LNA-amide monomer synthesis is efficient and scalable
Efficient synthesis of chemically modified oligonucleotides is essential to fuel fundamental studies and therapeutic applications. We have devised a strategy which enables the straightforward assembly of oligonucleotides containing LNA-flanked amides which could be easily automated and scaled up. A maximum of eight monomers are required to make any sequence that contains non-contiguous amide linkages (four carboxylic acids and four amines).
Attaching the required ethanoic acid moiety to the 3´-carbon of the LNA sugar is challenging as it requires removal of the 3´-oxygen and formation of a C-C bond. However, we were able to produce the required 5´-dimethoxytrityl (DMT)-protected 3´-ethanoic acid LNA-monomers in 8 steps from 1, the same number as for conventional LNA phosphoramidites43 (Fig. 2a). Moreover, we did this with minimal chromatography. We first built the sugar with the C3´-ester 5, before addition of the nucleobase. This approach avoids complications associated with forming a C-C bond at the 3´-side of a nucleoside using the Barton-McCombie reaction23-25, 44 or the hydrogenation step if a Wittig reaction is used45, 46, which would ultimately limit the range of heterocyclic bases that can be added. Key intermediate 5 was readily prepared on multiple gram scale in 81% overall yield from commercially available 1 without the need for chromatographic separation as follows. Hydrogenolysis of compound 1 afforded alcohol 2 which was subsequently oxidised to give compound 3. Olefination of 3 with (carbethoxymethylene)triphenylphosphorane (Wittig reaction) selectively yielded 4 as the (E)-stereoisomer (Fig. 2b). Catalytic hydrogenation of 4 using Pd/C and H2 gave 5 as a single stereoisomer (Fig. 2b). This stereoselectivity was predicted because the 1,2-O-isopropylidene groups on the α-face of furanosyl carbohydrate derivatives direct incoming H2 to the β-face47. Key intermediate 5 was then converted to the 1,2-di-O-acetate glycosyl donor 6 following a procedure reported by Arzel et al.48, avoiding the formation of a lactone which occurred when the acetonide was cleaved in the presence of water.
The pathways to each monomer then diverged, with Vorbrüggen conditions43, 49 utilised for addition of the nucleobases to access 7a-e. Subsequent simultaneous unmasking of the 3´-carboxyl and 2´-hydroxyl groups by treatment with hydroxide, followed by cyclisation to form the 2´- 4´-oxymethylene bridge, then 5´-mesyl deprotection, gave the hydroxy-LNA acid compounds 8a-e. The progress of the reaction was rapid; mesyl deprotection using hydroxide ion conventionally requires several days under reflux conditions43. We postulate that the acceleration in rate is due to neighbouring group participation whereby the carboxylate anion displaces the 5´-mesyl group, forming a lactone that is subsequently opened by hydrolysis (Supplementary Fig. 1). Finally, we treated the resulting hydroxy-acids 8a-e with 4,4´-dimethoxytrityl chloride (DMT-Cl) in pyridine to give the DMT-protected LNA analogue nucleosides 9a-e.
Using this strategy we were able to access all four canonical nucleoside analogues along with the 5-methylcytidine monomer which is used in place of cytidine in antisense experiments to increase target affinity and improve other therapeutic properties18. Additionally, we required 5´-MMT-amino LNA phosphoramidite 10. Whilst this had been previously synthesised40 we chose to develop a more efficient route (Supplementary Fig. 2). Commercially available 5´-MMT-amino dT 11 and 5´-DMT thymidine-3´‑ethanoic acid 1241, 42 (Fig. 2c) were also required to enable us to make oligonucleotides to compare the properties of DNA amide with those of LNA-amide.
Chimeric oligonucleotides can be synthesised in high purity
Our oligonucleotide synthesis strategy is shown in Fig. 3. A phosphoramidite monomer with an MMT-protected 5´-amino group, either LNA 1040 or deoxythymidyl 11, is added to the oligonucleotide, and the amine is deprotected using trichloroethanoic acid. An LNA-acid (or DNA-acid50) monomer is coupled to the free amine using PyBOP activating agent in the presence of a non-nucleophilic base (N-methylmorpholine) to form the amide bond. Oligonucleotide synthesis is then resumed, starting with removal of the DMT group.
The process is repeated to install multiple non-contiguous amides in the same oligonucleotide. To demonstrate this, DMT-protected LNA acids 9a-e, phosphoramidites 10 and 11, and DNA acid 1241, 42 (Fig. 2), were used to synthesise several oligonucleotides, some of which contain multiple additions of LNA-amide, 2´-OMe sugars and PS linkages (Supplementary Table 1). In all cases, we obtained the oligonucleotides in high purity. High-performance liquid chromatography (HPLC) and mass spectrometry demonstrated the high incorporation efficiency for the DMT-protected LNA acids 9a-e (Supplementary Fig. 3-7).
LNA sugars stabilise duplexes containing the amide backbone
To evaluate the ability of the LNA-amide combination to bind to complementary RNA with high affinity we synthesised a series of 13-mer ONs with a central amide with a T-T sequence. These constructs were composed of either no LNA (ON1DNA-Am-DNA), an LNA 5´ to the amide (ON2LNA-Am-DNA), an LNA 3´ to the amide (ON3DNA-Am-LNA), or LNA on both sides of the amide (ON4LNA-Am-LNA, Fig. 4a). Controls without amide (ON5LNA-LNA and ON6DNAcontrol) were also made. We compared duplex denaturation temperatures (Tms) after hybridisation with DNA and RNA complementary strands (Fig. 4b, Supplementary Table 2, Supplementary Fig. 8 and 9). ON2LNA‑Am-DNA produced a significant increase in DNA:RNA hybrid stability, (+3.0 ˚C) compared to the unmodified ON6DNAcontrol, and an increase of +3.4˚C compared to ‘amide only’ ON1DNA-Am-DNA. Importantly, ON4LNA‑Am-LNA, in which the amide is surrounded by LNA sugars, gave the greatest increase in stability (+5.1 ˚C). It is noteworthy that ON2LNA-Am-DNA and ON4LNA-Am-LNA provide the first examples of duplex stabilisation by an LNA sugar attached to a non-phosphorus artificial DNA backbone. The RNA target selectivity of LNA-amide-containing ONs was excellent; a single mismatched base pair greatly reduced duplex stability, in some cases by >14 ˚C (Supplementary Table 2, Supplementary Fig. 10-13) In summary, an amide linkage flanked by LNA on both sides gives strong DNA:RNA duplex stabilisation and good mismatch discrimination.
In duplexes with DNA targets, ONs with all combinations of LNA and DNA sugars around the amide linkage were slightly destabilising (between ‑0.1 ˚C to ‑2.6 ˚C), indicating the selectivity of the amide linkage for complementary RNA. The stabilisation induced by the LNA amide combination is cumulative and general (Fig. 4c, Supplementary Table 3, Supplementary Fig. 14-17). In a biologically relevant sequence context, four LNA-amides increase duplex stability against complementary RNA by an impressive 13.0 ˚C compared to only 5.1 ˚C for the DNA target. This large difference is important when developing oligonucleotides to interact with RNA. Oligonucleotides for in vivo studies usually contain 2´-OMe modified sugars and/or phosphorothioate backbones to prevent degradation by nucleases. In such oligonucleotides the combination of LNA and amide also greatly increases duplex stability (Supplementary Table 3).
Combining LNA and amide provides strong nuclease resistance
Therapeutic oligonucleotides must remain stable in cells for prolonged periods to remain active. To evaluate whether the combination of LNA and amide confers greater nuclease resistance than LNA alone, we incubated unmodified DNA (ON15DNA/17PO) and DNA with four LNA-amide linkages (ON14 DNA/4LAL/13PO) in a 1:1 mixture of phosphate buffered saline (PBS) and foetal bovine serum (FBS) to mimic the in vivo environment, and compared it to the equivalent construct with LNA but without amide linkages (ON25DNA/8LNA/17PO, Supplementary Fig. 18). The results show that the combination of LNA and amide confers extreme resistance to nucleases. Both the oligonucleotides lacking amide linkages had partially degraded within 1 hour whereas ON14DNA/4LAL/13PO remained intact after 8 hours. Interestingly ON14 shows stability at its 3´-end, even though this region has an unmodified sugar phosphate trimer. The 3´-terminal pentamer region, however, is highly modified. It has reduced charge due to the amide linkage and also contains two LNA sugars. It may therefore not be recognised by nucleases. This enhanced stability further illustrates the advantages of removing charge and including modified sugars in antisense oligonucleotides.
X-ray crystallography of the LNA-amide modification
We solved several X-ray structures to determine the effects of LNA and amide modifications on duplex structure and conformation. These are the first crystal structures of DNA:RNA hybrids that contain amide linkages. An amide-modified RNA:RNA duplex was analysed previously, but this had amides in both strands surrounded by multiple mismatched base pairs which cannot exist outside the solid-state at ambient temperature27. The sequence of the modified DNA:RNA hybrid duplexes (Supplementary Table 4) was based on the corresponding unmodified version (d-CTTTTCTTTG/rCAAAGAAAAG)51 (the location of the amide is underlined). Good quality crystals diffracting between 2.5-2.8 Å resolution were obtained for the DNA:RNA hybrid in which the DNA strand contains an amide linkage flanked by DNA on both sides, LNA on both sides and with LNA only on the 5´-side (Supplementary Fig. 19). The unmodified DNA:RNA duplex was also studied. The data collection and refinement statistics are given in Supplementary Table 5. Electron density maps at the modification position for the four nucleic acid crystal structures reported in this manuscript are given in Supplementary Fig. 20.
The hybrids with amide and LNA-amide backbones (Fig. 5a) are structurally very similar to the unmodified duplex (all-atom RMSD 0.4 Å) as shown by their superimposition (Fig. 5a, Supplementary Fig. 21). All structures adopt the A-conformation with sugar puckers clustering around C3´-endo (Supplementary Fig. 22). As expected, all duplexes are stabilised by canonical Watson-Crick base pairs, indicating that the thermodynamic improvements due to the LNA-amide backbones are not due to unusual changes in hydrogen bonding interactions. In agreement with the DNA:RNA hybrid NMR structure by Rosners22 in which the DNA strand contained multiple amides, our X-ray studies indicate that the amide linkage is a close mimic of the phosphodiester backbone (Fig. 5b). Both are four-atom linkages, hence similar in length, and the amide carbonyl is orientated in the same direction as one of the phosphodiester P‑O bonds.
In Fig. 5c, the structures of all amide backbones are overlaid to assess the effects of the LNA modifications. Between each structure, the orientation of the backbone is consistent, directing the amide oxygen into the major groove. Other atomic positions of the backbones also show close similarity, and the presence of 3´-LNA causes no significant distortion. 5´-LNA does however cause some structural displacement; the 5´-sugars in the LNA-amide-DNA and LNA-amide-LNA structures are shifted slightly outwards compared to the DNA-amide-DNA and unmodified strands. Despite this, the positioning of the amide backbone remains consistent between each structure. The amide adopts the expected trans-conformation, and LNA on the 5´-side of the amide has little effect on backbone torsion angles (Fig. 5c, Supplementary Fig. 23). In summary, combined LNA and amide modifications have minimal effect on the duplex structure, and are excellent mimics of natural phosphodiesters.
Combining LNA-amide and PS enhances gymnotic delivery
In a preliminary study we have evaluated the biological activity of the LNA-amide combination using the HeLa pLuc/705 cell line52 that carries a luciferase-encoding gene interrupted by a mutated ß-globin intron52. This mutation creates a 5´-splice site which activates a cryptic 3´-splice site, resulting in incorrect mRNA splicing and the production of non-functional luciferase. An oligonucleotide that hybridises to the mutant 5´-splice site prevents incorporation of the aberrant intron, restoring the pre-mRNA splicing to produce functional luciferase, which is quantified by luminometry. Oligonucleotides complementary to this aberrant splice site were synthesised with combinations of different modifications to determine their individual effects (Supplementary Table 3). ON14DNA/4LAL/13PO, ON162¢OMe/4LAL/13PO, and ON182¢OMe/4LAL/13PS were designed to evaluate LNA-amide with the DNA, 2´‑OMe/phosphodiester, and 2´-OMe/phosphorothioate backbones respectively, and ONs 14 and 16 were included to determine the degree to which LNA-amide influences delivery, activity and toxicity in the absence of PS linkages. LNA and amide linkages are incompatible with RNase-H so there is no risk of ON14 or ON16 inadvertently destroying the RNA target2,56. Three controls were included: ON202¢OMe/17PS (which represents the gold standard in the assay) to determine whether LNA-amide enhances biological activity52, ON172¢OMe/17PO to evaluate the effects of the PS linkage independently of LNA or amide linkages, and ON192¢OMe/8LNA/17PS with LNA sugars but no amide linkages to determine the effects of the enhanced duplex stability caused by LNA. A scrambled control with a 2´-OMe/PS backbone was also included to determine off target effects (ON312¢OMe/17PS scrambled, Fig. 6).
To compare biological activity independent of cell uptake, Lipofectamine 2000 (LF2000), a cationic liposome transfection/delivery reagent, was used. All three target-complementary PS-ONs were active in the assay (ON202¢OMe/17PS, ON182¢OMe/4LAL/13PS, and ON192¢OMe/8LNA/17PS), whereas all the PO-ONs (ON14DNA/4LAL/13PO, ON162¢OMe/4LAL/13PO and ON172¢OMe/17PO) were inactive at 100 nM (Fig. 6a). Hence, in agreement with previous studies, the phosphorothioate modification in a target-complementary sequence is necessary for splice-switching activity. This could result from the PS groups enhancing nuclear enrichment53 of the oligonucleotides, and/or recruiting ILF2/3 to the RNA transcript54. Notably, the addition of the amide linkage significantly improved the splice-switching activity of 2´-OMe/PS ONs at the lower concentrations (6.25 nM and 12.5 nM), probably due to improved target affinity (Fig. 6b).
Next, we compared the naked (gymnotic) uptake of the ONs. These conditions more closely represent in vivo applications where transfection agents such as LF2000 cannot be used. We seeded cells at low confluency, added the oligonucleotides in fresh media after 16 h, and measured luciferase activity after a further 96 h. The presence of just four LNA-amides (ON182¢OMe/4LAL/13PS) significantly increased activity in a dose-dependent manner compared to ON202¢OMe/17PS (Fig. 6c). Greater than 5-fold increase in activity was observed for gymnotic delivery compared to a maximum of 3-fold increase for the LF2000‑mediated transfection. This suggests that synergy between the PS and LNA-amide modifications leads to enhanced productive delivery into cells. The improved therapeutic properties of LNA-amide/PS oligonucleotides observed in these preliminary studies are possibly due to a combination of reduced charge from the LNA-amides and interactions of the PS backbone with cellular components. Improved cell uptake may result from the neutral amide linkages breaking up the poly-anionic backbone into short segments which penetrate cells more readily than long poly-anionic stretches. Interestingly, ON192¢OMe/8LNA/PS with LNA and no amides showed only slight dose response in activity, even at the highest concentration tested (Fig. 6c, Supplementary Fig. 24). This could be due to its binding to off-targets, altered rigidity, or undesirable secondary structures induced by the extreme stability caused by the LNA sugars, reducing the ability of the ON to interact with the cell surface, a mechanism for productive uptake. ON162¢OMe/4LAL/13PO with LNA-amides and no PS linkages also displayed slight gymnotic splice-switching activity (Supplementary Fig. 24).
We compared the viability of the HeLa cells following lipofection using a WST-1 cell proliferation assay (Fig. 6d). At the highest concentration tested (400 nM) the cells treated with ON202¢OMe/17PS were only 21% viable, whereas the cells treated with the same concentration of ON182¢OMe/4LAL/13PS were 50% viable, demonstrating that the LNA-amide linkage significantly reduces the cytotoxicity of ONs delivered with LF2000. This is verified by analysis of the protein levels (Supplementary Fig. 25) and visible cell death (Fig. 6e, and Supplementary Fig. 26). It supports the use of the combination of LNA-amide, 2´-OMe and PS modifications for in-vitro studies. Interestingly, the oligonucleotide containing eight LNA sugars without amide linkages (ON192¢OMe/8LNA/17PS) had a poor toxicity profile. This could be due to off-target effects and could also explain why, despite showing the highest affinity towards RNA in the UV melting studies, the LNA modified ON192¢OMe/8LNA/17PS was not the most active in the exon-skipping assay. Further detailed studies are required to determine whether this is a general or a sequence specific phenomenon, and to validate its relevance in terms of toxicity.
Given that cell uptake and toxicity remain major challenges when developing new therapeutic oligonucleotides the results in Fig. 6 suggest that our modification strategy could be advantageous. The synergistic effect of LNA-amide linkages and phosphorothioate modifications appear to produce oligonucleotides with enhanced biological properties. However, cell culture studies cannot address some of the major challenges in oligonucleotide therapeutics, particularly those that relate to pharmacokinetics, pharmacodynamics, biodistribution and aspects of toxicity. We are therefore planning further detailed biological studies on LNA-amide-phosphorothioate oligonucleotides to answer these important questions.