Making cyclic RNAs easily available Miriam Frieden, Anna Grandas and Enrique Pedroso* Departament de Qu´ ımica Org` anica, Facultat de Qu´ ımica, Universitat de Barcelona, Mart´ ı i Franqu` es 1-11, 08028 Barcelona, Spain. E-mail: pedroso@admin.qo.ub.es Received (in Cambridge, UK) 17th June 1999, Accepted 13th July 1999 A simple solid-phase procedure allows cyclic oligoribonu- cleotides to be obtained as long as the linear precursor attached to the support has a 2A-deoxyribonucleoside or a 2A- O-methylribonucleoside at the 3A-end. Some of the smallest oligoribonucleotides with known bio- logical activity are cyclic molecules. For instance, c(GG) is an activator of cellulose synthase in Acetobacter xilinum, 1 and c(UU) and c(AU) are inhibitors of the DNA dependent RNA polymerase of E. coli. 2 At the other end of the scale, large cyclic RNAs are formed in the splicing processes of ribonucleic acids in certain organisms 3 and viroids have circular single stranded RNA as their genomic material. The structure of cyclic RNAs has been determined either by X-ray diffraction 4 or by NMR 5 only for very small molecules. Cyclic RNAs may be useful models for the study of a variety of RNA structural motifs, 6 such as hairpin loops, which could bring new insights into the structure–function relationship of ribonucleic acids. In spite of their wide potential applications, few efforts have been dedicated so far to the development of chemical methods for the preparation of cyclic oligoribonucleotides. The synthesis of cyclic dinucleotides 7 and tetranucleotides 8,9 has been reported using the phosphotriester 7,8 or the H-phosphonate 9 methods in solution, whereas a solid-phase method has been employed for the preparation of cytidine homooligomers. 10 Larger cyclic RNAs have been obtained either by template- directed chemical 11 or enzymatic 12 ligation, and by rolling circle transcription and self-processing of circular DNA oligonucleotides encoding hairpin ribozymes. 13 We have recently described a straightforward solid-phase procedure for obtaining small to medium-sized cyclic oligo- deoxyribonucleotides. 14 The main advantage of the method is that, after the cyclization and cleavage reactions, non-cyclized chains, polymers and other by-products remain attached to the solid matrix. Therefore, fairly pure crude products are obtained, regardless of the size of the circles and the sometimes low cyclization yields. Scheme 1 shows the key steps in the extension of this methodology to the preparation of cyclic RNA. Synthesis of the ribonucleotide-resin 1 and chain elongation by the phosphite-triester approach using 2-cyanoethyl (CNE) phosphoramidites yields the oligonucleotide-resin 2 which, upon cyclization with 1-mesitylenesulfonyl-3-nitro-1,2,4-tria- zole (MSNT), affords 3. After oximate cleavage of 3 and deprotection the cyclic RNA is obtained. The choice of a suitable 2A-OH protecting group is the most crucial decision in RNA synthesis. It is well documented that the last step of the synthesis must be the deprotection of the 2A- hydroxy functions to avoid strand cleavage or 3A–5A to 2A–5A migration of the phosphodiester linkages. 15 The commonly used TBDMS group was discounted because of its lack of stability towards the oximate cleavage treatment. 16 Early elimination of the TBDMS group of the ribonucleoside directly attached to the support may result in the above mentioned undesirable side reactions. For this reason, we turned our attention to the acid-labile 1-(2-fluorophenyl)-4-methoxypiperidin-4-yl group (Fpmp), 17 whose stability to oximate was confirmed using 2A-O-Fpmp- uridine (data not shown). However, to our surprise, the first attempts to prepare small cyclic RNAs (2- to 6-mer) employing the Fpmp group were completely unsuccessful: very low yields and impure crude products were obtained. We reasoned that the key difference with respect to cyclic DNA synthesis was the presence of the bulky 2A-O-Fpmp group at the 3A-end of 2 that may be hindering the cyclization reaction. In order to test the validity of this assumption, three nucleotide resins 1 having differently hindered phosphate groups were prepared as pre- viously described 14 on an NH 2 -TentaGel™ resin, 1a (BA = T, RA = H), 1b (BA = U, RA = OMe) and 1c (BA = U, RA - OFpmp), and their homogeneity was assessed by gel-phase 31 P NMR.† On these nucleotide-resins a series of dinucleotides were assembled, cyclized, cleaved and deprotected.‡ Nucleoside sequences, yields and mass spectrometric data are indicated in Table 1 (entries 1 to 5). These results clearly indicate that the lowest yields were obtained when the nucleotide-resin 1c was employed (entries 4 and 5). In fact, c(UU) could not be isolated from the complex crude mixture, whilst the rest of the cyclic dinucleotides were easily purified by C18-HPLC chromatog- raphy and unequivocally characterized by mass spectrometry and enzymatic digestion. 14 Particularly striking is the difference in yield between the two syntheses of c(UT) [ = c(TU)] (entries 3 and 4): much lower yield was obtained from the most hindered resin 1c than from 1a, thus demonstrating that steric hindrance in the vicinity of the 3A terminus phosphate of linear precursor 2 plays a key role. Nevertheless, comparing the results of entries 4 and 5 we can assert that the rest of the chain has a non- negligible effect. Most probably, it hinders the dinucleotide in attaining a reactive conformation. Scheme 1 Chem. Commun., 1999, 1593–1594 1593