Generality-Driven Catalyst Development: A Universal Catalyst for Enantioselective Nitroalkene Reduction

Abstract

Cracking the selectivity-generality paradox is among the most pressing challenges in asymmetric catalysis. This obstacle prevents the immediate and successful translation of new methods to diverse small molecules. This is particularly rate-limiting for therapeutic development where availability and structural diversity are often critical components of successful campaigns. Here we describe the union of generality-driven enantioselective catalysis and the preparation of diverse peptidomimetics. A single new organocatalyst provides high selectivity and substrate generality that is matched only by a combination of metal- and organocatalysts. Within organocatalysis, this discovery breaks a 16-year monolithic paradigm, uncovering a powerful new scaffold for enantioselective reduction with behavior that suggests recognition of a nitroethylene minimal catalaphile.

Graphical Abstract

The study of small molecule modulators of biological targets is the basis for the development of new therapeutics, but it relies on rapid access to large numbers of stereochemically complex building blocks. Building block availability is dependent on highly selective and ‘general’ reactions.¹ Designing for generality (catalyst-substrate promiscuity), however, is often antithetical to designing for selectivity (catalyst-substrate tailoring)² (Figure 1A). Progress toward cracking this paradox has been slow, but advances defining and prospecting for generality in asymmetric catalysis have been reported very recently.³ A possible solution to this is the search for catalysts that recognize a minimal catalaphile, engaging substrates with precise functional group recognition, but nothing more (Figure 1A).⁴ Here we describe the successful development of a new enantioselective nitroalkene reduction, one that delivers a rare combination of selectivity and generality that only the combination of several metal and organocatalysts can equal. Furthermore, this diversified nitroalkane feedstock is integrated with peptidomimetic synthesis to provide a uniform solution for the structural and stereochemical challenges raised by noncanonical amino acids, β-amino acids, chiral carboxylic acids, and amines (Figure 1B). This discovery signals a catalyst-substrate recognition mechanism that differs fundamentally from all other known nitroalkene-binding catalysts, and suggests recognition of a nitroethylene minimal catalaphile.

Figure 1.

A: Minimal catalaphile paradigm as an answer to the generality-selectivity paradox. B: Traditional residue-by-residue approach, compared to nitroalkane Rosetta Stone approach relying on a general, highly enantioselective nitroalkene reduction.

A broad plan to develop nitroalkanes as Rosetta Stone intermediates⁵ for sustainable peptide synthesis required access to diverse enantioenriched β-chiral nitroalkanes (Figure 1B). Peptidomimetics are a compound class where unusual α- or β-side chains^6,7,8,9,10 can present challenges to conventional chiral carboxylic acid feedstock, requiring lengthy syntheses that precede amide couplings with unresolved sustainability issues (Figure 1B). Reduction by transfer hydrogenation¹¹ has been applied to the conjugate reduction of unsaturated nitroalkene substrates to form β-chiral nitroalkanes. A database of all existing reports of enantioselective nitroalkene reduction was created, including metal-catalyzed (Cu,¹² Rh,¹³ Ir¹⁴), organocatalyzed,^{15,16,17,18,19,20,21,22,23} and enzyme-catalyzed²⁴ methods. In each of the landmark reports, enantioselective nitroalkene reductions were exposed to a new catalyst type, but success in generalizing these tools in the intervening years has been remarkably limited as evaluated by a quantitative generality assessment tool developed by Reid (Figure 2B).²⁵ None of the known catalysts have a generality above 60%.

Figure 2. Assessments of AmA-Catalyzed Enantioselective Reductions of Nitroalkenes Using Machine Learning: Scope of Organocatalysts, and Generality of All Reported Catalysts^a.

^aMethods used as developed and defined by Reid (ref. 25). A) 2D UMAP visualization exhibiting the clustering of various nitroalkenes (details in SI). B) Generality calculation for individual catalysts: The generality results for Adamant-AmA·HNTf₂ (7·HNTf₂) are extracted from Table 1. Generality is defined as the equation, where K represents the number of clusters and 'Success' is a manually defined threshold (>80%). A catalyst is considered successful if the average enantiomeric excess (ee) of its cluster exceeds this threshold value. C) Catalyst group comparison: this section contrasts the performance of different catalyst groups. D) Experimental validation: summarizes experimental outcomes comparing the efficacy of thiourea catalysts to the AmA catalyst.

Within organocatalysis, thioureas are the only catalyst paradigm to develop in over 16 years, with modest gains in substrate scope (Figure 2B, T1-T7). This, and the speculative nature of catalyst-substrate models, did not portend ready translation to substrate classes of interest. Furthermore, a generality-driven approach to catalyst development might circumvent the need to use multiple catalysts for applications in therapeutic development, where structural diversity and synthetic expediency are critical. For example, efforts to identify more tolerable and safe proteasome inhibitors based on the dipeptide boronate bortezomib highlight the unmet need for new chemistry to prepare α-chiral amides and β-amino amides.²⁶

Our initial investigation focused on the reduction of β,β-disubstituted nitroalkenes, specifically 3, prepared from the corresponding terminal alkene (Scheme 1).²⁷ A rather broad inspection of diverse Bis(AMidine) (BAM) ligands was made,²⁸ searching for efficacy in this transformation. This evaluation included both free base and monoprotic acid salt forms²⁹ of many BAM ligands and several isomers (Table S1), but mindful of a bifunctional catalysis paradigm.³⁰ A promising start was uncovered by ligand 6, albeit unexpected since this is a 3-aminopyridine isomer³¹ of the most studied and successful BAM motif (which was not enantioselective: Table S1). The triflic acid salt of 6 provided the nitroalkane product in 50% ee and 40% yield (Scheme 1). A hypothesis-driven optimization phase ensued, one that evaluated the importance of ligand coordination features, symmetry, and Brønsted acid effects on enantioselection and yield, but also prioritized generality across a range of diverse nitroalkenes. Interestingly, Amidine-Amide (AmA) ligand 7 emerged from this search with substantial improvement to both yield and enantioselection. The free base was considerably less reactive and selective, providing the nitroalkane in 74% ee and 26% yield. The analogous ligand that lacks a pyrrolidine at the quinoline 4-position³² was less reactive as well.³³ Alternative reducing agents were evaluated but without improvement.

Scheme 1.

Lead results from a broad evaluation of BAM ligands in nitroalkene reduction.^a

Further optimization resulted in standard conditions that were then applied to nitroalkenes as listed in Table 1. Beginning with the most studied cases, the β-aryl/β-alkyl substituted nitroalkenes were found to give nitroalkane products in good yield and enantioselection (Table 1, 9a-9m). The absolute configuration of the product correlated with the geometry of the nitroalkene. This behavior suggests that the alkene geometric isomers bind to the catalyst in a conserved manner, a phenomenon observed by others as well. This was exploited when using similar aryl substituents at the β-position (Table 1, entries 9n-9r). In these examples, the enantioselection was high, regardless of β-aryl positioning relative to the nitro substituent. These diaryl-substituted nitroalkenes have not been reported previously, and although they exhibited lower apparent reactivity, extended time and an additional equivalent of Hantzsch ester provided the products in moderate to good yield and enantiomeric excess. The pairing of two β-alkyl substituents was also studied, producing findings analogous to those above (Table 1, entries 9s-9u). Heteroaryl-substituted nitroalkenes were reduced with similar yields and selectivities (Table 1, entries 9v-9x). The success realized in these reductions provides a single solution to the preparation (by subsequent nitro reduction) of enantioenriched β-chiral terminal amines.

Table 1.

Catalyzed enantioselective reductions of nitroalkenes.^a

^aGeneral experimental conditions: Nitroalkene (1 equiv.), Hantzsch ester (1.6 equiv.) and (R,R)-catalyst 7·HNTf₂ (10 mol%) were stirred under Ar at 0 °C for 48 h. Yields are isolated yields. Enantiomeric excess was determined by HPLC using a chiral stationary phase. Nitroalkene geometry: E for 9a-o, 9s-9x, 9ii-9kk and Z for 9p-9r, 9y-9hh.

^bReduced by (S,S)-catalyst.

^cReaction time = 72 h, temperature = 23 °C.

^dReaction time = 96 h, Hantzsch ester (2.6 equiv.).

^e5 mol % catalyst, 60 °C.

^f5 mol % catalyst, temperature = 40 °C.

Application of AmA catalyst 7 to vinylogous nitroamines was also prioritized as an avenue to α- and β-amino acids (Table 1, entries 9y-9kk). The dehydro β-aminoester substrates are easily prepared from either nitromethane addition to a nitrile,^34,35 followed by N-protection, or from unsaturated cinnamic esters using a nitrene equivalent.³⁶A higher temperature was generally required for these reductions for reasonable reaction times, but good yields and high enantioselection were typical, with few exceptions. The nitroalkenes endowed with a β-C-N bond (Boc, Fmoc, or Ac protecting groups) provided the desired nitroalkanes in moderate yield and high ee (Table 1, 9y, 9gg-9hh), and revealed that the N-protecting group had little effect on the enantioselectivity of the reduction. An initial examination of tetrasubstituted cases provided promising results with bromo and ester α-substituents (Table 1, entries 9ii-9kk). These tetrasubstituted nitroalkenes are challenging for thiourea catalyst-mediated reduction.²³

The expansion in substrate scope was analyzed by comparison to the state of the art using machine learning.²⁵ A data set was assembled from all reported enantioselective nitroalkene reductions and a virtual substrate scope built with nitroalkenes containing the most common functional groups in drug substances. The details of the data set are described in SI. The UMAP plot in Figure 2 is a two-dimensional depiction of multidimensional substrate structural data that provides a convenient depiction of catalyst scope. While thiourea catalysts can be tuned for high selectivity in localized areas of the map, 7·HNTf₂ is broadly effective across the diverse substrates yet explored. Generality values were also generated for individual catalysts and major classes of catalysts, alongside tabulation of average ee. From a reagent perspective, Figure 2B and C illustrate that no other single catalyst or catalyst class is sufficient to achieve the selectivity and generality of AmA 7·HNTf₂. The generality calculation was verified by direct comparisons using several challenging substrates with thiourea catalysts and 7·HNTf₂ (Figure 2D). In all cases, the AmA catalyst demonstrates significantly higher ee.

The broad effectiveness of catalyst 7 across a diverse slate of substrates creates a pool of nitroalkane products available from a single catalyst facilitator. We were drawn to a class of proteasome inhibitors, discovered by Zhu, Lei and coworkers, that exhibited similar potency to bortezomib but broader efficacy against a panel of tumor cells (10, Scheme 2).³⁷ Their work called attention to two unmet needs. First, there is lack of expedient routes to the β-amino acid (12) needed for preparation of the central amide, as evidenced by the preparation and evaluation of 10 as a diastereomeric mixture using racemic β-amino acids (from rac-12). Second, the tetrahydronaphthalene amide lacks a diverse pool of derivatives that could fuel a lead optimization campaign. For example, analog 11 would normally be sourced from the fluorine-containing acid 13. Notably, the benzylic nature of these carboxylic acids can lead to racemization during amide bond formation.

Scheme 2.

Synthesis of proteasome inhibitors using enantioselective reductions of nitroalkenes to prepare each stereoisomer.

We first addressed the need for the individual diastereomers of 10 through a nonconventional approach (Scheme 2). From nitroalkene 14, α-bromonitroalkane 15 could be enantioselectively prepared in 2 steps using (S,S)-AmA catalyst. Separately, nitroalkene 16 was used to prepare β-amino acid (R)-12. Umpolung Amide Synthesis (UmAS)³⁸ with (R)-12 and 15 under aerobic conditions³⁹ delivered amide (S,R)-17. Immediate coupling to the protected α-amino boronate ester and subsequent hydrolysis, furnished peptide (S,R,R)-10 as a single diastereomer. The complementary diastereomer was prepared using the (R,R)-catalyst to form 16 and, in a procedure parallel in form and outcomes, delivered (S,S,R)-10.⁴⁰ Along this pathway, the intermediate boronate ester was crystallized to confirm both relative and absolute stereochemistry. Finally, fluoropeptide 11 was prepared from 9j using an identical route, illustrating the interrelationship of general catalysis and synthesis.

In summary, a new catalyst has been discovered that single-handedly delivers nitroalkanes from β,β-disubstituted nitroalkenes with a generality otherwise achievable only by a combination of several organo- and organometallic catalysts. A machine learning protocol was used to analyze generality for the state of the art before and after this development, characterizing the increase in scope while maintaining efficacy for existing product classes. This catalyst (7·HNTf₂) improves yield and/or enantioselection of each major class of nitroalkene when compared directly to the leading alternative, ultimately enabling the concise synthesis of a class of established proteasome inhibitors. The impact of a general catalyst was demonstrated by the synthesis of α- and β-chiral amides. Developments such as these positively engage the generality-selectivity paradox to reduce the tension created within therapeutic development when unusual amide residues are required and the parent carboxylic acid is limited by availability or cost.