Leveraging a Structural Blueprint to Rationally Engineer the Rieske Oxygenase TsaM

Abstract

Rieske nonheme iron oxygenases use two metallocenters, a Rieske-type [2Fe-2S] cluster and a mononuclear iron center, to catalyze oxidation reactions on a broad range of substrates. These enzymes are widely used by microorganisms to degrade environmental pollutants and to build complexity in a myriad of biosynthetic pathways that are industrially interesting. However, despite the value of this chemistry, there is a dearth of understanding regarding the structure–function relationships in this enzyme class, which limits our ability to rationally redesign, optimize, and ultimately exploit the chemistry of these enzymes. Therefore, in this work, by leveraging a combination of available structural information and state-of-the-art protein modeling tools, we show that three “hotspot” regions can be targeted to alter the site selectivity, substrate preference, and substrate scope of the Rieske oxygenase p-toluenesulfonate methyl monooxygenase (TsaM). Through mutation of six to 10 residues distributed between three protein regions, TsaM was engineered to behave as either vanillate monooxygenase (VanA) or dicamba monooxygenase (DdmC). This engineering feat means that TsaM was rationally engineered to catalyze an oxidation reaction at the meta and ortho positions of an aromatic substrate, rather than its favored native para position, and that TsaM was redesigned to perform chemistry on dicamba, a substrate that is not natively accepted by the enzyme. This work thus contributes to unlocking our understanding of structure–function relationships in the Rieske oxygenase enzyme class and expands foundational principles for future engineering of these metalloenzymes.

Introduction

The Rieske nonheme iron oxygenases, or Rieske oxygenases, use a Rieske-type [2Fe-2S] cluster and a mononuclear iron center to activate molecular oxygen. These enzymes typically facilitate the functionalization of C–H bonds¹⁻⁶ and are often recognized for their ability to catalyze monooxygenation and dioxygenation chemistry. Rieske oxygenases have also been shown to catalyze a handful of other divergent reactions including formylation,⁷ C–C bond cleavage,⁸ and carbocyclization.⁹ These transformations are integral to the biosynthesis of medically, industrially, and environmentally relevant natural products and to the degradation of aromatic and polyaromatic hydrocarbons.¹⁻¹⁰ However, despite the valuable applications of this chemistry, remaining gaps in knowledge regarding the structure–function relationships in this class of enzymes have limited the practical use of Rieske oxygenases as biocatalysts.

Currently, the 70,000 members of the Rieske oxygenase enzyme class are represented by approximately 20 crystal structures of different enzymes (Figure 1A and Figure S1A).¹¹⁻³⁰ These structures have revealed that Rieske oxygenases exist in heterohexameric (α₃β₃ or α₃α′₃),¹¹⁻²¹ homohexameric (α₃α₃),²² and homotrimeric (α₃)²³⁻³⁰ quaternary architectures. Based on these structures, several pioneering studies have revealed that residues found in the active site,^{26,29,31−35} substrate entrance tunnel,³⁶ and flexible loop regions^21,37 of the catalytic α-subunit can be independently targeted to influence catalytic outcomes. Indeed, residues located in the active site of several Rieske oxygenases are documented to impact substrate positioning^19,27−29 and, in some cases, have been demonstrated to affect catalytic outcomes.^{26,31−33,38} Moving outside of the active site, a recent computational study revealed that the substrate scope of the α₃β₃ Rieske oxygenase naphthalene dioxygenase relies upon a tunnel that leads from the surface of the protein to the active site.³⁶ Further, the site selectivity of the caffeine-demethylating α₃α′₃ Rieske oxygenases NdmA and NdmB has been demonstrated to rely on residues found in the active site and a flexible loop. Specifically, an NdmA variant with two active site substitutions and the NdmB loop sequence has markedly increased activity on the native NdmB substrate and no activity on the native NdmA substrate²¹ (Figure S1B). Additional studies demonstrated that the activity of two α₃ Rieske oxygenases, SxtT and GxtA, which catalyze monooxygenation reactions at the C12 and C11 positions of the saxitoxin scaffold, respectively, can be attributed to the cooperative work of three protein regions^6,27,28 (Figure S1B). The mutation of six residues distributed across the active site, substrate entrance tunnel, and flexible β13-to-β14 connecting loop of SxtT into their GxtA counterparts results in creation of an SxtT variant that fully exhibits the site selectivity, substrate scope, and substrate specificity of GxtA.^6,27,28 The tunnel-lining residues of SxtT direct the substrate into the active site where it is held by active site and loop residues in the correct orientation for catalysis.^6,27,28

Figure 1.

Understanding the blueprint for Rieske oxygenase chemistry. In this work, we investigate whether the structural trends shown to confer the site selectivity, substrate scope, and substrate specificity of the Rieske oxygenases SxtT and GxtA are widely used in the Rieske oxygenase enzyme class. (A) Toward this investigation, three target enzymes, TsaM, VanA, and DdmC, were selected as a model system to test the impact of the three regions on catalysis. These enzymes were selected through the use of a sequence similarity network (SSN) and reside in a cluster that is distinct from where SxtT, GxtA, NdmA, and NdmB are found. In this SSN, structurally characterized enzymes are indicated with a purple diamond. Additional details regarding the SSN can be found in Figure S1. (B) A phylogenetic tree of the subcluster in the SSN that contains the structurally characterized enzyme DdmC, as well as TsaM and VanA, reveals that each of these enzymes occupies different clades. The reactions catalyzed by TsaM, VanA, and DdmC are shown on the right. (C) Through making iterative changes in the active site, tunnel, and loop regions (as described for SxtT and GxtA^27,28), the site selectivity and substrate specificity of TsaM were altered. (D) Specifically, the introduction of six to 10 mutations into the “hotspot” regions depicted in panel C allowed TsaM to be adapted to exhibit the catalytic activity and substrate specificity of VanA and DdmC, respectively.

Along with a structural analysis that revealed the presence of an analogous flexible loop and calculable substrate entrance tunnels in other structurally characterized α₃ Rieske oxygenases, these previous results suggest that residues found both inside and outside of the active site regions have a widespread influence on catalysis.^6,27,28 The identification of such a subset of positions that can be targeted to alter the site selectivity, substrate preference, substrate scope, or product distribution of a reaction is valuable because it limits the sequence space that must be explored to confer desired catalytic properties to a Rieske oxygenase of interest. However, at this point, to ensure the broad applicability of the observed trends, it is critical to explore whether the identified sequence–structure–function relationships are indeed valuable for enzyme engineering. Primarily, it is integral to determine whether the identified “hotspots” can be used to engineer systems with lower sequence identity and more disparate substrates than that of SxtT and GxtA (88% sequence identity) or NdmA and NdmB (50% sequence identity, Figure S1).

Thus, in this work, to determine whether targeting these “hotspots”, in an individual or cumulative manner, can be used to alter the reaction outcome of other divergent Rieske oxygenases, a “proof-of-principle” set of enzymes was identified using a sequence similarity network (SSN) of the Rieske oxygenase enzyme class (Figure 1A and Figure S1). The selected enzymes share approximately 35% sequence identity with one another, represent three distinct evolutionary clades of a phylogenetic tree, and perform monooxygenation reactions at the para, meta, and ortho positions of a carboxylate-containing aromatic substrate (Figure 1A,B and Figures S1–S3). These enzymes include p-toluenesulfonate methyl monooxygenase (TsaM), vanillate monooxygenase (VanA), and dicamba monooxygenase (DdmC), of which only DdmC has been structurally characterized.³⁰ TsaM natively catalyzes a monooxygenation reaction on the methyl groups of p-toluenesulfonate and p-methylbenzoate, with the aid of its reductase TsaB (Figure 1B).³⁹⁻⁴¹ DdmC and VanA, on the other hand, catalyze oxidative demethylation reactions on the methoxy substituents of dicamba and vanillate, in the presence of reductase proteins DdmA/DdmB and VanB, respectively^30,42−45 (Figure 1B). Each of these enzymes is an α₃ Rieske oxygenase, and in the SSN, each of these sequences is detached from the regions where SxtT, GxtA, NdmA, and NdmB reside, making them an excellent paradigm for testing whether the previously identified design principles can be targeted to alter reactivity (Figure 1 and Figures S1 and S3).

Through this work, it was determined that as demonstrated for SxtT and GxtA, TsaM can be rationally engineered to exhibit the non-native site selectivity and substrate specificity for both VanA and DdmC (Figure 1C,D). Remarkably, to affect these changes in TsaM, an enzyme that is composed of 347 residues, only six to 10 residues distributed between the active site, β13-to-β14 connecting loop, and substrate entrance tunnel needed to be altered. The small subset of mutations to simulate the wild-type properties of VanA and DdmC suggests that these architectural regions are also important determinants of reaction specificity in other members of the Rieske oxygenase enzyme class. Therefore, this work provides key details regarding how to predictively link protein sequence to function in the Rieske oxygenase enzyme class and contributes to our understanding of how to rationally manipulate Rieske oxygenase chemistry.

Materials and Methods

Creation of TsaM Variants

For all variants used in this work, mutated plasmids were produced using an Agilent QuikChange lightning site-directed mutagenesis kit, and the thermocycling steps were performed in a BioRad C1000 thermal cycler. The protocol was adapted from Agilent; briefly, 50 μL mutagenesis reactions were prepared to contain 100 ng of the dsDNA template (pET-28a(+)-TEV-tsaM) and 125 ng of each oligonucleotide primer. The primers were synthesized by IDT, and their sequences are listed in Table S1. Following PCR, the amplified products were digested with 2 μL of the DpnI restriction enzyme for approximately 30 min at 37 °C. The digested reaction mixtures were transformed into XL10-Gold ultracompetent cells and plated on Luria–Bertani (LB) agar with kanamycin (50 μg/mL). Sanger sequencing (Genewiz) was used to verify successful incorporation of the desired mutation.

Protein Expression Conditions of Wild-Type TsaM and TsaM Variants

A codon-optimized version of the gene encoding TsaM was synthesized and subsequently cloned into a pET-28a(+)-TEV plasmid (Genscript). The plasmid was transformed by heat shock into C41(DE3) chemically competent Escherichia coli (E. coli) cells (Novagen). C41(DE3) cells were used for expression because we typically observe better metal incorporation into Rieske oxygenases using this cell line. Transformed cells were plated on LB agar with kanamycin (50 μg/mL) and were allowed to incubate overnight. After incubation, a single colony containing the pET-28a(+)-TEV-tsaM plasmid was used to inoculate 8 mL of LB containing 50 μg/mL kanamycin. The inoculated culture was shaken at 200 rpm and incubated overnight at 37 °C. The overnight starting culture was then inoculated into 1 L of LB in a 2 L Erlenmeyer cell culture flask containing 50 μg/mL kanamycin. The culture was incubated at 37 °C and shaken at 200 rpm. The 1 L cultures were monitored until A₆₀₀ reached a value between 0.6 and 0.8. Upon reaching this desired density, cell cultures were removed from shakers and allowed to cool to 20 °C. After cooling, the cultures were induced with 0.1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). At this induction stage, 0.2 mg/mL ferric ammonium citrate and 0.4 mg/mL ferrous ammonium sulfate hexahydrate ((NH₄)₂Fe(SO₄)₂·6H₂O) were added to the cultures. The large-scale induced cultures were allowed to incubate for an additional 18 h at 18 °C with shaking at 160 rpm. The cells were harvested after the overnight incubation producing a typical wet mass of 6 g per 1 L culture of wild-type TsaM. For TsaM variants 8, 9, 12, and 15, the average pellet mass was 3–4 g per 1 L of cell culture. The cell mass for other TsaM variants was 4–6 g per 1 L of cell culture.

Purification Protocol of Wild-Type TsaM and TsaM Variants

After the 1 L cell cultures were incubated for 18 h, the cells containing the His-tagged TsaM protein were pelleted by centrifugation at 5000 rpm for 30 min at 4 °C. The supernatant was decanted, and the cell pellet was collected. Typically, the collected pellets from two 1 L cultures were resuspended in 60 mL of lysis buffer (50 mM Tris–HCl (pH 7.2) and 250 mM NaCl). The resuspended cells were incubated on ice and lysed using a Fisherbrand Model 120 Sonic Dismembrator and a 10 min program that cycles between 5 s sonication at a 30% amplitude and 10 s rest. The lysed cells were clarified by centrifugation using an Eppendorf centrifuge 5810R at 12,000g for 50 min at 4 °C. The clarified supernatant was collected and loaded onto a pre-equilibrated 5 mL HisTrap HP column (Cytiva). The HisTrap column was washed with 8 column volumes of buffer A (50 mM Tris–HCl (pH 7.2), 250 mM NaCl, and 10 mM imidazole), and the tagged protein was eluted with a 5-column volume gradient of buffer B (50 mM Tris–HCl (pH 7.2), 250 mM NaCl, and 200 mM imidazole). The eluted fractions were concentrated to a final volume of approximately 5 mL. The concentrated fractions were loaded onto a HiPrep 16/60 Sephacryl S200-HR (Cytiva) gel filtration column that was pre-equilibrated with buffer C (50 mM HEPES (pH 8.0), 200 mM NaCl, and 5% glycerol). The eluted trimeric TsaM protein from this column was collected and concentrated to 10 mg/mL. For wild-type TsaM, the typical yield was 2 mg of isolated protein from 1 L of cells. For most TsaM variants, the average yield was approximately 1.75 mg per 1 L of cells. However, for TsaM variants 8, 9, 12, and 16, the average yield was about 0.8 mg of isolated protein per 1 L of cell culture. The concentrated protein was aliquoted into 200 μL fractions and flash frozen using liquid nitrogen and stored at −80 °C.

Protein Expression Conditions for Vanillate O-Demethylase Oxygenase (VanA) and Dicamba Monooxygenase (DdmC)

Codon-optimized genes encoding vanA and ddmC were synthesized and cloned into pMCSG9 plasmids (Joint Genome Institute for vanA; Genscript for ddmC). These plasmids were transformed into C41(DE3) E. coli cells using the standard heat shock protocol. For both VanA and DdmC, single colonies were picked from the plate and used to inoculate 10 mL of LB with 100 μg/mL ampicillin. These 10 mL cultures were grown at 37 °C and 200 rpm overnight. The next day, each 10 mL culture was used to inoculate 1 L cultures of LB media containing 100 μg/mL ampicillin. These large cultures were grown at 37 °C and 200 rpm until A₆₀₀ reached 0.6–0.8. Cultures were cooled to 20 °C and induced by addition of 0.1 mM IPTG, 0.2 mg/mL ferric ammonium citrate, and 0.4 mg/mL ((NH₄)₂Fe(SO₄)₂·6H₂O). Cultures were incubated for an additional 18 h prior to harvesting. The average pellet for 1 L cultures of VanA and DdmC was 6 g of wet cell mass.

Protein Purification Protocol for VanA and DdmC

To obtain homogeneous VanA and DdmC protein for activity assay experiments, the cell pellets produced from 2 L cultures that contained either the overexpressed MBP-VanA or MBP-DdmC were resuspended in 60 mL of buffer A (50 mM Tris–HCl (pH 7.2), 200 mM NaCl, and 1 mM DTT). For each protein, the resuspended cell mixture was lysed and harvested as described for TsaM. The cell supernatant, which contained either MBP-VanA or MBP-DdmC, was loaded onto a 5 mL MBPTrap column (Cytiva) that was pre-equilibrated with buffer A. This column was washed with 8 column volumes of buffer A to get rid of nonspecific binding and eluted with 5 column volumes of buffer B (50 mM Tris–HCl (pH 7.2), 200 mM NaCl, 1 mM DTT, and 10 mM maltose). Fractions containing the MBP-VanA or MBP-DdmC were collected and mixed with 2 mg of tobacco etch virus (TEV) protease to cleave the MBP tag. This mixture was dialyzed in a 10,000 Da MWCO snakeskin dialysis tubing against buffer C (50 mM Tris–HCl (pH 7.2), 200 mM NaCl, and 1 mM DTT) at 4 °C overnight. The next day, the MBP tag-free VanA or DdmC proteins were loaded onto a pre-equilibrated 5 mL HisTrap (Cytiva) for the removal of both TEV protease and the cleaved His-MBP tags. The flow-through from the HisTrap column containing the tag-free VanA or DdmC was then buffer exchanged to buffer D (50 mM HEPES (pH 8.0), 200 mM NaCl, and 10% glycerol) using a PD-10 desalting column (BioRad). The final isolated VanA and DdmC proteins were concentrated to ∼250 μM and flash frozen by liquid nitrogen for storage at −80 °C.

Protein Expression Conditions of VanB and TsaB

Expression and purification methods for VanB were previously described and followed here.⁴⁶ In short, a codon-optimized gene encoding VanB was cloned into a pMCSG7 plasmid, transformed by heat shock into BL21(DE3) chemically competent E. coli cells (Novagen), and plated onto LB agar plates with ampicillin (100 μg/mL). After incubation overnight, a single colony containing the plasmid encoding VanB was used to grow a starter culture in 8 mL of LB containing ampicillin (100 μg/mL) at 37 °C overnight. The starter culture was used to inoculate 1 L of LB containing ampicillin (100 μg/mL) and was incubated at 37 °C, shaken at 200 rpm, and induced by the addition of 0.1 mM IPTG and incubated overnight at 18 °C.

For the expression of TsaB, a pET28a(+)-TEV-tsaB plasmid was generated by synthesizing a codon-optimized gene encoding TsaB and cloning the gene into a pET28a(+)-TEV plasmid (Genscript). The generated plasmid was transformed by heat shock into BL21(DE3) chemically competent E. coli cells (Novagen) and plated onto LB agar plates with kanamycin (50 μg/mL). As described for TsaM, after incubation, a single colony containing the pET28a(+)-TEV-tsaB plasmid was used to grow an 8 mL starter culture at 37 °C overnight. The starter culture was used to inoculate 1 L of Terrific Broth (TB) containing kanamycin (50 μg/mL) and was incubated at 37 °C and shaken at 200 rpm. The cultures were monitored and allowed to reach an A₆₀₀ in between 0.6 and 0.8. Once the optimal optical density was reached, the cultures were cooled to 20 °C. Upon cooling, expression in the cultures was induced by the addition of 0.1 mM IPTG and shaken overnight at 18 °C.

Purification Protocol of VanB and TsaB

After the 1 L cell cultures were incubated for 18 h, the cells containing the His-tagged protein were pelleted by centrifugation at 5000 rpm for 30 min at 4 °C. The supernatant was decanted, and the cell pellet was collected. Typically, the collected pellets from two 1 L cultures were resuspended in 60 mL of lysis buffer (50 mM Tris–HCl (pH 7.2), 250 mM NaCl, 10 mM imidazole, and 100 μM flavin adenine dinucleotide (FAD)). The resuspended cells were lysed, centrifuged, and purified using HisTrap and gel filtration chromatography as described for TsaM. The eluted reductase protein from this column was collected, concentrated to 200 μM, aliquoted into 200 μL fractions, flash frozen using liquid nitrogen, and stored at −80 °C. Incorporation of FAD into TsaB and VanB was monitored using the characteristic UV–vis absorbance maximum at 450 nm (extinction coefficient of 11,300 M^–1 cm^–1).⁴⁷

Protein Expression Conditions of DdmA and DdmB

These proteins were expressed and purified as previously described.⁴⁶ In short, the pMCSG-7 protein-containing plasmids were expressed in C41 E. coli cells, grown in LB-ampicillin media until an A₆₀₀ of 0.7 was reached, and inoculated with 1 mM IPTG. Cells were then incubated at 18 °C overnight for DdmA and for 72 h for DdmB.

Purification Protocol of DdmA and DdmB

Cell pellets were suspended in 4 mL of lysis buffer (50 mM HEPES (pH 8.0), 150 mM NaCl, and 5% glycerol, (pH 8.0)), 1 mM phenylmethylsulfonyl fluoride (PMSF), and DNAse per gram of cell pellet. Cells were then lysed via sonication for 7 min (7 s on at a 30% amplitude and 15 s off) and centrifuged for 50 min at 7500g. The supernatant was then loaded onto a HisTrap HP column (Cytiva) and eluted using a buffer that contained 50 mM Tris (pH 8.0), 150 mM NaCl, and 300 mM imidazole. The elution fractions were then pooled, buffer exchanged into lysis buffer, and flash frozen to be stored at −80 °C.

Iron Quantification for Isolated TsaM, DdmC, and VanA

Ferrozine analysis using a ferene reagent was used to quantify the iron content of isolated TsaM, VanA, and DdmC following the published protocol.⁴⁸ Briefly, 100 μL of a 100 μM enzyme was mixed with 125 μL of 0.04% ascorbate, 7.5 μL of concentrated HCl, and 12.5 μL of ferene solution. The mixture was vortexed for 30 s followed by the addition of 500 μL of 8 M Gu-HCl and 100 μL of ammonium acetate. The absorbance at 593 nm was measured after 1 h of incubation at room temperature. The number of iron ions incorporated for isolated TsaM, VanA, and DdmC was determined to be 2.6, 3.1, and 2.8 per monomer, respectively.

Circular Dichroism (CD) Experiments

Collection methods for CD spectra were previously described and adapted here.²⁷ In short, the stock solutions of wild-type TsaM and TsaM variants were used to make 350 μL samples by diluting to 10 μM using a buffer containing 5 mM HEPES (pH 8.0) and 20 mM NaCl. The prepared samples were transferred into a 10 mm quartz cuvette (Hellma), and the CD spectra were measured using a Jasco J-1500 CD spectrometer. Each sample spectrum was an average of three cumulative spectra.

Substrate and Product Standards Used in Enzymatic Reactions

Substrates and their corresponding products used in this work for assays were commercially purchased. Compounds are listed in Table S2. The purity of each compound used was greater than 95%.

Stock Preparation for Enzymatic Assays

Each of the enzymatic assays performed in this work used substrate and product standards listed in Table S2. All stock solutions of these compounds were prepared in dimethyl sulfoxide (DMSO, analytical grade) to a final concentration of 100 mM. A 5 mM stock of (NH₄)₂Fe(SO₄)₂·6H₂O was prepared in ultrapure H₂O and stored at −20 °C. A stock solution of 20 mM nicotinamide adenine dinucleotide hydride (NADH) was also prepared in ultrapure H₂O and stored at −20 °C. Wild-type TsaM, TsaM variants, wild-type VanA, and wild-type DdmC protein used for enzymatic assays were purified using the protocol described as above and stored at −80 °C. To minimize variation from degradation due to freeze–thaw cycles, the enzyme aliquots were discarded after one freeze–thaw cycle. The (NH₄)₂Fe(SO₄)₂·6H₂O and NADH aliquots were discarded after not more than five freeze–thaw cycles.

Total Turnover Number (TTN) Determination

Wild-type TsaM, VanA, DdmC, and TsaM variants were purified as described above. Enzymes were stored in −80 °C at a final concentration of 250 μM. Fresh protein aliquots were discarded after one freeze–thaw cycle. The reactions contained 5 μM oxygenase (wild-type TsaM, TsaM variants, VanA, or DdmC) and 20 μM reductase (VanB, TsaB, or DdmA/DdmB). To determine the total turnover number of each enzyme, a 50 μL reaction was prepared containing 5 μM oxygenase, 20 μM reductase, 2 mM substrate, 1 mM NADH, and 100 μM (NH₄)₂Fe(SO₄)₂. Once combined, the reactions were incubated at 30 °C for 3 h. Reactions were then quenched using a 100 μL solution of acetonitrile that contained our mass spectrometry internal standard (acetaminophen) and centrifuged at 17,000g for 15 min. A 50 μL aliquot of the supernatant was then transferred into a sample vial and run on LC–MS as described below. The TTN assay was also repeated with a 4 mM substrate or 2 mM NADH to make sure that the amount of the substrate or the electron source was not a limiting factor. All assays presented in this work were run in triplicate.

To calculate the total turnover number, standard curves of the product standard were used to calculate the total amount of products produced in enzymatic reactions. The product standard curves were constructed using the commercially purchased product following the equation

The generated product concentration was then obtained using the ratio calculated from the enzymatic assay. TTNs were calculated using the following equation:

All of the product standard curves and enzymatic reactions performed in this work were prepared and analyzed in triplicate. Product standard curves used in this work are shown in Figure S33.

Kinetic Analysis

The assessment of the steady-state kinetic parameters of wild-type TsaM, TsaM variants, VanA, and DdmC started with an investigation into the linear range of product formation. The 50 μL enzymatic reactions consisted of 400 μM substrate, 1 mM NADH, and 100 μM (NH₄)₂Fe(SO₄)₂·6H₂O. Different concentrations of oxygenase (wild-type TsaM, TsaM variants, VanA, and DdmC) at 2, 5, and 10 μM were used to make sure that the observed product signal was measurable. The reaction was quenched at 2, 5, 10, and 40 min time points with the addition of 100 μL of acetonitrile containing the mass spectrometry standard, acetaminophen.

Following the investigation of a linear range of product formation, the steady-state kinetic parameters were determined using the identified oxygenase concentration. The 50 μL enzymatic reactions contained 1 mM NADH, 100 μM (NH₄)₂Fe(SO₄)₂·6H₂O, and 0–1 mM substrate. These reactions were initiated by the addition of a predetermined concentration of the enzyme identified from the linear-range experiment described above. Reactions were quenched at the identified time point from an initial linear-range experiment with the addition of 100 μL of acetonitrile with an acetaminophen internal standard. These reactions were then centrifuged at 17,000g for 15 min. The supernatant was transferred into a sample vial and run on LC–MS using the protocol described below. The amount of products generated was calculated using the product standard curve (Figure S33). The data were plotted and fit to the Michaelis–Menten equation using GraphPad Prism 9. All reactions were performed in triplicate.

Substrate Specificity Experiments

To determine the substrate specificity of wild-type TsaM, TsaM variants, VanA, and DdmC, an enzymatic assay mixture containing equal amounts of two or three competing substrates was prepared. Enzymatic assays were performed on a scale of 50 μL containing 1 mM NADH, 100 μM (NH₄)₂Fe(SO₄)₂·6H₂O, 5 μM oxygenase (wild-type TsaM, TsaM variants, VanA, or DdmC), 20 μM reductase (VanB, DdmA, and DdmB), and 1 mM of the competing substrates. The reactions were initiated by the addition of the enzyme and then incubated at 30 °C for 3 h. Following incubation, the reaction mixtures were quenched with 100 μL of acetonitrile containing the mass spectrometry internal standard acetaminophen. The quenched reactions were centrifuged at 17,000g for 15 min, and the supernatant was run on LC–MS following the method described below.

O₂ Uncoupling Experiments

The amount of O₂ uncoupling in the wild-type TsaM and TsaM variant reactions was determined using an Invitrogen Amplex Red hydrogen peroxide/peroxidase assay kit. The experiment was performed following the vendor protocol. In brief, 5 mL of Amplex Red reagent/horse radish peroxidase (HRP) working solution was prepared, protected from light, and used within one day. The experiment was performed in a 96-well plate. A 50 μL working solution of the Amplex Red reagent/HRP was added to each microplate well containing a 50 μL enzymatic reaction that contained a 2 mM substrate, 1 mM NADH, 100 μM (NH₄)₂Fe(SO₄)₂, 5 μM oxygenase, and 20 μM reductase. The absorbance at 560 nm was measured every 5 min for a total of 190 min using a BioTek Epoch2 microplate reader. Of note, oxygenase was added last using a multichannel pipette to minimize experimental error.

Thermal Shift Assays

Thermal shift assays were conducted using a QuantStudio5 real-time PCR (Thermo Fisher Scientific) following the manufacturer protocol as described previously.⁴⁹ In brief, enzyme samples (5 μM) were combined with an 8-fold dilution of a Protein Thermal Shift Dye (Thermo Fisher Scientific) in buffer (15 mM HEPES (pH 8.0), 200 mM NaCl, and 5% glycerol) in a 96-well plate. Samples were then subjected to melting at a ramp rate of 0.05 °C/s over a temperature range of 25 to 99 °C. Fluorescence curves (570 nm) were then analyzed to determine the melting temperature (T_m) of each sample using Applied Biosystems Protein Thermal Shift Software v1.4 (Thermo Fisher Scientific). Each sample was measured in triplicate.

LC–MS Analysis of Enzymatic Reactions

LC–MS analysis used an Agilent G6545A quadrupole-time-of-flight (TOF) mass spectrometer equipped with a dual AJS ESI source and an Agilent 1290 Infinity series diode array detector, an autosampler, and a binary pump. The LC–MS solvents used were solvent A (5% acetonitrile, 95% water, and 10 mM ammonium acetate (pH 5.5)) and solvent B (95% acetonitrile, 5% water, and 10 mM ammonium acetate (pH 5.5)). An Agilent ZORBAX Rapid Resolution HT 3.5 μm, 4.6 × 75 mm SB-CN liquid column was used for the separation of all compounds. The chromatographic method used was typically 10% solvent B (0 to 1.0 min, to waste), a 10 to 98% solvent B gradient (1.0 to 5.0 min), and a 10% solvent B (1.0 min, to waste) at 0.4 mL/min. Typically, a 1–3.5 μL injection was made for each sample. Data collection was run in the negative ion mode.

Calculations of Active Site Tunnels in DdmC, TsaM, and VanA

The substrate entrance tunnels were calculated in DdmC, TsaM, and VanA using the MOLEonline server.^50,51 For the tunnel calculation of DdmC (PDB ID 3GKE), the structure was loaded into PyMol and exported as a monomer, which contained the nonheme iron center. As for TsaM and VanA, the monomers were generated with the AlphaFold Monomer v2.0 pipeline^52,53 and loaded into PyMol. The β13-to-β14 flexible connecting loops of the monomers were deleted (K188 to P224 for TsaM; L171 to P189 for VanA) due to their closed conformation. The channel parameter settings were modified to automatically compute pores from exits in all cavities. In each Rieske oxygenase, the facial triad residues were selected as the starting point for the calculation, and after calculation, the tunnels were visualized using PyMol (The PyMol Molecular Graphics System, version 2.0, Schrödinger, LLC).

Construction of a Sequence Similarity Network (SSN) and a Phylogenetic Tree

Sequence similarity networks (SSNs) were generated using the Enzyme Function Initiative Enzyme Similarity Tool (EFI-EST).⁵⁴⁻⁵⁸ An initial SSN, representing the Rieske oxygenase superfamily, was generated at an E-value of 5 from the following InterPro protein families:⁵⁹ IPR044043 (vanillate O-demethylase oxygenase-like, C-terminal catalytic domain), IPR001663 (aromatic-ring-hydroxylating dioxygenase, alpha subunit), IPR005805 (Rieske iron–sulfur protein, C-terminal), IPR013626 (pheophorbide a oxygenase), IPR037338 (aminopyrrolnitrin oxygenase PrnD, Rieske domain), IPR021028 (homotrimeric ring hydroxylase, catalytic domain), and IPR017941 (Rieske [2Fe-2S] iron–sulfur domain). Due to the large size of the network, UniRef50 cluster ID sequences were used, meaning that enzyme clusters were generated containing all UniProt IDs that shared greater than 50% sequence identity and had at least 80% overlap with the longest sequence in the enzyme cluster; these enzyme clusters were then used to generate the nodes in the SSN. Clusters containing only sequences with less than 150 amino acid residues were then excluded from the SSN, giving 11,906 nodes in the SSN. After initial separation of the SSN at an alignment score of 25, one main node cluster containing 8884 members emerged. Of the remaining smaller node clusters, several were identified as having Rieske-type [2Fe-2S] clusters but not belonging to the Rieske oxygenase superfamily (e.g., the Rieske subunits of the cytochrome bc₁ complex and the cytochrome b₆f complex). Therefore, the main node cluster was used to generate a new SSN representative of the Rieske oxygenase superfamily at an alignment score of 25. The Rieske oxygenase SSN was then analyzed at various alignment score thresholds. At an alignment score of 60, further curation was employed, as many clusters were observed to contain primarily enzymes not belonging to the Rieske oxygenase class. Curation at this stage entailed removal of any cluster containing membership annotated at greater than 50% of the ubiquinol-cytochrome c reductase iron–sulfur subunit, cytochrome b₆f complex iron–sulfur subunit, cytochrome P450, cytochrome bc₁ complex subunit Rieske/cytochrome bc₁ complex Rieske iron–sulfur subunit, nitrite reductase/ring-hydroxylating ferredoxin subunit, or MOSC domain-containing protein. SSNs were visualized with the Cytoscape software package, version 3.8.2.⁶⁰

Phylogenetic tree generation of the TsaM-VanA-DdmC node cluster, separated at an alignment score of 70, was conducted by first generating a multiple sequence alignment, using one representative sequence from each node, on the COBALT webserver.⁶¹ The multiple sequence alignment was then processed with the MEGA11 software package⁶² using a maximum likelihood method with the Jones–Taylor–Thornton (JTT) matrix model⁶³ to determine a maximum log likelihood tree following a heuristic search of generated initial trees.

Density Functional Theory Calculations

Initial structures were built and preoptimized using the Avogadro software package.⁶⁴ All calculations were carried out using the ORCA quantum chemistry software package, version 4.2.1.^65,66 All calculations consisted of geometry optimization followed by property calculation at the same level of theory, reported below. All calculations used the hybrid density functional B3LYP^67,68 with the Becke–Johnson damping scheme for dispersion correction.^69,70 The Ahlrichs TZVPP basis set⁷¹ and the def2-J auxiliary basis set⁷² were used on all atoms. The resolution of the identity approximation for Coulomb integrals⁷³ with the “chain of spheres” approximation for the exchange integrals,⁷⁴ RIJCOSX in ORCA nomenclature, was employed to facilitate faster calculations. All calculations were conducted in the gas phase with a slow convergence and tight self-consistent field criterion enforced. An increased integration grid, Grid6 in ORCA nomenclature, was used. Electrostatic potential surfaces were visualized using the Avogadro software package.⁶⁴

Results

Isolation of Three Rieske Oxygenase Systems

To begin this work, codon-optimized genes that encode the Rieske oxygenases and their partner reductase proteins, TsaM/TsaB from Comamonas testosteroni, DdmC/DdmA/DdmB from Stenotrophomonas maltophilia, and VanA/VanB from Pseudomonas sp. were synthesized. These proteins, which contained either N-terminal His- or a combination of His- and MBP-tags, were recombinantly expressed and purified (Figure S4). Once purified, the respective sets of proteins, TsaM/TsaB, DdmC/DdmA/DdmB, and VanA/VanB, were tested for their ability to catalyze their native transformations (Figure 2A and Figures S5–S7). Here, it was determined that whereas each set of proteins was active, the amount of products formed for the TsaM/TsaB and DdmC/DdmA/DdmB systems was low relative to the VanA/VanB combination (Figure 2B and Figures S5–S7). Therefore, VanB, a similar Rieske reductase to TsaB, was tested in each of the TsaM-catalyzed reactions (Figure 2B and Figure S5). Using VanB, rather than TsaB, in combination with TsaM, resulted in the formation of 8.75- and 14-fold higher quantities of p-(hydroxymethyl)benzenesulfonate and p-(hydroxymethyl)benzoate, respectively (Figure 2B and Figure S5). Likewise, VanB was determined to serve as a functional replacement for DdmA/DdmB. This piece of data reveals that VanB, a ferredoxin-NAD⁺-reductase, can substitute for a two-component ferredoxin-NAD⁺-reductase system that employs the [2Fe-2S]-type ferredoxin protein, DdmB. In this case, however, product formation is not amplified relative to the native reductase system (Figure 2B and Figure S7). Nevertheless, due to the ability of VanB to support the chemistry of TsaM, VanA, and DdmC, VanB was employed as the primary electron donor for this work (Figure 2B and Figure S7). Using VanB as an electron donor for TsaM, VanA, and DdmC, the activity on p-toluenesulfonate, p-methylbenzoate, vanillate, and dicamba was assessed (Figure 2C and Figure S8). Here, it was determined that all three Rieske oxygenase systems (TsaM/VanB, VanA/VanB, and DdmC/VanB) accept vanillate as a substrate, both TsaM and VanA accept and oxygenate p-toluenesulfonate and p-methylbenzoate substrates, and neither TsaM nor VanA performs chemistry on dicamba (Figure 2C and Figure S8).

Figure 2.

TsaM, VanA, and DdmC can each use VanB as an electron donor, and all enzymes show activity on non-native substrates. (A) Each of the three purified Rieske oxygenase systems (TsaM/TsaB, VanA/VanB, and DdmC/DdmA/DdmB) was tested for their ability to catalyze their native transformations using liquid chromatography–mass spectrometry (LC–MS) experiments. (B) A comparison of the total turnover numbers (TTN) for TsaM and DdmC using both the native reductases (TsaB and DdmA/DdmB) and a non-native reductase (VanB) was performed. Here, it was determined that VanB supports higher activity of TsaM. In the case of DdmC, when VanB is used in place of DdmA and DdmB, approximately the same amount of products is formed. (C) The activity of TsaM/VanB, VanA/VanB, and DdmC/VanB was assessed using p-toluenesulfonate and p-methylbenzoate (left panel), vanillate (middle panel), and dicamba substrates (right panel). In panels B and C, data were measured using n = 3 independent experiments and are presented as the mean value ± SD of these measurements.

Identification of Protein “Hotspots” for Engineering TsaM

Toward the goal of exploring whether targeting residues found in the presumed active site, loop, and substrate entrance channel of TsaM would be an effective strategy for engineering TsaM to exhibit altered site selectivity and substrate specificity, a sequence alignment of the three different proteins and AlphaFold^52,75 models of VanA and TsaM were produced (Figures S3 and S9A,B). Using the location of a previously calculated⁷ substrate entrance tunnel and the available structure of substrate-bound DdmC³⁰ as guides, 10 TsaM residues of interest were identified (Figure 3A,B and Figure S9C). Six of these residues were predicted to be located in the putative active site and were chosen because their DdmC counterparts either interact with dicamba (Met230, Ser257, and Gly291) or show different hydrophobic propensities (Tyr220, Thr232, and His255). Four of the chosen residues are equivalent to residues in DdmC that surround the entrance of the tunnel to the active site (His204, Arg216, Ser234, and Phe253). Of these residues, one (His204) is presumably located on the β13-to-β14 connecting loop (Figure 3A and Figures S3 and S9C,D). To determine whether these residues, which span the so-called “hotspot” regions, have a profound effect on the outcome of the TsaM-catalyzed reaction, 18 TsaM variants that contain different combinations of two, three, four, five, six, eight, nine, or 10 mutations at the identified locations were made (Table 1). The identified residues were mutated into the corresponding residues found in DdmC (H204M, R216A, Y220I, M230N, T232I, S234V, F253S, H255G, S257H, and G291W, Table 1). The DdmC primary sequence was chosen as the guide for engineering because wild-type DdmC shows activity on both vanillate and dicamba substrates (Figure 2C and Figures S3 and S8C,D).

Figure 3.

The site selectivity of the TsaM-catalyzed reaction can be rationally altered. (A) “Hotspot” residues in the tunnel and loop regions of TsaM were identified using a calculated substrate entrance tunnel and a combination of sequence alignments, AlphaFold models,^52,75 and the crystal structure of DdmC (PDB ID 3GKE).³⁰ (B) “Hotspot” residues in the active site were identified using the crystal structure of dicamba-bound DdmC.³⁰ Here, residues that interact with dicamba or differ in polarity from those in TsaM were chosen as mutagenesis targets. (C) TsaM performs a monooxygenation reaction on p-methylbenzoate and m-methylbenzoate. VanA performs a monooxygenation reaction on p-methylbenzoate, m-methylbenzoate, and o-methylbenzoate. DdmC catalyzes a monooxygenation reaction only on o-methylbenzoate and m-methylbenzoate. Substrates tested with the different enzymes are colored based on the structure shown on the left. (D) The combined introduction, but not the individual introduction, of two mutations into the active site of TsaM significantly improves the activity on m-methylbenzoate and o-methylbenzoate. (E) TsaM catalyzes oxidative demethylation reactions on p-(methoxy)benzoate and m-(methoxy)benzoate substrates but does not accept an o-(methoxy)benzoate substrate. VanA, on the other hand, can accept all three substrates, and DdmC performs chemistry on m-(methoxy)benzoate and o-(methoxy)benzoate. This trend resembles that shown in panel C. (F) The iterative introduction of mutations into the active site (Met230, His255, and Ser257) and substrate entrance tunnel (Arg218, Ser234, and Phe253) results in significantly decreased activity of TsaM on a p-(methoxy)benzoate substrate. (G) In contrast, the addition of mutations into the active site and tunnel of TsaM leads to improved activity on m-(methoxy)benzoate. Variant 14, which contains six mutations distributed between the active site and tunnel, is not significantly different from wild-type VanA with respect to its activity on an m-(methoxy)benzoate substrate. (H) The same six mutation-containing variant of TsaM that leads to the highest level of activity on m-(methoxy)benzoate is also active on o-(methoxy)benzoate. The activity of this variant is not significantly different from that of wild-type VanA but is greatly reduced relative to wild-type DdmC. The residues highlighted in panels A and B are labeled by their identities in DdmC/TsaM and are numbered based on their position in the TsaM sequence. The data shown in panels C–H were measured using n = 3 independent experiments and are presented as the mean value ± SD of these measurements. All numbered variants in the bar graphs are described in Table 1. In this figure, ****p < 0.0001, ***p < 0.001, **p < 0.01, and ns indicates no significant difference from an ordinary one-way ANOVA Tukey analysis.

Table 1. TsaM Variants Investigated in This Work.

variant number	mutations
1	M230N
2	S257H
3	M230N/S257H
4	F253S/S257H
5	H255G/S257H
6	M230N/S234V
7	R216A/F253S/S257H
8	R216A/H255G/S257H
9	R216A/M230N/S234V
10	M230N/S234V/F253S/S257H
11	M230N/S234V/H255G/S257H
12	R216A/M230N/S234V/F253S/S257H
13	R216A/M230N/S234V/H255G/S257H
14	R216A/M230N/S234V/F253S/H255G/S257H
15	R216A/Y220I/M230N/T232I/S234V/F253S/H255G/S257H
16	H204M/R216A/Y220I/M230N/T232I/S234V/F253S/H255G/S257H
17	R216A/Y220I/M230N/T232I/S234V/F253S/H255G/S257H/G291W
18	H204M/R216A/Y220I/M230N/T232I/S234V/F253S/H255G/S257H/G291W

Once each TsaM variant was made, expressed, purified, and assessed for correct folding using circular dichroism (CD) spectroscopy, activity was probed using a series of different substrates (Table S2 and Figures S10 and S11). The different substrates used in the assays were designed to iteratively probe the effect of the introduced mutations on the site selectivity (p-methylbenzoate, m-methylbenzoate, o-methylbenzoate, p-(methoxy)benzoate, m-(methoxy)benzoate, and o-(methoxy)benzoate), and substrate specificity of the catalyzed reaction (p-toluenesulfonate, p-methylbenzoate, vanillate, and dicamba, Table S2).

Rational Alteration of the Site Selectivity of the TsaM-Catalyzed Reaction

As a starting point toward evaluating the importance of the identified residues to the TsaM-catalyzed reaction, the ability of wild-type TsaM, VanA, and DdmC to perform chemistry on p-methylbenzoate, m-methylbenzoate, and o-methylbenzoate was analyzed using liquid chromatography–mass spectrometry (LC–MS) experiments (Figure 3C and Figure S12). In these experiments, it was revealed that whereas VanA shows activity on each of the three substrates, TsaM and DdmC show activity on only two (Figure 3C and Figure S12). TsaM accepts and oxygenates both p-methylbenzoate and m-methylbenzoate, whereas DdmC catalyzes a monooxygenation reaction only on o-methylbenzoate and m-methylbenzoate (Figure 3C and Figure S12). Once this baseline level of activity was established, the ability of the TsaM active site variants M230N, S257H, and M230N/S257H to oxygenate these substrates was probed (Table 1, Figure 3D, and Figure S13). Although creation of single M230N or S257H variants decreased the activity of TsaM on p-methylbenzoate and m-methylbenzoate, the combined introduction of M230N and S257H mutations led to improved activity on m-methylbenzoate and o-methylbenzoate, relative to wild-type TsaM (Figure 3D).

Surprisingly, despite the observed increase in activity with the latter double variant and the knowledge that wild-type DdmC performs chemistry on both of these molecules, no further amplification of activity was detected on the provided substrates using the other variants that contain two (variants 3–6), three (variants 7–9), four (variants 10–11), or five mutations (variants 12–13) distributed between the active site and tunnel regions (Table 1 and Figure S14). This inability to further increase the activity of TsaM on m-methylbenzoate and o-methylbenzoate suggested that either the “hotspots” or the guiding sequence for engineering was incorrectly chosen and may not be a valuable strategy for engineering divergent Rieske oxygenases. Regarding the latter possibility, as the native chemistry of DdmC is performed on a methoxy substituent, rather than a methyl group, it was hypothesized that the “hotspot” regions in DdmC may be biased to accept a dicamba substrate, which has markedly different chemical attributes from the methylated substrate options. Since the sizes and charge distributions of p-(methoxy)benzoate, m-(methoxy)benzoate, and o-(methoxy)benzoate more closely resemble dicamba and it was hypothesized that these properties are important for recognition by the variant enzymes, the activity of the TsaM variants was next tested on the methoxy suite of molecules (Figure S15). To begin these assays, again, the baseline activity of the wild-type enzymes on these substrates was established (Figure 3E and Figure S16). Here, it was determined that as previously described,⁴⁰ TsaM catalyzes oxidative demethylation reactions on p-(methoxy)benzoate and m-(methoxy)benzoate to produce p-hydroxybenzoate and m-hydroxybenzoate, respectively (Figure 3E and Figure S16). It was subsequently established that, as described for the methylated substrates, VanA accepts and oxygenates all three substrates, and DdmC performs oxidative demethylation reactions on m-(methoxy)benzoate and o-(methoxy)benzoate (Figure 3E and Figure S16).

Following these initial activity trials, the ability of variants that contain different combinations of mutations in the identified “hotspots” to perform chemistry at the p-, m-, and o-methoxy substrate positions was assessed. This experiment demonstrated that the introduction of single or multiple mutations into TsaM generally abolishes the activity on p-(methoxy)benzoate (Figure 3F and Figure S17A). In contrast and unlike that observed with the methylated substrates, TsaM variants 1–13, which contain rational mutations in the identified active site and tunnel regions, show improved activity on both m-(methoxy)benzoate and o-(methoxy)benzoate (Figure 3G,H and Figure S17B,C). With this noted increase in activity, variant 14, which contains six mutations distributed between the active site (M230N, H255G, and S257H) and substrate entrance tunnel (R216A, S234V, and F253S), was also tested for its ability to catalyze an oxidative demethylation reaction on these substrates (Table 1). Remarkably, this variant yielded the highest level of activity on the non-native m-(methoxy)benzoate substrate relative to wild-type TsaM, and the total turnover numbers of this variant on m-(methoxy)benzoate and o-(methoxy)benzoate show no significant difference to that observed with wild-type VanA (Figure 3G,H and Figure S17B,C). However, the latter activity on o-(methoxy)benzoate is significantly lower than that observed with wild-type DdmC (Figure 3H and Figure S17C). These findings illustrate that the site of TsaM-catalyzed oxidative demethylation can be adjusted to favor the meta position, rather than the para position, by cumulatively targeting residues distributed between the active site and tunnel. Therefore, these experiments support the notion that these “hotspots” are important dictators of selectivity across the enzyme class. Furthermore, as the sequence of DdmC provided the blueprint for this engineering campaign, these results also suggest that rational targeting of these residues may not only allow TsaM to be engineered to perform chemistry on dicamba but may also provide a mechanism to enhance the reactivity of TsaM on vanillate.

TsaM Can Be Rationally Engineered to Catalyze the VanA Reaction

As described above, wild-type TsaM performs an oxidative demethylation reaction on vanillate and m-(methoxy)benzoate substrates (Figures 2C and 3E and Figures S8C and S17B). Building on these findings and the knowledge that TsaM variant 14 is significantly more active on m-(methoxy)benzoate, relative to wild-type TsaM, TsaM variants 1–14 were iteratively assayed with a vanillate substrate (Table 1, Figure 4A, and Figure S18A). In this venture, it was determined that, as shown for the m-(methoxy)benzoate substrate, the total turnover numbers with a vanillate substrate did not appreciably increase until variants 12–14, which contain five or six mutations distributed through the tunnel and loop regions, were assayed (Figure 4A and Figure S18A). Again, the highest turnover is observed with variant 14, and therefore, using this variant, the apparent steady-state kinetic parameters were measured. As previously described,¹⁹ in this work, the apparent kinetic parameters are reported because saturating concentrations for NADH and O₂ in our enzyme systems have not been determined. In these experiments, variant 14 showed an approximate two-fold increase in the apparent k_cat and a 14-fold decrease in the apparent K_M relative to wild-type TsaM with a vanillate substrate (Tables 1 and 2 and Figure S19). As a result of these improved apparent kinetic parameters, the k_cat of TsaM variant 14 is remarkably close (3.8 ± 0.2 min^–1) to that of wild-type VanA (3.4 ± 0.3 min^–1), meaning that the catalytic efficiency of TsaM with a vanillate substrate can be amplified 22-fold by the introduction of six mutations.

Figure 4.

TsaM can be rationally engineered to exhibit the catalytic activity and substrate specificity of VanA. (A) The iterative introduction of mutations into the active site (Met230, His255, and Ser257) and substrate entrance tunnel (Arg218, Ser234, and Phe253) of TsaM revealed that TsaM variant 14 has a similar level of activity to wild-type VanA with a vanillate substrate. This activity is significantly higher than that measured for wild-type TsaM. (B) In contrast to wild-type TsaM, TsaM variant 14 shows a strong preference for a vanillate, rather than a p-toluenesulfonate substrate. Additional information for this panel is contained in Figure S20. This substrate preference mirrors that of wild-type VanA. In this figure, data were measured using n = 3 independent experiments and are presented as the mean value ± SD of these measurements. All numbered variants in the bar graph are described in Table 1. In this figure, ****p < 0.0001, and ns indicates no significant difference from an ordinary one-way ANOVA Tukey analysis.

Table 2. Summary of Apparent Kinetic Parameters Determined for the Wild-Type and Variant Enzymes Studied in This Work.

enzyme and substrate	KMapp (μM)	kcatapp (min–1)	kcatapp/KMapp (M–1 s–1)
TsaM, VanB, and vanillate	182 ± 3.7	1.8 ± 0.7	230 ± 48
TsaM variant 14, VanB, and vanillate	13 ± 0.8	3.8 ± 0.2	5100 ± 530
TsaM variant 15, VanB, and vanillate	20 ± 1.2	3.7 ± 0.2	3200 ± 260
VanA, VanB, and vanillate	6.4 ± 0.6	3.4 ± 0.3	8900 ± 800
TsaM, VanB, and dicamba	N/A	N/A	N/A
TsaM variant 15, VanB, and dicamba	100 ± 9.2	1.8 ± 0.2	300 ± 43
TsaM variant 18, VanB, and dicamba	42 ± 6.2	2.8 ± 0.3	1100 ± 200
TsaM variant 18, DdmA, DdmB, and dicamba	39 ± 7.2	2.7 ± 0.2	1200 ± 240
DdmC, DdmA, DdmB, and dicamba	28 ± 3.4	2.8 ± 0.1	2100 ± 240
DdmC, VanB, and dicamba	22 ± 4.1	2.8 ± 0.3	2300 ± 410

In addition, TsaM variant 14, when provided with an equimolar amount of p-toluenesulfonate and vanillate or p-methylbenzoate and vanillate, preferentially performs chemistry on vanillate, producing 99% of a 3,4-dihydroxybenzoate product (Figure 4B and Figure S20). This result stands in stark contrast to that observed for TsaM, for which more than 99% of the product made is p-(hydroxymethyl)benzenesulfonate or p-(hydroxymethyl)benzoate (Figure 4B and Figure S20). Thus, TsaM variant 14, which shares just 37% sequence identity with VanA, not only showcases similar catalytic activity to VanA but also exhibits the substrate specificity of VanA. Also noteworthy is the fact that TsaM variant 14 shows only a low level of activity on p-toluenesulfonate and p-methylbenzoate, and despite the use of the DdmC sequence to motivate the rational mutagenesis experiments, this variant is not active on dicamba (Figure 5A and Figure S18). These results show that the starting promiscuous oxidative demethylation activity on vanillate becomes the primary reaction catalyzed by TsaM and support our hypothesis that the introduced mutations bias the substrate preference away from the methylated molecules.

Figure 5.

TsaM can be rationally engineered to exhibit the catalytic activity and substrate specificity of DdmC. (A) TsaM variant 18, which contains six mutations in the active site (Tyr220, Met230, Thr232, His255, Ser257, and Gly291), three mutations in the substrate entrance tunnel (Arg218, Ser234, and Phe253), and one mutation on the β13-to-β14 connecting loop (H204M), shows a high level of activity on dicamba. (B) Here, wild-type TsaM and DdmC, as well as several TsaM variants, were provided with an equimolar amount of p-toluenesulfonate and dicamba. Akin to wild-type DdmC, TsaM variants 16–18 preferentially perform chemistry on dicamba 100% of the time. Variant 15 also shows a clear albeit slightly lower preference for dicamba (98.5%). In contrast, wild-type TsaM and variant 14 prefer p-toluenesulfonate (100%). (C) In addition to showing activity on dicamba, variant 18 is active on vanillate. This observed activity is not significantly different from wild-type DdmC but is significantly different from variant 14. (D) Similar to the total turnover number, the substrate specificity of TsaM variant 18 closely mirrors that of DdmC. More specifically, variant 18 and wild-type DdmC choose to hydroxylate dicamba, rather than vanillate, 97 and 98% of the time. This number is higher than that observed for TsaM variants 14 (0%), 15 (12%), 16 (68%), and 17 (36%). (E) When challenged with three different substrates, variant 14 (98% of chemistry happens on vanillate) and variant 18 (97% of chemistry happens on dicamba) closely resemble VanA (99% of chemistry happens on vanillate) and DdmC (98% of chemistry happens on dicamba), respectively. Additional information for panels B, D, and E is contained in Figures S23, S24, S27, and S28. In this figure, all data were measured using n = 3 independent experiments and are presented as the mean value ± SD of these measurements. All numbered variants in the bar graphs are described in Table 1. In this figure, ****p < 0.0001, ***p < 0.001, and ns indicates no significant difference from one-way an ordinary ANOVA Tukey analysis.

TsaM Can Be Rationally Adapted to Catalyze the DdmC Reaction

Based upon the success of rationally engineering TsaM to exhibit the activity and specificity of VanA, it was next investigated whether similar strategies could be used to engineer TsaM to fully behave like DdmC. For this undertaking, unlike that described for the vanillate substrate, wild-type TsaM, which lacks activity on dicamba, needed to be engineered into an enzyme that produces 3,6-dichlorosalicylic acid (Figure 2C and Figure S8D). Intriguingly, whereas TsaM variant 14 shows near wild-type VanA levels of activity on m-(methoxy)benzoate and vanillate, as described above, this variant shows only low levels of activity on o-(methoxy)benzoate and is unable to perform chemistry on dicamba (Figure 3H, Figure 5A, and Figure S18D,E). Based on these results, it was hypothesized that the chemical nature of dicamba and the presence of two chlorine atoms on the aromatic ring that are not found in o-(methoxy)benzoate are responsible for these differences in activity (Figure S15).

To test this hypothesis, the active site of DdmC was again examined (Figure 3B). As previously described,³⁰ the chlorine atoms of dicamba do not form any direct interactions in the active site of DdmC but instead appear to help orient dicamba such that chemistry occurs on the methoxy substituent (Figure 3B). Comparison of the equivalent pocket in the AlphaFold model of TsaM reveals the presence of two polar residues (Thr232 and Tyr220) in place of Ile232 and Ile220 (Figure S21). As these residues appear to change the microenvironment of the binding pocket and the side chain of Tyr220 projects closer to dicamba than the DdmC residue Ile220, TsaM variant 15 was tested for its ability to catalyze the conversion of dicamba into 3,6-dichlorosalicylic acid (Figure 5A and Figure S22A). This variant contains eight mutations distributed between the active site (Y220I, M230N, T232I, H255G, and S257H) and substrate entrance tunnel (R216A, S234V, and F253S) and shows a low level of activity on dicamba (Table 1, Figure 5A, and Figure S22A). Regarding substrate specificity, when challenged to perform chemistry on a native substrate of TsaM or dicamba, this variant oxidizes dicamba 99% of the time (Figure 5B and Figure S23). However, despite the added ability of this variant to perform chemistry on dicamba, this variant, which is similar to TsaM variant 14, still shows a strong preference for vanillate (Figure 5C,D and Figures S22B and S24). Consistent with this substrate preference, when provided with a vanillate substrate, TsaM variant 15 has relatively similar kinetic parameters to those measured for variant 14, further confirming that the targeted active site and tunnel residues are not enough to convert TsaM into an enzyme that has both the catalytic activity and substrate specificity of DdmC (Table 2, Figure 5B,D, and Figure S19E).

The total turnover number measured for dicamba is amplified in variants 16 and 17, which contain additional H204M and G291W mutations on the presumed loop and in the active site, respectively (Table 1, Figure 5A, and Figure S22A). These TsaM variants come not only with improved activity on dicamba but also have altered substrate specificities. Quite notably, as described for TsaM variant 15, when provided with a choice to perform chemistry on the native TsaM substrates or dicamba, these variants convert dicamba into 3,6-dichlorosalicylic acid 100% of the time (Figure 5B and Figures S22C,D and S23). However, these variants still fail to embrace the ability of DdmC to select dicamba over vanillate (Figure 5D and Figure S24). Nonetheless, TsaM variant 16 is the first that shows a preference for dicamba over vanillate, suggesting that the combined effect of targeting residues in the active site, tunnel, and loop is a promising strategy for conferring activity on dicamba (Figure 5D and Figures S22B and S24). As such, the apparent steady-state kinetic parameters for variant 16 were investigated (Table 2 and Figure S25A–C). Unlike wild-type TsaM, which shows no activity on dicamba, the k_cat of this variant is 76% of that measured for DdmC, and the apparent K_M is approximately 5 times higher than that measured for wild-type DdmC (Table 2 and Figures S22D and S25A–C).

As expected, based on the trend of improved activity on dicamba through the cumulative targeting of the three “hotspot” regions, the highest turnover is observed using TsaM variant 18, which contains 10 mutations, six in the active site (Y220I, M230N, T232I, H255G, S257H, and G291W), one on the loop (H204M), and three in the tunnel (R216A, S234V, and F253S, Figure 5A and Figure S22A). The kinetic parameters for this variant closely mirror those of DdmC (Table 2 and Figure S25). As described for TsaM variants 15, 16, and 17, this variant, which shares approximately 37% sequence identity with DdmC, is devoid of activity on both native TsaM substrates (Figure S22C,D). On the other hand, the activity and specificity of this variant with respect to a vanillate substrate, unlike the variants described above, are not significantly different from those of wild-type DdmC (Figure 5D and Figure S24). To further interrogate the catalytic activity of TsaM variants 16, 17, and 18, the total turnover numbers were also measured in the presence of a dicamba substrate and the native DdmA/DdmB reductase proteins (Figure S26A). Here, it was shown that there was no significant difference in the amount of the 3,6-dichlorosalicylic acid product formed using DdmA/DdmB in place of VanB (Figure S26A). Likewise, the apparent k_cat of variant 18 in the presence of DdmA/DdmB not only resembles that measured in the presence of VanB but also mirrors that of the wild-type DdmA/DdmB/DdmC system (Table 2 and Figure S26B,C).

Biochemical Assessment of Engineered Enzymes

In an additional investigation of the engineered VanA- and DdmC-like TsaM variants, the overall substrate preferences of the wild-type enzymes were compared to TsaM variants 14, 15, 16, 17, and 18 (Table 1). To perform this experiment, rather than supplying each enzyme with two substrate options, an equimolar amount of three substrate options (p-toluenesulfonate, vanillate, and dicamba or p-methylbenzoate, vanillate, and dicamba) was provided (Figure 5E and Figures S27 and S28). These experiments revealed that variant 14 closely resembles wild-type VanA, and variant 18 behaves in an equivalent manner to wild-type DdmC. TsaM variants 15–17 also deviate substantially from wild-type TsaM, and each shows varying levels of activity on vanillate and dicamba (Figure 5E). These results further confirm that wild-type TsaM can be rationally engineered to not only exhibit the chemistry of VanA and DdmC but also adapted to showcase the substrate specificity of these enzymes. Further assessment of the activity of variants 14, 15, 16, 17, and 18 using the aforementioned methylated substrates (p-methylbenzoate, m-methylbenzoate, and o-methylbenzoate) shows that the ability of the wild-type VanA and DdmC enzymes to perform monooxygenation reactions on m-methylbenzoate and o-methylbenzoate, respectively, has been conferred through our engineering campaign (Figure 3C and Figures S14 and S29). This result suggests that the TsaM variant enzymes also showcase similar substrate scopes to wild-type VanA and DdmC.

It was next explored whether these engineering endeavors resulted in variant enzymes that were less stable than wild-type TsaM. Through the use of differential scanning fluorimetry experiments, it was determined that wild-type TsaM has a melting temperature (T_m) of 53.7 ± 0.5 °C. Similarly, the VanA- and DdmC-like variants (variants 14 and 18) exhibit T_m values of 53.0 ± 0.1 and 53.0 ± 0.2 °C, respectively (Figure S30). These results importantly suggest that the stability, like the folding of these variants, was not deleteriously impacted by our engineering campaign (Figures S11 and S30).

Finally, it was investigated whether the engineering campaign resulted in uncoupling of the oxygen activation step from substrate functionalization. Uncoupling is a long-described but recently appreciated phenomenon that can result in lower product formation via the loss of the activated oxygen species as H₂O₂ or reactive-oxygen-species-induced enzyme modification and inactivation.^76,77 To probe whether the oxygen activation step had been perturbed or directed down an unproductive pathway in our engineering campaign, the extent of reaction uncoupling was compared between wild-type TsaM and the variant enzymes (Figures S31 and S32). To make these measurements, a previously described^76,78 assay that relies on the horseradish peroxidase and H₂O₂-mediated conversion of 10-acetyl-3,7-dihydroxypenoxazine into resorufin was implemented. Upon incubation of these reagents with wild-type TsaM or TsaM variants 14, 15, 16, 17, or 18 and a vanillate, dicamba, or native TsaM substrate, the absorbance of resorufin was monitored for 190 min. A comparison of the endpoints measured in each of these assays reveals that there is no significant difference in the amount of H₂O₂ formed by the different tested enzymes. This result suggests that the rational engineering did not disrupt the path of the activated oxygen intermediate, confirming that making targeted mutations in the active site, tunnel, and the β13-to-β14 connecting loop of a Rieske oxygenase is a valuable and robust engineering strategy (Figures S31 and S32).

Discussion

Enzymes are nature’s catalysts. They have evolved, both in nature and in the laboratory, to catalyze reactions that are relevant to medical, environmental, and industrial applications.⁷⁹⁻⁸¹ However, despite the emerging uses of enzymes in these fields, taking a natural enzyme into an industrial application often requires substantial engineering for greater stability, different catalytic properties, or new functionality.⁸² Since the entire polypeptide sequence constitutes the catalyst, manipulating individual functional aspects of an enzyme for biocatalysis is not a straightforward process. Indeed, it has previously been suggested that 60 to 70% of mutations introduced in protein engineering campaigns will be detrimental to catalytic activity.⁸³ In contrast, it has also been said that 5%, or even as low as 0.01%, of mutations are beneficial.^83,84 Sampling a sequence to identify the enigmatic 0.01 to 5% can be an arduous task, especially in protein superfamilies for which there is a lack of understanding regarding how function is dictated at the molecular level.^83,84 Furthermore, the fact that enzymatic properties are often not conferred by individual amino acids but rather dictated by the cooperative effort of residues distributed among different structural regions only compounds these challenges.⁸⁵ Second- and third-sphere residues⁸⁶⁻⁸⁹ and protein dynamics⁹⁰ play a role in catalytic outcomes and suggest that this low percentage of residues will include more than just those found in the active site.

In this work, toward finding this small subset of residues that confer the site selectivity and substrate specificity of TsaM, we used the available substrate-bound structure of DdmC,³⁰ modern bioinformatics and structure prediction methods,^52,75 and previous lessons learned^27,28 to create a focused library of 18 TsaM variants (Table 1 and Figure 3A,B). Through sampling of these variants, we first identified that TsaM could be engineered to preferentially perform chemistry on m-(methoxy)benzoate, effectively shifting the primary site of chemistry on the substrate away from that natively showcased by TsaM (Figure 3E–G). Importantly, this TsaM variant (variant 14) also exhibits a new functionality as it has a low level of activity on o-(methoxy)benzoate and o-methylbenzoate (Figure 3C–H and Figure S29). Along with the low activity exhibited on p-toluenesulfonate and p-methylbenzoate, the scope of substrates accepted by variant 14 closely resembles that of wild-type VanA (Figures 2C and 3C,E). Consistently, variant 14 is the variant that showcases the highest level of activity on vanillate. Therefore, by using the structure of DdmC as a guide, our engineering campaign exploited the promiscuous ability of TsaM and DdmC to perform chemistry on vanillate (Figures 2C and 4). This result supports the idea that subtle changes made to the identity of active site protein residues can amplify promiscuous enzyme activities by creating a better binding pocket for a non-native substrate.⁸³ As described here and in our previous work,^27,28 changes to the fit of a substrate within the active site mean little without commensurate changes to the substrate entrance tunnel that leads to the active site.

From variant 14, just four more rational mutations converted TsaM into an enzyme that is reprogrammed to exhibit the functionality and substrate specificity of DdmC (Figure 5A). This functional jump bestows TsaM the ability to perform an oxidative demethylation reaction on dicamba and was made using a variant that contains six mutations in the active site (Y220I, M230N, T232I, H255G, S257H, and G291W), one on the β13-to-β14 connecting loop (H204M), and three in the tunnel region (R216A, S234V, and F253S, Figures 2C and 5A). Relative to the VanA-like variant (variant 14), this variant (variant 18) contains additional rational mutations in the active site and one mutation on the loop, which as previously hypothesized for SxtT and GxtA^27,28 is likely involved in substrate recognition.

Collectively, these results show that rationally targeting between 1.7 and 3% of the residues found in wild-type TsaM, all of which are in the three identified “hotspot” regions, is enough to confer TsaM the catalytic function of VanA or DdmC, respectively (Table 2 and Figures 4A and 5A). Similarly, this small set of rationally targeted residues is enough to impart a 100% switch of the tested substrate specificity between Rieske oxygenases that come from three different clades of a phylogenetic tree (Figures 1B, 4B, and 5B, and Figure S2). In fact, six of the 18 variants produced show significant deviations in their reaction specificity from wild-type TsaM, endorsing the idea that alterations to individual residues found within the “hotspot” regions trend toward desired changes in a Rieske oxygenase (Table 1 and Figure 3A,B). Indicated by CD and differential scanning fluorescence assays that show nominal changes in stability, we have further demonstrated that “hotspot”-motivated rational engineering is a promising avenue for developing Rieske oxygenase biocatalysts (Figures S11 and S30). With the success of this engineering campaign, future work, guided by the SSN to determine whether the identified design principles can be used to alter the reaction specificity of even more divergent Rieske oxygenases, including those that catalyze reactions other than monooxygenation or adopt non-α₃ architectures, is paramount. We posit that understanding the widespread contribution of these “hotspots” to enzymatic function is one important step toward adapting these biocatalysts in a predictable and reliable manner.

Funding and Additional Information

The work in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number R35 GM138271 (J.B.-R.). The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of NIGMS or NIH. A construct used in this work (VanA) was produced by the Joint Genome Institute. This work (proposal DOI 10.46936/10.25585/60000495) conducted by the U.S. Department of Energy Joint Genome Institute, a DOE Office of Science User Facility, is supported by the Office of Science of the U.S. Department of Energy operated under Contract No. DE-AC02-05CH11231.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.biochem.3c00150.

• Protein sequences; tables of mutagenesis primers and substrates; bioinformatics analysis; SDS-PAGE, circular dichroism, and thermal stability experiments; activity assays with different reductases and/or substrates; calculated substrate tunnels and density functional calculations; substrate specificity experiments; O₂ uncoupling assays; standard curves for product formation; calculated optimized coordinates and absolute energies of substrates (PDF)

Author Contributions

All authors (J.T., A.A.G., P.H.D., and J.B.-R.) contributed to the design of the experiments and wrote the manuscript. J.T. and A.A.G. performed mutagenesis experiments, purified all proteins and variants used in this work, and performed all activity measurements. P.H.D. performed substrate calculations and bioinformatics analyses.

The authors declare no competing financial interest.