Unlocking P(V): Reagents for chiral phosphorothioate synthesis of nucleic acid based therapeutics.

[NNadir]

The paper I'll discuss in this post is this one:

Unlocking P(V): Reagents for chiral phosphorothioate synthesis

In many ways, I was a pretty stupid kid. I recall telling people that no one would ever make a protein drug.

Then they started being approved; and in recent years, they have come to dominate the pharmaceutical formulary, proving to be highly effective drugs.

In the past few years, we have begun to see drugs that are nucleic acids, designed to treat genetic diseases. They work quite well but tend to be very expensive because of the difficulty of synthesizing them on scale.

There was a time in my life that I was offered a job working on this sort of thing; I turned it down because I thought "that will never happen."

One of the pleasures of growing old is recognizing all the times you've been stupid.

One of the problems with DNA therapeutics is their kinetics; they tend to be broken down quickly by enzymatic processes in vivo. One mechanism for slowing the pharmacokinetics is to change the normal phosphate linkage group that builds the "sugar stairway" of nucleic acids to a thiophosphate, changing the oxygen in the phosphate to a sulfur.

An interesting thing about this is that doing so makes the phosphorous chiral, that is it is not superimposable on its mirror image. (If you have two hands, they are chiral, and the overwhelming majority of molecules other than water in your body are chiral.)

This allows for subtle control of the properties.

I just saw a lecture on this topic. I hadn't thought about "chiral phosphorous" in years, so it was cool to be reminded of this remarkable feature.

Mixed chirality can destroy a drugs effectiveness. An oligonucleotide that is 18 residues in length, for example, with 18 racemic (mixed chirality) thiophosphate groups would actually be 2^18 different compounds = 262,144 different compounds, only one of which would be effective, and many of which might be expected to be toxic.





From the paper:

Molecular-based gene therapy is rapidly emerging as one of the most promising modes of disease amelioration, as evidenced by the swell of clinical trials employing microRNA, small interfering RNA, and antisense oligonucleotides (ASOs) (1). In a recent example, the ASO nusinersen (trade name Spinraza) (Fig. 1A) was approved by the U.S. Food and Drug Administration (FDA) as the sole available treatment of spinal muscular atrophy, the leading genetic cause of infant mortality (2). This breakthrough therapy comprises 18 nucleosides bound by phosphorothioate (PS) linkages (i.e., those in which one of the nonbridging oxygen atoms has been substituted with a sulfur). Compared with the native phosphodiester linkage, this simple modification confers a higher degree of metabolic stability and improved cellular uptake through an observation known as the thio effect (3–6). In fact, most FDA-approved ASOs incorporate this moiety, with numerous candidates under clinical evaluation (7). Although the PS alteration enhances the pharmacological profile of these systems, it comes at the expense of substantially increased structural complexity—each phosphorus atom is now an uncontrolled stereogenic center (8). Thus, in the case of Spinraza, patients receive a mixture of more than 100,000 discrete diastereoisomers that, in principle, may bear distinct three-dimensional structures and pharmaceutical properties (9, 10). Despite recent studies suggesting that stereodefined systems may improve therapeutic efficacy, readily scalable preparation methods often disregard the stereochemistry at phosphorus, whereas most of the methods that enable stereocontrol are impractical for large-scale manufacturing (10–24). As outlined in Fig. 1B, three coupling paradigms have traditionally been employed to facilitate P–O bond formation, each relying on cumbersome P(III)-based reagent strategies. The first (Fig. 1B, i), which produces stereorandom PS ASOs, commences with a classic phosphoramidite loading step to append the first nucleoside, followed by deprotection, coupling to a second nucleoside, and then oxidative sulfurization of the resulting P(III) species to the P(V) state (11, 12). The second strategy (Fig. 1B, iii), capable of accessing stereodefined PS linkages, uses a phosphoramidate containing a P(III)-centered chiral auxiliary to control the phosphorous stereochemistry but is otherwise identical to the first. The most mature manifestation of this method stems from work developed by Wada and co-workers (further refined by WAVE Life Sciences) and requires a total of seven steps for the preparation of the chiral auxiliary (10, 13–15). Finally, an approach pioneered by Stec and co-workers (Fig. 1B, ii) relies on a similar strategy that begins with a P(III)-oxathiaphospholane loading reaction, followed (again) by oxidative sulfurization to furnish a chiral P(V)–oxathiaphospholane (OTP) sulfide auxiliary; this stable OTP can then be coupled to form the requisite internucleotide linkage using a simple amine activator [typically 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU)] (16–18). However, by Stec’s own admission, this strategy is scale limited, suffers poor coupling efficiency, requires resolution of stereorandom OTPs, and is extremely labor intensive (19). Taken together, the deficiencies of current methods point to an overreliance on low-valent P–O bond forming chemistry, as P(III) has historically been viewed as the sole gateway for loading (and primary access for coupling) reactions. Reported herein is a major departure from that dogma: the invention of simple, P(V)-based reagents for phosphorus-sulfur incorporation (PSI, stylized hereafter as ? ) (1) via reactions characterized by marked rapidity and scalability and complete stereocontrol at phosphorus.

Figure 1:



The caption:

(A) Inspiration, (B) historical context, (C) development, and (D) invention of the ? reagent platform. i-Pr, iso-propyl; Me, methyl; Ph, phenyl; Cy, cyclohexyl; Et, ethyl; LG, leaving group.

Typically the stereochemistry of thiophosphate groups has been controlled using intermediates in which phosphorous in a reduced state, oxidation state (III) as opposed to the (V) state in DNA.

The authors challenged themselves to better control the stereochemistry of phosphorous to offer an inexpensive route to making these compounds:

Challenging the lore of the field, we opted for an inventive approach with the mission of bypassing P(III) entirely; the reaction sequence should rely exclusively on a P(V)-based platform to forge all key P–heteroatom bonds. This would require the development of a readily accessible reagent with programmable stereochemistry, capable of achieving both loading and coupling steps with robust precision. The realization of this design required a systematic investigation of three key components: (i) an inexpensive chiral backbone to predictably impart stereocontrol, (ii) a safe source of P(V) to eliminate the previously unavoidable oxidative sulfurization step, and (iii) a reactive yet stable leaving group to enable a facile loading reaction. Ideally, a traceless approach would join the parent species in a direct and asymmetric fashion while liberating a single P(S)O unit with no evidence of the chemical method of installation remaining. These disparate attributes should harmoniously assemble in such a way as to appeal to both process and discovery chemists—traits such as scalability, stability, crystallinity, and rapid reaction times were deemed critical to success.

The authors designed a new reagent utilizing limonene, a chiral compound found in great abundance in orange peels, limonene. (I have heard that the orange peels from the manufacture of orange juice can produce hundreds of metric tons of limonene.)

The reagent, which the authors call the "? reagent" is prepared as in this graphic:

Fig. 2 Formalisms for the use of ? reagents in dinucleotide synthesis.
(A) Stereochemical assignments. (B) Loading of nucleoside monomers. (C) Coupling to produce stereopure PS dinucleotides. R, TBDPS (tert-butyldiphenylsilyl). Loading: nucleoside [1 equivalent (equiv.)], ? reagent (1.3 equiv.), DBU (1.3 equiv.), MeCN, 25°C, 30 min. Coupling: nucleoside–P(V) (1.0 equiv.), coupling partner (2.0 equiv.), DBU (3.0 equiv.), MeCN, 25°C, 30 min.

Some more talk on the topic:

Cyclic dinucleotides (CDNs) are macrocyclic natural products endowed with noteworthy biological activities and a storied history (25). Ample research indicates that these compounds play a number of prominent biological roles, most notably as secondary messengers in both the mammalian immune system and bacterial communication (26–31). Their role as agonists of innate immune response via binding and activation of the stimulator of interferon genes (STING) protein has led to a surge of interest across the pharmaceutical industry (26, 27, 32). As a consequence, numerous industrial and academic groups have targeted CDN constructs for evaluation as potential therapeutic agents. Just as the PS modification facilitates a marked improvement in ASO properties, a similar effect has been observed in CDNs, adding a layer of complexity to an already challenging class of natural products (8, 33–40). To be sure, the current impediment to rapid clinical progression of these compounds lies in the chemical synthesis. The modular strategies of traditional medicinal chemistry programs are stymied by both the poorly soluble nature of CDNs and the litany of engineered protecting-group schemes required to shield the array of heteroatoms found in the nucleobase, sugar core, and P(III)-based starting materials. As shown in Fig. 3A, nine steps are generally required from the parent nucleoside to prepare a single CDN as a mixture of all four possible diastereoisomers, with an observed ratio that appears to be substrate dependent (33–40). Of this arduous sequence, only four steps can be considered construction reactions, or those that contribute directly to P–O bond formation. The remainder account for various synthetic concessions, including protecting-group manipulations and P(III)-to-P(V) oxidations. As compared with PS linkage integration in ASOs, the challenge of incorporating chiral PS linkages into CDNs becomes far more profound, as one is at the mercy of substrate bias to obtain diastereoselectivity at phosphorus. Modular, stereocontrolled CDN synthesis has so far proven elusive (Fig. 3A). In contrast, ? reagents enable a marked divergence from the traditional P(III) approach, requiring only four or five steps from a starting nucleoside to arrive at a stereopure CDN. Two general protocols for CDN synthesis have been devised, as outlined in Fig. 3B.

Figure 3:



The caption:

Fig. 3 Applications of the ? reagent platform to the simplified synthesis of CDNs.
(A) Prior approach. (B) Stepwise and concerted macrocyclization using ?. *See (40) for the literature method. (C) Origin of observed stereochemistry in concerted protocol. PG, protecting group.

Automated nucleic acid synthesis using the reagent:



The caption:

Fig. 4 Automated synthesis of PS oligonucleotides.

(A) Crude HPLC trace of pentamer 23 (16 diastereoisomers) synthesized under standard P(III) automated conditions. (B) Crude HPLC trace of pentamer 23 (1 diastereoisomer) synthesized under unoptimized ? automated conditions. DMTr, dimethoxytrityl.

This sort of thing is an excellent tool for making drugs that are entirely unaffordable for the majority of the people who need them much, much cheaper, improving the quality of - and in some cases, saving - lives.

Have a nice evening.