Optimized Substrate Positioning Enables Switches in C–H Cleavage Site and Reaction Outcome in the Hydroxylation-Epoxidation Sequence Catalyzed by Hyoscyamine 6β-Hydroxylase

Abstract

Hyoscyamine 6β-hydroxylase (H6H) is an Fe(II)- and 2-oxoglutarate-dependent (Fe/2OG) oxygenase that produces the prolifically administered anti-nausea drug, scopolamine. After its namesake hydroxylation reaction, H6H then couples the newly installed C6 oxygen to C7 to produce the drug’s epoxide functionality. Oxoiron(IV) (ferryl) intermediates initiate both reactions by cleaving C–H bonds, but it remains unclear how the enzyme switches target site and promotes (C6)O–C7 coupling in preference to C7 hydroxylation in the second step. In one possible epoxidation mechanism, the C6 oxygen would – analogously to mechanisms proposed for the Fe/2OG halogenases and, in our more recent study, N-acetylnorloline synthase (LolO) – coordinate as alkoxide to the C7–H-cleaving ferryl intermediate to enable alkoxyl coupling to the ensuing C7 radical. Here we provide structural and kinetic evidence that H6H does not employ substrate coordination or repositioning for the epoxidation step but instead exploits the distinct spatial dependencies of competitive C–H-cleavage (C6 vs C7) and C–O-coupling (oxygen rebound vs cyclization) steps to promote the two-step sequence. Structural comparisons of ferryl-mimicking vanadyl complexes of wild-type H6H and a variant that preferentially 7-hydroxylates instead of epoxidizing 6β-hydroxyhyoscyamine suggest that a modest (~ 10°) shift in the Fe–O–H(C7) approach angle is sufficient to change the outcome. The observation that, in wild-type H6H, ²H₂O solvent also increases the C7-hydroxylation:epoxidation ratio by ~ 8-fold implies that the latter outcome requires cleavage of the alcohol O-H bond, which, unlike in the LolO oxacyclization, is not accomplished in advance of C–H cleavage.

Graphical Abstract

Introduction

The diversity and potential synthetic utility of reactions catalyzed by Fe(II)- and 2-oxoglutarate-dependent (Fe/2OG) oxygenases have made the enzyme superfamily an enticing platform for development of new biocatalysts.^1–3 The longest-known and most well-studied subclass of Fe/2OG enzymes, the hydroxylases (dioxygenases), can insert oxygen into strong C–H bonds (homolytic bond dissociation enthalpies of ~ 100 kcal/mol) with high regio- and stereo-selectivity.^4–5 Their conserved mechanism of action is largely understood (Scheme 1). In the reactions of Fe/2OG monooxygenases from plants, fungi, and bacteria, different fates of the Fe(III)–OH/R• state (alternatives to step 4) produced by hydrogen-atom transfer (HAT) from the substrate to the ferryl complex (step 3) lead to halogenation, desaturation, cyclization, ring-expansion and stereoinversion of aliphatic carbons.^6–10 With respect to biocatalysis and pharmaceutical synthesis, the abilities of certain Fe/2OG enzymes to install a halogen leaving group^6,11–12 for chemoselective fragment couplings or an epoxide electrophile^13–15 for covalent capture of biomolecular targets are among the most potentially valuable. The mechanism of the latter reaction type is less thoroughly understood than those of either hydroxylation or halogenation, providing the motivation for this study.

Scheme 1.

Consensus mechanism for hydroxylation catalyzed by Fe/2OG dioxygenases.

The structure of the Fe/2OG halogenase SyrB2 from the syringomycin E biosynthetic pathway showed that a crucial structural adaptation enabling halogenation is the absence of the carboxylate ligand of the His₂Asp/Glu “facial triad” of protein iron ligands¹⁶ that had been viewed as conserved across the enzyme superfamily.¹⁷ Coordination of halide in its place allows the R• produced by the HAT step to couple to the cis halogen (Cl•/Br•) of the cis-Cl/Br–Fe(III)–OH complex. The behavior of SyrB2 with altered substrates revealed how the halogenase favors this alternative radical coupling over the oxygen-rebound step that is facile in related hydroxylases: an unusual disposition of the prime substrate’s scissile C–H bond relative to the haloferryl complex results in very slow HAT to the intermediate but then selective attack of the resulting R• on the cis halogen.¹⁸ Substrate modifications perturbing this substrate-cofactor disposition (SCD) unleashed more rapid HAT and competing hydroxylation,¹⁸ and a particular substrate underwent faster HAT and hydroxylation at one site but slower HAT and primarily chlorination at the adjacent carbon.¹⁸ A spectroscopic study provided direct structural evidence for the correlation among SCD, HAT efficiency, and reaction outcome.¹⁹ Subsequent computational studies have generally comported with this hypothesis,^19–21 and several proposed that an “off-line” iron coordination geometry^22–23 might be the key to establishing the special chlorinating SCD.^21,24–26 In the off-line geometry, the oxygen of the key intermediates would occupy the site trans to the N-terminal histidine ligand (His_N) and be directed nearly perpendicular to the vector between the iron cofactor and prime substrate’s scissile C–H bond rather than the conventional “in-line” position trans to His_C and directed toward the C–H bond.

The two crucial control elements identified in halogenases – (1) an open coordination site on the iron cofactor suitably disposed to allow for coupling to the R• and (2) a special SCD that disfavors competing oxygen rebound – could also be relevant to oxacyclization reactions catalyzed by Fe/2OG enzymes (Scheme 2). Hyoscyamine 6β-hydroxylase (H6H), N-acetylnorloline synthase (LolO), and clavaminate synthase (CAS) all couple substrate oxygen atoms, which they install in prior hydroxylation reactions, to neighboring carbons, thus forming epoxide, oxolane, and oxazolidine moieties, respectively.^8,15,27–30 In analogy to the halogenases, coordination of the alcohol oxygen, leading to its deprotonation, could promote alkoxyl coupling to the carbon-centered radical of the substrate to form the oxacycle. Although the facial-triad carboxylate ligand is conserved in these three enzymes, decarboxylation of 2OG during formation of the ferryl intermediate vacates a site and could potentially allow for alkoxide coordination during the reaction. Prime substrate coordination could cause a change in the SCD to make oxygen rebound unfavorable and R• ↔ alkoxyl coupling favorable, possibly by driving the intermediate’s (hydr)oxo ligand off-line.

Scheme 2.

Epoxidation and oxacyclization reactions catalyzed by H6H, LolO, CAS.

These hydroxylase/oxacyclase enzymes overcome the additional challenge of removing hydrogens from two different substrate carbons in sequential reactions. In principle, a precise, static SCD could ensure cleavage of the C–H bonds in the proper sequence. In this scenario, the substrate of the hydroxylation step would have both carbons poised for HAT to the first ferryl complex, but the site of hydroxylation (the first C–H bond to be cleaved) would be positioned more favorably, ensuring that it is preferentially targeted in the first step. After installation of the hydroxyl group at the first site, the less well-positioned second site could then donate H• to the second ferryl complex, albeit less rapidly. In this scenario, one might anticipate some off-pathway hydroxylation of the second site in the first reaction. Indeed, prior work with both H6H and LolO showed that out-of-order hydroxylation of the second site does occur.^31–32

In this work, we used kinetic, structural, and spectroscopic approaches to interrogate both scopolamine-biosynthetic steps – 6β hydroxylation of hyoscyamine (Hyo) and 6,7-exo-epoxidation of 6β-hydroxyhyoscyamine (6-OH-Hyo) – catalyzed by H6H to determine how the enzyme switches both the C–H-cleavage site and the outcome in its consecutive reactions. The results imply that H6H uses an elegantly simple strategy, in which the different intrinsic spatial dependencies of the competitive C–H-cleavage (C6 versus C7) and C–O-coupling (rebound/hy-droxylation versus epoxidation) steps are leveraged to enforce the two-reaction sequence in the absence of either major adjustment of SCD or alkoxide coordination in the second reaction. The results thus extend the principal conclusion of our studies on the halogenases – that SCD is of primary importance in the control of Fe/2OG-oxygenase regioselectivity and outcome – even to this case of consecutive reactions targeting different carbons of the substrate for distinct outcomes.

Results

Analysis of hyoscyamine hydroxylation by Atropa belladonna (Ab) H6H.

We previously used stopped-flow absorption (SF-Abs) experiments to quantify the contributions of various pathways of ferryl-intermediate decay to its overall kinetics in the hydroxylation reaction catalyzed by H6H from Hyoscyamus niger (HnH6H).³² Here, we carried out most of the experiments on the 90% identical (95% similar) homolog from Atropa belladonna (AbH6H), because it afforded better yields in heterologous overexpression and purification (Figures S1–S4). Therefore, we repeated the published kinetic analysis of the hydroxylation step on AbH6H (Scheme S1). The SF-Abs experiments gave rate constants for HAT to the ferryl complex from C6 and C7 (k_C6H and k_C7H) of 10 ± 0.1 s⁻¹ and 0.3 ± 0.1 s⁻¹ (5 °C), respectively (Figure S5A, Table S1), and the ratio of C6 to C7 hydroxylation of 97:3 determined by LCMS analysis of the products agreed well with these rate constants (Figure S5B, Table S2). The Hn and Ab orthologs thus behave almost identically (k_C6H = 14 s⁻¹, k_C7H = 0.5 s⁻¹; C6:C7 hydroxylation = 98:2 for HnH6H).

Equivalence of the rate-constants for HAT from C7 in the hydroxylation and epoxidation reactions.

Rapid mixing of the epoxidation reactant complex (AbH6H•Fe^II•2OG•6-OH-Hyo) with a buffer solution containing sub-stoichiometric O₂ (i.e., under single-turnover conditions) resulted in the transient ultraviolet absorption with maximum change near 320 nm (Figure S6) characteristic of a ferryl intermediate (as confirmed by freeze-quench Mössbauer experiments below). Use of 6-OH-Hyo bearing deuteria on C7 (exo and endo) as well as C1, C5, and C6 (endo) (d₅-6-OH-Hyo, Scheme S2; see Synthesis of Deuteriologs in Materials and Methods) markedly increased the amplitude and slowed decay of the transient feature (Figure S6). This substrate D-KIE originates exclusively from the C7 deuteron, as traces from reactions with 7-exo-d₁-6-OH-Hyo and d₅-6-OH-Hyo were effectively indistinguishable (Figure S7). The sluggishness of the epoxidation reaction – especially with the 6-OH-Hyo deuteriologs – pushes the kinetic traces into a time regime (100–1000 s) that is plagued by a consistent physical artifact with our stopped-flow instrument. This artifact is more obvious in traces with small absorbance changes (Figure S6) than in those with larger absorbance changes (Figure S7). Although the crucial D-KIE is still visible in the raw data from the small-ΔA₃₂₀ single-turnover reactions of Figure S6, this artifact would have interfered with more quantitative analysis, and so we subtracted polynomial backgrounds (Figure S8, black solid lines) to obtain unadulterated traces (Figure 1).

Figure 1.

Impacts of substrate and/or solvent deuteration on the kinetics of formation and decay of the ferryl complex during the cyclization reaction catalyzed by AbH6H. The four traces were globally simulated according to a two-step model with a second-order formation rate constant of 0.11 mM⁻¹s⁻¹ and four different observed decay rate constants: 0.49 ± 0.05 s⁻¹ for 6-OH-Hyo in H₂O, 0.35 ± 0.04 s⁻¹ for 6-OH-Hyo in D₂O, 0.11 ± 0.01 s⁻¹ for d₅-6-OH-Hyo in H₂O, and 0.06 ± 0.01 s⁻¹ for d₅-6-OH-Hyo in D₂O. The final concentrations after mixing were 0.55 mM AbH6H, 0.5 mM Fe, 5 mM 2OG, 5 mM 6-OH or d₅-6-OH, and 0.19 mM O₂. A molar absorptivity of 1.7 mM⁻¹cm⁻¹ provided optimal agreement and is consistent with previous studies.^8,32–36

The rate constant for decay of the second (epoxidation) ferryl complex obtained by fitting and simulating the corrected SF-Abs trace for the reaction with protium-bearing 6-OH-Hyo is 0.5 ± 0.1 s⁻¹, ~ 20 times less than the 10 ± 0.1 s⁻¹ seen for decay of the first ferryl intermediate. Use of 6-OH-Hyo substrate bearing deuterium at C7 slowed decay of the ferryl complex to 0.1 ± 0.02 s⁻¹, an observed D-KIE of ~ 5. In this reaction, decay of the ferryl intermediate is largely unpro-ductive, as the Sco:2OG stoichiometry with d₅-6-OH-Hyo is only ~ 40% of the value obtained with the protium-bearing 6-OH-Hyo (Figure S9). The ~ 60% uncoupled decay implies that the actual rate constant for productive deuterium transfer is only ~ 0.04 s⁻¹ (0.1 s⁻¹ × 40%) and that there is at least one unproductive decay pathway contributing ~ 0.06 s⁻¹. This parsing implies an intrinsic D-KIE of > 10, which is more in line with those for HAT from carbon to ferryl complexes in other Fe/2OG enzymes.^7–8,33,37 It also implies that k_C7H in the epoxidation reaction (0.4 ± 0.1 s⁻¹) is essentially unchanged from k_C7H (0.3 ± 0.1 s⁻¹) in the hydroxylation reaction.

Single ferryl complexes with normal Mössbauer parameters in the two H6H reactions.

We used freeze-quench (FQ) Mössbauer spectroscopy to confirm the identities and compare the parameters of the UV-absorbing ferryl species in the two reactions of AbH6H (Figure 2 and Table S3–S6). Use of deuterated substrates and super-saturating O₂ (generated by the chlorite/chlorite dismutase system)^38–39 enabled accumulation of the intermediates to high levels. The Mössbauer spectra of these freeze-quenched samples acquired at 4.2 K without an applied magnetic field (Figure 2A, B) reveal decay of the quadrupole-doublet signature of the reactant complex (top spectra) and development of a new sharp quadrupole doublet – with parameters typical of ferryl complexes (middle spectra) – following initiation of either reaction. The integrated areas of the decaying and developing features match and, at reactions times of 0.45 s for the hydroxylation reaction (Figure 2A) and 3 s for the epoxidation reaction (Figure 2B), correspond to ~ 70% of the total ⁵⁷Fe absorption (Figure 2C). At increased reaction times, the narrow ferryl-associated doublets can be seen to decay and Fe(II)-associated features to re-develop (Figure 2A,B, middle and bottom spectra). In each reaction, the transient features can be adequately accounted for as a single quadrupole doublet with parameters typical of ferryl complexes detected in other Fe/2OG enzymes^{7–8,37,40–41} (with the exception of the oxacyclization step of LolO discussed in a parallel study⁴²): δ = 0.32 mm/s and |ΔE_Q| = 0.85 mm/s for the hydroxylation reaction, and δ = 0.27 mm/s |ΔE_Q| = 0.86 mm/s for the epoxidation reaction (Figure S11). An overlay of the derived spectra of the ferryl complexes from the two AbH6H reactions illustrates their similarity, which contrasts with the two distinct ferryl intermediates observed in LolO (Figure S12).⁴²

Figure 2.

FQ-Mössbauer spectroscopic analysis of the two reactions catalyzed by AbH6H. A-B: Spectra (4.2-K/0 mT) of samples freeze-quenched at relevant reaction times in the hydroxylation (A) and oxacyclization (B) reactions. The final concentrations after mixing were 1.0 mM AbH6H, 0.93 mM Fe(II), 4.0 mM 2OG, 24 μM Cld, 15 mM NaClO₂ and 2.0 mM of either 6-d₁-Hyo (A) or d₅-6-OH-Hyo (B). C: Fe^IV concentrations determined by analysis of the Mössbauer spectra in panels A and B. Figure S10 shows that these kinetics approximately match the predicted kinetics of accumulation and decay deduced from SF-Abs experiments under single-turnover conditions. Difficulty in determining the (supersaturating) O₂ concentration over time in experiments with the Cld/ClO₂^– O₂-generation system precludes a more precise assessment of the kinetic consistency of results from these multiple-turnover experiments with those from the single-turnover SF-Abs experiments.

Crystal structures of the AbH6H active site implying invariant SCD in its consecutive steps.

We assessed the possibility of different SCDs in the two reactions more directly by structural methods. X-ray crystallography afforded high-resolution global views of the relevant iron complexes, and complementary hyperfine sublevel correlation (HYSCORE) spectroscopy experiments on the H6H•vanadyl•succinate•(6-OH-)Hyo (succinate is hereafter abbreviated succ) complexes afforded both accurate active-site metrics relevant to reactivity and, importantly, a cross-check of the more global geometries seen in the crystal structures, but with the metal in the +IV oxidation state. A previously reported X-ray crystal structure of another H6H homolog from Datura metel (DmH6H) in complex with 2OG and Hyo had Ni(II) in place of Fe(II), presumably to prevent turnover of the complex during crystallization and data collection.⁴³ Neither the structure of the authentic O₂-reactive H6H•Fe^II•2OG•Hyo hydroxylation complex nor any structure with the oxacyclization substrate, 6-OH-Hyo, bound has been reported. Our six structures of wild-type (wt) AbH6H in complex with either its iron cofactor or the vanadium mimic [oxovanadium(IV), prior to X-ray exposure] and appropriate combinations of substrates and products range in resolution from 1.53–1.79 Å, and they all show well-defined electron density for the metal ion, its ligands, and the substrates/products (Figure 3; statistics shown in Tables S7 and S8). Together, they confirm the expectation that the prime substrates for the two reactions would share a common binding site, and they reveal a nearly invariant SCD. The visualized atoms of the common tropane core of the Hyo and 6-OH-Hyo substrates occupy almost precisely the same positions in all six structures (Figure S13). For example, comparison of the structures of the two reactant complexes [AbH6H•Fe^II•2OG•Hyo and AbH6H•Fe^II•2OG•6-OH-Hyo; Figure 3A, D] reveals only a very minor (0.3 Å) shift of C6 relative to its position in the Hyo complex as a result of an interaction of the distinguishing C6 hydroxyl group of the oxacyclization substrate with 2OG. As seen in the prior structure of the DmH6H•Ni^II•2OG•Hyo complex,⁴³ the 2OG carboxylate coordinates trans to H273 (His_C), facing the C6 position of Hyo (Figure 3A,D). This binding mode is characterized as off-line,^22–23 because it leaves the presumptive O₂-addition site (occupied in the structures by a water ligand) roughly trans to H217 (His_N) and perpendicular to the Fe-C6 vector.

Figure 3.

Active site of AbH6H from crystal structures of (or mimicking) six different states in its hydroxylation-epoxidation reaction sequence. The six panels show zoomed views of the active-site facial triad, the metal cofactor (or mimic), the co-substrate/co-product, and the prime substrate/product from structures of the reactant (A, D), vanadyl-derived (B, E), and product (C, F) complexes for the hydroxylation (A-C) and cyclization (D-F) reactions. In each panel, a 2F_o-F_c electron density map contoured to 1.0 σ is shown as gray mesh. Selected side chains, water molecules, cofactors, substrates, co-substrates, and products are shown in ball-and-stick or sphere representation and colored according to atom type.

Prior work on other Fe/2OG enzymes has shown that the stable vanadyl complexes structurally mimic the ferryl intermediates, both by adopting active-site configurations that correlate better with observed reaction outcomes than do the configurations of the reactant complexes and by reproducing substrate-association/dissociation dynamics of the ferryl complexes more closely than do the reactant complexes.^{7,19,25,44–47} We therefore assessed the possibility that outcome-controlling structural differences between the hydroxylation and oxacyclization complexes might emerge during ferryl intermediate formation by comparing the structures of H6H complexes harboring the co-product (succ), the (6-OH-)Hyo prime substrate, and oxovanadium(IV) (vanadyl) in place of iron. Structures obtained from crystals of the AbH6H•V^IVO•succ•Hyo and AbH6H•V^IVO•succ•6-OH-Hyo complexes (the latter representing the first use of the vanadium cofactor mimic to provide structural insight in a native non-hydroxylation reaction^45–46,48) indicate that, as in the prior studies, the otherwise stable vanadyl ion undergoes photoreduction to V(III) or V(II) during data collection (Figure 4A and B).^45–46 Specifically, the shortest V-O distance of 2.1 Å is longer than the expected 1.6-Å V^IV≡O bond of vanadyl,⁴⁴ implying a lower metal oxidation state.⁴⁴

Figure 4.

Comparison of the active sites in the vanadyl-derived complexes intended to mimic the ferryl intermediates of the hydroxylation (A) and epoxidation (B) reactions. The gray mesh illustrates 2F_o-F_c electron density maps (contoured to 1.0 σ) for the vanadium and its two non-protein oxygen ligands. Selected side chains, water molecules, cofactors, substrates, co-substrates, and products are shown in ball-and-stick or sphere representation and colored according to atom type.

Surprisingly, there are two distinct lobes of electron density attributable to non-protein (presumably, oxygen) ligands in both vanadyl-derived structures. One of the lobes occupies the in-line position, trans to H273 (His_C), and projects toward the substrate. The second is cis to the first, in the off-line position, trans to H217 (His_N). Interrogation of both ferryl-mimicking complexes by HYSCORE spectroscopy (presented below) shows that the in-line oxygen ligand is derived from the vanadyl oxo, implying that the off-line oxygen is derived from water. The distance between the metal and C6 hydroxyl oxygen in the structure of the AbH6H•V•succ•6-OH-Hyo complex (4.0 Å) and the conserved SCD with the two substrates imply that the oxygen does not coordinate to the cofactor in the cyclization reaction. This top-line conclusion contradicts two prior proposals of halogenase-like radical coupling mechanisms involving alkoxide coordination.^31,49 The C6 oxygen instead engages in hydrogen bonding with the in-line (vanadyl-derived) oxygen. Although the very short (2.3-Å) separation between them in the experimental structure (Figure 4B) might be, in part, an artifact of the V–O elongation that occurs upon photoreduction, the analogous substrate-cofactor interaction could also mirror events in the actual oxacyclization reaction mediated by the second ferryl complex, as discussed below.

HYSCORE spectra revealing in-line vanadyl and indistinguishable SCD in complexes with either prime substrate.

To determine the origins of the two oxygen ligands seen in the structures and obtain independent active site metrics, we used HYSCORE spectroscopy to interrogate the dipolar interaction between the S = 1/2 vanadyl ion and deuterium (²H) nuclei placed at different locations in the H6H prime substrates. For the hydroxylation complex (with Hyo), we used three different substrate deuteriologs – 6-d₁-, 7-d₁-, and 8-d₃-Hyo (C8 is the N-methyl). As described previously,^7,19,45 the splitting in the HYSCORE spectra can be related to the V–²H distance. Analysis of the spectra of the HnH6H•V^IVO•succ•Hyo complexes yielded distances of 3.8 ± 0.3, 3.3 ± 0.1, and 3.3 ± 0.1 Å from the metal to the deuterium atoms on C6, C7, and C8, respectively (Figure 5 shows representative data; Figures S15–S21 show the full set of field-dependent ²H-HYSCORE spectra with simulations). To compare these distances to those predicted from the X-ray crystal structure, we added hydrogens to C6–8 of Hyo at standard C–H distance and angles (Figure 5, right). The ²H-HYSCORE metrics agree well with the corresponding distances of 4.2 ± 0.2, 3.5 ± 0.2, and 3.6 ± 0.2 Å in the hydrogens-added model (Table 1). For the HnH6H•V^IVO•succ•7-d₁-Hyo complex, the angle relating the g-tensor of the vanadyl spin probe and the metal-deuterium vector was also accessible by analysis of the magnetic field-dependence of the spectra (Figure S15–S21). The g-tensor of vanadyl aligns with the V≡O bond, allowing the spin-frame angle to be straightforwardly related to the molecular frame (Figure 6). The O–V–²H(C7) angle obtained through this analysis is 40 ± 5° (Figure 6, A–D, middle).

Figure 5.

²H-HYSCORE spectra of HnH6H•VO•succ•d_x-Hyo complexes containing 6-d₁-Hyo (A), 7-d₁-Hyo (B), 8-d₃-Hyo (C), or unlabeled hyoscyamine (D). The red-dashed lines encircle the expected positions of signals arising from deuterium nuclei. Spectra were collected at 396 mT and 35 K with microwave frequency 9.434 GHz. (E) V–O-contracted AbH6H•V^IVO•succ•Hyo crystallographic model, with the three measured distances shown for comparison to the values determined from the HYSCORE spectra. The vanadium, oxygen ligands, and substrate are shown in ball-and-stick or sphere representation and are colored according to atom type.

Table 1.

Distances and angles measured by ²H-HYSCORE spectroscopy and cognate measurements derived from X-ray crystal structures for comparison.

	2H-HYSCORE	Crystal Structures
Hyo
V–2H(C6) (Å)	3.8 ± 0.3	4.2 ± 0.2
V–2H(C7) (Å)	3.3 ± 0.1	3.5 ± 0.2
V–2H(N-Me) (Å)	3.3 ± 0.1	3.6 ± 0.2
O–V–2H(C7) (°)	40 ± 5	49.2 ± 5a
V-O/C7-2H Angle (°)	25 ± 2	26.1 ± 0.2
6-OH-Hyo
V–2H(C7) (Å)	3.4 ± 0.1	3.5 ± 0.1
O–V–2H(C7) (°)	40 ± 5	51.8a
V-O/C7-2H Angle (°)	25 ± 2	25.4 ± 0.2

^aAfter contraction of the in-line V-O bond to 1.6 Å

Figure 6.

Comparison of ²H-HYSCORE patterns observed for the vanadyl complex of HnH6H with either 7-d₁-Hyo (upper spectra) or 7-d₁-6-OH-Hyo (lower spectra) at (A) 280 mT, (B) 396 mT, (C) 344 mT, and (D) 340 mT. Orientation selectivities probed by each magnetic field position (blue plots at top of A-D) are color coded (red - fully excited; blue, not excited). Magnetic field positions at which the spectra were acquired are indicated by black arrows on the one-dimensional EPR spectrum (very top). Panel E shows the metrics extracted from simulation of the ²H-HYSCORE data, all identical for both substrates, including the O–V–²H(C7) angle (40°), the V–²H separation (3.3 Å) and the angle between the C7–²H bond and the V–O vector (25°). Panels F and G show the V–O-contracted AbH6H•VO•succ•(6-OH-)Hyo crystallographic models, with measurements displayed to enable comparison to the values obtained by the ²H-HYSCORE analysis. The vanadium, oxygen ligands, and substrate are shown in ball-and-stick or sphere representation and colored according to atom type.

In the hydrogens-added AbH6H crystallographic model, the O–V–²H(C7) angle is ~ 50° for the in-line oxygen ligand but ~ 130° for the off-line oxygen ligand (Figure 6F and S22). Only the former value is consistent with the HYSCORE spectra (Figure 6, E–G and Table 1).^37,50–52 In the event that the off-line oxygen ligand were actually the vanadyl oxo, this ~ 130° angle would have resulted in patterns (Figure S23) very different from those observed in Figure 6. The HYSCORE data thus establish that, of the two non-protein oxygen ligands seen in the structure of the AbH6H•V•succ•Hyo complex, the in-line oxygen is derived from the vanadyl oxo group, implicating (as expected) an in-line ferryl complex in the hydroxylation reaction.

The equivalent field-dependent HYSCORE spectra for the HnH6H•V^IVO•succ•7-d₁-6-OH-Hyo (epoxidation) complex are essentially identical to those for the HnH6H•V^IVO•succ•7-d₁-Hyo (hydroxylation) complex (Figure 6, A–D, compare bottom to middle). Together with the crystal structures, the HYSCORE data establish that the hydroxylation and cyclization substrates are positioned almost identically. Moreover, the data are again irreconcilable with an off-line vanadyl, which would lead to very different HYSCORE spectra (Figure S24). Indeed, the HYSCORE data impose tight restraints on the possible positions of the vanadyl oxo in both complexes (Figure S25, red mesh).

The HYSCORE analysis of 7-d₁-(6-OH-)Hyo-containing complexes also allowed us to extract reasonably precise estimates of ²H nuclear quadrupole-coupling parameters, yielding the angle between the C7–²H and V–O vectors (25 ± 5°) in both complexes. This estimate agrees well with that from the crystal structures (26° in both complexes; Figure 6). The relatively weak signals for the C6 deuteron precluded a meaningful estimate of either the O–V–²H(C6) angle or the C6–²H/V–O vector angle, but the data are consistent with the small angles seen in the crystallographic models (10–20°). The ²H-HYSCORE results are thus fully consistent with those obtained by crystallography and, importantly, with an in-line binding mode for the vanadyl ion.

Target-hydrogen approach angles in vanadyl-derived crystal structures of H6H and other Fe/2OG enzymes.

To better correlate the observed chemistry and HYSCORE data with the crystal structures, we further analyzed the structures under the assumption that V(IV) photoreduction at 100 K in the crystal causes only V–O bond elongation. This approximation has support from a prior study.⁴⁵ We thus manually shortened the bond back to 1.6 Å, the M–O-bond length determined for both vanadyl and ferryl intermediates,^37,50–52 by moving the oxygen toward the metal (Figure 7A, B). Interestingly, although HAT from C6 of Hyo to the hydroxylation ferryl complex is considerably faster (10 s⁻¹) than HAT from C7 (0.3 s⁻¹), the C6 and C7 exo hydrogens are nearly equidistant from the oxo ligand in the resultant vanadyl/ferryl model (2.7 Å versus 2.8 Å, respectively), suggesting that the difference in their rate constants is related to their different approach angles. The approach angle for the C6 exo hydrogen [V–O–H(C6); note that this angle is different from, but straightforwardly related to, the O–V–²H angle afforded by HYSCORE and shown in Figure 8] is 153°, which is considerably larger than the approach angle of the C7 exo hydrogen [V–O–H(C7) = 104°]. The larger angle for the preferred hydrogen is consistent with computational studies suggesting that the angle between the Fe^IV=O bond and the substrate hydrogen controls the FMOs engaged (σ- versus π-pathway) and thus the HAT barrier height, with σ-pathway-enabling obtuse angles giving more efficient HAT.^21,53–54 More recent analysis has suggested that, even within the σ manifold, more acute angles lead to slower HAT.²⁶

Figure 7.

Comparison of key metrics in the structural models of the (A) AbH6H•V^IVO•succ•Hyo and (B) AbH6H•V^IVO•succ•6-OH-Hyo complexes obtained by adding hydrogens to the experimentally observed prime substrates and shortening the observed V-O distances to 1.6 Å. Selected side chains, water molecules, cofactors, substrates, co-substrates, and products are shown in ball-and-stick or sphere representation and colored according to atom type.

Figure 8.

Depiction of the disposition of targeted C–H bonds relative to the metal-oxo cofactor from either (A) vanadyl-derived, V–O-bond contracted crystallographic models of enzymes with known HAT rate constants or ¹⁸O incorporation^{32,35,45–46,48} or (B) ²H-HYSCORE spectroscopy.^7,19,45–46 Panel A shows the seven reactions shown in Table 2. The approach angle of H to the vanadyl/ferryl unit correlates with both the HAT rate and % ¹⁸O incorporation, while the O-H distance and the angle between the C–H and metal-oxygen (M–O) vectors do not. The O-H and O-C distances, V-O-H angles, and V–O vector/C–H vector angles are drawn to scale; the ferryl bond length is 1.6 Å and each C–H bond is 1.09 Å. Carbon and hydrogen atoms are depicted as white or colored circles, respectively. Panel B shows the distance and angle determined by ²H-HYSCORE spectroscopy utilizing either an [Fe-NO]⁷ complex (SyrB2¹⁹) or the vanadyl ion (all others^7,45–46) as a spin probe. ²H-HYSCORE directly measures the O-M-²H angle and the M-²H distance; these two metrics along with the metal-oxygen bond length of 1.6 Å forms a triangle that readily allows the M-O-²H angle to be determined, and these angles are depicted in the figure to simplify visual comparison to the metrics derived from crystallography. The experimentally determined O–M–H angle and the calculated M–O–H angle for C7 of H6H are both indicated as an example.

To further develop this correlation, we compared our vanadyl-derived ferryl models of H6H to those that we could generate from reported structures of other Fe/2OG enzymes^45–46 and one additional structure that we solved recently (PDB ID 6VWQ⁴⁸). The analogously formulated vanadyl-derived models of the hydroxylating ferryl complexes in TauD,⁴⁶ VioC,⁴⁵ and CAS⁴⁸ (in complex with deoxyguanidinoproclavaminate) all exhibit V–O–H angles (132–142°) and (V)O–H distances (2.0–2.5 Å) more similar to those for the C6 exo hydrogen than to those for the C7 exo hydrogen in the AbH6H hydroxylation complex (Table 2, Figure 8A and S26). Accordingly, the hydroxylation reactions of TauD,³³ VioC,³⁵ and CAS (Figure S27) are all associated with relatively fast HAT steps, in addition to high incorporation of ¹⁸O from ¹⁸O₂ into the prime product (Figure 8A and Table 2), which we previously showed to be a hallmark of efficiency in the HAT and oxygen-rebound steps.⁵⁵ In the vanadyl-derived model of the AbH6H oxacyclization complex with 6-OH-Hyo bound, the disposition of the C7 exo hydrogen remains nearly identical [O–H(C7) = 2.8 Å, V–O–H(C7) = 101°] to that in the hydroxylation model with Hyo bound, comporting with the similarity of the measured rate constants for C7–H cleavage in the two reactions.

Table 2.

Metrics from vanadyl-complex-derived crystal structures of Fe/2OG enzymes along with measured HAT rate constants and fractions of ¹⁸O incorporation from ¹⁸O₂ into hydroxylated products.

	Enzymea	TauD	VioC + L-Arg	CAS + DGPCb	AbH6H (C6) Hyo	AbH6H (C7) Hyo	AbH6H (C7) 6-OH-Hyo	L289F AbH6H (C7) 6-OH-Hyo
	V-O Distance (Å)	1.9d	1.9e	1.9f	2.1g	2.1g	2.1g	1.7g
Unaltered (V-O > 1.6 Å)	V-O-C Angle (°)	136	132	134	139	113	108	127
	(V)O-H(C) Distance (Å)	1.8	2.4	2.1	2.3	2.6	2.7	2.8
	V-O-H Angle (°)	125	129	137	151	97	91	111
Altered (V-O = 1.6 Å)	(V)O-H(C) Distance (Å)	2.0	2.5	2.3	2.7	2.8	2.7	2.8
	V-O-H Angle (°)	132	134	142	153	104	101	112
	V-O/C-H Angle (°)	28	57	53	78	26	26	16
	k HAT (s−1) c	15h	10i	18	10	0.3	0.4	N.D.
	18O from 18O2 (%)	N.D.	98i	96	87j	34j	13	42
	PDB	6EDH	6ALR	6VWR	8CV9	8CV9	8CVC	8CVH

^aFe/2OG enzymes structurally characterized in their V•succ•substrate complexes for which either % ¹⁸O or k_HAT reported here or previously;

^bHydroxylated at C3;

^c5°C;

^dReference 53;

^eReference 15;

^fReference 56;

^gThis work;

^hReference 8;

ⁱReference 44;

^jReference 43.

Structure of Sco product complex confirms that observed SCD is productive.

As one additional corroboration that the active-site configurations seen in the crystal structures of the reactant and ferryl-mimicking vanadium complexes with 6-OH-Hyo accurately reflect the configurations that yield the epoxidation outcome, we also solved the structure of the AbH6H•Fe^II•succ•Sco product complex (Figure 3F). The fact that the tropane core of the epoxide product closely overlays with that of the 6-OH-Hyo substrate in the reactant and intermediate-mimicking complexes (Figure S13C) again implies that substrate repositioning within the active site is not integral to the enzyme’s switch in reactivity to C7 targeting and epoxidation. It further suggests that this nearly constant SCD might be special – and possibly even unique – in conferring competency for both reactions in the sequence without significant active-site reconfiguration.

Sterically perturbing L289F substitution disables epoxidation, unleashes second hydroxylation.

To test this hypothesis, we introduced sterically perturbing amino acid substitutions close to the prime substrate (as described in the Supporting Information) in search of variants fully competent for C6 hydroxylation but disabled in the subsequent epoxidation. The L289F variant (Figure S28) exhibited the desired phenotype. We compared its product profile to that of wt AbH6H in reactions with varying 2OG:Hyo ratios. The wt enzyme exhibited the expected dominant conversion of Hyo to 6-OH-Hyo at 2OG:Hyo ratios less than one and conversion of 6-OH-Hyo to Sco as the 2OG:Hyo ratio was further increased (Figure S29). In this titration, a residual peak at the elution time and m/z value of the Hyo substrate remained, even at 2OG:Hyo = 3, suggesting the presence of a contaminant with a similar structure and identical mass (e.g., the enantiomer of Hyo). We also detected three minor products (Figure S29, middle panels), which we identified by their Δm/z values and co-elution with standards as 7-OH-Hyo (Δm/z = +16), 6,7-dehydro-Hyo (Δm/z = −2), and 6,7-(OH)₂-Hyo (Δm/z = +32). Because the small quantity of 7-OH-Hyo that is formed by out-of-order C7 targeting in the first reaction is not significantly consumed under these experimental conditions (Figure S29, yellow bars; H6H will convert 7-OH-Hyo to Sco under more forcing conditions^31,49), the 6,7-(OH)₂-Hyo product must reflect C7 hydroxylation in competition with 6,7-epoxidation in the second reaction. Indeed, the dihydroxylated product was also readily detected in AbH6H reactions with 6-OH-Hyo (Figure S29, bottom left panel). In either the two-step conversion starting from Hyo or the direct conversion of 6-OH-Hyo, the epox-ide product was found to predominate by a factor of ~ 25 over the 6,7-dihydroxylation product.

Analogous assays of the AbH6H L289F variant revealed that it retains 6β-hydroxylase activity toward Hyo and exhibits regioselectivity for C6 over C7 similar to that of wt AbH6H. LCMS analysis gave a product distribution of 95 ± 1.2% 6-OH-Hyo, 3 ± 0.7% 7-OH-Hyo and 2 ± 0.9% 6,7-dehydro-Hyo following reactions of the variant with 1 equiv Hyo and 0.3 equiv 2OG at 4 °C for 1 h under air. By contrast to its wt-like behavior toward Hyo, we found the L289F variant enzyme to be drastically compromised in epoxidation of 6-OH-Hyo and instead to promote primar-ily 7-hydroxylation in its second reaction, yielding 3:1 6,7-(OH)₂-Hyo:Sco (Figure 9), a ~ 70-fold increase in the partition ratio. The observation that the dihydroxylation product predominates in the variant weighs against prior proposals that stereoelectronic features of the 6-OH-Hyo epoxidation substrate itself – rather than how it interacts with the enzyme – are crucial for the high epoxidation chemoselectivity.⁴⁹ It appears, rather, that the enzyme actively impedes reaction flux away from the inherently feasible second hydroxylation pathway to direct the strain-incorporating epoxidation.

Figure 9.

LCMS analysis of products in reactions of wt and L289F AbH6H with either Hyo or 6-OH-Hyo. The red trace shows m/z = 306, corresponding to two products, 7-OH-Hyo, with retention time of 3.1 mins, and 6-OH-Hyo, with retention time of 3.8 mins. The orange trace shows m/z = 288, corresponding to the 6,7-dehydro-Hyo minor product. The green and purple traces depict m/z = 304 and m/z = 322, respectively, corresponding to the Sco and (OH)₂-Hyo products. The final concentrations were 0.12 mM enzyme, 0.1 mM Fe, 0.1 mM prime substrate, and 0.03 mM 2OG. Assays were allowed to proceed for 1 h at 4° C open to the air.

Previous work on HnH6H showed that out-of-order C7 hydroxylation of Hyo incorporates ¹⁸O from ¹⁸O₂ to a lesser extent than does the proper C6 hydroxylation, despite the fact that the two are initiated by a common ferryl intermediate.⁵⁵ We posited that this difference reflects greater solvent exchange of the oxo-derived hydroxo ligand in the Fe^III–OH/C7• intermediate state than in the Fe^III–OH/C6• intermediate state, which we attributed to their distinct SCDs.⁵⁵ We used this effect to assess whether the ~ 70-fold enhanced hydroxylation of C7 of 6-OH-Hyo by the L289F variant of AbH6H might result from accelerated oxygen rebound caused by a change in SCD. In reactions under ¹⁸O₂, we found only 13% ¹⁸O in the 6,7-(OH)₂-Hyo product formed from 6-OH-Hyo by the wt enzyme (Figure 10). By contrast, we measured 42% ¹⁸O incorporation from ¹⁸O₂ in the predominant 6,7-(OH)₂-Hyo product produced by the L289F variant in its second reaction, consistent with faster oxygen rebound to C7 in the variant.

Figure 10.

LCMS chromatograms showing the composition of the dihydroxylated product formed under an ¹⁸O₂ atmosphere from the hydroxylation substrate, hyoscyamine (showing that the dihydroxylated can incorporate up to two atoms from O₂), and from the 6-OH-Hyo substrate for both wt AbH6H and the L289F variant. For the lowest set of traces, the final concentrations were 0.12 mM AbH6H, 0.1 mM Fe, 0.1 mM Hyo, 0.3 mM 2OG, and 0.9 mM ¹⁸O₂. For the middle set of traces, the final concentrations were 0.12 mM AbH6H, 0.1 mM Fe, 0.1 mM 6-OH-Hyo, 0.3 mM 2OG, and 0.9 mM ¹⁸O₂. For the upper set of traces, the final concentrations were 0.12 mM L289F AbH6H, 0.1 mM Fe, 0.1 mM 6-OH-Hyo, 0.3 mM 2OG, and 0.9 mM ¹⁸O₂. The reactions were carried out in sealed vials within an anoxic chamber. After initiation by simultaneous addition of 2OG and ¹⁸O₂, reactions were allowed to proceed for 15 min at 25 °C.

To understand the structural basis for the unleashing of rebound by the L289F substitution, we solved four structures of the variant from crystals of its Fe^II•2OG•(6-OH-)Hyo and V^IVO•succ•(6-OH-)Hyo complexes. They range in resolution from 1.53–2.03 Å (Figure S30; statistics shown in Table S9). Even though L289 lines the prime substrate binding pocket (Figure S28), its substitution by the bulkier F barely perturbs the SCD in the hydroxylation reactant complex (Figure 11, compare A to E). This minimal impact comports with the minor effect of the substitution on the distribution of products generated from Hyo. Two of the three remaining pairwise comparisons – between the structures derived from the V^IVO•succ•Hyo and Fe^II•2OG•6-OH-Hyo complexes of the wt and L289F proteins – reveal similarly modest changes (Figure 11, compare B to F and C to G). Among the four pairwise comparisons, only the structures derived from the V^IVO•succ•6-OH-Hyo complexes reveal a significant change in SCD associated with the L289F substitution (Figure 11, compare D to H). The additional bulk of F289 causes the oxo-derived in-line ligand to shift by ~ 1.3 Å, so that it forms a 106° O–V–O angle with the solvent-derived off-line oxygen ligand. This angle is ~ 20° greater than that in the structure of wt AbH6H. The 6-OH of the substrate also shifts by ~ 0.8 Å along a vector roughly parallel to that of the in-line oxygen ligand shift, preserving the H-bonding interaction between the substrate and ligand oxygens. The resultant ~ 25° global tilt of the tropane core can be seen by superposition of the modeled atoms and associated electron density maps for the two proteins (Figure 11I, far right). This modest repositioning does not change the C7–V distance, but it increases the V–O–C7 angle by ~ 20° (from 108° to 127°) (Figure 11). This change makes the angle more similar to those seen in vanadyl-derived structures of the Fe/2OG hydroxylases TauD, VioC, and CAS (132–136°), in which both the initiating HAT and terminating oxygen-rebound steps are efficient (Table 2, Figure 8 and S26).^33,35,46 As such, the perturbed SCD of the L289F variant may explain both its diminished capacity for C7–O(C6) coupling (epoxidation) and the diminished fraction of solvent exchange of the initially O₂-derived oxo group during its C7 hydroxylation.

Figure 11.

Comparisons of metal-substrate dispositions in each of the four cognate pairs of wt and L289F AbH6H crystal structures. Three of the four pairs overlay within error (compare A-C to E-G, respectively); only the structures derived from the V^IVO•succ•6-OH-Hyo complexes reveal a shift resulting from the increased steric bulk in the active site (compare D to H). (I) These structures are overlaid with gray mesh that illustrates 2F_o-F_c electron density (contoured at 1σ) for the 6-OH-Hyo substrate, the metal and oxygen ligands, and the L289 and F289 sidechains. The lighter mesh corresponds to the wt protein and the darker mesh to the L289F protein. Selected side chains, water molecules, cofactors, and substrates are shown in ball-and-stick or sphere representation and colored according to atom type.

²H₂O favors C7 hydroxylation over epoxidation and slows ferryl decay only in H6H.

In the absence of a Lewis-acid role of the cofactor toward the C6 hydroxyl group, epoxidation requires cleavage of this O–H bond along with the C7–H bond, whereas hydroxylation lacks an analogous requirement. Therefore, to test the conclusion that the C6 oxygen of 6-OH-Hyo does not coordinate, we asked whether D₂O solvent would impact the (OH)₂-Hyo:Sco ratio. The partition ratio increases by a factor of 7 to 24:76 in D₂O (Table 3 and Figure 12), consistent with a selective slowing of the step that commits to the epoxide product. D₂O also potentiates the effect of the L289F substitution: in the reaction of the variant protein with 6-OH-Hyo, the (OH)₂-Hyo:Sco partition ratio increases from 75:25 in H₂O to 94:6 in D₂O (Table 3 and Figure 12).

Table 3.

Relative quantification of the products formed by wt and L289F AbH6H with 6-OH-Hyo and 7-d₁-6-OH-Hyo in H₂O and D₂O.

% (OH)2-Hyo Producta	WT H2O	WT D2O	L289F H2O	L289F D2O
6-OH	4.5 ± 0.3	23.7 ± 1.2	74.8 ± 1.4	93.8 ± 0.4
7-d1-6-OH	3.6 ± 0.3	18.3 ± 0.1	66.5 ± 2.2	91.7 ± 0.4

^aIn each case, the reported value is the percent of the total detected product [Sco + (OH)₂-Hyo] that is (OH)₂-Hyo. Values in the table were determined as the percentage of the area of each peak to the total area of the two peaks. Each value is an average of three trials.

Figure 12.

Chromatograms showing the product distributions in reactions of (A) wt AbH6H and (B) the L289F variant with 6-OH-Hyo and 7-d₁-6-OH-Hyo in H₂O and D₂O. The final concentrations were 0.25 mM AbH6H (L289F or wt), 0.2 mM Fe, 0.5 mM prime substrate, and 0.1 mM 2OG. Assays were allowed to proceed open to atmosphere at 4 °C for 1 h. Product percentages derived from peak integrations are given in Table 3.

A solvent D-KIE on (C6)O–H deprotonation occurring after a canonical HAT from C7 would not be expected to impact the rate of decay of the C7–H-cleaving ferryl complex (except under special circumstances; see Discussion). Surprisingly, we observed a significant normal solvent D-KIE on ferryl decay for both protium-bearing and deuterium-bearing substrates (Figure 1). Decay was slowed to 0.35 and 0.06 s⁻¹ for 6-OH-Hyo and 7-d₁-6-OH-Hyo, for observed solvent D-KIEs of 1.4 and 1.8, respectively. Because we had not, in our prior work, tested other Fe/2OG enzymes for such an effect, we did so here for the prototypical hydroxylase, TauD,⁵⁶ the non-native desaturation of deoxyproclavaminate by clavaminate synthase (which competes with C3 hydroxylation),⁵⁷ and the oxacyclization reaction of LolO. We observed no comparable normal solvent D-KIE on ferryl decay for either the hydroxylation or the desaturation reaction (Figure S31). More importantly, as reported in the separate study on LolO,⁴² its superficially analogous oxacyclization reaction exhibits (1) no discernible solvent D-KIE on the partition ratio [(OH)₂-AcAP:NANL] and (2) a negligible (perhaps slightly inverse) effect on ferryl decay. We conclude that, in the LolO reaction, the C2-bonded oxygen coordinates to the ferryl species, leading to its deprotonation in advance of HAT and obviating a PT step as part of C7–O(C2) coupling that forms the oxalane ring.⁴² The analogous Lewis-acid assistance by the cofactor is not operant in the H6H epoxidation.

Analysis of the kinetic traces in Figure 1 reveals that the substrate and solvent D-KIEs potentiate each other. In other words, the observed substrate D-KIE is larger in D₂O than in H₂O (5.8 versus 4.5), and the observed solvent D-KIE is larger with the deuterium-bearing substrate than with the protium-bearing substrate (1.8 versus 1.4). Such potentiation is generally not observed in simultaneous KIEs on distinct isotope-sensitive steps when those steps occur sequentially and has been used in extensive prior literature as an indication that the isotope-sensitive steps occur concertedly.^58–60 As discussed below, this last observation is challenging to rationalize by any variant of the canonical sequential HAT/radical-coupling mechanism generally considered for Fe/2OG enzymes.

Discussion

Optimized SCD obviates active-site reconfiguration between hydroxylation and epoxidation.

Results presented above from several different lines of investigation imply a surprisingly static SCD throughout the two chemically distinct, sequential reactions catalyzed by H6H. The position of the tropane core of (6-OH-)Hyo/Sco is virtually invariant in six crystal structures of reactant, product, and intermediate-mimicking (vanadyl) complexes for both reactions. This conserved SCD is also seen in HYSCORE data on the vanadyl complexes, from which calculated metrics agree well with those seen in the crystallographic models, thus ruling out confounding effects of X-ray induced photoreduction in the crystallography. The invariant SCD is inappropriate for coordination of the C6 oxygen of the second substrate to the cofactor in the epoxidation step, seemingly ruling out a previously proposed halogenase-like radical C–O coupling step in closure of the strained ring.^31,49 Both ferryl intermediates have Mössbauer parameters that are within the range set by analogous complexes in other Fe/2OG enzymes, and they cleave the exo C7–H bond of the (6-OH-)Hyo substrate with indistinguishable rate constants. It is apparent that, rather than remodeling the active site or ferryl complex (or both) between the first and second reactions to switch the C–H target and change the fate of the ensuing substrate radical, H6H largely pre-programs its two-reaction sequence by a very special – and possibly unique – SCD. This conclusion is starkly validated by the observation that a minor (10–20°) shift caused by the bulk-increasing L289F substitution near the substrate defaults the second reaction to primarily C7 hydroxylation, leading to enhanced incorporation of oxygen from O₂ into the (OH)₂-Hyo product as a result of more efficient rebound.

Implications for the chemical mechanism of epoxidation.

What does its extreme sensitivity to SCD say about the chemical mechanism of the epoxidation reaction? The possibility most in line with established precedent would begin with HAT from C7 to the second ferryl complex (Scheme 3). The observed substrate C7 D-KIE of 5 on ferryl decay, which is necessarily less than the intrinsic effect on C7–H cleavage by at least of factor of 2 (due to uncoupling), is consistent with this initiating step. In lieu of substrate-alkoxide coordination, a substrate-to-cofactor PCET step, in which the C6–OH is the proton donor, the hydroxo cofactor ligand is the proton acceptor, and the C7 radical is the electron donor, would generate a 1,3-zwitterion poised for polar C–O coupling (Scheme 3A, purple arrows). This mechanism would provide a possible functional rationale for the extraordinary proximity of the substrate oxygen to the in-line vanadium-coordinated oxygen in the structure of the AbH6H•V•succ•6-OH-Hyo complex: elongation of the Fe–O bond upon HAT from C7 to the ferryl complex could establish or strengthen a hydrogen bond between the substrate and ligand oxygens in preparation for the PCET step. Provided that both the initiating HAT and subsequent PCET steps are endergonic, making them reversible and disfavored, this mechanism can even account for the surprising solvent D-KIE on ferryl decay seen (uniquely among the enzymes that we tested) in the H6H epoxidation. In such a scenario (scheme at right in Figure S32), the HAT step would effectively consume the ferryl complex only after traversal of the entire three-step sequence. The large intrinsic D-KIEs associated with hydrogen/proton transfers (which have been explained by the different tunneling efficiencies of protium and deuterium) allow formulation of a set of four rate constants, of which three (k_{obs H-sub/H2O}, k_{obs D-sub/H2O,} k_{obs H-sub/D2O}) are consistent with values determined in this study (and their ratios, which are the observed substrate and solvent D-KIEs) as well as with measured values of analogous intrinsic D-KIEs. This kinetic scheme linking the C7 HAT and PCET steps would imply that the HAT step is endergonic by ~ 1.7 kcal/mol. The likelihood of such an uphill HAT step is difficult to assess. As noted, we are unaware of any experimental evidence for reversibility in HAT from an aliphatic carbon to a ferryl complex, but the absence of such evidence could imply either that the steps immediately following HAT (e.g., oxygen rebound for the hydroxylases) are invariably much faster than the reverse HAT steps (i.e., a high forward commitment) or that the possibility of reversibility has not been adequately explored. Moreover, DFT and QM-MM studies have not generally agreed as to whether the HAT steps are endergonic or exergonic.^21,61–63 With these caveats, the chemical and kinetic mechanism of Scheme 3 and Figure S32 can account for all but one of the experimental observations from this study (see below for the exception). According to this mechanism, the extreme sensitivity of the outcome to SCD demonstrated by the L289F variant would reflect the narrowness of the “SCD space” that can allow for a viably efficient HAT step while almost completely suppressing oxygen rebound, as originally laid out for the halogenases.

Scheme 3.

Possible mechanisms for the H6H-catalyzed epoxidation reaction.

The lone observation that appears to confound the HAT/PCET/polar-C–O-coupling mechanism in Scheme 3A is the combined effect of simultaneous substrate and solvent deuteration. In the case of a sequential mechanism, each substitution should diminish the observed effect of the other (Figure S32). As explained by O’Leary,⁵⁸ Cleland,⁵⁹ Cook,⁶⁰ and others in the context of steady-state-kinetic analysis using two isotopic substitutions, slowing one step by the heavy isotope diminishes the extent to which the other step is rate-limiting for the sequence, and so the intrinsic KIE on the other step becomes less “expressed” in the flux through the multi-step pathway. This predicted co-dependency contrasts starkly with the one shown in Figure 1, in which one effect is potentiated by the other. In prior studies, observations of this mutual potentiation have been taken as evidence that the two isotopic substitutions affect the same reaction step.^58–60,64 Applying this logic to our results on H6H would imply concerted cleavage of C7–H and (C6)O–H bonds in the epoxidation reaction (Scheme 3B, orange arrows), as previously speculated.⁴⁹ A concerted proton-coupled HAT step (H⁺ + H•) seems unlikely, as other well-studied cases of neutral hydrogen (H•) transfer are not known to require a PT step, and such a mechanism would result in charge separation. Charge-neutral proton-coupled hydride transfers (H⁺ + H^–, denoted PCHT) are known,^65–66 and recent studies have proposed hydride transfers to metal-dioxygen complexes of the sort commonly encountered in reactions involving initiating HAT steps.⁶⁷ However, it is difficult to conceive of such a mechanism in this reaction, which would involve transfer of a hydride from a carbon lacking an activating heteroatom to a high-spin (S = 2) ferryl complex. Computational analysis of the feasibility of a mechanism involving such a 2-electron/2-proton transfer (E²P²T) to a ferryl complex is perhaps warranted. If such a mechanism could occur, it might recast the demonstrated sensitivity to SCD in terms of the imperative to suppress the canonical HAT reactivity in favor of the unusual E²P²T leading directly to the postulated 1,3-zwitterion. Competition by HAT, leading to C7 hydroxylation instead of epoxidation (Scheme 3B, blue arrows), would be enhanced in D₂O, because only the unusual E²P²T step would involve transfer of a solvent exchangeable hydron. Disruption of the special SCD by the L289F substitution would allow HAT to outcompete the E²P²T, resulting in predominant hydroxylation, especially in D₂O. The likelihood of this alternative explanation is difficult to assess until (unless) a possible pathway has been identified and its activation barrier(s) evaluated.

Conclusion

The data presented demonstrate that neither coordination of the C6-OH to the iron center nor major reconfiguration of the active site is required for the switch in outcome from C6 hydroxylation in the first reaction of H6H to C7–O(C6) coupling in the second reaction. Instead, the enzyme achieves selective epoxidation by enforcing a possibly unique SCD that disfavors oxygen insertion (hydroxylation) and allows a chemically challenging, strain-incorporating, alternative C–O-coupling (epoxidation) step to prevail in the second reaction. Strategic introduction of additional steric bulk in a single residue near the active site tilts the substrate ~ 20° from this special binding mode and unleashes the default oxygen-insertion reactivity of the second ferryl complex, thus largely preempting epoxidation. The least radical interpretation of the data is in terms of a cyclization mechanism involving polar capture of a C7 carbocation – formed in a rapid and disfavored PT-coupled ET to the Fe(III) cofactor after HAT from C7 – by the C6 alkoxide, a mechanism that would align with proposals advanced in recent studies on aziridine formation by Fe/2OG enzymes.⁶⁸ The requirement for C6–OH deprotonation to allow for successful capture of the disfavored C7 carbocation would explain the large solvent D-KIE of ~ 8 on the epoxidation:hydroxylation partition ratio. The surprising and – by this mechanism – unexplained observation of a smaller solvent D-KIE also on ferryl decay leaves open the possibility of a more divergent mechanism involving highly coupled or concerted cleavage of the (C6)O–H and C7–H bonds by the second ferryl complex, which would recast the extraordinarily strict geometric requirements of the H6H epoxidation reaction in terms of the need to disfavor the conventional C7-HAT relative to the hypothetical E²P²T step. Irrespective of which scenario is operant, the dichotomy between the mechanisms deduced here for H6H and in a parallel study of LolO, in a pair of reactions that are superficially similar, reinforces the previously noted importance of substrate and product structures in constraining chemical pathways that may be available for enzymatic acceleration of such chemically challenging C–H-functionalization reactions.⁷