Baran et al. Science: An idealized platform for oligonucleotide synthesis

In traditional small molecule drug development, the determination of target specificity is often inseparable from the structural characteristics of DMPK (distribution, metabolism, and pharmacokinetics), making it difficult to design and optimize independently of each other. In contrast, the pharmacophores and pharmacokinetic properties of oligonucleotide drugs can theoretically be optimized separately, as the nucleoside sequence of oligonucleotide drugs directly determines the former, while the chemical properties of their phosphate backbone greatly affect the latter. At present, there are over 155 ongoing clinical trials of oligonucleotide drugs, and multiple drugs have been approved for market, most of which contain modified phosphate ester skeletons. Unfortunately, despite the profound impact of advances in organic synthesis on the development of modern medicinal chemistry, there have been few significant advances in the basic chemistry of oligonucleotide synthesis as the range of oligonucleotide sequences and possible phosphate backbone expands. As shown in Figure 1, a chimeric oligonucleotide sequence containing four different phosphate linkage structures and multiple sugar skeletons, if synthesized using existing methods, faces enormous difficulties and to some extent highlights the limitations of existing methods. Ideally, a wider range of combinations and variations can be installed in any order, making it even more unattainable. In this way, the commercialization of oligonucleotides, and more specifically, the highly anticipated phosphorothioate antisense oligonucleotides (PS-ASO), faces significant challenges.

In 2018, Professor Phil S. Baran from The Scripps Research Institute (TSRI) and researchers Ivar M. McDonald, Martin Eastgate, Michael Schmidt, and others from pharmaceutical giant Bristol Myers Squibb (BMS) collaborated to develop a phosphorus sulfur reagent based on P (V) (referred to as PSI or PSI) to solve the synthesis problem of stereopure thiophosphate (R-PS and S-PS, mainly in the DNA environment) oligonucleotides (Science, 2018, 361, 1234-1238, click to read more). However, this preliminary study also has unresolved issues, such as the installation of other types of phosphate backbone linking structures such as dithiophosphates and natural phosphodiesters, other sugar chemistry of lock-in nucleic acids (LNA), and their applicability to modern automated synthesis. To solve these problems, there are many challenges to be faced, such as: 1) P (V) reagent was once considered to have a too slow reaction rate to compete with P (III) disproportionation; 2) The influence of guanine and thymine bases can lead to different chemical selectivity; 3) The use of existing P (III) - based reagents to install dithiophosphate linkage structures is not ideal; 4) The current methods for modifying phosphate ester skeletons have complex steps, including oxidation, deprotection, and product separation. Given this, it is particularly important to develop a highly selective and compatible automated oligonucleotide synthesis method, which will have a profound impact on the development and commercialization of oligonucleotide drugs.

On the basis of previous work, Professor Baran and others have recently made significant progress. They developed a universal P (V) platform that is completely different from P (III) - based oligonucleotide synthesis by utilizing three new reagents, namely PSI 2 (3), rac - PSI (4), and PSI O (5), and combining them with the reported [(+) - PSI, (+) -2] and [(-) - PSI, (-) -2] systems, achieving arbitrary controlled synthesis of specific oligonucleotide sequences (Figure 1C). This method not only reduces the dependence on protecting group chemistry, special reagents, oxidizing reagents, and other issues, but also eliminates a complete step in the standard solid-phase oligonucleotide synthesis (SPOS) protocol, namely the phosphorus oxidation process. The relevant results were recently published in Science.

Compared to the synthesis of PSI 2 and PSI O, the synthesis of rac - PSI is relatively simple, and can be obtained by analogy with PSI using epoxycyclohexane as the raw material. For this purpose, the author explored the synthesis of a PSI O reagent containing dithiophosphate ester PSI 2 and natural phosphodiester (Figure 2). Although chemists reported in 1995 that dithiophosphlanes can be mounted on nucleosides and coupled to them to synthesize dinucleotides containing dithiophosphate linkage structures, the installation of sulfur requires a separate oxidation step and toxic and unstable (explosive) reagents (Figure 2A). In view of this, the author chose P (V) as the center to design and synthesize the dithiophosphate ester PSI 2 with the aim of removing oxidation steps and hazardous reagents. After determining the optimal leaving group and ring size, the author found that inexpensive P2S5 can be combined with pentafluorophenol, and then reacted with P (V) intermediate (6) using cyclohexane to synthesize PSI 2 (3) containing dithiophosphate structures on a large scale (>100 g). Similarly, the author evaluated nearly 30 different skeletons and 3 different leaving groups, among which skeleton optimization systematically assessed the effects of ring size, substituents, electronic effects, and stereochemistry on steric hindrance, coupling, and overall stability, and found that ψ O is a feasible reagent. Subsequently, the author synthesized the PSI reagent in a simple and scalable manner (>50 g) (Figure 2A). Specifically, cheap P2S5 reacts with 4-bromophenylthiophenol to obtain P (V) intermediate (7), which combines with hydrogenated cis limonene epoxide (8) to form PS reagent (9). Subsequently, it can be desulfurized with SeO2 to obtain PSI O (5).

Next, the author compared the P (V) platform with the currently advanced P (III) method (Figure 2B), and the results showed that the P (V) platform has simpler steps, higher product purity (>99%), and does not produce PS impurities compared to the P (III) method. In order to further test and compare the advantages and disadvantages of the two platforms, the author explored the synthesis of a mixed PO-PS main chain. Taking the P (III) method as an example, PS dimer (13) undergoes phosphoramidite coupling reaction with PA dT (18) to obtain a protected trimer (14), which is oxidized to obtain a mixture of the target product and desulfurization by-products. In contrast, the use of redox neutral P (V) method can successfully prepare unprotected mixed PS-PO trimer (19) without any sulfur loss.

Another major challenge faced by this P (V) method is whether it can overcome the slow coupling rate of classical P (V) reagents to adapt to traditional automated oligonucleotide synthesis schemes. The author evaluated the coupling performance of the P (V) platform based on a complete set of P (V) reagents using kinetic studies. The results showed that the classical P (V) - based phosphate triester method was very slow (Figure 2C, orange bar), while the P (V) reagent platform described in this article had the same performance as the industry standard P (III) scheme, with all reactions achieving complete conversion within two minutes.

Over the past 30 years, the SPOS method based on phosphoramidite synthesis has been continuously optimized. Although some of these methods may still be usable, existing solutions are not compatible with P (V) synthesis methods in certain aspects (Figure 3A). At present, the universal carrier is not stable enough for 1,8-diazabicyclo [5.4.0] undec-7-ene (DBU). Therefore, the authors used Pya protecting group instead of the standard amide protecting group to develop a universal carrier with significantly improved stability (20, Figure 3B). Meanwhile, better results were also obtained when Pom protection was applied to thymidine. Secondly, after determining the chemical selectivity and relative coupling rate of all P (V) reagents, the author systematically studied the efficiency of the redox neutral P (V) platform on automated SPOS. In the previous step, the DMT group of the resin bound nucleoside was deprotected to obtain free 5-ol, which reacted with the nucleotide of P (V), followed by capping and unblocking steps to complete the solid-phase cycle and lay the foundation for the next coupling.

Next is the practical test of synthesis. Any new reagent system used for oligonucleotide synthesis platforms must first demonstrate high fidelity and robustness when preparing different sequences using a single protocol. In order to test the purity of the new P (V) platform, the authors designed an oligonucleotide molecular structure "matrix" to introduce all possible combinations of nucleobases (A, C, G, T) and sugars (DNA and LNA) into the 3-10-3 DNA/RNA gapmer framework, which is currently one of the more advanced techniques for RNase H-activated antisense oligonucleotides (Figure 3C). Whether the P (V) monomer is used to evaluate the generality of the method or for sequence specific optimization, the authors used a single approach. Firstly, the author tested the generality of this method and synthesized homogeneous, chiral PS-ASO with alternating (21,22) and continuous stereochemical structures (23-26). It is worth mentioning that this method of using redox neutral P (V) - based reagents is the second industrially feasible platform for producing stereopure PS-ASO. With the achievement of this important goal, the author began to introduce PO2 linking structures into these molecular structures, obtaining chimeric sequences (27-30) with high purity without substantial loss of sulfur during the synthesis process. Next, the author prepared oligonucleotide sequences containing both PS and PS2 linking structures (31-34), as well as oligonucleotide sequences containing all four possible linking structures (35, 36). Later, in order to demonstrate the advantages of this P (V) oligonucleotide synthesis platform, the authors synthesized racemic PS oligonucleotides (37,38) on the platform based on the derivative rac - ψ of ψ (Figure 1) (Figure 3D).

summary

Oligonucleotide therapy can directly regulate gene expression and is considered a new hot topic in drug development after small molecule drugs and protein drugs. So far, the number of oligonucleotide modifications studied is very limited, and their clinical applications are even fewer. Ultimately, the reason is that the synthesis level cannot keep up. The synthesis of oligonucleotides based on P (V) reagents has long been available, but due to issues such as low chemical selectivity and poor reactivity, it cannot take on significant responsibilities. The work of Professor Baran et al. can be said to rejuvenate the P (V) strategy, making it possible for people to obtain desired oligonucleotide chimeric sequences. More importantly, this P (V) platform is compatible with commercial oligonucleotide automated solid-phase synthesis systems and has a bright application prospect. Although the separation yield of the platform (12-27%) is not too high, after some optimization, it should not be a big problem in the future. Even now, this level of yield is sufficient to support laboratory research in medicinal chemistry.