Molecular Basis of Iterative C─H Oxidation by TamI, a Multifunctional P450 monooxygenase from the Tirandamycin Biosynthetic Pathway

Abstract

Biocatalysis offers an expanding and powerful strategy to construct and diversify complex molecules by C─H bond functionalization. Due to their high selectivity, enzymes have become an essential tool for C─H bond functionalization and offer complementary reactivity to small-molecule catalysts. Hemoproteins, particularly cytochromes P450, have proven effective for selective oxidation of unactivated C─H bonds. Previously, we reported the in vitro characterization of an oxidative tailoring cascade in which TamI, a multifunctional P450 functions co-dependently with the TamL flavoprotein to catalyze regio- and stereoselective hydroxylations and epoxidation to yield tirandamycin A and tirandamycin B. TamI follows a defined order including 1) C10 hydroxylation, 2) C11/C12 epoxidation, and 3) C18 hydroxylation. Here we present a structural, biochemical, and computational investigation of TamI to understand the molecular basis of its substrate binding, diverse reactivity, and specific reaction sequence. The crystal structure of TamI in complex with tirandamycin C together with molecular dynamics simulations and targeted mutagenesis suggest that hydrophobic interactions with the polyene chain of its natural substrate are critical for molecular recognition. QM calculations and molecular dynamics simulations of TamI with variant substrates provided detailed information on the molecular basis of sequential reactivity, and pattern of regio- and stereo-selectivity in catalyzing the three-step oxidative cascade.

Graphical Abstract

INTRODUCTION

Selective functionalization of C─H bonds to form C─X or C─C bonds is a powerful strategy for the synthesis of complex organic compounds and other high-value chemicals. Although the past decade has seen expansive applications of C─H functionalization for chemical synthesis, controlling regio- and stereoselectivity remains a significant challenge. Most research has focused on catalyst-controlled methods for selectivity in C─H functionalization.¹ While several effective small molecule catalysts have been developed in the past few years, Nature has utilized selective C─H functionalization throughout evolution via enzymemediated catalysis. In various metabolic pathways, enzymes perform diverse C─H functionalization reactions such as halogenation, hydroxylation, alkylation, and amination with high regio- and stereo-selectivity; therefore, biosynthetic enzymes are an excellent resource to access biocatalysts that can be harnessed for efficient C─H functionalization with high selectivity over a broad range of substrates.

Most biological site-selective C─H functionalizations are catalyzed by cytochromes P450 (P450s) with high chemo-, regio-, and stereoselectivities. P450s are found in all domains of life including viruses,² and catalyze a broad range of oxidative transformations such as aliphatic and aromatic bond hydroxylation, double bond epoxidation, C─C bond formation, decarboxylation, desaturation, oxidative rearrangement of carbon skeletons and other transformations.³ These P450s are frequently integrated into biosynthetic gene clusters and perform oxidative tailoring of complex substrates to produce secondary metabolites.⁴ The remarkable versatility of P450s has driven increased applications for synthetic biology⁵ and for late-stage functionalization of several natural products including Nigelladine A, Juvenamicins, Vancomycin, Mycinamicin and Thaxtomin D.⁶ Taking advantage of inherent evolvability of these P450s, researchers have engineered P450s to catalyze new and non-natural reactions including C-Si, C-B, anti-Markovnikov alkene oxidation for which no enzymes exist in nature.^7-11 The continued discovery, characterization, and engineering of novel biosynthetic P450s provides provides new insights for biocatalyst development and expanding the toolbox for addressing the challenges of C-H functionalization.

TamI, is a multifunctional cytochrome P450 from Streptomyces sp. 307-9 that catalyzes late-stage oxidation at three positions on the bicyclic ketal moiety of tirandamycin C (1),¹² which is a required modification for bacterial RNA polymerase targeting by the tirandamycins.¹³ In distinction from flexible enzymes that act on a variety of substrates, multifunctional P450s catalyze a stepwise sequence of reactions on a particular substrate. This uncommon reactivity has been studied in bacterial enzymes MycG¹⁴ and AurH¹⁵, as well as other fungal¹⁶ and plant¹⁷ P450s. Previously, we reported in vitro functional characterization of TamI, and a partner flavin adenine dinucleotide (FAD)-dependent oxidase, TamL, and established their roles in the installation of the C10 keto, C11/C12 epoxide and C18 hydroxyl groups on native substrate 1 (Figure 1).¹² TamI catalyzes initial hydroxylation of 1 at C10, to form tirandamycin E (2); subsequently, TamL catalyzes oxidative conversion of C10 hydroxyl to carbonyl to yield tirandamycin D (3). Further epoxidation and hydroxylation catalyzed by TamI complete the oxidative catalytic cascade to yield antibiotics tirandamycin A (4) and tirandamycin B (5), respectively (Figure 1). Despite detailed functional studies, the molecular basis of the iterative nature of TamI remained unexplored, which we were motivated to pursue using structural biology and computational approaches. A key hypothesis that we sought to interrogate was whether TamI uses a discrete substrate binding mechanism in which the tetramic acid moiety anchors 1 in a position appropriate for the iterative oxidation reactions to occur. Substrate anchoring has notably been observed in the TamI homologue PikC (47% sequence identity), where presence of the dimethylamino sugar desosamine or a suitable surrogate drives substrate recognition, thus enabling oxidation of non-native substrate scaffolds bearing this recognition element.¹⁸ Presence of such a feature in TamI could provide additional opportunity for a tetramate-based anchoring strategy for oxidative tailoring of diverse chemical scaffolds.

Figure 1. TamI oxidations.

TamI catalyzes iterative oxidation of bicyclic ketal of 1. The C10 hydroxyl of 2 is oxidized to a ketone by the TamL flavoprotein.

Here we report the molecular characterization of TamI to understand in detail its structure, substrate molecular recognition, and the basis for its regio- and stereo-selectivities. The TamI structure together with molecular dynamics simulations and subsequent mutagenesis experiments indicate that hydrophobic interactions with the central polyene chain region of 1 is the critical element for substrate recognition and catalysis. QM calculations, together with molecular dynamics simulations, provide detailed information on the molecular basis of regio- and stereo-selectivity.

RESULTS AND DISCUSSION

Structural analysis of TamI in complex with tirandamycin C

The structure of TamI in complex with 1 was solved by molecular replacement at 2.7 Å resolution using the ligand free structure as a search model (Figure S1-S2). The overall structure exhibits the characteristic P450 fold with a single heme cofactor covalently bound via a Cys359 heme-thiolate (Figure 2A). The asymmetric unit is composed of eight unique chains which are differentiated by their crystal packing environment and show high conformational similarity (r.m.s. deviations < 0.3 Å) (Figure S3); therefore the best-ordered chain (chain A, average Bfactor 50 Ų) was used for analysis and interpretation. Structural homology analysis revealed CYP267B1 (6GK5), CYP154C5 (4J6C), CYP245A1 (2Z3U), and CreJ (5GWE) as the closest structural homologs with an average r.m.s. deviation of 2.1 Å (DALI).¹⁹ These enzymes are involved in oxidation of medium chain fatty acids, steroids, alkaloids, alkylphenols.^20-23

Figure 2. Structure of TamI in complex with 1.

(A) Cartoon representation of the complex. 1 is shown in yellow space-filling model; heme is shown in black. (B) Omit map (Fo-Fc) for 1. The bicyclic ketal is proximal to the heme.

Tirandamycin C is unambiguously bound in the active site with well-defined electron density for the entire molecule (Figure 2B). The bound substrate is enveloped completely within the enzyme without evidence of direct access to bulk solvent. Tirandamycin C makes close contacts with the enzyme throughout its structure (Figure 2B). The tetramic acid moiety is bound in a pocket composed of segments of four distinct loop regions of the protein: the β1-1/2 loop (residues Val42, Pro43, Val44), the BC loop (Phe92), the FG loop (Val185), and the β3-3/2 loop (Ser 397, Thr398, Leu399) (Figure 3A). Tetramic acid binding is mediated by several direct polar interactions via the carbonyl oxygen of Pro43, the amide nitrogen of Thr398 and Leu399, and the side chain hydroxyls of Ser397 and Thr398, and by hydrophobic contacts with the side chains of Val44 and Phe92 (Figure 3A).

Figure 3. Tetramic acid binding site.

(A) Stick representation of key contacts proximal to tetramic acid. (B) HPLC traces for endpoint assays with TamI mutants. Signal is UV absorbance at 340 nm. 2 shows peak shoulder due to tautomerization in methanol.

In addition to the tetramic acid component of the substrate there are several interactions along the length of the polyene chain and the bicyclic ketal. The linear polyene predominantly contacts residues from the β3-3/2 loop region (Leu399 and Ile400),FG region (Val185) and BC loop (Phe92). These hydrophobic side chains constrict the binding site in this region and lead to a hydrophobic pocket adjacent to the heme center in which the bicyclic ketal is bound. Numerous hydrophobic side chains line the binding site of the bicyclic ketal (Leu101, Leu244, Leu295) in addition to other non-polar residues His102, Gly248, Thr252 and Thr299 (Figure 4A). Consistent with the observed order of oxidation by TamI in which C10 is first hydroxylated to form 2, the C10 atom distance to the heme-iron is 4.1 Å, which is suitable for hydroxylation upon redox partner binding and iron-oxo formation. The conserved I-helix residues Glu250 and Thr251, which have been implicated in proton delivery during dioxygen activation²⁴ are also found in this region adjacent to the bicyclic ketal moiety.

Figure 4. Bicyclic ketal binding site.

(A) Stick representation of key contacts proximal to the tirandamycin bicyclic ketal. (B) HPLC traces for endpoint assays with TamI mutants. Signal is UV absorbance at 340 nm. 2 shows peak shoulder due to tautomerization in methanol.

Mutagenesis and biochemical characterization to probe the molecular basis of substrate binding and anchoring mechanisms

Based on the PikC crystal structure in which a salt-bridge interaction distal from the site of oxidation provided substrate anchoring,^{18, 25-26} we sought to investigate whether the tetramic acid moiety of 1 could have an analogous role in TamI. Numerous polar interactions about the tetramic acid moiety in the TamI active site appear consistent with a mechanism that permits iterative oxidation of the distal bicyclic ketal, which we probed experimentally and computationally (see below). Mutagenesis of TamI active site residues in proximity to 1 was performed to assess their specific roles in catalysis. TamI mutant assays were conducted using an optimized three component system composed of TamI, spinach ferredoxin and ferredoxin reductase. To probe whether the dienoyl tetramic acid moiety of the native substrate 1 plays a role in substrate anchoring, we performed mutagenesis of active site residues Ser397, Thr398, Pro43, Val44, and Val185 (Figure 3). Based on the complex structure, we hypothesized that the polar interactions mediated by the side chain hydroxyl groups of Ser397 and Thr398 could be required for proper substrate binding and perhaps could act as a substrate anchor similar to the desosamine sugar of narbomycin/YC-17 and Asp50/Glu85 in PikC.^{18, 25-26} Therefore, we prepared single (S397A, T398A, V185A) and double (P43A/V44A) amino acid variants. Analysis of these mutants, however, revealed no evident changes in product profile or abundance compared to wild type TamI in an end-point assay. Furthermore, these variants showed minor perturbations in K_d measured by UV spectral assay where addition of substrate was used to displace the axial water ligand (Table S2, Figure S4). This demonstrated that these polar side chain interactions near the dienoyl tetramic acid moiety have no direct influence on the TamI-mediated TirC to TirE conversion and refuted a tetramic acid based anchoring mechanism. Considering the array of backbone mediated interactions with the tetramic acid moiety, and also the numerous close contacts along the full substrate molecule, we expanded our analysis to alternative active site interactions with bound 1.

Thus, we examined a cluster of hydrophobic and aromatic residues (Leu399, Ile400, Phe92) located near the polyene region of the substrate. Substitution of Phe92 with alanine resulted in complete loss of catalytic activity in an end-point assay. The double mutant L399A/I400A demonstrated significantly reduced activity (Figure 3B). In these cases, the spectrophotometric assay showed no shift for these mutants (Figure S5) likely indicating perturbed substrate binding. These data indicated that the aromatic and hydrophobic residues surrounding the substrate are critical for C-10 hydroxylation.

Finally, we probed the active site residues (Leu101, His102, Leu295, Thr299) proximal to the bicyclic ketal moiety for their role in substrate binding and product profile. We created a double mutant (L101V/H102S) to examine the impact of increasing the active site volume. However, we did not observe any change in product profile that would indicate relaxed site selectivity for the first oxidation event; instead we observed a significant decrease in TirC to TirE conversion. Presumably, the Leu101 and His102 side chains are important for substrate binding through interactions with the C14 methyl group, which is evident in the crystal structure. Specifically, the Cδ atom of Leu101 is at ~3.4 Å, Nε of His102 is at ~4 Å in relation to the oxygen atoms of bicyclic ketal moiety in 1. These interactions may be required for proper positioning of the bicyclic moiety with respect to heme-iron for C10 hydroxylation. We also generated alanine substitutions at Leu295 and Thr299, which are located on the opposite side of the active site and in proximity of the bicyclic ketal moiety. However, this TamI double mutant displayed wild type activity in its ability to convert substrate to 2 (Figure 4).

In summary, mutagenesis studies revealed that the hydrophobic residues in proximity to substrate 1 (Leu399, Ile400, Phe92) are essential for C-10 hydroxylation. These data further suggested that the polyene chain of tirandamycin substrates are a critical component and that appropriate substrate orientation is achieved through hydrophobic and steric interactions with non-polar residues (Leu399, Ile400) and the aromatic ring of Phe92.

Computational investigation of regio- and stereo-selectivity for sequential reactivity and pattern of oxidation

A notable feature of the TamI enzyme is its ability to selectively oxidize multiple positions of tirandamycins in a specific and sequential manner. We used computational methods of density functional theory (DFT) and molecular dynamics (MD) simulations to characterize the intrinsic reactivity of these competing oxidations on the tirandamycin substrates and classify how the enzyme influences the regio- and stereo-selectivity.

Using a truncated model of the tirandamycins C, D, and A containing the biketal moiety and the reactive Fe(IV)-oxo radical cation, we used DFT methods to calculate the transition state barriers for all the competing oxidations, C10 hydroxylation, C11/12 epoxidation, and C18 hydroxylation (Figure 5A). For the first substrate in the oxidative cascade, 1, the lowest energy barrier is at the C10 position for allylic hydroxylation with no difference between abstracting the R or S hydrogen, as they both have C─H abstraction barriers of 14.5 kcal/mol (Figure 5A, Figure S6). This matches the experimentally observed regioselectivity, but it indicates that the active site environment of TamI is essential for distinguishing the selectivity between these diastereomeric transition states. For 3, the second substrate in the TamI oxidation cascade, the front facing epoxidation at C11/12 is lowest in energy by 1.9 kcal/mol, which matches the experimentally observed stereoselectivity and reaction specificity (Figure 5A, Figure S7). Lastly, we calculated the transition state barriers for the final substrate, 4, which has a barrier of 17.5 kcal/mol. These calclulated energetic barriers through the oxidation cascade match the trend reported for the TamI-RhFRED fusion protein, where k_cat drops significantly (0.11 min⁻¹) for the final oxidation of 4 compared to the first oxidation of 1 (40.5 min⁻¹) or second oxidation of 3 (83.8 min⁻¹).¹² While these energetic trends match the order of sequential reactivity without accounting for the affect of enzyme environment on the activation barriers, we analyzed more of the stereoselectivity and how TamI biases towards the experimentally observed reactions, particularly for the C10 allylic hydroxylation step.

Figure 5. Computational analysis of TamI iterative oxidation.

(A) Intrinsic energy for each of the three oxidation steps in tirandamycin biosynthesis. (B) MD simulations showing TamI interaction with 1. The pro-S hydrogen resides closer to the QM ideal transition state (green dot) geometry in the crystal structure (orange dot) and throughout the simulation. (C) Overlays of top five occupied clusters for 1 with WT and F92A TamI.

MD simulations were performed to analyze the influence of TamI on the regio- and stereoselectivity within the oxidative cascade. We performed the simulations beginning from the co-crystal structure of 1 with TamI. To ensure these simulations were representative the tautomeric state of the tetramic acid was examined computationally (Figure S8) with the enol tautomer 1 having the lowest relative energy. The distances of key atoms of the 1 bicyclic ketal (C10-C11-C12-C18) from the reactive Fe(IV)-oxo radical cation were tracked to explore the regioselectivity of the first oxidation event. Both C10 and C11 remained within ~4 Å throughout the 500 ns simulation, consistent with initial C10 hydroxylation (Figure S9). We next analyzed how the enzyme orients the substrate geometry at C10 relative to the Fe(IV)-oxo radical cation, and compared to the ideal transition state geometry calculated with QM. By plotting the O_heme–H10 distance and O_heme–H–C10 angle for both the S and R hydrogens relative to that of the ideal transition state geometry we find that TamI controls the orientation of 1 such that the S hydrogen is closer to a transition state geometry for the duration of the simulation (Figure 5B). This suggests that TamI controls the stereoselectivity for this C10 allylic hydroxylation step, which is likely due to the hydrophobic interactions according to the experimental mutational analysis, and is consistent with the isolation of a single C10 stereoisomer from the fermentation broth of Streptomyces sp. 307-9. MD simulations were similarly performed on both 3 and 4, which revealed that the orientation of the tirandamycin substrates change upon increasing oxidation. The geometry of the substrate throughout these simulations, including a dihedral angle for epoxidation of 3 and the angle for C─H abstraction of 4, were compared those of the ideal transition state geometry (Figure S10), which reveal that the tirandamycins can occupy the appropriate geometry for the expected reactivity in the enzyme active site. This is also demonstrated in visualizing the top occupied clusters of the MD simulations, which when overlayed, show different substrate binding modes as the oxidation cascade progresses and show that the expected reactive position for 3 and 4 (C12 and H18, respectively) to be closest to the reactive Fe(IV)-oxo radical cation in the corresponding simulations (Figure S10).

As reported, mutagenesis experiments highlighted the importance of the hydrophobic residues within TamI’s active site: F92, L399, and I400. We analyzed the influence of these residues on the substrate orientation by performing MD simulations of WT TamI with 1, as well as the single point mutants, F92A, L399A, and I400A. We found that the RMSD of 1 within these various simulations is much lower in the WT relative to the mutants, which shows that the substrate maintains a more stable geometry within the active site. (Figure 5C, Figure S11) However, 1 occupies many different positions in the active sites of the F92A, L399A, and I400A mutants, suggesting that these residues are essential for generating the proper geometry for catalysis.

Lastly the MD simulations provided insight into the binding mode of Tirandamycin C in comparison to other P450s regarding substrate anchoring, which has been a key component of substrate engineering efforts with biocatalytic P450s. Surprisingly the extensive polar interactions about the tetramic acid observed in the crystal structure (Figures 2-3) rapidly relax within the first few ns of the MD simulation (Figure S12), while proximity of the bicyclic ketal to the heme center is maintained. Consistent with the experimental data, this observation also indicates that hydrophobic interactions around the polyene region are most critical for productive substrate binding and catalysis in TamI. These combined biochemical, structural and computational analyses reveal key determinants for binding and catalysis that are distinct from the homologous P450 PikC (47% seq id) in which electrostractic interactions between carboxylate side chains and the desosamine dimethylamino group are paramount. This was unexpected, given the similarity between the active site architecture and apparent substrate binding mode in the complex structures (Figure 6). Given the work from related macrolide oxidizing P450s MycC1²⁷ (41% seq id-TamI) and TylH1²⁸ (35% seq id-TamI) we observe that natural product biosynthetic P450s employ a variety of strategies to accommodate substrate, and that subtle differences can have pronounced functional outcomes. In the case of TylH1, mutation of a few key residues in the BC and FG-loop region were sufficient to convert a “specialist” enzyme with narrow substrate scope into a “generalist” that could accommodate a variety of glycosylated (e.g. macrolide) and non-glycosylated (e.g. macrolactone) substrates reminiscent of its homolog MycC1. Future exploration of substrate scope and protein engineering in TamI will be enabled by the conceptual basis established in this work.

Figure 6. Overlay of TamI with PikC ana MycCI.

Despite the evident similarity in binding mode of their respective substrates, different factors drive molecular recognition in TamI (gray) compared to PikC (cyan, 2C7X) and MycCI (magenta, 5FOI). 1 is colored yellow.

CONCLUSION

The current structural and computational investigation of TamI demonstrated the overall structure, molecular basis of substrate binding and pattern of sequential oxidation of 1. Comparison of these structures highlights the conformational movement upon substrate binding. The substrate-bound form of TamI adopts a closed state in which 1 is buried inside the protein core consistent with previous reports of substrate-induced conformational changes in P450 enzymes. The tirandamycin C-bound structure of TamI revealed key residues responsible for catalysis. TamI mutational analysis demonstrated that steric and hydrophobic interactions mediated by Phe92, Leu399, Ile400 are critical for 1 oxidation. These experimental observations were further verified by MD simulations of the alanine mutants of Phe92, Leu399, Ile400 in which we observed that the substrate samples diverse binding orientations due to the loss of hydrophobic and π-π interactions. Overall, these data reveal the dominance of hydrophobic interactions with the entire substrate for adequate positioning in the active site. QM calculations describe the inherent preference of C10 (secondary allylic carbon) hydroxylation of 1 in the first oxidation step, but displayed no energetic preference for the stereoselectivity in C10 hydroxylation. MD simulations demonstrate that TamI orients the substrate in a specific geometry where the pro-S hydrogen is closer to the ferryl oxygen for abstraction. This observation verifies the single stereoisomer isolated from the fermentation broth of Streptomyces sp. 307-9. QM calculations provide further insights into the energetic preference for the subsequent reactions in the cascade that match the experimentally observed order of oxidation. MD simulations with 3 and 4, demonstrated that the substrate binding orientations consistent with idealized geometries are preferred in the active site environment. Together these results provide detailed information on physiochemical principles that control substrate binding, specificity and selectivity of TamI in catalyzing iterative multistep C─H oxidation of tirandamycin intermediates to produce increasingly potent antibiotic molecules.²⁹ This investigation further underscores the importance of identifying and exploring multi-functional natural product biosynthetic tailoring enzymes that catalyze late stage C─H functionalization of complex organic compounds.

MATERIALS AND METHODS

Materials & general experimental procedures:

Unless otherwise noted, chemicals, reagents, and solvents were purchased from EMD Millipore, Sigma-Aldrich, and Thermo Fisher Scientific. Agarose for gel electrophoresis was purchased from BioExpress (VWR). Kanamycin sulfate, isopropyl-β-Dthiogalactopyranoside (IPTG), and dithiothreitol (DTT) were obtained from Gold Biotechnology. Tris-HCl, ampicillin disodium salt and NADP⁺ were purchased from Amresco. Chloramphenicol was obtained from Roche. Thiamine and NADPH were purchased from Chem-Impex. δ-aminolevulinic acid was purchased from Oakwood Chemical. Lysozyme was purchased from RPI. Imidazole was purchased from AK Scientific. Amicon Ultra centrifugal filters used for protein concentration were from EMD Millipore. PD-10 columns were purchased from GE Healthcare. Glucose-6-phosphate was from Biosynth, and glucose-6-phosphate dehydrogenase (yeast) was from Alfa Aesar.

Deionized water was obtained from a Milli-Q system (EMD Millipore) using Q-Gard 2/Quantum Ex Ultrapure organex cartridges. Media components for E. coli growth were purchased from EMD Millipore, Sigma-Aldrich, and Thermo Fisher Scientific unless otherwise specified. Glycerol was purchased from BDH (VWR). LB broth (Miller) and LB agar (Miller) were obtained in pre-made granulated form from EMD Millipore. TB broth was made from individually purchased components and consisted of 4% (v/v) glycerol. Media and solutions were autoclaved or sterile filtered before use. For bacterial culture, all media solutions were autoclaved and buffers, and other solutions, pH was monitored using a VWR sympHony SB70P pH meter calibrated according to the manufacturer’s specifications. When specified, room temperature (rt) was ~22-23 °C.

DNA oligonucleotides for cloning and mutagenesis were purchased from Integrated DNA Technologies. PCR was performed using a Bio-Rad iCycler thermal cycler system. Restriction endonucleases and other associated molecular biology reagents were purchased from New England Biolabs. Invitrogen PCR cleanup/gel extraction and plasmid miniprep kits were purchased from Thermo Fisher Scientific. All DNA manipulations were accomplished following the manufacturer’s protocols. DNA concentrations and protein purity indexes were measured using a NanoDrop ND-1000 spectrophotometer. The DNA sequencing was performed at the University of Michigan DNA Sequencing Core. E. coli DH5α was used for plasmid preparation and maintenance while E. coli BL21(DE3) was used for protein overexpression. Chemically competent E. coli cells were prepared using the method of Inoue.³⁰ Optical density (OD₆₀₀) was measured using an Eppendorf BioPhotometer.

Protein purifications were performed manually using prepacked Ni²⁺⁻NTA columns with GE peristaltic pump P1, and automatically with GE columns and AKTA Pure FPLC machine, Reverse-phase high-performance liquid chromatography (HPLC) purification was performed using Luna C18 columns with the following specifications: dimensions, 150 x 4.6 mm; particle size, 5 μm; pore size, 100 Å; and a solvent system of acetonitrile and H₂O supplemented with 0.1% formic acid.

All protein figures were generated using Pymol software.

Cloning, site-directed mutagenesis, and preparation of recombinant enzymes:

The tamI_pSJ2 expression vector was constructed as described previously.¹² For crystallization, tamI was subcloned into a modified pET28 vector containing a TEV-cleavable N-terminal His₈ tag using the “Quikchange” cloning protocol (“tamI_pET28H8T”). Site-directed mutagenesis of TamI was performed by PCR using Phusion High-Fidelity DNA Polymerase and a standard site-directed mutagenesis protocol. All constructs and mutations were verified by Sanger DNA sequencing at the University of Michigan DNA Sequencing Core.

Overexpression and purification of TamI and it's recombinant mutant proteins:

Expression and purification of TamI from the tamI_pSJ2 expression vector was conducted as described previously.¹² The construct tamI_pET28H8T was used to transform E. coli BL21(DE3) for protein overexpression. Individual colonies were selected for overnight growth (37 °C) in 10 mL LB containing kanamycin (50 μg/mL). 6 x 1 L of TB (2.8 L Fernbach flasks) supplemented with kanamycin (50 μg/mL), and glycerol (4% v/v) were inoculated with 10 mL overnight cultures and incubated at 37 °C (200 rpm). When the OD₆₀₀ reached 0.6-1.0, the cultures cooled in an ice-water bath (10-20 min) before IPTG (0.2 mM) and δ-aminolevulinic acid (1 mM) were added to induce protein expression and to promote production of the heme cofactor in E. coli, respectively. The cultures were grown at 18 °C for 18-20 h before the cells were harvested by centrifugation and stored at −80 °C. The following purification steps were performed on ice or at 4 °C. The cells were thawed and resuspended in 30 mL of lysis buffer (50 mM Tris-HCl, pH 7.4 at room temperature , 50 mM NaCl, 10% (v/v) glycerol, 1 mM PMSF, 0.5 mg/mL lysozyme, 2 mM MgCl₂, 5-10 U/mL Benzonase nuclease) per 1 L of original overexpression culture (i.e., 2-3 mL per 1 g of cells). The cell suspension was incubated on a nutating shaker for 1 h prior to sonication using a Model 705 Sonic Dismembrator (Thermo Fisher Scientific) and centrifugation at 50,000 x g for 30 min to remove cellular debris. To the resulting clarified lysate 4 M solutions of NaCl and imidazole were added to final concentrations of 300 mM and 10 mM, respectively. The lysate was filtered through a syringe-operated 0.45 μm filter (Corning) before loading onto a prepacked column containing 7mL of Ni-NTA resin (Qiagen) using peristaltic pumps (GE Healthcare). The loaded material was washed with 5 CV of wash buffer (50 mM Tris-HCl, pH 7.4, 300 mM NaCl, 10 mM imidazole, 10% (v/v) glycerol) followed by 10 CV of wash buffer containing 5% elution buffer (50 mM Tris-HCl, pH 7.4, 300 mM NaCl, 300 mM imidazole, 10% (v/v) glycerol) to remove additional protein contaminants. TamI was eluted with elution buffer and the fractions containing pure material were assessed by SDS-PAGE and by monitoring absorbance A₄₂₀/A₂₈₀. Pooled fractions were concentrated using 30 kD MWCO centrifugal filters. Concentrated protein was dialyzed against storage buffer (50 mM Tris, pH 7.4, 1 mM EDTA, 0.2 mM DTT, 10% (v/v) glycerol). Aliquots of purified protein were flash frozen in liquid N₂ and stored at −80 °C. The concentration of functional P450 was assessed by obtaining CO difference spectra according to the established protocol.³¹ For crystallization, the His-tag was removed after Ni²⁺-NTA purification by incubation with purified TEV protease (1: 30 molar ratio) during dialysis in storage buffer overnight. Un-cleaved TamI and the His-tagged TEV were removed by applying the digest to a Ni²⁺-NTA column equilibrated with loading buffer. Cleaved TamI was collected in the flow-through fractions and concentrated using an Amicon Millipore concentrator (30 kD MWCO). The concentrated TamI solution was subjected to size exclusion purification via an ÄKTA FPLC system (GE Healthcare) with Superdex 200 Increase 10/300 GL column in storage buffer (50 mM Tris, pH 7.4, 1 mM EDTA, 0.2 mM DTT, 5% (v/v) glycerol). Purified fractions were flash frozen in liquid N2 and stored at −80 °C. To obtain a crystal structure of the TamI•tirandamycin C substrate complex, TamI was incubated overnight with 0.2 mM TirC at 4°C.

The UV–visible spectrum of the purified substrate-free TamI exhibited a major(γ) Soret peak at 417 and the smaller α and β bands at 567 and 533 nm, respectively, indicating that the TamI protein exists in its low spin state. The dithionite reduced spectrum of TamI displayed the absorption maximum of the Soret band at 410 nm. The reduced carbon monoxide bound form of the protein showed a typical peak maximum at 450 nm.

Expression and purification of other proteins:

The plasmid (pMAL-c, New England Biolabs) encoding spinach ferredoxin reductase linked to the C-terminus of E. coli maltose-binding protein (MBP) was obtained as a gift from Dr. Giuliana Zanetti. This construct was transformed into E. coli BL21(DE3) cells, and proteins were expressed and purified as described previously.²⁷ Spinach ferredoxin was expressed and purified as noted elsewhere.³²

Structure of substrate-free TamI:

Initial efforts to determine the TamI crystal structure were conducted using a tamI-pSJ2 expression vector containing an N-terminal His₈ tag.¹² Crystals grown with this construct were used to solve the TamI structure by molecular replacement and revealed a single polypeptide chain in the asymmetric unit; however, the P450 active site was occupied by sixteen N-terminal residues of a symmetryrelated protomer (Figure S1). The bound N-terminal segment was derived from the linker region between the His₈ tag and the EcoRI restriction site and prompted us to construct a tobacco etch virus protease (TEV)-cleavable construct (tamI-pET28H8T) where the affinity tag and linker could be removed by digestion. Crystals of substrate-free TamI were grown by vapor diffusion at 20 °C with a well solution composed of 3.5 M sodium formate, pH 7.0 and 0.1 M strontium chloride. Crystals were harvested using nylon loops and vitrified by rapid plunging into liquid nitrogen. Substrate-free TamI crystallized in space group I422 with unit cell dimensions of a = 154.5, b = 154.5, c = 91.8 Å and one chain in the asymmetric unit. X-ray data were collected at 100K on the Advanced Light Source 8.3.1 beamline in Berkeley CA, USA. Diffraction data were integrated and scaled using MOSFLM³³ and SCALA.³⁴ The structure of TamI was solved by molecular replacement with Phaser-MR³⁵ using PikC (PDBid: 2X5W, chain B) as a search model. Three rounds of autobuild using Buccaneer³⁴ were used to provide and initial model. Iterative rounds of manual building in Coot³⁶ and refinement with Refmac³⁷ and Phenix.refine³⁸ were used to furnish the final model. Data collection and refinement statistics are given in Table S2.

Structure of tirandamycin C-bound TamI:

TEV-cleaved TamI in the presence of tirandamycin C gave a different crystal form with eight chains in the asymmetric unit and substrate bound in each active site (Figure 2, Figure S2). Crystals of tirandamycin C-bound TamI were grown by vapor diffusion by mixing 2 μL of 7 mg/mL tirandamycin C-bound TamI with 2 μL of a well solution composed of 1.6 M ammonium sulfate, 0.9 M sodium chloride, 0.1 M bis-tris pH 6.5, 2.5% ethylene glycol. Narrow, rod-shaped crystals were cryoprotected by the addition of well solution containing 9% ethylene glycol. 0.2 mM tirandamycin C, 2% DMSO directly to the sitting droplets. Cyrstals were harvested using nylon loops and vitrified by rapid plunging into liquid nitrogen. Tam•tirandamycin C complex crystallized in spacegroup C2 with unit cell dimensions of a = 224.6, b = 57.2, c = 282.7 Å, α = 90°, β = 90.9°, γ = 90° and eight chains in the asymmetric unit. X-ray data were collected at 100 K on beamline 23ID-B at the General Medical Sciences and Cancer Institute Structural Biology Facility at the Advanced Photon Source in Argonne, IL, USA using helical data collection along the long axis of the rod-shaped crystal. Diffraction data were integrated and scaled using XDS.³⁹ The structure was solved by molecular replacement with Phaser-MR³⁵ using substrate-free TamI as a search model. To generate the intial model Phenix.autobuild³⁸ was performed using an eightfold-NCS averaged map. Iterative rounds of manual building in Coot³⁶ and refinement with Phenix.refine³⁸ were used to furnish the final model. Data collection and refinement statistics are given in Table S1.

UV-visible absorption spectroscopy:

Measurement of UV-visible spectra for TamI and mutant protein were measured at room temperature on Molecular Devices SpectraMax M5 UV-visible spectrophometer. Proteins were dissolved in 10mM potassium phosphate buffer, pH 7.4, were used for the measurement of spectra of the oxidized and the reduced form of TamI. The mutant proteins were reduced with a few grains of sodium dithionite and the concentration of the active P450 form of TamI was confirmed by CO-difference spectroscopy using ε(450-490) = 91 mM⁻¹cm⁻¹. After CO bubbling of the solution for 30-40 s, the spectrum of CO-bound reduced P450 species was recorded using the previously reduced spectrum for the same sample as a reference (Figure S13). The spectral shifts induced by tirandamycin binding were assayed at room temperature under aerobic conditions. Using a parent stock of tirandamycin C in DMSO, serial dilutions were prepared to obtain stock solutions of varying concentrations of substrate for binding experiments. Purified P450s were diluted in storage buffer to a final concentration of 1 μM and transferred to a 1 cm quartz cuvette (Beckman). Varying concentrations of substrate were independently titrated into the solution in small aliquots (<2 μL) to achieve the desired final concentrations. Scans were recorded from 365 nm to 429 nm (2 nm steps) until the P450s became saturated. The starting spectrum was subtracted from subsequent experimental spectra and average absorbance differences (ΔA = A_peak – A_trough) were plotted against concentration of substrate. The dissociation constant (K_d) for the P450s was calculated by fitting data points to either a nonlinear tight-binding quadratic equation (Equation 1) for high-affinity ligands (Kd ≤ 5 μM) or a rectangular hyperbolic function (Equation 2) for low-affinity ligands (Kd ≥ 5 μM) using the GraphPad Prism 8 software. All titrations were performed in triplicates and the K_d values reported represent the average of the three sets of data.

$$\Delta \mathrm{A}=({\mathrm{A}}_{\max }∕2[\mathrm{E}])({\mathrm{K}}_{\mathrm{d}}+[\mathrm{E}]+[\mathrm{S}])\text{-}({\mathrm{K}}_{\mathrm{d}}+[\mathrm{E}]+[\mathrm{S}]{)}^2-4[\mathrm{E}][\mathrm{S}{]}^{{}^{{}_1}∕_{{}^2}}$$ Equation 1

$$\Delta \mathrm{A}={\mathrm{A}}_{\max }[\mathrm{S}]∕({\mathrm{K}}_{\mathrm{d}}+[\mathrm{S}])$$ Equation 2

ΔA = Average absorbance difference (peak-to-trough) at each titration point; A_max= Maximum absorbance difference at substrate saturation; [E] = Total enzyme concentration; [S] = Substrate concentration

Enzyme assays:

Isolation of TirC has been described previously.²⁹ The in vitro enzymatic conversions of tirandamycins were performed in a total volume of 250 μl of conversion buffer (50 mM NaH₂PO₄, pH 7.4, 1 mM EDTA, 0.2 mM dithioerythritol, 10% glycerol) containing 500 μM substrates, 2 μM TamI, 4 mM spinach ferredoxin, 2 μM ferredoxin-NADP⁺ reductase, 1mM MgCl₂, 5 mM glucose-6-phosphate, and 1 U/mL glucose-6-phosphate dehydrogenase in storage buffer (50 mM NaH₂PO₄, pH 7.4, 1 mM EDTA, 0.2 mM DTT, 10% (v/v) glycerol). Reactions were incubated at 30 °C (100 rpm) for 3h before quenching by extraction, using 2 × 200 ml of CHCl₃. The resulting organic extract was dried, redissolved in 120 ml of methanol and analyzed by the addition of 100 μL of methanol with HPLC. All reactions were performed and analyzed in triplicate. HPLC analysis was performed on an Agilent 1100 Series HPLC system with UV detection at 354 nm using a Phenomenex Luna C18 column with the following specifications: dimensions, 150 x 4.6 mm; particle size, 5 μm; pore size, 100 Å. For analysis of reactions HPLC conditions we used mobile phase (A = deionized water + 0.1% formic acid, B = acetonitrile + 0.1% formic acid); and the gradient 10% B for 1 min, 10% to 75% B over 35 min, 95% B for 1 min, 10% B for 2 min.

Computational Methods:

Quantum Mechanics:

Conformational searches were performed with MMFF in Spartan. All QM calculations were performed with Gaussian 09.⁴⁰ Geometry optimizations and frequencies calculations were performed at the B3LYP level with LANL2DZ⁴¹ for iron and 6-31 G(d) for all other atoms. Transition structures contained one negative frequency. Enthalpies and free energies were computed at 1 atm and 298.15K. A correction to the entropy was applied in accordance with the work of Truhlar et al.⁴² Single point energy computations were performed at the B3LYP-D3(BJ) level with LANL2DZ for iron and 6-311+G(d,p) in gas phase for all calculations, except for the tetramic acid tautomer comparison, which included CPCM for water.

Molecular Dynamics (MD) Simulations:

The heme iron(IV)-oxo complex involved in the cytochromecatalyzed oxidative hydroxyiation and epoxidation cycle (compound I) was used to model the active form of the cofactor. Simulations were performed using the GPU code (pmemd)⁴³ of the Amber 12 package,⁴⁴ The Amber-compatible parameters developed by Cheatham et al. were used for compound I and its axial Cys ligand.⁴⁵ Parameters for tirandamycin substrates were generated within the antechamber module using the general AMBER force field (gaff)⁴⁶, with partial charges set to fit the electrostatic potential generated at the HF/6-31G(d) level by the RESP model.⁴⁷ The charges were calculated according to the Merz-Singh-Kollman^48-49 scheme using Gaussian 09. Each protein was immersed in a pre-equilibrated truncated cuboid box with a 10 Å buffer of TIP3P⁵⁰ water molecules using the tleap module, resulting in the addition of around 11370 solvent molecules. The systems were neutralized by addition of explicit counter ions (Na+ and Cl−). All subsequent calculations were done using the widely tested Stony Brook modification of the Amber 99 forced field (ff99sb).⁴⁶ The substrate and enzyme were optimized for total 1,000,000 steps, with 750,000 steepest descent steps and 250,000 conjugate gradient steps. The systems were gently heated using six 50 ps steps, incrementing the temperature by 50 K for each step (0-300 K) under constant-volume and periodic-boundary conditions. Water molecules were treated with the SHAKE algorithm such that the angle between hydrogen atoms were kept fixed. Long-range electrostatic effects were modelled using the particle-mesh-Ewald method.⁵¹ An 8 Å cutoff was applied to the Lennard-Jones and electrostatic interactions. Harmonic restraints of 30kcal/(mol Ų) were applied to the solute and the Andersen equilibration scheme was used to control and equalize the temperature. The time step was kept at 1 fs during the heating stages, allowing potential inhomogeneities to self-adjust. Each system was then equilibrated for 2 ns with a 2 fs time step at a constant volume. Production trajectories were then run for an additional 500 ns under the same simulation condtions.