Deracemization is an endergonic process, as converting a two-state racemate to a single enantiomeric state incurs an entropic penalty9,19 (Supplementary Fig. 3). Therefore, a catalytic deracemization requires input of an external energy such as stoichiometric chemical reagent20, light21 or mechanical energy22 to drive the process and overcome the unfavourable free-energy change. Despite the high barrier to rotation around the biaryl bond in BINOL, there are several reports demonstrating that external stimuli—especially oxidizing metal complexes—can induce conformational lability at ambient temperature owing to pyramidalization at the atropisomeric axis11,18,23,24,25,26 (Fig. 2b). However, in contrast to our observations, racemization of BINOL derivatives is the common outcome18,25,26. We suggested that, in our system, oxidation of BINOL by the P450 biocatalyst generates a radical intermediate with reduced conformational stability. The chiral environment provided by the P450 enzyme could then be the key to atropoenrichment14,18,27.
To identify which components drive this biocatalytic deracemization, we tested the role of each element (Fig. 3a). Given the established oxidation chemistry of P450s, we proposed that BINOL (5) could be oxidized by a reactive Fe-oxo (or Fe-peroxo) haem species that could initiate the process through hydrogen atom transfer (HAT)28,29,30 (Fig. 3a). Formation of this reactive intermediate requires the reduction of haem-bound oxygen by an external electron source, commonly a redox partner or reductase31,32 (see Supplementary Fig. 4). Our biocatalytic system15 uses a fused RhFRed reductase domain, which obtains its reducing equivalents from NADPH that is supplied in situ through a cofactor recycling system. Specifically, glucose-6-phosphate dehydrogenase (G6PDH) converts NADP+ to NADPH using glucose-6-phosphate (G6P) as the terminal electron source33. Furthermore, during optimization, we identified that the inclusion of sodium ascorbate was critical for high substrate recovery.
Fig. 3: Identification of the minimal set of reaction conditions. a, General roles of each component in the reaction mixture. b, Control experiments to exclude the reductase and recycling system. c, Identification of the terminal oxidants and reductants. d, Enantiomeric ratio of BINOLs 5 and 6 as a function of NaAsc concentration and time. All reactions performed in triplicate (n = 3). Error bars depict standard deviation. Full size image
Using the full cocktail of reaction components in our standard conditions with rac-5 (P450-fused reductase, NADPH recycling system, sodium ascorbate and oxygen), we obtained 5 in 93:7 er (Fig. 3b, entry 1). Omitting the A2-RhFRed variant (renamed BinDR1-RhFRed for BINOL deracemase-1) eliminated deracemization activity (Fig. 3b, entry 2); however, removing the fused reductase domain from BinDR1 did not have a negative impact (Fig. 3b, entry 3). Further elimination of the NADP+ cofactor and the NADPH recycling system did not have a notable impact on reactivity (Fig. 3b, entry 4). Although we initially considered that the reductase domain would be key to the oxygen activation mechanism31, these data suggested that generation of peroxide provided an alternative pathway. Inclusion of catalase, a peroxide scavenger, eliminated deracemization reactivity (see Supplementary Fig. 11). Together, these results identify a streamlined set of reaction components—P450, NaAsc and atmospheric oxygen—that bypass the traditional oxygen reduction pathway for P450s.
We then investigated the potential roles of oxygen or peroxide as the oxidant and NaAsc as the reductant powering deracemization. Under anaerobic conditions, P450 and NaAsc alone produced no change in the enantiomeric ratio of rac-5. However, addition of peroxide restored activity in the absence of oxygen (see Supplementary Fig. 5), suggesting that peroxide can serve as the active oxidant. Time-course analysis under ambient atmospheric conditions with BinDR1 and NaAsc gave 88% recovery and 84:16 er after 3 h (Fig. 3c, entry 1). Peroxide alone produced a rapid initial increase in enantiomeric ratio, followed by a plateau after 15 min of reaction time (Fig. 3c, entry 2). The high concentration of peroxide probably led to enzyme deactivation and substantial substrate loss, consistent with insufficient reductant present after the initial substrate oxidation34. Combining NaAsc and peroxide accelerated deracemization (Fig. 3c, entry 3), yet the high concentration of exogenous peroxide still reduced BINOL recovery compared with NaAsc alone. Thus, we focused our efforts on developing conditions using atmospheric oxygen and NaAsc to drive deracemization with the highest substrate recovery.
NaAsc concentration affects the rate and extent of deracemization (Fig. 3d). When incubating BinDR1 with BINOLs 5 and 6, increasing the concentration of NaAsc from 2 to 16 mM accelerated enrichment and increased the final enantiomeric ratio achieved. Beyond 16 mM, the maximum enantiomeric ratio (97:3 for substrate 6) did not improve; higher NaAsc only increased the rate of deracemization to this enantiomeric ratio. At 64 mM NaAsc, the reaction with 6 initially reached 97:3 er but eroded slightly at later time points (Fig. 3d). These results indicate that a concentration window exists in which increasing NaAsc increases the achievable enrichment. Once the maximum enantiomeric ratio is achieved, higher concentrations of NaAsc lead to faster reaction without increasing the final enantiomeric ratio (see Supplementary Fig. 10). These experiments define the minimal components for this chemical-driven deracemization: P450 biocatalyst, oxygen and NaAsc. We propose that oxygen serves as the terminal oxidant and NaAsc as the terminal reductant in this net-redox-neutral process. Reagent compatibility of the substrate, oxidant and reductant is critical (see Supplementary Fig. 6). In the absence of P450, oxidants such as oxygen and hydrogen peroxide react minimally with NaAsc35,36. This underscores an advantage of biocatalytic deracemization: the enzyme generates and localizes a reactive species, enabling redox-neutral deracemization in a single vessel despite the presence of both oxidant and reductant reagents11.
Having identified O 2 and NaAsc as the terminal oxidant and reductant that drive the uphill deracemization, we considered how (S)-5 is net converted to (R)-5 through mechanistically independent steps for oxidation and reduction (see Supplementary Fig. 7). We propose a cyclic deracemization in which an oxidizing P450 species generates a stereolabile phenoxy radical intermediate (I-5)29,30 (Fig. 4b). Both literature reports and our calculations indicate that BINOL radical species have a substantially decreased barrier of rotation about the chiral axis14,18. The computed transition state energy for the neutral radical I-5 is 25.5 kcal mol−1, comparable with values reported for related radical cations27 and consistent with accessible interconversion at ambient temperatures (see Supplementary Fig. 12). Binding interactions between this intermediate and the enzyme active site could further reduce this barrier. Reduction of the radical intermediate (I-5) through HAT or proton-coupled electron transfer (PCET) would then regenerate closed-shell BINOL 5, restoring the high rotational barrier and rigidifying the chiral axis in the favoured conformation. Possible reductants include direct reduction with NaAsc37,38, self-exchange with the phenol of another substrate molecule or enzyme-mediated PCET with a redox-active residue37. The large magnitude of the rate constant for phenol self-exchange is ascribed to the formation of a pre-PCET hydrogen bonding complex, which may be constrained in the active site39. Productive closure of the catalytic cycle probably involves a second HAT step by Cpd II to regenerate the resting ferric haem40. Unproductive pathways could lead to rebound hydroxylation and overoxidation of the substrate or enzyme inactivation oxidation of a proximal residue41. The observation that excess ascorbate improves BINOL recovery (Fig. 3c and Supplementary Fig. 1) supports a role for ascorbate in suppressing overoxidation and preserving enzyme activity, acting in part as a sacrificial reductant while permitting productive turnover.
Fig. 4: Plausible mechanisms for enantioenrichment. a, Thermodynamic/kinetic requirements for a deracemization contextualized with a cyclic deracemization. b, Proposed catalytic cycle for the BinDR-mediated deracemization of BINOL. Full size image
For a cyclic deracemization process to produce enrichment, at least one of the two mechanistic steps must be stereoselective. Although we cannot exclude a reductive step occurring within the active site, there is no evidence that ascorbate, a chiral molecule, influences the enantioselectivity, as the same enantiomeric ratio is achieved in the absence of sodium ascorbate (compare Fig. 2c with Fig. 3b, entry 4). By contrast, the oxidation event is anticipated to occur in the chiral enzyme active site and provide the necessary selectivity to achieve deracemization. The role of the P450 in controlling selectivity is supported by experiments with P450 variants that lead to different selectivity on the same substrate, which is highlighted in the substrate scope analysis of this deracemization with a set of BinDR variants (see Fig. 5).