Selectivity in the Addition of Electron Deficient Radicals to the C2 Position of Indoles.

Abstract

The addition of electron deficient radicals to the C2 position of indoles has been described in the literature as opposed to electrophilic addition at the C3 position. Density functional theory calculations were used to understand the switch in regioselectivity from C3 to C2 for indole to undergo radical additions. Electron deficient radicals have a lower barrier for reaction at C2 and a lower energy radical intermediate that benefits from benzylic radical stabilization. Trifluoromethyl radical addition has a lower energy barrier than acetonitrile radical, and the C3 addition transition state is just 0.8 kcal/mol higher than C2. This is supported by experimental observations.

Graphical Abstract

Computational methods are used to model the addition of electron deficient radicals (•CH₂CN or •CF₃) at 2 or 3-positon of indoles. Lower reaction barriers are found for reaction at C2 rather than C3. The resulting C2 substituted indole radical intermediate is also lower in energy leading to the experimentally observed C2 selectivity. This is in contrast to long established C3 selectivity for indole reactions with electrophiles.

Introduction

Indole is one of the most common heterocycles found in nature. A review article has estimated that over 80,000 papers have been published describing the synthesis or modification of indole, a structural element in tryptophan, serotonin and many alkaloids.^[1] Indoles are well-described carbon nucleophiles participating in polar 2 e⁻ additions by electrophiles at the C3-position.^[2] While there is HOMO character at C3 and C2 (Figure 1A), there is a higher natural charge from natural population analysis at C3 (Figure 1B: C3, −0.32 vs C2, −0.04). When both C2 and C3 are unsubstituted, reaction proceeds at C3 and as early as 1952 has been exploited by organic chemists.^[3] This reactivity has led to the invention of widely used transformations including Friedel-Crafts,^[4] conjugate addition, Mannich,^[5] Tsuji-Trost allylation^[6] or direct arylation^[7] reactions. Enantioselective alkylations are known using Cu-H,^[8] iminium^[9] or thiourea catalysts^[10] (Figure 1C). Alternatively, direct reaction with electrophiles at C2 in preference to C3 can be accomplished with an N-protecting group metallation via deprotonation^[11] or copper catalysis.^[12]

Figure 1.

A) Diagram of indole HOMO. B) Natural population analysis charges for indole. C) Divergent addition regiochemistry for electrophiles (C3) and D) electron deficient radicals (C2).

In contrast, electron deficient radicals are reported to react directly and selectively at the C2 position of unprotected or unsubstituted indoles (Figure 1D). Bromoacetonitrile^[13] or bromomalonates^[14] activated with Ir or Ru based photocatalysts provide the C2 addition products in high yield (Figure 2). Trifluoromethylation has been accomplished using CF₃SO₂Cl photoredox catalysis,^[15] CF₃I with Pt(ppy)(acac) as photocatalyst,^[16] CF₃I with Fe(II) peroxide,^[17] photoactivation of a Co-CF₃ reagent,^[18] TMSCF₃ with copper catalyst (with N-methylindole),^[19] or bistrifluoroacetyl peroxide.^[20] To understand this switch in regioselectivity, we carried out quantum mechanical calculations to describe the factors that determine the C2 selectivity in radical addition reactions to indoles.

Figure 2.

Examples of photoredox-mediated radical addition to the 2-position of indoles.

Results and Discussion

For both bromoacetonitrile^[13] and trifluoromethylation^[15], the mechanism has been described to first involve photocatalyst quenching to reduce the precursor (bromoacetonitrile or CF₃SO₂Cl) to generate radical. The electron deficient radical then adds to indole, and the indole is subsequently oxidized by single electron transfer back to the photocatalyst. The resulting cation is deprotonated to return aromaticity to afford the substituted product. This mechanism is supported experimentally by Stern-Volmer quenching studies where either CF₃SO₂Cl or bromoacetonitrile quench photocatalyst emission while indole does not. To understand the C2 selectivity of these reactions, we performed calculations^[21] on the radical addition to either the indole 2 or 3 position using M06-2X density functionals^[22] with the CC-PVQZ(-g) basis set.^[23] The M06-2X method with large basis set has been shown to perform well on prediction of transition state geometries and the energetics of radicals.^[24] Since secondary derivatives cannot be calculated with an f-basis function, the changes of free energies ΔG_f° at the M06-2X/CC-PVQZ(-g) level were estimated by frequency calculations using CC-PVTZ(-f) basis function on the M06-2X/CC-PVQZ(-g) optimized structures. For comparison we also provide calculations using B3LPY//6-311G**^[25] in Table 1.

Table 1.

Geometries and energetics of calculated structures at B3LPY/6–311G** or [M06-2X/CC-PVQZ(-G)].

Transition State		C … C distance (A)	ΔE‡ (kcal/mol)	ΔGf° (kcal/mol)	Radical Intermediate		CC bond length (Å)	ΔE (kcal/mol)	ΔGf° (kcal/mol)
	2a	2.179 [2.162]	10.8 [11.1]	11.7 [11.2]		2b	1.556 [1.539]	−5.8 [−8.9]	−2.9 [−7.4]
	3a	2.098 [2.106]	14.7 [13.8]	16.5 [14.1]		3b	1.560 [1.542]	1.6 [−3.0]	4.1 [−1.9]
	4a	2.140 [2.149]	14.7 [3.8]	15.5 [14.2]		4b	1.570 [1.553]	−0.9 [−6.4]	2.0 [−4.3]
	5a	2.063 [2.070]	17.3 [15.9]	18.4 [16.4]		5b	1.576 [1.558]	6.5 [1.7]	89 [3.5]
	6a	2.099 [2.105]	15.7 [14.5]	16.6 [15.0]		6b	1.572 [1.556]	2.8 [−2.0]	5.4 [−0.1]
	7a	2.115 [2.124]	12.7 [11.2]	14 3 [11.8]		7b	1.574 [1.555]	−0.6 [−6.7]	2.7 [−4.5]
	2e	2.339 [2.262]	2.3 [4.5]	3.5 [4.2]		2f	1.533 [1.524]	−26.1 [−26.0]	−22.0 [−26.1]
	3e	2.246 [2.203]	3.5 [5.3]	5.2 [5.0]		3f	1.528 [1.520]	−20.8 [−22.1]	−16.7 [−22.2]
	4e	2.285 [2.239]	3.8 [5.0]	5.6 [4.8]		4f	1.534 [1.526]	−23.3 [−25.2]	−19.0 [−24.5]
	5e	2.194 [2.159]	5.6 [6.8]	7.5 [6.5]		5f	1.537 [1.528]	−15.6 [−17.2]	−11.9 [−16.7]
	6e	2.238 [2.194]	4.6 [5.8]	6.3 [5.8]		6f	1.535 [1.526]	−19.3 [−20.7]	−15.4 [−20.1]
	7e	2.266 [2.217]	3.9 [5.6]	5.7 [5.1]		7f	1.533 [1.523]	−21.8 [−23.7]	−17.3 [−23.1]

The transition state for addition of •CH₂CN to C2 was lower by 2.9 kcal/mol via 2a with an energy barrier ΔG_f°of 11.1 kcal/mol (2a) compared to ΔG_f° 14.1 kcal/mol for C3 addition (3a, Figure 3). The resulting radical addition intermediate was significantly lower in energy for addition to C2 (2b, −7.4 kcal/mol) vs C3 (3b, −1.9 kcal/mol, Figure 3). For comparison, the Hansen group has computed the energy barrier for the addition of malonate radical to N-methyl indole C2 to be 4.5 kcal/mol lower than for addition at C3.^[26]

Figure 3.

Transition state and intermediate energies (estimated ΔG_f°) for the addition of acetonitrile radical to indole at either the C2 (2) or C3 (3) position using the M06-2X/CC-PVQZ(-g) method.

Shown in Table 1, the energetics of the indole addition transition states for positions 4 through 7 were also computed. The resulting dienyl radicals 4b and 7b were lower in energy than for addition at C3. Addition to the 5- and 6-positions was found to be slightly endothermic. The transition state energies follow the trend for radical intermediates 2a<<7a<3a, 4a<6a<5a. The lower energy found for reaction at the 2-position correlates with the experimentally observed selectivity for C2 functionalization.

Trifluormethylation of indoles at the 2-position has been well described using many different sources of •CF₃ in the literature^[15–20], and in most cases, C3 alkylation is observed as a minor product. Calculations at the M06-2X/CC-PVQZ(-g) level of theory reveal a small difference in energy barrier between addition at the 2-position (2e) compared with the 3-position (3e) (ΔG_f° 4.2 kcal/mol vs 5.0 kcal/mol for reaction, respectively). Of note is the reduced energy barrier as compared to •CH₂CN (11.2–14.1 kcal/mol). Shown in Figure 4, the transition state is earlier with a C…C bond length of 2.26 Å (•CF₃ addition, 2e) vs 2.16 Å (•CH₂CN addition, 2a). The resulting C2 adduct is lowest in energy (2f, –26.1 kcal/mol) compared to C3 (3f, –22.2 kcal/mol). Exergonic trifluoromethyl radical addition to C3 of 2-phenyl-N-methyl-indole has also been described (–8.9 kcal/mol).^[27] Houk and co-workers have also computationally studied the •CF₃ addition to enolates. They concluded that addition is near barrierless, as transition states could not be optimized and the product formation is exergonic (–24.9 kcal/mol).^[28] Addition of •CF₃ to other indole positions was found to have a slightly higher energy barrier (4.8–6.5 kcal/mol) (Table 1, 4e–7e). The lower barriers and small difference (0.8 kcal/mol) in transition state energies for 2e and 3e explain the formation of the experimentally observed C3 regioisomer (C2: C3, 4:1).^[15],[20] The radical intermediate 2f (–26.1 kcal/mol) is lowest in energy followed by 4f (–24.5 kcal/mol), 7f (–23.1 kcal/mol) and then 3f (–22.2 kcal/mol).

Figure 4.

Transition state and estimated intermediate energies (estimated ΔG_f°) for the addition of trifluoromethyl radical to indole at either the C2 (2) or C3 (3) position using the M06-2X/CC-PVQZ(-g) method.

The next step in the reaction involves oxidation of the radical intermediate to cation Figure 5. This process is more endothermic for C2 adduct 2b to 2c compared to 3b to 3c. Deprotonation then returns aromaticity and reaction products 2d and 3d. Loss of proton to return aromaticity in an analogous malonate system has been calculated to be exothermic.^[26]

Figure 5.

Oxidation of radical intermediates to M06-2X/CC-PVQZ(-g) fully optimized cations followed by deprotonation and rearomatization to provide indole addition products 2d,h or 3d,h.

We also calculated the spin density and relative energies in solution for intermediates 2 and 3 (Figure 6). In structures 3b and 3f radical character is mostly localized on C2 (3b, 0.71; 3f, 0.71). In structures 2b and 2f the radical is conjugated to the aromatic system and the radical spin density is delocalized from C3 (2b, 0.56; 2f, 0.57). Radicals 2b and 2f were found to be lower in energy than 3b,f in either gas, or using Poisson Boltzmann Finite element method (PBF)^[29] with water or DMSO as solvent. Delocalization of the radical spin density correlates with the benzylic stabilization^[30] of intermediates 2 vs 3.

Figure 6.

Spin density and relative energies in solution for 2b,f and 3b,f using the M06-2X/CC-PVQZ(-g) method.

For comparison trifluoromethyl cations are lower in energy when substituted at the 3-position. Shown in Figure 7 benzylic cation 2g is 7.0 kcal/mol higher in energy than aromatic iminium 3g. It requires breaking aromaticity for the amine lone pair electrons in 2g to stabilize the cation resulting from electrophilic addition at the 2-position. This is in contrast to the radicals 2b/2f which are lower in energy than 3b/3f. Further the C2 product 2h is higher in energy than the C3 product 3h by 3.6 kcal/mol. For comparison indole protonation with addition of H⁺ to C3 (Figure 7B, 9) is 7.1 kcal/mol lower in energy than the corresponding H⁺ addition to C2 (8). Placing the cation in the benzylic position is less advantageous than the cation stabilized by the nitrogen lone pair (iminium 9).

Figure 7.

A) Comparison of C2 or C3 trifluoromethyl substituted cations. B) Comparison of C2 or C3 protonation using the M06-2X/CC-PVQZ(-g) method.

Conclusions

In conclusion, we have performed density functional theory calculations that explain the regioselectivity in the addition of electron deficient radicals to indoles. Despite greater nucleophilicity at the 3-position of indoles, the barrier is lower for radical additions at the 2-position. This is in part due to the increased stability of the radical addition products for 2 versus 3. The very low reaction barriers for •CF₃ additions also explain why, depending on the synthetic method, regio-isomers are reported, whereas for •CH₂CN, the higher reaction barrier and greater separation in transition state energies imparts greater selectivity.

Experimental Section

Optimized structures in SDF format are provided in the supporting information. Jaguar (https://www.schrodinger.com/jaguar) was used for all calculations^[21] using M06-2X or B3LYP/6–311G** density functionals with the CC-PVQZ(-g) basis set.^[22] Solvation Model using Poisson-Boltzmann Finite element method (PBF) in Jaguar was applied to model the solvent effect in water or DMSO.^[29] Free energies ΔG_f° at the M06-2X/CC-PVQZ(-g) level were estimated by frequency calculations using CC-PVTZ(-f) basis function on the M06-2X/CC-PVQZ(-g) optimized structures.