Biocatalytic Carbon-Hydrogen and Carbon-Fluorine Bond Cleavage through Hydroxylation Promoted by a Histidyl-Ligated Heme Enzyme

Abstract

LmbB2 is a peroxygenase-like enzyme that hydroxylates L-tyrosine to L-3,4-dihydroxyphenylalanine (DOPA) in the presence of hydrogen peroxide. However, its heme cofactor is ligated by a proximal histidine, not cysteine. We show that LmbB2 can oxidize L-tyrosine analogs with ring-deactivated substituents such as 3-nitro-, fluoro-, chloro-, iodo-L-tyrosine. We also found that the 4-hydroxyl group of the substrate is essential for reacting with the heme-based oxidant and activating the aromatic C-H bond. The most interesting observation of this study was obtained with 3-fluoro-L-tyrosine as a substrate and mechanistic probe. The LmbB2-mediated catalytic reaction yielded two hydroxylated products with comparable populations, i.e., oxidative C-H bond cleavage at C5 to generate 3-fluoro-5-hydroxyl-L-tyrosine and oxygenation at C3 concomitant with a carbon-fluorine bond cleavage to yield DOPA and fluoride. An iron protein-mediated hydroxylation on both C-H and C-F bonds with multiple turnovers is unprecedented. Thus, this finding reveals a significant potential of biocatalysis in C-H/C-X bond (X = halogen) cleavage. Further ¹⁸O-labeling results suggest that the source of oxygen for hydroxylation is a peroxide, and that a commonly expected oxidation by a high-valent iron intermediate followed by hydrolysis is not supported for the C-F bond cleavage. Instead, the C-F bond cleavage is proposed to be initiated by a nucleophilic aromatic substitution mediated by the iron-hydroperoxo species. Based on the experimental results, two mechanisms are proposed to explain how LmbB2 hydroxylates the substrate and cleaves C-H/C-F bond. This study broadens the understanding of heme enzyme catalysis and sheds light on enzymatic applications in medicinal and environmental fields.

Graphical Abstract

Introduction

The enzymatic hydroxylation of an aromatic ring is a chemical reaction with vast history and utility. Enzymatic aromatic ring hydroxylation is an essential step in the activation and degradation of molecules by the human liver, and of natural and unnatural aromatic molecules by microorganisms in groundwater and soil.^{1, 2} It is known that aromatic rings are hydroxylated by NADH-dependent P450s,³ pterin- or α-ketoglutarate-dependent non-heme iron monooxygenases,⁴ di-iron hydroxylases,⁵ type III copper-mediated monooxygenases,⁶ and flavin-dependent monooxygenases⁷ with either molecular oxygen or hydrogen peroxide as their oxidants. Recently, reports of a new type of enzymatic hydroxylase activity on L-tyrosine have surfaced from the gene clusters of natural product biosynthesis.⁸ The first of these L-tyrosine hydroxylases was identified as LmbB2 from the biosynthetic gene cluster of the antibiotic, lincomycin;⁹ additional homologs were later identified in hormaomycin¹⁰ and four pyrrolo[1,4]benzodiazepines gene clusters of anthramycin,¹¹ sibiromycin,¹² tomaymycin,¹³ and poranthramycin.¹⁴ These clinically important natural products contain the structurally similar pyrroline moiety (Scheme 1A). The biosynthesis of such pyrroline moiety starts with L-tyrosine hydroxylation to L-3,4-dihydroxyphenylalanine (DOPA) by these LmbB2-like L-tyrosine hydroxylases (Scheme 1B).^{10, 15, 16} Preliminary biochemical studies on representative examples of these hydroxylases indicate that the hydroxylation was accomplished with a histidyl-ligated heme cofactor.^{11, 17} A heme ligated by histidine in the axial position distinguishes the LmbB2-like enzymes from other hemoproteins with established biological roles as hydroxylases. The cytochrome P450 enzymes and peroxygenases bind heme with axial cysteine-based thiol ligands, and it has been persuasively argued from a variety of experimental data that the identity and H-bonding environment of the axial ligand directly impact the redox potential of the heme and its ability to perform C-H bond activation and oxygen atom transfer.^18–21 Therefore, understanding the mechanism of L-tyrosine hydroxylases is not a simple extrapolation from previous work on aromatic hydroxylation systems, and careful study of reaction chemistry and mechanism is warranted.

Scheme 1.

(A) Natural antibiotics, including lincomycin, share the similar pyrroline moiety (highlighted in purple); (B) A new class of L-tyrosine hydroxylases (highlighted in navy) is involved in the first step of the pyrroline moiety biosynthesis, which converts L-tyrosine to DOPA in the presence of hydrogen peroxide.

In this study, our exploration of the enzymatic mechanism for tyrosine hydroxylase with unnatural substrate analogs and biophysical methods has revealed an unprecedented enzyme-catalyzed carbon-halogen (C-X) bond cleavage mechanism from halogenated L-tyrosine derivatives. EPR and HPLC results suggested that the histidyl-ligated enzyme can oxidize a range of tyrosine analogs even with strong ring-deactivating substituents. The product analyses demonstrated that LmbB2 can activate either C-H or C-X bond resulting in the departure of a proton or halide anion. Quantitative analyses indicated the C-H and C-X bond cleavage are from two independent reactions with multiple turnovers, and the highest yield of C-X cleavage was observed for fluorinated tyrosine. Analyzing isotope distribution on products with ¹⁸O-enriched peroxide or water suggests that the aromatic hydroxylation is driven by peroxide derivatives instead of water. Together, mechanistic pathways for C-H and C-X bond cleavage are proposed and discussed, and a biocatalytic C-H and C-F bond hydroxylation promoted by a histidyl-ligated heme enzyme is reported for the first time. Our characterization of this powerful chemistry from a naturally occurring heme-oxidant not only broadens our overall understanding of heme enzyme catalysis but also raises important implications for the pharmaceutical design of aryl-halogen containing drugs and bioremediation of environmental contaminants.

Results

LmbB2 requires 4-hydroxyl of phenolic amino acids for action

The tyrosine hydroxylase from Streptomyces lincolnensis (SlTyrH), LmbB2, was purified as described in the materials and methods and assayed for hydroxylation activity. The isolated enzyme was able to hydroxylate L-tyrosine (1) using H₂O₂ as the oxidant, producing DOPA (1a) (Scheme 2A) as shown by HPLC at room temperature (Figure 1A), under conditions similar to those previously reported for the homolog, Orf13.¹¹ In order to investigate the potential range of substrates, several substrate analogs were assayed. We first tested the requirement of a hydroxyl group at the 4 position during LmbB2-mediated hydroxylation by assaying L-phenylalanine (2) and O-methyl-L-tyrosine (3) as possible substrates. The former (2) does not contain a hydroxyl group, and the hydroxyl group of the latter (3) is protected in the form of a methoxy moiety (Scheme 2B). Initial observations from the assays indicated 2 and 3 were not substrates of the enzyme, as we did not detect any measurable reaction product from either analog regardless of concentration (up to 20 mM), pH (7 – 10) or oxygen source, i.e., H₂O₂ and peracetic acid (PAA) (Figure S1A and S1B). These results suggest an essential role for the 4-hydroxyl group during catalysis.

Scheme 2.

Summary of the LmbB2 reaction for (A) L-tyrosine (1, and product 1a), (C) 3-nitro-L-tyrosine (4, and product 4a), (D) 3-fluoro-, chloro-, iodo-L-tyrosine (5, 6, 7, and product 5a, 6a, 7a, respectively); (B) L-phenylalanine (2) and O-methyl-L-tyrosine (3) did not yield detectable reaction product.

Figure 1.

HPLC profile of the LmbB2 reaction shows L-tyrosine (1) hydroxylation to DOPA (1a) and the UV-Vis spectra (A). (i) 100 μM enzyme, 3 mM L-tyrosine and 3 mM H₂O₂ and three controls with (ii) 100 μM enzyme and 3 mM L-tyrosine, (iii) 3 mM L-tyrosine and 3 mM H₂O₂ and (iv) 100 μM enzyme and 3 mM H₂O₂. ITC binding assays showed LmbB2 binds 1 (B), 2 (C) and 3 (D) with K_D values of 1.35 ± 0.03, 318 ± 8, and 1294 ± 141 μM, respectively.

However, before drawing a firm conclusion regarding the catalytic role of the hydroxyl group, it is necessary to exclude the possibility that the non-reactive nature of these analogs is simply due to a binding problem. Therefore, binding of 1, 2 and 3 to LmbB2 was assessed by isothermal titration calorimetry (ITC), and the results revealed K_D values of 1.35 ± 0.03, 318 ± 8 and 1290 ± 140 μM, respectively (Figure 1B). Binding constants for analogs 2 and 3 imply substantially lower affinity compared to the native substrate; however, these K_D values are well below the actual substrate concentration used in the activity assay experiments, which guarantees sufficient binding of 2 and 3 under the conditions of the reaction.

To further probe whether 2 and 3 bind to the enzyme active site in a manner analogous to 1, we performed an electron paramagnetic resonance (EPR) analysis of the electronic structure of the ES complex. The EPR data collected on the enzyme bound to substrate 1 and analogs 2, 3 further supports the above interpretation. The LmbB2 protein was anaerobically reduced to the ferrous state and then exposed to an excess of nitric oxide (·NO) to form reduced enzyme-nitrosyl (E-NO·) complex. The E-NO· complex was then mixed with 1, 2 or 3 anaerobically in parallel experiments and frozen in liquid nitrogen before measurement by EPR spectroscopy. The EPR spectrum for an E-NO· complex in the absence of any other exogenous ligand exhibited a rhombic signal with an average g value, g_ave = 2.030 and some hyperfine splitting from ·NO (Figure 2A). This spectrum represents a typical low-spin six-coordinated S = 1/2 histidine-ligated Fe(II) heme-NO· complex, which has been observed in other biological systems.^{22, 23} In the presence of 20 mM of 1, the EPR spectrum shifted upfield. Such a small change indicates that 1 binds to a site adjacent to the heme iron. A direct ligation of 1 to the Fe would have a much large effect to the spectrum. The E-NO· EPR signal is very sensitive to a change at the active site, and therefore, even very small changes in the signal reflect alterations in electronic properties of the iron center. The anaerobic addition of 20 mM 2 or 3 to the E-NO· complex produced EPR spectra similar to that of 1 with the upfield shift (Figure 2B). The g_ave values after addition of 1, 2, or 3 were 2.027, 2.027 and 2.026, respectively. The nearly identical g_ave values and spectral changes suggest a similar microenvironment of ES-NO· after binding of those ligands, which all differ from the E-NO· alone signal. N-boc-L-tyrosine and p-cresol were selected as negative controls of the EPR experiments. The former of which has a bulky modification on its amino group and the later which lacks an amino acid moiety, and hence they should not show strong binding to the active site of LmbB2. The LmbB2 E-NO· complex in the presence of either compound had no significant EPR spectroscopic change compared with E-NO· alone, and the same g_ave values (Figure S2). These EPR data support the assertion that 2 and 3 bind to the active site in a manner consistent with natural substrate, 1.

Figure 2.

(A) EPR spectra of ferrous enzyme-NO· (E-NO·) complex alone and mixed with 1. In the absence of 1, E-NO· had a rhombic signal with hyperfine splitting from ·NO, g_ave = 2.030; after binding with 1, g_ave slightly shifted. (B) Spectra of E-NO· binding with 1, 2 and 3, the g_ave values were 2.027, 2.027 and 2.026, respectively. All spectra were obtained at 50 K, 9.6 GHz microwave frequency and 0.05 mW power.

Reaction of LmbB2 and L-tyrosine analogs with electron-deficient substitutions

We selected analogs with strong electron-withdrawing groups of nitro and three halide substitutions with increasing atom size and decreasing electronegativity. When 3-nitro-L-tyrosine (4) was used as a substrate of LmbB2, a small product peak was observed in the HPLC elution profile (Figure 3A) with absorption λ_max at 298 and a broad absorbance band around 356 nm (Figure 3B). High-resolution mass spectrometry (HRMS) analysis showed an m/z value of 243.0619 (Figure 3C), corresponding to 3-nitro-5-hydroxyl-L-tyrosine (4a, theoretical m/z = 243.0612 Da, mass accuracy = 2.88 ppm). ¹H NMR analysis further identified the product as 4a (Figure 3D). 4 had resonances at 7.97, 7.50, and 7.11 ppm arising from the aromatic hydrogens on C2, C5, and C6, respectively. The aromatic hydrogens of the eluted product were on C2 and C6 at 7.75 and 7.13 ppm, which confirmed that C5 was hydroxylated, generating 4a. Hydrogens on C_α and C_β had chemical shift around 3.9 and 3.1 ppm. These results established a reaction scheme shown in Scheme 2C. These data are consistent with previous reports and our data which assert that LmbB2 hydroxylation is specific to the C3, or C5 position of L-tyrosine yielding DOPA as the only product.^{11, 17} The inefficient conversion of 4 to 4a is likely due to the ineffective binding caused by the bulky nitro group, and the deactivation of the phenyl ring by the nitro substituent. Nevertheless, any production of 4a from such a heavily deactivated tyrosine ring implies the LmbB2 heme-based oxidant is highly potent.

Figure 3.

The LmbB2 reaction with 3-nitro-L-tyrosine (4) generated 3-nitro-5-hydroxyl-L-tyrosine (4a). (A) A small product peak was detected by HPLC. The setup for the catalytic reaction was the enzyme with 4 and H₂O₂ (i). Three controls were done with (ii) the enzyme and 4, (iii) 4 and H₂O₂ and (iv) the enzyme and H₂O₂. (B) Absorbance spectra of 4 (blue) and 4a (black). (C) High-resolution mass spectra of 4 and 4a. (D) ¹H NMR analysis 4 and 4a, hydrogen atoms are labeled according to the numbering of carbon.

Next, we tested LmbB2 reactivity towards three halogenated tyrosine analogs, 3-fluoro-L-tyrosine (5), 3-chloro-L-tyrosine (6), 3-iodo-L-tyrosine (7). Surprisingly, each of the compounds yielded two products in HPLC assays. The binding affinities for abovementioned analogs were assessed by ITC (Figure S3). The results showed that 5 has the highest binding affinity among four reactive analogs with a K_D of 22.2 ± 1.1 μM, around 16-fold weaker than the native substrate. The K_D values of 6 and 7 exponentially increases with the size of the halogen atom, at 346 ± 2 and 7820 ± 180 μM, respectively. The bulky nitro substitution of 4 resulted in no detectable thermal change by ITC experiment, even though some activity was observed. All the thermodynamic data from ITC are summarized in Table S1.

In the case of LmbB2 reaction with 5 (Figure 4A), the first eluted product had a retention time and absorption (2.2 min, 280 nm) consistent with 1a, while the second eluted product had an absorption maximum at 272 nm (Figure 4B). Based on the extinction coefficient and HPLC peak integration,^24–26 comparable amounts of products were produced with the estimated ratio of 2 : 1. Subsequent HRMS suggested that the first product lost a fluorine atom but gained a hydroxyl group as 1a (measured m/z = 198.0767 Da, theoretical m/z = 198.0761 Da, mass accuracy = 3.03 ppm), while the second product had only one additional oxygen atom (measured m/z = 216.0667 Da, theoretical m/z = 216.0667 Da, mass accuracy < 0.5 ppm) corresponding to 3-fluoro-5-hydroxyl-L-tyrosine (5a) (Figure 4C). Our ¹H NMR analysis confirmed the structure assignments (Figure 4D); the substrate 5 had a doublet at 7.00 ppm from the hydrogen at C2, a multiplet around 6.96 – 6.86 ppm from the two hydrogens on C5 and C6; the first eluted product corresponds to 1a, which had resonances at 6.87, 6.80 and 6.71 ppm arising from hydrogens on C5, C2, and C6, respectively; the later eluted product corresponds to 5a, which had resonances at 6.59 and 6.56 ppm from hydrogens on C2 and C6, respectively. The resonances around 3.9 ppm and 3.0 ppm corresponding to hydrogens on C_α and C_β had no significant chemical shift changes.

Figure 4.

Two products, DOPA (1a) and 3-fluoro-5-hydroxyl-L-tyrosine (5a), were detected by HPLC when LmbB2 reacted with 3-fluoro-L-tyrosine (5). (A) The HPLC profile. The setup for the enzymatic reaction was (i) LmbB2 with 5 and H₂O₂. Three controls were conducted with (ii) LmbB2 and 5, (iii) 5 and H₂O₂, and (iv) LmbB2 and H₂O₂. (B) Absorbance spectra of 5, 1a and 5a; (C) High-resolution mass spectra of 5, 1a and 5a; (D) ¹H NMR analysis for 5, 1a and 5a, hydrogen atoms are labeled according to the numbering of carbon.

To determine if LmbB2 would hydroxylate at both C3 and C5 for other C3-halogenated substrates, reactions using 6 and 7 as alternate substrates were conducted. The results obtained further elaborated the observation with 5. Indeed, LmbB2 catalyzes the expected hydroxylation at C5, and also the surprising hydroxylation at C3 requires apparent breakage of the carbon-halogen (C-X) bond to produce 1a. Examination of LmbB2 reaction with 6 revealed product 6a predominantly; only a small fraction was transformed into 1a (Figure S4). In the assay of 7, less substrate was hydroxylated. Most were converted to 7a, while 1a was present in trace amounts (Figure S5). Therefore, a reaction scheme for three halogen substituted analogs is summarized in Scheme 2D. The ratio of 1a formation among the reactions of substrates 5, 6, and 7 was 4.5 : 1.3 : 1 after normalizing for the percentage of substrate conversion (Table S2). Although 5 produced the most 1a, in both absolute and relative terms, 6 showed the most overall activity among the analogs tested, which may result from its relatively tight binding and weaker ring-deactivation. Overall, more 1a was formed in the reaction of 5 compared to 6 and 7 presumably because fluorine has the smallest steric effect among the halogens. Even though the C-F bond has the strongest strength and thermal stability among the three carbon-halogens, its cleavage is more effective than other carbon-halogen bonds.

Mechanistic investigation of C-H versus C-F bond cleavage by LmbB2

The formation of 5a is predictable as a result of an anticipated hydroxylation because the normal hydroxylation takes place on the only possible vacant site, i.e., C5 for 5. However, the formation of 1a is an entirely unexpected outcome, since it would require carbon-fluorine (C-F) bond cleavage. To verify the predicted loss of fluorine, ¹⁹F NMR analysis was conducted on the LmbB2 reaction of 5 with different equivalents of H₂O₂ (Figure 5A). The starting reaction mixture of LmbB2 and 5 without peroxide gave only a doublet of doublets at −137.68 ppm expected for 5. After addition of 0.5 equivalents of H₂O₂, the peak corresponding to 5 decreased while an adjacent doublet at −137.58 ppm and a singlet at −120.74 ppm were generated, corresponding to 5a and fluoride, respectively. With 1.5 equivalents of H₂O₂, the signal for 5 was completely converted to 5a and F⁻. These data are consistent with robust oxidation of 5 by LmbB2 at both C3 and C5. LmbB2 is able to hydroxylate at C5 to produce 5a, and also at C3, cleaving the aromatic C-F bond generating 1a and the loss of the fluorine atom as a fluoride anion. It is noticeable that at least 1.5 equivalents of H₂O₂ were required to perform the complete conversion. Taken together, the HPLC, MS and NMR evidence of 5a, 1a and F⁻ demonstrate the strong capacity for enzyme hydroxylation at C3 or C5 positions, despite the difficulty of C-F bond cleavage. Moreover, the continuous formation of 1a and F⁻ is evidence that the heme catalytic center returns to the resting state after each turnover of substrate-based C-F bond hydroxylation and the creation of 1a and F⁻ is the result of enzymatic catalysis.

Figure 5.

(A) ¹⁹F NMR spectra of in-tube reactions. (Black) 3-fluoro-tyrosine (5) and LmbB2; (Red) with 0.5 eq. of H₂O₂, 5 was partially converted to 3-fluoro-5-hydroxyl-L-tyrosine (5a) and fluoride; (Blue) with 1.5 eq. of H₂O₂, 5 was almost completely converted to 5a and fluoride. The insets show the zoomed-in spectra for 5 and 5a. (B) The HPLC analysis. The decay of 5 (black triangle) and the formation of DOPA (1a, blue circle) and 5a (blue square) had a linear relation to H₂O₂ concentration. (C) In the presence of 1 mM H₂O₂, the oxygen production rate as a function of the concentration of L-tyrosine (1) and 5. The fitting method is described in the Supporting Information.

The evidence of C3 hydroxylation with the C-F bond cleavage and formation of 1a as well as C5 hydroxylation and formation of 5a prompted further investigation into the mechanism. To investigate whether the formation of C3 and C5 hydroxylation products is processive versus an independent process, the enzymatic reaction of 5 was examined with increasing concentrations of peroxide and analyzed by HPLC. It should be noted that the peroxide addition for each targeted concentration was not made at once, but with multiple additions at a smaller quantity to avoid undesired protein damage which occurs at high peroxide concentrations. Plotting the integrated peak area of 5, 5a and 1a as a function of the H₂O₂ concentration revealed that substrate decay and the two product formations were linearly dependent on H₂O₂ concentration (Figure 5B). The two products were formed with constant rates during the peroxide titration, eliminating the possibility of multiple oxidations and showing that neither is intermediary of the other. Such an observation suggests that both products are derived directly from 5 via two, independent, non-processive enzymatic reactions. Regarding the overall chemical reactions, oxidative C-H bond cleavage and non-oxidative C-F bond cleavage differ by two electrons because fluorine leaves as an anion, implicating different mechanisms for the two reactivities. The most likely two-electron source for the C-F bond cleavage pathway would be the oxidation of H₂O₂ to oxygen (i.e., catalase-like activity) since more than one equivalent of H₂O₂ was required to achieve complete conversion of 5 to 1a (Figure 5A); therefore, the generation of 1a and fluoride could be accompanied by oxygen generation.

To detect the possibility of O₂ formation, we employed an oxygen electrode to compare the O₂ formation rate between the natural substrate 1 and analog 5 when mixed with LmbB2 at a saturated, 1 mM H₂O₂ concentration (Figure 5C). Without any organic substrate, LmbB2 shows a slow catalase activity of 0.9 s^-1. When substrate concentration was increased, less oxygen was generated due to the hydroxylation reaction of 1 competing with the catalase activity. The loss of the catalase-like activity upon substrate addition was more pronounced for the natural substrate 1 versus the substrate analog 5, which reflects the more rapid reaction and consumption of H₂O₂ for hydroxylation of 1. When substrate concentrations exceeded 200 µM, the rate of oxygen formation gradually approached zero and 10 min⁻¹ for 1 and 5, respectively. This result suggests a hypothesis that H₂O₂ is the primary two-electron source in the C-F bond cleavage pathway. When LmbB2 reacts with 5 to continuously form 1a and F⁻, more than one equivalent of H₂O₂ is reacted because it acts as both the oxygen source and the electron source in two half-reactions to bring the enzyme back to the resting state after each catalytic turnover.

After obtaining O₂ production during the catalytic turnover of 1 and 5, we sought to model the catalase inhibition assays kinetically. For the case of 1, the competition assay could be well described by a model in which 1 competes with H₂O₂ for the free enzyme (Scheme S1, Figure S6). However, in the competition assay using 5, the O₂ production rate did not approach zero, and the first model could not explain the catalytic behavior (Figure S6). Therefore, a second model which allows the production of DOPA concomitant with an oxidized enzyme was proposed to fit the change of the O₂ production rate of 5 (Scheme S2 and Figure S7). Currently, however, many of the microscopic rate constants for the reaction are unknown, so the modeling should be considered as qualitative rather than quantitative.

The source of oxygen determined by isotope labeling and redox potential measurement

To determine the oxygen source of the added hydroxyl in the LmbB2-catalyzed reaction, isotope labeling studies were performed. Reactions were carried out with 1 and 5 in ¹⁸O-enriched water (88% ¹⁸O) or H₂O₂ (≥ 90% ¹⁸O). When ¹⁸O-labeled H₂O₂ was used to react with 1 in normal ¹⁶O-water, the product 1a has over 90% enriched ¹⁸O in the newly incorporated hydroxyl group (Figure 6A, left panel), which indicates that H₂O₂ is the source of the new oxygen atom. Similarly, when using normal H₂O₂ and in ¹⁸O-labeled H₂O, 87.7% of the 1a produced was not isotopically enriched with ¹⁸O (Figure 6A, right panel). The 11.2% ¹⁸O enrichment when using labeled water derives from isotope scrambling presumably with either a substrate-based, possibly radical, intermediate or the heme iron-bound oxygen responsible for O-atom transfer.

Figure 6.

High-resolution mass spectrometry analyses of the isotope labeling reaction using ¹⁸O-enriched H₂O₂ (left panels) or H₂O (right panels). Only DOPA (1a) was produced when L-tyrosine was used as a substrate (A), while both 1a and 3-fluoro-5-hydroxyl-tyrosine (5a) (B) were observed using 3-fluoro-L-tyrosine as substrate. 1a and 5a have molecular weight of 198.0761 and 216.0667 Da with natural abundance.

Similar reactions were performed using 5, which yielded two products: 5a through hydroxylation and 1a via substitution of fluorine. When using labeled peroxide, the majority of both products contained one ¹⁸O atom, 84.5% and 69.7% (Figure 6B, left panel), respectively, similar as was observed for 1. Significantly more scrambling is observed when 5 was used as the substrate, presumably due to longer lifetimes of catalytic intermediates. When using labeled water, the majority of 5a produced was not enriched with ¹⁸O (¹⁶O¹⁶O, 57.3%). However, significant scrambling was observed (¹⁶O¹⁸O, 35.8%) including doubly labeled 5a (¹⁸O¹⁸O, 6.9%) (Figure 6B, right panel). Observation of a doubly labeled product implies that aromatic 4-hydroxyl group becomes very active in at least one intermediate and can exchange with bulk water. The dehalogenation reaction of 5 to 1a performed in labeled water showed the most significant degree of scrambling (Figure 6B, right panel), 51.5% singly labeled and 12.3% doubly labeled, presumably because it has the longest-lived intermediate(s). Nevertheless, observation of non-enriched 1a (36.2%) in excess of the residual unlabeled water (12%, calculated ¹⁶O percentage) and the majority of 1a labeled when using labeled peroxide confirms that the source of oxygen for the dehalogenation is indeed H₂O₂. Distribution of the ¹⁸O- incorporated product(s) is summarized in Table 1. As a control, the unreacted substrates, 1 and 5, remained as purely unlabeled in all reactions (Figure S8). To better understand the unusual reactivities for LmbB2 and explain the atypical C-H and C-F bond hydroxylation, we also performed Fe³⁺/Fe²⁺ redox potential measurement with a dye-coupled assay.²⁷ LmbB2 and Nile Blue (E_m = −116 mV) were reduced by receiving electrons from xanthine/xanthine oxidase simultaneously (Figure S9A). The linear fitting of the Nernst plot indicated the redox potential of LmbB2 as −98 ± 2 mV at pH 7.0, room temperature (Figure S9B). The determined redox potential value is at the high end of typical histidyl-ligated peroxidase reported (−250 to −100 mV) and higher than most thiolate-ligated heme enzyme (−340 to −6 mV).^{28, 29}

Table 1.

Distribution of the ¹⁸O-labeled product(s) after the LmbB2 reaction using ¹⁸O-enriched H₂O or H₂O₂. L-Tyrosine and 3-fluoro-L-tyrosine were used as the substrate in independent experiments. The standard deviation numbers were obtained from three replicate experiments.

Substrate	18O atom %		Hydroxylation			Dehalogenation and Hydroxylation
	Water	Peroxide	16O/16O	16O/18O	18O/18O	16O/16O	16O/18O	18O/18O
Tyrosine	-	≥ 90	7.2 ± 0.3	92.8 ± 0.3	-	-	-	-
	88	-	87.7 ± 0.9	11.2 ± 0.8	1.0 ± 0.1	-	-	-
3-Fluoro-L-tyrosiTane	-	≥ 90	15.5 ± 0.2	84.5 ± 0.2	-	30.3 ± 1.0	69.7 ± 1.0	-
	88	-	57.3 ± 1.6	35.8 ± 1.6	6.9 ± 0.7	36.2 ± 0.9	51.5 ± 0.7	12.3 ± 1.2

Discussion

Previous work described the LmbB2 protein and its function.^{11, 17} By rigorously studying enzymatic activity on a variety of substrates this study resulted in the mechanistic understanding. LmbB2 has been previously classified as a unique member of the peroxidase superfamily,¹⁷ whose typical function is to oxidize organic and inorganic substrates by H₂O₂ and yield one-electron oxidized products and water molecules. The data presented demonstrate that LmbB2 cleaves a C-H bond, or a C-X bond if it is present, during hydroxylation. A histidyl-ligated heme enzyme promoting hydroxylation is unusual, because such a catalytic function is commonly found in two superfamilies of enzymes, cytochrome P450s and peroxygenases, which both utilize a thiolate-ligated heme. Thus, there is a pressing need for mechanistic understanding of the tyrosine hydroxylase catalyzed reaction.

The observation of a histidyl-ligated heme enzyme to promote hydroxylation concomitant with both C-H and C-F bond cleavages is unprecedented. The aromatic C-F bond is believed to be one of the strongest bonds, and thus not easily dissociated.³⁰ Several synthetic nonheme iron complexes were reported to be able to perform C-F bond cleavage during oxygenation.^{31, 32,33} The cleavage of C-H and C-F bonds by a single biological iron center is also very rare. Such a reaction has been reported for nonheme iron-dependent thiol dioxygenases working on a protein-based cofactor with only a single turnover.^34–36 Thiolate-ligated cytochrome P450s are well known for hydroxylating aromatic C-H bond,³⁷ however, hydroxylation of an aromatic C-F bond is only reported for hexafluorobenzene or para-substituted anilines and phenols.^{38, 39} No histidyl-ligated heme enzymes have been reported for C-F bond hydroxylation, and certainly none for promoting both C-H and C-F bond hydroxylation with multiple turnovers as LmbB2 does.

The most relevant known example to this study is another histidyl-ligated heme enzyme, which catalyzes the two-electron oxidation of trihalophenols at the para position resulting in quinone and halides.^{40, 41} Studies have shown that trifluorophenol had the least activity among all the halogenated substrates, and high-valent heme iron species oxidizes substrate with water as the source of oxygen instead of peroxide.^{42, 43} In contrast, LmbB2 catalyzes oxidative C-H bond and non-oxidative C-X bond hydroxylation on the meta positions of L-tyrosine analogs to form quinol and halides using H₂O₂, and the fluoro-substituted tyrosine has the highest dehalogenation reactivity among all the halogenated substrates. Such differences indicate that the governing factor of the catalysis is distinct between LmbB2 and dehaloperoxidase. Thus, LmbB2 is unique for its chemistry since it is the first example of substrate-based C-H and C-F bond hydroxylation by a histidyl-ligated heme enzyme with multiple turnovers.

The fluorinated compound (5) yields the highest dehalogenated product among halogen-substituted substrates is quite intriguing because C-F bond is more durable than C-Cl and C-I bonds. Presumably, the small size of a fluorine atom allows the aromatic ring of L-tyrosine to access both binding orientations at the heme, i.e., fluorine can point toward or away from the heme center. Therefore, hydroxylation at both C3 and C5 of 5 would be possible. The formation of two hydroxylated products in LmbB2 reactions with 6 and 7 again indicates that the active site can accommodate the binding of L-tyrosine analogs with even larger halogen substituents in both possible orientations at the heme. However, the order of yield in 1a (F > Cl > I) inversely reflects the order of the size of the halogen atoms (F < Cl < I). Besides the binding affinities (F > Cl > I), one possible explanation is that the larger halogen atom favors binding in such a way that it is pointed away from the heme, and therefore, there is less of a chance to have the product of C-X hydroxylation. Additionally, the larger atom has a more significant steric effect making the 5- position even less accessible for bond cleavage by the active heme species. Another plausible reason is that fluorine is the most electronegative of the halogens which makes the adjacent carbon more electropositive and therefore more likely to undergo nucleophilic attack, ultimately generating 1a. Therefore, the order of 1a production (F > Cl > I) in the LmbB2 reaction accords with the order of electronegativity of halogen substituents (F > Cl > I).

Isotope ¹⁸O labeling experiment with the reaction of 1 indisputably indicated the oxygen source for normal hydroxylation comes from H₂O₂, while the intermediate in the natural reaction is short-lived with little chance to exchange with water molecules. Using 5 as the substrate, both defluoronated and hydroxylated products, 1a and 5a, have a significant scrambling effect and double ¹⁸O incorporation even though H₂O₂ is still the oxygen source of hydroxylation on both of the C-H and C-F sites. The ¹⁸O labeling study has provided important mechanistic insights into the LmbB2 pathways. The fact that 4-hydroxyl can exchange with water suggests the aromaticity is broken during catalysis at a certain stage; a ketone intermediate is plausible.⁴⁴ The newly formed hydroxyl was also exchangeable in both hydroxylation and dehalogenation, which implies a substrate-based intermediate or the iron-bound oxygen that is responsible for O-atom transfer. Reaction with 5 was slow, as compared to 1, by the fluorine substituent and future attempts to capture intermediates under certain conditions may be possible. Finally, together with the findings of consumption of more than 1 equivalent of peroxide and generation of fluoride and oxygen during the C-F bond cleavage with multiple turnovers, we proposed two separate mechanistic pathways for C3 versus C5 hydroxylation of L-tyrosine derivatives.

When the functional group points away from the heme center, hydroxylation occurs via the usual mechanism (Figure 7A). We propose that after organic substrate binding, peroxide binds to the ferric heme center and forms a histidyl-ligated Compound I (cpd I, an iron(IV) oxo porphyrin cation radical)-like species, which receives one electron from the tyrosine or the reactive analogs as the first step of the substrate oxidation. Initiating the reaction by one e⁻ oxidation may be preferred because histidyl-ligated hemes are typically considered competent for binding oxygen or electron transfer. However, its capacity to directly abstract a hydrogen atom-like thiolate-ligated heme is questionable.^45–47 The Cpd I species will be reduced to a deprotonated Compound II (Cpd II, an iron(IV) oxo porphyrin), after one electron transfer to generate a tyrosyl cation radical. A nearby active-site base could easily deprotonate the tyrosyl cation radical due to the dramatically decreased pK_a of the tyrosyl cation radical compared with tyrosine.⁴⁸ Then, the ferryl-oxo attacks the radical generating an intermediate-bound adduct and finally forms the derivative DOPA products.

Figure 7. Proposed reaction pathway for LmbB2 hydroxylation.

Oxidative aromatic C-H bond cleavage generates product DOPA or its derivatives. A Compound I-like intermediate performs electron transfer in this model (A). An alternative mechanism (B) like that of cytochrome P450s’ suggests that the Compound I-like intermediate abstracts a hydrogen atom. R = main chain of the amino acid; R’ = H, NO₂, F, Cl, or I.

An alternate reaction pathway is also proposed (Figure 7B), in which the Cpd I-like intermediate would directly abstract an H-atom from the 4-hydroxyl, like most P450s do, to generate protonated Cpd II and a substrate-based tyrosyl radical; the radical will migrate to C5 resulting in a ketone-based tyrosyl radical. The Cpd II-like intermediate then proceeds with hydroxyl radical rebound to the C5 position and finally, forms the hydroxylated product after a rearrangement. More mechanistic studies are necessary to reveal the potential of histidyl-ligated heme for hydroxylation and discriminate between these two mechanisms. Nevertheless, we propose that a quinone radical is a crucial substrate-based intermediate of the catalytic pathway based on the data from unreactive analogs, 2 and 3, and the scrambling result from isotope labeling.

Since halogens are electronegative, C3 is partially electropositive in 5, 6, and 7. Therefore, ferric heme-bound hydroperoxide can perform a nucleophilic attack at C3 once they are close enough, generating a high-valent heme species. Fluoride then leaves the ring to create a re-aromatized product, DOPA (Figure 8). The resulting Cpd I-like species can either activate another molecule of the organic substrate or react with another equivalent of H₂O₂ to return to the resting state via catalase-like activity, which likely precludes the possibility for the leaving halide to bind to the heme Fe to inactivate the enzyme. Our kinetic modeling (Scheme S2) suggests that organic substrates are unable to compete with H₂O₂ for the Cpd I-like species effectively. A flavin-dependent monooxygenase was proposed to achieve dehalogenation while forming a quinone via an electrophilic aromatic substitution (EAS) mechanism.⁴⁹ While an electrophilic substitution mechanism cannot be positively excluded in LmbB2 without further experimental evidence, the formation of quinol and F⁻ favors a nucleophilic substitution mechanism because the products are two electrons reduced compared to the EAS mechanism. Further, the fluorinated phenol ring is relatively electron-deficient and would make a weak nucleophile. The natural substrate of LmbB2 has a neutral C3 which makes C-H bond cleavage proceeding from nucleophilic attack very unlikely; however, the inductive effect caused by an electronegative group at C3 makes a nucleophilic attack at this position more likely, and nucleophilic substitution of fluorinated aromatics is common in organic chemistry.^{50, 51} Altogether, the steric effect and inductive effect of the halogen substituents distributes substrate between two distinct reaction pathways.

Figure 8.

Proposed mechanistic model for the carbon-halogen (C-X) bond cleavage. The iron-bound hydroperoxo intermediate is proposed to perform aromatic nucleophilic substitution, leading to cleavage of the aromatic C-X bond with DOPA formation as the only hydroxylation product. The Compound I-like intermediate subsequently performs a catalase-like reaction with O₂ liberation as a side product. R = main chain of amino acid, and X = F, Cl, or I.

The fluorine substituents on the substrate-like molecules are often considered when developing new drug candidates. Due to their unique steric and lipophilic properties, among others,^52–54 halogenated compounds have now been widely used against a large number of biomedically essential targets. Our findings here call attention to the reality that enzymes, especially metalloenzymes, are often powerful enough to cleave off the covalently bound halogens and, therefore, destroy such intentionally designed halogenated drugs. However, these findings may also be used to help develop new catalysts to mitigate the potential health threats of halogenated aromatic hydrocarbons (HAHs) that contaminate the natural environment as an unintended and undesired effect of civilization. HAHs are not only significant contaminants of the air, drinking water and sediment, but they have been linked to cardiotoxicity and the DNA damage that leads to cancer due to inherent reactivity, such as, aryl hydrocarbon receptor activation.^55–57 Designing catalysts and engineering biosystems capable of degrading such toxic and fatal HAHs have long been a goal of scientists,^{41, 58–60} and most catalysts activate C-Cl, C-Br or C-I bond but have very little or no reactivity on the C-F bond. However, LmbB2 reported here has precisely the opposite activation tendency that defluorination is actually the most efficient and only requires the inexpensive oxidant, H₂O₂, which provides a potential new scaffold for engineering enzymatic degradation of HAHs, especially fluorinated aromatics, through hydroxylation of the carbon-halogen bonds.

Conclusion

This study shows that the histidyl-ligated, heme-dependent L-tyrosine hydroxylase LmbB2 has a wide range of catalytic activities towards deactivated L-tyrosine analogs, as long as the 4-hydroxyl group is present. The mono-substituted tyrosine analogs represented by 3-fluoro-L-tyrosine presumably bind in two different orientations at the heme active site, substituents pointing away from or toward the heme center. The former orientation results in an oxidative C-H bond cleavage at C5 while the latter allows for different chemistry to happen on C3; for example, the carbon-fluorine bond hydroxylation to generate DOPA. Based on spectroscopic characterization and product analysis, we proposed two distinct pathways to explain how LmbB2 accomplishes the catalytic reaction upon different substrate orientations. The surprisingly robust heme-oxidant of LmbB2 sheds new light on the capacity for heme enzymes to perform a wide range of chemistries. Finally, this study is the first to describe a biocatalytic C-H and C–F bond cleavage through hydroxylation at the carbon site promoted by a histidyl-ligated heme enzyme. This finding broadens our understanding of fluorine chemistry and provides potentially novel medicinal and environmental applications for hemoproteins.

Materials and Methods:

Chemicals

L-Tyrosine and all of the analogs were purchased with the highest purity as listed below: 1, Alfa Aesar, 99%; 2, Acros Organics, 98.5%; 3, Sigma-Aldrich, 98%; 4, Sigma-Aldrich, crystalline; 5, TCI, 98%; 6, Sigma-Aldrich, 97%; 7, Alfa Aesar, 98%; N-boc-L-tyrosine, 98%, Sigma-Aldrich; p-Cresol, 99%, Sigma-Aldrich. Stock solutions were made in 0.1 M HCl.

Enzyme overexpression and purification

LmbB2 from Streptomyces lincolnensis with His₆-tag at the N-terminus cloned into a pET16 plasmid with ampicillin resistance was a generous gift from Prof. Wolfgang Piepersberg (Mikrobiologie, Bergische Universität GH Wuppertal, Germany). pET16B2 was transformed into an E. coli BL21 (DE3) overexpression system. Expression of LmbB2 was started in Luria Bertani (LB) medium with 100 μg/mL ampicillin at 37 °C. At OD_{600 nm} value of 0.3, δ-aminolevulinic acid and ferrous ammonium sulfate were added to a final concentration of 20 mg and 10 mg per liter culture. The cells were induced at an OD_{600 nm} of 0.6 using a final concentration of 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) with continued incubation overnight at 28 °C and 220 rpm agitation. The cells were harvested 12 hours after IPTG induction by centrifuging at 6000 × g for 20 mins at 4 °C.

To isolate LmbB2, the cells were re-suspended in buffer A (50 mM KPi buffer with 200 mM NaCl at pH 8) supplemented with 1 mM PMSF protease inhibitor and DNAse I (0.05 mg/mL) and lysed using a Microfluidizer LM20 cell disruptor. The supernatant was recovered after centrifugation (27,000 × g, 4 °C for 30 min) and applied to a Ni-affinity column pre-equilibrated with buffer A. LmbB2 was eluted by a gradient profile using buffer A and B (50 mM KPi buffer with 200 mM NaCl, 500 mM imidazole at pH 8). The eluted protein was collected and concentrated using an Amicon centrifugal filter (Ultra-15 10K, Millipore) and then desalted into 50 mM KPi pH 8 and 50 mM NaCl with additional 5% glycerol using Sephadex G-25 column (GE Healthcare) and stored at −80 °C for future use.

Catalytic reaction setup and HPLC analyses

The standard HPLC assays were conducted at room temperature in 100 mM KPi with 50 mM NaCl at pH 8. 100 µM enzyme was incubated with 3 mM L-tyrosine or analogs for 5 mins prior to H₂O₂ addition. To avoid any heme bleaching, 20 mM H₂O₂ stock solution was titrated to initiate the reaction by adding multiple small volumes each time until a final concentration of 3 mM was reached. Enzymatic assays and controls conducted were of four types: (i) 100 µM enzyme, 3 mM substrate and 3 mM H₂O₂ and three controls with (ii) 100 µM enzyme and 3 mM substrate, (iii) 3 mM substrate and 3 mM H₂O₂, and (iv) 100 µM enzyme and 3 mM H₂O₂. For 2 and 3, concentration up to 20 mM, pH 7–10 (out of this range, the enzyme became unstable) and two oxidants (H₂O₂ and peracetic acid) were examined. After reaction for 10 min, 10 µL of 6 M HCl was used to quench the reaction. The final volume for each reaction/control was 200 µL. After removing the precipitant by centrifugation, the supernatant was filtered using 10 kDa molecular weight cut off centrifugal filter unit (Millipore). 10 µL filtrate was injected to an InertSustain C18 column (5 µm particle size, 4.6 × 100 mm, GL Sciences Inc.), and then analyzed by a Thermo Scientific Ultimate-3000SD HPLC rapid separation system equipped with a photodiode array detector. The UV-Vis spectra were recorded at full range from 190 nm to 800 nm. The HPLC method was derived from a published method using isocratic elution profile (3/97/0.1, ACN/H₂O/FA).⁶¹ All the HPLC profiles are shown with absorbance at 280 nm except assays for 2 were at both 257 and 280 nm. The peaks eluted from HPLC were then collected and concentrated by SpeedVac (ThermoFisher) for further MS and NMR analysis.

Isothermal titration calorimetry (ITC) measurements

A Microcal VP-ITC system (Malvern Instruments) was used to conduct the ITC measurements as previously described.⁶² The titration buffer consisted of 100 mM KPi with 50 mM NaCl at pH 8. L-tyrosine or analogs were prepared in titration buffer and injected to the cell containing 2 mL of 100–120 μM purified enzyme. Binding of L-tyrosine and its analogs was assessed in a total volume of 300 µL at the following concentrations: 1.5 mM 1, 10 mM 2 and 20 mM 3. After the temperature was equilibrated to 20 °C, a total of 50 injections were performed with a reference power of 15 μcal/s and a stirring speed of 350 rpm. The ITC data were processed and analyzed using non-linear least squares curve fitting of one-site binding models with Origin version 7.0 (OriginLab Corp.) software.

Electron paramagnetic resonance (EPR) analysis on nitrosyl complex

Ferrous heme enzyme was made by reducing 200 µM argon-saturated enzyme with 1 mM sodium dithionite anaerobically. The argon-saturated enzyme was prepared using the published method.²² Reduced LmbB2 was exposed to excess NO released by diethylamine NONOate (Cayman) in a sealed O₂-free vial. Substrate 1, analogs (2 and 3) and unbounded phenols (N-boc-L-tyrosine and p-cresol) were added with a final concentration of 20 mM. All samples were frozen in 4-mm quartz EPR tubes by liquid nitrogen. X-band EPR spectra were recorded by a Bruker E560 spectrometer at 9.4 GHz microwave frequency with an SHQE-W resonator at 100 kHz modulation frequency equipped with a cryogen-free 4 K temperature system as described earlier.⁶³ Spectra for nitrosyl samples were collected at 50 K, 0.05 mW power. The g values reported were obtained by inspection of the EPR line shape.

High-resolution mass spectrometry

Mass spectra were collected on a maXis plus quadrupole-time of flight mass spectrometer equipped with an electrospray ionization source (Bruker Daltonics) and operated in the positive ionization mode. Samples were introduced via a syringe pump at a constant flow rate of 3 μL/min. Instrumental parameters used were standard preset values for small molecules. Important parameters are summarized as follows: capillary voltage, 3500 V with a set end plate offset of 500 V; nebulizer gas pressure, 0.4 bar; dry gas flow rate, 4.0 L/min; source temperature, 200 °C; quadrupole ion energy, 3.0 eV; collision energy, 5.0 eV. Scans were collected at a rate of one scan per second in the range of 50 ≤ m/z ≤ 1500, and 60 seconds of data were averaged to yield a final spectrum. Compass Data Analysis software version 4.3 (Bruker Daltonics) was used to process all mass spectra.

¹H and ¹⁹F nuclear magnetic resonance (NMR) spectroscopy

¹H and ¹⁹F NMR spectra were recorded on a Bruker (Billerica, MA) Avance III HD 500 MHz spectrometer equipped with a 5-mm Cryoprobe Prodigy at 300 K. Spectra were recorded in 90/10 buffer/D₂O. One-dimension ¹H spectra (zg30) were recorded with 1 s relaxation delay, 32 k data points, and multiplied with an exponential function for a line-broadening of 0.3 Hz before Fourier transformation. One-dimension ¹H-coupled ¹⁹F spectra (zgflqn) were recorded with 5 s relaxation delay, 128 k data points, and multiplied with an exponential function for a line-broadening of 5 Hz before Fourier transformation and referenced to internal trifluoroacetic acid (−76.5 ppm). Each peroxide concentration-dependent ¹⁹F NMR spectra were recorded with 64 scans. All NMR data were processed using MestReNova NMR v11.0.3 software. The detailed NMR data are reported in the Supporting Information.

Oxygen determination

An oxygen electrode (Oxygraph, Hansatech Instruments) was used to measure the oxygen production in LmbB2 reactions with substrate 1 or 5 at room temperature. A total volume of 1 mL reaction consisted of 0.5–2 μM enzyme, 1 mM H₂O₂ and various concentration of substrate (0–0.65 mM) in 50 mM KPi with 50 mM NaCl at pH 8. The enzyme was pre-incubated with the substrate in a sealed electrode chamber with constant stirring. The reaction was initiated by addition of H₂O₂ into the electrode chamber. Heated denatured enzyme reacting with peroxide was used as a blank for subtraction. Substrate concentration was recorded as [S]; enzyme concentration was recorded as [E]; oxygen production rate was recorded as ΔO₂. Oxygen production rate per unit of enzyme ΔO₂/[E] versus [S] was plotted to compare the difference between substrate 1 and 5. The detailed fitting procedure is shown in supporting information.

Redox potential determination

The method of redox potential measurement was derived from a reported method for determination of redox potential in heme proteins.²⁷ The assay contained 6 μM LmbB2, 300 μM xanthine and 40 μM Nile Blue in 1 mL of anaerobic 100 mM KPi, pH 7.0 at room temperature. 50 nM xanthine oxidase was added to initiate the reduction and UV-vis spectra were recorded every half minute for 25 min. 5 mM dithionite was added at the end to obtain a fully reduced spectrum. Absorbance changes corresponding to heme and dye reduction were measured at 404 and 636 nm. The potential was given versus a normal hydrogen electrode.