Introduction

Phenyl rings are ubiquitous in medicinal chemistry, featuring prominently in various biologically relevant molecules1,2. Departing from the concept of ‘flatland’, a range of saturated bridged bicyclic ring systems—such as bicyclo[1.1.1]-pentane (BCP), bicyclo[2.1.1]hexane (BCHex), bicyclo [3.1.1]heptane (BCHep), and bicyclo[2.2.2]octane (BCO)—have been developed to replace phenyl rings, yielding drug molecules with enhanced biological activity and improved physicochemical properties (Fig. 1a)3,4,5,6,7,8,9,10,11. Among the phenyl containing bioactive molecules, aniline stands as a prevalent structural motif, especially in high-throughput screening libraries (Fig. 1b)12,13. However, its prevalence poses substantial challenges in later pharmaceutical development stages, notably due to its tendency to drive compounds towards metabolically derived toxicities or heighten the risk of adverse drug-drug interactions14,15. For instance, Capesaris, a synthetic nonsteroidal estrogen initially developed by GTx, Inc. for advanced prostate cancer treatment, was found to have a high risk of venous thromboembolism in the clinical trials due to the reactive metabolite (RM) formation16,17. One of the approach to tackle this challenge involves leveraging sp3-rich saturated bridged bicyclic ring systems (e.g. [1.1.1]BCP to replace the phenyl ring) (Fig. 1b)1,10,18,19,20,21,22. These isosteric replacements often demonstrate heightened resistance against reactive metabolite formation, along with great opportunities for intellectual property development. However, current synthetic methodologies predominantly focus on bioisosteric analogs of non- or para-substituted anilines (e.g., amino-BCP with substituents at the bridgehead position), necessitating intricate de novo synthesis with lengthy steps and limited modularity22,23,24,25,26,27,28,29. Recently, small-ring cage bioisostere scaffolds like bridgehead-disubstituted [2.1.1]BCHex and [3.1.1]BCHep, and 1,2-disubstituted [2.2.1]BCHep have emerged to mimic meta- and ortho-substituted arenes (Fig. 1c)30,31,32,33,34,35,36. Considering that many drug candidates containing aniline motifs also display meta- or ortho-substitutions, the creation of a versatile synthetic approach with high modularity, capable of universally generating saturated bioisosteres of substituted anilines—encompassing para-, meta-, and ortho-substitutions—holds immense value.

Fig. 1: Background and introduction to saturated bioisosteres.
figure 1

a Examples of the phenyl bioisosteres core appearing in drugs and bioactive compounds. b The world’s best-selling small molecule drugs containing aniline and diaryl amine motifs. The toxicity of Capesaris associated with the metabolism of aniline could be mitigated by BCP isostere replacement. c Representative saturated bioisosteric scaffolds that faithfully replicate the geometry of substituted anilines. d Typical approaches to amino-bioisosteres from small-ring cage carboxylic acids require pre-activation of carboxylic acids with iodomesitylene diacetate. In this work, a photoelectrochemical decarboxylative platform enables modular access to saturated bioisosteres of anilines.

Full size image

Small-ring cage carboxylic acids such as 3-(methoxycarbonyl)bicyclo[1.1.1]pentane-1-carboxylic acid 1a and 4-(methoxycarbonyl)bicyclo[2.2.2]octane-1-carboxylic acid 1b are readily accessible from commercial sources or synthesized using established methods. This accessibility has sparked our interest in devising a catalytic direct decarboxylative amination reaction. Such a process would enable the streamlined and modular synthesis of saturated bioisosteres of substituted anilines directly from the corresponding carboxylic acids. An elegant example from the MacMillan group showcased the facile decarboxylative amination of certain small-ring cage carboxylic acids with indazole, employing a synergistic combination of photoredox catalysis and copper catalysis29,36. However, the protocol’s requirement for pre-activation of carboxylic acids with iodomesitylene diacetate to form redox-active iodonium carboxylates, crucial for generating key alkyl radical species such as BCP radical, presents inherent limitations within organic synthesis (Fig. 1d, top left). Direct decarboxylative amination offers synthetic appeal as it circumvents the need for preparing and isolating redox-active carboxylic acid derivatives37,38,39,40. Nevertheless, a significant challenge lies in effectively oxidizing carboxylic acids or carboxylates (e.g. hexanoate ion, E1/2red = +1.16 V vs SCE) without affecting the more readily oxidizable amine coupling partners (e.g. p-toluidine, E1/2red = +0.81 V vs SCE)41. Additionally, closing the net-oxidative catalytic cycle without resorting to stoichiometric chemical oxidants like DTBP or copper salts remains another key concern. The use of stoichiometric oxidants can indeed limit the range of substrates, posing challenges for large-scale preparation, and may not be applicable to saturated bridged bicyclic ring systems39,40,42. The interrupted Kolbe reaction allows for the direct anodic oxidation of alkyl carboxylic acids to generate carbocation species43. These species can engage in the subsequent nucleophilic replacement, enabling direct decarboxylative C(sp3)–N bond forming reactions44,45. However, applying this method to small-ring cage skeletons commonly found in saturated bioisosteres has proven challenging due to the highly unstable carbocations formed at the 3o bridgehead position, resulting in rapid ring fragmentation (Fig. 1d, bottom left)44,45,46.

Drawing inspiration from recent advancements in photoelectro-catalytic transformations47,48,49,50,51,52,53, we envisage to invent a modular assembly of aniline bioisosteres through photoelectro-catalytic decarboxylative C(sp3)–N coupling process. Specifically, our approach is centered on generating kinetically stable small-ring cage radical intermediates (e.g., BCP radical) using a photo-induced ligand-to-metal charge transfer (LMCT) process within the solution phase54,55,56,57,58. This strategic design aims to bypass the potential issue of direct electrode overoxidation of radicals into carbocations, consequently mitigating the risk of skeletal rearrangement in small-ring cage bioisosteres. On the other hand, harnessing the synergies of light and electrical energy enables a substantial reduction in the required electrode potential for the crucial redox event, thus offering mild reaction conditions that are suitable for the coupling of easily oxidizable amines. Additionally, employing an electrochemical potential instead of stoichiometric chemical oxidants in the net oxidative coupling process not only provides practical, economic, and sustainable advantages but also has potential to broaden the scope of both reactants42,47. Herein, we report the discovery of a photoelectro-catalysis enabled direct decarboxylative C(sp3)–N coupling reaction (Fig. 1d, right). The chemistry displays a high degree of practicality with a remarkable scope (>65 examples) across various valuable substrates (including many previously inaccessible saturated bioisosteres of meta-, ortho- substituted anilines) using inexpensive carbon-based electrodes. The utility of this approach was demonstrated through the direct conversion of natural products and drug molecules into bioisosteric aniline analogs, as well as in the synthesis of BCP analogs of clinically relevant drugs.

Our proposed mechanism for the direct decarboxylative C(sp3)–N Coupling is depicted in Fig. 2a. We aimed to merge the capacity of LMCT photocatalysis to generate alkyl radicals from carboxylic acids with copper’s established role in reductive elimination for C–N bond formation. Upon light irradiation, the Fe(III) complex A undergoes homolysis of the Fe-O2CR bond (LMCT decarboxylation), leading to CO2 extrusion and yielding Fe(II) photocatalyst B along with an alkyl radical. This alkyl radical could be rapidly captured by the copper(II)-amido complex C, derived from the coordination of the amine coupling partner with a copper(II) precatalyst, forming the crucial copper(III) complex D. Subsequent reductive elimination occurs to give the desired sp3 C–N coupling product and copper(I) catalyst E59,60. The copper(I) species E and Fe(II) species B are expected to undergo a facile anode oxidation to yield the higher valent metals to close the catalytic cycles. Meanwhile, to maintain electrochemical balance, hydrogen evolution occurs through proton reduction at the cathode.

Fig. 2: Catalytic cycles and reaction parameters evaluation.
figure 2

a Proposed catalytic cycles. b Investigation of reaction parameters, standard conditions: 1a (0.6 mmol, 3.0 equiv), 2a (0.2 mmol, 1.0 equiv), Fe(OAc)2 (15 mol%), Cu(acac)2 (15 mol%), Et3N (0.5 equiv), TBABF4 (1.2 equiv), DMF (4.0 mL), carbon felt as the anode, RVC as the cathode, undivided cell, I = 2.5 mA, 6 W LEDs (390 nm), N2, 20 °C, 15 h. bYield determined by 1H NMR. cIsolated yield.

Full size image

Results

Evaluation of the reaction conditions

Based on this mechanistic understanding, we explored the decarboxylative C(sp3)–N cross-coupling between BCP-derived carboxylic acid 1a and p-toluidine 2a (Fig. 2b). After extensive evaluation of the reaction parameters, we identified the optimal reaction conditions—simultaneous electrolysis and violet light-emitting diode (LED) irradiation (λmax  =  390 mm) of the reaction mixture—employing a cooperative iron mediated photocatalysis and copper catalysis. This led to the formation of the BCP-based saturated bioisosteric product 3 in an 83% yield, with 10% of 2a remaining unreacted (Entry 1). One of the major byproducts from 1a is the decarboxylative hydrogenation by-product, specifically methyl bicyclo[1.1.1]pentane-1-carboxylate, as detected by GC-MS analysis. Replacing the Fe(OAc)2 with commonly used LMCT decarboxylation catalysts, such as Ce(OTf)3 or CeCl3, resulted in complete reaction cessation (Entry 2). This observation could be attributed to the inherent higher oxidation potential of cerium(III), which renders it unsuitable for C–N bond formations with easily oxidizable amine coupling partners47,50. Similarly, no reactivity was observed using 4CzIPN as the photocatalyst (Entry 3), underscoring the crucial role of Fe(OAc)2 in this photoelectro-catalytic reaction. Alternative transition metal catalysts like Cu(hfacac)2 and Ferrous oxalate provided the product in moderate yields (Entries 4 and 5). Employing nickel foam as the cathode dramatically decreased the yield to 8% (Entry 6). When using RVC instead of carbon felt as the anode, a moderate yield of 48% was obtained (Entry 7). A shift to 365 nm LED reduced the yield to 20% (Entry 8). Furthermore, 2,4,6-collidine proved to be an effective base, affording 3 in 65% yield (Entry 9). It is worth noting that constant voltage electrolysis (0.8 V) could also deliver the product in 51% yield (Entry 10). Finally, control studies confirmed the indispensable roles of light irradiation, electricity, and Fe, Cu catalysts (Entries 11 − 12).

Scope of amines

We next turned our attention to exploring the scope of this direct decarboxylative C(sp3)–N bond forming reaction. As shown in Fig. 3, an extensive range of aryl and heterocyclic amines were successfully coupled with BCP carboxylic acid 1a under standard conditions, affording the corresponding amino-BCPs as products. Aryl amines bearing electron-donating groups such as Me (3, 6), PhO (7), Boc-protected amine (12) underwent the targeted reaction in moderate to good yields (35-81%). The reaction also displayed compatibility with electron-withdrawing groups including CF3O (8), HCF2O (9), CF3S (11), CF3 (13), CN (14), COOMe (15), SO2Me (16), POMe2 (17). Our reaction’s chemoselectivity was illustrated by its compatibility with various functional groups, including aryl halides (18-21), Bpin (22), ester (23), CN (24), unprotected OH (25), among others. Notably, secondary amines like indoline (27), methyl indoline-2-carboxylate (28) smoothly participated, yielding the desired tertiary amine products in good yields (60%, 66% respectively). Our methodology extended beyond aryl amines, demonstrating success even with pyridine or pyrimidine-containing heterocyclic amines which might otherwise cause catalyst poisoning, highlighting the protocol’s potential attractiveness in drug discovery. In addition, sulfoxide imines (37-38) and imine (39) also demonstrated a high degree of reaction compatibility, with sulfoxide imine 38 being prepared in 75% yield on a gram scale. Furthermore, the reaction exhibited exceptional broad functional-group tolerance, as evidenced by late-stage couplings of complex structures derived from natural products and drug molecules, including 4-amino-L-Phe-Gly (40), aminoglutethimide (41), florfenicol (42), uridine (43), estradiol benzoate (44), and podophyllotoxin (45).

Fig. 3: Substrate scope with amine coupling partners.
figure 3

Red indicates newly formed C–N bonds. Isolated yields (given as a percentage) are denoted for each product. Reaction conditions: 1a (0.6 mmol, 3.0 equiv), amine (0.2 mmol), Fe(OAc)2 (15 mol%), Cu(acac)2 (15 mol%), Et3N (0.5 equiv), TBABF4 (1.2 equiv), DMF (4.0 mL), carbon felt as the anode, RVC as the cathode, undivided cell, I = 1–2.5 mA, 6 W LEDs (390 nm), N2, 20 °C, 15–18 h. The current and reaction time vary slightly for each substrate; see Supplementary Information, Section 4 for the specific reaction conditions.

Full size image

Scope of carboxylic acids

To explore the versatility of our technology in creating saturated bioisosteres of anilines, we evaluated a diverse range of BCP carboxylic acids featuring various bridgehead substitutions (Fig. 4). Encouragingly, functionalities such as CF2H (46), phenyl (47), CN (48), silyl-protected oxygen (49), amides (50, 51), carbamate (52), and ester (53) were well accommodated, yielding the desired bioisosteres products with high efficiency (40-65%). The reactivity could be extended to [2.2.2]BCO (54) and [2.2.1]BCHep (55) scaffolds, offering alternative products suitable as para-substituted aniline bioisosteres. Of particular note, several intricate bicyclic ring systems, recently identified as meta-substituted phenyl bioisostere mimics, were also amenable to our current C(sp3)–N cross-coupling conditions, furnishing aniline bioisosteres that were challenging to obtain (56-58). Moreover, our method allowed for the synthesis of ortho-substituted aniline bioisostere in a synthetically useful yield (59). The successful incorporation of complex units from natural products and drug molecules into BCP-based bioisosteres underlines the potential of our approach in drug discovery (60-62). Beyond bioisosteric replacement scenarios, this newly developed reaction exhibited applicability in the decarboxylative functionalization of secondary and primary alkyl carboxylic acids, providing N-alkyl products in good yields (63-71).

Fig. 4: Substrate scope with alkyl carboxylic acid.
figure 4

Red indicates newly formed C–N bonds. Isolated yields (given as a percentage) are denoted for each product. Reaction conditions: alkyl carboxylic acid (0.6 mmol, 3.0 equiv), amine (0.2 mmol), Fe(OAc)2 (15 mol%), Cu(acac)2 (15 mol%), Et3N (0.5 equiv), TBABF4 (1.2 equiv), DMF (4.0 mL), carbon felt as the anode, RVC as the cathode, undivided cell, I = 1.0–3.0 mA, 6 W LEDs (390 nm), N2, 20 °C, 15 h. The catalysts loading for primary carboxylic acids are Fe(OAc)2 (20 mol%), Cu(acac)2 (20 mol%). The current vary slightly for each substrate; see Supplementary Information, Section 4 for the comprehensive details.

Full size image

Synthetic applications

The product derivatization is illustrated in Fig. 5a. Ester hydrolysis under basic conditions yielded 72 in nearly quantitative yield, offering a versatile carboxylate handle for diversification. For example, treatment of 72 with MeLi resulted in the formation of ketone-substituted BCP 73 in 90% yield. The Curtius rearrangement occurred smoothly to form the other bridgehead C(sp3)–N bond (74). Additionally, BH3 reduction of the carboxylate group produced BCP containing primary alcohol (75) in 80% yield. Condensation with Meldrum’s acid, followed by decarboxylation under heating conditions, provided 76 in moderate yield. The oxidation of 27 to indole substituted BCP 77 was easily within reach. The facile transformation of the indole-substituted BCP carboxylic acid 78 into 79 through a photo-induced decarboxylative Heck-type reaction further demonstrated the versatility of our method. We also explored integrating the BCP scaffold into drug molecules. As depicted in Fig. 5b, the BCP analog of the anticancer agent Capesaris was synthesized from decarboxylative amination product 18. It was achieved through a synthetic sequence involving amide formation, Weinreb ketone synthesis, Baeyer–Villiger oxidation, deprotection of the silyl, and hydrolysis (22% yield over 6 steps). This three-dimensional bioisosteric replacement has the potential to enhance absorption, distribution, and metabolism properties, while addressing issues associated with metabolically derived toxicities (vide supra).

Fig. 5: Synthetic applications and mechanistic studies.
figure 5

a Product derivatization. b This protocol enables the rapid preparation of saturated bioisosteres of anticancer drug Capesaris. c The radical clock and radical inhibition experiments reveal the involvement of alkyl radical intermediate. d The carbocation pathway is excluded by using water as nucleophile. e Experiments investigating the impact of Et3N suggest its potential role as a protector for both the amine substrate and product, shielding them from decomposition. f The reaction proceeds using stoichiometric metal catalysts even in the absence of electricity. g Prussian Blue detection experiment proved the existence of Fe(III) species. h Cyclic voltammetry experiments indicated that TolNH2 promoted the generation of Cu(II) species, while carboxylic acid facilitated the formation of Fe(III) intermediates (compound concentration:10 mM in DMF, reference Electrode: Ag/AgCl electrode, scan rate:100 mV s–1). i, UV−vis absorption experiments provided further evidence for photoexcited Fe(III) intermediate.

Full size image

Mechanistic studies

Mechanistic studies were conducted to unravel the reaction pathway. As shown in Fig. 5c, radical clock reaction with 2-cyclopropylacetic acid 83 produced the ring-opening product 84 in 30% yield. Introduction of 2.0 equiv of TEMPO in a radical inhibition experiment reduced the yield of 3–26%, and HRMS analysis confirmed the presence of TEMPO adduct product 85. Moreover, replacing amine 2a with H2O or MeOH failed to produce alcohol product 87 or ether product 88 (Fig. 5d). All these findings exclude the carbocation pathway and strongly suggest the participation of an alkyl radical intermediate in this C–N coupling process. The substantial increase in the recovery rate of both 2a and 3 upon adding Et3N suggested a dual role for Et3N, serving not only as a base but also protecting the amine substrate and product from decomposition (Fig. 5e). This stabilization is further corroborated by the detection of diethylamine by HRMS, indicating that Et3N plays a protective role for the reaction components. In another experiment, employing a stoichiometric amount of Fe(III) catalyst [Fe(OAc)2OH] and Cu(acac)2 without electricity, the desired product was obtained in 31% yield (Fig. 5f). In addition, a Prussian Blue detection experiment confirmed the formation of a Fe(III) species through the electrochemical oxidation of the Fe(OAc)2 precatalyst (Fig. 5g)61. This collective evidence indicates the pivotal role of Fe(III) as the active species, initiating LMCT to generate alkyl radicals that participate in the ensuing C–N bond formation. We also conducted cyclic voltammetry experiments to further elucidate the reaction mechanism (Fig. 5h). Cu(acac)2 exhibited no distinct redox features between –0.5 V and 1.0 V. However, the addition of TolNH2 introduced a new anodic event around 0.52 V, corresponding to the oxidation of XCu(I)NH2Tol to XCu(II)NHTol at approximately 0.52 V (vs. Ag/Ag+). An increase in current was observed upon the addition of Et3N, suggesting an interaction between Et3N and XCu(I)NH2Tol. In comparison to Fe(OAc)2, the inclusion of carboxylic acid 1a led to the appearance of a new anodic event with an onset potential around 0.04 V. These observations suggested that TolNH2 promoted the production of Cu(II) species, similar to how carboxylic acid facilitated the formation of Fe(III) intermediates. Additionally, UV−vis absorption experiments clearly showed the emergence of a new ultraviolet absorption peak upon the addition of carboxylic acid 1a, indicating the formation of the Fe(II)-carboxylic acid complex (Fig. 5i). After the electrolysis, the absorption was significantly enhanced, revealing the formation of a Fe(III) intermediate that was photoexcited, displaying LMCT transitions.

Discussion

In summary, we have demonstrated how the combination of photoelectro-catalysis and copper catalysis can be used to overcome the long-standing challenge of the direct decarboxylative C(sp3)–N coupling, providing a versatile and modular approach for the assembly of saturated bioisosteres of aniline-based drug molecules. Noteworthy features of our protocol include its mild reaction conditions, simple experimental operations, broad scope and generality across a wide number of counterparts, enabling the universal synthesis of saturated bioisosteres for various substituted anilines, encompassing para-, meta-, and ortho-substitutions. Given the importance of bioisosteric replacement in drug development, we anticipate that this reaction will significantly propel advancements within pharmaceutical research.

Methods

General procedure for synthesis of saturated anilines bioisosteres

To an ElectraSyn vial was added TBABF4 (0.24 mmol, 1.2 equiv), carboxylic acid (0.4 mmol, 2.0 equiv), Fe(OAc)2 (0.03 mmol, 15 mol%), Cu(acac)2 (0.03 mmol, 15 mol%) and aniline (0.20 mmol, 1.0 equiv). The vial was transferred into a N2-filled glovebox. Degassed DMF (4.0 mL) was added followed by Et3N (0.10 mmol, 50 mol%). The ElectraSyn vial was sealed with a cap which equipped with anode (carbon felt) and cathode (RVC), then removed from the glovebox. The reaction was irradiated with LEDs (6 W, 390 nm) under the vessel and electrolysis was initiated at a constant current of 1.0–2.5 mA. After 10 h at 20 oC, a solution of carboxylic acid (0.20 mmol, 1.0 equiv) in DMF (0.20 mL) was add to the reaction. After 5 h, the photolysis and electrolysis were terminated, the tube cap was removed and electrodes were rinsed with EtOAc, which was combined with the crude mixture. The organic layers were further washed with brine, dried over anhydrous Na2SO4 and concentrated in vacuo and purified by alkaline prepared thin layer chromatography plate.