1 Gmodified RNA ( Figure 4D, E). The c 1 G-modified RNA structure superimposed with the unmodified RNA displayed a root-mean-square deviation (rmsd) of 0.09 Å (within the errors on coordinates of 0.09 Å). Direct comparison of the base triples U2656-A2665-G2655 ( Figure 4C) and U2656-A2665-c 1 G2655 ( Figure 4D) reveals that with the weakening (or loss) of the G2655 N1−H···O−P H-bond, c 1 G slightly opens up by retaining the H-bond between c 1 G2655 2-NH 2 and O4 of U2656. A comparison of the melting profiles of wild-type and c 1 G modified SRL hairpins indicated a modest weakening of the fold (Supporting Figure S13).
Taken together, our crystallization experiments imply that c 1 G does not significantly affect an RNA fold as long as it is not replacing G in a Watson−Crick base pair. The weakening (or loss) of an H-bond to the phosphate backbone seems better tolerated and allowed crystallization and structure solution.
Base Pairing Mode Switched by c 1 G. In A-form RNA, Hoogsteen (HG) base pairs are energetically disfavored relative to Watson−Crick (WC) pairs. With 1-deazaguanosine in our hands, however, we were wondering if stable HG pairing might become favorable. We thereby focused on the human immunodeficiency virus type 2 (HIV-2) transactivation response element (TAR) RNA, where a G26-C39 WC bp is adjacent to a dinucleotide bulge ( Figure 5A). It was demonstrated earlier that upon replacement of G26 by 1methylguanosine (m 1 G26), the formation of a HG base pair with C39 occurs (Figure 5B, C). 49 While in this case N1methylation represents a severe steric block at the Watson Crick face, we intended to test the hypothesis if a simple shapecomplementary modification (such as c 1 G) is sufficient to switch the pairing mode ( Figure 5B, C). Indeed, our NMR spectroscopic analysis of HIV-2 TAR RNA containing c 1 G26 revealed that the HG base pair c 1 G26(syn)-C39H + forms in a comparable manner. Characteristically, we observed a downfield shifted imino proton at ∼15 ppm ( Figure 5D, Supporting Figure S14) and downfield shifted amino protons ( Figure 5E) of C39H + that is hydrogen-bonded to syn c 1 G26. Furthermore, a strong intra-nucleotide H1′−H8 NOE cross-peak ( Figure  5F) is consistent with the syn conformation of the c 1 G26 base.
Active-Site c 1 G Impedes Twister Ribozyme Cleavage. Deazanucleobase-modified RNAs are frequently applied in atomic mutagenesis studies of ribozymes to shed light on the mechanism of the chemical reactions they catalyze. 10,18,24,50 In particular, atomic mutagenesis experiments led to an in-depth understanding of general acid−base catalysis of small nucleolytic ribozymes that cleave their phosphodiester backbone, revealing the functionally crucial imino groups of purines and pyrimidines in the active site. For instance, this concerns the twister ribozyme 51 where proton transfer from the (protonated) N3 of a conserved adenine (A6) at the cleavage site to the 5′-O leaving group significantly contributes to reaction catalysis; the replacement of this adenine by c 3 A or c 1 c 3 A rendered the twister ribozyme inactive. 15,52 Another example is a phosphodiester cleavage by the pistol ribozyme. 53 Replacements of active site purines by c 3 A, c 1 A, and c 7 G revealed the key residue − a highly conserved guanine (G33) − that serves as inner sphere coordination site for a hydrated Mg 2+ ion, thereby likely providing a 1st shell water molecule as general acid for protonation of the 5′-O leaving group in the course of the reaction. 16  Access to c 1 G-modified RNA now allows evaluation of active site guanines that are suspected to be involved in reaction catalysis via their Watson−Crick face. To exemplify this, we picked the three-way junctional twister ribozyme, 53 for which several structures of precatalytic states were solved by X-ray crystallography 15,54−58 and structural dynamics elucidated by smFRET imaging. 10,59 Bases U5 and A6 at the cleavage site are splayed apart, with a guanine (G48 for PDB ID 4RGE 54 and 5DUN 15 ) in a hydrogen bond distance (2.6 Å) between the N1 atom and the nonbridging pro-R oxygen of the scissile phosphate ( Figure 6A). The structures implicate that G48 may play a significant role in phosphorane transition state stabilization. Furthermore, several studies propose the hypothesis of concerted general acid−base catalysis for twister in which G48 acts as the general base ( Figure 6B). 52,56,60  With c 1 G in hand, we can now probe whether or not the NH donor of the Watson−Crick face of the suspected G48 is indeed a determinant in reaction catalysis (i.e., β and/or γcatalysis according to refs 61, 62). Involvement of G48 as general base in phosphodiester cleavage catalysis implies either N1-deprotonation or enol tautomerization or at least hydrogen bonding to activate the attacking 2′-OH nucleophile. Also, stabilization of the pentavalent phosphorane transition state is conceivable as extrapolated from the G48N1−H···OP interaction seen in the crystal structure. 54 All these scenarios are severely affected upon the replacement of G48 by c 1 G; we therefore anticipated that cleavage becomes abolished. We, however, found that cleavage still occurs, albeit with a 275-fold reduction in rate ( Figure 6C−F and Supporting Figures S15, S16).
The remaining cleavage activity indicates that the other catalytic determinants (i.e., αand δcatalysis according to refs 61 and 62) are sufficient to achieve residual activity of c 1 G48modified twister ribozyme. We further note that the 6-OH group of c 1 G (possessing a comparable pK a to N1−H of G) is dislocated in comparison with N1 (in G) and therefore is not likely to be able to efficiently take over its role. However, the 2-NH 2 group of guanine is present also in c 1 G, and therefore, this NH 2 group can contribute to stabilization of the phosphorane transition state (together with a remaining weak stabilization originating from a C1−H interaction with phosphorane).

■ CONCLUSIONS
Our study introduces robust syntheses of c 1 G, the corresponding phosphoramidite, and c 1 G modified RNA. The synthetic foundation enabled comprehensive analysis of the biophysical properties of such modified RNA, and furthermore, enabled c 1 G atomic mutagenesis in functional RNA assays. This led to evidence for c 1 G switching the mode of base pairing from Watson−Crick to Hoogsteen. Moreover, the approach allows direct evaluation of ribozyme mechanistic proposals that claim a catalytic role for guanosines via their N1 position, the central H-donor of their Watson−Crick face. Beyond twister, such guanosines are found in many ribozymes including twister sister, pistol, hatchet, and the most recently discovered RNA methyltransferase ribozymes. 18,24,63 Functional atomic mutagenesis relying on c 1 G RNA modifications will contribute to achieve an in-depth understanding of RNA catalysis of ribozymes that exhibit a much broader reactivity scope than previously anticipated. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
We thank Marco Oberlechner (University of Innsbruck) for synthetic support. We thank Vincent Olieric for help during data collection at SLS, Villigen. We thank Daniel Fellner and Ulrike Schober (University of Innsbruck) for technical support. We thank Elisabeth Mairhofer (University of Innsbruck) for discussions.