Carbene and photocatalyst-catalyzed decarboxylative radical coupling of carboxylic acids and acyl imidazoles to form ketones

The carbene and photocatalyst co-catalyzed radical coupling of acyl electrophile and a radical precursor is emerging as attractive method for ketone synthesis. However, previous reports mainly limited to prefunctionalized radical precursors and two-component coupling. Herein, an N-heterocyclic carbene and photocatalyst catalyzed decarboxylative radical coupling of carboxylic acids and acyl imidazoles is disclosed, in which the carboxylic acids are directly used as radical precursors. The acyl imidazoles could also be generated in situ by reaction of a carboxylic acid with CDI thus furnishing a formally decarboxylative coupling of two carboxylic acids. In addition, the reaction is successfully extended to three-component coupling by using alkene as a third coupling partner via a radical relay process. The mild conditions, operational simplicity, and use of carboxylic acids as the reacting partners make our method a powerful strategy for construction of complex ketones from readily available starting materials, and late-stage modification of natural products and medicines.

Here we disclose an operationally simple strategy for coupling of acyl imidazoles and carboxylic acids to form ketones (Fig. 1e right), in which the carboxylic acids were directly used as radical precursors. The acyl imidazoles could also be generated in situ via the reaction of a carboxylic acid with CDI (carbonyldiimidazole), thus provides a formally decarboxylative coupling of two carboxylic acids. Importantly, our NHC and photocatalyst cocatalyzed coupling reaction could be extended to threecomponent coupling reactions with alkenes as the third radical coupling partner via a radical relay process (Fig. 1e, left), providing a straightforward strategy to construct ketones with high levels of complexity from readily available materials. To demonstrate the utility of this method in synthetic chemistry, late-stage modification of marketed drugs and direct coupling of fragments of two medicinal molecules were performed.

Results
Optimization of reaction conditions. To simplify our condition optimizations, we first evaluated the coupling reactions between activated acyl electrophiles (1a, 1b, 1c) and carboxylic acid (2a) (  75 , and a base (Cs 2 CO 3 ) in CH 3 CN as the solvent. The reaction was carried out under visible light (blue light LED, λ max = 427 nm) at 30-40°C. With 1a as precursor of the NHC-bound acyl azolium intermediate, the corresponding ketone product (3a) could be detected but with very low yield (entry 1). Switching 1a to acyl chloride (1b) under otherwise identical conditions led to slightly improved while still very low reaction yield (<5% yield, entry 2). When acyl imidazole (1c) was used, the ketone product (3a) could be obtained in an appreciable 22% yield (entry 3). We then moved to search for better NHC catalysts (entries 4-6) and found that with the use of triazolium B could give 3a in 42% yield (entry 4). The N-substituents on the triazolium pre-catalyst had a clear influence on the reaction efficiency, as replacing the N-mesityl substituent of B with a N-2,6-methoxyl (C) or N-2,6-chloro (D) substituents led to muchdropped yields (entries 5-6). With B as an NHC pre-catalyst, we then evaluated several metal and organic photocatalysts (entries 7-11, for details see Supplementary
Development of one-pot operation. We next moved to identify conditions for coupling of two carboxylic acids instead of using pre-prepared acyl imidazole (1c) as precursor of the NHC-bound acyl azolium intermediate (Fig. 2a). To our delight, we found that when carboxylic acid 1d and the carboxylic acid activation reagent CDI were mixed and stirred for 2-3 h followed by the addition of the other reagents (as those used in the optimal condition in Table 1, entry 8), 3a could be afforded in 68% yield. Several other acid activation reagents (HATU, DIC, and DCC) evaluated here did not give satisfactory results under the standard condition. It turned out the one-pot operation approach worked well for different substrates including marketed drugs, such as nalidixic acid and fenoprofen, giving the corresponding ketone products (3b-3d) with yields that were only slightly lower than those by using pre-prepared acyl imidazole substrates (Fig. 2b). This formally decarboxylative coupling of two carboxylic acids exhibits attractive applications in late-stage functionalization of marketed drugs and coupling of two drug molecules (Vide infra, Fig. 5).
Substrate scope. With optimized reaction condition in hands, we then moved to evaluate the scope of our reaction (Fig. 3).
To have a precise estimation on the reaction efficiency of this NHC and photocatalyst-catalyzed coupling process, we chose to use preprepared acyl imidazoles as precursors of the acyl azolium intermediates. As a technical note, the one-pot operation is recommended for practical synthetic applications, albeit with a slight loss on product yields. We first evaluated the scope of acyl imidazoles (1) by using 2a as model substrate (Fig. 3). Various substituents on para-position of aryl ring were tolerated to give the corresponding coupling ketone products with good to high yields, regardless of their electronic nature (3e-3h). For example, electron withdrawing groups such as fluorine and chlorine atom substituted acyl imidazoles gave 3a (Fig. 2b) and 3e in 78 and 58% yields respectively. Electron donating groups such as methoxyl and tertiary butyl substituted substrates gave desired 3c (Fig. 2b) and 3h in 71% and 69% yields. The position of substituents on aryl ring has little effect on the reaction outcomes. Meta-and orthosubstituted acyl imidazoles were all converted to the target ketone products with good yields (3i-3k). Cyano group was also tolerated, albeit with lower yield (3l). Acyl imidazole containing radical-sensitive C-C double bond was excellent substrate, leading to 3m in 83% yield. Notably, the ester and ketone moieties, which are typically incompatible in traditional methods for ketone synthesis such as Grignard reactions, can be tolerated in this mild coupling reactions (3n and 3o). Acyl imidazoles bearing heteroaryls such as quinoline (3p), furan (3q), thiophene (3r), and pyridine (3s) proceeded smoothly in this reaction to give the target ketone products in moderate to good yields. Potentially reactive functional groups such as boronic ester (3t), terminal alkynyl (3u), and alkyl chloride (3v) were all tolerated. To test the feasibility of our method for the synthesis of dialkyl ketones, a variety of aliphatic acyl imidazoles were investigated. Our method can be used to prepare methyl ketones by using acetic acid-derived acyl imidazole as an acetyl source. For example, flurbiprofen and naproxen underwent decarboxylative acetylation smoothly to afford methyl ketone 3x and 3y in 64% and 76% yield, respectively. Other aliphatic acyl imidazoles also proceed smoothly to afford the dialkyl ketones in high yields (3z-3aa). As a technical note, when aliphatic acyl imidazole substrates were used, a lower loading of Cs 2 CO 3 and a longer wavelength were optimal in order to avoid side reactions likely caused by base-mediated α-carbon deprotonation (of the acyl imidazole substrates) 73,74 and high energy irradiation.
We next examined the scope of the carboxylic acids by using para-methoxyl substituted acyl imidazole 1e as a model precursor of NHC-bound acyl azolium intermediate. Chlorine atom on aryl ring was tolerated to give 52% yield of target ketone (3ab). The methyl group at α position of carboxylic acids (2) could be replaced with ethyl (3ac), benzyl (3ad), cyclohexyl (3ae), and allyl (3af) groups without affecting the reaction outcomes dramatically. Cyclic alkyl units such as cyclopentyl (3ag) and cyclohexyl (3ah) could also be incorporated into the ketone product, albeit with a slight decrease in yield for the former one (3ag). Besides, the α-tertiary alkyl substituted carboxylic acids were also competent in this coupling reaction, which thus render this method to be an efficient tool to access sterically hindered ketones with an all-carbon quaternary center at α-position (3ai-3al, 3ap). Notably, these congested ketone products are generally challenging to synthesis [76][77][78][79] . To our delight, N-protected cyclic amino acids were excellent substrates for this coupling reaction, thus offering a straightforward method to access valuable α-amino ketones from readily available and abundant materials. For example, N-protected piperidine-2-carboxylic acid, proline, and their derivatives such as 4-fluornated proline, 2-methylproline, and bridged-ring containing piperidine-2-carboxylicacid were all competent substrates to give the corresponding α-amino ketones in moderate yields (3am-3aq). Three-component radical relay reactions. Multi-component radical relay reactions provide a powerful tool for the synthesis of complex skeletons from simple and readily accessible starting materials [80][81][82] . While significant progress has been achieved in NHC and photocatalyst co-catalyzed ketone synthesis 57-64,83 , three-component radical relay coupling that allows various carbon radicals involved is still limited 72 .
To test the feasibility of our coupling reaction in muticomponent radical relay reactions, alkenes were incorporated into the system as third coupling partners. We began the investigation with N-protected-4-methyl-4-carboxylic acid (2b), para-methoxyl substituted acyl imidazole 1, and 4-methyl styrene (4a) as model substrates (Fig. 4). To our delight, the three-component coupling ketone product (5a) was obtained exclusively in up to 90% yield with complete regioselectivity under our optimal conditions. The exceptional regioselectivity comes from the preference in forming aryl ring stabilized benzyl radical. Further investigation revealed that various styrenes bearing different functional groups such as methoxyl, fluorine, and chlorine atoms are tolerated to give desired ketones in high yields, regardless of their electronic nature (5b-5d). Other radical precursors (including amino acids) and acyl imidazoles (including hetero cyclic ones) were also competent for this three-component radical relay coupling (5e-5l). For example, tertiary and secondary radicals generated from cyclic amino acids were incorporated into a series of γ-amino ketones (5g-5j) in high yields. Marketed drugs such as nalidixic acid was transformed into ketone product (5l) with high level of complexity. These three-component radical relay coupling reaction further demonstrated the flexibility and utility of our NHC and photocatalyst co-catalyzed coupling reactions.
into the corresponding ketone products (6a-6d) in moderate to high yields under our one-pot operation approach. Besides, our method also allows direct decarboxylative coupling of two drug molecules that tethered to carboxylic acid group, affording ketone entities bearing two drug fragments. For example, nalidixic acid, a synthetic quinolone antibiotic, was coupled with flurbiprofen, and indoprofen smoothly to deliver corresponding new ketone entities (6f-6g) in good yields. In case of the coupling of nalidixic acid and ketoprofen, while the one-pot coupling gave poor yield, the target ketone product (6e) was obtained in 78% yield when preprepared acyl imidazole was used as starting material. Aliphatic carboxylic acids could also undergo the one-pot coupling to construct dialkyl ketones. For example, dehydrocholic acid and flurbiprofen were coupled with naproxen smoothly under the one-pot condition, giving the dialkyl ketones (6h, 6l) in 65% yields. Our method was also applied in the modification of adapalene, a third-generation topical retinoid. Nevertheless, the onepot coupling of adapalene and indoprofen only gave 41% yield of desired ketone product (6i). To our delight, the using of preprepared acyl imidazole could give moderate yield (56%). Under the same condition, adapalene coupled with ibuprofen, and flurbiprofen successfully to give the target coupling products (6j-6k) in moderate yields. These results revealed that our method exhibits remarkable potential application in rapid assembly of complex molecules.
Mechanistic studies. To gain insight into the mechanism, mechanistic studies were conducted next (Fig. 6). The model reaction was suppressed with 6% yield of the target product was observed in the presence of TEMPO (2,2,6,6-Tetramethylpiperidinooxy) under otherwise identical conditions. Meanwhile, TEMPO-adduct was isolated in 29% yield, suggesting a radical pathway (Fig. 6a). This was further supported by a radical clock reaction, in which a cyclopropyl group was installed in the carboxylic acid (Fig. 6b). The observation of the cyclopropane ring-opening product (7) strongly supports the presence of α-cyclopropyl carbon radical that generated from oxidative decarboxylative of carboxylic acid (2c) (Fig. 6b). It is well known that excited state of photocatalyst can promote the oxidative decarboxylation of carboxylic acids 84,85 . Meanwhile, the acyl azolium intermediate II (Int. II) were proved to be excited under visible light irradiation 64,86 and further act as strong oxidant 64 , which may also contribute to the decarboxylation process. Thus, photophysical behaviors of pre-prepared Int. II and PC-3 were investigated to identify these two possibilities. We firstly investigated the UV-Vis absorption spectra of Int. II and photocatalyst (PC-3) under the above reaction concentration (Fig. 6c). While Int. II can absorb visible light until around 430 nm (Fig. 6c, green line), PC-3 shows much more strong absorption (Fig. 6c, red line). Besides, Stern-Volmer quenching experiments of Int. II and PC-3 were conducted by using sodium carboxylate (8) as a quenching reagent. The results revealed that sodium carboxylate 8 could not quench the emission of Int. II (Fig. 6e, black line), suggesting that the Int. II was not responsible for the oxidative decarboxylation of carboxylic acid. In contrast, the emission of PC-3 (Fig. 6c, blue dashed line) could be effectively quenched by sodium carboxylate 8 ( Fig. 6d and the red line in Fig. 6e) with the quenching rate constant (k q ) calculated by using the reported lifetime (2280 ns) of PC-3 as 6.68 × 10 8 M −1 s −1 87 . This result is consistent with the crucial role of PC-3 in this decarboxylative coupling reaction (Table 1, entry 13), supporting the photocatalyst-initiated oxidative decarboxylation of carboxylic acid.

Discussion
In conclusion, we have developed a convenient approach for coupling of carboxylic acids and acyl imidazoles by merging NHC and photo catalysis. The acyl imidazoles can be generated in situ via the reaction of another carboxylic acid with CDI, thus providing a straightforward strategy for the synthesis of ketone moieties from two carboxylic acids. The mild condition, operational simplicity, and use of readily available carboxylic acid as both acyl source and radical precursors make our method much more convenient (than previous approaches) for broad applications. Notably, the coupling process can also be intercepted by the addition of an alkene as the third coupling partner, leading to a three-component radical reaction with the formation of sophisticated products from simple starting materials. Natural products and medicinal molecules and their fragments can be directly coupled using our methods. Ongoing studies in our laboratories include the preparation and study of drug conjugates and derivatives, and concise synthesis of complex molecules by applying and further developing our carboxylic acid radical coupling reactions.

Methods
General procedure for the decarboxylative coupling of carboxylic acids and acyl imidazoles. To a 10 mL Schlenk tube equipped with a stir bar was added acyl imidazole 1 (0.15 mmol), carboxylic acid 2 (0.10 mmol), NHC pre-catalyst B (0.02 mmol), photocatalyst PC-3 (0.001 mmol) and dry Cs 2 CO 3 (0.20 mmol). The Schlenk tube was sealed and placed under argon before 2 mL of dry MeCN was added. The reaction was stirred and irradiated with one/two blue LED Kessil lamp (λ max = 427 nm, 3 cm away from the Schlenk tube, with cooling fan to keep the reaction temperature at 30-40°C.) for 24 h. Then the reaction mixture was filtered through a pad of celite and washed with ethyl acetate. The filtrate was concentrated in vacuum to afford the crude material which was purified by column chromatography (silica gel, EtOAc/hexanes) to give product 3.
General procedure for the one-pot decarboxylative coupling of two carboxylic acids. In gloves box, to a 4 mL vial equipped with a stir bar was added 4-flurobenzoic acid 1d (0.2 mmol) and CDI (0.2 mmol). MeCN (2 mL) was added as solvent. The reaction mixture was stirred for 2.5 h in gloves box at room temperature until the solution become homogenous. The resulting reaction mixture was mixed with carboxylic acid 2a (0.10 mmol), NHC pre-catalyst B (0.02 mmol), photocatalyst PC-3 (0.001 mmol) and dry Cs 2 CO 3 (0.20 mmol). The resulting mixture was sealed and take out from the gloves box. Then the reaction was stirred and irradiated with one blue LED Kessil lamp (λ max = 427 nm, 3 cm away from the Schlenk tube, with cooling fan to keep the reaction temperature at 30-40°C.) for 12-24 h. The reaction mixture was filtered through a pad of celite and washed with ethyl acetate. The filtrate was concentrated in vacuum to afford the crude material which was purified by column chromatography (silica gel, EtOAc/hexanes) to give product 3a.

Data availability
All data generated in this study are provided in the Supplementary Information/Source Data file, and can be obtained from the authors upon request. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30583-2