Bacterial cytochrome P450 for oxidative halogenated biaryl coupling

Abstract

Biaryl motifs are fundamental structural elements in many pharmaceuticals, agrochemicals, and advanced materials. Traditional synthetic approaches for biaryl bond formation often require harsh conditions, costly catalysts, and pre-functionalized starting materials, which limit their efficiency, sustainability, and substrate scope. Enzymatic catalysis offers a greener alternative. However, biocatalysts capable of directly coupling halogenated biaryl compounds remain largely underexplored. Here, we report the functional characterization of the marine-derived cytochrome P450 enzyme Bmp7, which catalyzes the formation of halogenated biaryls. We first characterized the product profile of recombinant Bmp7 using its native substrate 2,4-dibromophenol (1) and confirmed the dominant ortho-ortho C-C homocoupled product as MC21-A. Screening a halogenated aromatic substrate library revealed that Bmp7 binds and catalyzes the coupling of 17 halogenated phenols, as evidenced by spectral shift assays, LC-HRMS, HRMS/MS and GC-MS analyses. Two homocoupled products were structurally confirmed by NMR analysis to possess ortho-ortho C-C linkages. In addition to efficient homocoupling, Bmp7 catalyzed heterocoupling reactions between substrate 1 and 16 other substrates, producing mixtures of homocoupled and heterocoupled halogenated biphenols. X-ray crystallography revealed the binding of two substrate 1 molecules within the active site, while DFT calculations supported a single-radical reaction mechanism, shedding light on the mechanistic basis of the coupling reaction. Together, these findings lay the groundwork for these findings establish a foundation for future efforts in enzyme engineering and the development of biocatalytic strategies for synthetic applications.

Graphical Abstract

Introduction

The biaryl motif is a key structural feature found in many pharmaceuticals, agrochemicals, and advanced materials (Figure 1), where it contributes critical properties such as stability, bioactivity, and electronic functionality. Traditional synthetic strategies for biaryl bond formation, particularly transition metal-catalyzed cross-coupling reactions (e.g., Suzuki, Heck, and Negishi), have significantly advanced the field by enabling the selective functionalization of halogenated arenes (Figure 1).^1–2 However, these methods often require harsh reaction conditions, including elevated temperatures, toxic solvents, and heavy metal catalysts such as palladium and nickel. These factors not only limit functional group tolerance but also pose environmental and economic concerns due to the generation of hazardous byproducts (e.g., organotins and organoboronates).³ As the demand for sustainable and efficient synthetic methodologies grows, the development of environmentally friendly alternatives is imperative to overcome these challenges and expand the accessibility of biaryl-containing molecules.

Figure 1.

A. Biaryl moieties in selected pharmaceuticals sotorasib and vericiguat and bioactive natural product nigerone. B. Representative chemical synthesis of biaryls via oxidative and reductive cross-coupling reactions. C. Biaryl couplings catalyzed by cytochrome P450s KtnC and DesC. D. Present work was designed to characterize cytochrome P450 Bmp7 for homo- and hetero-coupling of halogenated phenols.

Biaryl coupling mediated by multiple enzyme classes (Figure S1) plays a crucial role in the biosynthesis of structurally diverse natural products.^4–5 For example, cytochrome P450 enzymes catalyze oxidative C-C coupling reactions in natural product biosynthesis, as seen in the generation of vancomycin^6–8 and several families of ribosomally synthesized and post-translationally modified peptides (RiPPs).^9–11 Laccases, multicopper oxidases, mediate phenol coupling in lignin biosynthesis and plant secondary metabolism,^12–13 offering broad substrate scope and compatibility with mild reaction conditions. These enzymes have inspired the development of biocatalytic strategies for biaryl bond formation as sustainable alternatives to traditional chemical methods.^{17, 18} Notably, a seminal study by the Narayan group demonstrated the potential of the engineered fungal P450 enzyme KtnC (Figure 1),^14–15 which exhibits an expanded substrate scope and enhanced site- and regio-selectivity for biaryl synthesis in the whole cell platform. Furthermore, several bacterial P450s have been engineered to synthesize bioactive natural and unnatural compounds such as arylomycin and mycocyclosin (Figure S1B),^16–18 often through intramolecular C-C biaryl coupling. Despite these advances, the biocatalytic construction of halogenated biaryls, which serve as key intermediates for the synthesis of complex compounds,^19–20 remains largely underexplored.²¹ Addressing this gap could significantly expand the utility of biocatalytic biaryl coupling, providing a more sustainable and selective route for halogenated biaryl synthesis, particularly in the development of pharmaceuticals and agrochemicals.

While synthetic polyhalogenated aromatics are widely used in flame retardants, pesticides, and pharmaceuticals, their toxicity and bioaccumulation raise significant environmental concerns.²² In contrast, marine γ-proteobacteria, such as Pseudoalteromonas and Marinomonas, naturally produce polybrominated diphenyl ethers (PBDEs),^23–24 bioactive halogenated metabolites with ecological relevance.²⁵ The Moore group identified the cytochrome P450 enzyme Bmp7 as the key catalyst for biaryl coupling reactions in PBDE biosynthesis.²⁴ This enzyme converts 2,4-dibromophenol (1) and 2,3,4-tribromopyrrole into homocoupled and heterocoupled products with substantial regioflexibility (Figure S2), primarily forming C-C linkages, alongside minor C–O–C products. Recent studies further demonstrate that Bmp7 tolerates several structural analogs of 2,4-dibromophenol such as bromocatechols,²⁶ positioning it as a promising biocatalyst for halogenated biaryl synthesis. However, the substrate scope of Bmp7 toward halogenated aromatics, particularly halogenated phenols, remains to be characterized, and the molecular basis underlying its biaryl coupling mechanism is not well understood. These gaps limit the effective utilization and engineering of Bmp7 as a biocatalyst.

Here, we further characterize Bmp7 to aid its future biocatalytic applications. We first investigated its catalytic performance using the native substrate 1, followed by screening a library of 38 halogenated aromatic compounds, primarily halogenated phenols, to evaluate its substrate scope. LC-high-resolution mass spectrometry (HRMS) and GC-MS analyses confirmed biaryl homo- and hetero-coupling across both natural and synthetic substrates. NMR characterization of three isolated products confirmed that ortho–ortho C-C bond formation was the dominant coupling mode. X-ray crystallography provided key structural details of the binding of two molecules of substrate 1 within the active site, while DFT calculations supported a radical-mediated coupling mechanism. These findings pave the way for developing Bmp7 as a versatile biocatalyst for halogenated biaryl synthesis.

Results and Discussion

Biochemical characterization of recombinant Bmp7

Recombinant Bmp7 of Marinomonas mediterranea, as well as its redox partners Bmp9 and Bmp10, was expressed in E. coli and purified via single-step affinity chromatography (Figure S3A). The oligomeric state of Bmp7 was assessed as a monomer in solution by size-exclusion chromatography (Figure S3B). UV-Vis spectroscopy of Bmp7 revealed a characteristic 450 nm peak upon the addition of CO to the sample, confirming the presence of properly folded P450 enzyme (Figure S3C). To evaluate substrate binding, we performed isothermal titration calorimetry (ITC) and spectral shift analysis using 2,4-dibromophenol (1).²⁴ Initial ITC data appeared monophasic and fit well to a “one set of sites” model, yielding a dissociation constant (K_D) of 2.5 ± 0.4 μM (Figure S4). However, the observed stoichiometry of 1.80 suggested the binding of two substrate molecules per enzyme, implying a two-site binding mode. UV-Vis spectral shift analysis confirmed substrate binding with a similar overall K_D of 11.1 μM. In support of the two-site binding mode suggested by ITC analysis, a Hill coefficient of 4.2 ± 1.7 was determined, indicating cooperative binding (Figure S5). Consequently, the ITC data were reanalyzed using a sequential binding model with two sites, resulting in K_D values of 0.119 ± 0.004 μM and 9.3 ± 0.6 μM for the first and second binding events, respectively (Figure 2A). The absence of clear biphasic features in the integrated heat plot may reflect a poorly populated singly-bound state, wherein the binding of the first substrate molecule rapidly promotes the association of the second. This results in a composite exothermic signal that appears monophasic (Figure S4).

Figure 2.

Biochemical characterization of Bmp7 in the homocoupling of 1. A. Isothermal titration calorimetry (ITC) analysis of Bmp7 binding to 1. The integrated heat plot from successive injections of 1 (750 μM) into Bmp7 (50.4 μM) was fitted using a “sequential binding” model with the number of sites equal to 2. B. HPLC traces of the Bmp7-catalyzed reaction, including heat-inactivated enzyme control. The reactions were monitored at 214 nm. C. Extracted ion chromatogram (EIC) traces showing the dominant and minor homocoupling products in the Bmp7 reaction. The MS analysis was performed in negative mode. D. GC-MS traces of the Bmp7 reaction and heat-inactivated control. Commercially available MC21-A served as the authentic standard. All samples were derivatized with trimethylsilyl (TMS) groups prior to analysis.

We next examined Bmp7’s catalytic activity toward 1. Reactions were performed aerobically in the presence of Bmp9, Bmp10, and 500 μM NADH. HPLC analysis revealed the conversion of 1 (0.25 mM) into a new product within 10 hours (Figure 2B), which was absent in the heat-inactivated enzyme control. HRMS identified the product as a dimer of 1, with the most abundant [M-H]⁻ ion at m/z 500.6986, consistent with the expected isotopic pattern of polybrominated compounds (Figures 2C, S6A–B). The [M-H]⁻ ion of three minor trace species at m/z 420.7905 suggested debrominated product analogs (Figures S6A, S6C). Following the previous studies,²⁴ HRMS/MS fragmentation of the major product confirmed a stable C–C linkage (Figure S7) as no dibromobenzoquinone fragment was observed, ruling out C–O–C bond formation. Further characterization of the major product, particularly the number of its −OH groups, was conducted using chemical derivatization and GC-MS analysis (Figure S8). Comparison with an authentic standard confirmed the identity of the compound as MC21-A, a known antibiotic previously isolated from P. phenolica sp. nov. O-BC30T (Figures 2D and S9).²⁷ These results align with the previous study where Bmp7 was shown to form multiple regioisomeric coupling products from 1 with MC21-A as the major product (Figure S2).²⁴

Evaluation of Bmp7 substrate scope with halogenated phenol analogs

The predominant formation of MC21-A from 2,4-dibromophenol (1) by Bmp7 underscores its potential for brominated biaryl synthesis. To assess its substrate scope, we evaluated 38 structurally related aromatic compounds, primarily halogenated phenols (Figures 3 and S10), first using UV-Vis spectral shift analysis to probe substrate binding. Seventeen compounds (100 μM) induced a notable Type I spectral shift of Bmp7 (4.5 nM). Subsequent analysis revealed that binding affinities (K_D) of these compounds ranged from 3.2 to 26.9 μM, with Hill coefficients of up to 4.9 (Figures 3 and S5). Compound 1 exhibited a strong binding, comparable to 4-bromo-2-chlorophenol (11, K_D = 11.3 ± 1.2 μM) and 2,4-dichlorophenol (12, K_D = 11.6 ± 2.1 μM), whereas monohalogenated phenols (4–8) showed weaker interactions (K_D = 16.8–26.9 μM). On the other hand, eight halogenated phenol substrates (10, 13, 14, 18, 21-24) demonstrated tighter binding with Bmp7 than 1 (Figure 3). Other halogenated aromatic compounds induced minimal spectral shifts, suggesting weak or no binding to Bmp7 (Figures 3 and S10).

Figure 3.

Halogenated phenols evaluated for Bmp7 binding and catalytic activity. Substrate binding assays were performed by titrating 4.5 nM Bmp7 in 100 mM phosphate buffer (pH 7.0) with 100 μM substrate, followed by determination of binding affinities using increasing ligand concentrations (n = 2). Enzymatic reactions contained 250 μM substrate, 1 μM recombinant Bmp9 and Bmp10, and 2 mM NADH in 100 mM potassium phosphate buffer (pH 6.8). Reactions were initiated by adding 0.9 μM recombinant Bmp7 and incubated at 30 °C for 10 h (n = 3).

We next assessed Bmp7’s catalytic activity toward all compounds in the substrate library. HPLC and LC-HRMS analyses confirmed that all Type I binders were converted into their corresponding homocoupled dimers (Figures 3 and 4A, Table S2). After 10 hours, conversion ratios ranged from 3% to 86%, with dihalogenated phenols generally exhibiting higher catalytic efficiency than their monohalogenated counterparts. Most active substrates yielded a single predominant product through C-C bond formation as indicated by HRMS/MS²⁴ and GC-MS analysis (Figures S11–25). Notably, GC-MS analysis detected minor products bearing C–O–C linkage in a subset of Bmp7 reactions, except for 21, which showed comparable ion intensities of products with both linkages (Figure S22). To further elucidate the structures of the predominant products, we isolated 1.2 mg of the homocoupled product of 10 from the Bmp7 reaction by semi-preparative HPLC. ¹H and ¹³C NMR analysis confirmed the ortho-ortho C-C linkage (Figure 5A and S26–27). The same ortho-ortho C-C coupling pattern was observed in the homodimeric product of 18+18, as determined by ¹H NMR analysis (Figure 5B and S28). Collectively, HRMS/MS, GC-MS, and NMR analyses suggest that the ortho–ortho C–C linkage likely represents a common coupling mode among the major homocoupled products (Figure 4A). In addition, several minor dimeric products were detected, particularly in reactions involving 14, 21, 23 and 24 (Figure 4A). These isomers might arise from coupling pathways similar to those generating the minor products observed with 1, warranting further investigation.

Figure 4.

LC-HRMS extracted ion chromatogram (EIC) traces of Bmp7-catalyzed homocoupling (A) and heterocoupling (B) reactions with active substrates. The m/z values of selected [M-H]⁻ ions are shown. In the heterocoupling reaction, 1 (0.125 mM) was used as the constant substrate, paired with 0.125 mM of the second substrate. All reactions were incubated for 10 hours. The HRMS analysis was performed in negative mode. C. Relative distributions of homo- and heterocoupled products from 1 and another substrate in the Bmp7 reaction. Total ion intensity was obtained by summing the intensities of homo- and hetero-coupled products, and relative distributions were then calculated for individual ions in each reaction, assuming equal ionization efficiencies for all products.

Figure 5.

Characterization of the homo- and hetero-coupled products. A-B. GC-MS (I) and 1H NMR (II) analyses confirmed the ortho-ortho C–C coupling of homocoupled 10 (A) and 18 (B). Semi-preparative HPLC purification yielded 1.2 mg of the homocoupled product of 10. C. Structural characterization of heterocoupled product of 1+17. I: LC-HRMS EIC traces of the Bmp7 reaction with 1 and 17 as substrates, with the m/z values of selected [M-H]⁻ ions shown. II: GC-MS analysis of the Bmp7 reaction with 1 and 17 as substrates. The chemical structure of the identified heterocoupled dimer following TMS derivatization is shown. III: Structural elucidation of the heterodimer formed between 1 and 17 through ¹H NMR analysis.

The substrate screening provides a few insights into the substrate selectivity of Bmp7, although the library size and diversity were constrained by the commercial availability of relevant compounds. First, both hydroxyl and halogen substituents are essential for activity. Removal of these groups from compound 1 resulted in a complete loss of activity in 1,3-dibromobenzene (2) and phenol (3) (Figure 3A). Moreover, removing either the ortho- or para-bromine in 1 markedly impaired substrate binding and catalytic conversion (6: K_D = 26.7 ± 3.2 μM, 15% conversion; 7: K_D = 22.3 ± 9.8 μM, 13% conversion; Figure 3B). In contrast, Bmp7 exhibited no detectable binding toward 1-bromo-4-methoxybenzene (29) compared with 6 (Figure S10), underscoring the critical role of the substrate hydroxyl group in enzyme binding. Second, the identity of the halogen substituents strongly influences both binding affinity and catalytic efficiency. For instance, replacing the para-bromine in monohalogenated 6 with fluorine yielded an inactive substrate (9), whereas substitution with chlorine (7) or iodine (8) produced similar performance to 6 (Figure 3B). Interestingly, replacing the para-bromine in 1 with chlorine (10) or fluorine (18) enhanced binding affinity (Figure 3C–D), and Bmp7 exhibited its highest conversion toward 18. Although the enzyme tolerated replacement of the ortho-bromine in 1 with chlorine (11) or fluorine (17) in terms of binding, conversions for both were modestly reduced (Figure 3C–D). Similarly, replacing both bromines in 1 with chlorines (12) maintained binding but lowered activity, whereas 2,4-difluorophenol (16) was completely inactive. These results indicate that mono- or di-chlorine substitution is compatible with Bmp7 catalysis, while fluorine substitution is only tolerated when combined with bromine (Figure 3A–D). The combination of fluorine and chlorine substitution awaits further exploration. Third, Bmp7 tolerates methyl substitution. Although no binding was observed for compounds bearing halogen replacements with −NH₂ (27), −OMe (28), or −OH (25, 26), 4-bromo-2-methylphenol (22) remained an active substrate (K_D = 6.0 ± 0.8 μM; 51% conversion; Figure 3E–F). In contrast, replacing both bromines in 1 with methyl groups rendered 20 inactive. Remarkably, 4-bromo-2,3-dimethylphenol (21) displayed comparable binding and activity to 1, suggesting the potential of Bmp7 for generating structurally diverse halogenated biaryls. Fourth, Bmp7 accommodates halogenated phenols with meta-substitutions (Figure 3). For example, 2-bromo-3-chlorophenol (13) exhibited the tightest binding among all tested substrates, while 4-bromo-3-chlorophenol (14) retained strong binding but displayed markedly reduced conversion (Figure 3C). Additionally, 3-bromo-4-methylphenol (23) showed both tight binding and high catalytic turnover (K_D = 7.6 ± 0.8 μM; 80% conversion; Figure 3E), indicating Bmp7’s tolerance toward both methyl and meta-substitutions. The meta-brominated compound 24 was also accepted as a substrate. In contrast, Bmp7 was intolerant of disubstitution on both sides of the phenol ring, as observed with 5-bromo-2-chlorophenol (15). Consistent with this trend, 2,4,6-tribromophenol (31) was also inactive (Figure S10). Finally, Bmp7 exhibited low tolerance toward other classes of halogenated aromatics, including furans, pyridines, and pyrazoles (Figure S10). Due to commercial unavailability, 2,3,4-tribromopyrrole and related analogs were not examined in the present study. Collectively, these findings provide an initial molecular framework for future enzyme engineering aimed at expanding the synthetic utility of Bmp7 in halogenated biphenol formation.

Bmp7 catalyzed heterocoupling of compound 1 with other halogenated phenols

The advanced understanding of Bmp7’s substrate scope prompted us to investigate its capacity to catalyze heterocoupling reactions by incubating the enzyme with mixtures of compound 1 and a second substrate from the screening library. Compound 1 was selected as the constant partner because it is the native substrate of Bmp7 and exhibits high conversion efficiency (Figures 3 and 4A). Reactions were conducted under optimized conditions for 10 hours, using 0.125 mM of 1 and 0.125 mM of a second substrate, irrespective of their binding affinity differences. LC-HRMS and HRMS-MS analyses revealed the formation of heterocoupled dimers between 1 and 4–8, 10–14, 17, 18, 21–24 (Figures 4B, S29–43, Table S2). GC-MS analysis further confirmed C-C bond formation in the dominant heterocoupled products (Figures S44–S57), along with a subset of minor products exhibiting C–O–C linkages. This result aligns well with the earlier biochemical studies (Figures 2 and 3), suggesting that the active site of Bmp7 accommodates two substrates for the reaction, while the enzyme exhibits an overlapping substrate scope in both homo- and hetero-coupling reactions. Consistent with the enzyme homocoupling reactions (Figure 4A), most heterocoupling reactions yielded a single dominant product (Figure 4B). Monohalogenated substrates 4, 5, and 8 yielded noticeable amounts of isomeric heterocoupled products, with the MS signal intensity of the 1+8 heterocoupled product significantly lower than others. Similarly, multiple heterocoupled isomers were detected in the 1+24 reaction, likely arising from distinct regioisomeric C–C coupling modes. The trace abundance of these minor isomers precluded further structural characterization. In addition to heterocoupled products, the homocoupled product of 1 was consistently detected across all heterocoupling reactions (Figure 4C). Notable amounts of homocoupled 10+10, 11+11, and 12+12 were also identified in LC-HRMS analysis, while the homocoupled products of 4, 5, 7, 8, and 13 were barely detectable (Figure 4C).

To gain structural insights into the heterocoupled products, we analyzed the product profile of the heterocoupling reaction between 1 and 17 as a representative case (Figure 5C). LC-HRMS analyses confirmed the formation of coupled products, identifying MC21-A ([M-H]⁻ m/z = 496.7029) as the major homocoupled species and 1+17 ([M-H]⁻ m/z = 436.7829) as the predominant heterocoupled product (Figure 5C-I). Notably, while 1 was nearly consumed, a significant portion of 17 remained unreacted. Consistent with this, only trace levels of homocoupled 17 ([M-H]⁻ m/z = 376.8630) were detected by LC-HRMS (Figures 4C and 5C-I). GC-MS analysis further supported C-C bond formation in the heterocoupled product (Figurew 5C-II and S58). The isolated heterocoupled compound was elucidated as 3,5,5’-tribromo-3’-fluoro-[1,1’-biphenyl]-2,2’-diol by ¹H, 2D COSY and ¹⁹F NMR (Figures 5C-III and S59), confirming ortho-ortho C–C coupling analogous to that observed in the homocoupled products of 1, 10, and 18 (Figures 2 and 5A–B). Collectively, these findings indicate that Bmp7 displays preferential catalytic efficiency toward substrates with stronger active site binding, providing mechanistic insight into substrate competition dynamics during heterocoupling reactions.

Structural characterization of Bmp7

The crystal structure of Bmp7 was determined at 2.28 Å resolution (PDB ID: 9YV5). It adopts the characteristic triangular prism fold of cytochrome P450s, but with several distinct features (Figure 6A). In comparison to canonical bacterial P450s, Bmp7 features an extended F/G helical region with the inclusion of an additional α-helix, F’. This region is known to influence substrate selectivity and reaction regioselectivity in P450s.^28–29 A second additional short α-helix (I’) is located between the H and I helices, which shifts the I helix, typically positioned across the heme center, away from the active site, thereby enlarging the distal pocket. Finally, part of the β3 sheet transitions into a protruding loop, which extends toward the heme center from the opposite side of the I helix (Figure 6A and S60). A structural homology search using the DALI server³⁰ with Bmp7 as the query revealed that these distinct features recognized in Bmp7 are shared with some human P450s, particularly CYP1A1, 1A2, and 1B1 (Table S4). Given that these human enzymes also accommodate large, planar aromatic ligands, the structural resemblance is consistent with the observed substrate scope of Bmp7.

Figure 6.

Structural characterization of Bmp7. A. The overall structure of Bmp7 (PDB ID: 9YV5, 2.28 Å) colored in a rainbow spectrum from blue (N terminus) to red (C terminus). B. Two molecules of 1 bound distal to the heme with 1A and 1B in green and yellow, respectively (PDB ID: 9MS3, 2.25 Å). The F_o-F_c omit map is contoured at +3σ and is depicted by grey mesh. C. The interactions of 1A and 1B with protein residues. Hydrogen and halogen bonding interactions and the distance (Å) between the two substrate molecules are shown as grey dashed lines.

To gain structural insights into the mechanism of Bmp7-catalyzed biaryl coupling, we determined crystal structures of Bmp7 in complex with 1 using two different approaches: substrate soaking (PDB ID: 9MS3, 2.25 Å) and co-crystallization (PDB ID: 9YVV, 2.32 Å). During refinement of both enzyme-substrate (ES) complex structures, two distinct areas of strong positive electron density resembling the heads of “Mickey Mouse” were observed in the active site (Figure 6B and S61A). These densities were modelled with two molecules of 1, with the “ears” guiding assignment to the electron-rich bromo substituents. The substrates were refined to occupancies of 0.83 (1A) and 0.73 (1B) in the substrate-soaked structure (Figure S61B) and 0.64 (1A) and 0.59 (1B) in the co-crystallized structure (Figure S61C). Superposition of the two ES complex structures yielded an root-mean-square deviation (RMSD) of 0.132 Å over 428 C_α atoms, signifying high structural similarity. No notable differences were observed in the positioning of substrates or active site residues between the two structures (Figure S62), despite the different methods of substrate incorporation. Thus, we focus on the substrate-soaked ES structure (9MS3) for further discussion due to its improved resolution and higher substrate occupancies.

Minimal changes were observed between the substrate-free and substrate-bound forms of Bmp7. Superposition of the two structures yielded an RMSD of 0.152 Å over 423 C_α atoms, indicating that the majority of the protein backbone are aligned. The main difference between the two states is a shift of the F helix towards a more closed conformation in the ES complex (Figure S63). This region caps the distal pocket, effectively enclosing the active site and isolating the substrate molecules from bulk solvent. Another difference was observed near the heme center in the I helix, which appears to be slightly more relaxed in the substrate-free structure. Overall, the movements of these regions slightly reshape the active site to stabilize substrate molecules.

The two substrate molecules adopt a near-perpendicular orientation relative to the heme plane, with 1A sandwiched between the I helix and 1B, creating π-π stacking interactions between the two substrates (Figure 6C). 1A is stabilized by the I helix and a halogen bond with the backbone carbonyl of Leu194 from the F helix. In contrast, 1B lacks halogen bonding interactions, but its aromatic ring is stabilized by Ala365 in the protruding loop (Figure 6C and S60). This unique loop in the Bmp7 structure plays a crucial role in maintaining the upright positioning of the substrate via hydrophobic interactions.

Another key distinction between the two substrate binding sites is the relative positioning of 1A and 1B. 1A sits directly above the heme center, whereas 1B is positioned laterally within the active site. Notably, the phenol group of 1A forms a 3.4 Å hydrogen bond with a distal water molecule weakly coordinated to the heme, while 1B is too distant for direct interaction (Figure 6B). A series of ordered water molecules pass under 1B; however, they do not directly reach within hydrogen bonding distance of the 1B phenol. This may implicate substrate 1A as the initial site of hydrogen atom abstraction and radical generation by proximity. Additionally, the ordered waters and both substrate phenol groups participate in hydrogen bonding interactions with Glu105 and Gln292 (Figure 6C), ensuring proper substrate orientation with the phenol groups directed toward the heme center. In contrast, the opposite end of the active site contains hydrophobic residues Val300 and Phe477, which excludes polar −OH groups. The structural insights support the essential role of hydroxyl group for tight binding of halogenated phenol substrates (Figure 3).

The presence of two distinct substrate binding sites within the distal pocket is consistent with the ITC analysis (Figure 2A). Given that 1A forms more specific contacts with active-site residues, it likely corresponds to the tighter binding site identified from the ITC experiment. Furthermore, the sequential binding model suggests that 1A binds first, priming the active site for subsequent 1B binding.

The binding of dual substrate molecules has also been observed or proposed in other dimerizing CYPs, though with distinct binding orientations. A particularly relevant example is CYP158A2, which catalyzes the polymerization of flaviolin.³¹ In the active site of CYP158A2, two substrate molecules also engage in π–π stacking interactions, with one positioned close to the heme and the other more distant. However, both substrates adopt a nearly parallel orientation relative to the heme plane.³¹ One hydroxyl group of the proximal substrate is located 5.0 Å from the Fe center and plays a critical role in supporting oxygen binding and catalytic turnover.³² Additionally, water chains within the active site bridge the substrate hydroxyl groups and the heme-bound ligands. A comparable stacked substrate conformation, oriented parallel to the heme, has also been predicted by molecular docking for CYP90J6 in biflavonoid biosynthesis.³³ Collectively, structural investigations of dimerizing P450s suggest that their ability to accommodate two stacked substrates and position substrate phenols in close proximity to the heme is central to oxidative coupling, whereas the precise orientations modulated by surrounding residues and water networks likely dictate the catalytic regioselectivity.

Computational analysis of Bmp7 catalyzed heterocoupling

The biochemical and structural characterization of Bmp7 supports a plausible reaction mechanism for diaryl C-C bond formation (Figure 7A), informed by mechanistic paradigms established for other P450 catalyzed C–C coupling reactions.^{6, 33–38} We propose that Bmp7-mediated coupling proceeds through either a diradical coupling pathway (I) or a single-radical addition route (II). The reaction starts with a hydrogen atom transfer (HAT) from the phenolic hydroxyl group of substrate 1A (Figure 6B) to the high-valent iron-oxo species, Compound I (Cpd I),³⁹ generating a delocalized phenoxy radical on 1A. This HAT may occur via an active-site water molecule, analogous to the water-assisted C-O phenol coupling proposed for P450 Cih33/DmlH.³⁸ In the 1 bound Bmp7 complex structure (Figure 6B), a water molecule located approximately 3.5 Å above the heme lies in an ideal position to mediate electron or proton transfer, potentially functioning as a transient relay during radical initiation. However, since the water network is expected to rearrange upon Cpd I formation, its precise role in mediating electron transfer remains uncertain. Considering the expected Fe=O distances in ferryl species,⁴⁰ direct HAT from the substrate to Cpd I or II could also occur. In pathway I, radical transfer generates a 1B-centered radical, followed by a second HAT from 1A-OH mediated by Cpd II. The resulting 1A and 1B radicals subsequently undergo intermolecular radical-radical coupling to yield the product. Structural characterization of both homo- and hetero-coupled products indicates that the ortho-ortho C-C linkage is the dominant coupling mode (Figure 5). In pathway II, the 1A radical instead attacks 1B, likely its ortho-position, forming an aryl-aryl σ-bonded intermediate bearing the radical on the 1B moiety. Product formation in this scenario would require a second HAT mediated by Cpd II. Given the relatively long distance between the 1B-OH and the heme iron (Figure 6B), direct HAT from 1B appears geometrically unfavorable. We therefore propose an intramolecular radical relay from the 1B to 1A moiety within the coupled intermediate, enabling subsequent reduction by Cpd II and formation of the final C-C coupled product (Figure 7A).

Figure 7.

Mechanistic model of Bmp7-catalyzed diaryl formation via C–C bond coupling. A. Illustration of the proposed di- (I) and mono-radical (II) mechanisms underlying Bmp7-catalyzed coupling. B-C. Potential energy surfaces calculated by DFT for homo- and hetero-coupling reactions of 1 and 17 (B) and 1 and 7 (C). The solid lines show the single radical mechanism, while the dashed lines represent the diradical mechanism. The unit of free energy is electron volts. The proportions of different products in the coupling reactions were calculated according to the Boltzmann distribution.

To elucidate the factors underlying product profile variations in the Bmp7-mediated heterocoupling reactions, we performed density functional theory (DFT) calculations based on the reaction between 1 and 17 as a model system to simulate the reaction processes involved. For the formation of C–C coupled products, we evaluated two potential mechanisms, a single-radical pathway and a double-radical pathway (Figure 7A), represented by solid and dash lines, respectively (Figure 7B). Thermodynamic analysis of the heterocoupling of 1 and 17 revealed that the radical of 1 is slightly more stable than that of 17 (ΔG = 3.47 eV vs 3.49 eV), favoring the formation of 1-containing dimers such as the homodimer 1+1 (MC21-A, ΔG = 4.67 eV) and heterodimer 1+17 (ΔG = 4.67 eV). The observed similar intensities of MC21-A (1+1) and 3,5,5’-tribromo-3’-fluoro-[1,1’-biphenyl]-2,2’-diol (1+17) (Figure 5C) suggest that the reaction predominantly follows a single-radical mechanism, most likely proceeding via the radical of 1 (solid orange line 1·+1 or solid blue line 1·+17). In contrast, thermodynamic calculations indicate that the formation of 17+17 is less favorable due to the higher free energy barrier (ΔG = 4.70 eV, solid purple line), consistent with its low abundance in the product mixture (Figures 4C and 5C). On the other hand, the significantly higher energy states disfavor the diradical mechanism (Figure 7B), which is also incompatible with the observed product distribution (Figure 4C). Indeed, if the diradical pathway were operative, the homodimer of 1, which is calculated to be thermodynamically more stable, would be expected as the predominant product, in clear contrast to the experimental results. Similarly, for the reaction between 1 and 7 (Figure 7C), DFT calculation suggested that radical formation from 7 radical is significantly less favorable (ΔG = 3.54 eV), which accounts for the low yield of its homodimer (ΔG = 4.69 eV). Instead, the heterodimer product 1+7 is thermodynamically preferred (ΔG = 4.63 eV), aligning with experimental findings (Figure 4C). Furthermore, our calculation revealed that the formation of 1+1 through C–O–C linkage is highly unfavorable, with a free energy change of 5.36 eV, further supporting the predominant C–C coupling pathway (Table S3). Overall, DFT calculations conducted in the absence of the enzyme environment support that Bmp7-mediated halogenated biaryl coupling proceeds via a single-radical mechanism, with radical stability and product thermodynamics serving as key determinants of reaction outcomes. Further mechanistic insights will necessitate molecular dynamics simulations to account for substrate conformational flexibility and active site motions.

Conclusions

Halogenated biaryls are critical scaffolds of compounds with broad applications and developing biocatalysts capable of selectively synthesizing these motifs is essential for advancing sustainable and environmentally compatible synthetic approaches. In this work, we demonstrated that the marine proteobacterial cytochrome P450 enzyme Bmp7 exhibits considerable substrate flexibility, efficiently catalyzing homo- and hetero-coupling reactions of 17 mono- and di-halogenated phenols, primarily through ortho-ortho C–C bond formation. X-ray crystallographic analysis uncovered the binding of two substrate 1 molecules within the Bmp7 active site, providing one of the rare structural snapshots supporting intermolecular C–C coupling mechanisms in P450 enzymes. We proposed that Bmp7 can follow either di- or mono-radical reaction mechanism for its coupling reaction (Figure 7A). DFT analysis supported the mono-radical mechanism and revealed that the stability of radical intermediates and product thermodynamics govern the observed product distribution, offering mechanistic insight into coupling selectivity. These findings not only deepen our understanding of halogenated phenol coupling catalyzed by P450s but also establish a framework for rational engineering of Bmp7 and related enzymes to improve catalytic efficiency, regioselectivity, and substrate scope. Our work also lays a foundation for the development of next-generation biocatalysts for the sustainable synthesis of halogenated biaryls, with broad implications for green chemistry, pharmaceutical manufacturing, and synthetic methodology design.

Materials and methods.

Chemicals, strains, and general experimental procedures

Molecular biology reagents and chemicals were purchased from Fisher Scientific, Ambeed, Sigma-Aldrich, or New England Biolabs, Inc. if not specifically indicated. Genes were ordered from Twist Bioscience or provided by Dr. Vinayak Agarwal (Georgia Tech). DNA sequencing was performed at Eurofins Genomics or GENEWIZ. E. coli strains DH5α and BL21-GOLD (DE3) were used for routine molecular biology studies and protein expression, respectively. The strains were cultured in Luria-Bertani (LB) broth or Terrific broth (TB). GeneJET Plasmid Miniprep Kit and GeneJETGel Extraction Kit (Thermo Scientific) were used for plasmid preparation and DNA purification, respectively. Data analysis was performed using Microsoft Excel or GraphPad Prism. NMR samples were prepared in CDCl₃ or CD₃OD and were recorded. Spectroscopy data were collected using MestreNova software. Chemical shifts δ are given in ppm using tetramethylsilane as an internal standard. Multiplicities of NMR signals are designated as singlet (s), doublet (d), doublet of doublets (dd), triplet (t), quartet (q), and multiplet (m).

Protein expression and purification

Bmp9 and Bmp10 genes from Marinomonas mediterranea were synthesized by Twist Bioscience and Bmp7 gene was provided by Dr. Vinayak Agarwal (Table S5). These genes were subsequently cloned into the pET28b vector digested with NdeI and XhoI. The resulting constructs were sequenced to exclude any error and then transformed into E. coli BL21-GOLD (DE3) cells for protein expression. The overnight culture of transformed cells was diluted 1:1000 in TB medium containing appropriate antibiotics and incubated at 37 °C, 250 rpm, until the optical density at 600 nm (OD600) reached 0.5. The flasks were then cooled at 4 °C for 30 min before the induction of protein expression with 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG). For Bmp9 expression, cultures were supplemented with 10 mg of riboflavin; for Bmp10, 50 mg of FeSO₄ was added; and for Bmp7, 0.15 mM 5-aminolevulinic acid and 1 X trace metal solution were added. The trace metal solution consisted of the following 1000 X stocks: 50 mM FeCl₃, 20 mM CaCl₂, 10 mM MnSO₄, 10 mM ZnSO₄, 2 mM CoSO₄, 2 mM CuCl₂, 2 mM NiCl₂, 2 mM Na₂MoO₄, and 2 mM H₃BO₃. The cultures were then incubated at 16 °C, 200 rpm, for 16–48 hours. The cell pellets were collected by centrifugation at 5,000 × g for 10 min at 4 °C and stored at −80 °C until use. Thawed cell pellets were resuspended in the lysis buffer (20 mM Tris-HCl, pH 8.0, 500 mM NaCl,10% glycerol). Soluble proteins were released by sonication (5 s pulse and 10 s rest, 1 min in total), and the supernatants were clarified by centrifugation at 35,000 × g for 30 min at 4°C. Recombinant proteins were purified using HisPur Ni-NTA resin (Thermo Scientific) following the manufacturer’s protocol, with the elution buffer (20 mM Tris-HCl, pH 8.0, 1 M NaCl, 250 mM imidazole). Protein-containing fractions were combined and exchanged into storage buffer (100 mM phosphate buffer, pH 7.0, 10 % glycerol) using a PD-10 desalting column (GE Healthcare Life Sciences) according to the manufacturer’s instructions. Desalted proteins were concentrated to a final volume of < 1 mL using Amicon Ultra-15 Centrifugal Filter Units (3–10 kDa). Protein concentrations were determined by Bradford assay or Nanodrop spectrophotometry, and protein aliquots were stored at −80 °C until use.

UV-Vis spectroscopy and substrate binding assays

The absorbance spectrum (350–600 nm) of Bmp7 in 100 mM phosphate buffer, pH 7.0 were recorded by a Shimadzu UV2700 dual beam UV-vis spectrophotometer. To assess carbon monoxide binding, the enzyme solution was first saturated by bubbling carbon monoxide gas (Airgas) through the sample, and spectrum of the CO-saturated enzyme was recorded. Subsequently, 30 μL of 0.5 M sodium dithionite was added to reduce the ferric iron, and the spectrum of the reduced enzyme was collected. CO-reduced difference spectrum was generated by subtracting the spectrum of the reduced enzyme from that of the CO-bound reduced enzyme.

Data analysis was performed using Microsoft Excel. Substrate binding assays were conducted by incubating 300 μL of 4.5 nM Bmp7 in phosphate buffer (100 mM, pH 7.0) with increasing concentrations of ligands (two independent replicates). Binding was monitored by measuring the change in absorbance (ΔA), calculated as the difference between absorbance at ~390 nm and ~420 nm. The resulting binding curves were fitted to the Hill equation ΔA = ΔA_max[L]ⁿ/(K_Dⁿ + [L]ⁿ) using GraphPad Prism 9 to determine the dissociation constant (K_D) values and Hill coefficient (n).

Biocatalytic reaction

The Bmp7 reactions (100 μL) included substrate (250 μM), recombinant Bmp9 (1 μM), recombinant Bmp10 (1 μM) and 2 mM NADH in 100 mM potassium phosphate buffer, pH 6.8. The reactions were initiated by adding recombinant Bmp7 (0.9 μM) and then incubated at 30 °C for 10 h. The control reactions contained heat-inactivated Bmp7. All reactions were extracted twice with ethyl acetate. The organic layer was collected, combined and dried in vacuo. The residues were dissolved in 100 μl 20% acetonitrile. After centrifugation at 20,000 × g for 20 min, the clear supernatants were collected for HPLC and LC-HRMS analysis analyzed. All experiments were repeated independently at least three times.

HPLC, LC-HRMS, and HRMS/MS analysis

Samples were analyzed on a Shimadzu Prominence UHPLC system (Kyoto, Japan) coupled with a PDA detector. The compounds were separated on an Agilent Poroshell 120 EC-C18 column (4.6 × 50 mm, 2.7 μm) using the following HPLC program: 5% B for 1 min, 5–95% B gradient in 10 min, 95% B for 1 min, 95–5% in 1 min, and re-equilibration in 2% B for 1 min. Solvents A and B were water and acetonitrile, respectively. The flow rate was set at 0.6 mL/min. LC-HRMS and HRMS/MS experiments were conducted on Thermo Scientific^™ Q Exactive Focus mass spectrometer with Dionex^™ Ultimate^™ RSLC 3000 uHPLC system, equipped with H-ESI II probe on Ion Max API Source. Water (A) and acetonitrile (B) were used as mobile phases to separate analytes on an Agilent Poroshell 120 EC-C18 column (4.6 × 50 mm, 2.7 μm). A typical LC program with a 0.6 mL/min flow rate included 5% B for 2 min, 5–95% B in 13 min, 95% B for 2 min, 95 to 5% B in 0.5 min, and re-equilibration in 5% B for 2 min. MS1 signals were acquired under the Full MS negative ion mode covering a mass range of m/z 150–750, with a resolution at 35,000 and an AGC target at 1e⁶. Fragmentation was obtained using MS2 confirmation and Parallel Reaction Monitoring (PRM) mode using an inclusion list of calculated parental ions. Precursor ions were selected typically with an isolation width of 3.0 m/z and fragmented in the HCD cell with stepwise collision energies (CE) of 25 and 30.

HPLC purification and characterization

To obtain sufficient quantities of the heterocoupled product of substrates 1 and 17 for NMR analysis, a large-scale reaction was carried out using 500 microcentrifuge tubes (100 μL per tube). The reaction conditions were identical to those described above, except that the substrate ratio was adjusted to 1:2 (substrate 1:substrate 17). Similarly, large-scale (20 mL) reactions of 10 + 10 and 18 + 18 were conducted in 50 mL tubes under the same reaction conditions as described above. After 24 h, all reaction mixtures were combined and extracted with 50 mL ethyl acetate three times. Organic layers were collected, combined, and dried. The residue was dissolved in 20% acetonitrile/water and purified by semi-preparative HPLC with a C18 Agilent column (22 × 250 mm) using solvent A (water) and B (acetonitrile). The flow rate was set at 1 mL/min. The HPLC purification used the following method: 20% B for 3 min, 20–55% B over 7 min, 55–65% B over 12 min, 65–95% B over 2 min, 95% B hold for 2 min, then 95–20% B for 1 min. The peak fractions were collected, dried and the purified compound (0.2 mg for 1+17, 1.2 mg for 10+10, 0.5 mg for 18+18) was stored at −30 °C until NMR analysis.

Heterocoupled product of 1 and 17: white solid. HRMS (ESI) m/z: [M-H]⁻ calcd for C₁₂H₅Br₃FO₂⁻ 428.7808, found 438.7785. ¹H NMR (600 MHz, CD₃OD): δ 7.56 (d, J = 2.5 Hz, 1H), 7.35 (d, J = 2.5 Hz, 1H), 7.19 (m), 7.14 (dd, J = 10.0, 2.5 Hz, 1H). ¹⁹F NMR (565 MHz, CD₃OD) δ −135.38.

Self-coupled product 10: white solid. HRMS (ESI) m/z: [M-H]⁻ calcd for C₁₂H₅Br₂Cl₂O₂⁻ 410.8018, found 410.8012. ¹H NMR (600 MHz, Acetone) δ 7.4629 – 7.4285 (d, J = 2.634 Hz, 1H), 7.2500 – 7.2156 (d, J = 2.634 Hz, 1H). ¹³C NMR (151 MHz, Acetone) δ 156.0930, 130.5633, 129.2176, 129.1147, 120.2294, 113.7005.

Self-coupled product 18: white solid. HRMS (ESI) m/z: [M-H]⁻ calcd for C₁₂H₅Br₂F₂O₂⁻ 378.8609, found 378.8614. ¹H NMR (600 MHz, Acetone) δ 7.2639 – 7.2292 (dd, J = 7.922, 3.130 Hz, 2H), 7.1644 – 7.1312 (dd, J = 9.695, 3.128 Hz, 2H).

Gas chromatography-mass spectrometry (GC-MS) analysis

The samples and standards used in GC-MS analysis were prepared using the N-methyl trifluoroacetamide (MSTFA) derivatization method. For derivatization reactions, 40 μL of 2% methoxyamine hydrochloride (dissolved in pyridine) was added to samples, followed by shaking at 37 °C for 2 h. After a brief spin, 70 μL of MSTFA with 1% trimethylchlorosilane (TMCS) was added, followed by shaking at 37 °C for 30 min. After centrifugation at 17,000 g for 5 min, 80 μL of the clear supernatant was transferred into a GC vial. The derivatized samples were run on a Thermo GC-MS equipment coupled with the Zebron ZB-5MSplus column. The temperature program was set to 85 °C with an initial hold for 2 min, followed by 10 °C/min of increase till 110 °C, 5 °C/min of increase until 230 °C, and 15 °C/min of increase until 320 °C, with a final hold for 2 min. The MS scan was set for the range of 70–600 amu.

Isothermal titration calorimetry (ITC) to probe binding

The Soret band extinction coefficient of Bmp7 was determined to be 117.4 mM⁻¹ cm⁻¹ at 417 nm by a pyridine hemochromogen assay.⁴¹ This constant was used to determine the concentration of Bmp7. ITC experiments were performed at 20.1 °C using a MicroCal PEAQ-ITC instrument (Malvern Panalytical). The sample cell contained Bmp7 (50.4 μM) in 50 mM Tris-HCl buffer, pH 7.0, and the syringe was loaded with 750 μM of 2,4-dibromophenol (1) in the same buffer. The cell was stirred continuously at 750 rpm. Following a 60-second equilibration delay, one initial 0.4 μL injection was made, followed by eighteen 2.0 μL injections, each spaced 150 seconds apart. The first two injections were excluded from analysis. Data were analyzed using the MicroCal PEAQ-ITC Analysis Software and fitted to a sequential binding model with the number of sites (N) fixed at 2.0, consistent with structural and spectroscopic evidence supporting two substrate-binding events.

Crystallization of Bmp7-1 complex

Bmp7 for crystallization was concentrated and loaded onto a HiLoad 16/ 600 Superdex 75 pg column (Cytiva) equilibrated with 50 mM Tris/HCl, pH 7.0. After elution, Bmp7 was concentrated to 27 mg/mL based on its heme concentration. Purified enzyme (2 μL) was incubated with 2 μL of mother liquor (20% PEG 3350 and 0.2 M of sodium citrate tribasic dihydrate) in a hanging-drop tray and was stored at 5 °C. Small, rod-shaped red crystals appeared after 3 days and grew to their full size over approximately 2 weeks. To generate the co-crystallized ES complex, 2 eq of 1 was pre-mixed with purified enzyme before the drops were set. To generate the soaked ES complex, Bmp7 only crystals were soaked in mother liquor containing 3 mM of 1 for 3 hours prior to cryoprotection in 20% glycerol and flash freezing in liquid nitrogen. Data were collected at 22-ID-D beamline at SER-CAT, APS, Argonne National Laboratory. The datasets were collected at 100 K at a wavelength of 1.0000 Å and processed using HKL2000.⁴² An AlphaFold prediction of the Bmp7 structure was used as the search model for molecular replacement.⁴³ The structure model was built and refined using PHENIX software packages.⁴⁴

DFT calculations

All structures were optimized using Gaussian16⁴⁵ at the M06–2X-D3/6–31+G(d) level,^46–51 followed by vibrational frequency analysis to confirm the absence of imaginary frequencies. Single-point energy calculations were conducted at the M06–2X-D3/6–311++G(d,p) level, incorporating the implicit Solvation Model based on Density^52–56 with water as the solvent to approximate realistic solution conditions. The M06–2X functional with D3 dispersion correction has been demonstrated to provide robust and accurate predictions in radical-mediated reactions.⁵⁷

Supplementary Material

Supporting Information includes additional results, Tables S1–S5 and Figures S1–S63. Accession Codes: PDB IDs: 9YV5, 9YVV and 9MS3.