Due to the ubiquity of C-H bonds in organic molecules, selective C-H functionalization reactions hold the potential to significantly streamline organic synthesis. 1 Over the past decade, extensive efforts have been devoted to the development of catalytic asymmetric functionalization of C(sp 3 )-H bonds. 2 Among these C-H functionalization processes, asymmetric C-H amination is particularly attractive, as it allows for the rapid assembly of chiral amines that are found in a range of important pharmaceuticals and agrochemicals. 3 Owing to their exquisite control over reaction stereochemistry, biocatalytic methods have recently emerged as an appealing alternative to stereoselective C(sp 3 )-H functionalization. 4 Over the past 8 years, the groundbreaking work from Arnold, Fasan, Hartwig, and many other researchers has culminated in a range of enantioselective C(sp 3 )-H amination processes using a metalloenzyme-catalyzed nitrene transfer mechanism. 5 In these C(sp 3 )-H amination reactions, using nitrene precursors including organic azides and hydroxylamine esters, a putative metal nitrene intermediate forms in the enzyme's active site, enabling further C(sp 3 )-H amination in a stereoselective fashion. 6,7 In recent years, the Arnold group engineered "P411" enzymes, a class of cytochromes P450 featuring an ironbinding serine residue, to facilitate asymmetric C(sp 3 )-H amination processes. 6a-g In 2019, a set of P411 nitrene transferases were developed for the asymmetric intramolecular amination of primary, secondary, and tertiary C(sp 3 )-H bonds. 6e With P411 Diane2 , cyclic 1,2-diamine derivatives formed with excellent enantioselectivity via the asymmetric functionalization of secondary C(sp 3 )-H bonds (Figure 1A). Our previous density functional theory (DFT) calculations using an Fe-porphine model complex indicated a stepwise mechanism involving an Fe nitrene-mediated hydrogen atom transfer (HAT) and a subsequent radical rebound step, leading to C(sp 3 )-N bond formation products. 6e To date, the enantioselectivity-determining step in this intramolecular secondary C(sp 3 )-H amination process remains unresolved.
Specifically, two enantioinduction scenarios, including enantiodetermining HAT and enantiodetermining radical rebound, could account for this stereoselective C(sp 3 )-H amination. As described in Figure 1B, in the HAT event, either of the two prochiral C-H bonds (C-H S and C-H R ) can be cleaved by the Fe nitrene (I), giving rise to two prochiral radical intermediates (II and III), respectively. If the interconversion of the prochiral radicals is slower than the rebound step (k 1 , k -1 ≪ k S , k R ), retention of stereochemistry would be expected in the subsequent radical rebound step. In this scenario, the HAT step would constitute the enantioselectivity-determining step. On the other hand, if the configurational change of this prochiral carbon-centered radical is much faster relative to the rebound step (k 1 , k -1 ≫ k S , k R ), stereoablation 8 would take place at this carbon-centered radical. In this case, the subsequent radical rebound would account for the excellent enantioselectivity under the Curtin-Hammett conditions.
Although several computational studies on related P411catalyzed C(sp 3 )-H amination processes supported the HAT/ radical rebound mechanism, 6e-g,9 the enantioselectivitydetermining step and the origin of enantioselectivity in most reactions are still not well understood. It is often assumed that the HAT step is enantioselectivity-determining. 9b,c The enantioselectivity control in the radical rebound step has not been quantitatively investigated by computation or experiment. In particular, the relative rates of C-N bond forming radical rebound and the conformational reorganization of the carboncentered radical remain elusive. Therefore, it is unclear whether radical rebound has any impact on the enantioselectivity. Additionally, the impact of key active site residues on these individual steps in the catalytic cycle is also unexplored.
Here, we report a combined computational and experimental study to shed light on the mechanism and origin of stereoselectivity of this biocatalytic C(sp 3 )-H amination. To differentiate between the two mechanistic scenarios and gain insights into the origin of stereocontrol, we undertook computational studies using quantum mechanics/molecular mechanics (QM/MM) calculations, classical and ab initio molecular dynamics (MD) simulations. We investigated the reaction free energy profiles and the enantiocontrol of the HAT and the radical rebound steps. We also performed computational studies on the rate of the interconversion between the prochiral carbon-centered radical intermediates. Moreover, deuterium-labeling experiments were performed to support the computational results by quantitatively determining the levels of enantioselectivity in the HAT and radical rebound steps. Finally, the roles of key active site residues on catalytic activity and enantioselectivity were investigated computationally and validated experimentally using mutated enzymes generated from site-directed mutagenesis. Collectively, these studies revealed that the enantioselectivity of this enzymatic C(sp 3 )-H amination is determined in the radical rebound step, a scenario which is often overlooked in previous studies.
Classical MD Simulations. To account for the dynamic nature of the protein scaffold, 10 we implemented the protocol shown in Figure 2 to unravel the reaction mechanism and origin of stereoselectivity in the enzymatic C(sp 3 )-H amination. The initial geometry of P411 Diane2 used in the calculations was generated by modifying the X-ray crystal structure of a related P411 enzyme (PDB ID: 5UCW). Eight amino acid residues were mutated using the Mutagenesis tool in PyMOL 11 (see Figure S1 for mutated residues). The substrate was then docked into P411 Diane2 using AutoDock. 12 Classical MD simulations were carried out using the pmemd module 13 of the GPU-accelerated Amber 16 package. 14 Force field parameters for the Fe nitrene complex were generated using the MCPB. py module 15 with the general Amber force field (gaff), 16 and the Amber ff14SB force field 17 was used for standard residues. After initial equilibration (see the Supporting Information), a 500 ns MD simulation was performed using the isothermal-isobaric ensemble (NPT). Clustering analysis was carried out using the cpptraj module 18 to identify the most populated structure in the last 300 ns of the MD simulation [see Figure S2 for root-mean-square deviation (rmsd) along the 500 ns MD trajectory]. The rmsd of the backbone was used as the distance metrics in the clustering analysis. Conformational samplings of the Fe nitrene intermediate and all the HAT and radical rebound transition states were then carried out using 50 ns of classical MD simulations. For transition states, the breaking and forming bonds were constrained by applying a force constant of 1000 kcal•mol -1 •Å -2 in the MD simulations. In the last 20 ns of each MD simulation, snapshots were extracted every 5 ns, giving four structures from each simulation and a total of 20 snapshots. Subsequently, QM/MM calculations were performed using the 20 snapshots as input geometries for each stationary point along the reaction pathway. 19 QM/MM Calculations. The ONIOM method 20 implemented in Gaussian 16 21 was used in all QM/MM calculations. Water molecules and counterions within 5 Å from the enzyme were included in the QM/MM calculations. The QM region includes the Fe-porphine complex, the substrate, and boundary hydrogen atoms, with a total of 66 atoms and a total charge of -1 (Scheme 1). The deprotonated axial serine ligand was used because the calculated pK aH values of a methoxy-ligated Fe(porphine)-nitrene complex (3.07) and a methoxy-ligated Fe(porphine)-NHR complex (1.89) indicate relative high acidity of the protonated serine ligand. The pK aH calculations were performed in an aqueous solution using the computational method described by Smith et al. 22 (see the Supporting Information for details). A deprotonated axial ligand was also used in several previous computational studies of P411 enzymes. 6e-g,9a,b, 23 For the QM region, the B3LYP 24 /6-31+G(d)-LANL2DZ-(Fe) level of theory was used in geometry optimization and vibrational frequency calculations, and the B3LYP-D3(BJ)/6-311+G(d,p)-LANL2TZ(f)(Fe) level of theory was used in single-point energy calculations because of its good agreement with CCSD(T) benchmark results. 6e For the MM region, the same force field parameters from the classical MD simulations discussed above were used. Residues greater than 15 Å away from the QM region were fixed during geometry optimization. The quadratic coupled algorithm 25 and the mechanical embedding scheme were used in geometry optimization. Single-point energy calculations were performed with the electronic embedding scheme, which better describes the electrostatic interactions between QM and MM regions. 26 Open-shell singlet, triplet, and quintet spin states for each structure were calculated using QM/MM. Wave function stability of the open-shell singlet spin state was confirmed by using the "stable = opt" keyword in Gaussian. Boltzmannweighted average Gibbs free energies 19a,27 of the 20 snapshots were calculated using
where n is the number of structures (n = 20), k B is the Boltzmann constant, and T is the temperature (T = 298.15 K). Gibbs free energies computed from individual snapshots (ΔG i ) are provided in the Supporting Information (Table S1). AIMD Simulations. Ab initio MD (AIMD) simulations were performed using the QUICKSTEP module with the hybrid Gaussian and plane waves (GPW) method 28 implemented in the CP2K package. 29 Because the entire system is treated using DFT, to reduce computational costs, the enzyme is truncated into a smaller cluster model 30 composed of the Fe-porphine complex, the substrate, and side chains of amino acid residues within 5 Å of the substrate based on the QM/MM-optimized geometry of 4 pro-R (Scheme 2). The backbone atoms were constrained in the AIMD simulations. The BLYP functional 24b,31 with D3 dispersion correction 32 and the DZVP basis set 33 with Goedecker-Teter-Hutter pseudopotentials 34 were used. The plane wave cutoff and the convergence criterion were 280 Ry and 10 -5 au, respectively. Metadynamics simulations 35 were carried out at 298 K with a time step of 0.5 fs. The N 1 -C 1 -C 2 -C 3 dihedral angle of 4 pro-R was used in the metadynamics simulations as a collective variable. The Gibbs free energy profile was obtained using thermodynamic integration. 36
Expression of P411 Variants. Escherichia coli [E. cloni Bl21 (DE3)] cells carrying plasmid encoding P411 variant were grown overnight in 4 mL of Luria broth with ampicillin. Preculture (1.5 mL, 5% v/v) was used to inoculate 28.5 mL of hyper broth with ampicillin in a 125 mL Erlenmeyer flask. This culture was incubated at 37 °C, Scheme 1. QM Region Used in the QM/MM Calculations Journal of the American Chemical Society 230 rpm for 2 h in a New Brunswick Innova 44R shaker. The culture was then cooled on ice for 20 min and induced with 0.5 mM IPTG and 1.0 mM 5-aminolevulinic acid (final concentrations). Expression was conducted at 22 °C, 150 rpm for 20 h. Cells were then transferred to a 50 mL conical centrifuge tube and pelleted by centrifugation (3000g, 5 min, 4 °C) using an Eppendorf 5910R tabletop centrifuge. The supernatant was removed and the resulting cell pellet was resuspended in M9-N buffer to OD 600 = 30. An aliquot of this cell suspension (2 mL) was taken to determine the protein concentration by the hemochrome assay after lysis sonication.
Biotransformations Using Whole E. coli Cells. Suspensions of E. coli [E. cloni BL21(DE3)] cells expressing the appropriate P411 variant in M9-N buffer (OD 600 = 30) were kept on ice. In another conical tube, a solution of D-glucose (500 mM in M9-N buffer) was prepared. To a 2 mL vial were added the suspension of E. coli cells expressing P411 (OD 600 = 30, 345 μL) and D-glucose (40 μL of 500 mM stock solution in M9-N buffer). This 2 mL vial was then transferred into an anaerobic chamber, where the azide substrate (15 μL of a 270 mM stock solution in EtOH) was added. The final reaction volume was 400 μL; final concentrations were 10 mM substrate and 50 mM D-glucose (note: reaction performed with E. coli cells resuspended to OD 600 = 30 indicates that 345 μL of OD 600 = 30 cells were added, and likewise for other reaction OD 600 descriptions.) The vials were sealed and shaken in a Corning digital microplate shaker at room temperature and 680 rpm for 12 h. The reaction mixture was then extracted with EtOAc and analyzed by chiral highperformance liquid chromatography (HPLC) using an internal standard.
Reaction Mechanism. The Gibbs free energy profile of the P411 Diane2 -catalyzed C-H amination of sulfamoyl azide substrate 1 obtained from QM/MM calculations is shown in Figure 3 (see Figure S4 in the Supporting Information for QM/MM-optimized structures). The quintet (high spin), triplet (intermediate spin), and open-shell singlet (low spin) spin states of each intermediate and transition state structure were considered in the calculations. Here, only the pathways involving the cleavage of the C-H S bond of the substrate to form the favored (S)-enantiomer of the product are shown. The cleavage of the C-H R bond follows the same mechanism (see Figure S5). The origin of enantioselectivity will be discussed in detail in a later section of the article. Our QM/ MM calculations reveal that Fe nitrene species 3 has substantial radical character on the nitrogen center, facilitating for the C-H bond activation (see Figure S6 for spin densities of all computed structures). The C-H S cleavage via TS1 gives rise to the benzylic radical intermediate 4 pro-S , where the (Si)face of the benzylic radical points toward the nitrenoid nitrogen. Based on our QM/MM studies, this HAT step is exergonic and irreversible, which is consistent with DFT investigations with a model Fe-porphine complex (Figure S7). In the HAT step, the triplet and open-shell singlet spin states are very similar in energy, whereas the quintet is substantially higher in energy. The similar energy profiles suggest that the HAT step may involve both triplet and singlet spin states, a "two-state reactivity" reminiscent of the HAT step of the native C-H hydroxylation with P450 enzymes, which involves both doublet and quartet states of the active Fe oxo species. 37 However, unlike the P450-catalyzed C-H hydroxylation, where the low-spin doublet Fe IV -OH intermediate (compound II) undergoes barrierless radical rebound, both the lowand intermediate-spin intermediates of the Fe III -NHR intermediate 4 pro-S require substantial barriers to radical rebound via TS2. Instead, the quintet spin state becomes the most favorable in the radical rebound transition state (TS2), indicating a spin-crossover 6f,38 event from the triplet and singlet intermediates to the quintet spin state prior to the radical rebound (see Figure S8 for the calculated minimum energy crossing point between the triplet and quintet surfaces).
These computed energy profiles indicated a key difference between the P411 Diane2 -catalyzed C-H amination and the P450-catalyzed C-H hydroxylation. 39 In the P450-catalyzed hydroxylation, low-spin pathways are effectively concerted with ultrashort-lived intermediates, while high-spin pathways are stepwise. By contrast, the P411 Diane2 -catalyzed C-H amination favors the stepwise mechanism regardless of whether the singlet or triplet Fe nitrene is involved in HAT. The relatively high barrier in the C-N bond forming radical rebound step is due to the spin-state change from the near-degenerate singlet/ triplet carbon-centered radical 4 pro-S to the quintet state in TS2. The radical rebound from singlet and triplet 4 pro-S is disfavored due to several factors. First, in the triplet and singlet rebound transition states, 3 TS2 and 1 TS2, the Fe-N bond is elongated (2.33 and 2.09 Å, respectively) relative to 4 pro-S (2.01 Å), requiring significant distortion. In contrast, the Fe-N bond in quintet 5 4 pro-S is already elongated (2.22 Å). This predistortion of the Fe-N bond facilitates the nitrogen rebound via the quintet state. The elongated Fe-N bond in the quintet 5 TS2 reduces the steric repulsions between the secondary benzylic radical and the Fe-porphyrin during radical rebound. Second, the ground state of the product complex is the high-spin quintet Fe(II)-porphyrin 5. This leads to smaller thermodynamic driving force for the radical rebound on the singlet and triplet surfaces compared to that involving the quintet intermediate 5 4 pro-S . Lastly, the fivemembered cyclic transition state TS2 suffers from relatively high ring-strain energy. Based on computed values derived from hypothetical homodesmotic reactions 40 (Figure S9), the ring strain energies of the five-membered ring transition state 5 TS2 and the cyclic amination product 2 are 7.7 and 5.1 kcal/ mol, respectively. Therefore, the ring strain energy of the cyclization transition state makes the intramolecular C-N rebound slower than the corresponding intermolecular process. 6f,g,9b,c Taken together, the QM/MM-computed reaction energy profiles revealed an unusual mechanism with a high-barrier radical rebound, indicating relatively long lifetimes of the carbon-centered radical intermediates, which may lead to the ablation of stereochemistry prior to radical rebound.
Enantioselectivity in the HAT and Radical Rebound Steps. Because the high radical rebound barrier shown above indicates that either the HAT or the radical rebound step can be enantioselectivity-determining, we computed the enantioselectivity in both steps in the amination of the two prochiral benzylic C-H bonds in 1 (Figure 4). From the Fe nitrene species 3, HAT with the C-H S (TS1) and C-H R (TS3) bonds require activation free energies of 12.2 and 14.0 kcal/ mol, respectively. After the formation of the prochiral benzylic radical intermediates 4 pro-R and 4 pro-S , subsequent C-N bond forming radical rebound takes place via quintet transition states, TS2 and TS4, to form the two enantiomers of the amination product. The computed activation free energy difference (ΔΔG ⧧ ) between radical rebound transition states TS2 and TS4 is 6.5 kcal/mol, much higher than the moderate ΔΔG ⧧ of the HAT step, which is only 1.8 kcal/mol.
The computed energy profiles have some interesting implications on the enantiocontrol of the asymmetric amination. Although the HAT step is exergonic and irreversible, the relatively high barriers to radical rebound suggest that the rate of interconversion between benzylic radical intermediates 4 pro-R and 4 pro-S may be faster than the C-N bond forming radical rebound. This scenario is reminiscent of the Curtin-Hammett principle, where the enantioselectivity is only affected by the energy difference between the radical rebound transition states, TS2 and TS4. Additionally, the computed ΔΔG ⧧ for HAT (1.8 kcal/mol) corresponds to 91% ee, much lower than the experimentally observed enantioselectivity (>99.9% ee). On the other hand, the high enantioselectivity in the radical rebound step (ΔΔG ⧧ = 6.5 kcal/mol, which corresponds to >99.9% ee) is consistent Only the moststable spin state of each species, denoted by superscript, is reported. ΔG ave and ΔH ave values are Boltzmann-weighted Gibbs free energies and enthalpies computed from initial structures taken from 20 MD snapshots. Blue: pathway leading to the major enantiomeric product 2. Red: pathway leading to the minor enantiomeric product ent-2.
with the high levels of enantioselectivity observed in the experiment.
Deuterium-Labeling Experiments. Our QM/MM calculations revealed an unusual enantioinduction mechanism, wherein the HAT step is moderately enantioselective and the nitrogen-rebound step is highly enantioselective. To further shed light on the enantioselectivity-determining step in this intramolecular C-H amination, we prepared a set of enantiopure monodeutero, monoprotio substrates and subjected them to the enzymatic reaction conditions (Figure 5A). When (R)-1-d 1 was applied, the biocatalytic C-H amination furnished the product with >99% ee, and the 2-d 1 /2 ratio was determined to be 96:4 on the basis of 1 H NMR analysis. When (S)-1-d 1 was applied, this biocatalytic amination reaction still furnished the same enantiomeric product with >99% ee, and the 2/2-d 1 ratio was found to be 67:33.
Quantitative activation free energy analysis of these reactions was next performed (Figure 5B. See Figures S10 andS11 for detailed calculations). Based on our QM/MM-computed energy profiles with evolved P411 Diane2 (Figure 4) and DFT calculations on a model complex (Figure S7), the HAT step in this C-H amination is irreversible. Thus, the 2/2-d 1 ratio, that is, ΔΔG ⧧ H/D (HAT), is controlled by two energy terms, including ΔG KIE , which reflects the kinetic isotope effect (KIE) in the HAT step, and ΔG enantioselectivity , which reflects the enzyme-induced enantioselectivity in the HAT event with the non-deuterated substrate. With (R)-1-d 1 as the substrate, both the KIE and the enzymatic stereocontrol favor the abstraction of the pro-(S)-H, thus leading to a higher 2-d 1 /2 ratio. On the other hand, when (S)-1-d 1 was applied, enzymatic stereocontrol overrides the inherent KIE effect, resulting in a lower 2/2-d 1 ratio. Activation free energy analysis allowed us to dissect the two effects, and the two energy terms of ΔG KIE and ΔG enantioselectivity were determined to be -0.73 and -1.14 kcal/ mol, respectively. This ΔG KIE value corresponds to a k H /k D of 3.4. This KIE is consistent with previously measured KIEs with related enzymatic intramolecular C-H amination processes. 6b,l The primary intramolecular KIE suggests an irreversible HAT step, which is consistent with the computed reaction energy profiles. Moreover, the small value of ΔG enantioselectivity for the HAT step corresponds to an enantiomeric ratio (e.r.) of 87:13 (74% ee), clearly showing that the HAT step with P411 Diane2 is only moderately enantioselective, and the very high level of product enantiopurity was controlled by the radical rebound step. Therefore, these results suggest that the benzylic radical is relatively long-lived, and the excellent enantioselectivity observed in this enzymatic C-H amination is likely controlled by the C-N bond-forming radical rebound step.
Stereoablation after HAT via Conformational Change of Benzylic Radical Species. The experimental and computational results discussed above indicate that although HAT is irreversible, it is not enantioselectivity-determining. This means the prochiral benzylic radical intermediates 4 pro-R and 4 pro-S must undergo rapid interconversion before the radical rebound. This mechanistic scenario is not well understood for radical-mediated enzymatic reactions, particularly new-to-nature enzymatic reactions, considering the short lifetime of radical intermediates.
In order to investigate the rate of the interconversion between benzylic radical intermediates 4 pro-R and 4 pro-S via conformational reorganization, classical MD and AIMD simulations were carried out. 41 Using the QM/MM-optimized structure of 4 pro-R as the starting geometry, a 500 ns classical MD simulation was performed. Within the first 6.5 ns of the MD simulation, a snapshot with geometry akin to 4 pro-S was obtained (see Figure S12). The rotations about the S-N 1 and C 1 -C 2 bonds (Figure 6A,B) occur at approximately the same time, from 6.0 to 6.3 ns (see Figure S13 for rotation about the N 1 -C 1 bond). This process allows the bulky Ph group on the substrate to point toward the same direction without clashing with the Fe-porphyrin. Throughout the 500 ns classical MD simulation, frequent rotations about the S-N 1 and C 1 -C 2 bonds were observed (see Figure S14). These results suggest that within the enzyme's active site, the benzylic radical is conformationally flexible and can rapidly rotate to expose either prochiral face of the carbon-centered radical toward Fe III -NHR prior to the C-N bond forming radical rebound.
Next, we performed AIMD metadynamics simulations to quantitively determine the Gibbs free energy barrier to the isomerization of the benzylic radical intermediates. The AIMD simulations suggest that the transformation of 4 pro-R to 4 pro-S is facile, with a low barrier of 3.6 kcal/mol with respect to 4 pro-R (Figure 6C). This result is consistent with the rapid conformational change observed in the classical MD simulations. Most importantly, this conformational change from 4 pro-R to 4 pro-S occurs at a time scale faster than the C-N bond-forming radical rebound, highlighting the essential role of radical rebound in determining the enantioselectivity of this enzymatic C-H amination. 42 Origin of Enantioselectivity and the Roles of Key Residues. The QM/MM-optimized structures of intermediates and transition states in the C-H amination pathways are shown in Figure 7. In all structures, C-H/π or π/π interactions between the substrate and the W263 residue, a key mutation previously introduced during the directed evolution of P411 Diane2 , are observed. This indicates that W263 plays an essential role in enhancing substrate binding and stabilizing intermediates and both HAT and radical rebound transition states. With the C-H/π interaction with W263 and a steric effect of A87, the orientation of the substrate is constrained, pointing the Ph group on the substrate toward V328, a key residue for enantioinduction (vide infra).
In the transition state stereoisomers leading to the opposite enantiomeric products, the benzylic C-H bonds in TS1 and TS3 and the benzylic radical in TS2 and TS4 approach from the opposite faces of the Fe nitrene or Fe III -NHR complex. In transition states TS3 and TS4, the benzyl group is placed close to V328, leading to steric repulsions between these two groups, whereas in TS1 and TS2, the benzyl group approaches from the opposite side of V328, and thus, the steric repulsion is diminished. Although TS3 and TS4 are both destabilized by steric repulsions with V328, the magnitude of steric effect is differentthe Ph group is placed much closer to V328 in the radical rebound transition state TS4 [d(H•••H) = 2.14 Å] because of the shorter distance between the benzylic carbon and the Fe-porphyrin (2.33 Å) than that in TS3. The stronger steric repulsions that destabilize TS4 lead to the greater enantioselectivity in the radical rebound step (ΔΔG ⧧ rebound = 6.5 kcal/mol) compared to that in the HAT step (ΔΔG ⧧ HAT = 1.8 kcal/mol).
To validate the computationally revealed roles of key residues, such as W263, V328, and A87, we carried out sitedirected mutagenesis and tested the catalytic activity and enantioselectivity of these P411 Diane2 variants (Tables 1 andS4). We found that mutations W263A and W263F did not affect the enantioselectivity but substantially reduced the enzyme activity (Table 1, entries 1-3). This finding is consistent with the stabilization effect of W263 observed from QM/MM calculations. The P411 Diane2 A87V and A87L variants also led to lower yields without impacting enantioselectivity (entries 4-5). This suggests that a larger residue 87 may hinder substrate binding and decrease conversion without altering the mode of enantioinduction. We further validated the role of V328 by evaluating the enantioselectivity of P411 Diane2 V328A generated by sitedirected mutagenesis. Indeed, mutant P411 Diane2 V328A provided product 2 with 97% ee and lower activity, suggesting a larger residue at 328 is needed for enantiocontrol (entry 6). Furthermore, double mutants P411 Diane2 V328A A87V and P411 Diane2 V328A A87L were generated and found to furnish further decreased enantioselectivity (81 and 55% ee, respectively, entries 7-8). Our QM/MM calculations using the V328A A87V variant indeed showed a decreased enantioselectivity in the radical rebound step (ΔΔG ⧧ = 1.7 kcal/mol, Figure S17) due to reduced steric repulsions between the substrate and residue 328. These results are consistent with our computational model and demonstrate the importance of residues 328 and 87 in crafting a substrate binding pocket for excellent enantiodifferentiation.
Using a combined QM/MM and experimental approach, we investigated the mechanism and origin of enantioselectivity of the recently developed biocatalytic asymmetric C(sp 3 )-H amination process. This C-H amination was found to occur via an irreversible HAT step and a C-N bond-forming radical rebound step. Contrary to previous understanding on enantioinduction mechanisms, we showed that radical rebound, rather than HAT, is enantioselectivity-determining in this biocatalytic intramolecular C-H amination. Our QM/ MM calculations indicated that the radical rebound is relatively slow due to the spin-state change and ring strain in the intramolecular cyclization transition state. Classical and AIMD simulations suggested that the carbon-centered radical undergoes much faster conformational change, allowing for stereoablation at the carbon-centered radical intermediate. Therefore, under the Curtin-Hammett conditions, the enantioselectivity is determined in the radical rebound step. The QM/ MM-computed activation free energy differences between the two stereoisomeric pathways indicated a moderate level of stereoselectivity in the HAT step (ΔΔG ⧧ = 1.8 kcal/mol) and an excellent level of enantioinduction (ΔΔG ⧧ = 6.5 kcal/mol) at the radical rebound stage that can account for the high levels of enantioselectivity observed in the experiment. These computational findings are corroborated by experimental results employing enantiopure and deuterium-labeled substates, which also indicated moderate enantioselectivity in an irreversible C-H cleavage event, and thus, the excellent enantioselectivity of the overall amination must be controlled in the subsequent radical rebound step. The roles of several key active site residues, including W263, A87, and V328, in confining substrate orientation and manifesting enantioinduction via steric effects, have been demonstrated computationally and validated experimentally by evaluating the activity and enantioselectivity of enzyme variants generated by site-directed mutagenesis. Together, this study highlights an unusual enantioinduction mechanism in metalloenzyme-catalyzed asymmetric transformations involving radical intermediates. We expect that these insights will guide the further development of stereoselective new-to-nature biocatalytic reactions using a radical mechanism. Yields and ee's were determined by HPLC analysis. Reactions were carried out using whole E. coli cells harboring P411 Diane2 mutants.
https://doi.org/10.1021/jacs.2c02283 J. Am. Chem. Soc. 2022, 144, 11215-11225