Over the past three decades, considerable progress has been made in the development of methods to construct sp2 carbon–nitrogen (C–N) bonds using palladium, copper or nickel catalysis1,2. However, the incorporation of alkyl substrates to form sp3 C–N bonds remains one of the major challenges in the field of cross-coupling chemistry. Here we demonstrate that the synergistic combination of copper catalysis and photoredox catalysis can provide a general platform from which to address this challenge. This cross-coupling system uses naturally abundant alkyl carboxylic acids and commercially available nitrogen nucleophiles as coupling partners. It is applicable to a wide variety of primary, secondary and tertiary alkyl carboxylic acids (through iodonium activation), as well as a vast array of nitrogen nucleophiles: nitrogen heterocycles, amides, sulfonamides and anilines can undergo C–N coupling to provide N-alkyl products in good to excellent efficiency, at room temperature and on short timescales (five minutes to one hour). We demonstrate that this C–N coupling protocol proceeds with high regioselectivity using substrates that contain several amine groups, and can also be applied to complex drug molecules, enabling the rapid construction of molecular complexity and the late-stage functionalization of bioactive pharmaceuticals.
Within the field of organic chemistry, the efficient construction of C–N bonds is important owing to the prevalence of nitrogen-containing motifs in a wide array of natural products, pharmaceuticals and functional materials3,4,5. There are several notable routes to the formation of sp2 C–N bonds, including the Buchwald–Hartwig reaction1, Ullmann coupling6 and Chan–Lam amination7. However, sp3 C–N bond formation typically relies on classical methods, such as nucleophilic substitution reactions between nitrogen nucleophiles and alkyl halides8, Mitsunobu alkylations of alcohols using nitrogen nucleophiles9, reductive amination with carbonyls10 or olefin hydroamination11. Recently, several research groups have reported transition-metal-catalysed variants of the alkylation reaction of nitrogen nucleophiles with aliphatic halides12,13. In 1894, Curtius reported the rearrangement of acyl azides to form C–N-containing aliphatic substrates14. We questioned whether it might be possible to expand this concept to complex, medicinally relevant nitrogen-bearing fragments, thereby accelerating access to drug-like complexity in one step.
The synergistic merger of photoredox15 and transition metal catalysis (termed metallaphotoredox catalysis) has resulted in the development of many cross-coupling reactions that are now being widely adopted within the pharmaceutical sector16. The combination of nickel and photoredox catalysis has enabled the efficient construction of C(sp3)–C(sp2) and C(sp3)–C(sp3) bonds using abundant alkyl carboxylic acids17 and alcohols18. Here we show that metallaphotoredox catalysis involving copper in place of nickel enables alkyl sp3 C–N bond formation in a generic sense without the use of alkyl halides or other prototypical electrophiles. More specifically, we hoped to merge the capacity of photoredox reactions to form alkyl radicals from iodonium carboxylates (derived in situ from carboxylic acids) with the long-established propensity of copper to participate in reductive elimination to form carbon–heteroatom bonds2. By taking advantage of the widely abundant nature of alkyl carboxylic acids and nitrogen nucleophiles such as heteroaromatics, sulfonamides, amides and anilines, we hoped to devise a new fragment coupling reaction that would be broadly useful yet mechanistically orthogonal to established alkylation reactions (Fig. 1). Recently, two methods have been reported with this aim in mind19,20, and these illustrate the capacity of copper to function in decarboxylative mechanisms.
A detailed mechanism for the proposed decarboxylative coupling of sp3 carbon with nitrogen nucleophiles is outlined in Fig. 2a. Excitation of photocatalyst Ir(F-Meppy)2(dtbbpy)PF6 (1) (F-Meppy = 2-(4-fluorophenyl)-5-(methyl)pyridine, dtbbpy = 4,4ʹ-di-tert-butyl-2,2ʹ-bipyridine) is known to generate the long-lived triplet-excited-state *Iriii complex 2 (with a lifetime, τ, of 1.1 μs)21. At the same time, we proposed that coordination of the nitrogen nucleophile 11 with a copper(i) precatalyst followed by deprotonation would readily form the copper(i)-amido species 3. The excited state *Iriii complex 2 (E1/2red [*Iriii/Irii] = 0.94 V versus the standard calomel electrode (SCE) in CH3CN)21 should rapidly oxidize this copper(i) complex 3 (E1/2red [Cuii(BPhen)2/Cui(BPhen)2] = 0.08 V versus SCE in DMF, BPhen = 4,7-diphenyl-1,10-phenanthroline)22 to generate the corresponding copper(ii)-amido system 4 and the corresponding iridium(ii) complex 5. At this stage we considered that iodomesitylene dicarboxylate 8 (which is preformed via the mixing of carboxylic acid 6 and iodomesitylene diacetate 7, see Supplementary Information) would be readily reduced by the newly formed iridium(ii) species 5 (E1/2red [Iriii/Irii] = −1.50 V versus SCE in CH3CN21, Ep [8/8•−] = −1.14 V versus SCE in CH3CN) to generate a carboxyl radical, which upon CO2 extrusion23,24 would produce the desired alkyl radical 9, while reconstituting the ground-state photocatalyst 1. At this stage, we anticipated that copper(ii)-amido complex 4 would capture alkyl radical 9 to form copper(iii) complex 10, which upon reductive elimination25 would yield the desired fragment-coupled sp3 C–N bearing adduct 12 and regenerate copper(i) catalyst 3.
We first examined the proposed sp3 C–N coupling using three electronically disparate nitrogen nucleophiles (indole 11a, azaindole 11b and indazole 11c, Fig. 2b), along with cyclohexyl carboxylic acid 6 as the alkylating reagent, and a wide range of copper(i) and photoredox catalysts. The desired decarboxylative sp3 C–N coupling can be achieved in good to excellent efficiency for all three substrates (60%, 76% and 90% yield, respectively) using Ir(F-Meppy)2(dtbbpy)PF6 as the photocatalyst, CuTC (TC = thiophene-2-carboxylate) as the copper catalyst, BPhen or dOMe-Phen (dOMe-Phen = 4,7-dimethoxy-1,10-phenanthroline) as the ligand, BTMG (BTMG = 2-tert-butyl-1,1,3,3-tetramethylguanidine) as the base, with exposure to 34-W blue light-emitting diodes (LEDs). Notably, a series of control experiments revealed that although the copper(i) catalyst is essential for the desired C–N bond formation in all cases, the absence of light and/or photocatalyst has a profound effect on reaction efficiency depending on the nitrogen nucleophile used. More specifically, the alkylation of indazole 11c is successful with or without photocatalysis (90% compared with 86% yield, respectively), whereas indole 11a and azaindole 11b achieve markedly improved yields and reaction times when light and the iridium photocatalyst are combined with copper (indole 11a, 60% compared with 7% yield; azaindole 11b, 76% compared with 47% yield). We speculate that a non-photonic mechanism is possible when copper(i)-amido species 3 is sufficiently electron-rich to undergo direct electron-transfer with the iodomesitylene dicarboxylate 8, thereby removing the requirement for a redox catalyst to act as an electron shuttle. However, when copper(i)-amido species 3 is not sufficiently reducing, or is formed slowly, the presence of a photoexcited electron-shuttle catalyst becomes essential to achieve useful efficiencies. Indeed, this latter case was found to be the most common, with non-photonic conditions leading to useful yields with only a small subset of substrates (yields for non-photonic conditions are reported in parentheses in Figs. 3 and 4). As such, the combination of copper and photoredox catalysts with light was identified as the superior protocol with which to evaluate a broad range of sp3 C–N cross-coupling reactions.
Having established the preferred reaction conditions, we next examined the generality of this new C–N fragment coupling by exploring the scope of the carboxylic-acid alkylation partner (Fig. 3). Notably, a diverse range of alkyl carboxylic acids can be used in this new protocol to deliver the N-alkyl heteroaryl derivatives in good to excellent efficiency. It is important to highlight that in all cases the reactions were complete at room temperature within 1 h. Another feature of this protocol is that only one regioisomer of the product is produced for all cases shown in Fig. 3, which offers a notable advantage when compared to traditional N-alkylation reactions. Indeed, a broad series of differentially substituted primary alkyl acids can readily participate in this new C–N coupling (13–21, 50–82% yield). Moreover, these mild reaction conditions are compatible with a range of common functional groups, such as terminal olefins (17, 64% yield), terminal alkynes (18, 80% yield), nitro groups (19, 50% yield), esters (21, 80% yield) and protected amines (16 and 20, 63% and 82% yield, respectively). In contrast to many established alkylation reactions, steric encumbrance proximal to the acid group is also well tolerated, as exemplified by the successful incorporation of a neopentyl system (15, 54% yield). Moreover, direct N-methylation (13, 65% yield) can also be realized by carrying out the C–N coupling protocol using commercial MesI(OAc)2 7. Finally, we successfully applied this coupling technology to three complex natural products that contain alcohols, ketones and internal alkenes. All were found to participate in decarboxylative N-alkylation with high efficiency (22–24, 54%–76% yield).
We next sought to examine the scope of secondary alkyl carboxylic acids as alkylating agents. Importantly, a large array of ring-bearing carboxylates can be used to access N-cycloalkylation adducts in good to excellent efficiency (26–31, 40%–85% yield). Moreover, this new transformation is not limited to cyclic systems, as exemplified by the rapid incorporation of acyclic secondary alkyl groups (25, 61% yield), a transformation that is not readily achieved using established alkylation protocols. An additional six examples using other secondary alkyl acids for this sp3 C–N coupling are detailed in Supplementary Information and Extended Data Fig. 1.
A notable feature of this decarboxylative C–N coupling method is the number of tertiary-carbon-bearing carboxylic acids that can be readily used to prepare heteroaryl N-tertiary alkyl derivatives with good to excellent efficiency (32–36, 53%–80% yield). This is especially important given that the N-alkylation using tertiary alkyl halides can often be elusive, if not impossible, using traditional technologies. Given the recent interest in the incorporation of rigid bicyclic structures into drug-like compounds, such as the bicyclo[1.1.1]pentane core (as shown in product 36)26, we expect this light-mediated N-alkylation protocol to be immediately applicable to a large range of medicinal chemistry programs. Indeed, the capacity to access these bridged bicyclic aryl isosteres in only one operation using commercial carboxylic acids and iodomesitylene diacetate (at room temperature in less than 1 h) should enable the rapid adoption of this method.
Next, we turned our attention to the scope of the nitrogen-nucleophile component in this new catalytic alkylation protocol (Fig. 4). Almost every class of medicinally relevant nitrogen heterocycle, including, but not limited to, indazoles (37 and 38, 81% and 73% yield, respectively), azaindoles (39 and 40, 89% and 75% yield, respectively), indoles (41, 58% yield), pyrazoles (42 and 43, 98% and 68% yield, respectively), pyrroles (51, 75% yield), imidazoles (52, 68% yield), triazoles (53, 90% yield), benzimidazoles (54, 67% yield), benzotriazoles (55, 80% yield), purines (56, 60% yield) and carbazoles (57, 46% yield), can be successfully used to deliver N-alkyl products in good to excellent efficiency. Moreover, an additional 42 examples using other nitrogen heterocycles are detailed in Supplementary Information and Extended Data Figs. 2 and 3. For nucleophiles that are as acidic as or more acidic than pyrazoles (such as indazoles, triazoles and imidazoles), the addition of an exogenous base is not generally necessary. The carboxylate anion, which is generated upon the reduction of iodomesitylene dicarboxylate 8 (Fig. 2a), can function as a weak base. One notable feature is that this heterocycle C–N forming mechanism exhibits excellent regioselectivity (Fig. 4) with substrates that possess several N-alkylation sites (for example, pyrazoles, imidazoles, triazoles, indazoles and benzimidazoles). At this stage, we presume that the considerable steric bulk of the ligated copper complex ensures that the sp3 C–N coupling takes place at the least hindered site of these heteroaromatic nucleophiles. This outcome is in sharp contrast to classical alkylation methods that typically lead to regioisomeric mixtures. Several of these nitrogen heterocycles were also tested using our non-photonic conditions, and the corresponding yields are shown in parentheses in Fig. 4. It is clear that although C–N formation can be achieved with these nucleophiles in the absence of light (51–57, 7–39% yield), the dual copper and photoredox protocol enables a more extensive scope and substantially higher efficiencies across the board, providing a more general protocol for this new C–N heterocyclic coupling reaction.
As shown in Fig. 4, this decarboxylative C–N coupling method is not limited to the cross-coupling of nitrogen heterocycles. Under our optimized conditions, a large selection of electron-deficient (and also less acidic) nitrogen nucleophiles, including anilines (44 and 45, 55% and 70%, respectively), aryl sulfonamides (46, 59% yield), alkyl sulfonamides (47, 51% yield), aryl amides (48, 79% yield), phthalimides (49, 61% yield) and cyclic carbamates (50, 71% yield), were found to participate readily in this sp3 C–N coupling. Notably, functional groups including aryl iodides (44, 55% yield), aryl bromides (46, 59% yield) and ketones (45, 70% yield) were readily tolerated, a useful feature with respect to further synthetic manipulation.
A long-established problem in the functionalization of primary amines with alkyl halides is the formation of polyalkylation products, given that the initial secondary amine adduct is more nucleophilic than the starting amine. Notably, as shown in Fig. 4, only monoalkylated products are obtained when primary amides, sulfonamides and anilines are used in this new copper-catalysed protocol (44–48, also see Supplementary Information for additional examples using primary alkyl acids). Moreover, an additional 27 examples using other electron-deficient nitrogen nucleophiles are detailed in Supplementary Information and Extended Data Fig. 3. As already highlighted, this coupling protocol does not appear to be negatively influenced by steric constraints, as ortho-substituted and ortho-,ortho-disubstituted anilines, sulfonamides and amides can be used readily (44, 46, and 48, 55%–79% yield).
To illustrate the utility of this new transformation with respect to drug discovery, we examined this decarboxylative sp3 C–N coupling using six known pharmaceuticals (Celebrex, Axitinib, Zelboraf, Actos, Rilutek and Skelaxin). Using three separate carboxylic acids, we were able to achieve decarboxylative alkylation in all cases in good to excellent yields (58–63, 44%–90% yield). An additional ten examples of pharmaceutical functionalization using this technology are described in Supplementary Information and Extended Data Fig. 4.
For substrates with several nucleophilic sites, achieving regioselective monofunctionalization has been a long-standing challenge27. Therefore, we were pleased to find that this new alkylation technology can be applied sequentially to the same drug molecule to achieve selective alkylation at two discrete nitrogen positions through the judicious choice of reaction conditions. As demonstrated in Fig. 5a, heterocycle 64 contains both an indazole nitrogen and a primary amide; however, when the coupling protocol is performed without an exogenous base, regioselective N-alkylation of the indazole was observed in 80% yield. Moreover, we have carried out this regioselective N-alkylation step on a 7.4-mmol scale to prepare 1.45 g of the N-cyclohexyl indazole derivative 65. The origins of regioselectivity in this decarboxylative coupling arise from the relative acidity of the two N–H moieties in substrate 64 (that is, indazole and amide). More specifically, when no exogenous base is used, the carboxylate anion (formed by reduction of the iodomesitylene dicarboxylate 8) can function as a weak base and thereby selectively deprotonate the more acidic indazole nitrogen (pKa (indazole) <19.8 in DMSO28,29; pKa (phenyl acetamide) = 23.3 in DMSO28) upon coordination to the copper catalyst, which in turn leads to regioselective N-alkylation of the indazole N1 position. Perhaps most important, subsequent exposure of the resulting N-cyclohexyl indazole 65 to our coupling protocol with a second carboxylic acid in the presence of a strong organic base, BTTP (BTTP = tert-butyliminotri(pyrrolidino)phosphorane) leads to a second N-alkylation on the remaining amide nitrogen, again with a useful yield. We anticipate that the capacity to perform regioselective and sequential C–N coupling steps as a function of relative N–H acidities will have substantial benefit with respect to the step economy of building complex molecules, by removing the need to install and remove nitrogen protecting groups.
Finally, to further showcase the utility of this new sp3 C–N coupling method, we performed a series of experiments to compare the decarboxylative protocol with traditional nucleophilic substitution reactions (Fig. 5b). Under various classical SN2 and SN1 alkylation conditions, tertiary alkyl bromide 67 and bromocyclopropane (68) failed to react with 3-chloroindazole to generate the desired N-alkyl products, which is consistent with a lack of literature precedent for using these alkyl bromides with indazole nucleophiles. By contrast, through decarboxylative C–N couplings using commercially available carboxylic acids 69 and 70 and our mild catalytic protocol, the desired N-alkyl products can be obtained in good yields and excellent regioselectivities within 30 min at room temperature using an integrated photoreactor30. As such, we anticipate that this new decarboxylative coupling strategy will provide a useful, complementary new approach to sp3 C–N alkylation.
Here we describe a typical procedure for the decarboxylative sp3 C–N coupling reaction; a summary of general conditions is included in Supplementary Information (Supplementary Fig. 36) and further experimental details are also provided in Supplementary Information.
Procedure for decarboxylative sp 3 C–N couplings
To a 20 ml or 40 ml vial equipped with a stir bar was added photocatalyst, nitrogen nucleophile, iodomesitylene dicarboxylate, copper salt, and ligand. Dioxane was added followed by addition of the base. The solution was sonicated for 1–3 min until it became homogeneous. Next, the solution was degassed by sparging with nitrogen for 5–10 min before sealing with Parafilm. The reaction was stirred and irradiated using two 34-W blue LED lamps (3 cm away, with cooling fan to keep the reaction at room temperature) for 1 h. The reaction mixture was removed from the light, cooled to ambient temperature, diluted with water (15 ml) and ethyl acetate (25 ml), and the aqueous layer was extracted with ethyl acetate (3 × 25 ml). The combined organic layers were washed with brine, dried over Na2SO4, filtered and concentrated. The residue was purified by flash chromatography on silica gel to afford the desired decarboxylative C–N coupling product. For aniline substrates, a solution of these nitrogen nucleophiles in dioxane was used; additionally, if the iodomesitylene dicarboxylate is a liquid, its solution in dioxane was used.
The findings of this study are available within the paper and its Supplementary Information.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research was supported by the National Institutes of Health National Institute of General Medical Sciences (R01 GM103558-03) and gifts from Merck, Bristol-Myers Squibb, Eli Lilly, Genentech and Johnson & Johnson. X.Z. is grateful for a postdoctoral fellowship from the Shanghai Institute of Organic Chemistry. We thank C. Liu and Y. Y. Loh for assistance in preparing this manuscript.
Nature thanks S. Bagley and the other anonymous reviewer(s) for their contribution to the peer review of this work.