Determining the Inherent Selectivity for Carbon Radical Hydroxylation versus Halogenation with Fe^III(OH)(X) Complexes: Relevance to the Rebound Step in Non-heme Iron Halogenases

Abstract

The first structural models of the proposed cis-Fe^III(OH)(halide) intermediate in the non-heme iron halogenases were synthesized and examined for their inherent reactivity with tertiary carbon radicals. Selective hydroxylation occurs for these cis-Fe^III(OH)(X) (X = Cl, Br) complexes in a radical rebound-like process. In contrast, a cis-Fe^III(Cl)₂ complex reacts with carbon radicals to give halogenation. These results are discussed in terms of the inherent reactivity of the analogous rebound intermediate in both enzymes and related catalysts.

Non-heme iron enzymes utilize dioxygen to carry out C–H bond transformations, including hydroxylation and halogenation. These enzymes include the αKG-dependent hydroxylase TauD, which carries out the hydroxylation of taurine,¹ and αKG-dependent SyrB2,² CytC3,³ and WelO5,^4,5 which carry out the halogenation of natural products.⁶ The proposed mechanisms of these enzymes are similar, relying on an Fe^IV(O) oxidant to abstract a hydrogen atom from the C–H substrate. The nascent carbon radical then recombines, or “rebounds”, with either the hydroxyl (FeOH) group in the case of the hydroxylases or the halide (FeX) group in the case of the halogenases.⁷ A simplifed halogenase mechanism is shown in Scheme 1. Halogenase activity appears to derive from factors that control the selective transfer of Cl· (or X·) to the nearby carbon radical.

Scheme 1.

Proposed Rebound Mechanism

The key factors that control the radical rebound process remain under debate,^8–13 in part because of the lack of a direct methods to study the presumably short-lived, unobserved ferric rebound intermediate. A range of different proposals^5,14–23 attempting to describe these factors for the halogenases have been put forth. Synthetic iron model complexes also exhibit some halogenation activity^24–28 together with hydroxylation activity (Scheme 1). For both enzymes and models, computational studies have indicated that the thermodynamically preferred product should be the alcohol, not the halogenated compound.^14,19,22 Thus, halogenation is predicted to arise from a kinetic pathway, in which the reaction barrier for rebound of the halide is significantly lower than that for the hydroxide.^14,22 However, addressing the inherent reactivity of Fe^III(OH)(X) species experimentally has been out of reach until now.

Herein we report the synthesis of two mononuclear Fe^III(OH)(X) (X = Cl, Br) complexes that to our knowledge are the first structural models of the key ferric intermediate proposed for the non-heme iron halogenases. The isolation of these complexes allowed us to examine their inherent reactivity with tertiary carbon radicals. In earlier work, we examined the rebound reactivity of non-heme Fe^III(OMe)²⁹ and Fe^III(OH)³⁰ complexes, and the latter complex contained pendent H-bond donors to stabilize a terminal OH ligand as well as a labile cis coordination site that has now been exploited to prepare the elusive cis-Fe^III(OH)(X) complexes. Reactions with tertiary carbon radicals give exclusively the hydroxylated products, showing a preference for the thermodynamic pathway, while an analogous Fe^III(Cl)₂ complex is competent to halogenate the radicals.

The tetradentate ligand BNPA^Ph2O⁻ was used previously to prepare Fe^III(BNPA^Ph2O)(OH)(OTf), a terminal Fe^III(OH) complex stabilized by the steric protection and H-bonding of the pendent neopentylamine groups.³⁰ The triflate ligand is cis to the OH⁻, and we hypothesized that it could be replaced by halide (X⁻) under suitable conditions. Reaction of BNPA^Ph2O⁻ with ferrous halide salts gives Fe^II(BNPA^Ph2O)(X) (1, X = Cl; 2, X = Br). X-ray diffraction (XRD) revealed the five-coordinate Fe^II complexes shown in Figure 1. The halide ligands in these approximately trigonal-bipyramidal complexes (τ = 0.626 for 1, τ = 0.635 for 2)³¹ occupy the axial positions and form H-bonds with the pendent neopentylamine groups: N(H)⋯X(ave) = 3.222 Å in 1 and 3.363 Å in 2 and N–H–X(ave) = 171° in 1 and 166° in 2 (Figure 1). The remaining Fe–N and Fe–O distances are in the typical ranges for high-spin (hs) (S = 2) Fe^II complexes^32–34(Table S1).

Figure 1.

Displacement ellipsoid plots (50% probability level) for 1 and 2 at 110(2) K. Hydrogen atoms (except for N–H) have been omitted for clarity.

The reactions of 1 and 2 with excess O₂ in THF at 23 °C lead to a rapid color change from yellow to bright orange. Removal of the volatiles gives dark-orange powders, and recrystallization from THF/n-pentane yields crystals of 3 and 4, respectively, as dark-orange blocks suitable for XRD. The crystal structures of these new species (Figure 2) reveal the six-coordinate complexes Fe^III(BNPA^Ph2O)(OH)(X) (3: X = Cl, 4: X = Br), whose metrical parameters are consistent with hs ferric (S = ⁵/₂) ions.^33–37 The OH ligands occupy the sites involved in H-bonding interactions with the pendent amines (3: N(H)⋯O = 2.78 Å, N–H–O = 162°; 4: N(H)···O = 2.77 Å, N–H–O = 159°). These H-bonds are relatively tight, as seen in previously reported Fe^III(OH) complexes.^{32–36,38,39} The halide ligand is bound cis to the OH ligand and exhibits an Fe–X distance that is in the range of those for other hs Fe^III(X) compounds.^28,40,41 Density functional theory (DFT) calculations at the B3LYP/6–311G* level give optimized geometries for 1–4 that are in good agreement with the crystal structures (Figure S22 and Tables S1 and S2). The formation of 3 and 4 from O₂ (Scheme 2) likely follows a mechanism similar to that proposed for Fe^III(OH)(OTf)-(BNPA^Ph2O)³⁰ and related compounds.^42,43

Figure 2.

Displacement ellipsoid plots (50% probability level) for 3 and 4 at 110(2) K. Hydrogen atoms (except for N–H and O–H) have been omitted for clarity.

Scheme 2.

Synthesis of Fe^III(OH)(X) Complexes

The UV–vis spectra of 1–4 are relatively featureless in the visible region (Figures S34–S37). The ¹H NMR and Mössbauer spectra are more informative. The ¹H NMR spectra for complexes 1 and 2 in CD₃CN exhibit similar paramagnetically shifted peaks between −12 and +85 ppm (Figure 3). In contrast, the hs Fe^III complexes 3 and 4 exhibit fewer and much more broadened peaks (Figures S3 and S4). Enrichment of 1–4 with ⁵⁷Fe was accomplished by developing an alternative synthetic route involving ligand metathesis of ⁵⁷Fe^II(BNPA^Ph2O)(OTf) or ⁵⁷Fe^III(BNPA^Ph2O)(OH)(OTf)³⁰ with Bu₄NX (X = Cl⁻, Br⁻) in 2-MeTHF to give ⁵⁷Fe-labeled 1–4 in situ. Substitution of X⁻ for OTf⁻ was confirmed by ¹H NMR spectroscopy, which gave spectra identical to those seen for independently prepared 3 and 4 (Figures S28 and S29). Mössbauer spectroscopy of ⁵⁷Fe-labeled 1 and 2 in 2-MeTHF shows a sharp quadrupole doublet for both complexes, with δ = 1.02 mm s⁻¹ and |ΔE_Q| = 2.65 mm s⁻¹ for 1 and δ = 1.03 mm s⁻¹ and |ΔE_Q| = 2.55 mm s⁻¹ for 2 (Figure 4), indicative of hs Fe^II (S = 2). As anticipated, the Mössbauer spectra of 3 and 4 show broadened quadrupole doublets with parameters typical of hs Fe^III: δ = 0.42 mm s⁻¹ and |ΔE_Q| = 1.01 mm s⁻¹ for 3 and δ = 0.41 mm s⁻¹ and |ΔE_Q| = 1.10 mm s⁻¹ for 4 (Figure 4). The significant line broadening seen for 3 and 4 can be attributed to population of an intermediate relaxation regime, as reported previously^44,45,24 (see the Supporting Information for fitting details).

Figure 3.

¹H NMR spectra in CD₃CN of (a) 1, (b) 3 + (p-MeOC₆H₄)₃C·, (c) 2, and (d) 4 + (p-MeOC₆H₄)₃C·.

Figure 4.

Zero-field ⁵⁷Fe Mössbauer spectra (80 K) of (a, a′) 1 and 2, (b, b′) 3 and 4, and (c, c′) 3 and 4 + (p-MeOC₆H₄)₃C·.

To our knowledge, complexes 3 and 4 are the first structural models of the Fe^III(OH)(X) intermediates proposed for the non-heme iron halogenases and related catalysts, as shown in Figure 1. They also nicely reproduce the H-bonding proposed for the active site in SyrB2.^14,19 These complexes are excellent starting points for the study of the inherent rebound reactivity of Fe^III(OH)(X) species.

Reactivity studies for 3 and 4 were initiated with the triphenylmethyl radical derivatives (p-YC₆H₄)₃C· (Y = OMe, H, Cl). These tertiary carbon radicals are relatively stable and have been used to show “rebound”-like reactions with Fe and Cu complexes.^{29,30,39,46–50} Addition of (p-YC₆H₄)₃C· to 3 or 4 was followed by stirring for 1–4 h (depending on the identity of Y) at 23 °C in THF/toluene (5/2 v/v). The volatiles were removed under vacuum, and the reaction mixture was dissolved in CD₃CN for analysis by ¹H NMR spectroscopy. The disappearance of the peaks for 3 and 4 occurred with the appearance of the spectra for the one-electron-reduced complexes 1 and 2 (Figure 3). These data suggested that the OH group was selectively transferred to the carbon radical.

The reactions with the triphenylmethyl radical derivatives were also monitored by Mössbauer spectroscopy. The final reaction mixture of ⁵⁷Fe-enriched 3 and excess (p-MeOC₆H₄)₃C·, after removal of the volatiles and dissolution of the resultant solid in 2-MeTHF, showed a sharp quadrupole doublet with δ = 1.03 mm s⁻¹ and |ΔE_Q|= 2.69 mm s⁻¹ (94% of the total fit), a close match to that seen for 1 (Figure 4, left panel). Similarly, the ferric complex 4 shows complete conversion to the Fe^II(Br) complex 2 upon reaction with (p-MeOC₆H₄)₃C· (Figure 4, right panel).

The tertiary alcohols (p-YC₆H₄)₃COH were identified by ¹H NMR spectroscopy of the crude reaction mixtures. Peaks arising from the newly formed hydroxyl groups appeared as well-separated singlets, and integration gave a high (>90%) yield of the alcohol in all cases (Figures S13–S18). The formation of Ph₃COH (m/z 260.1) and (p-ClC₆H₄)₃COH (m/z 362.0) was confirmed by mass spectrometry. Isotopic substitution of the OH group in 3 via synthesis with ¹⁸O₂ led to Ph₃C¹⁸OH (m/z 262.1) with ≥99% isotopic purity. The same reaction with the (p-ClC₆H₄)₃C· and isotopically labeled 4 led to (p-ClC₆H₄)₃C¹⁸OH (m/z 364.0) (Figures S30–S33). The data show that selective hydroxylation occurs in a rebound-type process with 3 and 4 (Scheme 3), and the ¹⁸O labeling supports a mechanism involving direct transfer of OH·from Fe^III to the carbon radical.

Scheme 3.

Reaction of 3 and 4 with Triphenylmethyl Radical Derivatives

Metalation of BNPA^Ph2O⁻ with FeCl₃ led to the formation of the dichloro analogue Fe^III(BNPA^Ph2O)(Cl)₂ (Figure 5). The structure of 5 shows that the alkoxide ligand is bound trans to the equatorial halide, whereas in complexes 3 and 4 a pyridine arm is bound trans to the equatorial halide. It is not yet known whether the nature of the trans ligand has a significant impact on the reactivity of these complexes. Reaction of 5 with the p-OMe and p-Cl trityl radical derivatives showed complete conversion to 1 by paramagnetic NMR spectroscopy. The formation of the halogenated (p-YC₆H₄)₃CCl was initially monitored by thin-layer chromatography (TLC), and then the products were isolated by silica gel chromatography in good yields (75% for Y = Cl and 84% for Y = OMe). Although it is not possible to determine which of the Cl ligands is transferred, it is clear from these results that halogen transfer occurs easily in the absence of a terminal hydroxide (Scheme 4). This result provides the first example, to our knowledge, of the direct reaction of an isolated Fe^III(Cl) complex with a carbon radical to give Fe^IIand a new C–Cl bond, a process that mimics halogen rebound.

Figure 5.

Displacement ellipsoid plot (50% probability level) for 5 at 110(2) K. Hydrogen atoms (except for N–H) have been omitted for clarity.

Scheme 4.

Reaction of 5 with Triphenylmethyl Radical Derivatives

Our findings indicate that the reactions of tertiary carbon radicals with 3 and 4 give hydroxylated products with no evidence of halogenation, although the analogous dichloro complex 5 is competent to transfer the halogen. The lack of halogen transfer for 3 and 4 contrasts with what is seen for both enzymes and halogenase models and thus deserves further comment. The selectivity seen for 3 and 4 is in line with thermodynamic expectations, which are supported by calculations on both enzymes and models that uniformly predict hydroxylation to be thermodynamically favored over halogenation by at least 15 kcal/mol.^14,21,22,26 However, the same computational studies also predict that the reaction barriers to halogenation are lower, leading to kinetically favored halogenated products. A range of factors have been offered as contributing to the lower kinetic barriers for halogenation, including substrate orientation,^16,23 relocation of the positioning of the OH group on the metal,^5,19,20 relative ordering of frontier molecular orbitals,¹⁴ relative spin densities on OH versus Cl,¹⁵ and steric effects.²⁸ It has also been suggested that H-bonds between the OH group of the putative Fe^III(OH)(Cl) intermediate and Arg₂₅₄/Glu₁₀₂/water molecules in SyrB2 raise the hydroxylation barrier.^14,19

The lack of halogenation seen for 3 and 4 occurs even though OH ligand is held in place and deactivated toward the incoming carbon radical by two tight H-bonds. This inherent reactivity may be a result of the nature of complexes 3 and 4 or may be related to the nature of the radical. It was shown that [Fe^IV(O)(Cl)(PyTACN)]⁺ reacts with triphenylmethane to give Ph₃ COH,²⁵ suggesting that the tertiary carbon radical may favor hydroxylation. However, [Fe^IV(O)(Cl)(PyTACN)]⁺ reacts with other C–H bonds (e.g., cyclohexane), but still no halogenated products were observed. In contrast, some halogenation has been seen for cyclohexane and toluene with a non-heme Fe^IV(O)(X) (X = Cl, Br) oxidant.²⁴ Reaction of Ph₃C· with a low-coordinate Fe^III(NHAr)(Cl) complex disfavored halogenation, leading to amination instead.⁴⁷ The importance of the nature of the carbon radical remains to be determined.

Complexes 3 and 4 are analogous to the key ferric intermediates in non-heme iron halogenases and related catalysts. Future plans include computational studies and other mechanistic investigations of these rebound reactions to understand the key factors that control the rebound selectivity in both enzymes and models.