Building C(sp3)-rich complexity by combining cycloaddition and C-C cross-coupling reactions.

Prized for their ability to rapidly generate chemical complexity by building new ring systems and stereocentres1, cycloaddition reactions have featured in numerous total syntheses2 and are a key component in the education of chemistry students3. Similarly, carbon-carbon (C-C) cross-coupling methods are integral to synthesis because of their programmability, modularity and reliability4. Within the area of drug discovery, an overreliance on cross-coupling has led to a disproportionate representation of flat architectures that are rich in carbon atoms with orbitals hybridized in an sp2 manner5. Despite the ability of cycloadditions to introduce multiple carbon sp3 centres in a single step, they are less used6. This is probably because of their lack of modularity, stemming from the idiosyncratic steric and electronic rules for each specific type of cycloaddition. Here we demonstrate a strategy for combining the optimal features of these two chemical transformations into one simple sequence, to enable the modular, enantioselective, scalable and programmable preparation of useful building blocks, natural products and lead scaffolds for drug discovery.

Retrosynthetic analysis is built upon the strategic identification of the reactions (transforms) that offer the greatest potential to simplify target preparation (“T-goal”)[3]. To this end, the capacity for generating complex ring systems and multiple bonds with precise stereochemical control via cycloaddition is unmatched. In contrast, C–C cross couplings such as Heck, Suzuki, and Negishi reactions, capable of making only one bond at a time (most often between flat sp2-systems)[4], are the most used C–C bond forming methods in the patent literature[7]. To understand this phenomenon, it is worthwhile to compare the features of these two diverse reaction classes (). Cycloadditions form two new sigma bonds generally through concerted pathways that follow precise rules for predicting stereo- and regiochemistry. This rapidly accesses complexity starting with specialized building blocks, designed to enable the reaction to take place cleanly[8]. On the other hand, C–C cross couplings form one new sigma bond between easily accessible building blocks using a transition metal catalyst to reliably and controllably produce new connections. What C–C cross coupling lacks in terms of complexity generation it makes up in the form of sheer reliability and modularity. These distinct features are illustrated by the syntheses of the alkaloid epibatidine (1, )[9] and the commercial antihypertensive medicine Cozaar (2)[10]. 1, coveted for its pronounced analgesic properties, has been prepared by total synthesis over sixty times (see SI section E for a complete listing). Of these, at least 31 have utilized cycloaddition chemistry as their key ring-constructing step. The vast majority of these approaches involve formation of the bridged pyrrolidine core followed by stepwise, and often lengthy, pyridine incorporation. Medicinal agent 2, bereft of any stereogenic centers or topological complexity, was heralded as one of the first examples of a “blockbuster” drug. Both its discovery and eventual manufacture utilized C–C cross coupling (Ullmann and Suzuki) for the key bond cleavage[11,12]. This facilitated both a convergent assembly and modularity that permitted the rapid exploration of hundreds of analogs. We therefore sought to combine the innate complexity-generation of cycloaddition with the simplicity and modularity of C–C coupling. When applied to structures such as 1, this strategy would permit the rapid generation of analogs and when applied to medicinal programs such as 2, it would allow for rapid exploration of otherwise challenging complex chemical space.

Building block 3, of hypothetical value in medicinal chemistry, represents the manifestation of this idea (). Although its structure beckons for the use of a Diels–Alder (DA) reaction, the corresponding building blocks 4 and 5 are not electronically matched and therefore one would expect no reaction to take place. Even if a workable enantioselective DA could be invented to achieve this transform, the strategy would suffer from a lack of modularity. In order to solve this problem, one could envisage a vicinal dihaloethylene 6 in place of 5, and a fumarate-type dienophile such as 7 could serve as a viable synthetic equivalent. The favorable matched electronics of the dienophile should allow for a facile DA and subsequent radical cross coupling (RCC). To address the enantioselectivity challenge, a combination of transforms was proposed (). Maleic anhydride was chosen as a surrogate for the hypothetical chiral (pseudo)dihalide 6 given its inexpensive nature, ready participation in most cycloaddition modes ([2+1], [2+2], [3+2], [4+2]), and known desymmetrization through chiral Lewis-base mediated hydrolysis. Such a sequence for generating complexity in a modular fashion would involve five simple steps: 1. Cycloaddition to build a scaffold, 2. Desymmetrization to set absolute stereochemistry, 3. RCC to install a new C–C bond, 4. Hydrolysis, and 5. RCC to forge another new C–C bond. The known versatility of the RCC enables a range of functionalities to be installed, from aryl[13] and heteroaryl[14] systems to alkenes[15], alkynes[16] and alkyl groups[17]. We describe the preparation of >80 synthetic examples and multiple applications covering a range of natural products (including the synthesis of 1) and real-world examples from industrial settings.

Studies commenced with the DA cycloaddition () where a large variety (6) of enantiopure scaffolds could be produced in a simple modular fashion. First, scaffolds A-A (A is a DA adduct and all others are derived from DA/hydrogenation) were desymmetrized using Deng’s conditions with either quinine or quinidine as Lewis-base to deliver mixed acid/esters in excellent enantioselectivity[18]. Next, the mono-acid substrate was subjected to successive Negishi[13,15-17] and Suzuki[14] type RCC reactions based on starting material availability or individual preferences. In this manner, some of the most simple and inexpensive DA products known (A is $38.54/mol and A is $9.52/mol) were transformed into enantiopure products of high value. Indeed, none of these entities can currently be prepared using a DA reaction (racemic or enantiopure) as in most cases the analogous DA would require both an electron-rich diene and electron-rich dienophile which is electronically unfavored. Furthermore, controlling the chemoselectivity when there are multiple alkenes present in the dienophile is challenging using tradition DA chemistry; however, with our method diverse alkenes and alkynes can be easily installed post-cyclization with excellent isomeric and geometrical purity (12, 15, 19, 21, 25, 31). Typically scaffolds derived from DA-adducts A-A have a clear retrosynthetic signature wherein the electron withdrawing groups on the dienophile have been homologated, alkylated, or degraded. Therefore, our approach allows access to previously unexplored chemical space of electron-rich, chiral, 1,2-disubstituted DA scaffolds. Using the tactical combination outlined above a virtually limitless array of substituents are now easily accessed including halogenated arenes, Lewis-basic heterocycles, α,β-unsaturated carbonyl groups, cyclopropanes, cyclobutanes, sulfur moieties, and alkenes. In the case of exo and endo isomers A and A (both commercially available), the products of the sequence were convergent to an identical product using the same coupling partners (27 could be prepared from A and A, respectively).

As shown in , the strategy outlined above could also be applied to [3+2], [2+2], and [2+1] cycloadditions. The vast scope observed with DA chemistry was also seen in these cases, accessing substituents such as substituted olefins, terminal alkynes, homoallyl groups, and heterocycles. Pyrrolidine-containing systems could be derived from simple building blocks B and B (accessed through dipolar cycloaddition)[19] to furnish 64-82 in high ee. Quick access to pyrrolidine-containing drug scaffolds is useful as historically these have been extremely relevant to medicinal chemists with ca. 40 pharmaceuticals containing this motif[20]. As with the DA adducts, none of these structures have been accessed before. Therefore, our approach serves as a modular entry to chiral variants of these coveted scaffolds. In a similar vein, enantiopure cyclopentenes derived from Pd-catalyzed TMM-based formal [3+2] cycloaddition adduct B[21] were easily produced. Such structures are highly challenging to access in any other way. Scaffolds C and C, representing entry to [2+2]-derived systems, could be similarly processed. Modular access to enantiopure cyclobutanes is significant and useful given the strict electronic requirements for photochemical cycloaddition and the documented challenge in achieving general asymmetric induction[22]. Access to 1,2-disubstituted cyclobutanes (83-86, derived from C) compares favorably with photochemical approaches to such systems that frequently give inseparable racemic mixtures of regio- and diastereomers. Furthermore, access to tetrasubstituted, chiral cyclobutanes is highly useful as multiple families of dimeric and pseudodimeric cyclobutane natural products contain such structural motifs[23]. Structures such as 87, which would be otherwise extremely difficult to access through conventional photochemistry, can be easily prepared in enantiopure manner from maleic anhydride heterodimerization adducts such as C. Finally, the strategy outlined above when applied to [2+1] cycloadditions using scaffolds D and D is a significant departure from common retrosynthetic logic normally applied to such structures. Conventional approaches, usually involving late-stage cyclopropanation of an olefin, suffer from lack of enantiocontrol in the absence of directing groups or specialized carbenoid donors and complex catalysts[24]. Structures 94 and 95 (D-derived) are particularly illustrative of this fact as it would be extremely challenging to access either of these in an enantioselective fashion with current synthetic technology (cyclopropanation or C–C cross coupling). As a testament to the power of this strategy to access diverse libraries, an additional 48 enantioriched compounds were synthesized in a similar manner (see Extended Data Figures for details).

To further demonstrate the potential of this approach to simplify synthesis, six applications are presented in natural product total synthesis and both early- and late-stage medicinal chemistry programs (). As mentioned above, alkaloid 1 has been a popular target by both academic and industrial scientists. By application of cycloaddition and cross coupling the native carboxylate required for DA can be employed directly to produce epibatidine 1 in five steps (for optimization and in depth analysis of previous approaches, see SI section E) in 38% gram–scale overall yield (). It is worth noting that the key decarboxylative cross coupling takes place in 95% isolated yield (gram–scale, 72% yield). Saphris™ (asenapine, 102, ), an FDA–approved antipsychotic, is currently marketed as a racemic mixture (although the (+)-isomer has superior pharmacokinetic properties)[25]. This near-symmetric molecule has been challenging to prepare enantioselectively as the two aromatic systems differ only in one chlorine substituent. It is therefore hard to envisage a cycloaddition that could be rendered enantioselective for the preparation of 102, and only one enantioselective approach has been reported (16 steps)[26]. In contrast, 102 could be prepared in formally six steps from B with complete enantiocontrol by a strategy that could also be used to make an array of near-symmetric analogs. Epothilone, a famous microtubule inhibitor natural product that inspired the FDA–approved medicine Ixabepilone[27], has been the subject of numerous synthetic studies and analog campaigns[28]. Intermediate 105 () was utilized during Nicolaou’s synthesis of a cyclobutyl-containing analog and required 15 steps (24% overall yield) to be prepared in enantiopure form starting from C via enzymatic desymmetrization and a series of carefully choreographed homologations[28]. Using the same starting material C could be prepared in only eight steps through sequential desymmetrization, RCC with alkyne 103, hydrolysis, RCC with alkyl zinc 104, protecting group exchange, and finally hydroboration/oxidation. Differentially substituted norbornene rings have been shown to be useful phenyl bioisosteres in medicinal chemistry however their broad implementation is hindered by a lack of methods for their rapid modular construction (unlike aryl systems)[29]. In collaboration with LEO Pharma, a key target for an ongoing program (109, ) was prepared from DA adduct 106 via alkene hydration and RCC with the pyrazole-boronic ester 108 in 44% yield. It is worth noting that this is the first report of a RCC reaction using a BPin derivative rather than a boronic acid. This advance was achieved using an in situ prepared ate complex prepared by adding one equivalent of n-BuLi relative to the aryl-BPin donor[30]. This modification of our Suzuki-RCC protocol was also employed to prepare cyclopropanes 91, 92 and 93. Finally, an ongoing program at Eisai necessitated the preparation of complex scaffold 114 () wherein the key SAR to be explored resided at the aryl (green) portion of the molecule. This is a particularly powerful application of the strategy outlined herein since a carboxylate, needed to achieve the diastereoselective DA to construct the decalin framework, served as a gateway for the exploration of chemical space at the desired position. Thus, an asymmetric DA reaction using Corey’s oxazaborolidine[31] catalyst followed by functional group manipulations (See SI section I) led to intermediate 112 which could be cross coupled with boronic acid 113 in a particularly challenging context to deliver 114 and enable biological follow-up.

The strategy outlined here could be useful for more than just forging C–C linkages through cross-coupling. The incorporation of classic nucleophilic (ketone synthesis) and decarboxylative functionalizations (Hunsdiecker reaction, Curtius and Wolff rearrangements) of intermediate adducts opens innumerable possibilities for diversification rendering access to previously inaccessible building blocks (). As an example, a chiral 2π enamine synthon was employed for the preparation of the EED protein–protein inhibitor 118 (Ki=4 nM) that was previously prepared by AbbVie[32] through a non-modular, racemic [3+2] approach in 1.9% overall yield. In contrast, optically pure 118 (97% ee) was prepared from meso scaffold B2 utilizing RCC followed by Curtius rearrangement in 13% overall yield.

The advance described herein is mainly strategic in nature and thus the underlying limitations are mainly tied to individual parameters of a particular cycloaddition and ensuing RCC reactions. That said, cis-1,2-disubstituted products are not currently accessible unless downstream isomerization reactions are pursued which need to be evaluated on a case-by-case basis or decarboxylative non-isomerizing processes are employed (e.g. Curtius). While ligand-controlled RCC reactions might eventually address this issue, in its current form this platform for modular molecular assembly holds great promise for accessing new areas of chemical space. The combination of classic cycloaddition chemistry with newly emerging radical C–C coupling offers a powerful way to repurpose the most classic skeleton-building reactions of organic synthesis to simplify the enantioselective preparation of building blocks, natural products, and medicines.

Complete substrate scope of [4+2] cycloaddtion/cross-couplings. See Supplementary Information for synthetic details. R1,R2 = (Het)Aryl, alkyl, alkenyl, alknnyl. X-ray structure data is available for compounds 11, 19, 23, 25, 28, 44, 45, 49,

Complete substrate scope of A [3+2], B [2+2] and C [2+1] sections. See Supplementary Information for synthetic details. R1,R2 = (Het)Aryl, alkyl, alkenyl, alknnyl. X-ray structure data is available for compounds C2 and 65.