Refining the Substrate-Cofactor Disposition Model of Hyoscyamine 6β-Hydroxylase Catalysis Using Hyoscyamine Analogs

Abstract

Hyoscyamine 6β-hydroxylase (H6H) is a mononuclear nonheme iron and 2-oxoglutrate dependent oxidase responsible for the biosynthesis of the tropane alkaloid scopolamine. H6H is an ideal model system to study the nonheme iron enzymes as it has dual functions catalyzing both C6-hydroxylation of hyoscyamine and epoxidation of the resulting alcohol. Previous crystallographic and spectroscopic studies have led to a model where substrate disposition relative to the intermediary iron complexes determines the reaction outcome. Herein, a number of H6H substrate analogs with cyclopropyl, methylidene, fluoro, methoxy and trifluoromethoxy substituents among others at C6 and C7 are assayed with H6H to determine whether the disposition model accurately predicts the reaction outcome. The results suggest that additional factors, which include H-bonding interactions that stabilize intermediary hydroxy-ferric complexes against rebound, are likely at play in determining the fate of the individual substrate radicals. These effects are both important and expected to correlate with substrate-cofactor disposition.

Graphical Abstract

INTRODUCTION

Scopolamine (3) is a clinically useful tropane alkaloid produced by the Solanaceae family of plants.¹ It has been used as a sedative and an antispasmodic agent to treat disorders characterized by restlessness and agitation.^1,2 The biosynthesis of scopolamine has also been of much interest due to two sequential oxidation reactions catalyzed by the nonheme iron and 2-oxoglutarate dependent (Fe/2OG) enzyme hyoscyamine 6β-hydroxylase (H6H, Figure 1A).^3-10 Like other Fe/2OG enzymes,^11-16 H6H coordinates a single ferrous iron in its active site via a conserved His/His/Asp(Glu) facial triad.^17,18 Upon binding 2OG, O₂ and substrate (e.g., 1), 2OG is oxidized to CO₂ and succinate concomitant with formation of the reactive ferryl complex (4) as in the case of other Fe/2OG enzymes (Figure 1B).^19,20 Regioselective H atom abstraction from C6 of 1 by the ferryl complex thus yields a C6 radical intermediate (6) that is subsequently hydroxylated upon rebound from the resulting hydroxy-ferric complex (5).⁹ While the 6β-hydroxylated product (2) is a substrate for H6H as well, this reaction proceeds with changes in both the regiochemistry and reaction outcome leading to H atom abstraction from C7 and epoxidation to yield scopolamine (4 + 2 → 3).^5,8,21

Figure 1.

(A) Biosynthesis of scopolamine (3) involves sequential 6β-hydroxylation and epoxidation of hyoscyamine (1) catalyzed by H6H. (B) H6H is an Fe/2OG enzyme that utilizes a strongly oxidizing ferryl (4) intermediary complex to activate the otherwise inert C6–Hβ and C7–Hβ bonds. (C) H6H also catalyzes the epoxidation of 6,7-dehydrohyoscyamine (9) and 7β-hydroxyhyoscyamine (10); however, it does not catalyze hydroxylation of the latter compound.

In addition to H6H, other enzymes such as LolO^22,23 and clavaminate synthase^24-26 have also been shown to catalyze intermolecular followed by intramolecular oxygen insertion into otherwise unreactive C–H bonds to form heterocyclic products. This switching between canonical hydroxylation and noncanonical oxacyclization as part of their normal biosynthetic activity has made these enzymes of particular interest for studying the determinants of Fe/2OG catalytic outcomes. Moreover, H6H has been shown to catalyze the oxidation of a wide variety of tropanes and related compounds in addition to its biosynthetic substrates 1 and 2. For example, H6H catalyzes epoxidation of 9,^6,9 which is the 6,7-desaturated derivative of hyoscyamine, as well as epoxidation of 7β-hydroxylated hyoscyamine (10) with little to no apparent hydroxylation at C6β (Figure 1C).²⁷ Explaining these changes in the regiochemistry and product distributions of oxacyclases such as H6H has thus become an outstanding question in the study of Fe/2OG enzymes.

One possible explanation involves the substrate-cofactor disposition model, which was originally developed from studies of Fe/2OG halogenases.²⁸ This model distinguishes two limiting configurations or reaction channels for H atom abstraction that appear to correlate with the subsequent fate of the substrate radical.^28-30 The σ-channel involves apical positioning of the cleaved C–H bond with respect to the iron-oxo ligand and a more parallel orientation of the C–H and Fe=O bonding vectors, whereas the alternative π-channel places the substrate more lateral and orthogonal to the Fe=O complex (Figure 2A). According to this model, the σ-channel exhibits the lower barrier to H atom transfer but also leaves the resulting substrate radical more susceptible to immediate hydroxyl rebound (Figure 2B). In contrast, the π-channel results in less efficient, slower H atom abstraction; however, it places the subsequent substrate radical adjacent to alternative radical acceptors such as halide ligands coordinated to the iron center thereby facilitating halogenation.³⁰ While some computational studies support the hypothesis of faster H atom transfer via the σ-channel,^31,32 others have suggested little difference in activation energy.^33,34 In contrast, experimental assessments of ferryl decay have generally shown more than an order of magnitude faster H atom transfer via the σ versus π-channels.^18,30,35 Therefore, slower H atom transfer via the π-channel indeed appears to be a catalytically acceptable trade-off for improved substrate positioning to facilitate halogenation of the substrate radical.³⁵

Figure 2.

(A) The substrate-cofactor disposition model describes two limiting channels for H atom abstraction from the substrate by the ferryl iron-oxo complex based on the relative positioning and orientation of the two interacting components. H atom abstraction via the σ-channel is proposed to be more efficient and thus faster compared to H atom abstraction via the π-channel. (B) In the case of Fe/2OG halogenases, the H atom abstraction channel correlates with the subsequent fate of the substrate radical. (C,D) Proposed geometries of the H6H ferryl complexes bound with (C) hyoscyamine (1) and (D) 6β-hydroxyhyoscyamine (2) inferred from crystal structures of the corresponding vanadium(IV) oxide complexes.¹⁸

The success of the disposition model in describing Fe/2OG halogenases has led to the proposal that it may also explain H6H catalyzed epoxidation.^17,18 Indeed, crystallographic studies as well as spectroscopic measurements coupled with computational analyses have suggested placement of the C6Hβ and C7Hβ centers of hyoscyamine (1) and 6β-hydroxyhyoscyamine (2) respectively more apical and lateral to the iron-oxo ligand (Figure 2C,D).^17,18 Such a configuration would thus predict facile C6 H atom abstraction via the σ-channel and favor subsequent hydroxylation of the substrate radical 6. In contrast, slower H atom abstraction from C7 of 2 via the π-channel would presumably correlate with slower hydroxyl rebound to the C7 radical (7) permitting electron transfer to yield a secondary carbocation (8) that undergoes intramolecular nucleophilic addition to generate the epoxide in scopolamine (Figure 1B).^17,18 This is consistent with crystal structures of H6H bound with 2 and vanadium(IV) oxide as a mimic of the ferryl complex and led to previous suggestions of epoxidation taking place concerted with H atom abstraction from C7^21,27 being otherwise dismissed.¹⁸

H6H catalysis, however, may not be quite so simple. In particular, the disposition model does not explain why H atom abstraction from C7 via the π-channel necessarily leads to epoxidation in the subsequent step. Furthermore, while decomposition of the ferryl complex during epoxidation has been shown to proceed concerted with transfer of a solvent Hydron,¹⁸ H atom transfer concerted with deprotonation of the 6β-hydroxyl upon cyclization would be inconsistent with the otherwise stepwise disposition model (Figure 1B). Substrate-cofactor placement would likewise predict H6H catalyzed hydroxylation of 7β-hydroxyhyoscyamine (10) to yield the 6,7-dihydroxylated product (11), because the C6 H atom should remain apically positioned given the purported resiliency of substrate binding configurations in the H6H active site.¹⁸ However, precisely the opposite is observed (see Figure 1C).²⁷ Herein, several additional substrate analogs are employed in an effort to further refine the model of H6H catalyzed hydroxylation versus epoxidation. A modification is proposed that accounts for all experimental data acquired to date and emphasizes the importance of interactions between the substrate and intermediary iron complexes.

RESULTS AND DISCUSSION

Lifetime of the C6 Radical.

Recent computations based on H6H crystal structures have implied the existence of the C6 (6) radical intermediate with a half-life of ca. 0.16 ns during the hydroxylation reaction at 298 K.¹⁷ In order to provide an experimental assessment of this prediction, the half-life of the C6 radical was measured using the C7-spirocyclopropane radical clock 12 (Figure 3A). Based on empirical values, the C6 radical 13 should undergo cyclopropyl ring opening with a half-life of ca. 10 ns (60-fold longer than the computed half-life for rebound).^36-38 Therefore, the computations predict that the reaction of 13 with H6H should result in significant formation of the hydroxylated product 18.

Figure 3.

(A) H6H reaction outcome with the C7-spirocyclopropane analog (12) does not lead to observable C6 hydroxylation. (B) H6H reaction with the 6,7-desaturated-7-hydroxyethyl derivative 15 results in 8,9-epoxidation rather than C6 hydroxylation. (C) The 7-methylidene analog 21 of hyoscyamine will yield an allylic radical upon H atom abstraction from C6 that undergoes C8 rather than C6 hydroxylation in addition to 7,8-epoxidation.

Incubation of 1.0 mM 12 with 92 μM H6H from Hyoscyamus niger in the presence of 0.40 mM FeSO₄, 5.0 mM 2OG, 4.0 mM ascorbate and 50 mM tris(hydroxyl-methyl)aminomethane (Tris) buffer (pH 7.4) for 1 h indeed produced only a single monooxygenated product (calcd m/z for C₁₉H₂₅NO₄ [M + H]⁺ 332.1856, obsd 332.1860, Figure S1). Contrary to expectation, however, ¹H NMR analysis showed no evidence of a cyclopropane signal in the 0–1 ppm region where it would typically be observed (see Supporting Information). Instead, an olefinic resonance at 5.78 ppm was identified, which along with additional ¹³C NMR analysis established the structure of the product as that of 15 (Figure 3A). These results indicate that the half-life for hydroxyl rebound to 13 is much greater than the 10 ns for radical-mediated ring opening and thus more than an order of magnitude longer than the 0.16 ns calculated for hydroxyl rebound to 6.¹⁷

This discrepancy between computed and observed rates may be due to the C7 spiro moiety perturbing the binding orientation of 12 compared to that of 1. However, an argument has been made that small changes in substrate structure are unlikely to induce significant changes in binding configuration, which is consistent with the observation that scopolamine among other derivatives bind in essentially the same configuration as hyoscyamine.¹⁸ In contrast, the authors of the original computational study proposed that other substrate binding conformations may be available in order to explain additional discrepancies between observed and computed partitioning between C6 versus C7 hydroxylation of hyoscyamine.¹⁷ This additional entropy, which may be difficult to appreciate in low temperature crystal structures, may help to explain the apparent reduction in hydroxylation rate versus the computed optimum. In other words, the H6H ferryl complexes are likely to be more dynamic than the static crystal structures might otherwise suggest.

Determinants of Reaction Outcome.

Prolonged incubation of 12 with H6H exceeding 1 h showed the gradual disappearance of the newly formed product 15 concomitant with the formation of a new product having an MS profile indicative of dehydrogenation (calcd m/z for C₁₉H₂₃NO₄ [M + H]⁺ 330.1700, obsd 330.1698, Figure S1). This species was identified as the epoxide 16 due to the presence of an olefinic proton resonance at 6.15 ppm together with three oxirane proton signals at 3.45 and 2.84–2.91 ppm in the ¹H NMR spectrum. Consistent with this assignment, the product was readily hydrolyzed to the diol 17 (calcd m/z for C₁₉H₂₅NO₅ [M + H]⁺ 348.1805, obsd 348.1804) upon exposure to dilute trifluoroacetic acid during chromatographic purification (16 → 17, Figure 3A). Therefore, H6H is also able to catalyze the epoxidation of 15 to yield 16. If the oxidative cyclization of 15 to 16 proceeds through a stepwise mechanism, then it presumably would involve H atom transfer from C8 to the ferryl complex via the π-channel (Figure 3B). However, the resulting intermediate would be a stabilized allylic radical with significant electron density at both C8 and C6 (19a vs 19b). According to the substrate-cofactor disposition model, this should result in preferential hydroxyl rebound to C6 to yield 20, which was not observed.

While the contradiction could be addressed by either a concerted mechanism such that an allylic radical never actually forms or a stepwise epoxidation involving stabilization of the hydroxy-ferric complex via H bonding with the C9 hydroxyl group, comparison with the epoxidation of 6β-hyoscyamine may not be justified given the structural differences between 15 and 2. Therefore, an additional test was performed using the C7 methylidene analog 21 (Figure 3C). Hydrogen atom abstraction from C6 in this substrate analog will also produce an allylic radical in which the spin density is delocalized over C6 and C8. Therefore, the disposition model would predict C6 hydroxylation possibly in competition with epoxidation of the 7,8-double bond of 21 given the reported ability of H6H to also catalyze epoxidation of alkenes such as 6,7-dehydrohyo-scyamine (9, Figure 1C).^6,9

When H6H was incubated with 21 under the aforementioned conditions for 1 h, two new monooxygenated products were indeed observed (calcd m/z C₁₈H₂₃NO₄ [M + H]⁺ 318.1700, obsd 318.1701 and 318.1702) (Figure S2). Whereas NMR characterization indicated one of these products to be a diastereomer of the anticipated 7,8-epoxide 24, the other was determined to be the C8-hydroxylated species 23 in a ratio of 64:36. The C6-hydroxylated species 25, however, was not observed nor was further oxidation of 23 during the time course of the assay. Hence, apical versus lateral disposition of the C–H bond or subsequent carbon radical does not appear to be the only factor that determines the regiochemistry of rebound.

Consequences of C6 Substitution.

A C6O···OV distance of ca. 2.7 Å has been inferred from crystal structures of H6H bound with vanadium(IV) oxides and 6β-hydroxyhyoscyamine (2) (see Figure 2C),¹⁸ which is consistent with a potential H-bonding interaction between the C6 hydroxyl and the vanadium oxo ligand. If maintained in the ferryl complex or more importantly the subsequent hydroxy-ferric complex during turnover, then it could help to stabilize the latter against rebound during a stepwise epoxidation (Figure 1A). While HYSCORE studies of LolO catalysis have indicated direct coordination of the substrate to the ferryl iron predisposes the complex to cyclization,³⁹ coordination of the C6 hydroxyl of 2 directly to the ferryl iron as previously proposed²¹ seems unlikely, because the substrate appears to bind trans to the only other coordination site.¹⁸ However, second-sphere interactions between the iron and the C6 functional group may very well participate in mediating the epoxidation of 6β-hydroxyhyoscyamine (2).

In an effort to investigate the importance of these interactions more carefully, two analogs were prepared possessing either a methoxy (26) or a trifluoromethoxy (32) substituent at C6 (Figure 4). When 1.0 mM 26 was incubated with H6H under the aforementioned conditions including at least a 5-fold excess of 2OG, it was indeed almost completely consumed within 12 h without observable formation of the 7β-hydroxylated species 31. Instead, HPLC and electrospray ionization mass spectroscopy (ESI-MS) of the incubation mixture revealed the production of 2 and 3, which were confirmed by coelution with the corresponding standards (Figure S3). These observations revealed that the C6-methoxy group of 26 was removed in the presence of H6H to give 2, which was further oxidized to give 3 as per the established chemistry of H6H. Moreover, the production of formaldehyde (29) during the oxidation of 26 was confirmed by the detection of hydrazone (29-DNPH) during HPLC analysis of the product mixture following treatment with 2,4-dinitrophenylhydrazine (DNPH, Figure S3). Assuming apical positioning of the C6 methoxy substituent, its hydroxylation and subsequent elimination as formaldehyde from the hemiacetal 28 (Figure 4A) is fully consistent with the substrate-cofactor disposition model.

Figure 4.

(A) H6H reaction with the 6-methoxy derivative 26 results in demethylation of the methoxy moiety without C7 hydroxylation. (B) H6H reaction with the 6-trifluoromethoxy derivative 32 results in demethylation of the N-methyl moiety without C7 hydroxylation.

The trifluoromethoxy moiety of 32, however, cannot be further oxidized. Therefore, it was anticipated based on the disposition model that catalysis would be redirected to C7 hydroxylation (Figure 4B). However, 32 did not react to yield the expected product 38, despite being consumed in the presence of H6H. HPLC analysis of the reaction mixture instead revealed a product having a mass indicative of demethylation (calcd m/z for C₁₇H₂₀F₃NO₄ [M + H]⁺ 360.1417, obsd 360.1395, Figure S4). Upon treatment with excess acetic anhydride in pyridine (37 °C, 1 h), a species having a mass consistent with diacetylation was obtained (calcd m/z for C₂₁H₂₄F₃NO₆ [M + H]⁺ 444.1628, obsd 444.1630, Figure S4) indicating the presence of a second nucleophile in addition to the original secondary hydroxyl group on the side chain. Based on these results, the major product from incubation of 32 with H6H was assigned as 35 and the diacetylated derivative as 36. Consequently, the C6-OCF₃ functionality appears to redirect oxidation from C7 to the N-methyl. This would yield the hemiaminal 34 that subsequently eliminates formaldehyde, which was also detected by HPLC following treatment with DNPH (Figure S4).

H6H catalyzed N-demethylation of the natural substrates 1 or 2 has never been reported. Nevertheless, this result is consistent with the reported crystal structures, which place the N-methyl carbon of 1 at a similar distance to the metal (ca. 4.7 Å) and oxo ligand (ca. 3.8 Å) compared to carbons 6 and 7 (ca. 4.7 Å to the metal and ca. 3.3 Å to the oxo ligand).^17,18 Therefore, C6 substitution does appear to influence the regiochemistry of H atom abstraction between sites that are seemingly equally accessible. It should also be noted that all three sites (i.e., C6Hβ, C7Hβ & NCH₃) are too far from the putative oxo ligand observed in the crystal structures with the vanadium(IV) oxide complexes for H atom transfer.¹⁸ Indeed, computations suggest an O···H distance of less than 1.3 Å in the transition states for H atom transfer from both C6 as well as C7.¹⁷ This would require a roughly 1.5 Å displacement of the substrate from its position in the crystal structures to allow H atom transfer from either the N-methyl or the C7 methylene. Therefore, the H bond established between the C6 hydroxyl group and the iron oxo ligand may act as a pivot to guide the C7 hydrogen within sufficient proximity of the oxo ligand for H atom transfer.

Importance of the C6 H-Bond Donor.

In order to further test the importance of the putative H-bonding interaction between the C6 substituent and the oxo ligand, the substrate analog 6β-fluorohyoscyamine (39) was also prepared (see Supporting Information). Like 32, analog 39 has a nonoxidizable substituent at C6 that better approximates the size of a hydroxyl group but cannot serve as a H-bond donor to the oxo ligand of the ferryl (4) or hydroxy-ferric complexes (5). When 1.0 mM 39 was incubated with H6H and at least a 5-fold excess of 2OG, it was fully consumed under the standard reaction conditions; however, neither a demethylated product such as 44 nor a monooxygenated product such as 45 was observed (Figure 5A). Instead, LC-MS analysis showed the generation of two new products consistent with the ketone derivative 42 in addition to the C7 hydroxylated ketone derivative 43 (Figure S5).

Figure 5.

(A) H6H reaction with 6β-fluorohyoscyamine (39) results in defluorination via C6 hydroxylation (only one of the two possible C6-configurations of 41 is shown). The resulting 6-ketohyoscyamine subsequently undergoes H6H catalyzed hydroxylation at C7. Neither C7-hydroxylation nor N-demethylation of 39 is observed. (B) At least two possible mechanisms for C7-hydroxylation of 6-ketohyoscyamine are possible. Isotope labeling experiments, however, suggest hydroxylation via rebound (b) rather than epoxidation (a).

Synthetic preparation of 42 was used to verify H6H catalyzed oxidation of 39 to 42 (Supporting Information). Likewise, ¹⁹F NMR analysis of the incubation mixture revealed the release of free fluoride as a singlet resonance at −121 ppm with almost complete depletion of the multiplet fluorine peak of the substrate around −199 ppm (Figure S5c). Furthermore, incubation of 42 directly with H6H also led to production of the second product (calcd m/z for C₁₇H₂₁NO₅ [M + H]⁺ 320.1492, obsd 320.1510), and its assignment as 43 was further affirmed by analysis of acetylated derivatives (Figure S6). Finally, ¹H NMR spectra of the second product after partial purification demonstrated a singlet resonance at 3.49 ppm for C7–H indicative of the exo-configuration at C7. Thus, H6H catalyzes two sequential two-electron oxidations of 39 to yield 43 (Figure 5A).

Oxidation of 39 to 42 is consistent with H atom abstraction from C6 followed by hydroxyl rebound (40 → 41, Figure 5A). Hydrogen atom abstraction from C7 followed by [1,2]-migration of a H atom or hydride from C6 was ruled out when the C6-deuterated isotopolog of 39 (6α-[²H]39) reacted with H6H to again give 42 at natural abundance (Figure S7). Therefore, the putative H-bonding interaction indeed appears to play a role in guiding H atom abstraction from C7. In its absence, the C6α hydrogen may occupy an apical position suitable for H atom abstraction via the preferred σ-channel, which could be blocked by a large bulky moiety such as C6–OCH₃ or C6–OCF₃ resulting in direct hydroxylation of the former (26 → 27 → 28) and redirection to N–CH₃ hydroxylation by the latter (32 → 33 → 34) as discussed above. While the ketone product 42 lacks both a C6 H-bond donor and a blocking group, it also lacks an oxidizable C6 carbon resulting in C7 hydroxylation to yield 43 (Figure 5B). Alternatively, 42 may instead bind H6H as the hydrate (46) thereby again establishing the H-bonding interaction and promoting dehydrogenation of 46 to the unstable epoxide 49. The latter would then readily decompose to the observed product 43.

In an effort to distinguish between these possibilities, the reaction with 42 was run in the presence of either ¹⁸O₂/H₂O or O₂/H₂¹⁸O (97% ¹⁸O). These experiments respectively led to 27% and 67% incorporation of a single ¹⁸O center into 43 without the appearance of doubly labeled product (Figure S8). Had the reaction proceeded via the epoxide 49, then no oxygen should have been incorporated from O₂ (unless solvent exchange of the reduced remnants of O₂ in the H6H active site is very fast versus product release). Partial incorporation of the ¹⁸O₂ label suggests solvent exchange of the iron-oxo (or possibly the hydroxy-ferric ligand) prior to rebound, which has previously been reported with H6H and found to correlate inversely with the rate of H atom abstraction.⁴⁰ Similarly, the absence of multiple incorporation of ¹⁸O from H₂¹⁸O suggests formation and hydroxylation of the more stable keto radical 50 as opposed to the hydrated radical 47.

CONCLUSIONS

While epoxidation versus hydroxylation appears to correlate with angular positioning of the substrate radical following H atom abstraction,¹⁸ the relationship is not necessarily causative. Instead, the fate of the substrate radical once generated more likely stems from a combination of radical proximity and stabilization of the hydroxy-ferric complex. In particular, the putative H-bonding interaction between C6–OH of 6β-hydroxyhyoscyamine (2) and the ferryl complex implied by the crystal structures with vanadium(IV) oxides¹⁸ not only may function as a pivot to guide H atom abstraction from C7 as opposed to the N-methyl group but also inhibit subsequent rebound from the resulting hydroxy-ferric complex (Figure 6A,B). Disruption of these interactions can thus result in facile rebound to the most proximal locus of unpaired spin density. Nevertheless, formation of the H-bond between the C6-hydroxyl and the hydroxide ligand of the ferric iron requires lateral positioning of the C7 β-hydrogen during the epoxidation reaction. As a consequence, disposition of the substrate radical and the H atom abstraction channel (i.e., σ vs π) in the preceding step correlate with the observed reaction outcome.

Figure 6.

Mechanistic proposal for the two sequential oxidations catalyzed by H6H during the biosynthesis of scopolamine. (A) Hydroxylation at C6β proceeds via hydroxyl rebound from an unstabilized hydroxy-ferric complex. This involves displacement of the substrate toward the iron complex to reduce the distance between C6Hβ and the iron(IV)-oxo ligand, which may require proton transfer from the bridging tertiary ammonium ion to Glu116 in order to avoid charge separation as the salt bridge is disrupted. (B) Epoxidation proceeds stepwise with a H-bonding interaction between the C6OH and ferryl complex serving as a pivot to promote H_β atom abstraction from C7. This also may require proton transfer between the bridging ammonium ion and Glu116. The H-bonding interaction is maintained with the hydroxy-ferric complex stabilizing it against rebound. Cyclization of the radical 56 following H atom transfer may proceed stepwise (56 → 57 → 58) as shown or via a concerted proton coupled electron transfer event (56 → 58) without the intermediacy of 57.

This model chemistry also identifies the iron hydroxide ligand as the base which deprotonates the C6 hydroxyl during the cyclization reaction that follows H atom transfer (Figure 6B). While this stepwise mechanism is consistent with the reported solvent deuterium kinetic isotope effect that slows cyclization (57 → 58) in favor of hydroxide addition to the C7 cation, it does not explain the observed solvent deuterium kinetic isotope effect on decay of the ferryl complex.¹⁸ As demonstrated crystallographically, however, binding of hyoscyamine and its derivatives is primarily via hydrophobic interactions with the exception of a salt bridge between Glu116 and the protonated tertiary amine of the tropane ring.^17,18 Maintenance of this salt bridge draws the C6–C7 fragment away from the catalytic iron such that approach of C6 or C7 toward the ferryl oxo ligand would require either an overall compression of the active site or proton transfer from the nitrogen bridge to Glu116 in order to avoid charge separation (55 → 56). The latter event would induce a solvent deuterium isotope effect on H atom abstraction that is distinct from the effect on cyclization and thus consistent with the stepwise mechanism.

H6H thus remains an important case study in mechanistic enzymology, which continues to be investigated using a variety of experimental and computational approaches. While the details of its chemistry have yet to be definitively elucidated, a model that explains its mechanism of catalysis and the partitioning of radical intermediates between different reaction outcomes continues to form with increasing clarity. This model suggests a complex interplay between dynamic motions of the substrate, favored channels for H atom transfer based on substrate disposition and critical interactions between the substrate and intermediary iron complexes that contribute to determining the reaction outcome.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c17484.

• Additional experimental details, materials and methods, supporting results, spectra, and characterization of all novel chemical compounds (PDF)