Radical-Mediated Regiodivergent C(sp³)–H Functionalization of N-Substituted Indolines via Enzymatic Carbene Transfer

Summary

Indolines are ubiquitous structural motifs occurring in pharmaceuticals and natural products. Here, we report a strategy for regio- and stereoselective C(sp³)–H functionalization of N-substituted indolines via carbene transfer chemistry mediated by engineered CYP119-based catalysts. These systems offer high enantioselectivity, high catalytic efficiency, as well as regiodivergent selectivity, furnishing an efficient and convenient route for diversification of these important scaffolds via direct C(sp³)–H functionalization. Selective functionalization of exocyclic C(sp³)–H bond in N-methyl indolines was also achieved and a biocatalytic cascade combining enzyme-mediated α- and β-C(sp³)–H functionalization yielded a polycyclic indoline-containing motif found in drugs. Mechanistic and computational studies support a radical-mediated C–H functionalization pathway and provide insights into protein-mediated regiodivergent selectivity. Altogether, this work offers a direct and tunable strategy to access functionalized indolines as key building blocks for medicinal chemistry and natural product synthesis and provides first insights into the mechanism of P450-catalyzed C(sp³)–H carbene insertion.

Introduction

Given the privileged nature of functionalized indolines in pharmaceuticals and their ubiquitous presence within biological systems (Figure 1a),^1–4 many efforts have been made toward the synthesis of such scaffolds. Existing methodologies largely rely on either reduction of the indole counterpart^5–7 or catalytic intramolecular cyclization reactions^8–11 to produce functionalized indolines. Comparatively, direct C–H functionalization offers an attractive approach for the synthesis and diversification of indoline-based scaffolds, although methods for their regio- and stereoselective direct C–H functionalization have been elusive.¹²

Figure 1. Biological significance of indolines and enzyme-catalyzed regiodivergent C(sp³)-H functionalization.

(a) Representative indoline containing drugs and natural products. (b) P450-catalyzed Regiodivergent C(sp3)–H hydroxylation. (c) Engineered P450-catalyzed regiodivergent C(sp³)–H alkylation.

Biocatalysis has attracted increasing attention as a strategy to address important challenges in chemical synthesis.^13–21 In the area of C–H functionalization, various classes of oxidizing enzymes, including cytochromes P450, unspecific peroxygenases, flavin-dependent halogenases and non-heme iron-dependent dioxygenases and halogenases, have proven useful for the selective functionalization of aromatic and aliphatic C–H bonds in small molecules.^{13, 14, 16, 18, 21–29} Furthermore, through protein engineering, regiodivergent selectivity has been achieved, for example, for the halogenation of aromatic compounds using evolved flavin-dependent halogenases^{22, 30, 31} or for late-stage C(sp³)-H hydroxylation of complex natural products using engineered P450 enzymes and halogenases (Figure 1B),^32–39 thus creating new opportunities for the diversification and/or chemoenzymatic synthesis of these molecules. More recently, important progress has been made also in the development of biological catalysts for C–H functionalization via ‘non-native’ chemistry such as carbene transfer catalysis.^40–50 Among them, iron-based cytochrome P450s have constituted attractive systems for achieving selective C(sp³)–H functionalization in the presence of diazoester-based carbene donor reagents.^{46, 47, 50–52} Despite this progress, and in stark contrast to the scope and regiodivergent selectivity achieved via these enzymes’ native reactivity (Figure 1B), these systems and methodologies have been largely limited to a narrow set of substrates and to functionalization of a single C(sp³)–H site in the target substrate.^{45–47, 50–52}

Herein, we describe the development of a biocatalytic strategy for the direct C(sp³)–H functionalization of indolines, a structural subclass found in many bioactive molecules, via P450-catalyzed carbene transfer. In addition to being compatible with a broad range of indoline substrates and diazo reagents, this methodology is shown to allow regiodivergent access to three distinct C(sp³)–H bonds in the substrate in a highly regio- and enantioselective manner (Figure 1c). As such, this strategy can provide efficient and direct strategy to the synthesis and diversification of these medicinally important scaffolds, whose synthesis is neither straightforward⁵³ not readily accessible using metal-catalyzed carbene transfer chemistry. The synthetic utility of the present biocatalysts and methodology is further exemplified through the synthesis of a polycyclic indoline-based core structure akin to that found in many pharmacologically active molecules.^{3, 54–56} Mechanistic studies provide first-time insights into the mechanism of the present reaction and hemoprotein-catalyzed C(sp³)–H carbene insertion.

Results and Discussion

CYP119 biocatalyst for α-C–H functionalization of N-Methyl Indoline with EDA

Motivated by our recent success in developing engineered CYP119 variants for C(sp³)–H bond functionalization of N-aryl-pyrrolidines via carbene transfer with diazoacetone and ethyl diazoacetate,⁵⁰ we sought to develop a biocatalytic method for the C(sp³)–H functionalization of N-substituted indolines, which are privileged scaffolds in medicinal chemistry. Of note, this transformation poses a unique challenge in terms of regioselectivity given the presence of two C(sp³)–H sites with similar reactivities, i.e., at the α and β position with respect to the nitrogen atom.⁵⁷ To this end, we began our investigation by screening a small panel (~100) of engineered CYP119 variants derived from our previous work,⁵⁰ including CYP119 (F153G, A209G, T213G, V254A, C317S) (called CYP119–137). The latter variant features an expanded active site as a result of four space-creating active site mutations and it was previously found to exhibit a pronounced substrate promiscuity toward C–H functionalization of N-aryl-pyrrolidines via carbene transfer. These CYP119-based variants also harbor a non-native axial serine ligation, a mutation shown to be beneficial for catalysis of non-native carbene transfer reactions in this⁵⁰ and other P450 scaffolds.⁵⁸

The C(sp³)–H functionalization of N-methyl indoline (1a) in the presence of ethyl diazoacetate (2a, EDA) was investigated as model reaction for this work (Table 1). While wild-type CYP119 shows no activity, the promiscuous variant CYP119–137 was found capable of converting N-methyl indoline (1a) in 17% yield (210 turnovers or TON) to yield a mixture of the α-amine C(sp³)–H functionalization product 3a and β-C(sp³)–H functionalization product 4a in approximately 2:1 ratio (Table 1, Entry 4). Along with the desired C–H functionalization products, the CYP119–137-catalyzed reaction was accompanied with the formation of various demethylation and desaturation/N-H insertion by-products (SI Table S5), which have implications with respect to the mechanism of this transformation as discussed later. Importantly, the isolated cofactor hemin along with various organometallic carbene transfer catalysts⁵⁹, such as Rh₂(OAc)₄, Ru(BPY)₂, Co(TPP), Cu(OTf)₂, and Fe(TPP), showed no product formation (SI Table S1), highlighting the peculiar role of the protein matrix in enhancing the enzyme’s reactivity toward this challenging transformation. In addition, the ability of CYP119–137 to target each of the three distinct C(sp³)–H bonds in the indoline substrate (i.e., benzylic and the endo and exocyclic α-amino C-H bonds) held promise toward tuning the enzyme’s regioselectivity via protein engineering.

Table 1.

Intermolecular C–H functionalization of N-methyl Indoline (1a) with EDA (2a) using hemoprotein CYP119 and variants thereof.^[a]

Entry	Catalyst	Yield (3a)[b]	3a:4a	TON (3a)[c]	e.r. (3a)[d]
1	Hemin	0	0	-	-
2	‘Empty’ E. coli cells	0	0	-	-
3	CYP119 (WT)	0	0	-	-
4	[CYP119–137] CYP119 (F153A, A209G, T213G, V254A, C317S)	17%	66:34	210	53:47
5	CYP119 (F153G, T213A, V254I, C317S)	26%	70:30	840	58:42
6	CYP119 (F153G, T213A, V254L, C317S)	20%	76:24	1,150	60:40
7	CYP119 (F153G, T213A, V254F, C317S)	73%	87:13	4,180	62:38
8	CYP119 (F153G, T213A, V254Y, C317S)	66%	84:16	3,210	78:12
9	[CYP119–168] CYP119 (F153G, T213A, V254W, C317S)	92%	92:8	5,270	91:9
10[e]	[CYP119–168] CYP119 (F153G, T213A, V254W, C317S)	42%	92:8	8,930	91:9
11[f]	[CYP119–168] CYP119 (F153G, T213A, V254W, C317S)	92%	92:8	460	91:9
12[g]	[CYP119–168] CYP119 (F153G, T213A, V254W, C317S)	92%	92:8	460	91:9

^[a]Standard reaction conditions: protein expressing C41(DE3) E. coli cells (OD₆₀₀ = 60), 10 mM 1a, 80 mM EDA (2a), in KPi buffer (50 mM, pH 7), room temperature, 16 hours, anaerobic chamber.

^[b]Assay yields as determined by GC using calibration curves with isolated product 3a.

^[c]TON for product 3a as calculated based on the protein concentration measured in cell lysate using the CO-binding assay.

^[d]Enantiomeric ratio (e.r.) of the major product as determined by chiral HPLC.

^[e]OD₆₀₀ = 15.

^[f]Using 20 μM purified protein and 10 mM Na₂S₂O₄.

^[g]Using 20 μM lyophilized purified protein and 10 mM Na₂S₂O₄. N.d. = not determined.

Encouraged by these results, we aimed to identify CYP119 catalysts capable of favoring the formation of the α-amine C(sp³)–H functionalization product 3a with higher regioselectivity as well as higher chemoselectivity against formation of the undesired demethylation/unsaturation byproducts. To this end, we extended our screening to an in-house library of CYP119-derived variants generated in previous evolution campains⁵⁰, targeting the partial mutagenesis of active site residues F153, L205, A209, and V254 using (mostly) apolar amino acids of variable size (Ala, Phe, Ile, Leu, Pro, Ser, Thr, Val). The enzyme variants were expressed in E. coli C41(DE3) and screened as whole cell reactions in multi-well plates. These experiments revealed a set of structurally related variants, i.e., CYP119 (F153G, T213A, V254X, C317S, where X is Ile, Leu or Phe), that show a clear beneficial effect of increased steric bulk at the level of position 254 (Phe>Leu>Ile) toward increasing product yield as well as favoring formation of the α-amine C(sp³)–H functionalization product 3a over 4a (Table 1, Entries 5–7). Among them, the V254F containing variant showing the highest levels of catalytic activity and regioselectivity among them (73% yield, 4,180 TON and 87:13 r.r.; Table 1, Entry 7; Figure 2B). These variants also exhibited appreciable enantiopreference for formation of the S-enantiomer (20–24% ee; Table 1, Entry 5–7). Based on this insightful structure-activity data, we chose to investigate the effect of larger aromatic substituents, i.e., tyrosine and tryptophan, at the 254 position. Among them, and in line with the aforementioned trend, the V254W-containing variant CYP119 (F153G, T213A, V254W, C317S), referred to as CYP119–168, showed further enhanced catalytic activity (4,180 → 5,270 TON) and regioselectivity (87:13 → 92:8 r.r.) for formation of 3a. In addition, CYP119–168 showed improved enantioselectivity in the reaction (78:12 → 91:9 e.r.) and it catalyzes nearly quantitative conversion of the indoline substrate to the α-C–H functionalized product 3a with no formation of undesirable by-products (SI Table S5).

Figure 2. Directed evolution of CYP119 catalysts for regioselective C–H functionalization of N-methyl Indoline (1a) with EDA (2a).

(a) General scheme for C–H functionalization of N-methyl Indoline (1a) with EDA (2a) to form the α-functionalization product 3a and β-functionalization product 4a. (b) Reconstructed directed evolution of CYP119 catalysts for regioselective C–H functionalization of N-methyl Indoline (1a) with EDA (2a). Yields as determined under standard reaction conditions with EDA (Table 1). (c) X-ray crystal structure of CYP119 from Sulfolobus solfataricus (PDB 1IO7).¹⁰¹ Active site residues targeted for mutagenesis are highlighted in dark green, conserved active site residues are highlighted in light green, active site residue V254 is highlighted in dark blue, and the heme cofactor is shown in teal.

Additional experiments indicated that up to 8,900 total turnovers (TTN) could be obtained for the CYP119–168-catalyzed α-C–H functionalization reaction using whole cells under catalyst limiting conditions (OD₆₀₀ = 15) (Table 1, Entry 10). Furthermore, this reaction was determined to proceed with equally high yields and enantioselectivity using purified protein at 0.2 mol% as catalyst (Table 1, Entry 11) and similar results could be obtained using the same CYP119 variant in lyophilized form (Table 1, Entry 12), thus demonstrating the robust nature of this biocatalyst to lyophilization and long-term storage, which are desirable attributes for preparative scale and industrial applications.

Given our prior findings on the C–H carbene insertion reactivity of cobalt-substituted myoglobins using phthalan as the substrate⁵¹, we investigated the effect of a similar metal substitution in the present CYP119-based biocatalysts, which were prepared via recombinant expression in the presence of Co-ppIX.^51,60 Interestingly, all of these Co-substitued variants showed activity toward the C–H functionalization of N-methyl indoline (1a) with EDA (2a) (SI Table S2). Although the activity and regioselectivity of these metallo-substituted variants were only comparable or slightly inferior compared to those of the iron-containing counterparts, the functionality of these enzymes as carbene transferases is notable and it could prove useful for other types of non-native transformations.

C–H Functionalization of Indoline-based Substrates

Focusing on the best biocatalyst identified for α-C–H functionalization of 1a, CYP119–168, we next explored the substrate scope of this enzyme using a range of aryl and N-substituted indoline compounds (1a-1o). Notably, each of these substrates underwent α-C–H functionalization in the presence of EDA with excellent regio- and chemoselectivity, i.e., showing no formation of the potentially competing β-C–H functionalization product and byproducts, respectively. However, these reactions were characterized by variable yields and enantioselectivity (Figure 3c), indicating a certain degree of substrate specificity as observed in other P450-catalyzed native and non-native reactions.^{50, 61, 62} To address this limitation, we employed a substrate versus enzyme library approach, where a set of selected CYP119-derived carbene transferases from the CYP119–168 lineage and other generations were screened against the substrate panel in a high-throughput manner (~2,000 substrate/enzyme combinations). From these experiments, a subset of evolved CYP119 variants was shown to catalyze the α-C–H functionalization of each indoline-based substrates with high activity (2,900–6,540 TON) and good to excellent enantioselectivity (up to 96% ee; Figure 3a).

Figure 3. Activity and selectivity of CYP119 biocatalysts for α-C–H functionalization of indoline-based substrates derivatives (1a-o) with EDA.

(a) Yields, TON, and enantioselectivity were determined from whole cell reactions under standard reaction conditions with EDA as described in Table 1. Analytical yields were determined using GC/LC using calibration curve prepared with isolated products. Isolated yield are indicated in brackets. TON as determined based on P450 concentration in cell lysate. [a] Using 240 mM EDA (2a) at 40 °C. [b] Using 240 mM diazoacetone (2b) or diazoacetonitrile (2c). (b) Chemodivergent reactivity of CYP119 CYP119–253 and Rh₂(OAc)₄ (10 mol%) with 1-(but-3-en-1-yl)indoline (1j) and EDA (2a). (c) Heat map depicting intermolecular C–H functionalization of N-substituted Indolines (1a-n) with EDA (2a) using hemoprotein CYP119 variants. White dots (•) indicate the best variant for the corresponding substrate. Details about the variants are provided in SI Figure S1.

In particular, both electron-withdrawing (3c and 3f) and electron-donating (3d and 3g) substitutions on the aryl ring were well tolerated, resulting in high levels of activity and enantioselectivity (85–94% yield, 5,740–6,310 TON, 83:17–97:3 e.r.). Racemic 3-methyl substituted indoline derivative 1b was efficiently converted into the desired product 3b with both excellent diastereo- and enantioselectivity (99:1 d.r., 98:2 e.r.). This demonstrates the ability of the enzyme to induce kinetic resolution of racemic starting materials, a valuable trait in the context of stereoselective catalysis. In addition to methyl group as N-substituent, a variety of linear (3f-h, 3j), branched (3i), and cyclic (3k-n) alkyl groups were found to be tolerated at this position, with a slight decrease in yield and/or enantioselectivity for bulkier N-cycloalkyl substitutions (i.e., 3m-n vs. 3h-k). In contrast, N-aryl-indolines such as 3o were converted only with trace activity. Enzymatic conversion of N-alkyl indolines was initially affected by the low solubility of these compounds in aqueous media even in the presence of organic cosolvent (10% v/v EtOH). Improved yields for these reactions (e.g., 27→44% yield 3n) could be achieved by raising the reaction temperature to 40°C, a readily applicable condition thanks to the thermostability of the CYP119 variants. Notably, substrate 1j could be converted to the desired α-C–H functionalization product 3j with high regio- and chemoselectivity and without affecting the terminal olefinic group, highlighting the ability of the CYP119 based catalyst to favor the more challenging C–H carbene insertion reaction over cyclopropanation, unlike organometallic catalysts used for carbene transfer reactions (Figure 3b).

Whereas different CYP119 variants can be selected for optimal yield and enantioselectivity for each substrate (Figure 3a), each CYP119 variant is capable of catalyzing the desired reaction (Figure 3c). These factors show the generality of our engineered CYP119 variants towards the α-C–H functionalization of indoline-based derivatives. These results also revealed the role of steric bulk in promoting favorable binding conformations to target α-C–H functionalization compared to β-C–H functionalization. In the presence of bulkier N-alkyl substitutions, no β-C–H functionalization product is formed (3f-o). We also noted an inverse correlation between steric bulk in the substrate and in the catalyst active site, as catalysts that contain bulkier residues within the active site tend to exhibit lower activity toward bulkier substrates. For example, CYP119–168 and CYP119–156, which contain a V254W and A209W mutation, showed basal activity towards substrate 1n, bearing a N-cyclohexyl moiety.

Finally, efficient and selective α-C–H functionalization of N-methyl indoline (1a) could be achieved using both diazoacetone (2b) and diazoacetonitrile (2c) as carbene donors, yielding the respective products (6a and 7a) in quantitative yields (97–99%) and with high catalytic activity (>6,000 TON), regioselectivity (100:0 r.r.) and enantioselectivity (92–96% ee) (Figure 3a). Altogether, these results demonstrate the generality of this CYP119-based methodology for α-C–H functionalization of a broad range of N-substituted indolines with high catalytic efficiency and selectivity.

Regiodivergent CYP119 catalysts for indoline C–H functionalization

During initial catalyst development for achieving selective α-C–H functionalization of N-methyl-indoline, we noted that early generations of enzymes bearing more open active sites showed appreciable regioselective toward formation of the β-C–H functionalization product (4a) (e.g., CYP119–137: 66:34 3a:4a; Table 1, Entry 4). Building upon this finding, we sought to develop a regiocomplementary catalyst capable of selectively targeting the β-position (Figure 4a). To this end, we selected variants from the in-house CYP119 library that contained bulky residues (Phe, Tyr, Trp) at positions L205, L69, and A205, which are located on the opposite site of the active site compared to the Val254 residue shown to be instrumental in tuning regioselectivity to the α-C–H functionalization product (Figure 2b–c).

Figure 4. CYP119-catalyzed regiodivergent C–H functionalization of N-methyl indoline (1a) with EDA (2a).

(a) Regiodivergent pathway for C–H functionalization of N-methyl Indoline (1a) with EDA to form α-alkylated product 3a and β–alkylated product 4a. (b-c) Active site views of representative Rosetta models for CYP119–168 in complex with DFT-calculated α-pro(S)-C–H transition state (b) and of CYP119–235 in complex with DFT-calculated β-pro(R)-C–H transition state (c). The heme cofactor is shown as line model, while mutated active site residues (green), N-methyl indoline (orange) and heme-bound carbene (blue) are shown as stick models. See SI Figure S10 for additional data. (d) Benzylic and N-methyl C–H functionalization products. Yields, TON, and enantioselectivity were determined from whole cell reactions under standard reaction conditions as described in Table 1. (e) Secondary functionalization pathway for the intramolecular C–H functionalization of diazo tethered N-methyl indole (8) into product 9a.

Through this approach, we identified CYP119 (F153G, L205W, T213A, V254A, C317S), named CYP119–235, as an efficient and selective biocatalyst for functionalization of the β-C–H position in N-methyl-indoline (1a), producing 4a in high yield and TON (78% yield, 5,380 TON) as well as high enantio- and regioselectivity (94:6 e.r., 16:84 3a:4a; Figure 4d). As anticipated, the active site mutations in CYP119–235 include a bulky substitution (i.e., L205W) (Figure 4c) on the opposite side of the enzyme active site compared to a similar bulky substitution (i.e., V254W) harbored by the α-C–H selective variant CYP119–168 (Figure 4b).

Given the value of 2,3-difunctionalized indolines in medicinal chemistry (Figure 1) and our success in the kinetic resolution of racemic 1b with 96% ee and 98% de (3b, Figure 3a), we challenged the β-C–H selective variant CYP119–235 with racemic N-methyl-2-methyl-indoline (1p). Importantly, substitution at the α-position was tolerated by the enzyme and the racemic substrate could be converted to the optically active product 4b in good diastereomeric and enantiomeric excess (67:33 d.r., 79:21 e.r.), albeit low efficiency (12% yield). Using a different CYP119 variant (CYP119–137), this transformation could be carried out with further increased enantio- and diastereoselectivity (92:8 e.r., 81:19 d.r.), illustrating the potential value of this biocatalytic systems for kinetic resolution applications.

Upon exploration of the evolved variant CYP119–235, we observed that the regioselectivity of this enzyme could be directed to the functionalization of the N-methyl C–H bond in the presence of substitutions at the C₅ (5e) and C₆ (5f) position on the aryl ring (Figure 4d). These reactions were found to proceed with full regiocontrol and excellent catalytic activity (89–92% yield, 6,140–6,340 TON), complementing the scope of the α- and β-C–H functionalization reactions catalyzed by the CYP119 catalysts. Altogether, these results demonstrated the capability of the present methodology to target as many as three different C(sp³)–H sites in a substrate for C–H functionalization via carbene transfer, a feature unprecedented for carbene transferases and rarely achieved with synthetic carbene transfer catalysts.^{63, 64}

Polycyclic Indolines via Tandem Enzyme-Catalyzed C(sp³)-H Carbene Insertion

As described above, our investigation of the substrate scope of the engineered CYP119 variants established that both α-C–H and β-C–H selective biocatalysts can tolerate substitutions in the adjacent position in the indoline substrate targeted for functionalization (i.e., 3b and 4b). In addition, with product 6a, we established that diazoketones are accepted as carbene donors by these catalysts. Based on these results, we envisioned the possibility to construct a tricyclic indoline-based scaffold via a tandem CYP119-mediated C–H functionalization strategy, in which the enzymatic β-C–H insertion product 4a is converted into the corresponding diazo ketone (8) and then subjected to an intramolecular C–H insertion to form the desired polycyclic compound 9a (Figure 4e). The latter is akin to the indoline-based core structure found in various natural products and pharmaceuticals (Figure 1a). Toward this goal, we performed a preparative scale (500 mg, 3.75 mmol) reaction of N-methyl indoline (1a) in the presence of EDA (2a) was carried out using the β-C–H selective CYP119 variant CYP119–235, to afford 4a in 67% isolated yield. The ester group of the enantioenriched 4a was then chemically converted to the respective diazoketone 8 in 45% yield over 3 steps using established chemistry (see SI for details). Gratifyingly, the diazoketone intermediate 8 could be then effectively cyclized by the α-C–H selective variant, CYP119–168, to afford the desired polycyclic product 9a in high yield (64%) and high enantio- and diastereomeric excess (>500:1 d.r.; 97:3 e.r.; Figure 4e). These results highlight the utility of the engineered CYP119 library for the synthesis of biologically relevant synthons.

Mechanistic Insights into CYP119-Catalyzed Indoline C–H Functionalization

As noted earlier, the C–H functionalization reaction with N-methyl-indoline and EDA catalyzed by the early, unoptimized CYP119 catalysts is accompanied by the formation of three distinct by-products, which were determined to correspond to N-methyl-indole (1a-DS), C₃-functionalized N-methyl indole (4a-DS), and the N-alkylated-indoline 13a (SI Table S5, Entry 2). While these side-reactions could be suppressed in the presence of the optimized biocatalysts and reaction conditions, they bear important mechanistic implications. Indeed, as summarized in Figure 5, all of these side products can be explained based on desaturation reactions involving substrate-derived radical intermediates, which suggests a radical, stepwise mechanism for the present C–H carbene insertion reaction also in line with the results from computational analyses described further below.

Figure 5.

Proposed stepwise radical-mediated mechanism for the C–H functionalization of N-methyl indoline (1a) in the presence of engineered CYP119 catalysts.

For the productive pathway leading to the C–H functionalized products, the reactive heme-carbene species derived from reaction of ferrous protein with EDA is proposed to abstract a hydrogen atom from either the α- or β-C–H site with respect to the indoline N-H, with regioselectivity being controlled by the enzyme’s active site configuration (vide infra) and resulting in the C-centered radical intermediates IM_I or IM_II, respectively. Radical rebound with the iron-carbenoid species yields the corresponding C–H functionalization products (e.g., 3a or 4a). For the side-reaction leading to the desaturation product 1a-DS, we envision the radical intermediate IM_I or IM_II may undergo radical polar crossover (RPC), e.g., via single-electron transfer to the heme cofactor, to give the respective carbocation IM_IV or IM_V, which upon deprotonation, yield N-methyl-indole 1a-DS (Figure 5). Alternatively, the latter can be produced from the radical intermediate IM_I (or IM_II) via a second HAA mediated by the Fe(III)-alkyl species (vide infra). Regardless of the nature of this step, the ensuing desaturation product N-methyl-indole (1a-DS) can be then converted by the enzyme to 4a-DS via a indole C(3)-H functionalization reaction with EDA akin to that previously reported by our group⁴⁴ and others^{65, 66} for other engineered hemoproteins.

On the other hand, formation of the N-alkylated-indoline product 13a can be rationalized based on a first step involving enzymatic N-demethylation of N-methyl-indoline 1a, followed by N–H carbene insertion of the resulting indoline to give 13a (Figure 5). Consistent with our results with 5e and 5f, hydrogen atom abstraction at the level of N-methyl group is also accessible to this biocatalyst, resulting in the carbon-based radical IM_III which, upon RPC, is expected to produce the iminium intermediate (IM_VI). The latter can then undergo hydrolysis to form indoline, which can then give rise to 13a via N–H carbene insertion with EDA, a known reaction for engineered hemoproteins.^67–69

Various lines of experimental evidence support the proposed reaction pathways. Since the model substrate N-methyl-indoline was found to be susceptible to (slow) desaturation to indole in the presence of air and further modification of the CYP119 product could occur by action of other enzymes in whole cell reactions, control experiments were first performed to confirm the enzymatic origin of the observed side-products. To this end, time-course experiments were carried out in the presence of N-cyclopropyl-indole (1k), which is stable toward oxidative desaturation, and the unselective variant CYP119–282 (in purified form) as the catalyst. Under catalytic (air-free) conditions, formation of the C–H insertion product 3k is accompanied by accumulation of desaturated by-products 1k-DS and 4k-DS in approximately 73:9:18 ratio over 90 minutes, with corresponding initial formation rates (TOF) of 28, 5.7, and 2.5 turnovers/minute for 3k, 1k-DS, and 4k-DS, respectively (SI Figure S2). In contrast, no formation of either desaturation byproduct was observed in the absence of the enzyme or EDA within the same conditions (SI Figure S2), clearly indicating that these species are enzymatic products. To further investigate the sequence of reactions leading to the C₃-functionalized N-alkyl-indole product (i.e., 4a-DS from 1a or 4k-DS from 1k), products 4a and 1a-DS were used as substrates under standard catalytic conditions in the presence of CYP119–137. While 4a was fully preserved in the reaction mixture, 1a-DS was consumed to give rise to the C₃-functionalized product 4a-DS (SI Figure S8). Although the mechanism of this step was not investigated, previous studies with engineered myoglobin support a stepwise mechanism involving a zwitterionic intermediate.^{44, 70}

N-Demethylation of N-methyl-indoline via the proposed mechanism in Figure 5 implies the release of formaldehyde as byproduct. To test this, the enzymatic reaction mixture was added with the formaldehyde trapping agent O-(2,3,4,5,6-pentafluorobenzyl)hydroxylamine (PFBHA, 14), followed by GC analysis to detect the corresponding formaldehyde adduct (oxime). As anticipated, detectable amount of the PFBHA-derived oxime was detected in this reaction and this species was found to be proportional to the concentration of the demethylated/N-alkylated product 13a (SI Figure S7). In separate experiments, we further determined that indoline is readily converted by the enzyme to the N-alkylated product 13a under standard catalytic conditions in the presence of EDA (SI Figure S8). Collectively, these results support the proposed mechanistic pathways for formation of the experimentally observed products and byproducts in this reaction (Figure 5).

While the application of reducing conditions (i.e., sodium dithionite or intracellular environment) in the current protocol entails the involvement of ferrous hemoprotein as carbene transfer catalyst, we further explored the importance of the redox state of the protein in this reaction, also in light of recent reports on the activity of iron(III)-metalloporphyrins for carbene transfer reactions.⁷¹ In the absence of the reductant, the CYP119–168-catalyzed reaction with 1a and EDA proceeds with reduced catalytic activity (45→32% yield; SI Table S6, Entry 1–2), but identical regio- and enantioselectivity as compared to that under reducing conditions. Further experiments were conducted to discern whether the former activity stems from the ferric protein or if the hemoprotein is reduced in situ by the diazo compound, as previously observed by our group for certain axial substituted myoglobin-based carbene transferases.⁷² Accordingly, the same reactions were carried out in the presence of CO-saturated buffer, with CO being expected to bind with high affinity to only the ferrous form of the protein, thus inhibiting its reactivity. Under these conditions, with or without an external reductant, no carbene transfer activity was observed (SI Table S6, Entry 3–4), indicating that ferrous CYP119 is the catalytic species responsible for formation of the C–H functionalization products.

Finally, further mechanistic insights were gained using the deuterium labeled substrate N-methyl-2-D₂-indoline (1a-D₂). In the presence of the α-C–H selective variant CYP119–168, enzymatic transformation of 1a-D₂ in the presence of EDA (2a) showed a notable change in the overall product distribution (vs. reaction with 1a) to favor formation of the β-C–H functionalized product 4a-D₂, along with the N-methyl-indole byproducts 1a-DS-D₂ and 4a-DS-D₂ (92:8 →68:19:4:9; Figure 6). Insightfully, and in line with the overall mechanism of Figure 5, these results indicate that (i) the H atom abstraction (HAA) step is regioselectivity determining and (ii) that the carbon radical intermediate derived from β-HAA represents a pathway branching point toward formation of the desaturation byproduct (indole). Indeed, as the energy barrier for α-HAA is increased by the H→D substitution, it is conceivable that the reaction pathway is partially diverged to favor the β-HAA manifold, thereby leading to 4a-D₂ via the productive pathway, or to 1a-DS-D₂, via the unproductive, side-reaction. Since no desaturation byproduct is observed in CYP119–168-catalyzed reaction with 1a, it can be further evinced that the desaturation products 1a-DS-D₂ (and 4a-DS-D₂) primarily derive from the β-HAA pathway (vs. α-HAA pathway), possibly due to a slower radical rebound step and/or more favorable conformation for RPC.

Figure 6. Kinetic isotope effect experiments.

(a) Intermolecular non-competitive KIE experiment with N-methylindoline-2,2-d₂ (1a-D₂) and EDA (2a). (b) Intermolecular competitive and non-competitive KIE experiments with N-phenylpyrrolidine (10a) or N-phenylpyrrolidine-2,2,5,5-d₄ (10-D₄) and EDA (2a) or DA (2b). (c) Intermolecular competitive and non-competitive KIE experiments with N-phenylpyrrolidine (10a) or N-phenylpyrrolidine-2,2,5,5-d₄ (10-D₄) and EDA (2a) or DA (2b). Elaborated KIE data in SI Figure S3 and S4.

It is worth noting that these desaturation reactions are reminiscent of those described for certain cytochromes P450 in non-classical oxidation reactions under aerobic conditions.^73–75 While the cationic mechanism proposed in Figure 5 is plausible based on computational studies of these analogous reactions,⁷⁶ an alternative desaturation mechanism entailing a second HAA on the radical intermediate IM_I or IM_II by the Fe(III)-alkyl species may be also operative, in analogy to desaturation reactions catalyzed by other metalloenzymes.⁷⁷ Further studies will be thus required to elucidate this aspect in more detail.

Mechanistic Studies

To further illuminate the mechanism of the present P450-catalyzed C–H carbene insertion, we investigated our previously reported C(sp³)–H functionalization reaction of aryl pyrrolidines with EDA and diazoacetone, as these and the present reaction share the same catalytic system (i.e., serine-ligated CYP119).⁵⁰ Importantly, and unlike the indoline reaction, kinetic isotope effect (KIE) values can be more readily measured for the pyrrolidine reactions due to the absence of side products and pathway branching points. Accordingly, KIE values were calculated from both competitive and non-competitive intermolecular KIE experiments using deuterated and non-deuterated N-phenyl pyrrolidine (10a-D₄ and 10a) in the presence of both EDA (2a) and diazoacetone (2b) and in combination with the respective optimized CYP119 catalysts (CHI-EDA and CHI-DA). In both cases, a relatively large primary KIE value of 4.36–4.43 was measured for both competitive and non-competitive (parallel) kinetic experiments (Figure 6b/c). The similar values in both experimental settings indicate that the C–H bond cleavage step is part of the rate-determining step. In addition, the nature of the carbene donor (i.e., diazoester vs. diazoacetone), and thus of the corresponding heme-carbene, has no noticeable effect on the kinetic role of the HAA step. To further investigate the role of the non-native serine axial ligand, the same KIE experiments were also carried out using cysteine-ligated versions of the aforementioned CYP119 variants, i.e., CHI-EDA (S317C) and CHI-DA (S317C) (Figure 6b/c). Interestingly, much smaller KIE values were obtained in both cases compared to the serine-ligated counterparts. For the reaction with EDA, KIE values of 2.34–2.79 were obtained from competitive and non-competitive reactions against a value of 4.36–4.43 measured with CHI-EDA. Similarly, the reactions of the cysteine-ligated CHI-DA (S317C) enzyme with diazoacetone yielded moderate to no KIE of 1.39 (compet.) and 1.06 (non-compet.), against a much larger KIE of 4.37–4.42 observed for the serine-ligated counterpart. These differences indicate a notable change in the kinetic impact of C–H bond cleavage as a result of the change in heme axial coordination environment, further highlighting the often critical role of the axial ligand for influencing reactivity and the mechanism of hemoprotein-based carbene transferases.^78–80

A quantum chemical study was performed to further examine the mechanism of C–H carbene insertion catalyzed by the engineered P450s described here. DFT calculations were carried out using the previous method that accurately predicted various experimental structures and reactivities of heme carbenes,^81–87 using [Fe(Por)(MeO⁻)] to mimic the active site core part of the biocatalyst as done previously,^85–87 where Por is a non-substituted porphyrin and MeO⁻ is the model for the Ser ligand.^88–91 Since conformations and spin states may influence the reaction results,^{83, 84, 86, 92} we first conducted a detailed study of these effects involving both the substrate and heme catalyst to select the most favorable conformations and spin states of the species along the reaction pathways, as described in more detail in the Supporting Information.

As shown in Figure 7a, both concerted and stepwise mechanisms were investigated. The concerted reaction pathway is similar to that reported previously on C–H insertions catalyzed by a different iron porphyrin carbene⁸³ and it features a simultaneous hydride transfer (as evidenced by a significant negative charge transfer from substrate to carbene, −0.441 e), C–C’ bond formation, and Fe–C bond breaking. In contrast, the stepwise reaction pathway entails a hydrogen atom transfer (HAT) from the substrate to the heme carbene to form a carbon-center radical intermediate, followed by radical rebound to form the new C–C’ bond, while breaking the Fe–C bond (Figure 7b). In the present system, a hydrogen atom transfer feature can be seen from the increase in the C’–H₁ bond length from 1.095 Å in R₂ to 1.354 Å in TS₁ and then to 3.399 Å in Int, while C–H₁ bond length decreases from 1.338 Å in TS₁ to 1.086 Å in Int (See SI Table S20). The most favorable spin state calculated for TS₁ is the open-shell singlet, where the radical is equally shared between C (-0.496 e) and C’ (-0.445 e), whereas the hydrogen atom carries spin densities in the opposite direction (0.090 e), indicating partial transfer of the hydrogen atom. In this electronic state, Fe is in the ferric form with spin densities of +0.995 e (SI Table S22). In the intermediate Int, after donating the hydrogen atom, the substrate C’ has the radical with spin densities of 0.995 e, see SI Table S22). A subsequent radical rebound lead to the formation of the C–C’ bond in TS₂, which has relatively lower energy compared to the hydrogen atom transfer step (-3.77 kcal/mol, see Figure 7b). In this step, again both C and C’ show radical feature, but with opposite spin directions (-0.438 e and 0.641 e respectively), ready for a radical coupling to facilitate the formation of the final product. While partial C–C’ bond formation can be seen by its distance shortening from 4.436 Å in Int to 2.543 Å in TS₂, a concomitant Fe–C bond elongation of ~0.3 Å in this step also indicates partial cleavage of that bond, which proceeds to the final release of the product.

Figure 7. Computational analysis of concerted vs. radical-mediated pathways.

(a) Concerted and stepwise pathways for heme catalyzed C–H insertion of N-phenylpyrrolidine. Oval represents porphyrin. (b) Schematic free energy diagram for heme catalyzed C–H insertion. Spin densities of the transition states are shown. Atom color scheme: C-cyan, N-blue, O- red, H-grey; Fe, black. (spin density contour value: ± 0.004 au).

Overall, our calculations indicate that the stepwise reaction pathway exhibits a significantly lower energy barrier (>9 kcal/mol difference) compared to the concerted reaction pathway (See Figure 7b), suggesting that the stepwise radical mechanism is preferred. Furthermore, the calculated KIE values for the concerted C–H carbene insertion step via TS and for the first step of stepwise pathway (TS₁) are 2.14 and 4.63, respectively, the latter being closer to the experimentally determined KIE values of 4.3–4.6 for this reaction (Figure 6). To further understand the origin of the reactivity differences between these two mechanisms, the geometric data were examined in more detail. As seen in SI Table S11, the overall structures of TS and TS₁ are similar except that TS has a significantly longer Fe–L (1.987 Å) and shorter Fe–C (1.946 Å) bonds compared to TS₁ (Fe–L: 1.912Å, Fe–C: 1.982 Å). On the one hand, the shorter Fe–L distance in TS₁ results in a stronger trans effect that pushes the carbene moiety away from iron center (longer Fe–C bond), thus facilitating attack on the substrate. On the other hand, breaking of the shorter Fe–C bond (1.946 Å) in the concerted TS is associated with a higher energy cost compared to cleavage of the Fe–C bond in the radical pathway TS₂, which is longer (2.345 Å). Altogether, these structural features contribute to favor the stepwise radical mechanism for this biocatalytic transformation.

In addition, we computationally studied the rate-determining steps of the C–H insertion reaction with EDA and N-methyl indoline 1a as the substrate, using the most favorable spin states for concerted and radical pathways as derived from the analyses above. As shown in Table S19, these two transition states have barriers similar to those calculated above for the reaction with the pyrrolidine substrate 10a (within ~1 kcal/mol differences). Therefore, these results also showed preference for the stepwise radical reaction pathway in the presence of 1a as the substrate, featuring a lower barrier by 7.44 kcal/mol compared to the concerted pathway. Overall, these data further support a radical reaction mechanism for the C-H functionalization reaction investigated in this work.

Analysis of regio- and stereocontrol in the regiodivergent biocatalysts

Further computational studies were performed to gain insights into the origin of the divergent regioselectivity (i.e., α-C–H vs β-C–H) of the engineered CYP119–168 and CYP119–235 biocatalysts. To this end, the most energetically favorable binding pose of the transition states (TS) for the regio- and stereoselectivity-determining hydrogen atom abstraction (HAA) step were modeled in the two enzymes. Briefly, transition states corresponding to near-attack conformations (NACs) for the insertion of the heme-bound carbene into the pro-(S) and pro-(R) C–H bond at the α- and β-C–H site of the N-methyl indoline substrate were calculated and optimized by DFT (see Supporting Information for further details and results). These analyses showed that the α-regio-isomers TS are about 5 to 9 kcal·mol⁻¹ lower in energy than the β-regio-isomers TS, indicating a generally higher reactivity of the α-C–H site vs. β-C–H site, and no stereoselectivity, in the absence of the protein environment (SI Table S23). Models of the CYP119–168 and CYP119–235 variants were generated using Rosetta⁹³ based on available crystal structures of this enzyme in both ligand-free form and in complex with imidazole-based ligands (SI Figure S9). As done previously with other systems,⁹⁴ the DFT-optimized TS regio-, stereo- and conformational isomers were these superimposed onto the CYP119–168 and CYP119–235 structures and the resulting protein-TS complexes were optimized using Rosetta to obtain the corresponding energies (SI Table S24–S25; see Supporting Information for Rosetta modeling details).

Inspection of most representative, low-energy models of these complexes revealed that mutations F153G and T213A, which are shared by both variants, both contribute to significantly enlarge the active site cavity to accommodate the bulky TS complex. The location and orientation of the N-methyl indoline substrate, however, were found to differ drastically in the two enzymes. In the α-C–H selective variant CYP119-168, the indoline substrate occupies a space near residue Leu205, with an orientation that favors carbene attack (from the re-face) to the α-pro(S)-C–H bond (Figure 4B), consistent with the regio and stereoselectivity of this biocatalyst. In CYP119-235, this arrangement is disfavored by mutation L205W, which inserts a bulky indole ring into the cavity occupied by the indoline molecule in CYP119-168 (SI Figure S10C). Conversely, in the β-C–H selective variant CYP119-235, the N-methyl-indoline substrate occupies the cavity between Ala213 and Ala254 (Figure 4C), which is greatly enlarged by mutations T213A and V254A, respectively, compared to the wild-type enzyme. In this complex, the N-methyl-indoline ring is oriented so that carbene attack (from the si-face) is directed toward the β-pro(R)-C–H bond of the substrate (Figure 4C), in line with the selectivity of this variant. In CYP119-168, this arrangement is disfavored by the presence of a bulky Trp residue (vs. Ala in CYP119-235) in position 254 (SI Figure S10B).

Overall, these analyses reveal how mutation of Thr213 and of the ‘gating’ residue Phe153 (SI Figure S9) both contribute to make CYP119 active site more accessible to carbene/indoline complex, thus enhancing its non-native carbene transferase reactivity. On the other hand, modulation of steric encumbrance at the 254 and 205 positions, which are located at opposing sides of the heme pocket, plays a critical role toward dictating the α- vs. β-C–H selectivity in the C–H alkylation reaction catalyzed by the two regiodivergent biocatalysts. These findings also provide a rationale for explaining the experimentally observed beneficial effect of increasing steric bulk in position 254 toward favoring α-C–H regioselectivity and (S)-stereoselectivity (Trp > Tyr ≈ Phe >> Ile/Leu/Val; Table 1).

Conclusion

In summary, we have developed a new, efficient biocatalytic strategy for the enantioselective C(sp³)–H functionalization of indoline scaffolds via P450-mediated carbene transfer. This method is amenable to the transformation of variously substituted indolines with diazoacetate and its scope extends to include other carbene donor reagents, such as diazoacetonitrile and diazoketone. Importantly, this work also demonstrates the possibility of achieving regiodivergent selectivity in enzyme-catalyzed C–H carbene insertion, through the regio- and enantioselective functionalization of up to three distinct C(sp³)–H sites in an indoline substrate (Figure 1c). While regiodivergent P450-catalyzed C(sp³)–H oxidation has been previously achieved through the native monooxygenase reactivity of these enzymes,^{32, 33, 35, 37} regiodivergent C(sp³)–H functionalization via carbene transfer has represented a major challenge^{45–47, 50–52} and has been so far largely elusive to engineered biological catalysts. On the one hand, this capability can provide rapid access to enantioenriched indoline-derived constitutional isomers, which can be valuable building blocks for medicinal chemistry and whose synthesis would require multi-step syntheses.⁵³ On the other hand, as demonstrated through the chemoenzymatic synthesis of 9a, these regiodivergent carbene transferases can be leveraged to afford more complex, polycyclic indoline-based scaffolds akin to those found in various bioactive molecules.^{3, 54–56} Regiodivergent selectivity is also of particular interest in the construction of compound libraries for drug discovery campaigns.^{95, 96}

Our mechanistic studies provide valuable, first-time insights into the mechanism of hemoprotein-catalyzed carbene C(sp³)–H insertion, and collectively support the involvement of a radical, stepwise pathway, akin to the mechanism of native P450-catalyzed hydroxylation reactions⁹⁷ and non-native C–H amination reactions via nitrene transfer catalyzed by engineered P450s and other hemoproteins.^98–100 Our molecular modeling studies clarify the role of key active site mutations in controlling the α- vs. β-C–H selectivity and the enantiopreference of the two regiodivergent biocatalysts developed for this transformation. Overall, this work expands the methodological toolbox for realizing selective C(sp³)–H functionalization via enzyme-mediated carbene transfer. This advance, along with the insights gained into the productive and non-productive pathways in these reactions, and molecular basis of protein-controlled regioselectivity, are expected to guide future development of new and improved biocatalysts for this important class of transformations.

Experimental Procedures

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Rudi Fasan (rudi.fasan@utdallas.edu).

Materials availability

All reagents in this study are commercially available or can be prepared as described in the supplemental information.

Data and code availability

There is no dataset or code associated with this publication. All supporting data and findings are available in the manuscript or supplemental information.