Electrochemical reactor dictates site selectivity in N-heteroarene carboxylations

Abstract

Pyridines and related N-heteroarenes are commonly found in pharmaceuticals, agrochemicals, and other bioactive compounds^1,2. Site-selective C─H functionalization would provide a direct way of making these medicinally-active products^3,4,5. For example, nicotinic acid derivatives could be made by C─H carboxylation, but this remains an elusive transformation^6,7,8. Here, we describe the development of an electrochemical strategy for the direct carboxylation of pyridines using CO₂. The choice of electrolysis setup gives rise to divergent site selectivity: a divided electrochemical cell leads to C5-carboxylation, whereas an undivided cell promotes C4-carboxylation. The undivided cell reaction is proposed to operate via a paired electrolysis mechanism^9,10, wherein both cathodic and anodic events play critical roles in altering the site selectivity. Specifically, anodically-generated iodine preferentially reacts with a key radical anion intermediate in the C4-carboxylation pathway via hydrogen-atom transfer, thus diverting the reaction selectivity via the Curtin-Hammett principle¹¹. The scope of the transformation was expanded to a wide range of N-heteroarenes including bi- and terpyridines, pyrimidines, pyrazines and quinolines.

N-heteroarenes are among the most common structural units in pharmaceuticals, agrochemicals, and other biologically active molecules^1,2. The direct functionalization of N-heteroarenes provides an efficient approach to the synthesis and modification of medicinal agents and has thus attracted widespread interest from both academia and the pharmaceutical industry^3,4,5. A major yet challenging objective in the C─H functionalization of N-heteroarenes is achieving precise control over the regioselectivity. In this regard, existing methods are frequently limited to intrinsic regiochemical outcomes determined by the electronic and steric properties of the substrates^12,13. Elegant advances in catalysis have led to methods that can attain either improved regioselectivity or new regiochemical outcomes through the influence of covalent^14,15,16 or noncovalent directing groups¹⁷. Despite these major contributions, achieving tunable and high regioselectivity remains a critical challenge.

Pyridine (1) is the second most prevalent heterocycle among US FDA-approved pharmaceuticals¹. The regioselective C─H carboxylation of pyridines and related N-heteroarenes (e.g., pyrazoles, quinolines, pyrimidines) with CO₂ is an attractive transformation, as it provides compounds containing core structures of many medicinally relevant agents (Fig. 1a). Furthermore, upgrading CO₂ to value-added organic products is an attractive synthetic approach from both economic and sustainability perspectives^{18,19,20,21,22,23,24,25}. In stark contrast to the catalogs of transformations available for arylation, alkylation, and borylation^{3,12,26,27,28,29,30,31,32,33,34} (Fig. 1b), synthetic methods for the carboxylation of pyridine derivatives remain scarce (Fig. 1c). Sole precedents in this regard include Zare’s work on Ni-catalyzed carboxylation of unsubstituted pyridine⁶, Fuchs’s report on electrochemical carboxylation of quinolines with low yield and selectivity⁷, and Li’s development of Rh-catalyzed pyridine carboxylation via a directing group strategy^8,35. Thus, methods that directly incorporate CO₂ into pyridine with precise and tunable site selectivity are highly desirable and will facilitate the discovery and synthesis of bioactive molecules. Against this backdrop, we report here methods for the regioselective C─H carboxylation of pyridines and related N-heteroarenes by means of electrochemical activation (Fig. 1d). By varying the type of electrolysis cell from divided to undivided, we were able to completely reverse the site selectivity by enabling Curtin-Hammett control¹¹ via paired electrolysis^9,10.

Fig. 1. Significance and background.

a, Examples of medicinally relevant molecules containing pyridinecarboxylic acid motif. b, Examples of common strategies for pyridine C─H functionalization. c, Precedents for C─H carboxylation of pyridine and related N-heteroarenes with CO₂. d, Regiodivergent electrochemical C─H carboxylation of pyridines (this work).

In recent years, electrochemistry—a technique that has long been used for studying redox processes—has gained increasing attention in the synthetic organic chemistry community^36,37,38. Through the use of a pair of electrodes and a power source, electrochemistry provides chemists with a direct source of electrons and holes of continuously variable potential. This feature has been explored in synthetic contexts for the activation of common and inert organic molecules for the generation of highly reactive intermediates. Related to our current work, electrochemical methods have been developed for the oxidative functionalization of C─H bonds. For instance, seminal recent reports detailed the use of anodic chemistry for sp² C─H functionalization of arenes via the directing group approach^39,40. Furthermore, the direct oxidation of sp³ C─H bonds has been demonstrated in the context of complex target derivatization^41,42,43. In light of the unique features of electrochemistry, we envisioned an electroreductive strategy toward the desired pyridine C─H carboxylation. Pyridine derivatives can typically undergo electroreduction under moderate potentials^6,44,45, and we aimed to explore this feature to achieve their electrochemical activation and carboxylation.

Through initial exploration, we found that the carboxylation of 2-phenylpyridine (2) could indeed be realized under electrochemical conditions using a divided electrolysis cell (see Supplementary Information Section 7 for initial reaction discovery). Further optimization led to a highly selective C5-carboxylation protocol in the presence of Cu(OTf)₂ and KO^tBu as additives, Fe and Zn as cathode and sacrificial anode, respectively, in an ⁿBu₄NI/NMP electrolyte solution saturated with CO₂ under a constant current of 5 mA. Under these conditions, the desired product was obtained as the methyl ester (3) upon methylation in 72% yield (Fig. 2a, Method A and entry 1) with effectively complete regioselectivity. C4-carboxylation (4) was observed as a minor product (3/4 > 20:1) by gas chromatography (GC). During optimization, we found that the addition of a small amount of H₂O and O₂ was beneficial to the reaction yield. To counter the desired pyridine reduction, Zn electrode in the anodic chamber is oxidized to Zn²⁺ into the solution to complete the circuit.

Fig. 2. Electrochemical carboxylation of pyridines.

a, Optimal conditions and control experiments. b, Mechanistic hypothesis. c, Electrogeneration of H-atom acceptor I₂. GC yields of the major regioisomeric product are given using tetradecane as internal standard. The ratio between 3 and 4 was determined by GC analysis. In Fig. 2b, the proposed reaction between Int4 and I₂ could proceed via HAT, wherein either I₂ or iodine atom generated from I₂ (likely promoted by KO^tBu⁴⁸) serves as the hydrogen-atom acceptor. Alternatively, PCET may take place with I₂ as the electron acceptor, and the carboxylate group, incipient I⁻, or ^tBuO⁻ as the proton acceptor. Computed natural atomic charges are given for Int1, and key free energy data are given in unit of kcal/mol, both using SMD(NMP)/(U)M06-2X/6-311+G(d,p) level of theory. Footnotes: ^[a]Reaction was set up under air before a CO₂ balloon was adapted for electrolysis; no exogenous O₂ was added. ^[b]For reactions with ⁿBu₄NBF₄, ⁿBu₄NClO₄, and ⁿBu₄NPF₄ as electrolyte, results were obtained without adding H₂O, which provided higher yield than conditions with H₂O. ^[c]At room temperature (21 °C) for 48 h.

Various control experiments were carried out to elucidate the necessity of each reaction component in the optimal system. In the absence of either Cu(OTf)₂, H₂O, or KO^tBu, the reaction remained productive, giving 3 in moderate yield and high regioselectivity (Fig. 2a, entry 2). Thus, these three additives do not play explicit roles in the reaction mechanism. Cyclic voltammetry studies showed that Cu(OTf)₂ is preferentially reduced to metallic Cu particulates on the cathode before substrate reduction, which may serve as a heterogeneous material to enhance reaction rate⁴⁶. The role of the base KO^tBu is likely to increase the effective concentration of CO₂ in the reaction medium through formation of carbonates [e.g., ^tBuOCO₂K (BocOK)]⁴⁷. However, under anaerobic conditions, substantially lower yield was observed (entry 3). In light of the above information, we found that the reaction could be directly set up in air without the addition of exogenous O₂ when BocOK was used as the base, affording 3 in yield similar to the optimal conditions (entry 4). Various electrolytes besides ⁿBu₄NI proved to be compatible with the reaction, yielding 39–62% product 3 (entry 5). In addition, sacrificial anode Zn could be replaced with Pt without significantly decreasing the yield (entry 6), in which case the counter reaction is likely the oxidation of the electrolyte I⁻ to I₂ in the anodic chamber. Finally, the electrical input is critical, as no product was obtained without an applied current. Importantly, running the same reaction under thermal chemical conditions using a reducing metal (e.g., Zn or Mg) instead of electrochemistry did not give the desired carboxylation.

We hypothesized that the observed carboxylation starts with the one-electron reduction of pyridine 2, giving rise to radical anion Int1 (Fig. 2b, top). Density functional theory (DFT) calculations suggested that Int1 bears the highest electron population at C5 position among the carbon atoms of the pyridine ring, which will thus engage in nucleophilic addition to CO₂ to furnish Int2. Subsequently, a second cathodic reduction furnishes dianion Int3, which is followed by oxidative rearomatization by O₂ to provide desired carboxylation product. Additional DFT calculations further supported this mechanistic hypothesis. The addition of Int1 to CO₂ at the C5 position is predicted to be endergonic by 8.9 kcal/mol with a Gibbs free energy of activation (ΔG^‡) of 13.7 kcal/mol, which is more favorable than the observed minor pathway with carboxylation at the C4 position (ΔG^‡ = 14.5 kcal/mol, ΔG = 12.3 kcal/mol).

The mechanism for the C5-carboxylation indicated that the reaction regioselectivity is dictated by the intrinsic electronic properties of key intermediate Int1, which precludes the selective functionalization at other sites via the same pathway. However, DFT calculations suggested that the reaction regioselectivity could potentially be altered via the Curtin-Hammett principle. Because the nucleophilic addition of Int1 to CO₂ is endergonic and reversible, the carboxylation regioselectivity could be kinetically controlled if a follow-up irreversible step favors an alternative pathway. Our DFT data revealed that if addition of CO₂ occurs at the C4 position, the C─H bond at C4 of Int4 has a significantly lower bond dissociation free energy (BDFE) than the C─H bond at C5 of Int2 (ΔBDFE = 3.8 kcal/mol; Fig. 2b). Thus, if a hydrogen-atom acceptor is introduced to the reaction to promote the formation of Int5 via hydrogen-atom transfer, the reaction regioselectivity could be shifted to favor C4-carboxylation.

Our initial attempts to explore this strategy proved unfruitful; various agents that can provide a hydrogen-atom acceptor in situ (e.g., halogens, peroxides) were preferentially and unproductively consumed on the cathode in the initial stage of electrolysis, thus precluding them from participating in the desired pathway. While searching for a solution to this issue, we serendipitously discovered that the use of an undivided cell in lieu of a divided cell under otherwise nearly identical conditions provided a drastic change in the carboxylation regioselectivity, and isonicotinic acid methyl ester 4 was obtained as the predominant product in 50% yield and good regioselectivity (4/3 = 6:1; Fig. 2a, entry 7 vs. entry 1). We reasoned that under these conditions, anodically generated I₂ (in the form of I₃⁻) from ⁿBu₄NI played the role of a hydrogen-atom acceptor, either through direct hydrogen-atom transfer (HAT)⁴⁸ or by means of proton-coupled electron transfer (PCET)⁴⁹. Critically, the use of an undivided cell allowed for the constant replenishment of I₂ via anodic oxidation of I⁻, thus compensating for the unproductive cathodic reduction of I₂ (Fig. 2c). Indeed, the reaction solution remained brown through the reaction course. Thus, the high C4-carboxylation selectivity was achieved at the expense of lower current efficiency (<5%; vs. 23% in a divided cell) and longer reaction time (48 h). Further experiments led to simplified and optimal conditions for C4-carboxylation employing only a substoichiometric amount of KO^tBu, generating 4 in excellent yield and regioselectivity (Fig. 2a, Method B and entries 8–9).

Control experiments showed that in the absence of KO^tBu as base, product 4 was formed but in substantially lower yield (Fig. 2a, entry 10). In stark contrast to the C5-carboxylation, the efficiency of the C4-carboxylation is highly sensitive to the identity of the electrolyte. While ⁿBu₄NBr provided the product in moderate yield (entry 11), none of the other electrolytes tested gave any product under the same conditions (entry 12). Changing the anode from Pt to sacrificial Zn provided only traces of 4 (entry 13), which supports the proposed role of anodically generated I₂ in the reaction mechanism. Finally, the passing of an electric current was again critical to the observed reactivity.

Under these optimal conditions, we investigated the scope of the electrochemical carboxylation reactions. Using the divided cell protocol (Fig. 3a), a collection of 2-arylpyridines with varying electronic properties underwent conversion to the corresponding nicotinic acids in good to excellent yields with high C5 selectivity. This transformation was compatible with diverse functional groups including thioethers (5), esters (10), aryl fluorides (12), trifluoromethyl ethers (6), tertiary amines (8, 18), alkenes (19), and electron-rich heterocycles (14–15, 22), some of which are potentially sensitive to strongly reducing conditions. In addition, 2,3-disubstituted pyridines were competent substrates, yielding C5-carboxylated product with excellent selectivity (16–19). We also investigated pyridines with 2-substituents other than aryl groups and found that those with phosphinoyl (20) and silyl (21) groups could be converted to carboxylation products in moderate yields, which could be further elaborated into other functionalized pyridines⁵⁰. Interestingly, 2-(methyldiphenylsilyl)pyridine underwent C4-carboxylation selectively (21), which was attributed to increased electron population at the C4 position of the corresponding radical anion intermediate caused by the silyl group as supported by DFT data (see Supplementary Information Section 8.4). Further, 2,2’-bipyridine as substrate furnished 5,5’-dicarboxylated product 23 in 17% yield along with a minor amount of mono-carboxylation. However, pyridine derivatives with other 2-substituents (including H) generally showed low reactivity (Fig. S11), which implicates the likely role of the 2-substituents in stabilizing key radical and ionic intermediates in the reaction mechanism (Fig. 2b). Under undivided cell conditions (Fig. 3b), 2-arylpyridines with a variety of substituents proved suitable substrates, yielding the corresponding isonicotinic acids (4, 24–42) with excellent regioselectivity (Method B). In addition, several 2,6-disubstituted pyridines reacted smoothly to furnish the desired C4-carboxylation product (34–37). 2-Cyanopyridine (38) and methyl picolinate (39) could also be converted into the C4 products as the major regioisomers. We also surveyed several polypyridyls and found that substrates 40–42 could transform into the corresponding carboxylation products selectively at C4 along with small quantities of regioisomers. Thus, our reaction could be employed as a general method for the modification of these important ligands in metal catalysis.

Fig. 3. Substrate scope.

a, Method A: C5-carboxylation. b, Method B: C4-carboxylation. c, Scope of other N-heteroarenes. Isolated yield of the major regioisomeric product is reported. The ratio between the major regioisomeric product and any identified minor regioisomeric products is reported in parentheses, which was determined by ¹H NMR of the crude reaction mixture. A regioisomeric ratio of >20:1 indicates that the minor product was either not observed by ¹H NMR or observed in a minute amount. In panel c, red and blue circles denote the major carboxylation site using Method A and Method B, respectively; purple circles denote the major carboxylation site shared by both Methods A and B. Footnotes: ^[a]Cu(OTf)₂ (0.2 equiv), KO^tBu (0.25 equiv), TBAI (0.15 M), without adding H₂O, 4 mA. ^[b]Yields were determined by ¹H NMR. ^[c]With 0.5 equiv KO^tBu. ^[d]With 0.15 equiv KO^tBu; unlike other quinoline substrates, C4-carboxylation was observed as the major regioisomer (see Supplementary Information Figure S20 for discussion). ^[e]With 0.05 equiv KO^tBu; 5% 2,5-di-CO₂Me product was also observed. ^[f]Regioselectivity was not determined.

We also expanded the reaction scope to various other N-heteroarenes given their prevalence in pharmaceutical and materials chemistry (Fig. 3c). Thus, 2-phenylpyrimidine (43), quinoxaline (44–45), and quinoline (46–52) derivatives were successfully transformed into the corresponding carboxylic acids with excellent regioselectivity. In most cases, selectivity switching depending on the reaction cell construction was observed, with Method A favoring carboxylation at the meta-position to the N whereas Method B providing ortho- or para-carboxylation. Specifically, quinolines 47–49, 51, 52 gave C2-carboxylation in an undivided cell (Method B), which could be attributed to energetically more favorable HAT at the C2 position vs. the C4 position (i.e., Int4 to Int5; see Supplementary Information Figure S20 for DFT data support). For substrates 46 and 53, we observed low reaction yields of unexpected C4-carboxylation under Method A, which was likely due to disruptive steric interactions introduced by the C2 substituent that deters the reaction from happening at the neighboring C3. 4,6-Dicarboxylated pyrazolo[1,5-a]pyridine (54) and 5-carboxylated 1-methyl-7-azaindole (55) were also obtained. Finally, we carried out product derivatization to showcase the synthetic utility of these new protocols using various directed and undirected aromatic C─H functionalization protocols reported in the literature (see Supplementary Information Section 16 for product derivatization).

Previous control experiments showed that the reactivity dependence on electrolyte ⁿBu₄NI was opposite in divided and undivided cell systems (Fig. 2a, entry 5 vs. entry 12). These findings are consistent with our mechanistic hypothesis that in an undivided cell setting, the observed reversal of regioselectivity resulted from a paired electrolysis scenario with participation from both cathodic and anodic reactions, the latter of which promotes the oxidation of I⁻. The low or lack of reactivity observed with ⁿBu₄NBr (entry 11) and other electrolytes such as ⁿBu₄NBF₄, ⁿBu₄NClO₄, and ⁿBu₄NPF₄ (entry 12) likely arose from more challenging anodic oxidation to effectively generate a hydrogen-atom acceptor. To further verify the participation of both electrodes in the C4-carboxylation, we carried out an alternating current (AC) electrolysis using a divided cell, wherein the current on the working electrode was altered between +6 mA (oxidation) and −6 mA (reduction) with a one-second pulse length. This operation emulated the undivided cell system and allowed both radical anion Int1 and I₂ to be generated on the same electrode in the same chamber, and C4-carboxylation was indeed observed as the predominant pathway (Fig. 4a). This selectivity reversal was however not observed by simply adding I₂ to the cathodic chamber of a divided cell in a direct current electrolysis (see Supplementary Information Section 14) due to competitive and unproductive consumption of I₂ on the cathode. This result highlights the importance of paired electrolysis in promoting C4-carboxylation, as it allows for hydrogen-atom acceptor I₂ to be continuously (re)generated on the anode (Fig. 2c).

Fig. 4. Mechanistic studies.

a, Alternating current electrolysis. b, Kinetic isotope effect study. Footnotes: ^[a]See Supplementary Information Section 8.8 and 8.9 for experimental details. ^[b]KIE data are collected in triplicates, with the error bar indicating the range of experimental values. ^[c]KIE from parallel experiments is not reported for Method A due to variability in reaction rate caused by divided cell setups.

Finally, we obtained kinetic isotope effect (KIE) data; Method A did not show any primary KIE at C5, whereas Method B displayed a k_H/k_D value of 2 at C4 (Fig. 4b). These findings suggested that the cleavage of the C─H bond at C4 is rate-determining in C4-carboxylation, which is in agreement with the proposed mechanism of selectivity control via the Curtin-Hammett principle. The small normal KIE observed for Method A indicated that the initial electron transfer step to generate Int1 is likely involved in the rate-determining process (see Supplementary Information Figure S23 for DFT prediction). Thus, this method presents a unique system wherein the regioselectivity of an electrochemical reaction can be controlled by the cell construction (divided vs. undivided cell), giving rise to divergent C─H carboxylation products.

Data availability:

All data supporting the findings of this work are available within the paper and its Supplementary Information.