Stereoselective synthesis of unnatural α-amino acid derivatives through photoredox catalysis

A protocol for stereoselective C-radical addition to a chiral glyoxylate-derived N-sulfinyl imine was developed through visible light-promoted photoredox catalysis, providing a convenient method for the synthesis of unnatural α-amino acids. The developed protocol allows the use of ubiquitous carboxylic acids as radical precursors without prior derivatization. The protocol utilizes near-stoichiometric amounts of the imine and the acid radical precursor in combination with a catalytic amount of an organic acridinium-based photocatalyst. Alternative mechanisms for the developed transformation are discussed and corroborated by experimental and computational studies.


Introduction
Unnatural a-amino acids constitute an important class of biologically relevant compounds that are widely used in both pharmaceutical industry and for fundamental research within molecular and structural biology. 1 A number of pharmaceuticals based on unnatural a-amino acids are currently available, including angiotensin-converting enzyme (ACE) inhibitors for the treatment of cardiovascular and renal diseases, 2 antiviral medicines, 3 and others. 4 Recently, peptidomimetic a-ketoamide inhibitors based on unnatural a-amino acids have received increased attention as drug candidates for treatment of COVID-19 disease caused by the SARS-CoV-2 coronavirus, 5 highlighting the high demand for such building blocks.
A variety of synthetic strategies to access unnatural amino acid derivatives have been developed over the years, with some notable methods being the catalytic asymmetric Strecker-type reactions, asymmetric hydrogenation of dehydroamino acids, and electrophilic and nucleophilic alkylation of glycine derivatives (Fig. 1A). 6 Among these, functionalization or reduction of a-imino esters offers a straightforward route to various enantiomerically enriched a-amino acids. 7 Traditionally, these strategies have employed polar retrosynthetic disconnections, which oen require the use of (super)stoichiometric amounts of toxic and highly sensitive reagents at low temperatures, thereby limiting the substrate scope and practicality for scale up of these reactions. These limitations have recently been challenged by re-introduction of free-radical reaction manifolds, aided by the developments in base-metal catalysis, 8 electrosynthesis 9 and photoredox catalysis, 10 leading to a vast array of strategies for light-induced modication and synthesis of amino acids and peptides. 11 Among these, radical addition to imines through photoredox catalysis was demonstrated in symmetric 12 and asymmetric 13 fashion (Fig. 1B). In 2017, Alemán and co-workers reported a protocol for asymmetric radical addition to imines mediated by visible light. 13a The developed catalytic system made use of a chiral sulfoxide auxiliary group, commonly employed in the synthesis of chiral amines. 14 Here, the C-centered radical was generated through visible light-mediated reductive cleavage of the N-O bond in a redox-active phthalimide ester, followed by radical addition to the N-sulnyl imine. The reductive nature of this protocol required the use of a stoichiometric amount of a reducing agent (Hanztsch ester). More recently, a related Ni-based catalytic system was described by Baran and co-workers. 15 This protocol employed a tetrachloro-substituted redox-active ester as the radical precursor, with Zn as a stoichiometric reducing agent and a Ni-based catalyst for mediating the C-C bond formation. Although this protocol displayed an impressive substrate scope, it is associated with moderate atom-economy, limiting its applicability for large-scale synthesis.

Results and discussion
Inspired by the catalytic systems developed by the Alemán 13a and Baran 15 groups, we sought to realize a protocol for diastereoselective decarboxylative radical addition to chiral N-sul-nyl imines that would utilize ubiquitous non-activated carboxylic acids as radical precursors. 16 A related direct decarboxylative addition process was attempted by the Alemán group for a benzaldehyde-derived N-sulnyl imine under reaction conditions reported by MacMillan; 17 however, no formation of the desired product was observed (see the ESI to ref. 13a). Similarly, we observed no desired product with pivalic acid 2a as the radical precursor and N-sulnyl imine 1 as the radical acceptor when the reaction was conducted in DMSO with [Ir(dF(CF 3 )ppy) 2 (dtbbpy)](PF 6 ) as photocatalyst (Table 1, entry 1), presumably due to fast decomposition of N-sulnyl imine 1. Gratifyingly, changing the solvent to a,a,a-triuorotoluene (PhCF 3 ) furnished the desired product 3a in fairly good yield of 65%, although with poor diastereoselectivity (Table 1, entry 2). Using other bases in place of Cs 2 CO 3 completely prohibited the reaction (for details on the optimization studies, see the ESI †), and the highly-oxidizing photocatalyst 4CzIPN 18 failed to deliver the radical addition product (Table 1, entry 3). Fortunately, the highly-oxidizing organic acridinium-based photocatalyst [Mes-Acr-Me](BF 4 ) delivered product 3a with excellent diastereoselectivity, although in poor yield (Table 1, entry 4). Increasing the catalyst loading from 1 to 5 mol% and switching to the more stable N-phenyl-substituted photocatalysts [Mes-Acr-Ph](BF 4 ) and [Mes-Me 2 Acr-Ph](BF 4 ) 19 dramatically increased the yield of the stereoselective radical addition product up to 78% (Table 1, entries 5-7). Conducting the reaction in more conventional solvents, such as MeCN, CH 2 Cl 2 , and 2,2,2-tri-uoroethanol (TFE) in place of PhCF 3 resulted in diminished yields (Table S1, see the ESI †), highlighting the documented inertness of PhCF 3 towards free-radical intermediates. 20 Changing the base to K 2 CO 3 and increasing the base loading further improved the yield up to 85% (Table 1, entry 11). Finally, utilizing a slight excess of the acid radical precursor 2a delivered the desired product 3a in excellent yields (91% and 95% for 1.2 and 1.5 equiv. of 2a, respectively; Table 1, entries 13 and 14). Consistently with the previous reports on radical additions to N-sulnyl imines, the tert-butyl-and para-tolyl-substituted N-sul-nyl imines 4 and 5 proved to be inefficient as radical acceptors (Table 1, entries 15 and 16). 13a, 15 In case of tert-butyl-substituted N-sulnyl imine 4, it is likely that the transiently formed aminyl radical intermediate underwent decomposition to form an iminosulfanone (-N]S]O), thereby disrupting the catalytic cycle. 21 The substrate scope of the developed transformation was evaluated with a variety of non-functionalized and functionalized tertiary, secondary, and primary carboxylic acids (Fig. 2). For all of the amino acid derivatives excellent diastereoselectivity at the a-position was observed (>95 : 5 dr). The radical precursors producing tertiary and secondary alkylsubstituted radicals provided the expected products in generally good to high yields (3a-e and 3k-n). The highly reactive primary alkyl radicals displayed lower selectivity for the addition reaction (3o and 3p), consistent with previous reports featuring unstable free-radical intermediates under related conditions. 22 Further optimization of the reaction conditions for the primary acids 2o and 2p did not result in improved yields (Tables S2 and S3 †), illustrating the intrinsic instability of the respective radical intermediates and/or the photocatalyst under the employed conditions. Benzylic-type radicals were generally inefficient (see Fig. 2, unsuccessful substrates); however, a cyclopropyl-substituted benzylic radical and an indole-derived benzylic-type radical provided the expected products 3f and 3q, respectively, in satisfactory yields.
The carboxylic acid radical precursors that furnish stabilized a-heteroatom C-radicals generally provided the addition products in good to excellent yields. Gratifyingly, N-Boc-protected aamino acid radical precursors based on pipecolic acid, proline, valine, and phenylalanine furnished the expected amino acid derivatives 3r-u in generally excellent yields, exemplifying a prominent synthetic route to biologically active a,b-diamino acids. 23 The a-O-substituted radicals derived from dialkyl (3v, 3w) and alkyl aryl ethers (3g-i, 3x) provided the expected products in moderate and excellent yields, respectively. To our delight, a primary a-S-substituted radical containing an aryl bromide functionality afforded the expected product 3y in satisfactory yield despite combining several structural features that can be deleterious under free-radical conditions. The sterically-demanding carbohydrate-based radical derived from diprogulic acid 2j delivered the monosaccharide-amino acid conjugate product 3j in satisfactory yield and excellent diastereoselectivity at both the aand b-stereocenters (>95 : 5 a dr, >95 : 5 b dr).
The N-sulnyl amide functionality in product 3a could be conveniently removed under mild acidic conditions in quantitative yield (Fig. 2E). The absolute conguration at the a-stereocenter in the obtained a-amino ester was then determined as (R), in complete agreement with the previous observations and the results of computational studies (for details, see the ESI †). Similarly, removal of the N-sulnyl amide functionality was carried out for more complex products 3q, 3x and 3z, and the corresponding a-amino esters 6q, 6x and 6z were isolated in excellent yields (>95%).
Based on literature precedents, a mechanism for the developed transformation was proposed and corroborated by uorescence quenching and computational studies (Fig. 3). 13a, 24 Initially, the acridinium photocatalyst [Mes-Me 2 Acr-Ph] + (Acr + ) [Mes-Me 2 Acr-Ph](BF 4 ), 5 mol% K 2 CO 3 , 0.5 equiv. 60 min 95% >95 : 5 15 As entry 13, but with t Bu-sulnyl imine 4 60 min --16 As entry 13, but with p-Tol-sulnyl imine 5 60 min is excited by visible light (l max z 425 nm) to a highly oxidizing excited state Acr + * (E(Acr + */Acrc) z 2.09 V vs. SCE). 25 In this state, the photocatalyst can abstract an electron from the deprotonated carboxylic acid via a single-electron transfer (SET) event to generate a carboxylate radical while being reduced to the acridinium radical Acrc. The steady-state and time-resolved uorescence quenching measurements for tetrabutylammonium pivalate as the model radical precursor demonstrated efficient quenching of the excited acridinium photocatalyst with Stern-Volmer quenching constant K SV ¼ 237.5 M À1 and bimolecular quenching constant k q ¼ 6.8 Â 10 9 M À1 s À1 , while no quenching was observed for the free pivalic acid (Fig. 3B, S3 and S4 †). The carboxyl radical formed via SET then extrudes CO 2 to yield a C-centered radical, which undergoes addition to the N-sulnyl imine 1 in the key step of the reaction, forming an aalkylated N-centered radical. Finally, the N-centered radical is reduced by acridinium radical Acrc, closing the photocatalytic cycle and furnishing the desired product 3 upon protonation (Fig. 3A).
In order to gain better understanding of the stereodetermining C-C bond forming step in the proposed mechanism, DFT calculations were performed on the M062X-D3/6-311+G(d,p) level of theory (for details, see the ESI †). First, the structure of the N-sulnyl imine radical acceptor 1 was evaluated. Previously, Alemán and co-workers tentatively suggested an s-cis conformation around the N-S bond as being more stable in such compounds due to the hydrogen bonding between the imine proton and the sulfoxide oxygen. 13a Such a conformational preference would then lead to the a-(R) product when the S(R)-sulnyl imine is employed as the radical acceptor. This stereochemical outcome was indeed observed for both Alemán's and our catalytic system. The calculations conrmed that the s-cis conformer is more stable compared to the s-trans-1 conformer by 3.8 kcal mol À1 , corresponding to >99.8 : 0.2 ratio between the conformers from the Boltzmann distribution at room temperature (Fig. 3C). In the s-cis conformer, the hydrogen bonding between the imine hydrogen and the sulfoxide oxygen could be observed from the noncovalent interaction (NCI) plots, while no hydrogen bonding was present in the s-trans-1 conformer (for a detailed discussion, see the ESI †).
Subsequently, the radical addition step was evaluated for the tert-butyl radical donor and the N-sulnyl imine radical acceptor 1. The computed Gibbs free energy and enthalpy diagrams for the reaction are presented in Fig. 3D. The formation of the (R,R)-diastereomer of 3a was found to be favored both kinetically and thermodynamically and the computed activation barrier was found to be 3.8 kcal mol À1 smaller for the re-addition compared to the si-addition, while the (R,R)-diastereomer product is 2.5 kcal mol À1 more stable compared to the (R,S)-diastereomer. Interestingly, the difference in the computed activation barriers for the re-and si-addition reactions originated almost exclusively from the enthalpic terms (DDG ‡ ¼ 3.8 kcal mol À1 , DDH ‡ ¼ 3.4 kcal mol À1 ). The better stabilization of the re-TS is in part due to the stronger hydrogen bonding between the imine hydrogen and the sulfoxide oxygen for this transition state, as evident from the calculated bond distances and the NCI plots ( Fig. 3D and S7 †). Additionally, signicant steric crowding occurs in the si-TS, where the incoming tert-butyl radical requires the mesityl group to become almost completely coplanar to the sulfoxide S]O bond. In contrast, the mesityl group and the S]O bond in the re-TS are out of plane by 50 while the incoming tert-butyl radical experiences no steric crowding.
An alternative mechanism for a related radical addition to imine derivatives was proposed by Ooi and co-workers. 26 In this mechanism, the key C-C bond-forming step was found to proceed through radical-radical coupling between a C-centered radical and an a-amino radical formed by one-electron oneproton reduction of an imine substrate. However, under our conditions such a mechanistic pathway seems unlikely due to weak reducing ability of the one-electron reduced form of the employed acridinium photocatalyst (E 1/2 (Acr + /Acrc) z À0.32 V vs. SCE, Fig. 3B). As opposed to the conditions reported by Ooi and co-workers, where strongly reducing [Ir(ppy) 2 (bpy)] + -type photocatalysts (E(Ir III /Ir II ) z À1.5 V vs. SCE) were used, electron transfer from Acrc to N-sulnyl imine 1 (E p/2 z À1.24 V vs. SCE, Fig. 3B) should not be favored. However, a contribution from the radical-radical coupling pathway would explain the low diastereoselectivity (4 : 1 dr) during formation of product 3a when the reaction was conducted with the [Ir(dF(CF 3 )ppy) 2 (dtbbpy)](PF 6 ) photocatalyst (Table 1, entry 2). Indeed, this photocatalyst displayed relatively low reduction potential (E 1/ 2 (Ir III /Ir II ) ¼ À1.10 V vs. SCE, Fig. 3B), sufficient to reduce the N-sulnyl imine substrate 1 to the corresponding a-amino radical. The conformation analysis of this radical then revealed nearly free rotation around the N-S bond with a barrier of ca. 2.5 kcal mol À1 , while N-sulnyl imine 1 displayed a signicantly higher rotation barrier of ca. 8.0 kcal mol À1 and only one dominant conformer. Addition of the tert-butyl radical to the aamino radical would therefore be expected to proceed with low, if any, diastereoselectivity. The low diastereoselectivity could also be explained by product epimerization during the reaction; however, no epimerization was observed when product 3a was subjected to the comparable reaction conditions with the Irbased photocatalyst.

Conclusions
In conclusion, a practical protocol for stereoselective synthesis of various a-amino acids has been developed, employing ubiquitous carboxylic acids as radical precursors and an organic photocatalyst under visible light irradiation. This protocol allows for synthesis of highly functionalized a-amino acids, which are challenging to prepare through traditional twoelectron reaction manifolds. The protocol utilizes nearstoichiometric amounts of reagents and does not produce large quantities of waste, which is an intrinsic disadvantage of the previously described systems utilizing redox-active esters as radical precursors.

Author contributions
A. S. performed optimization studies, major part of substrate scope investigation, electrochemical and spectroscopic studies, and wrote the manuscript. A. A. performed the computational studies and part of the substrate scope investigation. E. V. S., J.-Q. L., and B. B. performed part of the substrate scope investigation. A. Z. T. performed part of the analytic measurements. B. P. K. and J. M. G. assisted during data acquisition and analysis of the spectroscopical studies. M. D. K. conceived and directed the project. P. D. and M. D. K. supervised the project. All authors discussed the results and approved the nal version of the manuscript.

Conflicts of interest
There are no conicts to declare.