Delineating Physical Organic Parameters in Site-Selective C-H Functionalization of Indoles.

Site-selective C-H functionalization is a great challenge in homogeneous transition-metal catalysis. Herein, we present a physical organic approach to delineate the origin of regioselective amidation of N-acylindoles through Ir(III) catalysis. Bulkiness of N-directing groups of indole substrates and electronics of carboxylate additives were identified as two major factors in controlling C2 and C7 selectivity, and their microscopic mechanisms were studied with DFT-based transition state analysis. Computational insights led us to interrogate a linear free energy relationship, and parametrization of molecular determinants enabled the establishment of an intuitive yet robust statistical model that correlates an extensive number of validation data points in high accuracy. This mechanistic investigation eventually allowed the development of a new C2 amidation and alkenylation protocol of indoles, which affords the exclusive functionalization at the C2 position with up to >70:1 selectivity.

Introduction

Discriminating a specific C–H bond and selectively transforming it to valuable functional groups is a formidable challenge in synthetic organic chemistry.[1−3] A number of recent studies have led to the consideration of two tools based on substrate-guided[4−7] and catalytic system-controlled approaches.[8,9] Substrate modification has been widely utilized to drive the reaction toward the desired pathway, but its applications are rather confined to the specially engineered substrates.[10−13] In this context, the modulating catalytic system which includes a change of catalyst structure and/or reaction conditions is more ideal for versatile transformations, although identification of effective catalyst systems is much more challenging.[14−17] The difficulties in catalyst design mainly arise from a lack of mechanistic understanding behind the desired selectivity.

Given that indoles are widely present in pharmaceutical, natural, and biologically active molecules,[18] derivatization of this privileged scaffold in a selective manner could give a great impact in the related research areas.[19,20] In recent years, modulation of a N-carbonyl moiety in the indole skeleton has been scrutinized as an effective strategy to direct the C–H bond cleavage at the C2 and/or C7 position in the functionalization of indoles (Scheme a). As the C2–H bond is the most acidic in indole compounds,[18,21] lithiation of indoles readily takes place at the C2–H bond by applying strong bases.[22] Likewise, a number of transition-metal-catalyzed C–H activations of indoles are effective toward the C2 functionalization, and their C2 selectivity is believed to originate from the intrinsic C–H bond strength.[23−27] Modification of the directing group to N-pyrimidyl is another effective strategy for the C2 amidation of indoles,[28−30] but postmodification of substrates, such as removal of the directing group, often requires harsh conditions.[28] On the other hand, when sterically bulky N-acyl groups are installed, C7 metalation is known to be more favored,[31] and indeed, this strategy enabled the introduction new C–C and C–N bonds at the C7 position with high selectivity.[32−34] Despite these advances, divergent catalytic systems to finely modulate the C2/C7 selectivity are still elusive.

Scheme 1

(a) Substrate-Controlled Approaches in C–H Functionalization of Indoles and (b) Selectivity Reversal Achieved by the Integrated Substrate/Base Effects

Herein, we report an integrated substrate-guided/base-controlled approach to invert site selectivity in Ir(III)-catalyzed C–H functionalization of N-acylindoles (Scheme b). Examination of various reaction parameters led us to reveal microscopic mechanisms associated with site-selective C–H activation steps. Physical organic parametrization of substrates and additives enabled the construction of robust structure-selectivity relationships that can predict an extensive set of reaction outcomes. The series of analyses eventually led to the establishment of a new C2 amidation and alkenylation protocol offering an excellent level of site selectivity.

Results and Discussion

We recently reported versatile C–H amination procedures using organic azides as the amide source[35−37] that allow an exclusive C7 functionalization of N-pivaloylindoles through Ir(III) catalysis (Scheme a, R1 = tBu).[32] A key to achieving high reactivity was to utilize silver acetate as an additive, which is critical in accessing the aromatic C7–H bond by N-pivaloyl-directed C–H activation.[38,39] This previous result enabled us to envision that site-selective functionalization of other C–H bonds might also be viable if we can delineate underlying working modes related to the C–H cleavage step. In this context, the C2 position of the indole is an attractive synthetic target because it is proximal to the N-alkylcarbonyl directing group, potentially enabling chelation-assisted C–H bond activation. We sought to interrogate various reaction parameters possibly affecting the selectivity in an anticipation of achieving new C2-amidation protocols. First, four indole substrates having different sizes of directing groups were subjected to the previously optimized conditions by assuming that less bulky substituents might drive the rotational equilibrium toward (E)-isomer (vide supra).[31,34,40] As illustrated in Scheme a, all substrates were well-converted to the desired amidated products in the presence of silver acetate, but the site selectivity was highly dependent on the N-acyl groups. While exclusive C7 amidation was observed with N-pivaloylindole without any sign of C2 functionalization, subjecting N-isobutyrylindole, which has a smaller alkyl chain, gave rise to C2 amidation in 21% yield in addition to 54% of C7-amidated product. Indeed, a further decrease in the size of the N-substituent increased the portion of C2 products, and an almost nonbiased formation of two regioisomeric products was obtained with N-acetylindole (C7/C2 = 1.4:1).[41]

Scheme 2

Influence of Reaction Parameters on the Site Selectivity: (a) N-Directing Group Effects and (b) Silver Carboxylate Effects

Unless otherwise indicated, reactions were run with indole substrate (0.20 mmol), tosyl azide (0.22 mmol), [IrCp*Cl2]2 (5 mol %), silver carboxylate (20 mol %), and AgNTf2 (10 mol %) in 1,2-dichloroethane (0.5 mL) at 40 °C for 12 h. Ratios were determined by 1H NMR analysis of the crude mixture, and averaged over two runs.

Ratios were determined by HPLC analysis of the crude mixture.

Having identified the steric influence of the substrate, we further sought to examine the additive effects on the site selectivity (Scheme b). We hypothesized that the stereoelectronic property of the carboxylate additives could be important, since the C–H activation step involves a close interaction between the catalyst and the base in the selectivity-determining deprotonation process.[42−45] When more basic carboxylate salt, i.e., silver pivalate, was applied to the reaction in lieu of silver acetate, C7 selectivity was increased to give the ratio of 2.5:1. Interestingly, in sharp contrast, utilizing silver trifluoroacetate significantly suppressed the C7 functionalization, and highly favorably gave C2-amidated product 4 in 62:1 selectivity. These observations highlight that the nature of carboxylates directly affects the relative energy differences between selectivity-determining transition states.[46]

For a better understanding of the underlying origin of these effects, putative reaction intermediates and their working modes were evaluated with DFT calculation (Figure ). The most and least basic carboxylates (i.e. pivalate and trifluoroacetate) were representatively chosen for the case study. The pivalate-mediated C–H cleavage, as marked with the blue line, might be initiated from an iridium-substrate adduct species I, which contains κ2 coordination of the pivalate. This adduct species has been frequently proposed in the related C–H functionalization reactions.[42,44,45] Upon decoordination of one dative Ir–O bond and subsequent interaction between the metal center and either C2–H or C7–H bond, an agostic complex II or III can be readily formed.[39,42] As predicted, replacing the Ir–O bond with the agostic interaction is a thermodynamically uphill process. Complex I was destabilized by 11.2 kcal/mol to furnish complex II, while 7.5 kcal/mol was required for the formation of its C7 isomeric species III. These intermediates subsequently traverse transition states II-TS and III-TS via the concerted metalation–deprotonation (CMD) mechanism to generate iridacycle IV and V, respectively. The reaction barrier for the C2–H cleavage from the agostic complex II was 6.1 kcal/mol, whereas that of the C7–H cleavage was 8.8 kcal/mol. Interestingly, however, the overall activation barrier for the C7–H iridation was still lower in energy than the C2 pathway: the activation barrier from I to II-TS was 17.3 kcal/mol, whereas 16.3 kcal/mol was required to reach III-TS.

Figure 1

Competitive reaction energy profiles of C2–H and C7–H cleavage steps. Solid lines indicate kinetically favorable pathways. Dashed lines are disfavored ones. Free energy surfaces were constructed at the M06/SDD+6-311+G**/SMD(dichloroethane)//M06/Lanl2dz+6-31G** level of theory.

In sharp contrast, reverted selectivity was computationally estimated when similar reaction pathways were evaluated with trifluoroacetate as the base (red line, Figure ). The overall activation barrier for C2–H cleavage (I′ to II′-TS) was 2.3 kcal/mol lower in energy than the corresponding C7–H activation (I′ to III′-TS), which is consistent with the experimental observation in Scheme b. This dichotomy with two different carboxylates might be explained by considering two aspects: (i) stability of the agostic intermediates, and (ii) deprotonation barriers from them. In general, C7-agostic intermediates are more stable than analogous C2 complexes, mainly because of the higher covalency of less acidic C7–H bonds. However, their relative stability depends on the electron-richness of the iridium metal center: while pivalate-containing III is 3.7 kcal/mol more stable than II, trifluoroacetate complex III′ is slightly more stable than II′ by only 2.2 kcal/mol, suggesting that the electron-deficient metal center favors the agostic complexation with both C2–H and C7–H bonds, and in turn, the stability difference is only marginal with trifluoroacetate complexes. On the other hand, the deprotonation barrier for the C2–H cleavage is generally lower in energy than the C7–H event with both carboxylates, mainly because of the intrinsic acidity of weak C2–H bonds.[18,21] The reaction barrier is especially higher when weak trifluoroacetate base is utilized to cleave the strong C7–H bond (III′ to III′-TS), which requires a free energy of 13.4 kcal/mol. We propose that these two influences are convoluted and eventually reflected in C2 and C7 selectivity.

The above transition state analysis revealed that the electron-richness of carboxylate additives is directly associated with the stability of agostic complexes and deprotonation barriers. This implication led us to envision that thermodynamic descriptors of carboxylates might be useful to draw a quantitative relationship that accounts for the measured selectivity. We anticipated that a mathematical model would allow the prediction of the relative free energy differences and provide chemical insights for the selectivity. Among various quantitative physical parameters,[47−50] one modern approach is to utilize computational descriptors, which have been recently introduced by Sigman and co-workers.[51] This DFT-based approach is particularly effective when experimentally derived parameters are not available.[51] For the description of the stereoelectronic features, corresponding carboxylic acids were employed as the carboxylate surrogates, and their structures were computationally optimized with the M06-2X/jun-cc-pVTZ level of theory. From the optimized geometries, a number of stereoelectronic parameters, such as natural bond orbital (NBO) charges[52,53] of oxygens, IR frequencies[54,55] of carbonyl stretches, and the related Sterimol values,[56,57] were enumerated.[58]

For an extensive evaluation of the structure-selectivity relationship, an additional 20 silver carboxylates were prepared and subjected to the standard reaction condition (Figure a). Whereas the use of alkyl carboxylates (entries 1–4) preferred the formation of C7 amidation, phenylacetate derivatives reversed the selectivity (entries 5–7 and 10–13). Significantly, C2 selectivity became more pronounced when halogenated acetates were employed: while chloroacetate gave the C2 product in moderate selectivity (2.6:1), exclusive C2 amidation was observed with pentafluoropropionate (entry 23). These ratios were then converted to ΔΔG‡ in an assumption with the Curtin–Hammett situation and subjected into regression analysis with interrogated molecular descriptors.[59] Whereas no notable relationship was found with various steric parameters, such as Sterimol B5 values of carboxylic acids, excellent univariate correlation was found when the average charge of two carboxyl oxygens (NBOO-avg) was considered (Figure b).[60] In this model, 23 experimental outcomes were well-correlated with NBOO-avg with R2 of 0.91. Cross-validation of the model by the leave-one-out (LOO) method further suggested that the constructed model is highly robust (QL1O2 = 0.89). Remarkable linearity with a negative slope strongly suggests that carboxylic acids having more positive NBOO-avg charge may give rise to higher selectivity in favor of the C2 product.

Figure 2

(a) Dependence of site selectivity on the type of silver carboxylates (regiomeric ratio was determined by 1H NMR or HPLC analysis from an average of two runs). Measured ΔΔG‡ values were calculated from ΔΔG‡ = −RT ln(4/3) at 40 °C. (b) Plot of measured ΔΔG‡ versus Sterimol B5 value (left) and average NBO charges of two oxygens (NBOO-avg, right) of carboxylic acid surrogates.

Motivated by the remarkable correlation, we further wondered if two identified factors, which are N-directing groups of indole substrates and carboxylate additives, could be simultaneously accounted for in a quantitative manner (Scheme a). For this purpose, a series of reactions were examined with different directing groups and additive combinations, as listed in Figure S14. A total of 44 reaction outcomes were obtained and subjected to multivariate analysis. We selected 16 responses to adequately train simultaneous change of the substrates and additives (Figure S14a), and subsequent regression analysis revealed a two-parameter model displaying an excellent level of accuracy and robustness (blue squares, Figure b). This normalized model includes a linear combination of NBOO-avg, which was an effective parameter in Figure b, and the Sterimol B1 value of N-protecting groups that is a direct illustration of the minimum radius perpendicular to the substituent. The NBO charge displayed a higher contribution to the model, as noticed by a larger absolute value of coefficient (−0.97). Robustness of the relationship was further examined by applying a validation set having an extra 28 responses, and those outcomes are remarkably well-fitted in the developed model (red diamonds, R2 = 0.94). This quantitative relationship clearly suggests that both bulkiness of directing groups and electronic nature of carboxylate additives are directly involved, as predicted by transition state analysis.

Scheme 3

(a) Synthesis of C2 Iridacycle 5, (b) Stoichiometric C–N Coupling with Tosyl Azide, (c) Catalytic Reaction Using 5 as the Catalyst, and (d) Kinetic Isotope Effect (KIE) Experiments

Figure 3

(a) Hypothetical reaction mechanisms on the selectivity determining step. (b) Multivariate regression analysis for the regioselective C–H amidation of indole.

Having understood the selectivity-determining mechanism, experimental attempts to isolate putative reaction intermediates were made to support hypothesized working modes (Scheme ). Upon treatment of N-acetylindole with a dimeric iridium(III) complex in the presence of AgOTFA, the formation of an iridacycle 5 was observed. The solid state structure of this species was characterized by X-ray analysis, which unambiguously confirmed the selective C2–H activation. When azide 2 was added to 5, C–N coupling smoothly took place at room temperature to afford an amido-inserted complex 6 in high yield, and its structure was also confirmed by X-ray analysis (Scheme b). Importantly, an amidation reaction utilizing complex 5 as a catalyst afforded C2-amidated product 4 in an exclusive manner, strongly suggesting that the cyclometalated species 5 is an active intermediate in the catalytic cycle (Scheme c). In addition, primary kinetic isotope effect (KIE) values revealed that the C–H bond cleavage is irreversible and turnover-determining (Scheme d).[61]

The substrate generality of this mechanism-driven C2-amidation approach was evaluated with pregenerated Cp*Ir(OTFA)2 as the catalyst (Scheme ).[62] All reactions smoothly took place under mild conditions to furnish C2-amidated indoles in synthetically acceptable yields. While electron-donating substituents facilitated the amidation reactions (4, 7, 8), electron-withdrawing groups provided slightly lower reactivity (9–11). The position of substituents was less critical in the efficiency (12–17). The selective C2-amidation was not impeded even in the presence of C-3 substituents (18–20): the structure of compound 18 was fully characterized including its X-ray structure to confirm the site selectivity. The reaction was compatible with versatile functional groups as representatively demonstrated by the tryptophan derivative (19). In addition, a variety of amide sources were applicable to the present condition. Indeed, sulfonyl azides of benzene derivatives (20–24) and naphthalene (25) were readily reacted under the optimal conditions. Moreover, aliphatic variants also worked well without difficulty (26–28).

Scheme 4

Substrate Scope of Indoles and Sulfonyl Azides

Unless otherwise indicated, reactions were run with substrates (0.20 mmol), sulfonyl azides (0.22 mmol), IrCp*(OTFA)2 (10 mol %), and NaNTf2 (10 mol %) in 1,2-dichloroethane (0.5 mL) at 40 °C for 16 h; isolated yields.

Run at 80 °C.

To further highlight broad applicability of our strategy, a related C–H functionalization reaction was examined. Inspired by Ma’s approach on C7 alkenylation of N-pivaloylindoles,[33] we hypothesized that the site selectivity might be inverted to C2 functionalization on the basis of our mechanistic understanding. Indeed, as illustrated in Scheme a, the C2/C7 selectivity was significantly altered by the choice of carboxylate additives: while no preferential formation was observed with copper acetate, the more electron-rich pivalate gave an increase in C7 selectivity (C2:C7 = 1:1.9). Notably, selective C2 functionalization was achieved with electron-deficient copper trifluoroacetate in 13.2:1 selectivity. Ethyl and benzyl acrylates also gave excellent reactivity and selectivity with slightly modified conditions (Scheme b). These results clearly demonstrated that our mechanism-guided strategy could be broadly applicable to various C–H functionalization reactions of indoles.

Scheme 5

Application of the Strategy to C–H Alkenylation Reaction: (a) Electronic Effect of the Additive on the Selectivity and (b) C2-Selective C–H Alkenylation with Various Olefin Partners

Reactions were run with 3 (0.20 mmol), acrylates (1.0 mmol), [IrCp*Cl2]2 (5 mol %), NaNTf2 (20 mol %), and copper carboxylate (0.42 mmol) in methylene chloride (1.5 mL) at 80 °C for 36 h. Site selectivities were determined by 1H NMR analysis of the crude mixture.

Isolated yields.

Conclusion

In summary, we demonstrated that regioselectivity in the Ir(III)-catalyzed C–H amidation of indoles could be completely switched from the C7 to C2 position by modulating stereoelectronic parameters of substrates and catalysts. Transition state analysis shed light on the origin of the selectivity-determining step, and the understanding was leveraged to construct a quantitative structure-selectivity relationship by utilizing computationally derived descriptors. A robust two-parameter model enabled the correlation of 44 reaction outcomes with a high accuracy. A mechanistic understanding further allowed the establishment of a selective amidation/alkenylation protocol to access C2-amidated indoles in good to excellent yields.